1 This is doc/gcc.info, produced by makeinfo version 4.9 from
2 ../../../gcc-4.3.1/gcc/doc/gcc.texi.
4 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
5 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 2008 Free Software
8 Permission is granted to copy, distribute and/or modify this document
9 under the terms of the GNU Free Documentation License, Version 1.2 or
10 any later version published by the Free Software Foundation; with the
11 Invariant Sections being "GNU General Public License" and "Funding Free
12 Software", the Front-Cover texts being (a) (see below), and with the
13 Back-Cover Texts being (b) (see below). A copy of the license is
14 included in the section entitled "GNU Free Documentation License".
16 (a) The FSF's Front-Cover Text is:
20 (b) The FSF's Back-Cover Text is:
22 You have freedom to copy and modify this GNU Manual, like GNU
23 software. Copies published by the Free Software Foundation raise
24 funds for GNU development.
26 INFO-DIR-SECTION Software development
28 * gcc: (gcc). The GNU Compiler Collection.
29 * g++: (gcc). The GNU C++ compiler.
31 This file documents the use of the GNU compilers.
33 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
34 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 2008 Free Software
37 Permission is granted to copy, distribute and/or modify this document
38 under the terms of the GNU Free Documentation License, Version 1.2 or
39 any later version published by the Free Software Foundation; with the
40 Invariant Sections being "GNU General Public License" and "Funding Free
41 Software", the Front-Cover texts being (a) (see below), and with the
42 Back-Cover Texts being (b) (see below). A copy of the license is
43 included in the section entitled "GNU Free Documentation License".
45 (a) The FSF's Front-Cover Text is:
49 (b) The FSF's Back-Cover Text is:
51 You have freedom to copy and modify this GNU Manual, like GNU
52 software. Copies published by the Free Software Foundation raise
53 funds for GNU development.
57 File: gcc.info, Node: Top, Next: G++ and GCC, Up: (DIR)
62 This manual documents how to use the GNU compilers, as well as their
63 features and incompatibilities, and how to report bugs. It corresponds
64 to the compilers (GCC) version 4.3.1. The internals of the GNU
65 compilers, including how to port them to new targets and some
66 information about how to write front ends for new languages, are
67 documented in a separate manual. *Note Introduction: (gccint)Top.
71 * G++ and GCC:: You can compile C or C++ programs.
72 * Standards:: Language standards supported by GCC.
73 * Invoking GCC:: Command options supported by `gcc'.
74 * C Implementation:: How GCC implements the ISO C specification.
75 * C Extensions:: GNU extensions to the C language family.
76 * C++ Extensions:: GNU extensions to the C++ language.
77 * Objective-C:: GNU Objective-C runtime features.
78 * Compatibility:: Binary Compatibility
79 * Gcov:: `gcov'---a test coverage program.
80 * Trouble:: If you have trouble using GCC.
81 * Bugs:: How, why and where to report bugs.
82 * Service:: How to find suppliers of support for GCC.
83 * Contributing:: How to contribute to testing and developing GCC.
85 * Funding:: How to help assure funding for free software.
86 * GNU Project:: The GNU Project and GNU/Linux.
88 * Copying:: GNU General Public License says
89 how you can copy and share GCC.
90 * GNU Free Documentation License:: How you can copy and share this manual.
91 * Contributors:: People who have contributed to GCC.
93 * Option Index:: Index to command line options.
94 * Keyword Index:: Index of concepts and symbol names.
97 File: gcc.info, Node: G++ and GCC, Next: Standards, Prev: Top, Up: Top
99 1 Programming Languages Supported by GCC
100 ****************************************
102 GCC stands for "GNU Compiler Collection". GCC is an integrated
103 distribution of compilers for several major programming languages.
104 These languages currently include C, C++, Objective-C, Objective-C++,
105 Java, Fortran, and Ada.
107 The abbreviation "GCC" has multiple meanings in common use. The
108 current official meaning is "GNU Compiler Collection", which refers
109 generically to the complete suite of tools. The name historically stood
110 for "GNU C Compiler", and this usage is still common when the emphasis
111 is on compiling C programs. Finally, the name is also used when
112 speaking of the "language-independent" component of GCC: code shared
113 among the compilers for all supported languages.
115 The language-independent component of GCC includes the majority of the
116 optimizers, as well as the "back ends" that generate machine code for
119 The part of a compiler that is specific to a particular language is
120 called the "front end". In addition to the front ends that are
121 integrated components of GCC, there are several other front ends that
122 are maintained separately. These support languages such as Pascal,
123 Mercury, and COBOL. To use these, they must be built together with GCC
126 Most of the compilers for languages other than C have their own names.
127 The C++ compiler is G++, the Ada compiler is GNAT, and so on. When we
128 talk about compiling one of those languages, we might refer to that
129 compiler by its own name, or as GCC. Either is correct.
131 Historically, compilers for many languages, including C++ and Fortran,
132 have been implemented as "preprocessors" which emit another high level
133 language such as C. None of the compilers included in GCC are
134 implemented this way; they all generate machine code directly. This
135 sort of preprocessor should not be confused with the "C preprocessor",
136 which is an integral feature of the C, C++, Objective-C and
137 Objective-C++ languages.
140 File: gcc.info, Node: Standards, Next: Invoking GCC, Prev: G++ and GCC, Up: Top
142 2 Language Standards Supported by GCC
143 *************************************
145 For each language compiled by GCC for which there is a standard, GCC
146 attempts to follow one or more versions of that standard, possibly with
147 some exceptions, and possibly with some extensions.
152 GCC supports three versions of the C standard, although support for the
153 most recent version is not yet complete.
155 The original ANSI C standard (X3.159-1989) was ratified in 1989 and
156 published in 1990. This standard was ratified as an ISO standard
157 (ISO/IEC 9899:1990) later in 1990. There were no technical differences
158 between these publications, although the sections of the ANSI standard
159 were renumbered and became clauses in the ISO standard. This standard,
160 in both its forms, is commonly known as "C89", or occasionally as
161 "C90", from the dates of ratification. The ANSI standard, but not the
162 ISO standard, also came with a Rationale document. To select this
163 standard in GCC, use one of the options `-ansi', `-std=c89' or
164 `-std=iso9899:1990'; to obtain all the diagnostics required by the
165 standard, you should also specify `-pedantic' (or `-pedantic-errors' if
166 you want them to be errors rather than warnings). *Note Options
167 Controlling C Dialect: C Dialect Options.
169 Errors in the 1990 ISO C standard were corrected in two Technical
170 Corrigenda published in 1994 and 1996. GCC does not support the
173 An amendment to the 1990 standard was published in 1995. This
174 amendment added digraphs and `__STDC_VERSION__' to the language, but
175 otherwise concerned the library. This amendment is commonly known as
176 "AMD1"; the amended standard is sometimes known as "C94" or "C95". To
177 select this standard in GCC, use the option `-std=iso9899:199409'
178 (with, as for other standard versions, `-pedantic' to receive all
179 required diagnostics).
181 A new edition of the ISO C standard was published in 1999 as ISO/IEC
182 9899:1999, and is commonly known as "C99". GCC has incomplete support
183 for this standard version; see
184 `http://gcc.gnu.org/gcc-4.3/c99status.html' for details. To select this
185 standard, use `-std=c99' or `-std=iso9899:1999'. (While in
186 development, drafts of this standard version were referred to as "C9X".)
188 Errors in the 1999 ISO C standard were corrected in three Technical
189 Corrigenda published in 2001, 2004 and 2007. GCC does not support the
192 By default, GCC provides some extensions to the C language that on
193 rare occasions conflict with the C standard. *Note Extensions to the C
194 Language Family: C Extensions. Use of the `-std' options listed above
195 will disable these extensions where they conflict with the C standard
196 version selected. You may also select an extended version of the C
197 language explicitly with `-std=gnu89' (for C89 with GNU extensions) or
198 `-std=gnu99' (for C99 with GNU extensions). The default, if no C
199 language dialect options are given, is `-std=gnu89'; this will change to
200 `-std=gnu99' in some future release when the C99 support is complete.
201 Some features that are part of the C99 standard are accepted as
202 extensions in C89 mode.
204 The ISO C standard defines (in clause 4) two classes of conforming
205 implementation. A "conforming hosted implementation" supports the
206 whole standard including all the library facilities; a "conforming
207 freestanding implementation" is only required to provide certain
208 library facilities: those in `<float.h>', `<limits.h>', `<stdarg.h>',
209 and `<stddef.h>'; since AMD1, also those in `<iso646.h>'; and in C99,
210 also those in `<stdbool.h>' and `<stdint.h>'. In addition, complex
211 types, added in C99, are not required for freestanding implementations.
212 The standard also defines two environments for programs, a
213 "freestanding environment", required of all implementations and which
214 may not have library facilities beyond those required of freestanding
215 implementations, where the handling of program startup and termination
216 are implementation-defined, and a "hosted environment", which is not
217 required, in which all the library facilities are provided and startup
218 is through a function `int main (void)' or `int main (int, char *[])'.
219 An OS kernel would be a freestanding environment; a program using the
220 facilities of an operating system would normally be in a hosted
223 GCC aims towards being usable as a conforming freestanding
224 implementation, or as the compiler for a conforming hosted
225 implementation. By default, it will act as the compiler for a hosted
226 implementation, defining `__STDC_HOSTED__' as `1' and presuming that
227 when the names of ISO C functions are used, they have the semantics
228 defined in the standard. To make it act as a conforming freestanding
229 implementation for a freestanding environment, use the option
230 `-ffreestanding'; it will then define `__STDC_HOSTED__' to `0' and not
231 make assumptions about the meanings of function names from the standard
232 library, with exceptions noted below. To build an OS kernel, you may
233 well still need to make your own arrangements for linking and startup.
234 *Note Options Controlling C Dialect: C Dialect Options.
236 GCC does not provide the library facilities required only of hosted
237 implementations, nor yet all the facilities required by C99 of
238 freestanding implementations; to use the facilities of a hosted
239 environment, you will need to find them elsewhere (for example, in the
240 GNU C library). *Note Standard Libraries: Standard Libraries.
242 Most of the compiler support routines used by GCC are present in
243 `libgcc', but there are a few exceptions. GCC requires the
244 freestanding environment provide `memcpy', `memmove', `memset' and
245 `memcmp'. Finally, if `__builtin_trap' is used, and the target does
246 not implement the `trap' pattern, then GCC will emit a call to `abort'.
248 For references to Technical Corrigenda, Rationale documents and
249 information concerning the history of C that is available online, see
250 `http://gcc.gnu.org/readings.html'
255 GCC supports the ISO C++ standard (1998) and contains experimental
256 support for the upcoming ISO C++ standard (200x).
258 The original ISO C++ standard was published as the ISO standard
259 (ISO/IEC 14882:1998) and amended by a Technical Corrigenda published in
260 2003 (ISO/IEC 14882:2003). These standards are referred to as C++98 and
261 C++03, respectively. GCC implements the majority of C++98 (`export' is
262 a notable exception) and most of the changes in C++03. To select this
263 standard in GCC, use one of the options `-ansi' or `-std=c++98'; to
264 obtain all the diagnostics required by the standard, you should also
265 specify `-pedantic' (or `-pedantic-errors' if you want them to be
266 errors rather than warnings).
268 The ISO C++ committee is working on a new ISO C++ standard, dubbed
269 C++0x, that is intended to be published by 2009. C++0x contains several
270 changes to the C++ language, some of which have been implemented in an
271 experimental C++0x mode in GCC. The C++0x mode in GCC tracks the draft
272 working paper for the C++0x standard; the latest working paper is
273 available on the ISO C++ committee's web site at
274 `http://www.open-std.org/jtc1/sc22/wg21/'. For information regarding
275 the C++0x features available in the experimental C++0x mode, see
276 `http://gcc.gnu.org/gcc-4.3/cxx0x_status.html'. To select this standard
277 in GCC, use the option `-std=c++0x'; to obtain all the diagnostics
278 required by the standard, you should also specify `-pedantic' (or
279 `-pedantic-errors' if you want them to be errors rather than warnings).
281 By default, GCC provides some extensions to the C++ language; *Note
282 Options Controlling C++ Dialect: C++ Dialect Options. Use of the
283 `-std' option listed above will disable these extensions. You may also
284 select an extended version of the C++ language explicitly with
285 `-std=gnu++98' (for C++98 with GNU extensions) or `-std=gnu++0x' (for
286 C++0x with GNU extensions). The default, if no C++ language dialect
287 options are given, is `-std=gnu++98'.
289 2.3 Objective-C and Objective-C++ languages
290 ===========================================
292 There is no formal written standard for Objective-C or Objective-C++.
293 The most authoritative manual is "Object-Oriented Programming and the
294 Objective-C Language", available at a number of web sites:
297 `http://developer.apple.com/documentation/Cocoa/Conceptual/ObjectiveC/'
298 is a recent (and periodically updated) version;
300 * `http://www.toodarkpark.org/computers/objc/' is an older example;
302 * `http://www.gnustep.org' and `http://gcc.gnu.org/readings.html'
303 have additional useful information.
305 2.4 Treelang language
306 =====================
308 There is no standard for treelang, which is a sample language front end
309 for GCC. Its only purpose is as a sample for people wishing to write a
310 new language for GCC. The language is documented in
311 `gcc/treelang/treelang.texi' which can be turned into info or HTML
314 *Note GNAT Reference Manual: (gnat_rm)Top, for information on standard
315 conformance and compatibility of the Ada compiler.
317 *Note Standards: (gfortran)Standards, for details of standards
318 supported by GNU Fortran.
320 *Note Compatibility with the Java Platform: (gcj)Compatibility, for
321 details of compatibility between `gcj' and the Java Platform.
324 File: gcc.info, Node: Invoking GCC, Next: C Implementation, Prev: Standards, Up: Top
326 3 GCC Command Options
327 *********************
329 When you invoke GCC, it normally does preprocessing, compilation,
330 assembly and linking. The "overall options" allow you to stop this
331 process at an intermediate stage. For example, the `-c' option says
332 not to run the linker. Then the output consists of object files output
335 Other options are passed on to one stage of processing. Some options
336 control the preprocessor and others the compiler itself. Yet other
337 options control the assembler and linker; most of these are not
338 documented here, since you rarely need to use any of them.
340 Most of the command line options that you can use with GCC are useful
341 for C programs; when an option is only useful with another language
342 (usually C++), the explanation says so explicitly. If the description
343 for a particular option does not mention a source language, you can use
344 that option with all supported languages.
346 *Note Compiling C++ Programs: Invoking G++, for a summary of special
347 options for compiling C++ programs.
349 The `gcc' program accepts options and file names as operands. Many
350 options have multi-letter names; therefore multiple single-letter
351 options may _not_ be grouped: `-dr' is very different from `-d -r'.
353 You can mix options and other arguments. For the most part, the order
354 you use doesn't matter. Order does matter when you use several options
355 of the same kind; for example, if you specify `-L' more than once, the
356 directories are searched in the order specified. Also, the placement
357 of the `-l' option is significant.
359 Many options have long names starting with `-f' or with `-W'--for
360 example, `-fmove-loop-invariants', `-Wformat' and so on. Most of these
361 have both positive and negative forms; the negative form of `-ffoo'
362 would be `-fno-foo'. This manual documents only one of these two
363 forms, whichever one is not the default.
365 *Note Option Index::, for an index to GCC's options.
369 * Option Summary:: Brief list of all options, without explanations.
370 * Overall Options:: Controlling the kind of output:
371 an executable, object files, assembler files,
372 or preprocessed source.
373 * Invoking G++:: Compiling C++ programs.
374 * C Dialect Options:: Controlling the variant of C language compiled.
375 * C++ Dialect Options:: Variations on C++.
376 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
378 * Language Independent Options:: Controlling how diagnostics should be
380 * Warning Options:: How picky should the compiler be?
381 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
382 * Optimize Options:: How much optimization?
383 * Preprocessor Options:: Controlling header files and macro definitions.
384 Also, getting dependency information for Make.
385 * Assembler Options:: Passing options to the assembler.
386 * Link Options:: Specifying libraries and so on.
387 * Directory Options:: Where to find header files and libraries.
388 Where to find the compiler executable files.
389 * Spec Files:: How to pass switches to sub-processes.
390 * Target Options:: Running a cross-compiler, or an old version of GCC.
391 * Submodel Options:: Specifying minor hardware or convention variations,
392 such as 68010 vs 68020.
393 * Code Gen Options:: Specifying conventions for function calls, data layout
395 * Environment Variables:: Env vars that affect GCC.
396 * Precompiled Headers:: Compiling a header once, and using it many times.
397 * Running Protoize:: Automatically adding or removing function prototypes.
400 File: gcc.info, Node: Option Summary, Next: Overall Options, Up: Invoking GCC
405 Here is a summary of all the options, grouped by type. Explanations are
406 in the following sections.
409 *Note Options Controlling the Kind of Output: Overall Options.
410 -c -S -E -o FILE -combine -pipe -pass-exit-codes
411 -x LANGUAGE -v -### --help[=CLASS] --target-help
415 *Note Options Controlling C Dialect: C Dialect Options.
416 -ansi -std=STANDARD -fgnu89-inline
418 -fno-asm -fno-builtin -fno-builtin-FUNCTION
419 -fhosted -ffreestanding -fopenmp -fms-extensions
420 -trigraphs -no-integrated-cpp -traditional -traditional-cpp
421 -fallow-single-precision -fcond-mismatch -flax-vector-conversions
422 -fsigned-bitfields -fsigned-char
423 -funsigned-bitfields -funsigned-char
425 _C++ Language Options_
426 *Note Options Controlling C++ Dialect: C++ Dialect Options.
427 -fabi-version=N -fno-access-control -fcheck-new
428 -fconserve-space -ffriend-injection
429 -fno-elide-constructors
430 -fno-enforce-eh-specs
431 -ffor-scope -fno-for-scope -fno-gnu-keywords
432 -fno-implicit-templates
433 -fno-implicit-inline-templates
434 -fno-implement-inlines -fms-extensions
435 -fno-nonansi-builtins -fno-operator-names
436 -fno-optional-diags -fpermissive
437 -frepo -fno-rtti -fstats -ftemplate-depth-N
438 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++
439 -fno-default-inline -fvisibility-inlines-hidden
440 -fvisibility-ms-compat
441 -Wabi -Wctor-dtor-privacy
442 -Wnon-virtual-dtor -Wreorder
443 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel
444 -Wno-non-template-friend -Wold-style-cast
445 -Woverloaded-virtual -Wno-pmf-conversions
448 _Objective-C and Objective-C++ Language Options_
449 *Note Options Controlling Objective-C and Objective-C++ Dialects:
450 Objective-C and Objective-C++ Dialect Options.
451 -fconstant-string-class=CLASS-NAME
452 -fgnu-runtime -fnext-runtime
454 -fobjc-call-cxx-cdtors
455 -fobjc-direct-dispatch
458 -freplace-objc-classes
462 -Wno-protocol -Wselector
463 -Wstrict-selector-match
464 -Wundeclared-selector
466 _Language Independent Options_
467 *Note Options to Control Diagnostic Messages Formatting: Language
470 -fdiagnostics-show-location=[once|every-line]
471 -fdiagnostics-show-option
474 *Note Options to Request or Suppress Warnings: Warning Options.
475 -fsyntax-only -pedantic -pedantic-errors
476 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds
477 -Wno-attributes -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual
478 -Wchar-subscripts -Wclobbered -Wcomment
479 -Wconversion -Wcoverage-mismatch -Wno-deprecated-declarations
480 -Wdisabled-optimization -Wno-div-by-zero
481 -Wempty-body -Wno-endif-labels
483 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2
484 -Wno-format-extra-args -Wformat-nonliteral
485 -Wformat-security -Wformat-y2k -Wignored-qualifiers
486 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int
487 -Wimport -Wno-import -Winit-self -Winline
488 -Wno-int-to-pointer-cast -Wno-invalid-offsetof
489 -Winvalid-pch -Wlarger-than-LEN -Wunsafe-loop-optimizations
490 -Wlogical-op -Wlong-long
491 -Wmain -Wmissing-braces -Wmissing-field-initializers
492 -Wmissing-format-attribute -Wmissing-include-dirs
494 -Wno-multichar -Wnonnull -Wno-overflow
495 -Woverlength-strings -Wpacked -Wpadded
496 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast
498 -Wreturn-type -Wsequence-point -Wshadow
499 -Wsign-compare -Wsign-conversion -Wstack-protector
500 -Wstrict-aliasing -Wstrict-aliasing=n
501 -Wstrict-overflow -Wstrict-overflow=N
502 -Wswitch -Wswitch-default -Wswitch-enum
503 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized
504 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code
505 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
506 -Wunused-value -Wunused-variable
507 -Wvariadic-macros -Wvla
508 -Wvolatile-register-var -Wwrite-strings
510 _C and Objective-C-only Warning Options_
511 -Wbad-function-cast -Wmissing-declarations
512 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs
513 -Wold-style-declaration -Wold-style-definition
514 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion
515 -Wdeclaration-after-statement -Wpointer-sign
518 *Note Options for Debugging Your Program or GCC: Debugging Options.
519 -dLETTERS -dumpspecs -dumpmachine -dumpversion
520 -fdbg-cnt-list -fdbg-cnt=COUNTER-VALUE-LIST
521 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit[-N]
522 -fdump-class-hierarchy[-N]
523 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline
525 -fdump-tree-original[-N]
526 -fdump-tree-optimized[-N]
527 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias
529 -fdump-tree-ssa[-N] -fdump-tree-pre[-N]
530 -fdump-tree-ccp[-N] -fdump-tree-dce[-N]
531 -fdump-tree-gimple[-raw] -fdump-tree-mudflap[-N]
534 -fdump-tree-phiopt[-N]
535 -fdump-tree-forwprop[-N]
536 -fdump-tree-copyrename[-N]
537 -fdump-tree-nrv -fdump-tree-vect
543 -ftree-vectorizer-verbose=N
544 -fdump-tree-storeccp[-N]
545 -feliminate-dwarf2-dups -feliminate-unused-debug-types
546 -feliminate-unused-debug-symbols -femit-class-debug-always
547 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs
548 -frandom-seed=STRING -fsched-verbose=N
549 -ftest-coverage -ftime-report -fvar-tracking
550 -g -gLEVEL -gcoff -gdwarf-2
551 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+
552 -fno-merge-debug-strings -fdebug-prefix-map=OLD=NEW
553 -femit-struct-debug-baseonly -femit-struct-debug-reduced
554 -femit-struct-debug-detailed[=SPEC-LIST]
555 -p -pg -print-file-name=LIBRARY -print-libgcc-file-name
556 -print-multi-directory -print-multi-lib
557 -print-prog-name=PROGRAM -print-search-dirs -Q
558 -print-sysroot-headers-suffix
561 _Optimization Options_
562 *Note Options that Control Optimization: Optimize Options.
563 -falign-functions[=N] -falign-jumps[=N]
564 -falign-labels[=N] -falign-loops[=N] -fassociative-math
565 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize
566 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves
567 -fcheck-data-deps -fcprop-registers -fcrossjumping -fcse-follow-jumps
568 -fcse-skip-blocks -fcx-limited-range -fdata-sections -fdce -fdce
569 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse
570 -fearly-inlining -fexpensive-optimizations -ffast-math
571 -ffinite-math-only -ffloat-store -fforward-propagate
572 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm
573 -fgcse-sm -fif-conversion -fif-conversion2 -finline-functions
574 -finline-functions-called-once -finline-limit=N
575 -finline-small-functions -fipa-cp -fipa-marix-reorg -fipa-pta
576 -fipa-pure-const -fipa-reference -fipa-struct-reorg
577 -fipa-type-escape -fivopts -fkeep-inline-functions -fkeep-static-consts
578 -fmerge-all-constants -fmerge-constants -fmodulo-sched
579 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap
580 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline
581 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability
582 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2
583 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros
584 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss
585 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls
586 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays
587 -fprofile-generate -fprofile-use -fprofile-values -freciprocal-math
588 -fregmove -frename-registers -freorder-blocks
589 -freorder-blocks-and-partition -freorder-functions
590 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops
591 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks
592 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous
593 -fsched-stalled-insns-dep[=N] -fsched-stalled-insns[=N]
594 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee
595 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller
596 -fsplit-wide-types -fstack-protector -fstack-protector-all
597 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer -ftree-ccp
598 -ftree-ch -ftree-copy-prop -ftree-copyrename -ftree-dce
599 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im
600 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize
601 -ftree-parallelize-loops=N -ftree-pre -ftree-reassoc -ftree-salias
602 -ftree-sink -ftree-sra -ftree-store-ccp -ftree-ter
603 -ftree-vect-loop-version -ftree-vectorize -ftree-vrp -funit-at-a-time
604 -funroll-all-loops -funroll-loops -funsafe-loop-optimizations
605 -funsafe-math-optimizations -funswitch-loops
606 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb
609 -O -O0 -O1 -O2 -O3 -Os
611 _Preprocessor Options_
612 *Note Options Controlling the Preprocessor: Preprocessor Options.
618 -include FILE -imacros FILE
619 -iprefix FILE -iwithprefix DIR
620 -iwithprefixbefore DIR -isystem DIR
621 -imultilib DIR -isysroot DIR
622 -M -MM -MF -MG -MP -MQ -MT -nostdinc
623 -P -fworking-directory -remap
624 -trigraphs -undef -UMACRO -Wp,OPTION
625 -Xpreprocessor OPTION
628 *Note Passing Options to the Assembler: Assembler Options.
629 -Wa,OPTION -Xassembler OPTION
632 *Note Options for Linking: Link Options.
633 OBJECT-FILE-NAME -lLIBRARY
634 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic
635 -s -static -static-libgcc -shared -shared-libgcc -symbolic
636 -Wl,OPTION -Xlinker OPTION
640 *Note Options for Directory Search: Directory Options.
641 -BPREFIX -IDIR -iquoteDIR -LDIR
642 -specs=FILE -I- --sysroot=DIR
645 *Note Target Options::.
646 -V VERSION -b MACHINE
648 _Machine Dependent Options_
649 *Note Hardware Models and Configurations: Submodel Options.
653 -mmangle-cpu -mcpu=CPU -mtext=TEXT-SECTION
654 -mdata=DATA-SECTION -mrodata=READONLY-DATA-SECTION
657 -mapcs-frame -mno-apcs-frame
659 -mapcs-stack-check -mno-apcs-stack-check
660 -mapcs-float -mno-apcs-float
661 -mapcs-reentrant -mno-apcs-reentrant
662 -msched-prolog -mno-sched-prolog
663 -mlittle-endian -mbig-endian -mwords-little-endian
664 -mfloat-abi=NAME -msoft-float -mhard-float -mfpe
665 -mthumb-interwork -mno-thumb-interwork
666 -mcpu=NAME -march=NAME -mfpu=NAME
667 -mstructure-size-boundary=N
669 -mlong-calls -mno-long-calls
670 -msingle-pic-base -mno-single-pic-base
673 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns
676 -mtpcs-frame -mtpcs-leaf-frame
677 -mcaller-super-interworking -mcallee-super-interworking
681 -mmcu=MCU -msize -minit-stack=N -mno-interrupts
682 -mcall-prologues -mno-tablejump -mtiny-stack -mint8
685 -mcpu=CPU[-SIREVISION]
686 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer
687 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly
688 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library
689 -mno-id-shared-library -mshared-library-id=N
690 -mleaf-id-shared-library -mno-leaf-id-shared-library
691 -msep-data -mno-sep-data -mlong-calls -mno-long-calls
692 -mfast-fp -minline-plt
695 -mcpu=CPU -march=CPU -mtune=CPU
696 -mmax-stack-frame=N -melinux-stacksize=N
697 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects
698 -mstack-align -mdata-align -mconst-align
699 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt
700 -melf -maout -melinux -mlinux -sim -sim2
701 -mmul-bug-workaround -mno-mul-bug-workaround
707 -all_load -allowable_client -arch -arch_errors_fatal
708 -arch_only -bind_at_load -bundle -bundle_loader
709 -client_name -compatibility_version -current_version
711 -dependency-file -dylib_file -dylinker_install_name
712 -dynamic -dynamiclib -exported_symbols_list
713 -filelist -flat_namespace -force_cpusubtype_ALL
714 -force_flat_namespace -headerpad_max_install_names
716 -image_base -init -install_name -keep_private_externs
717 -multi_module -multiply_defined -multiply_defined_unused
718 -noall_load -no_dead_strip_inits_and_terms
719 -nofixprebinding -nomultidefs -noprebind -noseglinkedit
720 -pagezero_size -prebind -prebind_all_twolevel_modules
721 -private_bundle -read_only_relocs -sectalign
722 -sectobjectsymbols -whyload -seg1addr
723 -sectcreate -sectobjectsymbols -sectorder
724 -segaddr -segs_read_only_addr -segs_read_write_addr
725 -seg_addr_table -seg_addr_table_filename -seglinkedit
726 -segprot -segs_read_only_addr -segs_read_write_addr
727 -single_module -static -sub_library -sub_umbrella
728 -twolevel_namespace -umbrella -undefined
729 -unexported_symbols_list -weak_reference_mismatches
730 -whatsloaded -F -gused -gfull -mmacosx-version-min=VERSION
731 -mkernel -mone-byte-bool
734 -mno-fp-regs -msoft-float -malpha-as -mgas
735 -mieee -mieee-with-inexact -mieee-conformant
736 -mfp-trap-mode=MODE -mfp-rounding-mode=MODE
737 -mtrap-precision=MODE -mbuild-constants
738 -mcpu=CPU-TYPE -mtune=CPU-TYPE
739 -mbwx -mmax -mfix -mcix
740 -mfloat-vax -mfloat-ieee
741 -mexplicit-relocs -msmall-data -mlarge-data
742 -msmall-text -mlarge-text
743 -mmemory-latency=TIME
745 _DEC Alpha/VMS Options_
749 -mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64
750 -mhard-float -msoft-float
751 -malloc-cc -mfixed-cc -mdword -mno-dword
753 -mmedia -mno-media -mmuladd -mno-muladd
754 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic
755 -mlinked-fp -mlong-calls -malign-labels
756 -mlibrary-pic -macc-4 -macc-8
757 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move
758 -moptimize-membar -mno-optimize-membar
759 -mscc -mno-scc -mcond-exec -mno-cond-exec
760 -mvliw-branch -mno-vliw-branch
761 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec
762 -mno-nested-cond-exec -mtomcat-stats
770 -mrelax -mh -ms -mn -mint32 -malign-300
773 -march=ARCHITECTURE-TYPE
774 -mbig-switch -mdisable-fpregs -mdisable-indexing
775 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld
776 -mfixed-range=REGISTER-RANGE
777 -mjump-in-delay -mlinker-opt -mlong-calls
778 -mlong-load-store -mno-big-switch -mno-disable-fpregs
779 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
780 -mno-jump-in-delay -mno-long-load-store
781 -mno-portable-runtime -mno-soft-float
782 -mno-space-regs -msoft-float -mpa-risc-1-0
783 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
784 -mschedule=CPU-TYPE -mspace-regs -msio -mwsio
785 -munix=UNIX-STD -nolibdld -static -threads
787 _i386 and x86-64 Options_
788 -mtune=CPU-TYPE -march=CPU-TYPE
790 -masm=DIALECT -mno-fancy-math-387
791 -mno-fp-ret-in-387 -msoft-float
792 -mno-wide-multiply -mrtd -malign-double
793 -mpreferred-stack-boundary=NUM -mcld -mcx16 -msahf -mrecip
794 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4
795 -msse4a -m3dnow -mpopcnt -mabm -msse5
796 -mthreads -mno-align-stringops -minline-all-stringops
797 -mpush-args -maccumulate-outgoing-args -m128bit-long-double
798 -m96bit-long-double -mregparm=NUM -msseregparm
799 -mveclibabi=TYPE -mpc32 -mpc64 -mpc80 -mstackrealign
800 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs
802 -m32 -m64 -mlarge-data-threshold=NUM
803 -mfused-madd -mno-fused-madd
806 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic
807 -mvolatile-asm-stop -mregister-names -mno-sdata
808 -mconstant-gp -mauto-pic -minline-float-divide-min-latency
809 -minline-float-divide-max-throughput
810 -minline-int-divide-min-latency
811 -minline-int-divide-max-throughput
812 -minline-sqrt-min-latency -minline-sqrt-max-throughput
813 -mno-dwarf2-asm -mearly-stop-bits
814 -mfixed-range=REGISTER-RANGE -mtls-size=TLS-SIZE
815 -mtune=CPU-TYPE -mt -pthread -milp32 -mlp64
816 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec
817 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec
818 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose
819 -mno-sched-prefer-non-data-spec-insns
820 -mno-sched-prefer-non-control-spec-insns
821 -mno-sched-count-spec-in-critical-path
826 -malign-loops -mno-align-loops
829 -mmodel=CODE-SIZE-MODEL-TYPE
831 -mno-flush-func -mflush-func=NAME
832 -mno-flush-trap -mflush-trap=NUMBER
836 -mcpu=CPU -msim -memregs=NUMBER
839 -march=ARCH -mcpu=CPU -mtune=TUNE
840 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
841 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407
842 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020
843 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort
844 -mno-short -mhard-float -m68881 -msoft-float -mpcrel
845 -malign-int -mstrict-align -msep-data -mno-sep-data
846 -mshared-library-id=n -mid-shared-library -mno-id-shared-library
849 -m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12
850 -mauto-incdec -minmax -mlong-calls -mshort
851 -msoft-reg-count=COUNT
854 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates
855 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
856 -m4byte-functions -mno-4byte-functions -mcallgraph-data
857 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
858 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
861 -EL -EB -march=ARCH -mtune=ARCH
862 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64
863 -mips16 -mno-mips16 -mflip-mips16
864 -minterlink-mips16 -mno-interlink-mips16
865 -mabi=ABI -mabicalls -mno-abicalls
866 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64
867 -mfp32 -mfp64 -mhard-float -msoft-float
868 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2
869 -msmartmips -mno-smartmips
870 -mpaired-single -mno-paired-single -mdmx -mno-mdmx
871 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc
872 -mlong64 -mlong32 -msym32 -mno-sym32
873 -GNUM -mlocal-sdata -mno-local-sdata
874 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt
875 -membedded-data -mno-embedded-data
876 -muninit-const-in-rodata -mno-uninit-const-in-rodata
877 -mcode-readable=SETTING
878 -msplit-addresses -mno-split-addresses
879 -mexplicit-relocs -mno-explicit-relocs
880 -mcheck-zero-division -mno-check-zero-division
881 -mdivide-traps -mdivide-breaks
882 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls
883 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp
884 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400
885 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130
886 -mfix-sb1 -mno-fix-sb1
887 -mflush-func=FUNC -mno-flush-func
888 -mbranch-cost=NUM -mbranch-likely -mno-branch-likely
889 -mfp-exceptions -mno-fp-exceptions
890 -mvr4130-align -mno-vr4130-align
893 -mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu
894 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols
895 -melf -mbranch-predict -mno-branch-predict -mbase-addresses
896 -mno-base-addresses -msingle-exit -mno-single-exit
899 -mmult-bug -mno-mult-bug
902 -mreturn-pointer-on-d0
906 -mno-crt0 -mbacc -msim
910 -mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10
911 -mbcopy -mbcopy-builtin -mint32 -mno-int16
912 -mint16 -mno-int32 -mfloat32 -mno-float64
913 -mfloat64 -mno-float32 -mabshi -mno-abshi
914 -mbranch-expensive -mbranch-cheap
915 -msplit -mno-split -munix-asm -mdec-asm
917 _PowerPC Options_ See RS/6000 and PowerPC Options.
919 _RS/6000 and PowerPC Options_
922 -mpower -mno-power -mpower2 -mno-power2
923 -mpowerpc -mpowerpc64 -mno-powerpc
924 -maltivec -mno-altivec
925 -mpowerpc-gpopt -mno-powerpc-gpopt
926 -mpowerpc-gfxopt -mno-powerpc-gfxopt
927 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd
928 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp
929 -mnew-mnemonics -mold-mnemonics
930 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc
931 -m64 -m32 -mxl-compat -mno-xl-compat -mpe
932 -malign-power -malign-natural
933 -msoft-float -mhard-float -mmultiple -mno-multiple
934 -mstring -mno-string -mupdate -mno-update
935 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
936 -mstrict-align -mno-strict-align -mrelocatable
937 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
938 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
939 -mdynamic-no-pic -maltivec -mswdiv
940 -mprioritize-restricted-insns=PRIORITY
941 -msched-costly-dep=DEPENDENCE_TYPE
942 -minsert-sched-nops=SCHEME
943 -mcall-sysv -mcall-netbsd
944 -maix-struct-return -msvr4-struct-return
945 -mabi=ABI-TYPE -msecure-plt -mbss-plt
954 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double
955 -mprototype -mno-prototype
956 -msim -mmvme -mads -myellowknife -memb -msdata
957 -msdata=OPT -mvxworks -mwindiss -G NUM -pthread
959 _S/390 and zSeries Options_
960 -mtune=CPU-TYPE -march=CPU-TYPE
961 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128
962 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack
963 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle
964 -m64 -m31 -mdebug -mno-debug -mesa -mzarch
965 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd
966 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard
973 -mscore5 -mscore5u -mscore7 -mscore7d
976 -m1 -m2 -m2e -m3 -m3e
977 -m4-nofpu -m4-single-only -m4-single -m4
978 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al
979 -m5-64media -m5-64media-nofpu
980 -m5-32media -m5-32media-nofpu
981 -m5-compact -m5-compact-nofpu
982 -mb -ml -mdalign -mrelax
983 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave
984 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace
985 -mprefergot -musermode -multcost=NUMBER -mdiv=STRATEGY
986 -mdivsi3_libfunc=NAME
987 -madjust-unroll -mindexed-addressing -mgettrcost=NUMBER -mpt-fixed
994 -m32 -m64 -mapp-regs -mno-app-regs
995 -mfaster-structs -mno-faster-structs
996 -mfpu -mno-fpu -mhard-float -msoft-float
997 -mhard-quad-float -msoft-quad-float
998 -mimpure-text -mno-impure-text -mlittle-endian
999 -mstack-bias -mno-stack-bias
1000 -munaligned-doubles -mno-unaligned-doubles
1001 -mv8plus -mno-v8plus -mvis -mno-vis
1002 -threads -pthreads -pthread
1005 -mwarn-reloc -merror-reloc
1006 -msafe-dma -munsafe-dma
1008 -msmall-mem -mlarge-mem -mstdmain
1009 -mfixed-range=REGISTER-RANGE
1012 -Qy -Qn -YP,PATHS -Ym,DIR
1015 -mlong-calls -mno-long-calls -mep -mno-ep
1016 -mprolog-function -mno-prolog-function -mspace
1017 -mtda=N -msda=N -mzda=N
1018 -mapp-regs -mno-app-regs
1019 -mdisable-callt -mno-disable-callt
1028 -mrtp -non-static -Bstatic -Bdynamic
1029 -Xbind-lazy -Xbind-now
1031 _x86-64 Options_ See i386 and x86-64 Options.
1037 -mconst16 -mno-const16
1038 -mfused-madd -mno-fused-madd
1039 -mtext-section-literals -mno-text-section-literals
1040 -mtarget-align -mno-target-align
1041 -mlongcalls -mno-longcalls
1043 _zSeries Options_ See S/390 and zSeries Options.
1045 _Code Generation Options_
1046 *Note Options for Code Generation Conventions: Code Gen Options.
1047 -fcall-saved-REG -fcall-used-REG
1048 -ffixed-REG -fexceptions
1049 -fnon-call-exceptions -funwind-tables
1050 -fasynchronous-unwind-tables
1051 -finhibit-size-directive -finstrument-functions
1052 -finstrument-functions-exclude-function-list=SYM,SYM,...
1053 -finstrument-functions-exclude-file-list=FILE,FILE,...
1054 -fno-common -fno-ident
1055 -fpcc-struct-return -fpic -fPIC -fpie -fPIE
1057 -frecord-gcc-switches
1058 -freg-struct-return -fshort-enums
1059 -fshort-double -fshort-wchar
1060 -fverbose-asm -fpack-struct[=N] -fstack-check
1061 -fstack-limit-register=REG -fstack-limit-symbol=SYM
1062 -fno-stack-limit -fargument-alias -fargument-noalias
1063 -fargument-noalias-global -fargument-noalias-anything
1064 -fleading-underscore -ftls-model=MODEL
1065 -ftrapv -fwrapv -fbounds-check
1071 * Overall Options:: Controlling the kind of output:
1072 an executable, object files, assembler files,
1073 or preprocessed source.
1074 * C Dialect Options:: Controlling the variant of C language compiled.
1075 * C++ Dialect Options:: Variations on C++.
1076 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1078 * Language Independent Options:: Controlling how diagnostics should be
1080 * Warning Options:: How picky should the compiler be?
1081 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1082 * Optimize Options:: How much optimization?
1083 * Preprocessor Options:: Controlling header files and macro definitions.
1084 Also, getting dependency information for Make.
1085 * Assembler Options:: Passing options to the assembler.
1086 * Link Options:: Specifying libraries and so on.
1087 * Directory Options:: Where to find header files and libraries.
1088 Where to find the compiler executable files.
1089 * Spec Files:: How to pass switches to sub-processes.
1090 * Target Options:: Running a cross-compiler, or an old version of GCC.
1093 File: gcc.info, Node: Overall Options, Next: Invoking G++, Prev: Option Summary, Up: Invoking GCC
1095 3.2 Options Controlling the Kind of Output
1096 ==========================================
1098 Compilation can involve up to four stages: preprocessing, compilation
1099 proper, assembly and linking, always in that order. GCC is capable of
1100 preprocessing and compiling several files either into several assembler
1101 input files, or into one assembler input file; then each assembler
1102 input file produces an object file, and linking combines all the object
1103 files (those newly compiled, and those specified as input) into an
1106 For any given input file, the file name suffix determines what kind of
1107 compilation is done:
1110 C source code which must be preprocessed.
1113 C source code which should not be preprocessed.
1116 C++ source code which should not be preprocessed.
1119 Objective-C source code. Note that you must link with the
1120 `libobjc' library to make an Objective-C program work.
1123 Objective-C source code which should not be preprocessed.
1127 Objective-C++ source code. Note that you must link with the
1128 `libobjc' library to make an Objective-C++ program work. Note
1129 that `.M' refers to a literal capital M.
1132 Objective-C++ source code which should not be preprocessed.
1135 C, C++, Objective-C or Objective-C++ header file to be turned into
1136 a precompiled header.
1145 C++ source code which must be preprocessed. Note that in `.cxx',
1146 the last two letters must both be literally `x'. Likewise, `.C'
1147 refers to a literal capital C.
1151 Objective-C++ source code which must be preprocessed.
1154 Objective-C++ source code which should not be preprocessed.
1164 C++ header file to be turned into a precompiled header.
1169 Fixed form Fortran source code which should not be preprocessed.
1174 Fixed form Fortran source code which must be preprocessed (with
1175 the traditional preprocessor).
1179 Free form Fortran source code which should not be preprocessed.
1183 Free form Fortran source code which must be preprocessed (with the
1184 traditional preprocessor).
1187 Ada source code file which contains a library unit declaration (a
1188 declaration of a package, subprogram, or generic, or a generic
1189 instantiation), or a library unit renaming declaration (a package,
1190 generic, or subprogram renaming declaration). Such files are also
1194 Ada source code file containing a library unit body (a subprogram
1195 or package body). Such files are also called "bodies".
1202 Assembler code which must be preprocessed.
1205 An object file to be fed straight into linking. Any file name
1206 with no recognized suffix is treated this way.
1208 You can specify the input language explicitly with the `-x' option:
1211 Specify explicitly the LANGUAGE for the following input files
1212 (rather than letting the compiler choose a default based on the
1213 file name suffix). This option applies to all following input
1214 files until the next `-x' option. Possible values for LANGUAGE
1216 c c-header c-cpp-output
1217 c++ c++-header c++-cpp-output
1218 objective-c objective-c-header objective-c-cpp-output
1219 objective-c++ objective-c++-header objective-c++-cpp-output
1220 assembler assembler-with-cpp
1227 Turn off any specification of a language, so that subsequent files
1228 are handled according to their file name suffixes (as they are if
1229 `-x' has not been used at all).
1232 Normally the `gcc' program will exit with the code of 1 if any
1233 phase of the compiler returns a non-success return code. If you
1234 specify `-pass-exit-codes', the `gcc' program will instead return
1235 with numerically highest error produced by any phase that returned
1236 an error indication. The C, C++, and Fortran frontends return 4,
1237 if an internal compiler error is encountered.
1239 If you only want some of the stages of compilation, you can use `-x'
1240 (or filename suffixes) to tell `gcc' where to start, and one of the
1241 options `-c', `-S', or `-E' to say where `gcc' is to stop. Note that
1242 some combinations (for example, `-x cpp-output -E') instruct `gcc' to
1246 Compile or assemble the source files, but do not link. The linking
1247 stage simply is not done. The ultimate output is in the form of an
1248 object file for each source file.
1250 By default, the object file name for a source file is made by
1251 replacing the suffix `.c', `.i', `.s', etc., with `.o'.
1253 Unrecognized input files, not requiring compilation or assembly,
1257 Stop after the stage of compilation proper; do not assemble. The
1258 output is in the form of an assembler code file for each
1259 non-assembler input file specified.
1261 By default, the assembler file name for a source file is made by
1262 replacing the suffix `.c', `.i', etc., with `.s'.
1264 Input files that don't require compilation are ignored.
1267 Stop after the preprocessing stage; do not run the compiler
1268 proper. The output is in the form of preprocessed source code,
1269 which is sent to the standard output.
1271 Input files which don't require preprocessing are ignored.
1274 Place output in file FILE. This applies regardless to whatever
1275 sort of output is being produced, whether it be an executable file,
1276 an object file, an assembler file or preprocessed C code.
1278 If `-o' is not specified, the default is to put an executable file
1279 in `a.out', the object file for `SOURCE.SUFFIX' in `SOURCE.o', its
1280 assembler file in `SOURCE.s', a precompiled header file in
1281 `SOURCE.SUFFIX.gch', and all preprocessed C source on standard
1285 Print (on standard error output) the commands executed to run the
1286 stages of compilation. Also print the version number of the
1287 compiler driver program and of the preprocessor and the compiler
1291 Like `-v' except the commands are not executed and all command
1292 arguments are quoted. This is useful for shell scripts to capture
1293 the driver-generated command lines.
1296 Use pipes rather than temporary files for communication between the
1297 various stages of compilation. This fails to work on some systems
1298 where the assembler is unable to read from a pipe; but the GNU
1299 assembler has no trouble.
1302 If you are compiling multiple source files, this option tells the
1303 driver to pass all the source files to the compiler at once (for
1304 those languages for which the compiler can handle this). This
1305 will allow intermodule analysis (IMA) to be performed by the
1306 compiler. Currently the only language for which this is supported
1307 is C. If you pass source files for multiple languages to the
1308 driver, using this option, the driver will invoke the compiler(s)
1309 that support IMA once each, passing each compiler all the source
1310 files appropriate for it. For those languages that do not support
1311 IMA this option will be ignored, and the compiler will be invoked
1312 once for each source file in that language. If you use this
1313 option in conjunction with `-save-temps', the compiler will
1314 generate multiple pre-processed files (one for each source file),
1315 but only one (combined) `.o' or `.s' file.
1318 Print (on the standard output) a description of the command line
1319 options understood by `gcc'. If the `-v' option is also specified
1320 then `--help' will also be passed on to the various processes
1321 invoked by `gcc', so that they can display the command line options
1322 they accept. If the `-Wextra' option has also been specified
1323 (prior to the `--help' option), then command line options which
1324 have no documentation associated with them will also be displayed.
1327 Print (on the standard output) a description of target-specific
1328 command line options for each tool. For some targets extra
1329 target-specific information may also be printed.
1331 `--help=CLASS[,QUALIFIER]'
1332 Print (on the standard output) a description of the command line
1333 options understood by the compiler that fit into a specific class.
1334 The class can be one of `optimizers', `warnings', `target',
1335 `params', or LANGUAGE:
1338 This will display all of the optimization options supported
1342 This will display all of the options controlling warning
1343 messages produced by the compiler.
1346 This will display target-specific options. Unlike the
1347 `--target-help' option however, target-specific options of the
1348 linker and assembler will not be displayed. This is because
1349 those tools do not currently support the extended `--help='
1353 This will display the values recognized by the `--param'
1357 This will display the options supported for LANGUAGE, where
1358 LANGUAGE is the name of one of the languages supported in this
1362 This will display the options that are common to all
1365 It is possible to further refine the output of the `--help='
1366 option by adding a comma separated list of qualifiers after the
1367 class. These can be any from the following list:
1370 Display only those options which are undocumented.
1373 Display options which take an argument that appears after an
1374 equal sign in the same continuous piece of text, such as:
1378 Display options which take an argument that appears as a
1379 separate word following the original option, such as: `-o
1382 Thus for example to display all the undocumented target-specific
1383 switches supported by the compiler the following can be used:
1385 --help=target,undocumented
1387 The sense of a qualifier can be inverted by prefixing it with the
1388 ^ character, so for example to display all binary warning options
1389 (i.e., ones that are either on or off and that do not take an
1390 argument), which have a description the following can be used:
1392 --help=warnings,^joined,^undocumented
1394 A class can also be used as a qualifier, although this usually
1395 restricts the output by so much that there is nothing to display.
1396 One case where it does work however is when one of the classes is
1397 TARGET. So for example to display all the target-specific
1398 optimization options the following can be used:
1400 --help=target,optimizers
1402 The `--help=' option can be repeated on the command line. Each
1403 successive use will display its requested class of options,
1404 skipping those that have already been displayed.
1406 If the `-Q' option appears on the command line before the
1407 `--help=' option, then the descriptive text displayed by `--help='
1408 is changed. Instead of describing the displayed options, an
1409 indication is given as to whether the option is enabled, disabled
1410 or set to a specific value (assuming that the compiler knows this
1411 at the point where the `--help=' option is used).
1413 Here is a truncated example from the ARM port of `gcc':
1415 % gcc -Q -mabi=2 --help=target -c
1416 The following options are target specific:
1418 -mabort-on-noreturn [disabled]
1421 The output is sensitive to the effects of previous command line
1422 options, so for example it is possible to find out which
1423 optimizations are enabled at `-O2' by using:
1425 -O2 --help=optimizers
1427 Alternatively you can discover which binary optimizations are
1428 enabled by `-O3' by using:
1430 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1431 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1432 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1435 Display the version number and copyrights of the invoked GCC.
1438 Read command-line options from FILE. The options read are
1439 inserted in place of the original @FILE option. If FILE does not
1440 exist, or cannot be read, then the option will be treated
1441 literally, and not removed.
1443 Options in FILE are separated by whitespace. A whitespace
1444 character may be included in an option by surrounding the entire
1445 option in either single or double quotes. Any character
1446 (including a backslash) may be included by prefixing the character
1447 to be included with a backslash. The FILE may itself contain
1448 additional @FILE options; any such options will be processed
1452 File: gcc.info, Node: Invoking G++, Next: C Dialect Options, Prev: Overall Options, Up: Invoking GCC
1454 3.3 Compiling C++ Programs
1455 ==========================
1457 C++ source files conventionally use one of the suffixes `.C', `.cc',
1458 `.cpp', `.CPP', `.c++', `.cp', or `.cxx'; C++ header files often use
1459 `.hh', `.hpp', `.H', or (for shared template code) `.tcc'; and
1460 preprocessed C++ files use the suffix `.ii'. GCC recognizes files with
1461 these names and compiles them as C++ programs even if you call the
1462 compiler the same way as for compiling C programs (usually with the
1465 However, the use of `gcc' does not add the C++ library. `g++' is a
1466 program that calls GCC and treats `.c', `.h' and `.i' files as C++
1467 source files instead of C source files unless `-x' is used, and
1468 automatically specifies linking against the C++ library. This program
1469 is also useful when precompiling a C header file with a `.h' extension
1470 for use in C++ compilations. On many systems, `g++' is also installed
1471 with the name `c++'.
1473 When you compile C++ programs, you may specify many of the same
1474 command-line options that you use for compiling programs in any
1475 language; or command-line options meaningful for C and related
1476 languages; or options that are meaningful only for C++ programs. *Note
1477 Options Controlling C Dialect: C Dialect Options, for explanations of
1478 options for languages related to C. *Note Options Controlling C++
1479 Dialect: C++ Dialect Options, for explanations of options that are
1480 meaningful only for C++ programs.
1483 File: gcc.info, Node: C Dialect Options, Next: C++ Dialect Options, Prev: Invoking G++, Up: Invoking GCC
1485 3.4 Options Controlling C Dialect
1486 =================================
1488 The following options control the dialect of C (or languages derived
1489 from C, such as C++, Objective-C and Objective-C++) that the compiler
1493 In C mode, this is equivalent to `-std=c89'. In C++ mode, it is
1494 equivalent to `-std=c++98'.
1496 This turns off certain features of GCC that are incompatible with
1497 ISO C90 (when compiling C code), or of standard C++ (when
1498 compiling C++ code), such as the `asm' and `typeof' keywords, and
1499 predefined macros such as `unix' and `vax' that identify the type
1500 of system you are using. It also enables the undesirable and
1501 rarely used ISO trigraph feature. For the C compiler, it disables
1502 recognition of C++ style `//' comments as well as the `inline'
1505 The alternate keywords `__asm__', `__extension__', `__inline__'
1506 and `__typeof__' continue to work despite `-ansi'. You would not
1507 want to use them in an ISO C program, of course, but it is useful
1508 to put them in header files that might be included in compilations
1509 done with `-ansi'. Alternate predefined macros such as `__unix__'
1510 and `__vax__' are also available, with or without `-ansi'.
1512 The `-ansi' option does not cause non-ISO programs to be rejected
1513 gratuitously. For that, `-pedantic' is required in addition to
1514 `-ansi'. *Note Warning Options::.
1516 The macro `__STRICT_ANSI__' is predefined when the `-ansi' option
1517 is used. Some header files may notice this macro and refrain from
1518 declaring certain functions or defining certain macros that the
1519 ISO standard doesn't call for; this is to avoid interfering with
1520 any programs that might use these names for other things.
1522 Functions that would normally be built in but do not have semantics
1523 defined by ISO C (such as `alloca' and `ffs') are not built-in
1524 functions when `-ansi' is used. *Note Other built-in functions
1525 provided by GCC: Other Builtins, for details of the functions
1529 Determine the language standard. *Note Language Standards
1530 Supported by GCC: Standards, for details of these standard
1531 versions. This option is currently only supported when compiling
1534 The compiler can accept several base standards, such as `c89' or
1535 `c++98', and GNU dialects of those standards, such as `gnu89' or
1536 `gnu++98'. By specifing a base standard, the compiler will accept
1537 all programs following that standard and those using GNU
1538 extensions that do not contradict it. For example, `-std=c89'
1539 turns off certain features of GCC that are incompatible with ISO
1540 C90, such as the `asm' and `typeof' keywords, but not other GNU
1541 extensions that do not have a meaning in ISO C90, such as omitting
1542 the middle term of a `?:' expression. On the other hand, by
1543 specifing a GNU dialect of a standard, all features the compiler
1544 support are enabled, even when those features change the meaning
1545 of the base standard and some strict-conforming programs may be
1546 rejected. The particular standard is used by `-pedantic' to
1547 identify which features are GNU extensions given that version of
1548 the standard. For example `-std=gnu89 -pedantic' would warn about
1549 C++ style `//' comments, while `-std=gnu99 -pedantic' would not.
1551 A value for this option must be provided; possible values are
1555 Support all ISO C90 programs (certain GNU extensions that
1556 conflict with ISO C90 are disabled). Same as `-ansi' for C
1560 ISO C90 as modified in amendment 1.
1566 ISO C99. Note that this standard is not yet fully supported;
1567 see `http://gcc.gnu.org/gcc-4.3/c99status.html' for more
1568 information. The names `c9x' and `iso9899:199x' are
1572 GNU dialect of ISO C90 (including some C99 features). This is
1573 the default for C code.
1577 GNU dialect of ISO C99. When ISO C99 is fully implemented in
1578 GCC, this will become the default. The name `gnu9x' is
1582 The 1998 ISO C++ standard plus amendments. Same as `-ansi' for
1586 GNU dialect of `-std=c++98'. This is the default for C++
1590 The working draft of the upcoming ISO C++0x standard. This
1591 option enables experimental features that are likely to be
1592 included in C++0x. The working draft is constantly changing,
1593 and any feature that is enabled by this flag may be removed
1594 from future versions of GCC if it is not part of the C++0x
1598 GNU dialect of `-std=c++0x'. This option enables experimental
1599 features that may be removed in future versions of GCC.
1602 The option `-fgnu89-inline' tells GCC to use the traditional GNU
1603 semantics for `inline' functions when in C99 mode. *Note An
1604 Inline Function is As Fast As a Macro: Inline. This option is
1605 accepted and ignored by GCC versions 4.1.3 up to but not including
1606 4.3. In GCC versions 4.3 and later it changes the behavior of GCC
1607 in C99 mode. Using this option is roughly equivalent to adding the
1608 `gnu_inline' function attribute to all inline functions (*note
1609 Function Attributes::).
1611 The option `-fno-gnu89-inline' explicitly tells GCC to use the C99
1612 semantics for `inline' when in C99 or gnu99 mode (i.e., it
1613 specifies the default behavior). This option was first supported
1614 in GCC 4.3. This option is not supported in C89 or gnu89 mode.
1616 The preprocessor macros `__GNUC_GNU_INLINE__' and
1617 `__GNUC_STDC_INLINE__' may be used to check which semantics are in
1618 effect for `inline' functions. *Note Common Predefined Macros:
1619 (cpp)Common Predefined Macros.
1621 `-aux-info FILENAME'
1622 Output to the given filename prototyped declarations for all
1623 functions declared and/or defined in a translation unit, including
1624 those in header files. This option is silently ignored in any
1625 language other than C.
1627 Besides declarations, the file indicates, in comments, the origin
1628 of each declaration (source file and line), whether the
1629 declaration was implicit, prototyped or unprototyped (`I', `N' for
1630 new or `O' for old, respectively, in the first character after the
1631 line number and the colon), and whether it came from a declaration
1632 or a definition (`C' or `F', respectively, in the following
1633 character). In the case of function definitions, a K&R-style list
1634 of arguments followed by their declarations is also provided,
1635 inside comments, after the declaration.
1638 Do not recognize `asm', `inline' or `typeof' as a keyword, so that
1639 code can use these words as identifiers. You can use the keywords
1640 `__asm__', `__inline__' and `__typeof__' instead. `-ansi' implies
1643 In C++, this switch only affects the `typeof' keyword, since `asm'
1644 and `inline' are standard keywords. You may want to use the
1645 `-fno-gnu-keywords' flag instead, which has the same effect. In
1646 C99 mode (`-std=c99' or `-std=gnu99'), this switch only affects
1647 the `asm' and `typeof' keywords, since `inline' is a standard
1651 `-fno-builtin-FUNCTION'
1652 Don't recognize built-in functions that do not begin with
1653 `__builtin_' as prefix. *Note Other built-in functions provided
1654 by GCC: Other Builtins, for details of the functions affected,
1655 including those which are not built-in functions when `-ansi' or
1656 `-std' options for strict ISO C conformance are used because they
1657 do not have an ISO standard meaning.
1659 GCC normally generates special code to handle certain built-in
1660 functions more efficiently; for instance, calls to `alloca' may
1661 become single instructions that adjust the stack directly, and
1662 calls to `memcpy' may become inline copy loops. The resulting
1663 code is often both smaller and faster, but since the function
1664 calls no longer appear as such, you cannot set a breakpoint on
1665 those calls, nor can you change the behavior of the functions by
1666 linking with a different library. In addition, when a function is
1667 recognized as a built-in function, GCC may use information about
1668 that function to warn about problems with calls to that function,
1669 or to generate more efficient code, even if the resulting code
1670 still contains calls to that function. For example, warnings are
1671 given with `-Wformat' for bad calls to `printf', when `printf' is
1672 built in, and `strlen' is known not to modify global memory.
1674 With the `-fno-builtin-FUNCTION' option only the built-in function
1675 FUNCTION is disabled. FUNCTION must not begin with `__builtin_'.
1676 If a function is named this is not built-in in this version of
1677 GCC, this option is ignored. There is no corresponding
1678 `-fbuiltin-FUNCTION' option; if you wish to enable built-in
1679 functions selectively when using `-fno-builtin' or
1680 `-ffreestanding', you may define macros such as:
1682 #define abs(n) __builtin_abs ((n))
1683 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1686 Assert that compilation takes place in a hosted environment. This
1687 implies `-fbuiltin'. A hosted environment is one in which the
1688 entire standard library is available, and in which `main' has a
1689 return type of `int'. Examples are nearly everything except a
1690 kernel. This is equivalent to `-fno-freestanding'.
1693 Assert that compilation takes place in a freestanding environment.
1694 This implies `-fno-builtin'. A freestanding environment is one
1695 in which the standard library may not exist, and program startup
1696 may not necessarily be at `main'. The most obvious example is an
1697 OS kernel. This is equivalent to `-fno-hosted'.
1699 *Note Language Standards Supported by GCC: Standards, for details
1700 of freestanding and hosted environments.
1703 Enable handling of OpenMP directives `#pragma omp' in C/C++ and
1704 `!$omp' in Fortran. When `-fopenmp' is specified, the compiler
1705 generates parallel code according to the OpenMP Application
1706 Program Interface v2.5 `http://www.openmp.org/'. This option
1707 implies `-pthread', and thus is only supported on targets that
1708 have support for `-pthread'.
1711 Accept some non-standard constructs used in Microsoft header files.
1713 Some cases of unnamed fields in structures and unions are only
1714 accepted with this option. *Note Unnamed struct/union fields
1715 within structs/unions: Unnamed Fields, for details.
1718 Support ISO C trigraphs. The `-ansi' option (and `-std' options
1719 for strict ISO C conformance) implies `-trigraphs'.
1721 `-no-integrated-cpp'
1722 Performs a compilation in two passes: preprocessing and compiling.
1723 This option allows a user supplied "cc1", "cc1plus", or "cc1obj"
1724 via the `-B' option. The user supplied compilation step can then
1725 add in an additional preprocessing step after normal preprocessing
1726 but before compiling. The default is to use the integrated cpp
1729 The semantics of this option will change if "cc1", "cc1plus", and
1730 "cc1obj" are merged.
1734 Formerly, these options caused GCC to attempt to emulate a
1735 pre-standard C compiler. They are now only supported with the
1736 `-E' switch. The preprocessor continues to support a pre-standard
1737 mode. See the GNU CPP manual for details.
1740 Allow conditional expressions with mismatched types in the second
1741 and third arguments. The value of such an expression is void.
1742 This option is not supported for C++.
1744 `-flax-vector-conversions'
1745 Allow implicit conversions between vectors with differing numbers
1746 of elements and/or incompatible element types. This option should
1747 not be used for new code.
1750 Let the type `char' be unsigned, like `unsigned char'.
1752 Each kind of machine has a default for what `char' should be. It
1753 is either like `unsigned char' by default or like `signed char' by
1756 Ideally, a portable program should always use `signed char' or
1757 `unsigned char' when it depends on the signedness of an object.
1758 But many programs have been written to use plain `char' and expect
1759 it to be signed, or expect it to be unsigned, depending on the
1760 machines they were written for. This option, and its inverse, let
1761 you make such a program work with the opposite default.
1763 The type `char' is always a distinct type from each of `signed
1764 char' or `unsigned char', even though its behavior is always just
1765 like one of those two.
1768 Let the type `char' be signed, like `signed char'.
1770 Note that this is equivalent to `-fno-unsigned-char', which is the
1771 negative form of `-funsigned-char'. Likewise, the option
1772 `-fno-signed-char' is equivalent to `-funsigned-char'.
1774 `-fsigned-bitfields'
1775 `-funsigned-bitfields'
1776 `-fno-signed-bitfields'
1777 `-fno-unsigned-bitfields'
1778 These options control whether a bit-field is signed or unsigned,
1779 when the declaration does not use either `signed' or `unsigned'.
1780 By default, such a bit-field is signed, because this is
1781 consistent: the basic integer types such as `int' are signed types.
1784 File: gcc.info, Node: C++ Dialect Options, Next: Objective-C and Objective-C++ Dialect Options, Prev: C Dialect Options, Up: Invoking GCC
1786 3.5 Options Controlling C++ Dialect
1787 ===================================
1789 This section describes the command-line options that are only meaningful
1790 for C++ programs; but you can also use most of the GNU compiler options
1791 regardless of what language your program is in. For example, you might
1792 compile a file `firstClass.C' like this:
1794 g++ -g -frepo -O -c firstClass.C
1796 In this example, only `-frepo' is an option meant only for C++
1797 programs; you can use the other options with any language supported by
1800 Here is a list of options that are _only_ for compiling C++ programs:
1803 Use version N of the C++ ABI. Version 2 is the version of the C++
1804 ABI that first appeared in G++ 3.4. Version 1 is the version of
1805 the C++ ABI that first appeared in G++ 3.2. Version 0 will always
1806 be the version that conforms most closely to the C++ ABI
1807 specification. Therefore, the ABI obtained using version 0 will
1808 change as ABI bugs are fixed.
1810 The default is version 2.
1812 `-fno-access-control'
1813 Turn off all access checking. This switch is mainly useful for
1814 working around bugs in the access control code.
1817 Check that the pointer returned by `operator new' is non-null
1818 before attempting to modify the storage allocated. This check is
1819 normally unnecessary because the C++ standard specifies that
1820 `operator new' will only return `0' if it is declared `throw()',
1821 in which case the compiler will always check the return value even
1822 without this option. In all other cases, when `operator new' has
1823 a non-empty exception specification, memory exhaustion is
1824 signalled by throwing `std::bad_alloc'. See also `new (nothrow)'.
1827 Put uninitialized or runtime-initialized global variables into the
1828 common segment, as C does. This saves space in the executable at
1829 the cost of not diagnosing duplicate definitions. If you compile
1830 with this flag and your program mysteriously crashes after
1831 `main()' has completed, you may have an object that is being
1832 destroyed twice because two definitions were merged.
1834 This option is no longer useful on most targets, now that support
1835 has been added for putting variables into BSS without making them
1838 `-ffriend-injection'
1839 Inject friend functions into the enclosing namespace, so that they
1840 are visible outside the scope of the class in which they are
1841 declared. Friend functions were documented to work this way in
1842 the old Annotated C++ Reference Manual, and versions of G++ before
1843 4.1 always worked that way. However, in ISO C++ a friend function
1844 which is not declared in an enclosing scope can only be found
1845 using argument dependent lookup. This option causes friends to be
1846 injected as they were in earlier releases.
1848 This option is for compatibility, and may be removed in a future
1851 `-fno-elide-constructors'
1852 The C++ standard allows an implementation to omit creating a
1853 temporary which is only used to initialize another object of the
1854 same type. Specifying this option disables that optimization, and
1855 forces G++ to call the copy constructor in all cases.
1857 `-fno-enforce-eh-specs'
1858 Don't generate code to check for violation of exception
1859 specifications at runtime. This option violates the C++ standard,
1860 but may be useful for reducing code size in production builds,
1861 much like defining `NDEBUG'. This does not give user code
1862 permission to throw exceptions in violation of the exception
1863 specifications; the compiler will still optimize based on the
1864 specifications, so throwing an unexpected exception will result in
1869 If `-ffor-scope' is specified, the scope of variables declared in
1870 a for-init-statement is limited to the `for' loop itself, as
1871 specified by the C++ standard. If `-fno-for-scope' is specified,
1872 the scope of variables declared in a for-init-statement extends to
1873 the end of the enclosing scope, as was the case in old versions of
1874 G++, and other (traditional) implementations of C++.
1876 The default if neither flag is given to follow the standard, but
1877 to allow and give a warning for old-style code that would
1878 otherwise be invalid, or have different behavior.
1881 Do not recognize `typeof' as a keyword, so that code can use this
1882 word as an identifier. You can use the keyword `__typeof__'
1883 instead. `-ansi' implies `-fno-gnu-keywords'.
1885 `-fno-implicit-templates'
1886 Never emit code for non-inline templates which are instantiated
1887 implicitly (i.e. by use); only emit code for explicit
1888 instantiations. *Note Template Instantiation::, for more
1891 `-fno-implicit-inline-templates'
1892 Don't emit code for implicit instantiations of inline templates,
1893 either. The default is to handle inlines differently so that
1894 compiles with and without optimization will need the same set of
1895 explicit instantiations.
1897 `-fno-implement-inlines'
1898 To save space, do not emit out-of-line copies of inline functions
1899 controlled by `#pragma implementation'. This will cause linker
1900 errors if these functions are not inlined everywhere they are
1904 Disable pedantic warnings about constructs used in MFC, such as
1905 implicit int and getting a pointer to member function via
1906 non-standard syntax.
1908 `-fno-nonansi-builtins'
1909 Disable built-in declarations of functions that are not mandated by
1910 ANSI/ISO C. These include `ffs', `alloca', `_exit', `index',
1911 `bzero', `conjf', and other related functions.
1913 `-fno-operator-names'
1914 Do not treat the operator name keywords `and', `bitand', `bitor',
1915 `compl', `not', `or' and `xor' as synonyms as keywords.
1917 `-fno-optional-diags'
1918 Disable diagnostics that the standard says a compiler does not
1919 need to issue. Currently, the only such diagnostic issued by G++
1920 is the one for a name having multiple meanings within a class.
1923 Downgrade some diagnostics about nonconformant code from errors to
1924 warnings. Thus, using `-fpermissive' will allow some
1925 nonconforming code to compile.
1928 Enable automatic template instantiation at link time. This option
1929 also implies `-fno-implicit-templates'. *Note Template
1930 Instantiation::, for more information.
1933 Disable generation of information about every class with virtual
1934 functions for use by the C++ runtime type identification features
1935 (`dynamic_cast' and `typeid'). If you don't use those parts of
1936 the language, you can save some space by using this flag. Note
1937 that exception handling uses the same information, but it will
1938 generate it as needed. The `dynamic_cast' operator can still be
1939 used for casts that do not require runtime type information, i.e.
1940 casts to `void *' or to unambiguous base classes.
1943 Emit statistics about front-end processing at the end of the
1944 compilation. This information is generally only useful to the G++
1947 `-ftemplate-depth-N'
1948 Set the maximum instantiation depth for template classes to N. A
1949 limit on the template instantiation depth is needed to detect
1950 endless recursions during template class instantiation. ANSI/ISO
1951 C++ conforming programs must not rely on a maximum depth greater
1954 `-fno-threadsafe-statics'
1955 Do not emit the extra code to use the routines specified in the C++
1956 ABI for thread-safe initialization of local statics. You can use
1957 this option to reduce code size slightly in code that doesn't need
1961 Register destructors for objects with static storage duration with
1962 the `__cxa_atexit' function rather than the `atexit' function.
1963 This option is required for fully standards-compliant handling of
1964 static destructors, but will only work if your C library supports
1967 `-fno-use-cxa-get-exception-ptr'
1968 Don't use the `__cxa_get_exception_ptr' runtime routine. This
1969 will cause `std::uncaught_exception' to be incorrect, but is
1970 necessary if the runtime routine is not available.
1972 `-fvisibility-inlines-hidden'
1973 This switch declares that the user does not attempt to compare
1974 pointers to inline methods where the addresses of the two functions
1975 were taken in different shared objects.
1977 The effect of this is that GCC may, effectively, mark inline
1978 methods with `__attribute__ ((visibility ("hidden")))' so that
1979 they do not appear in the export table of a DSO and do not require
1980 a PLT indirection when used within the DSO. Enabling this option
1981 can have a dramatic effect on load and link times of a DSO as it
1982 massively reduces the size of the dynamic export table when the
1983 library makes heavy use of templates.
1985 The behavior of this switch is not quite the same as marking the
1986 methods as hidden directly, because it does not affect static
1987 variables local to the function or cause the compiler to deduce
1988 that the function is defined in only one shared object.
1990 You may mark a method as having a visibility explicitly to negate
1991 the effect of the switch for that method. For example, if you do
1992 want to compare pointers to a particular inline method, you might
1993 mark it as having default visibility. Marking the enclosing class
1994 with explicit visibility will have no effect.
1996 Explicitly instantiated inline methods are unaffected by this
1997 option as their linkage might otherwise cross a shared library
1998 boundary. *Note Template Instantiation::.
2000 `-fvisibility-ms-compat'
2001 This flag attempts to use visibility settings to make GCC's C++
2002 linkage model compatible with that of Microsoft Visual Studio.
2004 The flag makes these changes to GCC's linkage model:
2006 1. It sets the default visibility to `hidden', like
2007 `-fvisibility=hidden'.
2009 2. Types, but not their members, are not hidden by default.
2011 3. The One Definition Rule is relaxed for types without explicit
2012 visibility specifications which are defined in more than one
2013 different shared object: those declarations are permitted if
2014 they would have been permitted when this option was not used.
2016 In new code it is better to use `-fvisibility=hidden' and export
2017 those classes which are intended to be externally visible.
2018 Unfortunately it is possible for code to rely, perhaps
2019 accidentally, on the Visual Studio behavior.
2021 Among the consequences of these changes are that static data
2022 members of the same type with the same name but defined in
2023 different shared objects will be different, so changing one will
2024 not change the other; and that pointers to function members
2025 defined in different shared objects may not compare equal. When
2026 this flag is given, it is a violation of the ODR to define types
2027 with the same name differently.
2030 Do not use weak symbol support, even if it is provided by the
2031 linker. By default, G++ will use weak symbols if they are
2032 available. This option exists only for testing, and should not be
2033 used by end-users; it will result in inferior code and has no
2034 benefits. This option may be removed in a future release of G++.
2037 Do not search for header files in the standard directories
2038 specific to C++, but do still search the other standard
2039 directories. (This option is used when building the C++ library.)
2041 In addition, these optimization, warning, and code generation options
2042 have meanings only for C++ programs:
2044 `-fno-default-inline'
2045 Do not assume `inline' for functions defined inside a class scope.
2046 *Note Options That Control Optimization: Optimize Options. Note
2047 that these functions will have linkage like inline functions; they
2048 just won't be inlined by default.
2050 `-Wabi (C++ and Objective-C++ only)'
2051 Warn when G++ generates code that is probably not compatible with
2052 the vendor-neutral C++ ABI. Although an effort has been made to
2053 warn about all such cases, there are probably some cases that are
2054 not warned about, even though G++ is generating incompatible code.
2055 There may also be cases where warnings are emitted even though
2056 the code that is generated will be compatible.
2058 You should rewrite your code to avoid these warnings if you are
2059 concerned about the fact that code generated by G++ may not be
2060 binary compatible with code generated by other compilers.
2062 The known incompatibilities at this point include:
2064 * Incorrect handling of tail-padding for bit-fields. G++ may
2065 attempt to pack data into the same byte as a base class. For
2068 struct A { virtual void f(); int f1 : 1; };
2069 struct B : public A { int f2 : 1; };
2071 In this case, G++ will place `B::f2' into the same byte
2072 as`A::f1'; other compilers will not. You can avoid this
2073 problem by explicitly padding `A' so that its size is a
2074 multiple of the byte size on your platform; that will cause
2075 G++ and other compilers to layout `B' identically.
2077 * Incorrect handling of tail-padding for virtual bases. G++
2078 does not use tail padding when laying out virtual bases. For
2081 struct A { virtual void f(); char c1; };
2082 struct B { B(); char c2; };
2083 struct C : public A, public virtual B {};
2085 In this case, G++ will not place `B' into the tail-padding for
2086 `A'; other compilers will. You can avoid this problem by
2087 explicitly padding `A' so that its size is a multiple of its
2088 alignment (ignoring virtual base classes); that will cause
2089 G++ and other compilers to layout `C' identically.
2091 * Incorrect handling of bit-fields with declared widths greater
2092 than that of their underlying types, when the bit-fields
2093 appear in a union. For example:
2095 union U { int i : 4096; };
2097 Assuming that an `int' does not have 4096 bits, G++ will make
2098 the union too small by the number of bits in an `int'.
2100 * Empty classes can be placed at incorrect offsets. For
2110 struct C : public B, public A {};
2112 G++ will place the `A' base class of `C' at a nonzero offset;
2113 it should be placed at offset zero. G++ mistakenly believes
2114 that the `A' data member of `B' is already at offset zero.
2116 * Names of template functions whose types involve `typename' or
2117 template template parameters can be mangled incorrectly.
2119 template <typename Q>
2120 void f(typename Q::X) {}
2122 template <template <typename> class Q>
2123 void f(typename Q<int>::X) {}
2125 Instantiations of these templates may be mangled incorrectly.
2128 `-Wctor-dtor-privacy (C++ and Objective-C++ only)'
2129 Warn when a class seems unusable because all the constructors or
2130 destructors in that class are private, and it has neither friends
2131 nor public static member functions.
2133 `-Wnon-virtual-dtor (C++ and Objective-C++ only)'
2134 Warn when a class has virtual functions and accessible non-virtual
2135 destructor, in which case it would be possible but unsafe to delete
2136 an instance of a derived class through a pointer to the base class.
2137 This warning is also enabled if -Weffc++ is specified.
2139 `-Wreorder (C++ and Objective-C++ only)'
2140 Warn when the order of member initializers given in the code does
2141 not match the order in which they must be executed. For instance:
2146 A(): j (0), i (1) { }
2149 The compiler will rearrange the member initializers for `i' and
2150 `j' to match the declaration order of the members, emitting a
2151 warning to that effect. This warning is enabled by `-Wall'.
2153 The following `-W...' options are not affected by `-Wall'.
2155 `-Weffc++ (C++ and Objective-C++ only)'
2156 Warn about violations of the following style guidelines from Scott
2157 Meyers' `Effective C++' book:
2159 * Item 11: Define a copy constructor and an assignment
2160 operator for classes with dynamically allocated memory.
2162 * Item 12: Prefer initialization to assignment in constructors.
2164 * Item 14: Make destructors virtual in base classes.
2166 * Item 15: Have `operator=' return a reference to `*this'.
2168 * Item 23: Don't try to return a reference when you must
2172 Also warn about violations of the following style guidelines from
2173 Scott Meyers' `More Effective C++' book:
2175 * Item 6: Distinguish between prefix and postfix forms of
2176 increment and decrement operators.
2178 * Item 7: Never overload `&&', `||', or `,'.
2181 When selecting this option, be aware that the standard library
2182 headers do not obey all of these guidelines; use `grep -v' to
2183 filter out those warnings.
2185 `-Wno-deprecated (C++ and Objective-C++ only)'
2186 Do not warn about usage of deprecated features. *Note Deprecated
2189 `-Wstrict-null-sentinel (C++ and Objective-C++ only)'
2190 Warn also about the use of an uncasted `NULL' as sentinel. When
2191 compiling only with GCC this is a valid sentinel, as `NULL' is
2192 defined to `__null'. Although it is a null pointer constant not a
2193 null pointer, it is guaranteed to of the same size as a pointer.
2194 But this use is not portable across different compilers.
2196 `-Wno-non-template-friend (C++ and Objective-C++ only)'
2197 Disable warnings when non-templatized friend functions are declared
2198 within a template. Since the advent of explicit template
2199 specification support in G++, if the name of the friend is an
2200 unqualified-id (i.e., `friend foo(int)'), the C++ language
2201 specification demands that the friend declare or define an
2202 ordinary, nontemplate function. (Section 14.5.3). Before G++
2203 implemented explicit specification, unqualified-ids could be
2204 interpreted as a particular specialization of a templatized
2205 function. Because this non-conforming behavior is no longer the
2206 default behavior for G++, `-Wnon-template-friend' allows the
2207 compiler to check existing code for potential trouble spots and is
2208 on by default. This new compiler behavior can be turned off with
2209 `-Wno-non-template-friend' which keeps the conformant compiler code
2210 but disables the helpful warning.
2212 `-Wold-style-cast (C++ and Objective-C++ only)'
2213 Warn if an old-style (C-style) cast to a non-void type is used
2214 within a C++ program. The new-style casts (`dynamic_cast',
2215 `static_cast', `reinterpret_cast', and `const_cast') are less
2216 vulnerable to unintended effects and much easier to search for.
2218 `-Woverloaded-virtual (C++ and Objective-C++ only)'
2219 Warn when a function declaration hides virtual functions from a
2220 base class. For example, in:
2226 struct B: public A {
2230 the `A' class version of `f' is hidden in `B', and code like:
2235 will fail to compile.
2237 `-Wno-pmf-conversions (C++ and Objective-C++ only)'
2238 Disable the diagnostic for converting a bound pointer to member
2239 function to a plain pointer.
2241 `-Wsign-promo (C++ and Objective-C++ only)'
2242 Warn when overload resolution chooses a promotion from unsigned or
2243 enumerated type to a signed type, over a conversion to an unsigned
2244 type of the same size. Previous versions of G++ would try to
2245 preserve unsignedness, but the standard mandates the current
2250 A& operator = (int);
2259 In this example, G++ will synthesize a default `A& operator =
2260 (const A&);', while cfront will use the user-defined `operator ='.
2263 File: gcc.info, Node: Objective-C and Objective-C++ Dialect Options, Next: Language Independent Options, Prev: C++ Dialect Options, Up: Invoking GCC
2265 3.6 Options Controlling Objective-C and Objective-C++ Dialects
2266 ==============================================================
2268 (NOTE: This manual does not describe the Objective-C and Objective-C++
2269 languages themselves. See *Note Language Standards Supported by GCC:
2270 Standards, for references.)
2272 This section describes the command-line options that are only
2273 meaningful for Objective-C and Objective-C++ programs, but you can also
2274 use most of the language-independent GNU compiler options. For
2275 example, you might compile a file `some_class.m' like this:
2277 gcc -g -fgnu-runtime -O -c some_class.m
2279 In this example, `-fgnu-runtime' is an option meant only for
2280 Objective-C and Objective-C++ programs; you can use the other options
2281 with any language supported by GCC.
2283 Note that since Objective-C is an extension of the C language,
2284 Objective-C compilations may also use options specific to the C
2285 front-end (e.g., `-Wtraditional'). Similarly, Objective-C++
2286 compilations may use C++-specific options (e.g., `-Wabi').
2288 Here is a list of options that are _only_ for compiling Objective-C
2289 and Objective-C++ programs:
2291 `-fconstant-string-class=CLASS-NAME'
2292 Use CLASS-NAME as the name of the class to instantiate for each
2293 literal string specified with the syntax `@"..."'. The default
2294 class name is `NXConstantString' if the GNU runtime is being used,
2295 and `NSConstantString' if the NeXT runtime is being used (see
2296 below). The `-fconstant-cfstrings' option, if also present, will
2297 override the `-fconstant-string-class' setting and cause `@"..."'
2298 literals to be laid out as constant CoreFoundation strings.
2301 Generate object code compatible with the standard GNU Objective-C
2302 runtime. This is the default for most types of systems.
2305 Generate output compatible with the NeXT runtime. This is the
2306 default for NeXT-based systems, including Darwin and Mac OS X.
2307 The macro `__NEXT_RUNTIME__' is predefined if (and only if) this
2310 `-fno-nil-receivers'
2311 Assume that all Objective-C message dispatches (e.g., `[receiver
2312 message:arg]') in this translation unit ensure that the receiver
2313 is not `nil'. This allows for more efficient entry points in the
2314 runtime to be used. Currently, this option is only available in
2315 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2317 `-fobjc-call-cxx-cdtors'
2318 For each Objective-C class, check if any of its instance variables
2319 is a C++ object with a non-trivial default constructor. If so,
2320 synthesize a special `- (id) .cxx_construct' instance method that
2321 will run non-trivial default constructors on any such instance
2322 variables, in order, and then return `self'. Similarly, check if
2323 any instance variable is a C++ object with a non-trivial
2324 destructor, and if so, synthesize a special `- (void)
2325 .cxx_destruct' method that will run all such default destructors,
2328 The `- (id) .cxx_construct' and/or `- (void) .cxx_destruct' methods
2329 thusly generated will only operate on instance variables declared
2330 in the current Objective-C class, and not those inherited from
2331 superclasses. It is the responsibility of the Objective-C runtime
2332 to invoke all such methods in an object's inheritance hierarchy.
2333 The `- (id) .cxx_construct' methods will be invoked by the runtime
2334 immediately after a new object instance is allocated; the `-
2335 (void) .cxx_destruct' methods will be invoked immediately before
2336 the runtime deallocates an object instance.
2338 As of this writing, only the NeXT runtime on Mac OS X 10.4 and
2339 later has support for invoking the `- (id) .cxx_construct' and `-
2340 (void) .cxx_destruct' methods.
2342 `-fobjc-direct-dispatch'
2343 Allow fast jumps to the message dispatcher. On Darwin this is
2344 accomplished via the comm page.
2347 Enable syntactic support for structured exception handling in
2348 Objective-C, similar to what is offered by C++ and Java. This
2349 option is unavailable in conjunction with the NeXT runtime on Mac
2350 OS X 10.2 and earlier.
2357 @catch (AnObjCClass *exc) {
2364 @catch (AnotherClass *exc) {
2367 @catch (id allOthers) {
2376 The `@throw' statement may appear anywhere in an Objective-C or
2377 Objective-C++ program; when used inside of a `@catch' block, the
2378 `@throw' may appear without an argument (as shown above), in which
2379 case the object caught by the `@catch' will be rethrown.
2381 Note that only (pointers to) Objective-C objects may be thrown and
2382 caught using this scheme. When an object is thrown, it will be
2383 caught by the nearest `@catch' clause capable of handling objects
2384 of that type, analogously to how `catch' blocks work in C++ and
2385 Java. A `@catch(id ...)' clause (as shown above) may also be
2386 provided to catch any and all Objective-C exceptions not caught by
2387 previous `@catch' clauses (if any).
2389 The `@finally' clause, if present, will be executed upon exit from
2390 the immediately preceding `@try ... @catch' section. This will
2391 happen regardless of whether any exceptions are thrown, caught or
2392 rethrown inside the `@try ... @catch' section, analogously to the
2393 behavior of the `finally' clause in Java.
2395 There are several caveats to using the new exception mechanism:
2397 * Although currently designed to be binary compatible with
2398 `NS_HANDLER'-style idioms provided by the `NSException'
2399 class, the new exceptions can only be used on Mac OS X 10.3
2400 (Panther) and later systems, due to additional functionality
2401 needed in the (NeXT) Objective-C runtime.
2403 * As mentioned above, the new exceptions do not support handling
2404 types other than Objective-C objects. Furthermore, when
2405 used from Objective-C++, the Objective-C exception model does
2406 not interoperate with C++ exceptions at this time. This
2407 means you cannot `@throw' an exception from Objective-C and
2408 `catch' it in C++, or vice versa (i.e., `throw ... @catch').
2410 The `-fobjc-exceptions' switch also enables the use of
2411 synchronization blocks for thread-safe execution:
2413 @synchronized (ObjCClass *guard) {
2417 Upon entering the `@synchronized' block, a thread of execution
2418 shall first check whether a lock has been placed on the
2419 corresponding `guard' object by another thread. If it has, the
2420 current thread shall wait until the other thread relinquishes its
2421 lock. Once `guard' becomes available, the current thread will
2422 place its own lock on it, execute the code contained in the
2423 `@synchronized' block, and finally relinquish the lock (thereby
2424 making `guard' available to other threads).
2426 Unlike Java, Objective-C does not allow for entire methods to be
2427 marked `@synchronized'. Note that throwing exceptions out of
2428 `@synchronized' blocks is allowed, and will cause the guarding
2429 object to be unlocked properly.
2432 Enable garbage collection (GC) in Objective-C and Objective-C++
2435 `-freplace-objc-classes'
2436 Emit a special marker instructing `ld(1)' not to statically link in
2437 the resulting object file, and allow `dyld(1)' to load it in at
2438 run time instead. This is used in conjunction with the
2439 Fix-and-Continue debugging mode, where the object file in question
2440 may be recompiled and dynamically reloaded in the course of
2441 program execution, without the need to restart the program itself.
2442 Currently, Fix-and-Continue functionality is only available in
2443 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2446 When compiling for the NeXT runtime, the compiler ordinarily
2447 replaces calls to `objc_getClass("...")' (when the name of the
2448 class is known at compile time) with static class references that
2449 get initialized at load time, which improves run-time performance.
2450 Specifying the `-fzero-link' flag suppresses this behavior and
2451 causes calls to `objc_getClass("...")' to be retained. This is
2452 useful in Zero-Link debugging mode, since it allows for individual
2453 class implementations to be modified during program execution.
2456 Dump interface declarations for all classes seen in the source
2457 file to a file named `SOURCENAME.decl'.
2459 `-Wassign-intercept (Objective-C and Objective-C++ only)'
2460 Warn whenever an Objective-C assignment is being intercepted by the
2463 `-Wno-protocol (Objective-C and Objective-C++ only)'
2464 If a class is declared to implement a protocol, a warning is
2465 issued for every method in the protocol that is not implemented by
2466 the class. The default behavior is to issue a warning for every
2467 method not explicitly implemented in the class, even if a method
2468 implementation is inherited from the superclass. If you use the
2469 `-Wno-protocol' option, then methods inherited from the superclass
2470 are considered to be implemented, and no warning is issued for
2473 `-Wselector (Objective-C and Objective-C++ only)'
2474 Warn if multiple methods of different types for the same selector
2475 are found during compilation. The check is performed on the list
2476 of methods in the final stage of compilation. Additionally, a
2477 check is performed for each selector appearing in a
2478 `@selector(...)' expression, and a corresponding method for that
2479 selector has been found during compilation. Because these checks
2480 scan the method table only at the end of compilation, these
2481 warnings are not produced if the final stage of compilation is not
2482 reached, for example because an error is found during compilation,
2483 or because the `-fsyntax-only' option is being used.
2485 `-Wstrict-selector-match (Objective-C and Objective-C++ only)'
2486 Warn if multiple methods with differing argument and/or return
2487 types are found for a given selector when attempting to send a
2488 message using this selector to a receiver of type `id' or `Class'.
2489 When this flag is off (which is the default behavior), the
2490 compiler will omit such warnings if any differences found are
2491 confined to types which share the same size and alignment.
2493 `-Wundeclared-selector (Objective-C and Objective-C++ only)'
2494 Warn if a `@selector(...)' expression referring to an undeclared
2495 selector is found. A selector is considered undeclared if no
2496 method with that name has been declared before the
2497 `@selector(...)' expression, either explicitly in an `@interface'
2498 or `@protocol' declaration, or implicitly in an `@implementation'
2499 section. This option always performs its checks as soon as a
2500 `@selector(...)' expression is found, while `-Wselector' only
2501 performs its checks in the final stage of compilation. This also
2502 enforces the coding style convention that methods and selectors
2503 must be declared before being used.
2505 `-print-objc-runtime-info'
2506 Generate C header describing the largest structure that is passed
2511 File: gcc.info, Node: Language Independent Options, Next: Warning Options, Prev: Objective-C and Objective-C++ Dialect Options, Up: Invoking GCC
2513 3.7 Options to Control Diagnostic Messages Formatting
2514 =====================================================
2516 Traditionally, diagnostic messages have been formatted irrespective of
2517 the output device's aspect (e.g. its width, ...). The options described
2518 below can be used to control the diagnostic messages formatting
2519 algorithm, e.g. how many characters per line, how often source location
2520 information should be reported. Right now, only the C++ front end can
2521 honor these options. However it is expected, in the near future, that
2522 the remaining front ends would be able to digest them correctly.
2524 `-fmessage-length=N'
2525 Try to format error messages so that they fit on lines of about N
2526 characters. The default is 72 characters for `g++' and 0 for the
2527 rest of the front ends supported by GCC. If N is zero, then no
2528 line-wrapping will be done; each error message will appear on a
2531 `-fdiagnostics-show-location=once'
2532 Only meaningful in line-wrapping mode. Instructs the diagnostic
2533 messages reporter to emit _once_ source location information; that
2534 is, in case the message is too long to fit on a single physical
2535 line and has to be wrapped, the source location won't be emitted
2536 (as prefix) again, over and over, in subsequent continuation
2537 lines. This is the default behavior.
2539 `-fdiagnostics-show-location=every-line'
2540 Only meaningful in line-wrapping mode. Instructs the diagnostic
2541 messages reporter to emit the same source location information (as
2542 prefix) for physical lines that result from the process of breaking
2543 a message which is too long to fit on a single line.
2545 `-fdiagnostics-show-option'
2546 This option instructs the diagnostic machinery to add text to each
2547 diagnostic emitted, which indicates which command line option
2548 directly controls that diagnostic, when such an option is known to
2549 the diagnostic machinery.
2551 `-Wcoverage-mismatch'
2552 Warn if feedback profiles do not match when using the
2553 `-fprofile-use' option. If a source file was changed between
2554 `-fprofile-gen' and `-fprofile-use', the files with the profile
2555 feedback can fail to match the source file and GCC can not use the
2556 profile feedback information. By default, GCC emits an error
2557 message in this case. The option `-Wcoverage-mismatch' emits a
2558 warning instead of an error. GCC does not use appropriate
2559 feedback profiles, so using this option can result in poorly
2560 optimized code. This option is useful only in the case of very
2561 minor changes such as bug fixes to an existing code-base.
2565 File: gcc.info, Node: Warning Options, Next: Debugging Options, Prev: Language Independent Options, Up: Invoking GCC
2567 3.8 Options to Request or Suppress Warnings
2568 ===========================================
2570 Warnings are diagnostic messages that report constructions which are
2571 not inherently erroneous but which are risky or suggest there may have
2574 The following language-independent options do not enable specific
2575 warnings but control the kinds of diagnostics produced by GCC.
2578 Check the code for syntax errors, but don't do anything beyond
2582 Inhibit all warning messages.
2585 Make all warnings into errors.
2588 Make the specified warning into an error. The specifier for a
2589 warning is appended, for example `-Werror=switch' turns the
2590 warnings controlled by `-Wswitch' into errors. This switch takes a
2591 negative form, to be used to negate `-Werror' for specific
2592 warnings, for example `-Wno-error=switch' makes `-Wswitch'
2593 warnings not be errors, even when `-Werror' is in effect. You can
2594 use the `-fdiagnostics-show-option' option to have each
2595 controllable warning amended with the option which controls it, to
2596 determine what to use with this option.
2598 Note that specifying `-Werror='FOO automatically implies `-W'FOO.
2599 However, `-Wno-error='FOO does not imply anything.
2602 This option causes the compiler to abort compilation on the first
2603 error occurred rather than trying to keep going and printing
2604 further error messages.
2607 You can request many specific warnings with options beginning `-W',
2608 for example `-Wimplicit' to request warnings on implicit declarations.
2609 Each of these specific warning options also has a negative form
2610 beginning `-Wno-' to turn off warnings; for example, `-Wno-implicit'.
2611 This manual lists only one of the two forms, whichever is not the
2612 default. For further, language-specific options also refer to *Note
2613 C++ Dialect Options:: and *Note Objective-C and Objective-C++ Dialect
2617 Issue all the warnings demanded by strict ISO C and ISO C++;
2618 reject all programs that use forbidden extensions, and some other
2619 programs that do not follow ISO C and ISO C++. For ISO C, follows
2620 the version of the ISO C standard specified by any `-std' option
2623 Valid ISO C and ISO C++ programs should compile properly with or
2624 without this option (though a rare few will require `-ansi' or a
2625 `-std' option specifying the required version of ISO C). However,
2626 without this option, certain GNU extensions and traditional C and
2627 C++ features are supported as well. With this option, they are
2630 `-pedantic' does not cause warning messages for use of the
2631 alternate keywords whose names begin and end with `__'. Pedantic
2632 warnings are also disabled in the expression that follows
2633 `__extension__'. However, only system header files should use
2634 these escape routes; application programs should avoid them.
2635 *Note Alternate Keywords::.
2637 Some users try to use `-pedantic' to check programs for strict ISO
2638 C conformance. They soon find that it does not do quite what they
2639 want: it finds some non-ISO practices, but not all--only those for
2640 which ISO C _requires_ a diagnostic, and some others for which
2641 diagnostics have been added.
2643 A feature to report any failure to conform to ISO C might be
2644 useful in some instances, but would require considerable
2645 additional work and would be quite different from `-pedantic'. We
2646 don't have plans to support such a feature in the near future.
2648 Where the standard specified with `-std' represents a GNU extended
2649 dialect of C, such as `gnu89' or `gnu99', there is a corresponding
2650 "base standard", the version of ISO C on which the GNU extended
2651 dialect is based. Warnings from `-pedantic' are given where they
2652 are required by the base standard. (It would not make sense for
2653 such warnings to be given only for features not in the specified
2654 GNU C dialect, since by definition the GNU dialects of C include
2655 all features the compiler supports with the given option, and
2656 there would be nothing to warn about.)
2659 Like `-pedantic', except that errors are produced rather than
2663 This enables all the warnings about constructions that some users
2664 consider questionable, and that are easy to avoid (or modify to
2665 prevent the warning), even in conjunction with macros. This also
2666 enables some language-specific warnings described in *Note C++
2667 Dialect Options:: and *Note Objective-C and Objective-C++ Dialect
2670 `-Wall' turns on the following warning flags:
2673 -Warray-bounds (only with `-O2')
2677 -Wimplicit-function-declaration
2680 -Wmain (only for C/ObjC and unless `-ffreestanding')
2688 -Wsign-compare (only in C++)
2693 -Wuninitialized (only with `-O1' and above)
2700 Note that some warning flags are not implied by `-Wall'. Some of
2701 them warn about constructions that users generally do not consider
2702 questionable, but which occasionally you might wish to check for;
2703 others warn about constructions that are necessary or hard to
2704 avoid in some cases, and there is no simple way to modify the code
2705 to suppress the warning. Some of them are enabled by `-Wextra' but
2706 many of them must be enabled individually.
2709 This enables some extra warning flags that are not enabled by
2710 `-Wall'. (This option used to be called `-W'. The older name is
2711 still supported, but the newer name is more descriptive.)
2715 -Wignored-qualifiers
2716 -Wmissing-field-initializers
2717 -Wmissing-parameter-type (C only)
2718 -Wold-style-declaration (C only)
2722 -Wuninitialized (only with `-O1' and above)
2723 -Wunused-parameter (only with `-Wunused' or `-Wall')
2725 The option `-Wextra' also prints warning messages for the
2728 * A pointer is compared against integer zero with `<', `<=',
2731 * (C++ only) An enumerator and a non-enumerator both appear in a
2732 conditional expression.
2734 * (C++ only) A non-static reference or non-static `const' member
2735 appears in a class without constructors.
2737 * (C++ only) Ambiguous virtual bases.
2739 * (C++ only) Subscripting an array which has been declared
2742 * (C++ only) Taking the address of a variable which has been
2743 declared `register'.
2745 * (C++ only) A base class is not initialized in a derived
2746 class' copy constructor.
2750 Inhibit warning messages about the use of `#import'.
2753 Warn if an array subscript has type `char'. This is a common cause
2754 of error, as programmers often forget that this type is signed on
2755 some machines. This warning is enabled by `-Wall'.
2758 Warn whenever a comment-start sequence `/*' appears in a `/*'
2759 comment, or whenever a Backslash-Newline appears in a `//' comment.
2760 This warning is enabled by `-Wall'.
2763 Check calls to `printf' and `scanf', etc., to make sure that the
2764 arguments supplied have types appropriate to the format string
2765 specified, and that the conversions specified in the format string
2766 make sense. This includes standard functions, and others
2767 specified by format attributes (*note Function Attributes::), in
2768 the `printf', `scanf', `strftime' and `strfmon' (an X/Open
2769 extension, not in the C standard) families (or other
2770 target-specific families). Which functions are checked without
2771 format attributes having been specified depends on the standard
2772 version selected, and such checks of functions without the
2773 attribute specified are disabled by `-ffreestanding' or
2776 The formats are checked against the format features supported by
2777 GNU libc version 2.2. These include all ISO C90 and C99 features,
2778 as well as features from the Single Unix Specification and some
2779 BSD and GNU extensions. Other library implementations may not
2780 support all these features; GCC does not support warning about
2781 features that go beyond a particular library's limitations.
2782 However, if `-pedantic' is used with `-Wformat', warnings will be
2783 given about format features not in the selected standard version
2784 (but not for `strfmon' formats, since those are not in any version
2785 of the C standard). *Note Options Controlling C Dialect: C
2788 Since `-Wformat' also checks for null format arguments for several
2789 functions, `-Wformat' also implies `-Wnonnull'.
2791 `-Wformat' is included in `-Wall'. For more control over some
2792 aspects of format checking, the options `-Wformat-y2k',
2793 `-Wno-format-extra-args', `-Wno-format-zero-length',
2794 `-Wformat-nonliteral', `-Wformat-security', and `-Wformat=2' are
2795 available, but are not included in `-Wall'.
2798 If `-Wformat' is specified, also warn about `strftime' formats
2799 which may yield only a two-digit year.
2801 `-Wno-format-extra-args'
2802 If `-Wformat' is specified, do not warn about excess arguments to a
2803 `printf' or `scanf' format function. The C standard specifies
2804 that such arguments are ignored.
2806 Where the unused arguments lie between used arguments that are
2807 specified with `$' operand number specifications, normally
2808 warnings are still given, since the implementation could not know
2809 what type to pass to `va_arg' to skip the unused arguments.
2810 However, in the case of `scanf' formats, this option will suppress
2811 the warning if the unused arguments are all pointers, since the
2812 Single Unix Specification says that such unused arguments are
2815 `-Wno-format-zero-length (C and Objective-C only)'
2816 If `-Wformat' is specified, do not warn about zero-length formats.
2817 The C standard specifies that zero-length formats are allowed.
2819 `-Wformat-nonliteral'
2820 If `-Wformat' is specified, also warn if the format string is not a
2821 string literal and so cannot be checked, unless the format function
2822 takes its format arguments as a `va_list'.
2825 If `-Wformat' is specified, also warn about uses of format
2826 functions that represent possible security problems. At present,
2827 this warns about calls to `printf' and `scanf' functions where the
2828 format string is not a string literal and there are no format
2829 arguments, as in `printf (foo);'. This may be a security hole if
2830 the format string came from untrusted input and contains `%n'.
2831 (This is currently a subset of what `-Wformat-nonliteral' warns
2832 about, but in future warnings may be added to `-Wformat-security'
2833 that are not included in `-Wformat-nonliteral'.)
2836 Enable `-Wformat' plus format checks not included in `-Wformat'.
2837 Currently equivalent to `-Wformat -Wformat-nonliteral
2838 -Wformat-security -Wformat-y2k'.
2840 `-Wnonnull (C and Objective-C only)'
2841 Warn about passing a null pointer for arguments marked as
2842 requiring a non-null value by the `nonnull' function attribute.
2844 `-Wnonnull' is included in `-Wall' and `-Wformat'. It can be
2845 disabled with the `-Wno-nonnull' option.
2847 `-Winit-self (C, C++, Objective-C and Objective-C++ only)'
2848 Warn about uninitialized variables which are initialized with
2849 themselves. Note this option can only be used with the
2850 `-Wuninitialized' option, which in turn only works with `-O1' and
2853 For example, GCC will warn about `i' being uninitialized in the
2854 following snippet only when `-Winit-self' has been specified:
2861 `-Wimplicit-int (C and Objective-C only)'
2862 Warn when a declaration does not specify a type. This warning is
2865 `-Wimplicit-function-declaration (C and Objective-C only)'
2866 Give a warning whenever a function is used before being declared.
2867 In C99 mode (`-std=c99' or `-std=gnu99'), this warning is enabled
2868 by default and it is made into an error by `-pedantic-errors'.
2869 This warning is also enabled by `-Wall'.
2872 Same as `-Wimplicit-int' and `-Wimplicit-function-declaration'.
2873 This warning is enabled by `-Wall'.
2875 `-Wignored-qualifiers (C and C++ only)'
2876 Warn if the return type of a function has a type qualifier such as
2877 `const'. For ISO C such a type qualifier has no effect, since the
2878 value returned by a function is not an lvalue. For C++, the
2879 warning is only emitted for scalar types or `void'. ISO C
2880 prohibits qualified `void' return types on function definitions,
2881 so such return types always receive a warning even without this
2884 This warning is also enabled by `-Wextra'.
2887 Warn if the type of `main' is suspicious. `main' should be a
2888 function with external linkage, returning int, taking either zero
2889 arguments, two, or three arguments of appropriate types. This
2890 warning is enabled by `-Wall'.
2893 Warn if an aggregate or union initializer is not fully bracketed.
2894 In the following example, the initializer for `a' is not fully
2895 bracketed, but that for `b' is fully bracketed.
2897 int a[2][2] = { 0, 1, 2, 3 };
2898 int b[2][2] = { { 0, 1 }, { 2, 3 } };
2900 This warning is enabled by `-Wall'.
2902 `-Wmissing-include-dirs (C, C++, Objective-C and Objective-C++ only)'
2903 Warn if a user-supplied include directory does not exist.
2906 Warn if parentheses are omitted in certain contexts, such as when
2907 there is an assignment in a context where a truth value is
2908 expected, or when operators are nested whose precedence people
2909 often get confused about.
2911 Also warn if a comparison like `x<=y<=z' appears; this is
2912 equivalent to `(x<=y ? 1 : 0) <= z', which is a different
2913 interpretation from that of ordinary mathematical notation.
2915 Also warn about constructions where there may be confusion to which
2916 `if' statement an `else' branch belongs. Here is an example of
2927 In C/C++, every `else' branch belongs to the innermost possible
2928 `if' statement, which in this example is `if (b)'. This is often
2929 not what the programmer expected, as illustrated in the above
2930 example by indentation the programmer chose. When there is the
2931 potential for this confusion, GCC will issue a warning when this
2932 flag is specified. To eliminate the warning, add explicit braces
2933 around the innermost `if' statement so there is no way the `else'
2934 could belong to the enclosing `if'. The resulting code would look
2947 This warning is enabled by `-Wall'.
2950 Warn about code that may have undefined semantics because of
2951 violations of sequence point rules in the C and C++ standards.
2953 The C and C++ standards defines the order in which expressions in
2954 a C/C++ program are evaluated in terms of "sequence points", which
2955 represent a partial ordering between the execution of parts of the
2956 program: those executed before the sequence point, and those
2957 executed after it. These occur after the evaluation of a full
2958 expression (one which is not part of a larger expression), after
2959 the evaluation of the first operand of a `&&', `||', `? :' or `,'
2960 (comma) operator, before a function is called (but after the
2961 evaluation of its arguments and the expression denoting the called
2962 function), and in certain other places. Other than as expressed
2963 by the sequence point rules, the order of evaluation of
2964 subexpressions of an expression is not specified. All these rules
2965 describe only a partial order rather than a total order, since,
2966 for example, if two functions are called within one expression
2967 with no sequence point between them, the order in which the
2968 functions are called is not specified. However, the standards
2969 committee have ruled that function calls do not overlap.
2971 It is not specified when between sequence points modifications to
2972 the values of objects take effect. Programs whose behavior
2973 depends on this have undefined behavior; the C and C++ standards
2974 specify that "Between the previous and next sequence point an
2975 object shall have its stored value modified at most once by the
2976 evaluation of an expression. Furthermore, the prior value shall
2977 be read only to determine the value to be stored.". If a program
2978 breaks these rules, the results on any particular implementation
2979 are entirely unpredictable.
2981 Examples of code with undefined behavior are `a = a++;', `a[n] =
2982 b[n++]' and `a[i++] = i;'. Some more complicated cases are not
2983 diagnosed by this option, and it may give an occasional false
2984 positive result, but in general it has been found fairly effective
2985 at detecting this sort of problem in programs.
2987 The standard is worded confusingly, therefore there is some debate
2988 over the precise meaning of the sequence point rules in subtle
2989 cases. Links to discussions of the problem, including proposed
2990 formal definitions, may be found on the GCC readings page, at
2991 `http://gcc.gnu.org/readings.html'.
2993 This warning is enabled by `-Wall' for C and C++.
2996 Warn whenever a function is defined with a return-type that
2997 defaults to `int'. Also warn about any `return' statement with no
2998 return-value in a function whose return-type is not `void'
2999 (falling off the end of the function body is considered returning
3000 without a value), and about a `return' statement with a expression
3001 in a function whose return-type is `void'.
3003 For C++, a function without return type always produces a
3004 diagnostic message, even when `-Wno-return-type' is specified.
3005 The only exceptions are `main' and functions defined in system
3008 This warning is enabled by `-Wall'.
3011 Warn whenever a `switch' statement has an index of enumerated type
3012 and lacks a `case' for one or more of the named codes of that
3013 enumeration. (The presence of a `default' label prevents this
3014 warning.) `case' labels outside the enumeration range also
3015 provoke warnings when this option is used. This warning is
3019 Warn whenever a `switch' statement does not have a `default' case.
3022 Warn whenever a `switch' statement has an index of enumerated type
3023 and lacks a `case' for one or more of the named codes of that
3024 enumeration. `case' labels outside the enumeration range also
3025 provoke warnings when this option is used.
3028 Warn if any trigraphs are encountered that might change the
3029 meaning of the program (trigraphs within comments are not warned
3030 about). This warning is enabled by `-Wall'.
3033 Warn whenever a static function is declared but not defined or a
3034 non-inline static function is unused. This warning is enabled by
3038 Warn whenever a label is declared but not used. This warning is
3041 To suppress this warning use the `unused' attribute (*note
3042 Variable Attributes::).
3044 `-Wunused-parameter'
3045 Warn whenever a function parameter is unused aside from its
3048 To suppress this warning use the `unused' attribute (*note
3049 Variable Attributes::).
3052 Warn whenever a local variable or non-constant static variable is
3053 unused aside from its declaration. This warning is enabled by
3056 To suppress this warning use the `unused' attribute (*note
3057 Variable Attributes::).
3060 Warn whenever a statement computes a result that is explicitly not
3061 used. To suppress this warning cast the unused expression to
3062 `void'. This includes an expression-statement or the left-hand
3063 side of a comma expression that contains no side effects. For
3064 example, an expression such as `x[i,j]' will cause a warning, while
3065 `x[(void)i,j]' will not.
3067 This warning is enabled by `-Wall'.
3070 All the above `-Wunused' options combined.
3072 In order to get a warning about an unused function parameter, you
3073 must either specify `-Wextra -Wunused' (note that `-Wall' implies
3074 `-Wunused'), or separately specify `-Wunused-parameter'.
3077 Warn if an automatic variable is used without first being
3078 initialized or if a variable may be clobbered by a `setjmp' call.
3080 These warnings are possible only in optimizing compilation,
3081 because they require data flow information that is computed only
3082 when optimizing. If you do not specify `-O', you will not get
3083 these warnings. Instead, GCC will issue a warning about
3084 `-Wuninitialized' requiring `-O'.
3086 If you want to warn about code which uses the uninitialized value
3087 of the variable in its own initializer, use the `-Winit-self'
3090 These warnings occur for individual uninitialized or clobbered
3091 elements of structure, union or array variables as well as for
3092 variables which are uninitialized or clobbered as a whole. They do
3093 not occur for variables or elements declared `volatile'. Because
3094 these warnings depend on optimization, the exact variables or
3095 elements for which there are warnings will depend on the precise
3096 optimization options and version of GCC used.
3098 Note that there may be no warning about a variable that is used
3099 only to compute a value that itself is never used, because such
3100 computations may be deleted by data flow analysis before the
3101 warnings are printed.
3103 These warnings are made optional because GCC is not smart enough
3104 to see all the reasons why the code might be correct despite
3105 appearing to have an error. Here is one example of how this can
3121 If the value of `y' is always 1, 2 or 3, then `x' is always
3122 initialized, but GCC doesn't know this. Here is another common
3127 if (change_y) save_y = y, y = new_y;
3129 if (change_y) y = save_y;
3132 This has no bug because `save_y' is used only if it is set.
3134 This option also warns when a non-volatile automatic variable
3135 might be changed by a call to `longjmp'. These warnings as well
3136 are possible only in optimizing compilation.
3138 The compiler sees only the calls to `setjmp'. It cannot know
3139 where `longjmp' will be called; in fact, a signal handler could
3140 call it at any point in the code. As a result, you may get a
3141 warning even when there is in fact no problem because `longjmp'
3142 cannot in fact be called at the place which would cause a problem.
3144 Some spurious warnings can be avoided if you declare all the
3145 functions you use that never return as `noreturn'. *Note Function
3148 This warning is enabled by `-Wall' or `-Wextra' in optimizing
3149 compilations (`-O1' and above).
3152 Warn when a #pragma directive is encountered which is not
3153 understood by GCC. If this command line option is used, warnings
3154 will even be issued for unknown pragmas in system header files.
3155 This is not the case if the warnings were only enabled by the
3156 `-Wall' command line option.
3159 Do not warn about misuses of pragmas, such as incorrect parameters,
3160 invalid syntax, or conflicts between pragmas. See also
3161 `-Wunknown-pragmas'.
3164 This option is only active when `-fstrict-aliasing' is active. It
3165 warns about code which might break the strict aliasing rules that
3166 the compiler is using for optimization. The warning does not
3167 catch all cases, but does attempt to catch the more common
3168 pitfalls. It is included in `-Wall'. It is equivalent to
3169 `-Wstrict-aliasing=3'
3171 `-Wstrict-aliasing=n'
3172 This option is only active when `-fstrict-aliasing' is active. It
3173 warns about code which might break the strict aliasing rules that
3174 the compiler is using for optimization. Higher levels correspond
3175 to higher accuracy (fewer false positives). Higher levels also
3176 correspond to more effort, similar to the way -O works.
3177 `-Wstrict-aliasing' is equivalent to `-Wstrict-aliasing=n', with
3180 Level 1: Most aggressive, quick, least accurate. Possibly useful
3181 when higher levels do not warn but -fstrict-aliasing still breaks
3182 the code, as it has very few false negatives. However, it has
3183 many false positives. Warns for all pointer conversions between
3184 possibly incompatible types, even if never dereferenced. Runs in
3187 Level 2: Aggressive, quick, not too precise. May still have many
3188 false positives (not as many as level 1 though), and few false
3189 negatives (but possibly more than level 1). Unlike level 1, it
3190 only warns when an address is taken. Warns about incomplete
3191 types. Runs in the frontend only.
3193 Level 3 (default for `-Wstrict-aliasing'): Should have very few
3194 false positives and few false negatives. Slightly slower than
3195 levels 1 or 2 when optimization is enabled. Takes care of the
3196 common punn+dereference pattern in the frontend:
3197 `*(int*)&some_float'. If optimization is enabled, it also runs in
3198 the backend, where it deals with multiple statement cases using
3199 flow-sensitive points-to information. Only warns when the
3200 converted pointer is dereferenced. Does not warn about incomplete
3204 `-Wstrict-overflow=N'
3205 This option is only active when `-fstrict-overflow' is active. It
3206 warns about cases where the compiler optimizes based on the
3207 assumption that signed overflow does not occur. Note that it does
3208 not warn about all cases where the code might overflow: it only
3209 warns about cases where the compiler implements some optimization.
3210 Thus this warning depends on the optimization level.
3212 An optimization which assumes that signed overflow does not occur
3213 is perfectly safe if the values of the variables involved are such
3214 that overflow never does, in fact, occur. Therefore this warning
3215 can easily give a false positive: a warning about code which is not
3216 actually a problem. To help focus on important issues, several
3217 warning levels are defined. No warnings are issued for the use of
3218 undefined signed overflow when estimating how many iterations a
3219 loop will require, in particular when determining whether a loop
3220 will be executed at all.
3222 `-Wstrict-overflow=1'
3223 Warn about cases which are both questionable and easy to
3224 avoid. For example: `x + 1 > x'; with `-fstrict-overflow',
3225 the compiler will simplify this to `1'. This level of
3226 `-Wstrict-overflow' is enabled by `-Wall'; higher levels are
3227 not, and must be explicitly requested.
3229 `-Wstrict-overflow=2'
3230 Also warn about other cases where a comparison is simplified
3231 to a constant. For example: `abs (x) >= 0'. This can only be
3232 simplified when `-fstrict-overflow' is in effect, because
3233 `abs (INT_MIN)' overflows to `INT_MIN', which is less than
3234 zero. `-Wstrict-overflow' (with no level) is the same as
3235 `-Wstrict-overflow=2'.
3237 `-Wstrict-overflow=3'
3238 Also warn about other cases where a comparison is simplified.
3239 For example: `x + 1 > 1' will be simplified to `x > 0'.
3241 `-Wstrict-overflow=4'
3242 Also warn about other simplifications not covered by the
3243 above cases. For example: `(x * 10) / 5' will be simplified
3246 `-Wstrict-overflow=5'
3247 Also warn about cases where the compiler reduces the
3248 magnitude of a constant involved in a comparison. For
3249 example: `x + 2 > y' will be simplified to `x + 1 >= y'.
3250 This is reported only at the highest warning level because
3251 this simplification applies to many comparisons, so this
3252 warning level will give a very large number of false
3256 This option is only active when `-ftree-vrp' is active (default
3257 for -O2 and above). It warns about subscripts to arrays that are
3258 always out of bounds. This warning is enabled by `-Wall'.
3261 Do not warn about compile-time integer division by zero. Floating
3262 point division by zero is not warned about, as it can be a
3263 legitimate way of obtaining infinities and NaNs.
3266 Print warning messages for constructs found in system header files.
3267 Warnings from system headers are normally suppressed, on the
3268 assumption that they usually do not indicate real problems and
3269 would only make the compiler output harder to read. Using this
3270 command line option tells GCC to emit warnings from system headers
3271 as if they occurred in user code. However, note that using
3272 `-Wall' in conjunction with this option will _not_ warn about
3273 unknown pragmas in system headers--for that, `-Wunknown-pragmas'
3277 Warn if floating point values are used in equality comparisons.
3279 The idea behind this is that sometimes it is convenient (for the
3280 programmer) to consider floating-point values as approximations to
3281 infinitely precise real numbers. If you are doing this, then you
3282 need to compute (by analyzing the code, or in some other way) the
3283 maximum or likely maximum error that the computation introduces,
3284 and allow for it when performing comparisons (and when producing
3285 output, but that's a different problem). In particular, instead
3286 of testing for equality, you would check to see whether the two
3287 values have ranges that overlap; and this is done with the
3288 relational operators, so equality comparisons are probably
3291 `-Wtraditional (C and Objective-C only)'
3292 Warn about certain constructs that behave differently in
3293 traditional and ISO C. Also warn about ISO C constructs that have
3294 no traditional C equivalent, and/or problematic constructs which
3297 * Macro parameters that appear within string literals in the
3298 macro body. In traditional C macro replacement takes place
3299 within string literals, but does not in ISO C.
3301 * In traditional C, some preprocessor directives did not exist.
3302 Traditional preprocessors would only consider a line to be a
3303 directive if the `#' appeared in column 1 on the line.
3304 Therefore `-Wtraditional' warns about directives that
3305 traditional C understands but would ignore because the `#'
3306 does not appear as the first character on the line. It also
3307 suggests you hide directives like `#pragma' not understood by
3308 traditional C by indenting them. Some traditional
3309 implementations would not recognize `#elif', so it suggests
3310 avoiding it altogether.
3312 * A function-like macro that appears without arguments.
3314 * The unary plus operator.
3316 * The `U' integer constant suffix, or the `F' or `L' floating
3317 point constant suffixes. (Traditional C does support the `L'
3318 suffix on integer constants.) Note, these suffixes appear in
3319 macros defined in the system headers of most modern systems,
3320 e.g. the `_MIN'/`_MAX' macros in `<limits.h>'. Use of these
3321 macros in user code might normally lead to spurious warnings,
3322 however GCC's integrated preprocessor has enough context to
3323 avoid warning in these cases.
3325 * A function declared external in one block and then used after
3326 the end of the block.
3328 * A `switch' statement has an operand of type `long'.
3330 * A non-`static' function declaration follows a `static' one.
3331 This construct is not accepted by some traditional C
3334 * The ISO type of an integer constant has a different width or
3335 signedness from its traditional type. This warning is only
3336 issued if the base of the constant is ten. I.e. hexadecimal
3337 or octal values, which typically represent bit patterns, are
3340 * Usage of ISO string concatenation is detected.
3342 * Initialization of automatic aggregates.
3344 * Identifier conflicts with labels. Traditional C lacks a
3345 separate namespace for labels.
3347 * Initialization of unions. If the initializer is zero, the
3348 warning is omitted. This is done under the assumption that
3349 the zero initializer in user code appears conditioned on e.g.
3350 `__STDC__' to avoid missing initializer warnings and relies
3351 on default initialization to zero in the traditional C case.
3353 * Conversions by prototypes between fixed/floating point values
3354 and vice versa. The absence of these prototypes when
3355 compiling with traditional C would cause serious problems.
3356 This is a subset of the possible conversion warnings, for the
3357 full set use `-Wtraditional-conversion'.
3359 * Use of ISO C style function definitions. This warning
3360 intentionally is _not_ issued for prototype declarations or
3361 variadic functions because these ISO C features will appear
3362 in your code when using libiberty's traditional C
3363 compatibility macros, `PARAMS' and `VPARAMS'. This warning
3364 is also bypassed for nested functions because that feature is
3365 already a GCC extension and thus not relevant to traditional
3368 `-Wtraditional-conversion (C and Objective-C only)'
3369 Warn if a prototype causes a type conversion that is different
3370 from what would happen to the same argument in the absence of a
3371 prototype. This includes conversions of fixed point to floating
3372 and vice versa, and conversions changing the width or signedness
3373 of a fixed point argument except when the same as the default
3376 `-Wdeclaration-after-statement (C and Objective-C only)'
3377 Warn when a declaration is found after a statement in a block.
3378 This construct, known from C++, was introduced with ISO C99 and is
3379 by default allowed in GCC. It is not supported by ISO C90 and was
3380 not supported by GCC versions before GCC 3.0. *Note Mixed
3384 Warn if an undefined identifier is evaluated in an `#if' directive.
3387 Do not warn whenever an `#else' or an `#endif' are followed by
3391 Warn whenever a local variable shadows another local variable,
3392 parameter or global variable or whenever a built-in function is
3396 Warn whenever an object of larger than LEN bytes is defined.
3398 `-Wunsafe-loop-optimizations'
3399 Warn if the loop cannot be optimized because the compiler could not
3400 assume anything on the bounds of the loop indices. With
3401 `-funsafe-loop-optimizations' warn if the compiler made such
3405 Warn about anything that depends on the "size of" a function type
3406 or of `void'. GNU C assigns these types a size of 1, for
3407 convenience in calculations with `void *' pointers and pointers to
3408 functions. In C++, warn also when an arithmetic operation involves
3409 `NULL'. This warning is also enabled by `-pedantic'.
3412 Warn if a comparison is always true or always false due to the
3413 limited range of the data type, but do not warn for constant
3414 expressions. For example, warn if an unsigned variable is
3415 compared against zero with `<' or `>='. This warning is also
3416 enabled by `-Wextra'.
3418 `-Wbad-function-cast (C and Objective-C only)'
3419 Warn whenever a function call is cast to a non-matching type. For
3420 example, warn if `int malloc()' is cast to `anything *'.
3422 `-Wc++-compat (C and Objective-C only)'
3423 Warn about ISO C constructs that are outside of the common subset
3424 of ISO C and ISO C++, e.g. request for implicit conversion from
3425 `void *' to a pointer to non-`void' type.
3427 `-Wc++0x-compat (C++ and Objective-C++ only)'
3428 Warn about C++ constructs whose meaning differs between ISO C++
3429 1998 and ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will
3430 become keywords in ISO C++ 200x. This warning is enabled by
3434 Warn whenever a pointer is cast so as to remove a type qualifier
3435 from the target type. For example, warn if a `const char *' is
3436 cast to an ordinary `char *'.
3439 Warn whenever a pointer is cast such that the required alignment
3440 of the target is increased. For example, warn if a `char *' is
3441 cast to an `int *' on machines where integers can only be accessed
3442 at two- or four-byte boundaries.
3445 When compiling C, give string constants the type `const
3446 char[LENGTH]' so that copying the address of one into a
3447 non-`const' `char *' pointer will get a warning; when compiling
3448 C++, warn about the deprecated conversion from string literals to
3449 `char *'. This warning, by default, is enabled for C++ programs.
3450 These warnings will help you find at compile time code that can
3451 try to write into a string constant, but only if you have been
3452 very careful about using `const' in declarations and prototypes.
3453 Otherwise, it will just be a nuisance; this is why we did not make
3454 `-Wall' request these warnings.
3457 Warn for variables that might be changed by `longjmp' or `vfork'.
3458 This warning is also enabled by `-Wextra'.
3461 Warn for implicit conversions that may alter a value. This includes
3462 conversions between real and integer, like `abs (x)' when `x' is
3463 `double'; conversions between signed and unsigned, like `unsigned
3464 ui = -1'; and conversions to smaller types, like `sqrtf (M_PI)'.
3465 Do not warn for explicit casts like `abs ((int) x)' and `ui =
3466 (unsigned) -1', or if the value is not changed by the conversion
3467 like in `abs (2.0)'. Warnings about conversions between signed
3468 and unsigned integers can be disabled by using
3469 `-Wno-sign-conversion'.
3471 For C++, also warn for conversions between `NULL' and non-pointer
3472 types; confusing overload resolution for user-defined conversions;
3473 and conversions that will never use a type conversion operator:
3474 conversions to `void', the same type, a base class or a reference
3475 to them. Warnings about conversions between signed and unsigned
3476 integers are disabled by default in C++ unless `-Wsign-conversion'
3477 is explicitly enabled.
3480 Warn if an empty body occurs in an `if', `else' or `do while'
3481 statement. Additionally, in C++, warn when an empty body occurs
3482 in a `while' or `for' statement with no whitespacing before the
3483 semicolon. This warning is also enabled by `-Wextra'.
3486 Warn when a comparison between signed and unsigned values could
3487 produce an incorrect result when the signed value is converted to
3488 unsigned. This warning is also enabled by `-Wextra'; to get the
3489 other warnings of `-Wextra' without this warning, use `-Wextra
3493 Warn for implicit conversions that may change the sign of an
3494 integer value, like assigning a signed integer expression to an
3495 unsigned integer variable. An explicit cast silences the warning.
3496 In C, this option is enabled also by `-Wconversion'.
3499 Warn about suspicious uses of memory addresses. These include using
3500 the address of a function in a conditional expression, such as
3501 `void func(void); if (func)', and comparisons against the memory
3502 address of a string literal, such as `if (x == "abc")'. Such uses
3503 typically indicate a programmer error: the address of a function
3504 always evaluates to true, so their use in a conditional usually
3505 indicate that the programmer forgot the parentheses in a function
3506 call; and comparisons against string literals result in unspecified
3507 behavior and are not portable in C, so they usually indicate that
3508 the programmer intended to use `strcmp'. This warning is enabled
3512 Warn about suspicious uses of logical operators in expressions.
3513 This includes using logical operators in contexts where a bit-wise
3514 operator is likely to be expected.
3516 `-Waggregate-return'
3517 Warn if any functions that return structures or unions are defined
3518 or called. (In languages where you can return an array, this also
3522 Do not warn if an unexpected `__attribute__' is used, such as
3523 unrecognized attributes, function attributes applied to variables,
3524 etc. This will not stop errors for incorrect use of supported
3527 `-Wstrict-prototypes (C and Objective-C only)'
3528 Warn if a function is declared or defined without specifying the
3529 argument types. (An old-style function definition is permitted
3530 without a warning if preceded by a declaration which specifies the
3533 `-Wold-style-declaration (C and Objective-C only)'
3534 Warn for obsolescent usages, according to the C Standard, in a
3535 declaration. For example, warn if storage-class specifiers like
3536 `static' are not the first things in a declaration. This warning
3537 is also enabled by `-Wextra'.
3539 `-Wold-style-definition (C and Objective-C only)'
3540 Warn if an old-style function definition is used. A warning is
3541 given even if there is a previous prototype.
3543 `-Wmissing-parameter-type (C and Objective-C only)'
3544 A function parameter is declared without a type specifier in
3545 K&R-style functions:
3549 This warning is also enabled by `-Wextra'.
3551 `-Wmissing-prototypes (C and Objective-C only)'
3552 Warn if a global function is defined without a previous prototype
3553 declaration. This warning is issued even if the definition itself
3554 provides a prototype. The aim is to detect global functions that
3555 fail to be declared in header files.
3557 `-Wmissing-declarations'
3558 Warn if a global function is defined without a previous
3559 declaration. Do so even if the definition itself provides a
3560 prototype. Use this option to detect global functions that are
3561 not declared in header files. In C++, no warnings are issued for
3562 function templates, or for inline functions, or for functions in
3563 anonymous namespaces.
3565 `-Wmissing-field-initializers'
3566 Warn if a structure's initializer has some fields missing. For
3567 example, the following code would cause such a warning, because
3568 `x.h' is implicitly zero:
3570 struct s { int f, g, h; };
3571 struct s x = { 3, 4 };
3573 This option does not warn about designated initializers, so the
3574 following modification would not trigger a warning:
3576 struct s { int f, g, h; };
3577 struct s x = { .f = 3, .g = 4 };
3579 This warning is included in `-Wextra'. To get other `-Wextra'
3580 warnings without this one, use `-Wextra
3581 -Wno-missing-field-initializers'.
3583 `-Wmissing-noreturn'
3584 Warn about functions which might be candidates for attribute
3585 `noreturn'. Note these are only possible candidates, not absolute
3586 ones. Care should be taken to manually verify functions actually
3587 do not ever return before adding the `noreturn' attribute,
3588 otherwise subtle code generation bugs could be introduced. You
3589 will not get a warning for `main' in hosted C environments.
3591 `-Wmissing-format-attribute'
3592 Warn about function pointers which might be candidates for `format'
3593 attributes. Note these are only possible candidates, not absolute
3594 ones. GCC will guess that function pointers with `format'
3595 attributes that are used in assignment, initialization, parameter
3596 passing or return statements should have a corresponding `format'
3597 attribute in the resulting type. I.e. the left-hand side of the
3598 assignment or initialization, the type of the parameter variable,
3599 or the return type of the containing function respectively should
3600 also have a `format' attribute to avoid the warning.
3602 GCC will also warn about function definitions which might be
3603 candidates for `format' attributes. Again, these are only
3604 possible candidates. GCC will guess that `format' attributes
3605 might be appropriate for any function that calls a function like
3606 `vprintf' or `vscanf', but this might not always be the case, and
3607 some functions for which `format' attributes are appropriate may
3611 Do not warn if a multicharacter constant (`'FOOF'') is used.
3612 Usually they indicate a typo in the user's code, as they have
3613 implementation-defined values, and should not be used in portable
3616 `-Wnormalized=<none|id|nfc|nfkc>'
3617 In ISO C and ISO C++, two identifiers are different if they are
3618 different sequences of characters. However, sometimes when
3619 characters outside the basic ASCII character set are used, you can
3620 have two different character sequences that look the same. To
3621 avoid confusion, the ISO 10646 standard sets out some
3622 "normalization rules" which when applied ensure that two sequences
3623 that look the same are turned into the same sequence. GCC can
3624 warn you if you are using identifiers which have not been
3625 normalized; this option controls that warning.
3627 There are four levels of warning that GCC supports. The default is
3628 `-Wnormalized=nfc', which warns about any identifier which is not
3629 in the ISO 10646 "C" normalized form, "NFC". NFC is the
3630 recommended form for most uses.
3632 Unfortunately, there are some characters which ISO C and ISO C++
3633 allow in identifiers that when turned into NFC aren't allowable as
3634 identifiers. That is, there's no way to use these symbols in
3635 portable ISO C or C++ and have all your identifiers in NFC.
3636 `-Wnormalized=id' suppresses the warning for these characters. It
3637 is hoped that future versions of the standards involved will
3638 correct this, which is why this option is not the default.
3640 You can switch the warning off for all characters by writing
3641 `-Wnormalized=none'. You would only want to do this if you were
3642 using some other normalization scheme (like "D"), because
3643 otherwise you can easily create bugs that are literally impossible
3646 Some characters in ISO 10646 have distinct meanings but look
3647 identical in some fonts or display methodologies, especially once
3648 formatting has been applied. For instance `\u207F', "SUPERSCRIPT
3649 LATIN SMALL LETTER N", will display just like a regular `n' which
3650 has been placed in a superscript. ISO 10646 defines the "NFKC"
3651 normalization scheme to convert all these into a standard form as
3652 well, and GCC will warn if your code is not in NFKC if you use
3653 `-Wnormalized=nfkc'. This warning is comparable to warning about
3654 every identifier that contains the letter O because it might be
3655 confused with the digit 0, and so is not the default, but may be
3656 useful as a local coding convention if the programming environment
3657 is unable to be fixed to display these characters distinctly.
3659 `-Wno-deprecated-declarations'
3660 Do not warn about uses of functions (*note Function Attributes::),
3661 variables (*note Variable Attributes::), and types (*note Type
3662 Attributes::) marked as deprecated by using the `deprecated'
3666 Do not warn about compile-time overflow in constant expressions.
3668 `-Woverride-init (C and Objective-C only)'
3669 Warn if an initialized field without side effects is overridden
3670 when using designated initializers (*note Designated Initializers:
3673 This warning is included in `-Wextra'. To get other `-Wextra'
3674 warnings without this one, use `-Wextra -Wno-override-init'.
3677 Warn if a structure is given the packed attribute, but the packed
3678 attribute has no effect on the layout or size of the structure.
3679 Such structures may be mis-aligned for little benefit. For
3680 instance, in this code, the variable `f.x' in `struct bar' will be
3681 misaligned even though `struct bar' does not itself have the
3687 } __attribute__((packed));
3694 Warn if padding is included in a structure, either to align an
3695 element of the structure or to align the whole structure.
3696 Sometimes when this happens it is possible to rearrange the fields
3697 of the structure to reduce the padding and so make the structure
3701 Warn if anything is declared more than once in the same scope,
3702 even in cases where multiple declaration is valid and changes
3705 `-Wnested-externs (C and Objective-C only)'
3706 Warn if an `extern' declaration is encountered within a function.
3708 `-Wunreachable-code'
3709 Warn if the compiler detects that code will never be executed.
3711 This option is intended to warn when the compiler detects that at
3712 least a whole line of source code will never be executed, because
3713 some condition is never satisfied or because it is after a
3714 procedure that never returns.
3716 It is possible for this option to produce a warning even though
3717 there are circumstances under which part of the affected line can
3718 be executed, so care should be taken when removing
3719 apparently-unreachable code.
3721 For instance, when a function is inlined, a warning may mean that
3722 the line is unreachable in only one inlined copy of the function.
3724 This option is not made part of `-Wall' because in a debugging
3725 version of a program there is often substantial code which checks
3726 correct functioning of the program and is, hopefully, unreachable
3727 because the program does work. Another common use of unreachable
3728 code is to provide behavior which is selectable at compile-time.
3731 Warn if a function can not be inlined and it was declared as
3732 inline. Even with this option, the compiler will not warn about
3733 failures to inline functions declared in system headers.
3735 The compiler uses a variety of heuristics to determine whether or
3736 not to inline a function. For example, the compiler takes into
3737 account the size of the function being inlined and the amount of
3738 inlining that has already been done in the current function.
3739 Therefore, seemingly insignificant changes in the source program
3740 can cause the warnings produced by `-Winline' to appear or
3743 `-Wno-invalid-offsetof (C++ and Objective-C++ only)'
3744 Suppress warnings from applying the `offsetof' macro to a non-POD
3745 type. According to the 1998 ISO C++ standard, applying `offsetof'
3746 to a non-POD type is undefined. In existing C++ implementations,
3747 however, `offsetof' typically gives meaningful results even when
3748 applied to certain kinds of non-POD types. (Such as a simple
3749 `struct' that fails to be a POD type only by virtue of having a
3750 constructor.) This flag is for users who are aware that they are
3751 writing nonportable code and who have deliberately chosen to
3752 ignore the warning about it.
3754 The restrictions on `offsetof' may be relaxed in a future version
3755 of the C++ standard.
3757 `-Wno-int-to-pointer-cast (C and Objective-C only)'
3758 Suppress warnings from casts to pointer type of an integer of a
3761 `-Wno-pointer-to-int-cast (C and Objective-C only)'
3762 Suppress warnings from casts from a pointer to an integer type of a
3766 Warn if a precompiled header (*note Precompiled Headers::) is
3767 found in the search path but can't be used.
3770 Warn if `long long' type is used. This is default. To inhibit
3771 the warning messages, use `-Wno-long-long'. Flags `-Wlong-long'
3772 and `-Wno-long-long' are taken into account only when `-pedantic'
3776 Warn if variadic macros are used in pedantic ISO C90 mode, or the
3777 GNU alternate syntax when in pedantic ISO C99 mode. This is
3778 default. To inhibit the warning messages, use
3779 `-Wno-variadic-macros'.
3782 Warn if variable length array is used in the code. `-Wno-vla'
3783 will prevent the `-pedantic' warning of the variable length array.
3785 `-Wvolatile-register-var'
3786 Warn if a register variable is declared volatile. The volatile
3787 modifier does not inhibit all optimizations that may eliminate
3788 reads and/or writes to register variables.
3790 `-Wdisabled-optimization'
3791 Warn if a requested optimization pass is disabled. This warning
3792 does not generally indicate that there is anything wrong with your
3793 code; it merely indicates that GCC's optimizers were unable to
3794 handle the code effectively. Often, the problem is that your code
3795 is too big or too complex; GCC will refuse to optimize programs
3796 when the optimization itself is likely to take inordinate amounts
3799 `-Wpointer-sign (C and Objective-C only)'
3800 Warn for pointer argument passing or assignment with different
3801 signedness. This option is only supported for C and Objective-C.
3802 It is implied by `-Wall' and by `-pedantic', which can be disabled
3803 with `-Wno-pointer-sign'.
3806 This option is only active when `-fstack-protector' is active. It
3807 warns about functions that will not be protected against stack
3810 `-Woverlength-strings'
3811 Warn about string constants which are longer than the "minimum
3812 maximum" length specified in the C standard. Modern compilers
3813 generally allow string constants which are much longer than the
3814 standard's minimum limit, but very portable programs should avoid
3815 using longer strings.
3817 The limit applies _after_ string constant concatenation, and does
3818 not count the trailing NUL. In C89, the limit was 509 characters;
3819 in C99, it was raised to 4095. C++98 does not specify a normative
3820 minimum maximum, so we do not diagnose overlength strings in C++.
3822 This option is implied by `-pedantic', and can be disabled with
3823 `-Wno-overlength-strings'.
3826 File: gcc.info, Node: Debugging Options, Next: Optimize Options, Prev: Warning Options, Up: Invoking GCC
3828 3.9 Options for Debugging Your Program or GCC
3829 =============================================
3831 GCC has various special options that are used for debugging either your
3835 Produce debugging information in the operating system's native
3836 format (stabs, COFF, XCOFF, or DWARF 2). GDB can work with this
3837 debugging information.
3839 On most systems that use stabs format, `-g' enables use of extra
3840 debugging information that only GDB can use; this extra information
3841 makes debugging work better in GDB but will probably make other
3842 debuggers crash or refuse to read the program. If you want to
3843 control for certain whether to generate the extra information, use
3844 `-gstabs+', `-gstabs', `-gxcoff+', `-gxcoff', or `-gvms' (see
3847 GCC allows you to use `-g' with `-O'. The shortcuts taken by
3848 optimized code may occasionally produce surprising results: some
3849 variables you declared may not exist at all; flow of control may
3850 briefly move where you did not expect it; some statements may not
3851 be executed because they compute constant results or their values
3852 were already at hand; some statements may execute in different
3853 places because they were moved out of loops.
3855 Nevertheless it proves possible to debug optimized output. This
3856 makes it reasonable to use the optimizer for programs that might
3859 The following options are useful when GCC is generated with the
3860 capability for more than one debugging format.
3863 Produce debugging information for use by GDB. This means to use
3864 the most expressive format available (DWARF 2, stabs, or the
3865 native format if neither of those are supported), including GDB
3866 extensions if at all possible.
3869 Produce debugging information in stabs format (if that is
3870 supported), without GDB extensions. This is the format used by
3871 DBX on most BSD systems. On MIPS, Alpha and System V Release 4
3872 systems this option produces stabs debugging output which is not
3873 understood by DBX or SDB. On System V Release 4 systems this
3874 option requires the GNU assembler.
3876 `-feliminate-unused-debug-symbols'
3877 Produce debugging information in stabs format (if that is
3878 supported), for only symbols that are actually used.
3880 `-femit-class-debug-always'
3881 Instead of emitting debugging information for a C++ class in only
3882 one object file, emit it in all object files using the class.
3883 This option should be used only with debuggers that are unable to
3884 handle the way GCC normally emits debugging information for
3885 classes because using this option will increase the size of
3886 debugging information by as much as a factor of two.
3889 Produce debugging information in stabs format (if that is
3890 supported), using GNU extensions understood only by the GNU
3891 debugger (GDB). The use of these extensions is likely to make
3892 other debuggers crash or refuse to read the program.
3895 Produce debugging information in COFF format (if that is
3896 supported). This is the format used by SDB on most System V
3897 systems prior to System V Release 4.
3900 Produce debugging information in XCOFF format (if that is
3901 supported). This is the format used by the DBX debugger on IBM
3905 Produce debugging information in XCOFF format (if that is
3906 supported), using GNU extensions understood only by the GNU
3907 debugger (GDB). The use of these extensions is likely to make
3908 other debuggers crash or refuse to read the program, and may cause
3909 assemblers other than the GNU assembler (GAS) to fail with an
3913 Produce debugging information in DWARF version 2 format (if that is
3914 supported). This is the format used by DBX on IRIX 6. With this
3915 option, GCC uses features of DWARF version 3 when they are useful;
3916 version 3 is upward compatible with version 2, but may still cause
3917 problems for older debuggers.
3920 Produce debugging information in VMS debug format (if that is
3921 supported). This is the format used by DEBUG on VMS systems.
3929 Request debugging information and also use LEVEL to specify how
3930 much information. The default level is 2.
3932 Level 0 produces no debug information at all. Thus, `-g0' negates
3935 Level 1 produces minimal information, enough for making backtraces
3936 in parts of the program that you don't plan to debug. This
3937 includes descriptions of functions and external variables, but no
3938 information about local variables and no line numbers.
3940 Level 3 includes extra information, such as all the macro
3941 definitions present in the program. Some debuggers support macro
3942 expansion when you use `-g3'.
3944 `-gdwarf-2' does not accept a concatenated debug level, because
3945 GCC used to support an option `-gdwarf' that meant to generate
3946 debug information in version 1 of the DWARF format (which is very
3947 different from version 2), and it would have been too confusing.
3948 That debug format is long obsolete, but the option cannot be
3949 changed now. Instead use an additional `-gLEVEL' option to change
3950 the debug level for DWARF2.
3952 `-feliminate-dwarf2-dups'
3953 Compress DWARF2 debugging information by eliminating duplicated
3954 information about each symbol. This option only makes sense when
3955 generating DWARF2 debugging information with `-gdwarf-2'.
3957 `-femit-struct-debug-baseonly'
3958 Emit debug information for struct-like types only when the base
3959 name of the compilation source file matches the base name of file
3960 in which the struct was defined.
3962 This option substantially reduces the size of debugging
3963 information, but at significant potential loss in type information
3964 to the debugger. See `-femit-struct-debug-reduced' for a less
3965 aggressive option. See `-femit-struct-debug-detailed' for more
3968 This option works only with DWARF 2.
3970 `-femit-struct-debug-reduced'
3971 Emit debug information for struct-like types only when the base
3972 name of the compilation source file matches the base name of file
3973 in which the type was defined, unless the struct is a template or
3974 defined in a system header.
3976 This option significantly reduces the size of debugging
3977 information, with some potential loss in type information to the
3978 debugger. See `-femit-struct-debug-baseonly' for a more
3979 aggressive option. See `-femit-struct-debug-detailed' for more
3982 This option works only with DWARF 2.
3984 `-femit-struct-debug-detailed[=SPEC-LIST]'
3985 Specify the struct-like types for which the compiler will generate
3986 debug information. The intent is to reduce duplicate struct debug
3987 information between different object files within the same program.
3989 This option is a detailed version of `-femit-struct-debug-reduced'
3990 and `-femit-struct-debug-baseonly', which will serve for most
3993 A specification has the syntax
3994 [`dir:'|`ind:'][`ord:'|`gen:'](`any'|`sys'|`base'|`none')
3996 The optional first word limits the specification to structs that
3997 are used directly (`dir:') or used indirectly (`ind:'). A struct
3998 type is used directly when it is the type of a variable, member.
3999 Indirect uses arise through pointers to structs. That is, when
4000 use of an incomplete struct would be legal, the use is indirect.
4001 An example is `struct one direct; struct two * indirect;'.
4003 The optional second word limits the specification to ordinary
4004 structs (`ord:') or generic structs (`gen:'). Generic structs are
4005 a bit complicated to explain. For C++, these are non-explicit
4006 specializations of template classes, or non-template classes
4007 within the above. Other programming languages have generics, but
4008 `-femit-struct-debug-detailed' does not yet implement them.
4010 The third word specifies the source files for those structs for
4011 which the compiler will emit debug information. The values `none'
4012 and `any' have the normal meaning. The value `base' means that
4013 the base of name of the file in which the type declaration appears
4014 must match the base of the name of the main compilation file. In
4015 practice, this means that types declared in `foo.c' and `foo.h'
4016 will have debug information, but types declared in other header
4017 will not. The value `sys' means those types satisfying `base' or
4018 declared in system or compiler headers.
4020 You may need to experiment to determine the best settings for your
4023 The default is `-femit-struct-debug-detailed=all'.
4025 This option works only with DWARF 2.
4027 `-fno-merge-debug-strings'
4028 Direct the linker to merge together strings which are identical in
4029 different object files. This is not supported by all assemblers or
4030 linker. This decreases the size of the debug information in the
4031 output file at the cost of increasing link processing time. This
4034 `-fdebug-prefix-map=OLD=NEW'
4035 When compiling files in directory `OLD', record debugging
4036 information describing them as in `NEW' instead.
4039 Generate extra code to write profile information suitable for the
4040 analysis program `prof'. You must use this option when compiling
4041 the source files you want data about, and you must also use it when
4045 Generate extra code to write profile information suitable for the
4046 analysis program `gprof'. You must use this option when compiling
4047 the source files you want data about, and you must also use it when
4051 Makes the compiler print out each function name as it is compiled,
4052 and print some statistics about each pass when it finishes.
4055 Makes the compiler print some statistics about the time consumed
4056 by each pass when it finishes.
4059 Makes the compiler print some statistics about permanent memory
4060 allocation when it finishes.
4062 `-fpre-ipa-mem-report'
4064 `-fpost-ipa-mem-report'
4065 Makes the compiler print some statistics about permanent memory
4066 allocation before or after interprocedural optimization.
4069 Add code so that program flow "arcs" are instrumented. During
4070 execution the program records how many times each branch and call
4071 is executed and how many times it is taken or returns. When the
4072 compiled program exits it saves this data to a file called
4073 `AUXNAME.gcda' for each source file. The data may be used for
4074 profile-directed optimizations (`-fbranch-probabilities'), or for
4075 test coverage analysis (`-ftest-coverage'). Each object file's
4076 AUXNAME is generated from the name of the output file, if
4077 explicitly specified and it is not the final executable, otherwise
4078 it is the basename of the source file. In both cases any suffix
4079 is removed (e.g. `foo.gcda' for input file `dir/foo.c', or
4080 `dir/foo.gcda' for output file specified as `-o dir/foo.o').
4081 *Note Cross-profiling::.
4084 This option is used to compile and link code instrumented for
4085 coverage analysis. The option is a synonym for `-fprofile-arcs'
4086 `-ftest-coverage' (when compiling) and `-lgcov' (when linking).
4087 See the documentation for those options for more details.
4089 * Compile the source files with `-fprofile-arcs' plus
4090 optimization and code generation options. For test coverage
4091 analysis, use the additional `-ftest-coverage' option. You
4092 do not need to profile every source file in a program.
4094 * Link your object files with `-lgcov' or `-fprofile-arcs' (the
4095 latter implies the former).
4097 * Run the program on a representative workload to generate the
4098 arc profile information. This may be repeated any number of
4099 times. You can run concurrent instances of your program, and
4100 provided that the file system supports locking, the data
4101 files will be correctly updated. Also `fork' calls are
4102 detected and correctly handled (double counting will not
4105 * For profile-directed optimizations, compile the source files
4106 again with the same optimization and code generation options
4107 plus `-fbranch-probabilities' (*note Options that Control
4108 Optimization: Optimize Options.).
4110 * For test coverage analysis, use `gcov' to produce human
4111 readable information from the `.gcno' and `.gcda' files.
4112 Refer to the `gcov' documentation for further information.
4115 With `-fprofile-arcs', for each function of your program GCC
4116 creates a program flow graph, then finds a spanning tree for the
4117 graph. Only arcs that are not on the spanning tree have to be
4118 instrumented: the compiler adds code to count the number of times
4119 that these arcs are executed. When an arc is the only exit or
4120 only entrance to a block, the instrumentation code can be added to
4121 the block; otherwise, a new basic block must be created to hold
4122 the instrumentation code.
4125 Produce a notes file that the `gcov' code-coverage utility (*note
4126 `gcov'--a Test Coverage Program: Gcov.) can use to show program
4127 coverage. Each source file's note file is called `AUXNAME.gcno'.
4128 Refer to the `-fprofile-arcs' option above for a description of
4129 AUXNAME and instructions on how to generate test coverage data.
4130 Coverage data will match the source files more closely, if you do
4134 Print the name and the counter upperbound for all debug counters.
4136 `-fdbg-cnt=COUNTER-VALUE-LIST'
4137 Set the internal debug counter upperbound. COUNTER-VALUE-LIST is a
4138 comma-separated list of NAME:VALUE pairs which sets the upperbound
4139 of each debug counter NAME to VALUE. All debug counters have the
4140 initial upperbound of UINT_MAX, thus dbg_cnt() returns true always
4141 unless the upperbound is set by this option. e.g. With
4142 -fdbg-cnt=dce:10,tail_call:0 dbg_cnt(dce) will return true only
4143 for first 10 invocations and dbg_cnt(tail_call) will return false
4149 Says to make debugging dumps during compilation at times specified
4150 by LETTERS. This is used for debugging the RTL-based passes of
4151 the compiler. The file names for most of the dumps are made by
4152 appending a pass number and a word to the DUMPNAME. DUMPNAME is
4153 generated from the name of the output file, if explicitly
4154 specified and it is not an executable, otherwise it is the
4155 basename of the source file. These switches may have different
4156 effects when `-E' is used for preprocessing.
4158 Most debug dumps can be enabled either passing a letter to the `-d'
4159 option, or with a long `-fdump-rtl' switch; here are the possible
4160 letters for use in LETTERS and PASS, and their meanings:
4163 Annotate the assembler output with miscellaneous debugging
4168 Dump after block reordering, to `FILE.148r.bbro'.
4171 `-fdump-rtl-combine'
4172 Dump after the RTL instruction combination pass, to the file
4173 `FILE.129r.combine'.
4178 `-dC' and `-fdump-rtl-ce1' enable dumping after the first if
4179 conversion, to the file `FILE.117r.ce1'. `-dC' and
4180 `-fdump-rtl-ce2' enable dumping after the second if
4181 conversion, to the file `FILE.130r.ce2'.
4186 `-dd' and `-fdump-rtl-btl' enable dumping after branch target
4187 load optimization, to `FILE.31.btl'. `-dd' and
4188 `-fdump-rtl-dbr' enable dumping after delayed branch
4189 scheduling, to `FILE.36.dbr'.
4192 Dump all macro definitions, at the end of preprocessing, in
4193 addition to normal output.
4197 Dump after the third if conversion, to `FILE.146r.ce3'.
4202 `-df' and `-fdump-rtl-cfg' enable dumping after control and
4203 data flow analysis, to `FILE.116r.cfg'. `-df' and
4204 `-fdump-rtl-cfg' enable dumping dump after life analysis, to
4205 `FILE.128r.life1' and `FILE.135r.life2'.
4209 Dump after global register allocation, to `FILE.139r.greg'.
4214 `-dG' and `-fdump-rtl-gcse' enable dumping after GCSE, to
4215 `FILE.114r.gcse'. `-dG' and `-fdump-rtl-bypass' enable
4216 dumping after jump bypassing and control flow optimizations,
4217 to `FILE.115r.bypass'.
4221 Dump after finalization of EH handling code, to `FILE.02.eh'.
4224 `-fdump-rtl-sibling'
4225 Dump after sibling call optimizations, to `FILE.106r.sibling'.
4229 Dump after the first jump optimization, to `FILE.112r.jump'.
4233 Dump after conversion from GCC's "flat register file"
4234 registers to the x87's stack-like registers, to
4239 Dump after local register allocation, to `FILE.138r.lreg'.
4243 `-dL' and `-fdump-rtl-loop2' enable dumping after the loop
4244 optimization pass, to `FILE.119r.loop2',
4245 `FILE.120r.loop2_init', `FILE.121r.loop2_invariant', and
4246 `FILE.125r.loop2_done'.
4250 Dump after modulo scheduling, to `FILE.136r.sms'.
4254 Dump after performing the machine dependent reorganization
4255 pass, to `FILE.155r.mach' if that pass exists.
4259 Dump after register renumbering, to `FILE.147r.rnreg'.
4262 `-fdump-rtl-regmove'
4263 Dump after the register move pass, to `FILE.132r.regmove'.
4266 `-fdump-rtl-postreload'
4267 Dump after post-reload optimizations, to `FILE.24.postreload'.
4271 Dump after RTL generation, to `FILE.104r.expand'.
4275 Dump after the second scheduling pass, to `FILE.149r.sched2'.
4279 Dump after CSE (including the jump optimization that
4280 sometimes follows CSE), to `FILE.113r.cse'.
4284 Dump after the first scheduling pass, to `FILE.136r.sched1'.
4288 Dump after the second CSE pass (including the jump
4289 optimization that sometimes follows CSE), to `FILE.127r.cse2'.
4293 Dump after running tracer, to `FILE.118r.tracer'.
4297 `-fdump-rtl-vartrack'
4298 `-dV' and `-fdump-rtl-vpt' enable dumping after the value
4299 profile transformations, to `FILE.10.vpt'. `-dV' and
4300 `-fdump-rtl-vartrack' enable dumping after variable tracking,
4301 to `FILE.154r.vartrack'.
4305 Dump after the second flow pass, to `FILE.142r.flow2'.
4308 `-fdump-rtl-peephole2'
4309 Dump after the peephole pass, to `FILE.145r.peephole2'.
4313 Dump after live range splitting, to `FILE.126r.web'.
4317 Produce all the dumps listed above.
4320 Produce a core dump whenever an error occurs.
4323 Print statistics on memory usage, at the end of the run, to
4327 Annotate the assembler output with a comment indicating which
4328 pattern and alternative was used. The length of each
4329 instruction is also printed.
4332 Dump the RTL in the assembler output as a comment before each
4333 instruction. Also turns on `-dp' annotation.
4336 For each of the other indicated dump files (either with `-d'
4337 or `-fdump-rtl-PASS'), dump a representation of the control
4338 flow graph suitable for viewing with VCG to `FILE.PASS.vcg'.
4341 Just generate RTL for a function instead of compiling it.
4342 Usually used with `r' (`-fdump-rtl-expand').
4345 Dump debugging information during parsing, to standard error.
4348 When doing debugging dumps (see `-d' option above), suppress
4349 address output. This makes it more feasible to use diff on
4350 debugging dumps for compiler invocations with different compiler
4351 binaries and/or different text / bss / data / heap / stack / dso
4355 When doing debugging dumps (see `-d' option above), suppress
4356 instruction numbers and address output. This makes it more
4357 feasible to use diff on debugging dumps for compiler invocations
4358 with different options, in particular with and without `-g'.
4360 `-fdump-translation-unit (C++ only)'
4361 `-fdump-translation-unit-OPTIONS (C++ only)'
4362 Dump a representation of the tree structure for the entire
4363 translation unit to a file. The file name is made by appending
4364 `.tu' to the source file name. If the `-OPTIONS' form is used,
4365 OPTIONS controls the details of the dump as described for the
4366 `-fdump-tree' options.
4368 `-fdump-class-hierarchy (C++ only)'
4369 `-fdump-class-hierarchy-OPTIONS (C++ only)'
4370 Dump a representation of each class's hierarchy and virtual
4371 function table layout to a file. The file name is made by
4372 appending `.class' to the source file name. If the `-OPTIONS'
4373 form is used, OPTIONS controls the details of the dump as
4374 described for the `-fdump-tree' options.
4377 Control the dumping at various stages of inter-procedural analysis
4378 language tree to a file. The file name is generated by appending
4379 a switch specific suffix to the source file name. The following
4383 Enables all inter-procedural analysis dumps.
4386 Dumps information about call-graph optimization, unused
4387 function removal, and inlining decisions.
4390 Dump after function inlining.
4393 `-fdump-tree-SWITCH'
4394 `-fdump-tree-SWITCH-OPTIONS'
4395 Control the dumping at various stages of processing the
4396 intermediate language tree to a file. The file name is generated
4397 by appending a switch specific suffix to the source file name. If
4398 the `-OPTIONS' form is used, OPTIONS is a list of `-' separated
4399 options that control the details of the dump. Not all options are
4400 applicable to all dumps, those which are not meaningful will be
4401 ignored. The following options are available
4404 Print the address of each node. Usually this is not
4405 meaningful as it changes according to the environment and
4406 source file. Its primary use is for tying up a dump file
4407 with a debug environment.
4410 Inhibit dumping of members of a scope or body of a function
4411 merely because that scope has been reached. Only dump such
4412 items when they are directly reachable by some other path.
4413 When dumping pretty-printed trees, this option inhibits
4414 dumping the bodies of control structures.
4417 Print a raw representation of the tree. By default, trees are
4418 pretty-printed into a C-like representation.
4421 Enable more detailed dumps (not honored by every dump option).
4424 Enable dumping various statistics about the pass (not honored
4425 by every dump option).
4428 Enable showing basic block boundaries (disabled in raw dumps).
4431 Enable showing virtual operands for every statement.
4434 Enable showing line numbers for statements.
4437 Enable showing the unique ID (`DECL_UID') for each variable.
4440 Turn on all options, except `raw', `slim' and `lineno'.
4442 The following tree dumps are possible:
4444 Dump before any tree based optimization, to `FILE.original'.
4447 Dump after all tree based optimization, to `FILE.optimized'.
4450 Dump each function before and after the gimplification pass
4451 to a file. The file name is made by appending `.gimple' to
4452 the source file name.
4455 Dump the control flow graph of each function to a file. The
4456 file name is made by appending `.cfg' to the source file name.
4459 Dump the control flow graph of each function to a file in VCG
4460 format. The file name is made by appending `.vcg' to the
4461 source file name. Note that if the file contains more than
4462 one function, the generated file cannot be used directly by
4463 VCG. You will need to cut and paste each function's graph
4464 into its own separate file first.
4467 Dump each function after copying loop headers. The file name
4468 is made by appending `.ch' to the source file name.
4471 Dump SSA related information to a file. The file name is
4472 made by appending `.ssa' to the source file name.
4475 Dump structure aliasing variable information to a file. This
4476 file name is made by appending `.salias' to the source file
4480 Dump aliasing information for each function. The file name
4481 is made by appending `.alias' to the source file name.
4484 Dump each function after CCP. The file name is made by
4485 appending `.ccp' to the source file name.
4488 Dump each function after STORE-CCP. The file name is made by
4489 appending `.storeccp' to the source file name.
4492 Dump trees after partial redundancy elimination. The file
4493 name is made by appending `.pre' to the source file name.
4496 Dump trees after full redundancy elimination. The file name
4497 is made by appending `.fre' to the source file name.
4500 Dump trees after copy propagation. The file name is made by
4501 appending `.copyprop' to the source file name.
4504 Dump trees after store copy-propagation. The file name is
4505 made by appending `.store_copyprop' to the source file name.
4508 Dump each function after dead code elimination. The file
4509 name is made by appending `.dce' to the source file name.
4512 Dump each function after adding mudflap instrumentation. The
4513 file name is made by appending `.mudflap' to the source file
4517 Dump each function after performing scalar replacement of
4518 aggregates. The file name is made by appending `.sra' to the
4522 Dump each function after performing code sinking. The file
4523 name is made by appending `.sink' to the source file name.
4526 Dump each function after applying dominator tree
4527 optimizations. The file name is made by appending `.dom' to
4528 the source file name.
4531 Dump each function after applying dead store elimination.
4532 The file name is made by appending `.dse' to the source file
4536 Dump each function after optimizing PHI nodes into
4537 straightline code. The file name is made by appending
4538 `.phiopt' to the source file name.
4541 Dump each function after forward propagating single use
4542 variables. The file name is made by appending `.forwprop' to
4543 the source file name.
4546 Dump each function after applying the copy rename
4547 optimization. The file name is made by appending
4548 `.copyrename' to the source file name.
4551 Dump each function after applying the named return value
4552 optimization on generic trees. The file name is made by
4553 appending `.nrv' to the source file name.
4556 Dump each function after applying vectorization of loops.
4557 The file name is made by appending `.vect' to the source file
4561 Dump each function after Value Range Propagation (VRP). The
4562 file name is made by appending `.vrp' to the source file name.
4565 Enable all the available tree dumps with the flags provided
4568 `-ftree-vectorizer-verbose=N'
4569 This option controls the amount of debugging output the vectorizer
4570 prints. This information is written to standard error, unless
4571 `-fdump-tree-all' or `-fdump-tree-vect' is specified, in which
4572 case it is output to the usual dump listing file, `.vect'. For
4573 N=0 no diagnostic information is reported. If N=1 the vectorizer
4574 reports each loop that got vectorized, and the total number of
4575 loops that got vectorized. If N=2 the vectorizer also reports
4576 non-vectorized loops that passed the first analysis phase
4577 (vect_analyze_loop_form) - i.e. countable, inner-most, single-bb,
4578 single-entry/exit loops. This is the same verbosity level that
4579 `-fdump-tree-vect-stats' uses. Higher verbosity levels mean
4580 either more information dumped for each reported loop, or same
4581 amount of information reported for more loops: If N=3, alignment
4582 related information is added to the reports. If N=4,
4583 data-references related information (e.g. memory dependences,
4584 memory access-patterns) is added to the reports. If N=5, the
4585 vectorizer reports also non-vectorized inner-most loops that did
4586 not pass the first analysis phase (i.e., may not be countable, or
4587 may have complicated control-flow). If N=6, the vectorizer
4588 reports also non-vectorized nested loops. For N=7, all the
4589 information the vectorizer generates during its analysis and
4590 transformation is reported. This is the same verbosity level that
4591 `-fdump-tree-vect-details' uses.
4593 `-frandom-seed=STRING'
4594 This option provides a seed that GCC uses when it would otherwise
4595 use random numbers. It is used to generate certain symbol names
4596 that have to be different in every compiled file. It is also used
4597 to place unique stamps in coverage data files and the object files
4598 that produce them. You can use the `-frandom-seed' option to
4599 produce reproducibly identical object files.
4601 The STRING should be different for every file you compile.
4604 On targets that use instruction scheduling, this option controls
4605 the amount of debugging output the scheduler prints. This
4606 information is written to standard error, unless `-dS' or `-dR' is
4607 specified, in which case it is output to the usual dump listing
4608 file, `.sched' or `.sched2' respectively. However for N greater
4609 than nine, the output is always printed to standard error.
4611 For N greater than zero, `-fsched-verbose' outputs the same
4612 information as `-dRS'. For N greater than one, it also output
4613 basic block probabilities, detailed ready list information and
4614 unit/insn info. For N greater than two, it includes RTL at abort
4615 point, control-flow and regions info. And for N over four,
4616 `-fsched-verbose' also includes dependence info.
4619 Store the usual "temporary" intermediate files permanently; place
4620 them in the current directory and name them based on the source
4621 file. Thus, compiling `foo.c' with `-c -save-temps' would produce
4622 files `foo.i' and `foo.s', as well as `foo.o'. This creates a
4623 preprocessed `foo.i' output file even though the compiler now
4624 normally uses an integrated preprocessor.
4626 When used in combination with the `-x' command line option,
4627 `-save-temps' is sensible enough to avoid over writing an input
4628 source file with the same extension as an intermediate file. The
4629 corresponding intermediate file may be obtained by renaming the
4630 source file before using `-save-temps'.
4633 Report the CPU time taken by each subprocess in the compilation
4634 sequence. For C source files, this is the compiler proper and
4635 assembler (plus the linker if linking is done). The output looks
4641 The first number on each line is the "user time", that is time
4642 spent executing the program itself. The second number is "system
4643 time", time spent executing operating system routines on behalf of
4644 the program. Both numbers are in seconds.
4647 Run variable tracking pass. It computes where variables are
4648 stored at each position in code. Better debugging information is
4649 then generated (if the debugging information format supports this
4652 It is enabled by default when compiling with optimization (`-Os',
4653 `-O', `-O2', ...), debugging information (`-g') and the debug info
4656 `-print-file-name=LIBRARY'
4657 Print the full absolute name of the library file LIBRARY that
4658 would be used when linking--and don't do anything else. With this
4659 option, GCC does not compile or link anything; it just prints the
4662 `-print-multi-directory'
4663 Print the directory name corresponding to the multilib selected by
4664 any other switches present in the command line. This directory is
4665 supposed to exist in `GCC_EXEC_PREFIX'.
4668 Print the mapping from multilib directory names to compiler
4669 switches that enable them. The directory name is separated from
4670 the switches by `;', and each switch starts with an `@' instead of
4671 the `-', without spaces between multiple switches. This is
4672 supposed to ease shell-processing.
4674 `-print-prog-name=PROGRAM'
4675 Like `-print-file-name', but searches for a program such as `cpp'.
4677 `-print-libgcc-file-name'
4678 Same as `-print-file-name=libgcc.a'.
4680 This is useful when you use `-nostdlib' or `-nodefaultlibs' but
4681 you do want to link with `libgcc.a'. You can do
4683 gcc -nostdlib FILES... `gcc -print-libgcc-file-name`
4685 `-print-search-dirs'
4686 Print the name of the configured installation directory and a list
4687 of program and library directories `gcc' will search--and don't do
4690 This is useful when `gcc' prints the error message `installation
4691 problem, cannot exec cpp0: No such file or directory'. To resolve
4692 this you either need to put `cpp0' and the other compiler
4693 components where `gcc' expects to find them, or you can set the
4694 environment variable `GCC_EXEC_PREFIX' to the directory where you
4695 installed them. Don't forget the trailing `/'. *Note Environment
4698 `-print-sysroot-headers-suffix'
4699 Print the suffix added to the target sysroot when searching for
4700 headers, or give an error if the compiler is not configured with
4701 such a suffix--and don't do anything else.
4704 Print the compiler's target machine (for example,
4705 `i686-pc-linux-gnu')--and don't do anything else.
4708 Print the compiler version (for example, `3.0')--and don't do
4712 Print the compiler's built-in specs--and don't do anything else.
4713 (This is used when GCC itself is being built.) *Note Spec Files::.
4715 `-feliminate-unused-debug-types'
4716 Normally, when producing DWARF2 output, GCC will emit debugging
4717 information for all types declared in a compilation unit,
4718 regardless of whether or not they are actually used in that
4719 compilation unit. Sometimes this is useful, such as if, in the
4720 debugger, you want to cast a value to a type that is not actually
4721 used in your program (but is declared). More often, however, this
4722 results in a significant amount of wasted space. With this
4723 option, GCC will avoid producing debug symbol output for types
4724 that are nowhere used in the source file being compiled.
4727 File: gcc.info, Node: Optimize Options, Next: Preprocessor Options, Prev: Debugging Options, Up: Invoking GCC
4729 3.10 Options That Control Optimization
4730 ======================================
4732 These options control various sorts of optimizations.
4734 Without any optimization option, the compiler's goal is to reduce the
4735 cost of compilation and to make debugging produce the expected results.
4736 Statements are independent: if you stop the program with a breakpoint
4737 between statements, you can then assign a new value to any variable or
4738 change the program counter to any other statement in the function and
4739 get exactly the results you would expect from the source code.
4741 Turning on optimization flags makes the compiler attempt to improve
4742 the performance and/or code size at the expense of compilation time and
4743 possibly the ability to debug the program.
4745 The compiler performs optimization based on the knowledge it has of
4746 the program. Optimization levels `-O' and above, in particular, enable
4747 _unit-at-a-time_ mode, which allows the compiler to consider
4748 information gained from later functions in the file when compiling a
4749 function. Compiling multiple files at once to a single output file in
4750 _unit-at-a-time_ mode allows the compiler to use information gained
4751 from all of the files when compiling each of them.
4753 Not all optimizations are controlled directly by a flag. Only
4754 optimizations that have a flag are listed.
4758 Optimize. Optimizing compilation takes somewhat more time, and a
4759 lot more memory for a large function.
4761 With `-O', the compiler tries to reduce code size and execution
4762 time, without performing any optimizations that take a great deal
4763 of compilation time.
4765 `-O' turns on the following optimization flags:
4772 -fguess-branch-probability
4775 -finline-small-functions
4784 -ftree-dominator-opts
4791 `-O' also turns on `-fomit-frame-pointer' on machines where doing
4792 so does not interfere with debugging.
4795 Optimize even more. GCC performs nearly all supported
4796 optimizations that do not involve a space-speed tradeoff. The
4797 compiler does not perform loop unrolling or function inlining when
4798 you specify `-O2'. As compared to `-O', this option increases
4799 both compilation time and the performance of the generated code.
4801 `-O2' turns on all optimization flags specified by `-O'. It also
4802 turns on the following optimization flags:
4804 -falign-functions -falign-jumps
4805 -falign-loops -falign-labels
4808 -fcse-follow-jumps -fcse-skip-blocks
4809 -fdelete-null-pointer-checks
4810 -fexpensive-optimizations
4812 -foptimize-sibling-calls
4815 -freorder-blocks -freorder-functions
4816 -frerun-cse-after-loop
4817 -fsched-interblock -fsched-spec
4818 -fschedule-insns -fschedule-insns2
4819 -fstrict-aliasing -fstrict-overflow
4823 Please note the warning under `-fgcse' about invoking `-O2' on
4824 programs that use computed gotos.
4827 Optimize yet more. `-O3' turns on all optimizations specified by
4828 `-O2' and also turns on the `-finline-functions',
4829 `-funswitch-loops', `-fpredictive-commoning',
4830 `-fgcse-after-reload' and `-ftree-vectorize' options.
4833 Reduce compilation time and make debugging produce the expected
4834 results. This is the default.
4837 Optimize for size. `-Os' enables all `-O2' optimizations that do
4838 not typically increase code size. It also performs further
4839 optimizations designed to reduce code size.
4841 `-Os' disables the following optimization flags:
4842 -falign-functions -falign-jumps -falign-loops
4843 -falign-labels -freorder-blocks -freorder-blocks-and-partition
4844 -fprefetch-loop-arrays -ftree-vect-loop-version
4846 If you use multiple `-O' options, with or without level numbers,
4847 the last such option is the one that is effective.
4849 Options of the form `-fFLAG' specify machine-independent flags. Most
4850 flags have both positive and negative forms; the negative form of
4851 `-ffoo' would be `-fno-foo'. In the table below, only one of the forms
4852 is listed--the one you typically will use. You can figure out the
4853 other form by either removing `no-' or adding it.
4855 The following options control specific optimizations. They are either
4856 activated by `-O' options or are related to ones that are. You can use
4857 the following flags in the rare cases when "fine-tuning" of
4858 optimizations to be performed is desired.
4860 `-fno-default-inline'
4861 Do not make member functions inline by default merely because they
4862 are defined inside the class scope (C++ only). Otherwise, when
4863 you specify `-O', member functions defined inside class scope are
4864 compiled inline by default; i.e., you don't need to add `inline'
4865 in front of the member function name.
4868 Always pop the arguments to each function call as soon as that
4869 function returns. For machines which must pop arguments after a
4870 function call, the compiler normally lets arguments accumulate on
4871 the stack for several function calls and pops them all at once.
4873 Disabled at levels `-O', `-O2', `-O3', `-Os'.
4875 `-fforward-propagate'
4876 Perform a forward propagation pass on RTL. The pass tries to
4877 combine two instructions and checks if the result can be
4878 simplified. If loop unrolling is active, two passes are performed
4879 and the second is scheduled after loop unrolling.
4881 This option is enabled by default at optimization levels `-O2',
4884 `-fomit-frame-pointer'
4885 Don't keep the frame pointer in a register for functions that
4886 don't need one. This avoids the instructions to save, set up and
4887 restore frame pointers; it also makes an extra register available
4888 in many functions. *It also makes debugging impossible on some
4891 On some machines, such as the VAX, this flag has no effect, because
4892 the standard calling sequence automatically handles the frame
4893 pointer and nothing is saved by pretending it doesn't exist. The
4894 machine-description macro `FRAME_POINTER_REQUIRED' controls
4895 whether a target machine supports this flag. *Note Register
4896 Usage: (gccint)Registers.
4898 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4900 `-foptimize-sibling-calls'
4901 Optimize sibling and tail recursive calls.
4903 Enabled at levels `-O2', `-O3', `-Os'.
4906 Don't pay attention to the `inline' keyword. Normally this option
4907 is used to keep the compiler from expanding any functions inline.
4908 Note that if you are not optimizing, no functions can be expanded
4911 `-finline-small-functions'
4912 Integrate functions into their callers when their body is smaller
4913 than expected function call code (so overall size of program gets
4914 smaller). The compiler heuristically decides which functions are
4915 simple enough to be worth integrating in this way.
4917 Enabled at level `-O2'.
4919 `-finline-functions'
4920 Integrate all simple functions into their callers. The compiler
4921 heuristically decides which functions are simple enough to be worth
4922 integrating in this way.
4924 If all calls to a given function are integrated, and the function
4925 is declared `static', then the function is normally not output as
4926 assembler code in its own right.
4928 Enabled at level `-O3'.
4930 `-finline-functions-called-once'
4931 Consider all `static' functions called once for inlining into their
4932 caller even if they are not marked `inline'. If a call to a given
4933 function is integrated, then the function is not output as
4934 assembler code in its own right.
4936 Enabled if `-funit-at-a-time' is enabled.
4939 Inline functions marked by `always_inline' and functions whose
4940 body seems smaller than the function call overhead early before
4941 doing `-fprofile-generate' instrumentation and real inlining pass.
4942 Doing so makes profiling significantly cheaper and usually
4943 inlining faster on programs having large chains of nested wrapper
4949 By default, GCC limits the size of functions that can be inlined.
4950 This flag allows coarse control of this limit. N is the size of
4951 functions that can be inlined in number of pseudo instructions.
4953 Inlining is actually controlled by a number of parameters, which
4954 may be specified individually by using `--param NAME=VALUE'. The
4955 `-finline-limit=N' option sets some of these parameters as follows:
4957 `max-inline-insns-single'
4960 `max-inline-insns-auto'
4963 See below for a documentation of the individual parameters
4964 controlling inlining and for the defaults of these parameters.
4966 _Note:_ there may be no value to `-finline-limit' that results in
4969 _Note:_ pseudo instruction represents, in this particular context,
4970 an abstract measurement of function's size. In no way does it
4971 represent a count of assembly instructions and as such its exact
4972 meaning might change from one release to an another.
4974 `-fkeep-inline-functions'
4975 In C, emit `static' functions that are declared `inline' into the
4976 object file, even if the function has been inlined into all of its
4977 callers. This switch does not affect functions using the `extern
4978 inline' extension in GNU C89. In C++, emit any and all inline
4979 functions into the object file.
4981 `-fkeep-static-consts'
4982 Emit variables declared `static const' when optimization isn't
4983 turned on, even if the variables aren't referenced.
4985 GCC enables this option by default. If you want to force the
4986 compiler to check if the variable was referenced, regardless of
4987 whether or not optimization is turned on, use the
4988 `-fno-keep-static-consts' option.
4991 Attempt to merge identical constants (string constants and
4992 floating point constants) across compilation units.
4994 This option is the default for optimized compilation if the
4995 assembler and linker support it. Use `-fno-merge-constants' to
4996 inhibit this behavior.
4998 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5000 `-fmerge-all-constants'
5001 Attempt to merge identical constants and identical variables.
5003 This option implies `-fmerge-constants'. In addition to
5004 `-fmerge-constants' this considers e.g. even constant initialized
5005 arrays or initialized constant variables with integral or floating
5006 point types. Languages like C or C++ require each non-automatic
5007 variable to have distinct location, so using this option will
5008 result in non-conforming behavior.
5011 Perform swing modulo scheduling immediately before the first
5012 scheduling pass. This pass looks at innermost loops and reorders
5013 their instructions by overlapping different iterations.
5015 `-fmodulo-sched-allow-regmoves'
5016 Perform more aggressive SMS based modulo scheduling with register
5017 moves allowed. By setting this flag certain anti-dependences
5018 edges will be deleted which will trigger the generation of
5019 reg-moves based on the life-range analysis. This option is
5020 effective only with `-fmodulo-sched' enabled.
5022 `-fno-branch-count-reg'
5023 Do not use "decrement and branch" instructions on a count register,
5024 but instead generate a sequence of instructions that decrement a
5025 register, compare it against zero, then branch based upon the
5026 result. This option is only meaningful on architectures that
5027 support such instructions, which include x86, PowerPC, IA-64 and
5030 The default is `-fbranch-count-reg'.
5033 Do not put function addresses in registers; make each instruction
5034 that calls a constant function contain the function's address
5037 This option results in less efficient code, but some strange hacks
5038 that alter the assembler output may be confused by the
5039 optimizations performed when this option is not used.
5041 The default is `-ffunction-cse'
5043 `-fno-zero-initialized-in-bss'
5044 If the target supports a BSS section, GCC by default puts
5045 variables that are initialized to zero into BSS. This can save
5046 space in the resulting code.
5048 This option turns off this behavior because some programs
5049 explicitly rely on variables going to the data section. E.g., so
5050 that the resulting executable can find the beginning of that
5051 section and/or make assumptions based on that.
5053 The default is `-fzero-initialized-in-bss'.
5055 `-fmudflap -fmudflapth -fmudflapir'
5056 For front-ends that support it (C and C++), instrument all risky
5057 pointer/array dereferencing operations, some standard library
5058 string/heap functions, and some other associated constructs with
5059 range/validity tests. Modules so instrumented should be immune to
5060 buffer overflows, invalid heap use, and some other classes of C/C++
5061 programming errors. The instrumentation relies on a separate
5062 runtime library (`libmudflap'), which will be linked into a
5063 program if `-fmudflap' is given at link time. Run-time behavior
5064 of the instrumented program is controlled by the `MUDFLAP_OPTIONS'
5065 environment variable. See `env MUDFLAP_OPTIONS=-help a.out' for
5068 Use `-fmudflapth' instead of `-fmudflap' to compile and to link if
5069 your program is multi-threaded. Use `-fmudflapir', in addition to
5070 `-fmudflap' or `-fmudflapth', if instrumentation should ignore
5071 pointer reads. This produces less instrumentation (and therefore
5072 faster execution) and still provides some protection against
5073 outright memory corrupting writes, but allows erroneously read
5074 data to propagate within a program.
5077 Perform optimizations where we check to see if a jump branches to a
5078 location where another comparison subsumed by the first is found.
5079 If so, the first branch is redirected to either the destination of
5080 the second branch or a point immediately following it, depending
5081 on whether the condition is known to be true or false.
5083 Enabled at levels `-O2', `-O3', `-Os'.
5085 `-fsplit-wide-types'
5086 When using a type that occupies multiple registers, such as `long
5087 long' on a 32-bit system, split the registers apart and allocate
5088 them independently. This normally generates better code for those
5089 types, but may make debugging more difficult.
5091 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5093 `-fcse-follow-jumps'
5094 In common subexpression elimination (CSE), scan through jump
5095 instructions when the target of the jump is not reached by any
5096 other path. For example, when CSE encounters an `if' statement
5097 with an `else' clause, CSE will follow the jump when the condition
5100 Enabled at levels `-O2', `-O3', `-Os'.
5103 This is similar to `-fcse-follow-jumps', but causes CSE to follow
5104 jumps which conditionally skip over blocks. When CSE encounters a
5105 simple `if' statement with no else clause, `-fcse-skip-blocks'
5106 causes CSE to follow the jump around the body of the `if'.
5108 Enabled at levels `-O2', `-O3', `-Os'.
5110 `-frerun-cse-after-loop'
5111 Re-run common subexpression elimination after loop optimizations
5114 Enabled at levels `-O2', `-O3', `-Os'.
5117 Perform a global common subexpression elimination pass. This pass
5118 also performs global constant and copy propagation.
5120 _Note:_ When compiling a program using computed gotos, a GCC
5121 extension, you may get better runtime performance if you disable
5122 the global common subexpression elimination pass by adding
5123 `-fno-gcse' to the command line.
5125 Enabled at levels `-O2', `-O3', `-Os'.
5128 When `-fgcse-lm' is enabled, global common subexpression
5129 elimination will attempt to move loads which are only killed by
5130 stores into themselves. This allows a loop containing a
5131 load/store sequence to be changed to a load outside the loop, and
5132 a copy/store within the loop.
5134 Enabled by default when gcse is enabled.
5137 When `-fgcse-sm' is enabled, a store motion pass is run after
5138 global common subexpression elimination. This pass will attempt
5139 to move stores out of loops. When used in conjunction with
5140 `-fgcse-lm', loops containing a load/store sequence can be changed
5141 to a load before the loop and a store after the loop.
5143 Not enabled at any optimization level.
5146 When `-fgcse-las' is enabled, the global common subexpression
5147 elimination pass eliminates redundant loads that come after stores
5148 to the same memory location (both partial and full redundancies).
5150 Not enabled at any optimization level.
5152 `-fgcse-after-reload'
5153 When `-fgcse-after-reload' is enabled, a redundant load elimination
5154 pass is performed after reload. The purpose of this pass is to
5155 cleanup redundant spilling.
5157 `-funsafe-loop-optimizations'
5158 If given, the loop optimizer will assume that loop indices do not
5159 overflow, and that the loops with nontrivial exit condition are not
5160 infinite. This enables a wider range of loop optimizations even if
5161 the loop optimizer itself cannot prove that these assumptions are
5162 valid. Using `-Wunsafe-loop-optimizations', the compiler will
5163 warn you if it finds this kind of loop.
5166 Perform cross-jumping transformation. This transformation unifies
5167 equivalent code and save code size. The resulting code may or may
5168 not perform better than without cross-jumping.
5170 Enabled at levels `-O2', `-O3', `-Os'.
5173 Combine increments or decrements of addresses with memory accesses.
5174 This pass is always skipped on architectures that do not have
5175 instructions to support this. Enabled by default at `-O' and
5176 higher on architectures that support this.
5179 Perform dead code elimination (DCE) on RTL. Enabled by default at
5183 Perform dead store elimination (DSE) on RTL. Enabled by default
5187 Attempt to transform conditional jumps into branch-less
5188 equivalents. This include use of conditional moves, min, max, set
5189 flags and abs instructions, and some tricks doable by standard
5190 arithmetics. The use of conditional execution on chips where it
5191 is available is controlled by `if-conversion2'.
5193 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5196 Use conditional execution (where available) to transform
5197 conditional jumps into branch-less equivalents.
5199 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5201 `-fdelete-null-pointer-checks'
5202 Use global dataflow analysis to identify and eliminate useless
5203 checks for null pointers. The compiler assumes that dereferencing
5204 a null pointer would have halted the program. If a pointer is
5205 checked after it has already been dereferenced, it cannot be null.
5207 In some environments, this assumption is not true, and programs can
5208 safely dereference null pointers. Use
5209 `-fno-delete-null-pointer-checks' to disable this optimization for
5210 programs which depend on that behavior.
5212 Enabled at levels `-O2', `-O3', `-Os'.
5214 `-fexpensive-optimizations'
5215 Perform a number of minor optimizations that are relatively
5218 Enabled at levels `-O2', `-O3', `-Os'.
5220 `-foptimize-register-move'
5222 Attempt to reassign register numbers in move instructions and as
5223 operands of other simple instructions in order to maximize the
5224 amount of register tying. This is especially helpful on machines
5225 with two-operand instructions.
5227 Note `-fregmove' and `-foptimize-register-move' are the same
5230 Enabled at levels `-O2', `-O3', `-Os'.
5233 If supported for the target machine, attempt to reorder
5234 instructions to exploit instruction slots available after delayed
5235 branch instructions.
5237 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5240 If supported for the target machine, attempt to reorder
5241 instructions to eliminate execution stalls due to required data
5242 being unavailable. This helps machines that have slow floating
5243 point or memory load instructions by allowing other instructions
5244 to be issued until the result of the load or floating point
5245 instruction is required.
5247 Enabled at levels `-O2', `-O3', `-Os'.
5250 Similar to `-fschedule-insns', but requests an additional pass of
5251 instruction scheduling after register allocation has been done.
5252 This is especially useful on machines with a relatively small
5253 number of registers and where memory load instructions take more
5256 Enabled at levels `-O2', `-O3', `-Os'.
5258 `-fno-sched-interblock'
5259 Don't schedule instructions across basic blocks. This is normally
5260 enabled by default when scheduling before register allocation, i.e.
5261 with `-fschedule-insns' or at `-O2' or higher.
5264 Don't allow speculative motion of non-load instructions. This is
5265 normally enabled by default when scheduling before register
5266 allocation, i.e. with `-fschedule-insns' or at `-O2' or higher.
5269 Allow speculative motion of some load instructions. This only
5270 makes sense when scheduling before register allocation, i.e. with
5271 `-fschedule-insns' or at `-O2' or higher.
5273 `-fsched-spec-load-dangerous'
5274 Allow speculative motion of more load instructions. This only
5275 makes sense when scheduling before register allocation, i.e. with
5276 `-fschedule-insns' or at `-O2' or higher.
5278 `-fsched-stalled-insns'
5279 `-fsched-stalled-insns=N'
5280 Define how many insns (if any) can be moved prematurely from the
5281 queue of stalled insns into the ready list, during the second
5282 scheduling pass. `-fno-sched-stalled-insns' means that no insns
5283 will be moved prematurely, `-fsched-stalled-insns=0' means there
5284 is no limit on how many queued insns can be moved prematurely.
5285 `-fsched-stalled-insns' without a value is equivalent to
5286 `-fsched-stalled-insns=1'.
5288 `-fsched-stalled-insns-dep'
5289 `-fsched-stalled-insns-dep=N'
5290 Define how many insn groups (cycles) will be examined for a
5291 dependency on a stalled insn that is candidate for premature
5292 removal from the queue of stalled insns. This has an effect only
5293 during the second scheduling pass, and only if
5294 `-fsched-stalled-insns' is used. `-fno-sched-stalled-insns-dep'
5295 is equivalent to `-fsched-stalled-insns-dep=0'.
5296 `-fsched-stalled-insns-dep' without a value is equivalent to
5297 `-fsched-stalled-insns-dep=1'.
5299 `-fsched2-use-superblocks'
5300 When scheduling after register allocation, do use superblock
5301 scheduling algorithm. Superblock scheduling allows motion across
5302 basic block boundaries resulting on faster schedules. This option
5303 is experimental, as not all machine descriptions used by GCC model
5304 the CPU closely enough to avoid unreliable results from the
5307 This only makes sense when scheduling after register allocation,
5308 i.e. with `-fschedule-insns2' or at `-O2' or higher.
5310 `-fsched2-use-traces'
5311 Use `-fsched2-use-superblocks' algorithm when scheduling after
5312 register allocation and additionally perform code duplication in
5313 order to increase the size of superblocks using tracer pass. See
5314 `-ftracer' for details on trace formation.
5316 This mode should produce faster but significantly longer programs.
5317 Also without `-fbranch-probabilities' the traces constructed may
5318 not match the reality and hurt the performance. This only makes
5319 sense when scheduling after register allocation, i.e. with
5320 `-fschedule-insns2' or at `-O2' or higher.
5323 Eliminate redundant sign extension instructions and move the
5324 non-redundant ones to optimal placement using lazy code motion
5327 `-freschedule-modulo-scheduled-loops'
5328 The modulo scheduling comes before the traditional scheduling, if
5329 a loop was modulo scheduled we may want to prevent the later
5330 scheduling passes from changing its schedule, we use this option
5334 Enable values to be allocated in registers that will be clobbered
5335 by function calls, by emitting extra instructions to save and
5336 restore the registers around such calls. Such allocation is done
5337 only when it seems to result in better code than would otherwise
5340 This option is always enabled by default on certain machines,
5341 usually those which have no call-preserved registers to use
5344 Enabled at levels `-O2', `-O3', `-Os'.
5347 Perform reassociation on trees. This flag is enabled by default
5351 Perform partial redundancy elimination (PRE) on trees. This flag
5352 is enabled by default at `-O2' and `-O3'.
5355 Perform full redundancy elimination (FRE) on trees. The difference
5356 between FRE and PRE is that FRE only considers expressions that
5357 are computed on all paths leading to the redundant computation.
5358 This analysis is faster than PRE, though it exposes fewer
5359 redundancies. This flag is enabled by default at `-O' and higher.
5362 Perform copy propagation on trees. This pass eliminates
5363 unnecessary copy operations. This flag is enabled by default at
5367 Perform structural alias analysis on trees. This flag is enabled
5368 by default at `-O' and higher.
5371 Discover which functions are pure or constant. Enabled by default
5375 Discover which static variables do not escape cannot escape the
5376 compilation unit. Enabled by default at `-O' and higher.
5378 `-fipa-struct-reorg'
5379 Perform structure reorganization optimization, that change C-like
5380 structures layout in order to better utilize spatial locality.
5381 This transformation is affective for programs containing arrays of
5382 structures. Available in two compilation modes: profile-based
5383 (enabled with `-fprofile-generate') or static (which uses built-in
5384 heuristics). Require `-fipa-type-escape' to provide the safety of
5385 this transformation. It works only in whole program mode, so it
5386 requires `-fwhole-program' and `-combine' to be enabled.
5387 Structures considered `cold' by this transformation are not
5388 affected (see `--param struct-reorg-cold-struct-ratio=VALUE').
5390 With this flag, the program debug info reflects a new structure
5394 Perform interprocedural pointer analysis.
5397 Perform interprocedural constant propagation. This optimization
5398 analyzes the program to determine when values passed to functions
5399 are constants and then optimizes accordingly. This optimization
5400 can substantially increase performance if the application has
5401 constants passed to functions, but because this optimization can
5402 create multiple copies of functions, it may significantly increase
5405 `-fipa-matrix-reorg'
5406 Perform matrix flattening and transposing. Matrix flattening
5407 tries to replace a m-dimensional matrix with its equivalent
5408 n-dimensional matrix, where n < m. This reduces the level of
5409 indirection needed for accessing the elements of the matrix. The
5410 second optimization is matrix transposing that attemps to change
5411 the order of the matrix's dimensions in order to improve cache
5412 locality. Both optimizations need fwhole-program flag.
5413 Transposing is enabled only if profiling information is avaliable.
5416 Perform forward store motion on trees. This flag is enabled by
5417 default at `-O' and higher.
5420 Perform sparse conditional constant propagation (CCP) on trees.
5421 This pass only operates on local scalar variables and is enabled
5422 by default at `-O' and higher.
5425 Perform sparse conditional constant propagation (CCP) on trees.
5426 This pass operates on both local scalar variables and memory
5427 stores and loads (global variables, structures, arrays, etc).
5428 This flag is enabled by default at `-O2' and higher.
5431 Perform dead code elimination (DCE) on trees. This flag is
5432 enabled by default at `-O' and higher.
5434 `-ftree-dominator-opts'
5435 Perform a variety of simple scalar cleanups (constant/copy
5436 propagation, redundancy elimination, range propagation and
5437 expression simplification) based on a dominator tree traversal.
5438 This also performs jump threading (to reduce jumps to jumps). This
5439 flag is enabled by default at `-O' and higher.
5442 Perform dead store elimination (DSE) on trees. A dead store is a
5443 store into a memory location which will later be overwritten by
5444 another store without any intervening loads. In this case the
5445 earlier store can be deleted. This flag is enabled by default at
5449 Perform loop header copying on trees. This is beneficial since it
5450 increases effectiveness of code motion optimizations. It also
5451 saves one jump. This flag is enabled by default at `-O' and
5452 higher. It is not enabled for `-Os', since it usually increases
5455 `-ftree-loop-optimize'
5456 Perform loop optimizations on trees. This flag is enabled by
5457 default at `-O' and higher.
5459 `-ftree-loop-linear'
5460 Perform linear loop transformations on tree. This flag can
5461 improve cache performance and allow further loop optimizations to
5465 Compare the results of several data dependence analyzers. This
5466 option is used for debugging the data dependence analyzers.
5469 Perform loop invariant motion on trees. This pass moves only
5470 invariants that would be hard to handle at RTL level (function
5471 calls, operations that expand to nontrivial sequences of insns).
5472 With `-funswitch-loops' it also moves operands of conditions that
5473 are invariant out of the loop, so that we can use just trivial
5474 invariantness analysis in loop unswitching. The pass also includes
5477 `-ftree-loop-ivcanon'
5478 Create a canonical counter for number of iterations in the loop
5479 for that determining number of iterations requires complicated
5480 analysis. Later optimizations then may determine the number
5481 easily. Useful especially in connection with unrolling.
5484 Perform induction variable optimizations (strength reduction,
5485 induction variable merging and induction variable elimination) on
5488 `-ftree-parallelize-loops=n'
5489 Parallelize loops, i.e., split their iteration space to run in n
5490 threads. This is only possible for loops whose iterations are
5491 independent and can be arbitrarily reordered. The optimization is
5492 only profitable on multiprocessor machines, for loops that are
5493 CPU-intensive, rather than constrained e.g. by memory bandwidth.
5494 This option implies `-pthread', and thus is only supported on
5495 targets that have support for `-pthread'.
5498 Perform scalar replacement of aggregates. This pass replaces
5499 structure references with scalars to prevent committing structures
5500 to memory too early. This flag is enabled by default at `-O' and
5504 Perform copy renaming on trees. This pass attempts to rename
5505 compiler temporaries to other variables at copy locations, usually
5506 resulting in variable names which more closely resemble the
5507 original variables. This flag is enabled by default at `-O' and
5511 Perform temporary expression replacement during the SSA->normal
5512 phase. Single use/single def temporaries are replaced at their
5513 use location with their defining expression. This results in
5514 non-GIMPLE code, but gives the expanders much more complex trees
5515 to work on resulting in better RTL generation. This is enabled by
5516 default at `-O' and higher.
5519 Perform loop vectorization on trees. This flag is enabled by
5522 `-ftree-vect-loop-version'
5523 Perform loop versioning when doing loop vectorization on trees.
5524 When a loop appears to be vectorizable except that data alignment
5525 or data dependence cannot be determined at compile time then
5526 vectorized and non-vectorized versions of the loop are generated
5527 along with runtime checks for alignment or dependence to control
5528 which version is executed. This option is enabled by default
5529 except at level `-Os' where it is disabled.
5532 Enable cost model for vectorization.
5535 Perform Value Range Propagation on trees. This is similar to the
5536 constant propagation pass, but instead of values, ranges of values
5537 are propagated. This allows the optimizers to remove unnecessary
5538 range checks like array bound checks and null pointer checks.
5539 This is enabled by default at `-O2' and higher. Null pointer check
5540 elimination is only done if `-fdelete-null-pointer-checks' is
5544 Perform tail duplication to enlarge superblock size. This
5545 transformation simplifies the control flow of the function
5546 allowing other optimizations to do better job.
5549 Unroll loops whose number of iterations can be determined at
5550 compile time or upon entry to the loop. `-funroll-loops' implies
5551 `-frerun-cse-after-loop'. This option makes code larger, and may
5552 or may not make it run faster.
5554 `-funroll-all-loops'
5555 Unroll all loops, even if their number of iterations is uncertain
5556 when the loop is entered. This usually makes programs run more
5557 slowly. `-funroll-all-loops' implies the same options as
5560 `-fsplit-ivs-in-unroller'
5561 Enables expressing of values of induction variables in later
5562 iterations of the unrolled loop using the value in the first
5563 iteration. This breaks long dependency chains, thus improving
5564 efficiency of the scheduling passes.
5566 Combination of `-fweb' and CSE is often sufficient to obtain the
5567 same effect. However in cases the loop body is more complicated
5568 than a single basic block, this is not reliable. It also does not
5569 work at all on some of the architectures due to restrictions in
5572 This optimization is enabled by default.
5574 `-fvariable-expansion-in-unroller'
5575 With this option, the compiler will create multiple copies of some
5576 local variables when unrolling a loop which can result in superior
5579 `-fpredictive-commoning'
5580 Perform predictive commoning optimization, i.e., reusing
5581 computations (especially memory loads and stores) performed in
5582 previous iterations of loops.
5584 This option is enabled at level `-O3'.
5586 `-fprefetch-loop-arrays'
5587 If supported by the target machine, generate instructions to
5588 prefetch memory to improve the performance of loops that access
5591 This option may generate better or worse code; results are highly
5592 dependent on the structure of loops within the source code.
5594 Disabled at level `-Os'.
5598 Disable any machine-specific peephole optimizations. The
5599 difference between `-fno-peephole' and `-fno-peephole2' is in how
5600 they are implemented in the compiler; some targets use one, some
5601 use the other, a few use both.
5603 `-fpeephole' is enabled by default. `-fpeephole2' enabled at
5604 levels `-O2', `-O3', `-Os'.
5606 `-fno-guess-branch-probability'
5607 Do not guess branch probabilities using heuristics.
5609 GCC will use heuristics to guess branch probabilities if they are
5610 not provided by profiling feedback (`-fprofile-arcs'). These
5611 heuristics are based on the control flow graph. If some branch
5612 probabilities are specified by `__builtin_expect', then the
5613 heuristics will be used to guess branch probabilities for the rest
5614 of the control flow graph, taking the `__builtin_expect' info into
5615 account. The interactions between the heuristics and
5616 `__builtin_expect' can be complex, and in some cases, it may be
5617 useful to disable the heuristics so that the effects of
5618 `__builtin_expect' are easier to understand.
5620 The default is `-fguess-branch-probability' at levels `-O', `-O2',
5624 Reorder basic blocks in the compiled function in order to reduce
5625 number of taken branches and improve code locality.
5627 Enabled at levels `-O2', `-O3'.
5629 `-freorder-blocks-and-partition'
5630 In addition to reordering basic blocks in the compiled function,
5631 in order to reduce number of taken branches, partitions hot and
5632 cold basic blocks into separate sections of the assembly and .o
5633 files, to improve paging and cache locality performance.
5635 This optimization is automatically turned off in the presence of
5636 exception handling, for linkonce sections, for functions with a
5637 user-defined section attribute and on any architecture that does
5638 not support named sections.
5640 `-freorder-functions'
5641 Reorder functions in the object file in order to improve code
5642 locality. This is implemented by using special subsections
5643 `.text.hot' for most frequently executed functions and
5644 `.text.unlikely' for unlikely executed functions. Reordering is
5645 done by the linker so object file format must support named
5646 sections and linker must place them in a reasonable way.
5648 Also profile feedback must be available in to make this option
5649 effective. See `-fprofile-arcs' for details.
5651 Enabled at levels `-O2', `-O3', `-Os'.
5654 Allows the compiler to assume the strictest aliasing rules
5655 applicable to the language being compiled. For C (and C++), this
5656 activates optimizations based on the type of expressions. In
5657 particular, an object of one type is assumed never to reside at
5658 the same address as an object of a different type, unless the
5659 types are almost the same. For example, an `unsigned int' can
5660 alias an `int', but not a `void*' or a `double'. A character type
5661 may alias any other type.
5663 Pay special attention to code like this:
5674 The practice of reading from a different union member than the one
5675 most recently written to (called "type-punning") is common. Even
5676 with `-fstrict-aliasing', type-punning is allowed, provided the
5677 memory is accessed through the union type. So, the code above
5678 will work as expected. However, this code might not:
5687 Enabled at levels `-O2', `-O3', `-Os'.
5690 Allow the compiler to assume strict signed overflow rules,
5691 depending on the language being compiled. For C (and C++) this
5692 means that overflow when doing arithmetic with signed numbers is
5693 undefined, which means that the compiler may assume that it will
5694 not happen. This permits various optimizations. For example, the
5695 compiler will assume that an expression like `i + 10 > i' will
5696 always be true for signed `i'. This assumption is only valid if
5697 signed overflow is undefined, as the expression is false if `i +
5698 10' overflows when using twos complement arithmetic. When this
5699 option is in effect any attempt to determine whether an operation
5700 on signed numbers will overflow must be written carefully to not
5701 actually involve overflow.
5703 This option also allows the compiler to assume strict pointer
5704 semantics: given a pointer to an object, if adding an offset to
5705 that pointer does not produce a pointer to the same object, the
5706 addition is undefined. This permits the compiler to conclude that
5707 `p + u > p' is always true for a pointer `p' and unsigned integer
5708 `u'. This assumption is only valid because pointer wraparound is
5709 undefined, as the expression is false if `p + u' overflows using
5710 twos complement arithmetic.
5712 See also the `-fwrapv' option. Using `-fwrapv' means that integer
5713 signed overflow is fully defined: it wraps. When `-fwrapv' is
5714 used, there is no difference between `-fstrict-overflow' and
5715 `-fno-strict-overflow' for integers. With `-fwrapv' certain types
5716 of overflow are permitted. For example, if the compiler gets an
5717 overflow when doing arithmetic on constants, the overflowed value
5718 can still be used with `-fwrapv', but not otherwise.
5720 The `-fstrict-overflow' option is enabled at levels `-O2', `-O3',
5724 `-falign-functions=N'
5725 Align the start of functions to the next power-of-two greater than
5726 N, skipping up to N bytes. For instance, `-falign-functions=32'
5727 aligns functions to the next 32-byte boundary, but
5728 `-falign-functions=24' would align to the next 32-byte boundary
5729 only if this can be done by skipping 23 bytes or less.
5731 `-fno-align-functions' and `-falign-functions=1' are equivalent
5732 and mean that functions will not be aligned.
5734 Some assemblers only support this flag when N is a power of two;
5735 in that case, it is rounded up.
5737 If N is not specified or is zero, use a machine-dependent default.
5739 Enabled at levels `-O2', `-O3'.
5743 Align all branch targets to a power-of-two boundary, skipping up to
5744 N bytes like `-falign-functions'. This option can easily make
5745 code slower, because it must insert dummy operations for when the
5746 branch target is reached in the usual flow of the code.
5748 `-fno-align-labels' and `-falign-labels=1' are equivalent and mean
5749 that labels will not be aligned.
5751 If `-falign-loops' or `-falign-jumps' are applicable and are
5752 greater than this value, then their values are used instead.
5754 If N is not specified or is zero, use a machine-dependent default
5755 which is very likely to be `1', meaning no alignment.
5757 Enabled at levels `-O2', `-O3'.
5761 Align loops to a power-of-two boundary, skipping up to N bytes
5762 like `-falign-functions'. The hope is that the loop will be
5763 executed many times, which will make up for any execution of the
5766 `-fno-align-loops' and `-falign-loops=1' are equivalent and mean
5767 that loops will not be aligned.
5769 If N is not specified or is zero, use a machine-dependent default.
5771 Enabled at levels `-O2', `-O3'.
5775 Align branch targets to a power-of-two boundary, for branch targets
5776 where the targets can only be reached by jumping, skipping up to N
5777 bytes like `-falign-functions'. In this case, no dummy operations
5780 `-fno-align-jumps' and `-falign-jumps=1' are equivalent and mean
5781 that loops will not be aligned.
5783 If N is not specified or is zero, use a machine-dependent default.
5785 Enabled at levels `-O2', `-O3'.
5788 Parse the whole compilation unit before starting to produce code.
5789 This allows some extra optimizations to take place but consumes
5790 more memory (in general). There are some compatibility issues
5791 with _unit-at-a-time_ mode:
5792 * enabling _unit-at-a-time_ mode may change the order in which
5793 functions, variables, and top-level `asm' statements are
5794 emitted, and will likely break code relying on some particular
5795 ordering. The majority of such top-level `asm' statements,
5796 though, can be replaced by `section' attributes. The
5797 `fno-toplevel-reorder' option may be used to keep the ordering
5798 used in the input file, at the cost of some optimizations.
5800 * _unit-at-a-time_ mode removes unreferenced static variables
5801 and functions. This may result in undefined references when
5802 an `asm' statement refers directly to variables or functions
5803 that are otherwise unused. In that case either the
5804 variable/function shall be listed as an operand of the `asm'
5805 statement operand or, in the case of top-level `asm'
5806 statements the attribute `used' shall be used on the
5809 * Static functions now can use non-standard passing conventions
5810 that may break `asm' statements calling functions directly.
5811 Again, attribute `used' will prevent this behavior.
5813 As a temporary workaround, `-fno-unit-at-a-time' can be used, but
5814 this scheme may not be supported by future releases of GCC.
5816 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5818 `-fno-toplevel-reorder'
5819 Do not reorder top-level functions, variables, and `asm'
5820 statements. Output them in the same order that they appear in the
5821 input file. When this option is used, unreferenced static
5822 variables will not be removed. This option is intended to support
5823 existing code which relies on a particular ordering. For new
5824 code, it is better to use attributes.
5827 Constructs webs as commonly used for register allocation purposes
5828 and assign each web individual pseudo register. This allows the
5829 register allocation pass to operate on pseudos directly, but also
5830 strengthens several other optimization passes, such as CSE, loop
5831 optimizer and trivial dead code remover. It can, however, make
5832 debugging impossible, since variables will no longer stay in a
5835 Enabled by default with `-funroll-loops'.
5838 Assume that the current compilation unit represents whole program
5839 being compiled. All public functions and variables with the
5840 exception of `main' and those merged by attribute
5841 `externally_visible' become static functions and in a affect gets
5842 more aggressively optimized by interprocedural optimizers. While
5843 this option is equivalent to proper use of `static' keyword for
5844 programs consisting of single file, in combination with option
5845 `--combine' this flag can be used to compile most of smaller scale
5846 C programs since the functions and variables become local for the
5847 whole combined compilation unit, not for the single source file
5850 This option is not supported for Fortran programs.
5853 After register allocation and post-register allocation instruction
5854 splitting, we perform a copy-propagation pass to try to reduce
5855 scheduling dependencies and occasionally eliminate the copy.
5857 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5859 `-fprofile-generate'
5860 Enable options usually used for instrumenting application to
5861 produce profile useful for later recompilation with profile
5862 feedback based optimization. You must use `-fprofile-generate'
5863 both when compiling and when linking your program.
5865 The following options are enabled: `-fprofile-arcs',
5866 `-fprofile-values', `-fvpt'.
5869 Enable profile feedback directed optimizations, and optimizations
5870 generally profitable only with profile feedback available.
5872 The following options are enabled: `-fbranch-probabilities',
5873 `-fvpt', `-funroll-loops', `-fpeel-loops', `-ftracer'
5875 By default, GCC emits an error message if the feedback profiles do
5876 not match the source code. This error can be turned into a
5877 warning by using `-Wcoverage-mismatch'. Note this may result in
5878 poorly optimized code.
5880 The following options control compiler behavior regarding floating
5881 point arithmetic. These options trade off between speed and
5882 correctness. All must be specifically enabled.
5885 Do not store floating point variables in registers, and inhibit
5886 other options that might change whether a floating point value is
5887 taken from a register or memory.
5889 This option prevents undesirable excess precision on machines such
5890 as the 68000 where the floating registers (of the 68881) keep more
5891 precision than a `double' is supposed to have. Similarly for the
5892 x86 architecture. For most programs, the excess precision does
5893 only good, but a few programs rely on the precise definition of
5894 IEEE floating point. Use `-ffloat-store' for such programs, after
5895 modifying them to store all pertinent intermediate computations
5899 Sets `-fno-math-errno', `-funsafe-math-optimizations',
5900 `-ffinite-math-only', `-fno-rounding-math', `-fno-signaling-nans'
5901 and `-fcx-limited-range'.
5903 This option causes the preprocessor macro `__FAST_MATH__' to be
5906 This option is not turned on by any `-O' option since it can
5907 result in incorrect output for programs which depend on an exact
5908 implementation of IEEE or ISO rules/specifications for math
5909 functions. It may, however, yield faster code for programs that do
5910 not require the guarantees of these specifications.
5913 Do not set ERRNO after calling math functions that are executed
5914 with a single instruction, e.g., sqrt. A program that relies on
5915 IEEE exceptions for math error handling may want to use this flag
5916 for speed while maintaining IEEE arithmetic compatibility.
5918 This option is not turned on by any `-O' option since it can
5919 result in incorrect output for programs which depend on an exact
5920 implementation of IEEE or ISO rules/specifications for math
5921 functions. It may, however, yield faster code for programs that do
5922 not require the guarantees of these specifications.
5924 The default is `-fmath-errno'.
5926 On Darwin systems, the math library never sets `errno'. There is
5927 therefore no reason for the compiler to consider the possibility
5928 that it might, and `-fno-math-errno' is the default.
5930 `-funsafe-math-optimizations'
5931 Allow optimizations for floating-point arithmetic that (a) assume
5932 that arguments and results are valid and (b) may violate IEEE or
5933 ANSI standards. When used at link-time, it may include libraries
5934 or startup files that change the default FPU control word or other
5935 similar optimizations.
5937 This option is not turned on by any `-O' option since it can
5938 result in incorrect output for programs which depend on an exact
5939 implementation of IEEE or ISO rules/specifications for math
5940 functions. It may, however, yield faster code for programs that do
5941 not require the guarantees of these specifications. Enables
5942 `-fno-signed-zeros', `-fno-trapping-math', `-fassociative-math'
5943 and `-freciprocal-math'.
5945 The default is `-fno-unsafe-math-optimizations'.
5947 `-fassociative-math'
5948 Allow re-association of operands in series of floating-point
5949 operations. This violates the ISO C and C++ language standard by
5950 possibly changing computation result. NOTE: re-ordering may
5951 change the sign of zero as well as ignore NaNs and inhibit or
5952 create underflow or overflow (and thus cannot be used on a code
5953 which relies on rounding behavior like `(x + 2**52) - 2**52)'.
5954 May also reorder floating-point comparisons and thus may not be
5955 used when ordered comparisons are required. This option requires
5956 that both `-fno-signed-zeros' and `-fno-trapping-math' be in
5957 effect. Moreover, it doesn't make much sense with
5960 The default is `-fno-associative-math'.
5963 Allow the reciprocal of a value to be used instead of dividing by
5964 the value if this enables optimizations. For example `x / y' can
5965 be replaced with `x * (1/y)' which is useful if `(1/y)' is subject
5966 to common subexpression elimination. Note that this loses
5967 precision and increases the number of flops operating on the value.
5969 The default is `-fno-reciprocal-math'.
5971 `-ffinite-math-only'
5972 Allow optimizations for floating-point arithmetic that assume that
5973 arguments and results are not NaNs or +-Infs.
5975 This option is not turned on by any `-O' option since it can
5976 result in incorrect output for programs which depend on an exact
5977 implementation of IEEE or ISO rules/specifications for math
5978 functions. It may, however, yield faster code for programs that do
5979 not require the guarantees of these specifications.
5981 The default is `-fno-finite-math-only'.
5984 Allow optimizations for floating point arithmetic that ignore the
5985 signedness of zero. IEEE arithmetic specifies the behavior of
5986 distinct +0.0 and -0.0 values, which then prohibits simplification
5987 of expressions such as x+0.0 or 0.0*x (even with
5988 `-ffinite-math-only'). This option implies that the sign of a
5989 zero result isn't significant.
5991 The default is `-fsigned-zeros'.
5993 `-fno-trapping-math'
5994 Compile code assuming that floating-point operations cannot
5995 generate user-visible traps. These traps include division by
5996 zero, overflow, underflow, inexact result and invalid operation.
5997 This option requires that `-fno-signaling-nans' be in effect.
5998 Setting this option may allow faster code if one relies on
5999 "non-stop" IEEE arithmetic, for example.
6001 This option should never be turned on by any `-O' option since it
6002 can result in incorrect output for programs which depend on an
6003 exact implementation of IEEE or ISO rules/specifications for math
6006 The default is `-ftrapping-math'.
6009 Disable transformations and optimizations that assume default
6010 floating point rounding behavior. This is round-to-zero for all
6011 floating point to integer conversions, and round-to-nearest for
6012 all other arithmetic truncations. This option should be specified
6013 for programs that change the FP rounding mode dynamically, or that
6014 may be executed with a non-default rounding mode. This option
6015 disables constant folding of floating point expressions at
6016 compile-time (which may be affected by rounding mode) and
6017 arithmetic transformations that are unsafe in the presence of
6018 sign-dependent rounding modes.
6020 The default is `-fno-rounding-math'.
6022 This option is experimental and does not currently guarantee to
6023 disable all GCC optimizations that are affected by rounding mode.
6024 Future versions of GCC may provide finer control of this setting
6025 using C99's `FENV_ACCESS' pragma. This command line option will
6026 be used to specify the default state for `FENV_ACCESS'.
6028 `-frtl-abstract-sequences'
6029 It is a size optimization method. This option is to find identical
6030 sequences of code, which can be turned into pseudo-procedures and
6031 then replace all occurrences with calls to the newly created
6032 subroutine. It is kind of an opposite of `-finline-functions'.
6033 This optimization runs at RTL level.
6036 Compile code assuming that IEEE signaling NaNs may generate
6037 user-visible traps during floating-point operations. Setting this
6038 option disables optimizations that may change the number of
6039 exceptions visible with signaling NaNs. This option implies
6042 This option causes the preprocessor macro `__SUPPORT_SNAN__' to be
6045 The default is `-fno-signaling-nans'.
6047 This option is experimental and does not currently guarantee to
6048 disable all GCC optimizations that affect signaling NaN behavior.
6050 `-fsingle-precision-constant'
6051 Treat floating point constant as single precision constant instead
6052 of implicitly converting it to double precision constant.
6054 `-fcx-limited-range'
6055 When enabled, this option states that a range reduction step is not
6056 needed when performing complex division. The default is
6057 `-fno-cx-limited-range', but is enabled by `-ffast-math'.
6059 This option controls the default setting of the ISO C99
6060 `CX_LIMITED_RANGE' pragma. Nevertheless, the option applies to
6064 The following options control optimizations that may improve
6065 performance, but are not enabled by any `-O' options. This section
6066 includes experimental options that may produce broken code.
6068 `-fbranch-probabilities'
6069 After running a program compiled with `-fprofile-arcs' (*note
6070 Options for Debugging Your Program or `gcc': Debugging Options.),
6071 you can compile it a second time using `-fbranch-probabilities',
6072 to improve optimizations based on the number of times each branch
6073 was taken. When the program compiled with `-fprofile-arcs' exits
6074 it saves arc execution counts to a file called `SOURCENAME.gcda'
6075 for each source file. The information in this data file is very
6076 dependent on the structure of the generated code, so you must use
6077 the same source code and the same optimization options for both
6080 With `-fbranch-probabilities', GCC puts a `REG_BR_PROB' note on
6081 each `JUMP_INSN' and `CALL_INSN'. These can be used to improve
6082 optimization. Currently, they are only used in one place: in
6083 `reorg.c', instead of guessing which path a branch is mostly to
6084 take, the `REG_BR_PROB' values are used to exactly determine which
6085 path is taken more often.
6088 If combined with `-fprofile-arcs', it adds code so that some data
6089 about values of expressions in the program is gathered.
6091 With `-fbranch-probabilities', it reads back the data gathered
6092 from profiling values of expressions and adds `REG_VALUE_PROFILE'
6093 notes to instructions for their later usage in optimizations.
6095 Enabled with `-fprofile-generate' and `-fprofile-use'.
6098 If combined with `-fprofile-arcs', it instructs the compiler to add
6099 a code to gather information about values of expressions.
6101 With `-fbranch-probabilities', it reads back the data gathered and
6102 actually performs the optimizations based on them. Currently the
6103 optimizations include specialization of division operation using
6104 the knowledge about the value of the denominator.
6106 `-frename-registers'
6107 Attempt to avoid false dependencies in scheduled code by making use
6108 of registers left over after register allocation. This
6109 optimization will most benefit processors with lots of registers.
6110 Depending on the debug information format adopted by the target,
6111 however, it can make debugging impossible, since variables will no
6112 longer stay in a "home register".
6114 Enabled by default with `-funroll-loops'.
6117 Perform tail duplication to enlarge superblock size. This
6118 transformation simplifies the control flow of the function
6119 allowing other optimizations to do better job.
6121 Enabled with `-fprofile-use'.
6124 Unroll loops whose number of iterations can be determined at
6125 compile time or upon entry to the loop. `-funroll-loops' implies
6126 `-frerun-cse-after-loop', `-fweb' and `-frename-registers'. It
6127 also turns on complete loop peeling (i.e. complete removal of
6128 loops with small constant number of iterations). This option
6129 makes code larger, and may or may not make it run faster.
6131 Enabled with `-fprofile-use'.
6133 `-funroll-all-loops'
6134 Unroll all loops, even if their number of iterations is uncertain
6135 when the loop is entered. This usually makes programs run more
6136 slowly. `-funroll-all-loops' implies the same options as
6140 Peels the loops for that there is enough information that they do
6141 not roll much (from profile feedback). It also turns on complete
6142 loop peeling (i.e. complete removal of loops with small constant
6143 number of iterations).
6145 Enabled with `-fprofile-use'.
6147 `-fmove-loop-invariants'
6148 Enables the loop invariant motion pass in the RTL loop optimizer.
6149 Enabled at level `-O1'
6152 Move branches with loop invariant conditions out of the loop, with
6153 duplicates of the loop on both branches (modified according to
6154 result of the condition).
6156 `-ffunction-sections'
6158 Place each function or data item into its own section in the output
6159 file if the target supports arbitrary sections. The name of the
6160 function or the name of the data item determines the section's name
6163 Use these options on systems where the linker can perform
6164 optimizations to improve locality of reference in the instruction
6165 space. Most systems using the ELF object format and SPARC
6166 processors running Solaris 2 have linkers with such optimizations.
6167 AIX may have these optimizations in the future.
6169 Only use these options when there are significant benefits from
6170 doing so. When you specify these options, the assembler and
6171 linker will create larger object and executable files and will
6172 also be slower. You will not be able to use `gprof' on all
6173 systems if you specify this option and you may have problems with
6174 debugging if you specify both this option and `-g'.
6176 `-fbranch-target-load-optimize'
6177 Perform branch target register load optimization before prologue /
6178 epilogue threading. The use of target registers can typically be
6179 exposed only during reload, thus hoisting loads out of loops and
6180 doing inter-block scheduling needs a separate optimization pass.
6182 `-fbranch-target-load-optimize2'
6183 Perform branch target register load optimization after prologue /
6186 `-fbtr-bb-exclusive'
6187 When performing branch target register load optimization, don't
6188 reuse branch target registers in within any basic block.
6191 Emit extra code to check for buffer overflows, such as stack
6192 smashing attacks. This is done by adding a guard variable to
6193 functions with vulnerable objects. This includes functions that
6194 call alloca, and functions with buffers larger than 8 bytes. The
6195 guards are initialized when a function is entered and then checked
6196 when the function exits. If a guard check fails, an error message
6197 is printed and the program exits.
6199 `-fstack-protector-all'
6200 Like `-fstack-protector' except that all functions are protected.
6203 Try to reduce the number of symbolic address calculations by using
6204 shared "anchor" symbols to address nearby objects. This
6205 transformation can help to reduce the number of GOT entries and
6206 GOT accesses on some targets.
6208 For example, the implementation of the following function `foo':
6211 int foo (void) { return a + b + c; }
6213 would usually calculate the addresses of all three variables, but
6214 if you compile it with `-fsection-anchors', it will access the
6215 variables from a common anchor point instead. The effect is
6216 similar to the following pseudocode (which isn't valid C):
6220 register int *xr = &x;
6221 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6224 Not all targets support this option.
6226 `--param NAME=VALUE'
6227 In some places, GCC uses various constants to control the amount of
6228 optimization that is done. For example, GCC will not inline
6229 functions that contain more that a certain number of instructions.
6230 You can control some of these constants on the command-line using
6231 the `--param' option.
6233 The names of specific parameters, and the meaning of the values,
6234 are tied to the internals of the compiler, and are subject to
6235 change without notice in future releases.
6237 In each case, the VALUE is an integer. The allowable choices for
6238 NAME are given in the following table:
6240 `salias-max-implicit-fields'
6241 The maximum number of fields in a variable without direct
6242 structure accesses for which structure aliasing will consider
6243 trying to track each field. The default is 5
6245 `salias-max-array-elements'
6246 The maximum number of elements an array can have and its
6247 elements still be tracked individually by structure aliasing.
6250 `sra-max-structure-size'
6251 The maximum structure size, in bytes, at which the scalar
6252 replacement of aggregates (SRA) optimization will perform
6253 block copies. The default value, 0, implies that GCC will
6254 select the most appropriate size itself.
6256 `sra-field-structure-ratio'
6257 The threshold ratio (as a percentage) between instantiated
6258 fields and the complete structure size. We say that if the
6259 ratio of the number of bytes in instantiated fields to the
6260 number of bytes in the complete structure exceeds this
6261 parameter, then block copies are not used. The default is 75.
6263 `struct-reorg-cold-struct-ratio'
6264 The threshold ratio (as a percentage) between a structure
6265 frequency and the frequency of the hottest structure in the
6266 program. This parameter is used by struct-reorg optimization
6267 enabled by `-fipa-struct-reorg'. We say that if the ratio of
6268 a structure frequency, calculated by profiling, to the
6269 hottest structure frequency in the program is less than this
6270 parameter, then structure reorganization is not applied to
6271 this structure. The default is 10.
6273 `max-crossjump-edges'
6274 The maximum number of incoming edges to consider for
6275 crossjumping. The algorithm used by `-fcrossjumping' is
6276 O(N^2) in the number of edges incoming to each block.
6277 Increasing values mean more aggressive optimization, making
6278 the compile time increase with probably small improvement in
6281 `min-crossjump-insns'
6282 The minimum number of instructions which must be matched at
6283 the end of two blocks before crossjumping will be performed
6284 on them. This value is ignored in the case where all
6285 instructions in the block being crossjumped from are matched.
6286 The default value is 5.
6288 `max-grow-copy-bb-insns'
6289 The maximum code size expansion factor when copying basic
6290 blocks instead of jumping. The expansion is relative to a
6291 jump instruction. The default value is 8.
6293 `max-goto-duplication-insns'
6294 The maximum number of instructions to duplicate to a block
6295 that jumps to a computed goto. To avoid O(N^2) behavior in a
6296 number of passes, GCC factors computed gotos early in the
6297 compilation process, and unfactors them as late as possible.
6298 Only computed jumps at the end of a basic blocks with no more
6299 than max-goto-duplication-insns are unfactored. The default
6302 `max-delay-slot-insn-search'
6303 The maximum number of instructions to consider when looking
6304 for an instruction to fill a delay slot. If more than this
6305 arbitrary number of instructions is searched, the time
6306 savings from filling the delay slot will be minimal so stop
6307 searching. Increasing values mean more aggressive
6308 optimization, making the compile time increase with probably
6309 small improvement in executable run time.
6311 `max-delay-slot-live-search'
6312 When trying to fill delay slots, the maximum number of
6313 instructions to consider when searching for a block with
6314 valid live register information. Increasing this arbitrarily
6315 chosen value means more aggressive optimization, increasing
6316 the compile time. This parameter should be removed when the
6317 delay slot code is rewritten to maintain the control-flow
6321 The approximate maximum amount of memory that will be
6322 allocated in order to perform the global common subexpression
6323 elimination optimization. If more memory than specified is
6324 required, the optimization will not be done.
6327 The maximum number of passes of GCSE to run. The default is
6330 `max-pending-list-length'
6331 The maximum number of pending dependencies scheduling will
6332 allow before flushing the current state and starting over.
6333 Large functions with few branches or calls can create
6334 excessively large lists which needlessly consume memory and
6337 `max-inline-insns-single'
6338 Several parameters control the tree inliner used in gcc.
6339 This number sets the maximum number of instructions (counted
6340 in GCC's internal representation) in a single function that
6341 the tree inliner will consider for inlining. This only
6342 affects functions declared inline and methods implemented in
6343 a class declaration (C++). The default value is 450.
6345 `max-inline-insns-auto'
6346 When you use `-finline-functions' (included in `-O3'), a lot
6347 of functions that would otherwise not be considered for
6348 inlining by the compiler will be investigated. To those
6349 functions, a different (more restrictive) limit compared to
6350 functions declared inline can be applied. The default value
6353 `large-function-insns'
6354 The limit specifying really large functions. For functions
6355 larger than this limit after inlining inlining is constrained
6356 by `--param large-function-growth'. This parameter is useful
6357 primarily to avoid extreme compilation time caused by
6358 non-linear algorithms used by the backend. This parameter is
6359 ignored when `-funit-at-a-time' is not used. The default
6362 `large-function-growth'
6363 Specifies maximal growth of large function caused by inlining
6364 in percents. This parameter is ignored when
6365 `-funit-at-a-time' is not used. The default value is 100
6366 which limits large function growth to 2.0 times the original
6370 The limit specifying large translation unit. Growth caused
6371 by inlining of units larger than this limit is limited by
6372 `--param inline-unit-growth'. For small units this might be
6373 too tight (consider unit consisting of function A that is
6374 inline and B that just calls A three time. If B is small
6375 relative to A, the growth of unit is 300\% and yet such
6376 inlining is very sane. For very large units consisting of
6377 small inlineable functions however the overall unit growth
6378 limit is needed to avoid exponential explosion of code size.
6379 Thus for smaller units, the size is increased to `--param
6380 large-unit-insns' before applying `--param
6381 inline-unit-growth'. The default is 10000
6383 `inline-unit-growth'
6384 Specifies maximal overall growth of the compilation unit
6385 caused by inlining. This parameter is ignored when
6386 `-funit-at-a-time' is not used. The default value is 30
6387 which limits unit growth to 1.3 times the original size.
6390 The limit specifying large stack frames. While inlining the
6391 algorithm is trying to not grow past this limit too much.
6392 Default value is 256 bytes.
6394 `large-stack-frame-growth'
6395 Specifies maximal growth of large stack frames caused by
6396 inlining in percents. The default value is 1000 which limits
6397 large stack frame growth to 11 times the original size.
6399 `max-inline-insns-recursive'
6400 `max-inline-insns-recursive-auto'
6401 Specifies maximum number of instructions out-of-line copy of
6402 self recursive inline function can grow into by performing
6405 For functions declared inline `--param
6406 max-inline-insns-recursive' is taken into account. For
6407 function not declared inline, recursive inlining happens only
6408 when `-finline-functions' (included in `-O3') is enabled and
6409 `--param max-inline-insns-recursive-auto' is used. The
6410 default value is 450.
6412 `max-inline-recursive-depth'
6413 `max-inline-recursive-depth-auto'
6414 Specifies maximum recursion depth used by the recursive
6417 For functions declared inline `--param
6418 max-inline-recursive-depth' is taken into account. For
6419 function not declared inline, recursive inlining happens only
6420 when `-finline-functions' (included in `-O3') is enabled and
6421 `--param max-inline-recursive-depth-auto' is used. The
6424 `min-inline-recursive-probability'
6425 Recursive inlining is profitable only for function having
6426 deep recursion in average and can hurt for function having
6427 little recursion depth by increasing the prologue size or
6428 complexity of function body to other optimizers.
6430 When profile feedback is available (see `-fprofile-generate')
6431 the actual recursion depth can be guessed from probability
6432 that function will recurse via given call expression. This
6433 parameter limits inlining only to call expression whose
6434 probability exceeds given threshold (in percents). The
6435 default value is 10.
6438 Specify cost of call instruction relative to simple
6439 arithmetics operations (having cost of 1). Increasing this
6440 cost disqualifies inlining of non-leaf functions and at the
6441 same time increases size of leaf function that is believed to
6442 reduce function size by being inlined. In effect it
6443 increases amount of inlining for code having large
6444 abstraction penalty (many functions that just pass the
6445 arguments to other functions) and decrease inlining for code
6446 with low abstraction penalty. The default value is 12.
6448 `min-vect-loop-bound'
6449 The minimum number of iterations under which a loop will not
6450 get vectorized when `-ftree-vectorize' is used. The number
6451 of iterations after vectorization needs to be greater than
6452 the value specified by this option to allow vectorization.
6453 The default value is 0.
6455 `max-unrolled-insns'
6456 The maximum number of instructions that a loop should have if
6457 that loop is unrolled, and if the loop is unrolled, it
6458 determines how many times the loop code is unrolled.
6460 `max-average-unrolled-insns'
6461 The maximum number of instructions biased by probabilities of
6462 their execution that a loop should have if that loop is
6463 unrolled, and if the loop is unrolled, it determines how many
6464 times the loop code is unrolled.
6467 The maximum number of unrollings of a single loop.
6470 The maximum number of instructions that a loop should have if
6471 that loop is peeled, and if the loop is peeled, it determines
6472 how many times the loop code is peeled.
6475 The maximum number of peelings of a single loop.
6477 `max-completely-peeled-insns'
6478 The maximum number of insns of a completely peeled loop.
6480 `max-completely-peel-times'
6481 The maximum number of iterations of a loop to be suitable for
6484 `max-unswitch-insns'
6485 The maximum number of insns of an unswitched loop.
6487 `max-unswitch-level'
6488 The maximum number of branches unswitched in a single loop.
6491 The minimum cost of an expensive expression in the loop
6494 `iv-consider-all-candidates-bound'
6495 Bound on number of candidates for induction variables below
6496 that all candidates are considered for each use in induction
6497 variable optimizations. Only the most relevant candidates
6498 are considered if there are more candidates, to avoid
6499 quadratic time complexity.
6501 `iv-max-considered-uses'
6502 The induction variable optimizations give up on loops that
6503 contain more induction variable uses.
6505 `iv-always-prune-cand-set-bound'
6506 If number of candidates in the set is smaller than this value,
6507 we always try to remove unnecessary ivs from the set during
6508 its optimization when a new iv is added to the set.
6510 `scev-max-expr-size'
6511 Bound on size of expressions used in the scalar evolutions
6512 analyzer. Large expressions slow the analyzer.
6515 The maximum number of variables in an Omega constraint system.
6516 The default value is 128.
6519 The maximum number of inequalities in an Omega constraint
6520 system. The default value is 256.
6523 The maximum number of equalities in an Omega constraint
6524 system. The default value is 128.
6526 `omega-max-wild-cards'
6527 The maximum number of wildcard variables that the Omega
6528 solver will be able to insert. The default value is 18.
6530 `omega-hash-table-size'
6531 The size of the hash table in the Omega solver. The default
6535 The maximal number of keys used by the Omega solver. The
6536 default value is 500.
6538 `omega-eliminate-redundant-constraints'
6539 When set to 1, use expensive methods to eliminate all
6540 redundant constraints. The default value is 0.
6542 `vect-max-version-for-alignment-checks'
6543 The maximum number of runtime checks that can be performed
6544 when doing loop versioning for alignment in the vectorizer.
6545 See option ftree-vect-loop-version for more information.
6547 `vect-max-version-for-alias-checks'
6548 The maximum number of runtime checks that can be performed
6549 when doing loop versioning for alias in the vectorizer. See
6550 option ftree-vect-loop-version for more information.
6552 `max-iterations-to-track'
6553 The maximum number of iterations of a loop the brute force
6554 algorithm for analysis of # of iterations of the loop tries
6557 `hot-bb-count-fraction'
6558 Select fraction of the maximal count of repetitions of basic
6559 block in program given basic block needs to have to be
6562 `hot-bb-frequency-fraction'
6563 Select fraction of the maximal frequency of executions of
6564 basic block in function given basic block needs to have to be
6567 `max-predicted-iterations'
6568 The maximum number of loop iterations we predict statically.
6569 This is useful in cases where function contain single loop
6570 with known bound and other loop with unknown. We predict the
6571 known number of iterations correctly, while the unknown
6572 number of iterations average to roughly 10. This means that
6573 the loop without bounds would appear artificially cold
6574 relative to the other one.
6577 Select fraction of the maximal frequency of executions of
6578 basic block in function given basic block will get aligned.
6580 `align-loop-iterations'
6581 A loop expected to iterate at lest the selected number of
6582 iterations will get aligned.
6584 `tracer-dynamic-coverage'
6585 `tracer-dynamic-coverage-feedback'
6586 This value is used to limit superblock formation once the
6587 given percentage of executed instructions is covered. This
6588 limits unnecessary code size expansion.
6590 The `tracer-dynamic-coverage-feedback' is used only when
6591 profile feedback is available. The real profiles (as opposed
6592 to statically estimated ones) are much less balanced allowing
6593 the threshold to be larger value.
6595 `tracer-max-code-growth'
6596 Stop tail duplication once code growth has reached given
6597 percentage. This is rather hokey argument, as most of the
6598 duplicates will be eliminated later in cross jumping, so it
6599 may be set to much higher values than is the desired code
6602 `tracer-min-branch-ratio'
6603 Stop reverse growth when the reverse probability of best edge
6604 is less than this threshold (in percent).
6606 `tracer-min-branch-ratio'
6607 `tracer-min-branch-ratio-feedback'
6608 Stop forward growth if the best edge do have probability
6609 lower than this threshold.
6611 Similarly to `tracer-dynamic-coverage' two values are
6612 present, one for compilation for profile feedback and one for
6613 compilation without. The value for compilation with profile
6614 feedback needs to be more conservative (higher) in order to
6615 make tracer effective.
6617 `max-cse-path-length'
6618 Maximum number of basic blocks on path that cse considers.
6622 The maximum instructions CSE process before flushing. The
6626 Maximum number of virtual operands per function allowed to
6627 represent aliases before triggering the alias partitioning
6628 heuristic. Alias partitioning reduces compile times and
6629 memory consumption needed for aliasing at the expense of
6630 precision loss in alias information. The default value for
6631 this parameter is 100 for -O1, 500 for -O2 and 1000 for -O3.
6633 Notice that if a function contains more memory statements
6634 than the value of this parameter, it is not really possible
6635 to achieve this reduction. In this case, the compiler will
6636 use the number of memory statements as the value for
6640 Average number of virtual operands per statement allowed to
6641 represent aliases before triggering the alias partitioning
6642 heuristic. This works in conjunction with
6643 `max-aliased-vops'. If a function contains more than
6644 `max-aliased-vops' virtual operators, then memory symbols
6645 will be grouped into memory partitions until either the total
6646 number of virtual operators is below `max-aliased-vops' or
6647 the average number of virtual operators per memory statement
6648 is below `avg-aliased-vops'. The default value for this
6649 parameter is 1 for -O1 and -O2, and 3 for -O3.
6652 GCC uses a garbage collector to manage its own memory
6653 allocation. This parameter specifies the minimum percentage
6654 by which the garbage collector's heap should be allowed to
6655 expand between collections. Tuning this may improve
6656 compilation speed; it has no effect on code generation.
6658 The default is 30% + 70% * (RAM/1GB) with an upper bound of
6659 100% when RAM >= 1GB. If `getrlimit' is available, the
6660 notion of "RAM" is the smallest of actual RAM and
6661 `RLIMIT_DATA' or `RLIMIT_AS'. If GCC is not able to
6662 calculate RAM on a particular platform, the lower bound of
6663 30% is used. Setting this parameter and `ggc-min-heapsize'
6664 to zero causes a full collection to occur at every
6665 opportunity. This is extremely slow, but can be useful for
6669 Minimum size of the garbage collector's heap before it begins
6670 bothering to collect garbage. The first collection occurs
6671 after the heap expands by `ggc-min-expand'% beyond
6672 `ggc-min-heapsize'. Again, tuning this may improve
6673 compilation speed, and has no effect on code generation.
6675 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit
6676 which tries to ensure that RLIMIT_DATA or RLIMIT_AS are not
6677 exceeded, but with a lower bound of 4096 (four megabytes) and
6678 an upper bound of 131072 (128 megabytes). If GCC is not able
6679 to calculate RAM on a particular platform, the lower bound is
6680 used. Setting this parameter very large effectively disables
6681 garbage collection. Setting this parameter and
6682 `ggc-min-expand' to zero causes a full collection to occur at
6685 `max-reload-search-insns'
6686 The maximum number of instruction reload should look backward
6687 for equivalent register. Increasing values mean more
6688 aggressive optimization, making the compile time increase
6689 with probably slightly better performance. The default value
6692 `max-cselib-memory-locations'
6693 The maximum number of memory locations cselib should take
6694 into account. Increasing values mean more aggressive
6695 optimization, making the compile time increase with probably
6696 slightly better performance. The default value is 500.
6698 `max-flow-memory-locations'
6699 Similar as `max-cselib-memory-locations' but for dataflow
6700 liveness. The default value is 100.
6702 `reorder-blocks-duplicate'
6703 `reorder-blocks-duplicate-feedback'
6704 Used by basic block reordering pass to decide whether to use
6705 unconditional branch or duplicate the code on its
6706 destination. Code is duplicated when its estimated size is
6707 smaller than this value multiplied by the estimated size of
6708 unconditional jump in the hot spots of the program.
6710 The `reorder-block-duplicate-feedback' is used only when
6711 profile feedback is available and may be set to higher values
6712 than `reorder-block-duplicate' since information about the
6713 hot spots is more accurate.
6715 `max-sched-ready-insns'
6716 The maximum number of instructions ready to be issued the
6717 scheduler should consider at any given time during the first
6718 scheduling pass. Increasing values mean more thorough
6719 searches, making the compilation time increase with probably
6720 little benefit. The default value is 100.
6722 `max-sched-region-blocks'
6723 The maximum number of blocks in a region to be considered for
6724 interblock scheduling. The default value is 10.
6726 `max-sched-region-insns'
6727 The maximum number of insns in a region to be considered for
6728 interblock scheduling. The default value is 100.
6731 The minimum probability (in percents) of reaching a source
6732 block for interblock speculative scheduling. The default
6735 `max-sched-extend-regions-iters'
6736 The maximum number of iterations through CFG to extend
6737 regions. 0 - disable region extension, N - do at most N
6738 iterations. The default value is 0.
6740 `max-sched-insn-conflict-delay'
6741 The maximum conflict delay for an insn to be considered for
6742 speculative motion. The default value is 3.
6744 `sched-spec-prob-cutoff'
6745 The minimal probability of speculation success (in percents),
6746 so that speculative insn will be scheduled. The default
6749 `max-last-value-rtl'
6750 The maximum size measured as number of RTLs that can be
6751 recorded in an expression in combiner for a pseudo register
6752 as last known value of that register. The default is 10000.
6754 `integer-share-limit'
6755 Small integer constants can use a shared data structure,
6756 reducing the compiler's memory usage and increasing its
6757 speed. This sets the maximum value of a shared integer
6758 constant. The default value is 256.
6760 `min-virtual-mappings'
6761 Specifies the minimum number of virtual mappings in the
6762 incremental SSA updater that should be registered to trigger
6763 the virtual mappings heuristic defined by
6764 virtual-mappings-ratio. The default value is 100.
6766 `virtual-mappings-ratio'
6767 If the number of virtual mappings is virtual-mappings-ratio
6768 bigger than the number of virtual symbols to be updated, then
6769 the incremental SSA updater switches to a full update for
6770 those symbols. The default ratio is 3.
6773 The minimum size of buffers (i.e. arrays) that will receive
6774 stack smashing protection when `-fstack-protection' is used.
6776 `max-jump-thread-duplication-stmts'
6777 Maximum number of statements allowed in a block that needs to
6778 be duplicated when threading jumps.
6780 `max-fields-for-field-sensitive'
6781 Maximum number of fields in a structure we will treat in a
6782 field sensitive manner during pointer analysis.
6785 Estimate on average number of instructions that are executed
6786 before prefetch finishes. The distance we prefetch ahead is
6787 proportional to this constant. Increasing this number may
6788 also lead to less streams being prefetched (see
6789 `simultaneous-prefetches').
6791 `simultaneous-prefetches'
6792 Maximum number of prefetches that can run at the same time.
6794 `l1-cache-line-size'
6795 The size of cache line in L1 cache, in bytes.
6798 The size of L1 cache, in kilobytes.
6801 The size of L2 cache, in kilobytes.
6803 `use-canonical-types'
6804 Whether the compiler should use the "canonical" type system.
6805 By default, this should always be 1, which uses a more
6806 efficient internal mechanism for comparing types in C++ and
6807 Objective-C++. However, if bugs in the canonical type system
6808 are causing compilation failures, set this value to 0 to
6809 disable canonical types.
6811 `max-partial-antic-length'
6812 Maximum length of the partial antic set computed during the
6813 tree partial redundancy elimination optimization
6814 (`-ftree-pre') when optimizing at `-O3' and above. For some
6815 sorts of source code the enhanced partial redundancy
6816 elimination optimization can run away, consuming all of the
6817 memory available on the host machine. This parameter sets a
6818 limit on the length of the sets that are computed, which
6819 prevents the runaway behaviour. Setting a value of 0 for
6820 this paramter will allow an unlimited set length.
6822 `sccvn-max-scc-size'
6823 Maximum size of a strongly connected component (SCC) during
6824 SCCVN processing. If this limit is hit, SCCVN processing for
6825 the whole function will not be done and optimizations
6826 depending on it will be disabled. The default maximum SCC
6831 File: gcc.info, Node: Preprocessor Options, Next: Assembler Options, Prev: Optimize Options, Up: Invoking GCC
6833 3.11 Options Controlling the Preprocessor
6834 =========================================
6836 These options control the C preprocessor, which is run on each C source
6837 file before actual compilation.
6839 If you use the `-E' option, nothing is done except preprocessing.
6840 Some of these options make sense only together with `-E' because they
6841 cause the preprocessor output to be unsuitable for actual compilation.
6843 You can use `-Wp,OPTION' to bypass the compiler driver and pass
6844 OPTION directly through to the preprocessor. If OPTION contains
6845 commas, it is split into multiple options at the commas. However,
6846 many options are modified, translated or interpreted by the
6847 compiler driver before being passed to the preprocessor, and `-Wp'
6848 forcibly bypasses this phase. The preprocessor's direct interface
6849 is undocumented and subject to change, so whenever possible you
6850 should avoid using `-Wp' and let the driver handle the options
6853 `-Xpreprocessor OPTION'
6854 Pass OPTION as an option to the preprocessor. You can use this to
6855 supply system-specific preprocessor options which GCC does not
6856 know how to recognize.
6858 If you want to pass an option that takes an argument, you must use
6859 `-Xpreprocessor' twice, once for the option and once for the
6863 Predefine NAME as a macro, with definition `1'.
6865 `-D NAME=DEFINITION'
6866 The contents of DEFINITION are tokenized and processed as if they
6867 appeared during translation phase three in a `#define' directive.
6868 In particular, the definition will be truncated by embedded
6871 If you are invoking the preprocessor from a shell or shell-like
6872 program you may need to use the shell's quoting syntax to protect
6873 characters such as spaces that have a meaning in the shell syntax.
6875 If you wish to define a function-like macro on the command line,
6876 write its argument list with surrounding parentheses before the
6877 equals sign (if any). Parentheses are meaningful to most shells,
6878 so you will need to quote the option. With `sh' and `csh',
6879 `-D'NAME(ARGS...)=DEFINITION'' works.
6881 `-D' and `-U' options are processed in the order they are given on
6882 the command line. All `-imacros FILE' and `-include FILE' options
6883 are processed after all `-D' and `-U' options.
6886 Cancel any previous definition of NAME, either built in or
6887 provided with a `-D' option.
6890 Do not predefine any system-specific or GCC-specific macros. The
6891 standard predefined macros remain defined.
6894 Add the directory DIR to the list of directories to be searched
6895 for header files. Directories named by `-I' are searched before
6896 the standard system include directories. If the directory DIR is
6897 a standard system include directory, the option is ignored to
6898 ensure that the default search order for system directories and
6899 the special treatment of system headers are not defeated . If DIR
6900 begins with `=', then the `=' will be replaced by the sysroot
6901 prefix; see `--sysroot' and `-isysroot'.
6904 Write output to FILE. This is the same as specifying FILE as the
6905 second non-option argument to `cpp'. `gcc' has a different
6906 interpretation of a second non-option argument, so you must use
6907 `-o' to specify the output file.
6910 Turns on all optional warnings which are desirable for normal code.
6911 At present this is `-Wcomment', `-Wtrigraphs', `-Wmultichar' and a
6912 warning about integer promotion causing a change of sign in `#if'
6913 expressions. Note that many of the preprocessor's warnings are on
6914 by default and have no options to control them.
6918 Warn whenever a comment-start sequence `/*' appears in a `/*'
6919 comment, or whenever a backslash-newline appears in a `//' comment.
6920 (Both forms have the same effect.)
6923 Most trigraphs in comments cannot affect the meaning of the
6924 program. However, a trigraph that would form an escaped newline
6925 (`??/' at the end of a line) can, by changing where the comment
6926 begins or ends. Therefore, only trigraphs that would form escaped
6927 newlines produce warnings inside a comment.
6929 This option is implied by `-Wall'. If `-Wall' is not given, this
6930 option is still enabled unless trigraphs are enabled. To get
6931 trigraph conversion without warnings, but get the other `-Wall'
6932 warnings, use `-trigraphs -Wall -Wno-trigraphs'.
6935 Warn about certain constructs that behave differently in
6936 traditional and ISO C. Also warn about ISO C constructs that have
6937 no traditional C equivalent, and problematic constructs which
6941 Warn the first time `#import' is used.
6944 Warn whenever an identifier which is not a macro is encountered in
6945 an `#if' directive, outside of `defined'. Such identifiers are
6949 Warn about macros defined in the main file that are unused. A
6950 macro is "used" if it is expanded or tested for existence at least
6951 once. The preprocessor will also warn if the macro has not been
6952 used at the time it is redefined or undefined.
6954 Built-in macros, macros defined on the command line, and macros
6955 defined in include files are not warned about.
6957 _Note:_ If a macro is actually used, but only used in skipped
6958 conditional blocks, then CPP will report it as unused. To avoid
6959 the warning in such a case, you might improve the scope of the
6960 macro's definition by, for example, moving it into the first
6961 skipped block. Alternatively, you could provide a dummy use with
6964 #if defined the_macro_causing_the_warning
6968 Warn whenever an `#else' or an `#endif' are followed by text.
6969 This usually happens in code of the form
6977 The second and third `FOO' should be in comments, but often are not
6978 in older programs. This warning is on by default.
6981 Make all warnings into hard errors. Source code which triggers
6982 warnings will be rejected.
6985 Issue warnings for code in system headers. These are normally
6986 unhelpful in finding bugs in your own code, therefore suppressed.
6987 If you are responsible for the system library, you may want to see
6991 Suppress all warnings, including those which GNU CPP issues by
6995 Issue all the mandatory diagnostics listed in the C standard.
6996 Some of them are left out by default, since they trigger
6997 frequently on harmless code.
7000 Issue all the mandatory diagnostics, and make all mandatory
7001 diagnostics into errors. This includes mandatory diagnostics that
7002 GCC issues without `-pedantic' but treats as warnings.
7005 Instead of outputting the result of preprocessing, output a rule
7006 suitable for `make' describing the dependencies of the main source
7007 file. The preprocessor outputs one `make' rule containing the
7008 object file name for that source file, a colon, and the names of
7009 all the included files, including those coming from `-include' or
7010 `-imacros' command line options.
7012 Unless specified explicitly (with `-MT' or `-MQ'), the object file
7013 name consists of the name of the source file with any suffix
7014 replaced with object file suffix and with any leading directory
7015 parts removed. If there are many included files then the rule is
7016 split into several lines using `\'-newline. The rule has no
7019 This option does not suppress the preprocessor's debug output,
7020 such as `-dM'. To avoid mixing such debug output with the
7021 dependency rules you should explicitly specify the dependency
7022 output file with `-MF', or use an environment variable like
7023 `DEPENDENCIES_OUTPUT' (*note Environment Variables::). Debug
7024 output will still be sent to the regular output stream as normal.
7026 Passing `-M' to the driver implies `-E', and suppresses warnings
7027 with an implicit `-w'.
7030 Like `-M' but do not mention header files that are found in system
7031 header directories, nor header files that are included, directly
7032 or indirectly, from such a header.
7034 This implies that the choice of angle brackets or double quotes in
7035 an `#include' directive does not in itself determine whether that
7036 header will appear in `-MM' dependency output. This is a slight
7037 change in semantics from GCC versions 3.0 and earlier.
7040 When used with `-M' or `-MM', specifies a file to write the
7041 dependencies to. If no `-MF' switch is given the preprocessor
7042 sends the rules to the same place it would have sent preprocessed
7045 When used with the driver options `-MD' or `-MMD', `-MF' overrides
7046 the default dependency output file.
7049 In conjunction with an option such as `-M' requesting dependency
7050 generation, `-MG' assumes missing header files are generated files
7051 and adds them to the dependency list without raising an error.
7052 The dependency filename is taken directly from the `#include'
7053 directive without prepending any path. `-MG' also suppresses
7054 preprocessed output, as a missing header file renders this useless.
7056 This feature is used in automatic updating of makefiles.
7059 This option instructs CPP to add a phony target for each dependency
7060 other than the main file, causing each to depend on nothing. These
7061 dummy rules work around errors `make' gives if you remove header
7062 files without updating the `Makefile' to match.
7064 This is typical output:
7066 test.o: test.c test.h
7071 Change the target of the rule emitted by dependency generation. By
7072 default CPP takes the name of the main input file, deletes any
7073 directory components and any file suffix such as `.c', and appends
7074 the platform's usual object suffix. The result is the target.
7076 An `-MT' option will set the target to be exactly the string you
7077 specify. If you want multiple targets, you can specify them as a
7078 single argument to `-MT', or use multiple `-MT' options.
7080 For example, `-MT '$(objpfx)foo.o'' might give
7082 $(objpfx)foo.o: foo.c
7085 Same as `-MT', but it quotes any characters which are special to
7086 Make. `-MQ '$(objpfx)foo.o'' gives
7088 $$(objpfx)foo.o: foo.c
7090 The default target is automatically quoted, as if it were given
7094 `-MD' is equivalent to `-M -MF FILE', except that `-E' is not
7095 implied. The driver determines FILE based on whether an `-o'
7096 option is given. If it is, the driver uses its argument but with
7097 a suffix of `.d', otherwise it takes the name of the input file,
7098 removes any directory components and suffix, and applies a `.d'
7101 If `-MD' is used in conjunction with `-E', any `-o' switch is
7102 understood to specify the dependency output file (*note -MF:
7103 dashMF.), but if used without `-E', each `-o' is understood to
7104 specify a target object file.
7106 Since `-E' is not implied, `-MD' can be used to generate a
7107 dependency output file as a side-effect of the compilation process.
7110 Like `-MD' except mention only user header files, not system
7114 When using precompiled headers (*note Precompiled Headers::), this
7115 flag will cause the dependency-output flags to also list the files
7116 from the precompiled header's dependencies. If not specified only
7117 the precompiled header would be listed and not the files that were
7118 used to create it because those files are not consulted when a
7119 precompiled header is used.
7122 This option allows use of a precompiled header (*note Precompiled
7123 Headers::) together with `-E'. It inserts a special `#pragma',
7124 `#pragma GCC pch_preprocess "<filename>"' in the output to mark
7125 the place where the precompiled header was found, and its
7126 filename. When `-fpreprocessed' is in use, GCC recognizes this
7127 `#pragma' and loads the PCH.
7129 This option is off by default, because the resulting preprocessed
7130 output is only really suitable as input to GCC. It is switched on
7133 You should not write this `#pragma' in your own code, but it is
7134 safe to edit the filename if the PCH file is available in a
7135 different location. The filename may be absolute or it may be
7136 relative to GCC's current directory.
7141 `-x assembler-with-cpp'
7142 Specify the source language: C, C++, Objective-C, or assembly.
7143 This has nothing to do with standards conformance or extensions;
7144 it merely selects which base syntax to expect. If you give none
7145 of these options, cpp will deduce the language from the extension
7146 of the source file: `.c', `.cc', `.m', or `.S'. Some other common
7147 extensions for C++ and assembly are also recognized. If cpp does
7148 not recognize the extension, it will treat the file as C; this is
7149 the most generic mode.
7151 _Note:_ Previous versions of cpp accepted a `-lang' option which
7152 selected both the language and the standards conformance level.
7153 This option has been removed, because it conflicts with the `-l'
7158 Specify the standard to which the code should conform. Currently
7159 CPP knows about C and C++ standards; others may be added in the
7162 STANDARD may be one of:
7165 The ISO C standard from 1990. `c89' is the customary
7166 shorthand for this version of the standard.
7168 The `-ansi' option is equivalent to `-std=c89'.
7171 The 1990 C standard, as amended in 1994.
7177 The revised ISO C standard, published in December 1999.
7178 Before publication, this was known as C9X.
7181 The 1990 C standard plus GNU extensions. This is the default.
7185 The 1999 C standard plus GNU extensions.
7188 The 1998 ISO C++ standard plus amendments.
7191 The same as `-std=c++98' plus GNU extensions. This is the
7192 default for C++ code.
7195 Split the include path. Any directories specified with `-I'
7196 options before `-I-' are searched only for headers requested with
7197 `#include "FILE"'; they are not searched for `#include <FILE>'.
7198 If additional directories are specified with `-I' options after
7199 the `-I-', those directories are searched for all `#include'
7202 In addition, `-I-' inhibits the use of the directory of the current
7203 file directory as the first search directory for `#include "FILE"'.
7204 This option has been deprecated.
7207 Do not search the standard system directories for header files.
7208 Only the directories you have specified with `-I' options (and the
7209 directory of the current file, if appropriate) are searched.
7212 Do not search for header files in the C++-specific standard
7213 directories, but do still search the other standard directories.
7214 (This option is used when building the C++ library.)
7217 Process FILE as if `#include "file"' appeared as the first line of
7218 the primary source file. However, the first directory searched
7219 for FILE is the preprocessor's working directory _instead of_ the
7220 directory containing the main source file. If not found there, it
7221 is searched for in the remainder of the `#include "..."' search
7224 If multiple `-include' options are given, the files are included
7225 in the order they appear on the command line.
7228 Exactly like `-include', except that any output produced by
7229 scanning FILE is thrown away. Macros it defines remain defined.
7230 This allows you to acquire all the macros from a header without
7231 also processing its declarations.
7233 All files specified by `-imacros' are processed before all files
7234 specified by `-include'.
7237 Search DIR for header files, but do it _after_ all directories
7238 specified with `-I' and the standard system directories have been
7239 exhausted. DIR is treated as a system include directory. If DIR
7240 begins with `=', then the `=' will be replaced by the sysroot
7241 prefix; see `--sysroot' and `-isysroot'.
7244 Specify PREFIX as the prefix for subsequent `-iwithprefix'
7245 options. If the prefix represents a directory, you should include
7249 `-iwithprefixbefore DIR'
7250 Append DIR to the prefix specified previously with `-iprefix', and
7251 add the resulting directory to the include search path.
7252 `-iwithprefixbefore' puts it in the same place `-I' would;
7253 `-iwithprefix' puts it where `-idirafter' would.
7256 This option is like the `--sysroot' option, but applies only to
7257 header files. See the `--sysroot' option for more information.
7260 Use DIR as a subdirectory of the directory containing
7261 target-specific C++ headers.
7264 Search DIR for header files, after all directories specified by
7265 `-I' but before the standard system directories. Mark it as a
7266 system directory, so that it gets the same special treatment as is
7267 applied to the standard system directories. If DIR begins with
7268 `=', then the `=' will be replaced by the sysroot prefix; see
7269 `--sysroot' and `-isysroot'.
7272 Search DIR only for header files requested with `#include "FILE"';
7273 they are not searched for `#include <FILE>', before all
7274 directories specified by `-I' and before the standard system
7275 directories. If DIR begins with `=', then the `=' will be replaced
7276 by the sysroot prefix; see `--sysroot' and `-isysroot'.
7279 When preprocessing, handle directives, but do not expand macros.
7281 The option's behavior depends on the `-E' and `-fpreprocessed'
7284 With `-E', preprocessing is limited to the handling of directives
7285 such as `#define', `#ifdef', and `#error'. Other preprocessor
7286 operations, such as macro expansion and trigraph conversion are
7287 not performed. In addition, the `-dD' option is implicitly
7290 With `-fpreprocessed', predefinition of command line and most
7291 builtin macros is disabled. Macros such as `__LINE__', which are
7292 contextually dependent, are handled normally. This enables
7293 compilation of files previously preprocessed with `-E
7296 With both `-E' and `-fpreprocessed', the rules for
7297 `-fpreprocessed' take precedence. This enables full preprocessing
7298 of files previously preprocessed with `-E -fdirectives-only'.
7300 `-fdollars-in-identifiers'
7301 Accept `$' in identifiers.
7303 `-fextended-identifiers'
7304 Accept universal character names in identifiers. This option is
7305 experimental; in a future version of GCC, it will be enabled by
7306 default for C99 and C++.
7309 Indicate to the preprocessor that the input file has already been
7310 preprocessed. This suppresses things like macro expansion,
7311 trigraph conversion, escaped newline splicing, and processing of
7312 most directives. The preprocessor still recognizes and removes
7313 comments, so that you can pass a file preprocessed with `-C' to
7314 the compiler without problems. In this mode the integrated
7315 preprocessor is little more than a tokenizer for the front ends.
7317 `-fpreprocessed' is implicit if the input file has one of the
7318 extensions `.i', `.ii' or `.mi'. These are the extensions that
7319 GCC uses for preprocessed files created by `-save-temps'.
7322 Set the distance between tab stops. This helps the preprocessor
7323 report correct column numbers in warnings or errors, even if tabs
7324 appear on the line. If the value is less than 1 or greater than
7325 100, the option is ignored. The default is 8.
7327 `-fexec-charset=CHARSET'
7328 Set the execution character set, used for string and character
7329 constants. The default is UTF-8. CHARSET can be any encoding
7330 supported by the system's `iconv' library routine.
7332 `-fwide-exec-charset=CHARSET'
7333 Set the wide execution character set, used for wide string and
7334 character constants. The default is UTF-32 or UTF-16, whichever
7335 corresponds to the width of `wchar_t'. As with `-fexec-charset',
7336 CHARSET can be any encoding supported by the system's `iconv'
7337 library routine; however, you will have problems with encodings
7338 that do not fit exactly in `wchar_t'.
7340 `-finput-charset=CHARSET'
7341 Set the input character set, used for translation from the
7342 character set of the input file to the source character set used
7343 by GCC. If the locale does not specify, or GCC cannot get this
7344 information from the locale, the default is UTF-8. This can be
7345 overridden by either the locale or this command line option.
7346 Currently the command line option takes precedence if there's a
7347 conflict. CHARSET can be any encoding supported by the system's
7348 `iconv' library routine.
7350 `-fworking-directory'
7351 Enable generation of linemarkers in the preprocessor output that
7352 will let the compiler know the current working directory at the
7353 time of preprocessing. When this option is enabled, the
7354 preprocessor will emit, after the initial linemarker, a second
7355 linemarker with the current working directory followed by two
7356 slashes. GCC will use this directory, when it's present in the
7357 preprocessed input, as the directory emitted as the current
7358 working directory in some debugging information formats. This
7359 option is implicitly enabled if debugging information is enabled,
7360 but this can be inhibited with the negated form
7361 `-fno-working-directory'. If the `-P' flag is present in the
7362 command line, this option has no effect, since no `#line'
7363 directives are emitted whatsoever.
7366 Do not print column numbers in diagnostics. This may be necessary
7367 if diagnostics are being scanned by a program that does not
7368 understand the column numbers, such as `dejagnu'.
7370 `-A PREDICATE=ANSWER'
7371 Make an assertion with the predicate PREDICATE and answer ANSWER.
7372 This form is preferred to the older form `-A PREDICATE(ANSWER)',
7373 which is still supported, because it does not use shell special
7376 `-A -PREDICATE=ANSWER'
7377 Cancel an assertion with the predicate PREDICATE and answer ANSWER.
7380 CHARS is a sequence of one or more of the following characters,
7381 and must not be preceded by a space. Other characters are
7382 interpreted by the compiler proper, or reserved for future
7383 versions of GCC, and so are silently ignored. If you specify
7384 characters whose behavior conflicts, the result is undefined.
7387 Instead of the normal output, generate a list of `#define'
7388 directives for all the macros defined during the execution of
7389 the preprocessor, including predefined macros. This gives
7390 you a way of finding out what is predefined in your version
7391 of the preprocessor. Assuming you have no file `foo.h', the
7394 touch foo.h; cpp -dM foo.h
7396 will show all the predefined macros.
7398 If you use `-dM' without the `-E' option, `-dM' is
7399 interpreted as a synonym for `-fdump-rtl-mach'. *Note
7400 Debugging Options: (gcc)Debugging Options.
7403 Like `M' except in two respects: it does _not_ include the
7404 predefined macros, and it outputs _both_ the `#define'
7405 directives and the result of preprocessing. Both kinds of
7406 output go to the standard output file.
7409 Like `D', but emit only the macro names, not their expansions.
7412 Output `#include' directives in addition to the result of
7416 Inhibit generation of linemarkers in the output from the
7417 preprocessor. This might be useful when running the preprocessor
7418 on something that is not C code, and will be sent to a program
7419 which might be confused by the linemarkers.
7422 Do not discard comments. All comments are passed through to the
7423 output file, except for comments in processed directives, which
7424 are deleted along with the directive.
7426 You should be prepared for side effects when using `-C'; it causes
7427 the preprocessor to treat comments as tokens in their own right.
7428 For example, comments appearing at the start of what would be a
7429 directive line have the effect of turning that line into an
7430 ordinary source line, since the first token on the line is no
7434 Do not discard comments, including during macro expansion. This is
7435 like `-C', except that comments contained within macros are also
7436 passed through to the output file where the macro is expanded.
7438 In addition to the side-effects of the `-C' option, the `-CC'
7439 option causes all C++-style comments inside a macro to be
7440 converted to C-style comments. This is to prevent later use of
7441 that macro from inadvertently commenting out the remainder of the
7444 The `-CC' option is generally used to support lint comments.
7447 Try to imitate the behavior of old-fashioned C preprocessors, as
7448 opposed to ISO C preprocessors.
7451 Process trigraph sequences. These are three-character sequences,
7452 all starting with `??', that are defined by ISO C to stand for
7453 single characters. For example, `??/' stands for `\', so `'??/n''
7454 is a character constant for a newline. By default, GCC ignores
7455 trigraphs, but in standard-conforming modes it converts them. See
7456 the `-std' and `-ansi' options.
7458 The nine trigraphs and their replacements are
7460 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??-
7461 Replacement: [ ] { } # \ ^ | ~
7464 Enable special code to work around file systems which only permit
7465 very short file names, such as MS-DOS.
7469 Print text describing all the command line options instead of
7470 preprocessing anything.
7473 Verbose mode. Print out GNU CPP's version number at the beginning
7474 of execution, and report the final form of the include path.
7477 Print the name of each header file used, in addition to other
7478 normal activities. Each name is indented to show how deep in the
7479 `#include' stack it is. Precompiled header files are also
7480 printed, even if they are found to be invalid; an invalid
7481 precompiled header file is printed with `...x' and a valid one
7486 Print out GNU CPP's version number. With one dash, proceed to
7487 preprocess as normal. With two dashes, exit immediately.
7490 File: gcc.info, Node: Assembler Options, Next: Link Options, Prev: Preprocessor Options, Up: Invoking GCC
7492 3.12 Passing Options to the Assembler
7493 =====================================
7495 You can pass options to the assembler.
7498 Pass OPTION as an option to the assembler. If OPTION contains
7499 commas, it is split into multiple options at the commas.
7501 `-Xassembler OPTION'
7502 Pass OPTION as an option to the assembler. You can use this to
7503 supply system-specific assembler options which GCC does not know
7506 If you want to pass an option that takes an argument, you must use
7507 `-Xassembler' twice, once for the option and once for the argument.
7511 File: gcc.info, Node: Link Options, Next: Directory Options, Prev: Assembler Options, Up: Invoking GCC
7513 3.13 Options for Linking
7514 ========================
7516 These options come into play when the compiler links object files into
7517 an executable output file. They are meaningless if the compiler is not
7521 A file name that does not end in a special recognized suffix is
7522 considered to name an object file or library. (Object files are
7523 distinguished from libraries by the linker according to the file
7524 contents.) If linking is done, these object files are used as
7525 input to the linker.
7530 If any of these options is used, then the linker is not run, and
7531 object file names should not be used as arguments. *Note Overall
7536 Search the library named LIBRARY when linking. (The second
7537 alternative with the library as a separate argument is only for
7538 POSIX compliance and is not recommended.)
7540 It makes a difference where in the command you write this option;
7541 the linker searches and processes libraries and object files in
7542 the order they are specified. Thus, `foo.o -lz bar.o' searches
7543 library `z' after file `foo.o' but before `bar.o'. If `bar.o'
7544 refers to functions in `z', those functions may not be loaded.
7546 The linker searches a standard list of directories for the library,
7547 which is actually a file named `libLIBRARY.a'. The linker then
7548 uses this file as if it had been specified precisely by name.
7550 The directories searched include several standard system
7551 directories plus any that you specify with `-L'.
7553 Normally the files found this way are library files--archive files
7554 whose members are object files. The linker handles an archive
7555 file by scanning through it for members which define symbols that
7556 have so far been referenced but not defined. But if the file that
7557 is found is an ordinary object file, it is linked in the usual
7558 fashion. The only difference between using an `-l' option and
7559 specifying a file name is that `-l' surrounds LIBRARY with `lib'
7560 and `.a' and searches several directories.
7563 You need this special case of the `-l' option in order to link an
7564 Objective-C or Objective-C++ program.
7567 Do not use the standard system startup files when linking. The
7568 standard system libraries are used normally, unless `-nostdlib' or
7569 `-nodefaultlibs' is used.
7572 Do not use the standard system libraries when linking. Only the
7573 libraries you specify will be passed to the linker. The standard
7574 startup files are used normally, unless `-nostartfiles' is used.
7575 The compiler may generate calls to `memcmp', `memset', `memcpy'
7576 and `memmove'. These entries are usually resolved by entries in
7577 libc. These entry points should be supplied through some other
7578 mechanism when this option is specified.
7581 Do not use the standard system startup files or libraries when
7582 linking. No startup files and only the libraries you specify will
7583 be passed to the linker. The compiler may generate calls to
7584 `memcmp', `memset', `memcpy' and `memmove'. These entries are
7585 usually resolved by entries in libc. These entry points should be
7586 supplied through some other mechanism when this option is
7589 One of the standard libraries bypassed by `-nostdlib' and
7590 `-nodefaultlibs' is `libgcc.a', a library of internal subroutines
7591 that GCC uses to overcome shortcomings of particular machines, or
7592 special needs for some languages. (*Note Interfacing to GCC
7593 Output: (gccint)Interface, for more discussion of `libgcc.a'.) In
7594 most cases, you need `libgcc.a' even when you want to avoid other
7595 standard libraries. In other words, when you specify `-nostdlib'
7596 or `-nodefaultlibs' you should usually specify `-lgcc' as well.
7597 This ensures that you have no unresolved references to internal GCC
7598 library subroutines. (For example, `__main', used to ensure C++
7599 constructors will be called; *note `collect2': (gccint)Collect2.)
7602 Produce a position independent executable on targets which support
7603 it. For predictable results, you must also specify the same set
7604 of options that were used to generate code (`-fpie', `-fPIE', or
7605 model suboptions) when you specify this option.
7608 Pass the flag `-export-dynamic' to the ELF linker, on targets that
7609 support it. This instructs the linker to add all symbols, not only
7610 used ones, to the dynamic symbol table. This option is needed for
7611 some uses of `dlopen' or to allow obtaining backtraces from within
7615 Remove all symbol table and relocation information from the
7619 On systems that support dynamic linking, this prevents linking
7620 with the shared libraries. On other systems, this option has no
7624 Produce a shared object which can then be linked with other
7625 objects to form an executable. Not all systems support this
7626 option. For predictable results, you must also specify the same
7627 set of options that were used to generate code (`-fpic', `-fPIC',
7628 or model suboptions) when you specify this option.(1)
7632 On systems that provide `libgcc' as a shared library, these options
7633 force the use of either the shared or static version respectively.
7634 If no shared version of `libgcc' was built when the compiler was
7635 configured, these options have no effect.
7637 There are several situations in which an application should use the
7638 shared `libgcc' instead of the static version. The most common of
7639 these is when the application wishes to throw and catch exceptions
7640 across different shared libraries. In that case, each of the
7641 libraries as well as the application itself should use the shared
7644 Therefore, the G++ and GCJ drivers automatically add
7645 `-shared-libgcc' whenever you build a shared library or a main
7646 executable, because C++ and Java programs typically use
7647 exceptions, so this is the right thing to do.
7649 If, instead, you use the GCC driver to create shared libraries,
7650 you may find that they will not always be linked with the shared
7651 `libgcc'. If GCC finds, at its configuration time, that you have
7652 a non-GNU linker or a GNU linker that does not support option
7653 `--eh-frame-hdr', it will link the shared version of `libgcc' into
7654 shared libraries by default. Otherwise, it will take advantage of
7655 the linker and optimize away the linking with the shared version
7656 of `libgcc', linking with the static version of libgcc by default.
7657 This allows exceptions to propagate through such shared
7658 libraries, without incurring relocation costs at library load time.
7660 However, if a library or main executable is supposed to throw or
7661 catch exceptions, you must link it using the G++ or GCJ driver, as
7662 appropriate for the languages used in the program, or using the
7663 option `-shared-libgcc', such that it is linked with the shared
7667 Bind references to global symbols when building a shared object.
7668 Warn about any unresolved references (unless overridden by the
7669 link editor option `-Xlinker -z -Xlinker defs'). Only a few
7670 systems support this option.
7673 Pass OPTION as an option to the linker. You can use this to
7674 supply system-specific linker options which GCC does not know how
7677 If you want to pass an option that takes an argument, you must use
7678 `-Xlinker' twice, once for the option and once for the argument.
7679 For example, to pass `-assert definitions', you must write
7680 `-Xlinker -assert -Xlinker definitions'. It does not work to write
7681 `-Xlinker "-assert definitions"', because this passes the entire
7682 string as a single argument, which is not what the linker expects.
7685 Pass OPTION as an option to the linker. If OPTION contains
7686 commas, it is split into multiple options at the commas.
7689 Pretend the symbol SYMBOL is undefined, to force linking of
7690 library modules to define it. You can use `-u' multiple times with
7691 different symbols to force loading of additional library modules.
7693 ---------- Footnotes ----------
7695 (1) On some systems, `gcc -shared' needs to build supplementary stub
7696 code for constructors to work. On multi-libbed systems, `gcc -shared'
7697 must select the correct support libraries to link against. Failing to
7698 supply the correct flags may lead to subtle defects. Supplying them in
7699 cases where they are not necessary is innocuous.
7702 File: gcc.info, Node: Directory Options, Next: Spec Files, Prev: Link Options, Up: Invoking GCC
7704 3.14 Options for Directory Search
7705 =================================
7707 These options specify directories to search for header files, for
7708 libraries and for parts of the compiler:
7711 Add the directory DIR to the head of the list of directories to be
7712 searched for header files. This can be used to override a system
7713 header file, substituting your own version, since these
7714 directories are searched before the system header file
7715 directories. However, you should not use this option to add
7716 directories that contain vendor-supplied system header files (use
7717 `-isystem' for that). If you use more than one `-I' option, the
7718 directories are scanned in left-to-right order; the standard
7719 system directories come after.
7721 If a standard system include directory, or a directory specified
7722 with `-isystem', is also specified with `-I', the `-I' option will
7723 be ignored. The directory will still be searched but as a system
7724 directory at its normal position in the system include chain.
7725 This is to ensure that GCC's procedure to fix buggy system headers
7726 and the ordering for the include_next directive are not
7727 inadvertently changed. If you really need to change the search
7728 order for system directories, use the `-nostdinc' and/or
7732 Add the directory DIR to the head of the list of directories to be
7733 searched for header files only for the case of `#include "FILE"';
7734 they are not searched for `#include <FILE>', otherwise just like
7738 Add directory DIR to the list of directories to be searched for
7742 This option specifies where to find the executables, libraries,
7743 include files, and data files of the compiler itself.
7745 The compiler driver program runs one or more of the subprograms
7746 `cpp', `cc1', `as' and `ld'. It tries PREFIX as a prefix for each
7747 program it tries to run, both with and without `MACHINE/VERSION/'
7748 (*note Target Options::).
7750 For each subprogram to be run, the compiler driver first tries the
7751 `-B' prefix, if any. If that name is not found, or if `-B' was
7752 not specified, the driver tries two standard prefixes, which are
7753 `/usr/lib/gcc/' and `/usr/local/lib/gcc/'. If neither of those
7754 results in a file name that is found, the unmodified program name
7755 is searched for using the directories specified in your `PATH'
7756 environment variable.
7758 The compiler will check to see if the path provided by the `-B'
7759 refers to a directory, and if necessary it will add a directory
7760 separator character at the end of the path.
7762 `-B' prefixes that effectively specify directory names also apply
7763 to libraries in the linker, because the compiler translates these
7764 options into `-L' options for the linker. They also apply to
7765 includes files in the preprocessor, because the compiler
7766 translates these options into `-isystem' options for the
7767 preprocessor. In this case, the compiler appends `include' to the
7770 The run-time support file `libgcc.a' can also be searched for using
7771 the `-B' prefix, if needed. If it is not found there, the two
7772 standard prefixes above are tried, and that is all. The file is
7773 left out of the link if it is not found by those means.
7775 Another way to specify a prefix much like the `-B' prefix is to use
7776 the environment variable `GCC_EXEC_PREFIX'. *Note Environment
7779 As a special kludge, if the path provided by `-B' is
7780 `[dir/]stageN/', where N is a number in the range 0 to 9, then it
7781 will be replaced by `[dir/]include'. This is to help with
7782 boot-strapping the compiler.
7785 Process FILE after the compiler reads in the standard `specs'
7786 file, in order to override the defaults that the `gcc' driver
7787 program uses when determining what switches to pass to `cc1',
7788 `cc1plus', `as', `ld', etc. More than one `-specs=FILE' can be
7789 specified on the command line, and they are processed in order,
7793 Use DIR as the logical root directory for headers and libraries.
7794 For example, if the compiler would normally search for headers in
7795 `/usr/include' and libraries in `/usr/lib', it will instead search
7796 `DIR/usr/include' and `DIR/usr/lib'.
7798 If you use both this option and the `-isysroot' option, then the
7799 `--sysroot' option will apply to libraries, but the `-isysroot'
7800 option will apply to header files.
7802 The GNU linker (beginning with version 2.16) has the necessary
7803 support for this option. If your linker does not support this
7804 option, the header file aspect of `--sysroot' will still work, but
7805 the library aspect will not.
7808 This option has been deprecated. Please use `-iquote' instead for
7809 `-I' directories before the `-I-' and remove the `-I-'. Any
7810 directories you specify with `-I' options before the `-I-' option
7811 are searched only for the case of `#include "FILE"'; they are not
7812 searched for `#include <FILE>'.
7814 If additional directories are specified with `-I' options after
7815 the `-I-', these directories are searched for all `#include'
7816 directives. (Ordinarily _all_ `-I' directories are used this way.)
7818 In addition, the `-I-' option inhibits the use of the current
7819 directory (where the current input file came from) as the first
7820 search directory for `#include "FILE"'. There is no way to
7821 override this effect of `-I-'. With `-I.' you can specify
7822 searching the directory which was current when the compiler was
7823 invoked. That is not exactly the same as what the preprocessor
7824 does by default, but it is often satisfactory.
7826 `-I-' does not inhibit the use of the standard system directories
7827 for header files. Thus, `-I-' and `-nostdinc' are independent.
7830 File: gcc.info, Node: Spec Files, Next: Target Options, Prev: Directory Options, Up: Invoking GCC
7832 3.15 Specifying subprocesses and the switches to pass to them
7833 =============================================================
7835 `gcc' is a driver program. It performs its job by invoking a sequence
7836 of other programs to do the work of compiling, assembling and linking.
7837 GCC interprets its command-line parameters and uses these to deduce
7838 which programs it should invoke, and which command-line options it
7839 ought to place on their command lines. This behavior is controlled by
7840 "spec strings". In most cases there is one spec string for each
7841 program that GCC can invoke, but a few programs have multiple spec
7842 strings to control their behavior. The spec strings built into GCC can
7843 be overridden by using the `-specs=' command-line switch to specify a
7846 "Spec files" are plaintext files that are used to construct spec
7847 strings. They consist of a sequence of directives separated by blank
7848 lines. The type of directive is determined by the first non-whitespace
7849 character on the line and it can be one of the following:
7852 Issues a COMMAND to the spec file processor. The commands that can
7856 Search for FILE and insert its text at the current point in
7859 `%include_noerr <FILE>'
7860 Just like `%include', but do not generate an error message if
7861 the include file cannot be found.
7863 `%rename OLD_NAME NEW_NAME'
7864 Rename the spec string OLD_NAME to NEW_NAME.
7868 This tells the compiler to create, override or delete the named
7869 spec string. All lines after this directive up to the next
7870 directive or blank line are considered to be the text for the spec
7871 string. If this results in an empty string then the spec will be
7872 deleted. (Or, if the spec did not exist, then nothing will
7873 happened.) Otherwise, if the spec does not currently exist a new
7874 spec will be created. If the spec does exist then its contents
7875 will be overridden by the text of this directive, unless the first
7876 character of that text is the `+' character, in which case the
7877 text will be appended to the spec.
7880 Creates a new `[SUFFIX] spec' pair. All lines after this directive
7881 and up to the next directive or blank line are considered to make
7882 up the spec string for the indicated suffix. When the compiler
7883 encounters an input file with the named suffix, it will processes
7884 the spec string in order to work out how to compile that file.
7890 This says that any input file whose name ends in `.ZZ' should be
7891 passed to the program `z-compile', which should be invoked with the
7892 command-line switch `-input' and with the result of performing the
7893 `%i' substitution. (See below.)
7895 As an alternative to providing a spec string, the text that
7896 follows a suffix directive can be one of the following:
7899 This says that the suffix is an alias for a known LANGUAGE.
7900 This is similar to using the `-x' command-line switch to GCC
7901 to specify a language explicitly. For example:
7906 Says that .ZZ files are, in fact, C++ source files.
7909 This causes an error messages saying:
7911 NAME compiler not installed on this system.
7913 GCC already has an extensive list of suffixes built into it. This
7914 directive will add an entry to the end of the list of suffixes, but
7915 since the list is searched from the end backwards, it is
7916 effectively possible to override earlier entries using this
7920 GCC has the following spec strings built into it. Spec files can
7921 override these strings or create their own. Note that individual
7922 targets can also add their own spec strings to this list.
7924 asm Options to pass to the assembler
7925 asm_final Options to pass to the assembler post-processor
7926 cpp Options to pass to the C preprocessor
7927 cc1 Options to pass to the C compiler
7928 cc1plus Options to pass to the C++ compiler
7929 endfile Object files to include at the end of the link
7930 link Options to pass to the linker
7931 lib Libraries to include on the command line to the linker
7932 libgcc Decides which GCC support library to pass to the linker
7933 linker Sets the name of the linker
7934 predefines Defines to be passed to the C preprocessor
7935 signed_char Defines to pass to CPP to say whether `char' is signed
7937 startfile Object files to include at the start of the link
7939 Here is a small example of a spec file:
7944 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7946 This example renames the spec called `lib' to `old_lib' and then
7947 overrides the previous definition of `lib' with a new one. The new
7948 definition adds in some extra command-line options before including the
7949 text of the old definition.
7951 "Spec strings" are a list of command-line options to be passed to their
7952 corresponding program. In addition, the spec strings can contain
7953 `%'-prefixed sequences to substitute variable text or to conditionally
7954 insert text into the command line. Using these constructs it is
7955 possible to generate quite complex command lines.
7957 Here is a table of all defined `%'-sequences for spec strings. Note
7958 that spaces are not generated automatically around the results of
7959 expanding these sequences. Therefore you can concatenate them together
7960 or combine them with constant text in a single argument.
7963 Substitute one `%' into the program name or argument.
7966 Substitute the name of the input file being processed.
7969 Substitute the basename of the input file being processed. This
7970 is the substring up to (and not including) the last period and not
7971 including the directory.
7974 This is the same as `%b', but include the file suffix (text after
7978 Marks the argument containing or following the `%d' as a temporary
7979 file name, so that that file will be deleted if GCC exits
7980 successfully. Unlike `%g', this contributes no text to the
7984 Substitute a file name that has suffix SUFFIX and is chosen once
7985 per compilation, and mark the argument in the same way as `%d'.
7986 To reduce exposure to denial-of-service attacks, the file name is
7987 now chosen in a way that is hard to predict even when previously
7988 chosen file names are known. For example, `%g.s ... %g.o ... %g.s'
7989 might turn into `ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s'. SUFFIX
7990 matches the regexp `[.A-Za-z]*' or the special string `%O', which
7991 is treated exactly as if `%O' had been preprocessed. Previously,
7992 `%g' was simply substituted with a file name chosen once per
7993 compilation, without regard to any appended suffix (which was
7994 therefore treated just like ordinary text), making such attacks
7995 more likely to succeed.
7998 Like `%g', but generates a new temporary file name even if
7999 `%uSUFFIX' was already seen.
8002 Substitutes the last file name generated with `%uSUFFIX',
8003 generating a new one if there is no such last file name. In the
8004 absence of any `%uSUFFIX', this is just like `%gSUFFIX', except
8005 they don't share the same suffix _space_, so `%g.s ... %U.s ...
8006 %g.s ... %U.s' would involve the generation of two distinct file
8007 names, one for each `%g.s' and another for each `%U.s'.
8008 Previously, `%U' was simply substituted with a file name chosen
8009 for the previous `%u', without regard to any appended suffix.
8012 Substitutes the name of the `HOST_BIT_BUCKET', if any, and if it is
8013 writable, and if save-temps is off; otherwise, substitute the name
8014 of a temporary file, just like `%u'. This temporary file is not
8015 meant for communication between processes, but rather as a junk
8020 Like `%g', except if `-pipe' is in effect. In that case `%|'
8021 substitutes a single dash and `%m' substitutes nothing at all.
8022 These are the two most common ways to instruct a program that it
8023 should read from standard input or write to standard output. If
8024 you need something more elaborate you can use an `%{pipe:`X'}'
8025 construct: see for example `f/lang-specs.h'.
8028 Substitutes .SUFFIX for the suffixes of a matched switch's args
8029 when it is subsequently output with `%*'. SUFFIX is terminated by
8030 the next space or %.
8033 Marks the argument containing or following the `%w' as the
8034 designated output file of this compilation. This puts the argument
8035 into the sequence of arguments that `%o' will substitute later.
8038 Substitutes the names of all the output files, with spaces
8039 automatically placed around them. You should write spaces around
8040 the `%o' as well or the results are undefined. `%o' is for use in
8041 the specs for running the linker. Input files whose names have no
8042 recognized suffix are not compiled at all, but they are included
8043 among the output files, so they will be linked.
8046 Substitutes the suffix for object files. Note that this is
8047 handled specially when it immediately follows `%g, %u, or %U',
8048 because of the need for those to form complete file names. The
8049 handling is such that `%O' is treated exactly as if it had already
8050 been substituted, except that `%g, %u, and %U' do not currently
8051 support additional SUFFIX characters following `%O' as they would
8052 following, for example, `.o'.
8055 Substitutes the standard macro predefinitions for the current
8056 target machine. Use this when running `cpp'.
8059 Like `%p', but puts `__' before and after the name of each
8060 predefined macro, except for macros that start with `__' or with
8061 `_L', where L is an uppercase letter. This is for ISO C.
8064 Substitute any of `-iprefix' (made from `GCC_EXEC_PREFIX'),
8065 `-isysroot' (made from `TARGET_SYSTEM_ROOT'), `-isystem' (made
8066 from `COMPILER_PATH' and `-B' options) and `-imultilib' as
8070 Current argument is the name of a library or startup file of some
8071 sort. Search for that file in a standard list of directories and
8072 substitute the full name found.
8075 Print STR as an error message. STR is terminated by a newline.
8076 Use this when inconsistent options are detected.
8079 Substitute the contents of spec string NAME at this point.
8082 Like `%(...)' but put `__' around `-D' arguments.
8085 Accumulate an option for `%X'.
8088 Output the accumulated linker options specified by `-Wl' or a `%x'
8092 Output the accumulated assembler options specified by `-Wa'.
8095 Output the accumulated preprocessor options specified by `-Wp'.
8098 Process the `asm' spec. This is used to compute the switches to
8099 be passed to the assembler.
8102 Process the `asm_final' spec. This is a spec string for passing
8103 switches to an assembler post-processor, if such a program is
8107 Process the `link' spec. This is the spec for computing the
8108 command line passed to the linker. Typically it will make use of
8109 the `%L %G %S %D and %E' sequences.
8112 Dump out a `-L' option for each directory that GCC believes might
8113 contain startup files. If the target supports multilibs then the
8114 current multilib directory will be prepended to each of these
8118 Process the `lib' spec. This is a spec string for deciding which
8119 libraries should be included on the command line to the linker.
8122 Process the `libgcc' spec. This is a spec string for deciding
8123 which GCC support library should be included on the command line
8127 Process the `startfile' spec. This is a spec for deciding which
8128 object files should be the first ones passed to the linker.
8129 Typically this might be a file named `crt0.o'.
8132 Process the `endfile' spec. This is a spec string that specifies
8133 the last object files that will be passed to the linker.
8136 Process the `cpp' spec. This is used to construct the arguments
8137 to be passed to the C preprocessor.
8140 Process the `cc1' spec. This is used to construct the options to
8141 be passed to the actual C compiler (`cc1').
8144 Process the `cc1plus' spec. This is used to construct the options
8145 to be passed to the actual C++ compiler (`cc1plus').
8148 Substitute the variable part of a matched option. See below.
8149 Note that each comma in the substituted string is replaced by a
8153 Remove all occurrences of `-S' from the command line. Note--this
8154 command is position dependent. `%' commands in the spec string
8155 before this one will see `-S', `%' commands in the spec string
8156 after this one will not.
8159 Call the named function FUNCTION, passing it ARGS. ARGS is first
8160 processed as a nested spec string, then split into an argument
8161 vector in the usual fashion. The function returns a string which
8162 is processed as if it had appeared literally as part of the
8165 The following built-in spec functions are provided:
8168 The `getenv' spec function takes two arguments: an environment
8169 variable name and a string. If the environment variable is
8170 not defined, a fatal error is issued. Otherwise, the return
8171 value is the value of the environment variable concatenated
8172 with the string. For example, if `TOPDIR' is defined as
8173 `/path/to/top', then:
8175 %:getenv(TOPDIR /include)
8177 expands to `/path/to/top/include'.
8180 The `if-exists' spec function takes one argument, an absolute
8181 pathname to a file. If the file exists, `if-exists' returns
8182 the pathname. Here is a small example of its usage:
8185 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8188 The `if-exists-else' spec function is similar to the
8189 `if-exists' spec function, except that it takes two
8190 arguments. The first argument is an absolute pathname to a
8191 file. If the file exists, `if-exists-else' returns the
8192 pathname. If it does not exist, it returns the second
8193 argument. This way, `if-exists-else' can be used to select
8194 one file or another, based on the existence of the first.
8195 Here is a small example of its usage:
8198 crt0%O%s %:if-exists(crti%O%s) \
8199 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8202 The `replace-outfile' spec function takes two arguments. It
8203 looks for the first argument in the outfiles array and
8204 replaces it with the second argument. Here is a small
8205 example of its usage:
8207 %{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)}
8209 ``print-asm-header''
8210 The `print-asm-header' function takes no arguments and simply
8211 prints a banner like:
8216 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8218 It is used to separate compiler options from assembler options
8219 in the `--target-help' output.
8222 Substitutes the `-S' switch, if that switch was given to GCC. If
8223 that switch was not specified, this substitutes nothing. Note that
8224 the leading dash is omitted when specifying this option, and it is
8225 automatically inserted if the substitution is performed. Thus the
8226 spec string `%{foo}' would match the command-line option `-foo'
8227 and would output the command line option `-foo'.
8230 Like %{`S'} but mark last argument supplied within as a file to be
8234 Substitutes all the switches specified to GCC whose names start
8235 with `-S', but which also take an argument. This is used for
8236 switches like `-o', `-D', `-I', etc. GCC considers `-o foo' as
8237 being one switch whose names starts with `o'. %{o*} would
8238 substitute this text, including the space. Thus two arguments
8242 Like %{`S'*}, but preserve order of `S' and `T' options (the order
8243 of `S' and `T' in the spec is not significant). There can be any
8244 number of ampersand-separated variables; for each the wild card is
8245 optional. Useful for CPP as `%{D*&U*&A*}'.
8248 Substitutes `X', if the `-S' switch was given to GCC.
8251 Substitutes `X', if the `-S' switch was _not_ given to GCC.
8254 Substitutes `X' if one or more switches whose names start with
8255 `-S' are specified to GCC. Normally `X' is substituted only once,
8256 no matter how many such switches appeared. However, if `%*'
8257 appears somewhere in `X', then `X' will be substituted once for
8258 each matching switch, with the `%*' replaced by the part of that
8259 switch that matched the `*'.
8262 Substitutes `X', if processing a file with suffix `S'.
8265 Substitutes `X', if _not_ processing a file with suffix `S'.
8268 Substitutes `X', if processing a file for language `S'.
8271 Substitutes `X', if not processing a file for language `S'.
8274 Substitutes `X' if either `-S' or `-P' was given to GCC. This may
8275 be combined with `!', `.', `,', and `*' sequences as well,
8276 although they have a stronger binding than the `|'. If `%*'
8277 appears in `X', all of the alternatives must be starred, and only
8278 the first matching alternative is substituted.
8280 For example, a spec string like this:
8282 %{.c:-foo} %{!.c:-bar} %{.c|d:-baz} %{!.c|d:-boggle}
8284 will output the following command-line options from the following
8285 input command-line options:
8289 -d fred.c -foo -baz -boggle
8290 -d jim.d -bar -baz -boggle
8293 If `S' was given to GCC, substitutes `X'; else if `T' was given to
8294 GCC, substitutes `Y'; else substitutes `D'. There can be as many
8295 clauses as you need. This may be combined with `.', `,', `!',
8296 `|', and `*' as needed.
8299 The conditional text `X' in a %{`S':`X'} or similar construct may
8300 contain other nested `%' constructs or spaces, or even newlines. They
8301 are processed as usual, as described above. Trailing white space in
8302 `X' is ignored. White space may also appear anywhere on the left side
8303 of the colon in these constructs, except between `.' or `*' and the
8306 The `-O', `-f', `-m', and `-W' switches are handled specifically in
8307 these constructs. If another value of `-O' or the negated form of a
8308 `-f', `-m', or `-W' switch is found later in the command line, the
8309 earlier switch value is ignored, except with {`S'*} where `S' is just
8310 one letter, which passes all matching options.
8312 The character `|' at the beginning of the predicate text is used to
8313 indicate that a command should be piped to the following command, but
8314 only if `-pipe' is specified.
8316 It is built into GCC which switches take arguments and which do not.
8317 (You might think it would be useful to generalize this to allow each
8318 compiler's spec to say which switches take arguments. But this cannot
8319 be done in a consistent fashion. GCC cannot even decide which input
8320 files have been specified without knowing which switches take arguments,
8321 and it must know which input files to compile in order to tell which
8324 GCC also knows implicitly that arguments starting in `-l' are to be
8325 treated as compiler output files, and passed to the linker in their
8326 proper position among the other output files.
8329 File: gcc.info, Node: Target Options, Next: Submodel Options, Prev: Spec Files, Up: Invoking GCC
8331 3.16 Specifying Target Machine and Compiler Version
8332 ===================================================
8334 The usual way to run GCC is to run the executable called `gcc', or
8335 `<machine>-gcc' when cross-compiling, or `<machine>-gcc-<version>' to
8336 run a version other than the one that was installed last. Sometimes
8337 this is inconvenient, so GCC provides options that will switch to
8338 another cross-compiler or version.
8341 The argument MACHINE specifies the target machine for compilation.
8343 The value to use for MACHINE is the same as was specified as the
8344 machine type when configuring GCC as a cross-compiler. For
8345 example, if a cross-compiler was configured with `configure
8346 arm-elf', meaning to compile for an arm processor with elf
8347 binaries, then you would specify `-b arm-elf' to run that cross
8348 compiler. Because there are other options beginning with `-b', the
8349 configuration must contain a hyphen.
8352 The argument VERSION specifies which version of GCC to run. This
8353 is useful when multiple versions are installed. For example,
8354 VERSION might be `4.0', meaning to run GCC version 4.0.
8356 The `-V' and `-b' options work by running the
8357 `<machine>-gcc-<version>' executable, so there's no real reason to use
8358 them if you can just run that directly.
8361 File: gcc.info, Node: Submodel Options, Next: Code Gen Options, Prev: Target Options, Up: Invoking GCC
8363 3.17 Hardware Models and Configurations
8364 =======================================
8366 Earlier we discussed the standard option `-b' which chooses among
8367 different installed compilers for completely different target machines,
8368 such as VAX vs. 68000 vs. 80386.
8370 In addition, each of these target machine types can have its own
8371 special options, starting with `-m', to choose among various hardware
8372 models or configurations--for example, 68010 vs 68020, floating
8373 coprocessor or none. A single installed version of the compiler can
8374 compile for any model or configuration, according to the options
8377 Some configurations of the compiler also support additional special
8378 options, usually for compatibility with other compilers on the same
8386 * Blackfin Options::
8390 * DEC Alpha Options::
8391 * DEC Alpha/VMS Options::
8393 * GNU/Linux Options::
8396 * i386 and x86-64 Options::
8409 * RS/6000 and PowerPC Options::
8410 * S/390 and zSeries Options::
8415 * System V Options::
8420 * Xstormy16 Options::
8425 File: gcc.info, Node: ARC Options, Next: ARM Options, Up: Submodel Options
8430 These options are defined for ARC implementations:
8433 Compile code for little endian mode. This is the default.
8436 Compile code for big endian mode.
8439 Prepend the name of the cpu to all public symbol names. In
8440 multiple-processor systems, there are many ARC variants with
8441 different instruction and register set characteristics. This flag
8442 prevents code compiled for one cpu to be linked with code compiled
8443 for another. No facility exists for handling variants that are
8444 "almost identical". This is an all or nothing option.
8447 Compile code for ARC variant CPU. Which variants are supported
8448 depend on the configuration. All variants support `-mcpu=base',
8449 this is the default.
8451 `-mtext=TEXT-SECTION'
8452 `-mdata=DATA-SECTION'
8453 `-mrodata=READONLY-DATA-SECTION'
8454 Put functions, data, and readonly data in TEXT-SECTION,
8455 DATA-SECTION, and READONLY-DATA-SECTION respectively by default.
8456 This can be overridden with the `section' attribute. *Note
8457 Variable Attributes::.
8461 File: gcc.info, Node: ARM Options, Next: AVR Options, Prev: ARC Options, Up: Submodel Options
8466 These `-m' options are defined for Advanced RISC Machines (ARM)
8470 Generate code for the specified ABI. Permissible values are:
8471 `apcs-gnu', `atpcs', `aapcs', `aapcs-linux' and `iwmmxt'.
8474 Generate a stack frame that is compliant with the ARM Procedure
8475 Call Standard for all functions, even if this is not strictly
8476 necessary for correct execution of the code. Specifying
8477 `-fomit-frame-pointer' with this option will cause the stack
8478 frames not to be generated for leaf functions. The default is
8482 This is a synonym for `-mapcs-frame'.
8485 Generate code which supports calling between the ARM and Thumb
8486 instruction sets. Without this option the two instruction sets
8487 cannot be reliably used inside one program. The default is
8488 `-mno-thumb-interwork', since slightly larger code is generated
8489 when `-mthumb-interwork' is specified.
8492 Prevent the reordering of instructions in the function prolog, or
8493 the merging of those instruction with the instructions in the
8494 function's body. This means that all functions will start with a
8495 recognizable set of instructions (or in fact one of a choice from
8496 a small set of different function prologues), and this information
8497 can be used to locate the start if functions inside an executable
8498 piece of code. The default is `-msched-prolog'.
8501 Generate output containing floating point instructions. This is
8505 Generate output containing library calls for floating point.
8506 *Warning:* the requisite libraries are not available for all ARM
8507 targets. Normally the facilities of the machine's usual C
8508 compiler are used, but this cannot be done directly in
8509 cross-compilation. You must make your own arrangements to provide
8510 suitable library functions for cross-compilation.
8512 `-msoft-float' changes the calling convention in the output file;
8513 therefore, it is only useful if you compile _all_ of a program with
8514 this option. In particular, you need to compile `libgcc.a', the
8515 library that comes with GCC, with `-msoft-float' in order for this
8519 Specifies which ABI to use for floating point values. Permissible
8520 values are: `soft', `softfp' and `hard'.
8522 `soft' and `hard' are equivalent to `-msoft-float' and
8523 `-mhard-float' respectively. `softfp' allows the generation of
8524 floating point instructions, but still uses the soft-float calling
8528 Generate code for a processor running in little-endian mode. This
8529 is the default for all standard configurations.
8532 Generate code for a processor running in big-endian mode; the
8533 default is to compile code for a little-endian processor.
8535 `-mwords-little-endian'
8536 This option only applies when generating code for big-endian
8537 processors. Generate code for a little-endian word order but a
8538 big-endian byte order. That is, a byte order of the form
8539 `32107654'. Note: this option should only be used if you require
8540 compatibility with code for big-endian ARM processors generated by
8541 versions of the compiler prior to 2.8.
8544 This specifies the name of the target ARM processor. GCC uses
8545 this name to determine what kind of instructions it can emit when
8546 generating assembly code. Permissible names are: `arm2', `arm250',
8547 `arm3', `arm6', `arm60', `arm600', `arm610', `arm620', `arm7',
8548 `arm7m', `arm7d', `arm7dm', `arm7di', `arm7dmi', `arm70', `arm700',
8549 `arm700i', `arm710', `arm710c', `arm7100', `arm7500', `arm7500fe',
8550 `arm7tdmi', `arm7tdmi-s', `arm8', `strongarm', `strongarm110',
8551 `strongarm1100', `arm8', `arm810', `arm9', `arm9e', `arm920',
8552 `arm920t', `arm922t', `arm946e-s', `arm966e-s', `arm968e-s',
8553 `arm926ej-s', `arm940t', `arm9tdmi', `arm10tdmi', `arm1020t',
8554 `arm1026ej-s', `arm10e', `arm1020e', `arm1022e', `arm1136j-s',
8555 `arm1136jf-s', `mpcore', `mpcorenovfp', `arm1156t2-s',
8556 `arm1176jz-s', `arm1176jzf-s', `cortex-a8', `cortex-r4',
8557 `cortex-m3', `xscale', `iwmmxt', `ep9312'.
8560 This option is very similar to the `-mcpu=' option, except that
8561 instead of specifying the actual target processor type, and hence
8562 restricting which instructions can be used, it specifies that GCC
8563 should tune the performance of the code as if the target were of
8564 the type specified in this option, but still choosing the
8565 instructions that it will generate based on the cpu specified by a
8566 `-mcpu=' option. For some ARM implementations better performance
8567 can be obtained by using this option.
8570 This specifies the name of the target ARM architecture. GCC uses
8571 this name to determine what kind of instructions it can emit when
8572 generating assembly code. This option can be used in conjunction
8573 with or instead of the `-mcpu=' option. Permissible names are:
8574 `armv2', `armv2a', `armv3', `armv3m', `armv4', `armv4t', `armv5',
8575 `armv5t', `armv5te', `armv6', `armv6j', `armv6t2', `armv6z',
8576 `armv6zk', `armv7', `armv7-a', `armv7-r', `armv7-m', `iwmmxt',
8582 This specifies what floating point hardware (or hardware
8583 emulation) is available on the target. Permissible names are:
8584 `fpa', `fpe2', `fpe3', `maverick', `vfp'. `-mfp' and `-mfpe' are
8585 synonyms for `-mfpu'=`fpe'NUMBER, for compatibility with older
8588 If `-msoft-float' is specified this specifies the format of
8589 floating point values.
8591 `-mstructure-size-boundary=N'
8592 The size of all structures and unions will be rounded up to a
8593 multiple of the number of bits set by this option. Permissible
8594 values are 8, 32 and 64. The default value varies for different
8595 toolchains. For the COFF targeted toolchain the default value is
8596 8. A value of 64 is only allowed if the underlying ABI supports
8599 Specifying the larger number can produce faster, more efficient
8600 code, but can also increase the size of the program. Different
8601 values are potentially incompatible. Code compiled with one value
8602 cannot necessarily expect to work with code or libraries compiled
8603 with another value, if they exchange information using structures
8606 `-mabort-on-noreturn'
8607 Generate a call to the function `abort' at the end of a `noreturn'
8608 function. It will be executed if the function tries to return.
8612 Tells the compiler to perform function calls by first loading the
8613 address of the function into a register and then performing a
8614 subroutine call on this register. This switch is needed if the
8615 target function will lie outside of the 64 megabyte addressing
8616 range of the offset based version of subroutine call instruction.
8618 Even if this switch is enabled, not all function calls will be
8619 turned into long calls. The heuristic is that static functions,
8620 functions which have the `short-call' attribute, functions that
8621 are inside the scope of a `#pragma no_long_calls' directive and
8622 functions whose definitions have already been compiled within the
8623 current compilation unit, will not be turned into long calls. The
8624 exception to this rule is that weak function definitions,
8625 functions with the `long-call' attribute or the `section'
8626 attribute, and functions that are within the scope of a `#pragma
8627 long_calls' directive, will always be turned into long calls.
8629 This feature is not enabled by default. Specifying
8630 `-mno-long-calls' will restore the default behavior, as will
8631 placing the function calls within the scope of a `#pragma
8632 long_calls_off' directive. Note these switches have no effect on
8633 how the compiler generates code to handle function calls via
8636 `-mnop-fun-dllimport'
8637 Disable support for the `dllimport' attribute.
8640 Treat the register used for PIC addressing as read-only, rather
8641 than loading it in the prologue for each function. The run-time
8642 system is responsible for initializing this register with an
8643 appropriate value before execution begins.
8645 `-mpic-register=REG'
8646 Specify the register to be used for PIC addressing. The default
8647 is R10 unless stack-checking is enabled, when R9 is used.
8649 `-mcirrus-fix-invalid-insns'
8650 Insert NOPs into the instruction stream to in order to work around
8651 problems with invalid Maverick instruction combinations. This
8652 option is only valid if the `-mcpu=ep9312' option has been used to
8653 enable generation of instructions for the Cirrus Maverick floating
8654 point co-processor. This option is not enabled by default, since
8655 the problem is only present in older Maverick implementations.
8656 The default can be re-enabled by use of the
8657 `-mno-cirrus-fix-invalid-insns' switch.
8659 `-mpoke-function-name'
8660 Write the name of each function into the text section, directly
8661 preceding the function prologue. The generated code is similar to
8665 .ascii "arm_poke_function_name", 0
8668 .word 0xff000000 + (t1 - t0)
8669 arm_poke_function_name
8671 stmfd sp!, {fp, ip, lr, pc}
8674 When performing a stack backtrace, code can inspect the value of
8675 `pc' stored at `fp + 0'. If the trace function then looks at
8676 location `pc - 12' and the top 8 bits are set, then we know that
8677 there is a function name embedded immediately preceding this
8678 location and has length `((pc[-3]) & 0xff000000)'.
8681 Generate code for the Thumb instruction set. The default is to
8682 use the 32-bit ARM instruction set. This option automatically
8683 enables either 16-bit Thumb-1 or mixed 16/32-bit Thumb-2
8684 instructions based on the `-mcpu=NAME' and `-march=NAME' options.
8687 Generate a stack frame that is compliant with the Thumb Procedure
8688 Call Standard for all non-leaf functions. (A leaf function is one
8689 that does not call any other functions.) The default is
8693 Generate a stack frame that is compliant with the Thumb Procedure
8694 Call Standard for all leaf functions. (A leaf function is one
8695 that does not call any other functions.) The default is
8696 `-mno-apcs-leaf-frame'.
8698 `-mcallee-super-interworking'
8699 Gives all externally visible functions in the file being compiled
8700 an ARM instruction set header which switches to Thumb mode before
8701 executing the rest of the function. This allows these functions
8702 to be called from non-interworking code.
8704 `-mcaller-super-interworking'
8705 Allows calls via function pointers (including virtual functions) to
8706 execute correctly regardless of whether the target code has been
8707 compiled for interworking or not. There is a small overhead in
8708 the cost of executing a function pointer if this option is enabled.
8711 Specify the access model for the thread local storage pointer.
8712 The valid models are `soft', which generates calls to
8713 `__aeabi_read_tp', `cp15', which fetches the thread pointer from
8714 `cp15' directly (supported in the arm6k architecture), and `auto',
8715 which uses the best available method for the selected processor.
8716 The default setting is `auto'.
8720 File: gcc.info, Node: AVR Options, Next: Blackfin Options, Prev: ARM Options, Up: Submodel Options
8725 These options are defined for AVR implementations:
8728 Specify ATMEL AVR instruction set or MCU type.
8730 Instruction set avr1 is for the minimal AVR core, not supported by
8731 the C compiler, only for assembler programs (MCU types: at90s1200,
8732 attiny10, attiny11, attiny12, attiny15, attiny28).
8734 Instruction set avr2 (default) is for the classic AVR core with up
8735 to 8K program memory space (MCU types: at90s2313, at90s2323,
8736 attiny22, at90s2333, at90s2343, at90s4414, at90s4433, at90s4434,
8737 at90s8515, at90c8534, at90s8535).
8739 Instruction set avr3 is for the classic AVR core with up to 128K
8740 program memory space (MCU types: atmega103, atmega603, at43usb320,
8743 Instruction set avr4 is for the enhanced AVR core with up to 8K
8744 program memory space (MCU types: atmega8, atmega83, atmega85).
8746 Instruction set avr5 is for the enhanced AVR core with up to 128K
8747 program memory space (MCU types: atmega16, atmega161, atmega163,
8748 atmega32, atmega323, atmega64, atmega128, at43usb355, at94k).
8751 Output instruction sizes to the asm file.
8754 Specify the initial stack address, which may be a symbol or
8755 numeric value, `__stack' is the default.
8758 Generated code is not compatible with hardware interrupts. Code
8759 size will be smaller.
8762 Functions prologues/epilogues expanded as call to appropriate
8763 subroutines. Code size will be smaller.
8766 Do not generate tablejump insns which sometimes increase code size.
8769 Change only the low 8 bits of the stack pointer.
8772 Assume int to be 8 bit integer. This affects the sizes of all
8773 types: A char will be 1 byte, an int will be 1 byte, an long will
8774 be 2 bytes and long long will be 4 bytes. Please note that this
8775 option does not comply to the C standards, but it will provide you
8776 with smaller code size.
8779 File: gcc.info, Node: Blackfin Options, Next: CRIS Options, Prev: AVR Options, Up: Submodel Options
8781 3.17.4 Blackfin Options
8782 -----------------------
8784 `-mcpu=CPU[-SIREVISION]'
8785 Specifies the name of the target Blackfin processor. Currently,
8786 CPU can be one of `bf522', `bf523', `bf524', `bf525', `bf526',
8787 `bf527', `bf531', `bf532', `bf533', `bf534', `bf536', `bf537',
8788 `bf538', `bf539', `bf542', `bf544', `bf547', `bf548', `bf549',
8789 `bf561'. The optional SIREVISION specifies the silicon revision
8790 of the target Blackfin processor. Any workarounds available for
8791 the targeted silicon revision will be enabled. If SIREVISION is
8792 `none', no workarounds are enabled. If SIREVISION is `any', all
8793 workarounds for the targeted processor will be enabled. The
8794 `__SILICON_REVISION__' macro is defined to two hexadecimal digits
8795 representing the major and minor numbers in the silicon revision.
8796 If SIREVISION is `none', the `__SILICON_REVISION__' is not
8797 defined. If SIREVISION is `any', the `__SILICON_REVISION__' is
8798 defined to be `0xffff'. If this optional SIREVISION is not used,
8799 GCC assumes the latest known silicon revision of the targeted
8802 Support for `bf561' is incomplete. For `bf561', Only the
8803 processor macro is defined. Without this option, `bf532' is used
8804 as the processor by default. The corresponding predefined
8805 processor macros for CPU is to be defined. And for `bfin-elf'
8806 toolchain, this causes the hardware BSP provided by libgloss to be
8807 linked in if `-msim' is not given.
8810 Specifies that the program will be run on the simulator. This
8811 causes the simulator BSP provided by libgloss to be linked in.
8812 This option has effect only for `bfin-elf' toolchain. Certain
8813 other options, such as `-mid-shared-library' and `-mfdpic', imply
8816 `-momit-leaf-frame-pointer'
8817 Don't keep the frame pointer in a register for leaf functions.
8818 This avoids the instructions to save, set up and restore frame
8819 pointers and makes an extra register available in leaf functions.
8820 The option `-fomit-frame-pointer' removes the frame pointer for
8821 all functions which might make debugging harder.
8824 When enabled, the compiler will ensure that the generated code
8825 does not contain speculative loads after jump instructions. If
8826 this option is used, `__WORKAROUND_SPECULATIVE_LOADS' is defined.
8828 `-mno-specld-anomaly'
8829 Don't generate extra code to prevent speculative loads from
8833 When enabled, the compiler will ensure that the generated code
8834 does not contain CSYNC or SSYNC instructions too soon after
8835 conditional branches. If this option is used,
8836 `__WORKAROUND_SPECULATIVE_SYNCS' is defined.
8838 `-mno-csync-anomaly'
8839 Don't generate extra code to prevent CSYNC or SSYNC instructions
8840 from occurring too soon after a conditional branch.
8843 When enabled, the compiler is free to take advantage of the
8844 knowledge that the entire program fits into the low 64k of memory.
8847 Assume that the program is arbitrarily large. This is the default.
8850 Do stack checking using information placed into L1 scratchpad
8851 memory by the uClinux kernel.
8853 `-mid-shared-library'
8854 Generate code that supports shared libraries via the library ID
8855 method. This allows for execute in place and shared libraries in
8856 an environment without virtual memory management. This option
8857 implies `-fPIC'. With a `bfin-elf' target, this option implies
8860 `-mno-id-shared-library'
8861 Generate code that doesn't assume ID based shared libraries are
8862 being used. This is the default.
8864 `-mleaf-id-shared-library'
8865 Generate code that supports shared libraries via the library ID
8866 method, but assumes that this library or executable won't link
8867 against any other ID shared libraries. That allows the compiler
8868 to use faster code for jumps and calls.
8870 `-mno-leaf-id-shared-library'
8871 Do not assume that the code being compiled won't link against any
8872 ID shared libraries. Slower code will be generated for jump and
8875 `-mshared-library-id=n'
8876 Specified the identification number of the ID based shared library
8877 being compiled. Specifying a value of 0 will generate more
8878 compact code, specifying other values will force the allocation of
8879 that number to the current library but is no more space or time
8880 efficient than omitting this option.
8883 Generate code that allows the data segment to be located in a
8884 different area of memory from the text segment. This allows for
8885 execute in place in an environment without virtual memory
8886 management by eliminating relocations against the text section.
8889 Generate code that assumes that the data segment follows the text
8890 segment. This is the default.
8894 Tells the compiler to perform function calls by first loading the
8895 address of the function into a register and then performing a
8896 subroutine call on this register. This switch is needed if the
8897 target function will lie outside of the 24 bit addressing range of
8898 the offset based version of subroutine call instruction.
8900 This feature is not enabled by default. Specifying
8901 `-mno-long-calls' will restore the default behavior. Note these
8902 switches have no effect on how the compiler generates code to
8903 handle function calls via function pointers.
8906 Link with the fast floating-point library. This library relaxes
8907 some of the IEEE floating-point standard's rules for checking
8908 inputs against Not-a-Number (NAN), in the interest of performance.
8911 Enable inlining of PLT entries in function calls to functions that
8912 are not known to bind locally. It has no effect without `-mfdpic'.
8915 File: gcc.info, Node: CRIS Options, Next: CRX Options, Prev: Blackfin Options, Up: Submodel Options
8920 These options are defined specifically for the CRIS ports.
8922 `-march=ARCHITECTURE-TYPE'
8923 `-mcpu=ARCHITECTURE-TYPE'
8924 Generate code for the specified architecture. The choices for
8925 ARCHITECTURE-TYPE are `v3', `v8' and `v10' for respectively
8926 ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is `v0' except for
8927 cris-axis-linux-gnu, where the default is `v10'.
8929 `-mtune=ARCHITECTURE-TYPE'
8930 Tune to ARCHITECTURE-TYPE everything applicable about the generated
8931 code, except for the ABI and the set of available instructions.
8932 The choices for ARCHITECTURE-TYPE are the same as for
8933 `-march=ARCHITECTURE-TYPE'.
8935 `-mmax-stack-frame=N'
8936 Warn when the stack frame of a function exceeds N bytes.
8938 `-melinux-stacksize=N'
8939 Only available with the `cris-axis-aout' target. Arranges for
8940 indications in the program to the kernel loader that the stack of
8941 the program should be set to N bytes.
8945 The options `-metrax4' and `-metrax100' are synonyms for
8946 `-march=v3' and `-march=v8' respectively.
8948 `-mmul-bug-workaround'
8949 `-mno-mul-bug-workaround'
8950 Work around a bug in the `muls' and `mulu' instructions for CPU
8951 models where it applies. This option is active by default.
8954 Enable CRIS-specific verbose debug-related information in the
8955 assembly code. This option also has the effect to turn off the
8956 `#NO_APP' formatted-code indicator to the assembler at the
8957 beginning of the assembly file.
8960 Do not use condition-code results from previous instruction;
8961 always emit compare and test instructions before use of condition
8965 Do not emit instructions with side-effects in addressing modes
8966 other than post-increment.
8974 These options (no-options) arranges (eliminate arrangements) for
8975 the stack-frame, individual data and constants to be aligned for
8976 the maximum single data access size for the chosen CPU model. The
8977 default is to arrange for 32-bit alignment. ABI details such as
8978 structure layout are not affected by these options.
8983 Similar to the stack- data- and const-align options above, these
8984 options arrange for stack-frame, writable data and constants to
8985 all be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit
8988 `-mno-prologue-epilogue'
8989 `-mprologue-epilogue'
8990 With `-mno-prologue-epilogue', the normal function prologue and
8991 epilogue that sets up the stack-frame are omitted and no return
8992 instructions or return sequences are generated in the code. Use
8993 this option only together with visual inspection of the compiled
8994 code: no warnings or errors are generated when call-saved
8995 registers must be saved, or storage for local variable needs to be
9000 With `-fpic' and `-fPIC', don't generate (do generate) instruction
9001 sequences that load addresses for functions from the PLT part of
9002 the GOT rather than (traditional on other architectures) calls to
9003 the PLT. The default is `-mgotplt'.
9006 Legacy no-op option only recognized with the cris-axis-aout target.
9009 Legacy no-op option only recognized with the cris-axis-elf and
9010 cris-axis-linux-gnu targets.
9013 Only recognized with the cris-axis-aout target, where it selects a
9014 GNU/linux-like multilib, include files and instruction set for
9018 Legacy no-op option only recognized with the cris-axis-linux-gnu
9022 This option, recognized for the cris-axis-aout and cris-axis-elf
9023 arranges to link with input-output functions from a simulator
9024 library. Code, initialized data and zero-initialized data are
9025 allocated consecutively.
9028 Like `-sim', but pass linker options to locate initialized data at
9029 0x40000000 and zero-initialized data at 0x80000000.
9032 File: gcc.info, Node: CRX Options, Next: Darwin Options, Prev: CRIS Options, Up: Submodel Options
9037 These options are defined specifically for the CRX ports.
9040 Enable the use of multiply-accumulate instructions. Disabled by
9044 Push instructions will be used to pass outgoing arguments when
9045 functions are called. Enabled by default.
9048 File: gcc.info, Node: Darwin Options, Next: DEC Alpha Options, Prev: CRX Options, Up: Submodel Options
9050 3.17.7 Darwin Options
9051 ---------------------
9053 These options are defined for all architectures running the Darwin
9056 FSF GCC on Darwin does not create "fat" object files; it will create
9057 an object file for the single architecture that it was built to target.
9058 Apple's GCC on Darwin does create "fat" files if multiple `-arch'
9059 options are used; it does so by running the compiler or linker multiple
9060 times and joining the results together with `lipo'.
9062 The subtype of the file created (like `ppc7400' or `ppc970' or `i686')
9063 is determined by the flags that specify the ISA that GCC is targetting,
9064 like `-mcpu' or `-march'. The `-force_cpusubtype_ALL' option can be
9065 used to override this.
9067 The Darwin tools vary in their behavior when presented with an ISA
9068 mismatch. The assembler, `as', will only permit instructions to be
9069 used that are valid for the subtype of the file it is generating, so
9070 you cannot put 64-bit instructions in an `ppc750' object file. The
9071 linker for shared libraries, `/usr/bin/libtool', will fail and print an
9072 error if asked to create a shared library with a less restrictive
9073 subtype than its input files (for instance, trying to put a `ppc970'
9074 object file in a `ppc7400' library). The linker for executables, `ld',
9075 will quietly give the executable the most restrictive subtype of any of
9079 Add the framework directory DIR to the head of the list of
9080 directories to be searched for header files. These directories are
9081 interleaved with those specified by `-I' options and are scanned
9082 in a left-to-right order.
9084 A framework directory is a directory with frameworks in it. A
9085 framework is a directory with a `"Headers"' and/or
9086 `"PrivateHeaders"' directory contained directly in it that ends in
9087 `".framework"'. The name of a framework is the name of this
9088 directory excluding the `".framework"'. Headers associated with
9089 the framework are found in one of those two directories, with
9090 `"Headers"' being searched first. A subframework is a framework
9091 directory that is in a framework's `"Frameworks"' directory.
9092 Includes of subframework headers can only appear in a header of a
9093 framework that contains the subframework, or in a sibling
9094 subframework header. Two subframeworks are siblings if they occur
9095 in the same framework. A subframework should not have the same
9096 name as a framework, a warning will be issued if this is violated.
9097 Currently a subframework cannot have subframeworks, in the
9098 future, the mechanism may be extended to support this. The
9099 standard frameworks can be found in `"/System/Library/Frameworks"'
9100 and `"/Library/Frameworks"'. An example include looks like
9101 `#include <Framework/header.h>', where `Framework' denotes the
9102 name of the framework and header.h is found in the
9103 `"PrivateHeaders"' or `"Headers"' directory.
9106 Like `-F' except the directory is a treated as a system directory.
9107 The main difference between this `-iframework' and `-F' is that
9108 with `-iframework' the compiler does not warn about constructs
9109 contained within header files found via DIR. This option is valid
9110 only for the C family of languages.
9113 Emit debugging information for symbols that are used. For STABS
9114 debugging format, this enables `-feliminate-unused-debug-symbols'.
9115 This is by default ON.
9118 Emit debugging information for all symbols and types.
9120 `-mmacosx-version-min=VERSION'
9121 The earliest version of MacOS X that this executable will run on
9122 is VERSION. Typical values of VERSION include `10.1', `10.2', and
9125 If the compiler was built to use the system's headers by default,
9126 then the default for this option is the system version on which the
9127 compiler is running, otherwise the default is to make choices which
9128 are compatible with as many systems and code bases as possible.
9131 Enable kernel development mode. The `-mkernel' option sets
9132 `-static', `-fno-common', `-fno-cxa-atexit', `-fno-exceptions',
9133 `-fno-non-call-exceptions', `-fapple-kext', `-fno-weak' and
9134 `-fno-rtti' where applicable. This mode also sets `-mno-altivec',
9135 `-msoft-float', `-fno-builtin' and `-mlong-branch' for PowerPC
9139 Override the defaults for `bool' so that `sizeof(bool)==1'. By
9140 default `sizeof(bool)' is `4' when compiling for Darwin/PowerPC
9141 and `1' when compiling for Darwin/x86, so this option has no
9144 *Warning:* The `-mone-byte-bool' switch causes GCC to generate
9145 code that is not binary compatible with code generated without
9146 that switch. Using this switch may require recompiling all other
9147 modules in a program, including system libraries. Use this switch
9148 to conform to a non-default data model.
9150 `-mfix-and-continue'
9151 `-ffix-and-continue'
9153 Generate code suitable for fast turn around development. Needed to
9154 enable gdb to dynamically load `.o' files into already running
9155 programs. `-findirect-data' and `-ffix-and-continue' are provided
9156 for backwards compatibility.
9159 Loads all members of static archive libraries. See man ld(1) for
9162 `-arch_errors_fatal'
9163 Cause the errors having to do with files that have the wrong
9164 architecture to be fatal.
9167 Causes the output file to be marked such that the dynamic linker
9168 will bind all undefined references when the file is loaded or
9172 Produce a Mach-o bundle format file. See man ld(1) for more
9175 `-bundle_loader EXECUTABLE'
9176 This option specifies the EXECUTABLE that will be loading the build
9177 output file being linked. See man ld(1) for more information.
9180 When passed this option, GCC will produce a dynamic library
9181 instead of an executable when linking, using the Darwin `libtool'
9184 `-force_cpusubtype_ALL'
9185 This causes GCC's output file to have the ALL subtype, instead of
9186 one controlled by the `-mcpu' or `-march' option.
9188 `-allowable_client CLIENT_NAME'
9190 `-compatibility_version'
9195 `-dylinker_install_name'
9197 `-exported_symbols_list'
9200 `-force_flat_namespace'
9201 `-headerpad_max_install_names'
9205 `-keep_private_externs'
9208 `-multiply_defined_unused'
9210 `-no_dead_strip_inits_and_terms'
9217 `-prebind_all_twolevel_modules'
9221 `-sectobjectsymbols'
9225 `-sectobjectsymbols'
9228 `-segs_read_only_addr'
9229 `-segs_read_write_addr'
9231 `-seg_addr_table_filename'
9234 `-segs_read_only_addr'
9235 `-segs_read_write_addr'
9240 `-twolevel_namespace'
9243 `-unexported_symbols_list'
9244 `-weak_reference_mismatches'
9246 These options are passed to the Darwin linker. The Darwin linker
9247 man page describes them in detail.
9250 File: gcc.info, Node: DEC Alpha Options, Next: DEC Alpha/VMS Options, Prev: Darwin Options, Up: Submodel Options
9252 3.17.8 DEC Alpha Options
9253 ------------------------
9255 These `-m' options are defined for the DEC Alpha implementations:
9259 Use (do not use) the hardware floating-point instructions for
9260 floating-point operations. When `-msoft-float' is specified,
9261 functions in `libgcc.a' will be used to perform floating-point
9262 operations. Unless they are replaced by routines that emulate the
9263 floating-point operations, or compiled in such a way as to call
9264 such emulations routines, these routines will issue floating-point
9265 operations. If you are compiling for an Alpha without
9266 floating-point operations, you must ensure that the library is
9267 built so as not to call them.
9269 Note that Alpha implementations without floating-point operations
9270 are required to have floating-point registers.
9274 Generate code that uses (does not use) the floating-point register
9275 set. `-mno-fp-regs' implies `-msoft-float'. If the floating-point
9276 register set is not used, floating point operands are passed in
9277 integer registers as if they were integers and floating-point
9278 results are passed in `$0' instead of `$f0'. This is a
9279 non-standard calling sequence, so any function with a
9280 floating-point argument or return value called by code compiled
9281 with `-mno-fp-regs' must also be compiled with that option.
9283 A typical use of this option is building a kernel that does not
9284 use, and hence need not save and restore, any floating-point
9288 The Alpha architecture implements floating-point hardware
9289 optimized for maximum performance. It is mostly compliant with
9290 the IEEE floating point standard. However, for full compliance,
9291 software assistance is required. This option generates code fully
9292 IEEE compliant code _except_ that the INEXACT-FLAG is not
9293 maintained (see below). If this option is turned on, the
9294 preprocessor macro `_IEEE_FP' is defined during compilation. The
9295 resulting code is less efficient but is able to correctly support
9296 denormalized numbers and exceptional IEEE values such as
9297 not-a-number and plus/minus infinity. Other Alpha compilers call
9298 this option `-ieee_with_no_inexact'.
9300 `-mieee-with-inexact'
9301 This is like `-mieee' except the generated code also maintains the
9302 IEEE INEXACT-FLAG. Turning on this option causes the generated
9303 code to implement fully-compliant IEEE math. In addition to
9304 `_IEEE_FP', `_IEEE_FP_EXACT' is defined as a preprocessor macro.
9305 On some Alpha implementations the resulting code may execute
9306 significantly slower than the code generated by default. Since
9307 there is very little code that depends on the INEXACT-FLAG, you
9308 should normally not specify this option. Other Alpha compilers
9309 call this option `-ieee_with_inexact'.
9311 `-mfp-trap-mode=TRAP-MODE'
9312 This option controls what floating-point related traps are enabled.
9313 Other Alpha compilers call this option `-fptm TRAP-MODE'. The
9314 trap mode can be set to one of four values:
9317 This is the default (normal) setting. The only traps that
9318 are enabled are the ones that cannot be disabled in software
9319 (e.g., division by zero trap).
9322 In addition to the traps enabled by `n', underflow traps are
9326 Like `u', but the instructions are marked to be safe for
9327 software completion (see Alpha architecture manual for
9331 Like `su', but inexact traps are enabled as well.
9333 `-mfp-rounding-mode=ROUNDING-MODE'
9334 Selects the IEEE rounding mode. Other Alpha compilers call this
9335 option `-fprm ROUNDING-MODE'. The ROUNDING-MODE can be one of:
9338 Normal IEEE rounding mode. Floating point numbers are
9339 rounded towards the nearest machine number or towards the
9340 even machine number in case of a tie.
9343 Round towards minus infinity.
9346 Chopped rounding mode. Floating point numbers are rounded
9350 Dynamic rounding mode. A field in the floating point control
9351 register (FPCR, see Alpha architecture reference manual)
9352 controls the rounding mode in effect. The C library
9353 initializes this register for rounding towards plus infinity.
9354 Thus, unless your program modifies the FPCR, `d' corresponds
9355 to round towards plus infinity.
9357 `-mtrap-precision=TRAP-PRECISION'
9358 In the Alpha architecture, floating point traps are imprecise.
9359 This means without software assistance it is impossible to recover
9360 from a floating trap and program execution normally needs to be
9361 terminated. GCC can generate code that can assist operating
9362 system trap handlers in determining the exact location that caused
9363 a floating point trap. Depending on the requirements of an
9364 application, different levels of precisions can be selected:
9367 Program precision. This option is the default and means a
9368 trap handler can only identify which program caused a
9369 floating point exception.
9372 Function precision. The trap handler can determine the
9373 function that caused a floating point exception.
9376 Instruction precision. The trap handler can determine the
9377 exact instruction that caused a floating point exception.
9379 Other Alpha compilers provide the equivalent options called
9380 `-scope_safe' and `-resumption_safe'.
9383 This option marks the generated code as IEEE conformant. You must
9384 not use this option unless you also specify `-mtrap-precision=i'
9385 and either `-mfp-trap-mode=su' or `-mfp-trap-mode=sui'. Its only
9386 effect is to emit the line `.eflag 48' in the function prologue of
9387 the generated assembly file. Under DEC Unix, this has the effect
9388 that IEEE-conformant math library routines will be linked in.
9391 Normally GCC examines a 32- or 64-bit integer constant to see if
9392 it can construct it from smaller constants in two or three
9393 instructions. If it cannot, it will output the constant as a
9394 literal and generate code to load it from the data segment at
9397 Use this option to require GCC to construct _all_ integer constants
9398 using code, even if it takes more instructions (the maximum is
9401 You would typically use this option to build a shared library
9402 dynamic loader. Itself a shared library, it must relocate itself
9403 in memory before it can find the variables and constants in its
9408 Select whether to generate code to be assembled by the
9409 vendor-supplied assembler (`-malpha-as') or by the GNU assembler
9420 Indicate whether GCC should generate code to use the optional BWX,
9421 CIX, FIX and MAX instruction sets. The default is to use the
9422 instruction sets supported by the CPU type specified via `-mcpu='
9423 option or that of the CPU on which GCC was built if none was
9428 Generate code that uses (does not use) VAX F and G floating point
9429 arithmetic instead of IEEE single and double precision.
9432 `-mno-explicit-relocs'
9433 Older Alpha assemblers provided no way to generate symbol
9434 relocations except via assembler macros. Use of these macros does
9435 not allow optimal instruction scheduling. GNU binutils as of
9436 version 2.12 supports a new syntax that allows the compiler to
9437 explicitly mark which relocations should apply to which
9438 instructions. This option is mostly useful for debugging, as GCC
9439 detects the capabilities of the assembler when it is built and
9440 sets the default accordingly.
9444 When `-mexplicit-relocs' is in effect, static data is accessed via
9445 "gp-relative" relocations. When `-msmall-data' is used, objects 8
9446 bytes long or smaller are placed in a "small data area" (the
9447 `.sdata' and `.sbss' sections) and are accessed via 16-bit
9448 relocations off of the `$gp' register. This limits the size of
9449 the small data area to 64KB, but allows the variables to be
9450 directly accessed via a single instruction.
9452 The default is `-mlarge-data'. With this option the data area is
9453 limited to just below 2GB. Programs that require more than 2GB of
9454 data must use `malloc' or `mmap' to allocate the data in the heap
9455 instead of in the program's data segment.
9457 When generating code for shared libraries, `-fpic' implies
9458 `-msmall-data' and `-fPIC' implies `-mlarge-data'.
9462 When `-msmall-text' is used, the compiler assumes that the code of
9463 the entire program (or shared library) fits in 4MB, and is thus
9464 reachable with a branch instruction. When `-msmall-data' is used,
9465 the compiler can assume that all local symbols share the same
9466 `$gp' value, and thus reduce the number of instructions required
9467 for a function call from 4 to 1.
9469 The default is `-mlarge-text'.
9472 Set the instruction set and instruction scheduling parameters for
9473 machine type CPU_TYPE. You can specify either the `EV' style name
9474 or the corresponding chip number. GCC supports scheduling
9475 parameters for the EV4, EV5 and EV6 family of processors and will
9476 choose the default values for the instruction set from the
9477 processor you specify. If you do not specify a processor type,
9478 GCC will default to the processor on which the compiler was built.
9480 Supported values for CPU_TYPE are
9485 Schedules as an EV4 and has no instruction set extensions.
9489 Schedules as an EV5 and has no instruction set extensions.
9493 Schedules as an EV5 and supports the BWX extension.
9498 Schedules as an EV5 and supports the BWX and MAX extensions.
9502 Schedules as an EV6 and supports the BWX, FIX, and MAX
9507 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX
9511 Set only the instruction scheduling parameters for machine type
9512 CPU_TYPE. The instruction set is not changed.
9514 `-mmemory-latency=TIME'
9515 Sets the latency the scheduler should assume for typical memory
9516 references as seen by the application. This number is highly
9517 dependent on the memory access patterns used by the application
9518 and the size of the external cache on the machine.
9520 Valid options for TIME are
9523 A decimal number representing clock cycles.
9529 The compiler contains estimates of the number of clock cycles
9530 for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9531 (also called Dcache, Scache, and Bcache), as well as to main
9532 memory. Note that L3 is only valid for EV5.
9536 File: gcc.info, Node: DEC Alpha/VMS Options, Next: FRV Options, Prev: DEC Alpha Options, Up: Submodel Options
9538 3.17.9 DEC Alpha/VMS Options
9539 ----------------------------
9541 These `-m' options are defined for the DEC Alpha/VMS implementations:
9543 `-mvms-return-codes'
9544 Return VMS condition codes from main. The default is to return
9545 POSIX style condition (e.g. error) codes.
9548 File: gcc.info, Node: FRV Options, Next: GNU/Linux Options, Prev: DEC Alpha/VMS Options, Up: Submodel Options
9554 Only use the first 32 general purpose registers.
9557 Use all 64 general purpose registers.
9560 Use only the first 32 floating point registers.
9563 Use all 64 floating point registers
9566 Use hardware instructions for floating point operations.
9569 Use library routines for floating point operations.
9572 Dynamically allocate condition code registers.
9575 Do not try to dynamically allocate condition code registers, only
9576 use `icc0' and `fcc0'.
9579 Change ABI to use double word insns.
9582 Do not use double word instructions.
9585 Use floating point double instructions.
9588 Do not use floating point double instructions.
9591 Use media instructions.
9594 Do not use media instructions.
9597 Use multiply and add/subtract instructions.
9600 Do not use multiply and add/subtract instructions.
9603 Select the FDPIC ABI, that uses function descriptors to represent
9604 pointers to functions. Without any PIC/PIE-related options, it
9605 implies `-fPIE'. With `-fpic' or `-fpie', it assumes GOT entries
9606 and small data are within a 12-bit range from the GOT base
9607 address; with `-fPIC' or `-fPIE', GOT offsets are computed with 32
9608 bits. With a `bfin-elf' target, this option implies `-msim'.
9611 Enable inlining of PLT entries in function calls to functions that
9612 are not known to bind locally. It has no effect without `-mfdpic'.
9613 It's enabled by default if optimizing for speed and compiling for
9614 shared libraries (i.e., `-fPIC' or `-fpic'), or when an
9615 optimization option such as `-O3' or above is present in the
9619 Assume a large TLS segment when generating thread-local code.
9622 Do not assume a large TLS segment when generating thread-local
9626 Enable the use of `GPREL' relocations in the FDPIC ABI for data
9627 that is known to be in read-only sections. It's enabled by
9628 default, except for `-fpic' or `-fpie': even though it may help
9629 make the global offset table smaller, it trades 1 instruction for
9630 4. With `-fPIC' or `-fPIE', it trades 3 instructions for 4, one
9631 of which may be shared by multiple symbols, and it avoids the need
9632 for a GOT entry for the referenced symbol, so it's more likely to
9633 be a win. If it is not, `-mno-gprel-ro' can be used to disable it.
9635 `-multilib-library-pic'
9636 Link with the (library, not FD) pic libraries. It's implied by
9637 `-mlibrary-pic', as well as by `-fPIC' and `-fpic' without
9638 `-mfdpic'. You should never have to use it explicitly.
9641 Follow the EABI requirement of always creating a frame pointer
9642 whenever a stack frame is allocated. This option is enabled by
9643 default and can be disabled with `-mno-linked-fp'.
9646 Use indirect addressing to call functions outside the current
9647 compilation unit. This allows the functions to be placed anywhere
9648 within the 32-bit address space.
9651 Try to align labels to an 8-byte boundary by inserting nops into
9652 the previous packet. This option only has an effect when VLIW
9653 packing is enabled. It doesn't create new packets; it merely adds
9654 nops to existing ones.
9657 Generate position-independent EABI code.
9660 Use only the first four media accumulator registers.
9663 Use all eight media accumulator registers.
9666 Pack VLIW instructions.
9669 Do not pack VLIW instructions.
9672 Do not mark ABI switches in e_flags.
9675 Enable the use of conditional-move instructions (default).
9677 This switch is mainly for debugging the compiler and will likely
9678 be removed in a future version.
9681 Disable the use of conditional-move instructions.
9683 This switch is mainly for debugging the compiler and will likely
9684 be removed in a future version.
9687 Enable the use of conditional set instructions (default).
9689 This switch is mainly for debugging the compiler and will likely
9690 be removed in a future version.
9693 Disable the use of conditional set instructions.
9695 This switch is mainly for debugging the compiler and will likely
9696 be removed in a future version.
9699 Enable the use of conditional execution (default).
9701 This switch is mainly for debugging the compiler and will likely
9702 be removed in a future version.
9705 Disable the use of conditional execution.
9707 This switch is mainly for debugging the compiler and will likely
9708 be removed in a future version.
9711 Run a pass to pack branches into VLIW instructions (default).
9713 This switch is mainly for debugging the compiler and will likely
9714 be removed in a future version.
9717 Do not run a pass to pack branches into VLIW instructions.
9719 This switch is mainly for debugging the compiler and will likely
9720 be removed in a future version.
9723 Enable optimization of `&&' and `||' in conditional execution
9726 This switch is mainly for debugging the compiler and will likely
9727 be removed in a future version.
9729 `-mno-multi-cond-exec'
9730 Disable optimization of `&&' and `||' in conditional execution.
9732 This switch is mainly for debugging the compiler and will likely
9733 be removed in a future version.
9735 `-mnested-cond-exec'
9736 Enable nested conditional execution optimizations (default).
9738 This switch is mainly for debugging the compiler and will likely
9739 be removed in a future version.
9741 `-mno-nested-cond-exec'
9742 Disable nested conditional execution optimizations.
9744 This switch is mainly for debugging the compiler and will likely
9745 be removed in a future version.
9748 This switch removes redundant `membar' instructions from the
9749 compiler generated code. It is enabled by default.
9751 `-mno-optimize-membar'
9752 This switch disables the automatic removal of redundant `membar'
9753 instructions from the generated code.
9756 Cause gas to print out tomcat statistics.
9759 Select the processor type for which to generate code. Possible
9760 values are `frv', `fr550', `tomcat', `fr500', `fr450', `fr405',
9761 `fr400', `fr300' and `simple'.
9765 File: gcc.info, Node: GNU/Linux Options, Next: H8/300 Options, Prev: FRV Options, Up: Submodel Options
9767 3.17.11 GNU/Linux Options
9768 -------------------------
9770 These `-m' options are defined for GNU/Linux targets:
9773 Use the GNU C library instead of uClibc. This is the default
9774 except on `*-*-linux-*uclibc*' targets.
9777 Use uClibc instead of the GNU C library. This is the default on
9778 `*-*-linux-*uclibc*' targets.
9781 File: gcc.info, Node: H8/300 Options, Next: HPPA Options, Prev: GNU/Linux Options, Up: Submodel Options
9783 3.17.12 H8/300 Options
9784 ----------------------
9786 These `-m' options are defined for the H8/300 implementations:
9789 Shorten some address references at link time, when possible; uses
9790 the linker option `-relax'. *Note `ld' and the H8/300:
9791 (ld)H8/300, for a fuller description.
9794 Generate code for the H8/300H.
9797 Generate code for the H8S.
9800 Generate code for the H8S and H8/300H in the normal mode. This
9801 switch must be used either with `-mh' or `-ms'.
9804 Generate code for the H8S/2600. This switch must be used with
9808 Make `int' data 32 bits by default.
9811 On the H8/300H and H8S, use the same alignment rules as for the
9812 H8/300. The default for the H8/300H and H8S is to align longs and
9813 floats on 4 byte boundaries. `-malign-300' causes them to be
9814 aligned on 2 byte boundaries. This option has no effect on the
9818 File: gcc.info, Node: HPPA Options, Next: i386 and x86-64 Options, Prev: H8/300 Options, Up: Submodel Options
9820 3.17.13 HPPA Options
9821 --------------------
9823 These `-m' options are defined for the HPPA family of computers:
9825 `-march=ARCHITECTURE-TYPE'
9826 Generate code for the specified architecture. The choices for
9827 ARCHITECTURE-TYPE are `1.0' for PA 1.0, `1.1' for PA 1.1, and
9828 `2.0' for PA 2.0 processors. Refer to `/usr/lib/sched.models' on
9829 an HP-UX system to determine the proper architecture option for
9830 your machine. Code compiled for lower numbered architectures will
9831 run on higher numbered architectures, but not the other way around.
9836 Synonyms for `-march=1.0', `-march=1.1', and `-march=2.0'
9840 Generate code suitable for big switch tables. Use this option
9841 only if the assembler/linker complain about out of range branches
9842 within a switch table.
9845 Fill delay slots of function calls with unconditional jump
9846 instructions by modifying the return pointer for the function call
9847 to be the target of the conditional jump.
9850 Prevent floating point registers from being used in any manner.
9851 This is necessary for compiling kernels which perform lazy context
9852 switching of floating point registers. If you use this option and
9853 attempt to perform floating point operations, the compiler will
9856 `-mdisable-indexing'
9857 Prevent the compiler from using indexing address modes. This
9858 avoids some rather obscure problems when compiling MIG generated
9862 Generate code that assumes the target has no space registers.
9863 This allows GCC to generate faster indirect calls and use unscaled
9864 index address modes.
9866 Such code is suitable for level 0 PA systems and kernels.
9868 `-mfast-indirect-calls'
9869 Generate code that assumes calls never cross space boundaries.
9870 This allows GCC to emit code which performs faster indirect calls.
9872 This option will not work in the presence of shared libraries or
9875 `-mfixed-range=REGISTER-RANGE'
9876 Generate code treating the given register range as fixed registers.
9877 A fixed register is one that the register allocator can not use.
9878 This is useful when compiling kernel code. A register range is
9879 specified as two registers separated by a dash. Multiple register
9880 ranges can be specified separated by a comma.
9883 Generate 3-instruction load and store sequences as sometimes
9884 required by the HP-UX 10 linker. This is equivalent to the `+k'
9885 option to the HP compilers.
9887 `-mportable-runtime'
9888 Use the portable calling conventions proposed by HP for ELF
9892 Enable the use of assembler directives only GAS understands.
9894 `-mschedule=CPU-TYPE'
9895 Schedule code according to the constraints for the machine type
9896 CPU-TYPE. The choices for CPU-TYPE are `700' `7100', `7100LC',
9897 `7200', `7300' and `8000'. Refer to `/usr/lib/sched.models' on an
9898 HP-UX system to determine the proper scheduling option for your
9899 machine. The default scheduling is `8000'.
9902 Enable the optimization pass in the HP-UX linker. Note this makes
9903 symbolic debugging impossible. It also triggers a bug in the
9904 HP-UX 8 and HP-UX 9 linkers in which they give bogus error
9905 messages when linking some programs.
9908 Generate output containing library calls for floating point.
9909 *Warning:* the requisite libraries are not available for all HPPA
9910 targets. Normally the facilities of the machine's usual C
9911 compiler are used, but this cannot be done directly in
9912 cross-compilation. You must make your own arrangements to provide
9913 suitable library functions for cross-compilation. The embedded
9914 target `hppa1.1-*-pro' does provide software floating point
9917 `-msoft-float' changes the calling convention in the output file;
9918 therefore, it is only useful if you compile _all_ of a program with
9919 this option. In particular, you need to compile `libgcc.a', the
9920 library that comes with GCC, with `-msoft-float' in order for this
9924 Generate the predefine, `_SIO', for server IO. The default is
9925 `-mwsio'. This generates the predefines, `__hp9000s700',
9926 `__hp9000s700__' and `_WSIO', for workstation IO. These options
9927 are available under HP-UX and HI-UX.
9930 Use GNU ld specific options. This passes `-shared' to ld when
9931 building a shared library. It is the default when GCC is
9932 configured, explicitly or implicitly, with the GNU linker. This
9933 option does not have any affect on which ld is called, it only
9934 changes what parameters are passed to that ld. The ld that is
9935 called is determined by the `--with-ld' configure option, GCC's
9936 program search path, and finally by the user's `PATH'. The linker
9937 used by GCC can be printed using `which `gcc
9938 -print-prog-name=ld`'. This option is only available on the 64
9939 bit HP-UX GCC, i.e. configured with `hppa*64*-*-hpux*'.
9942 Use HP ld specific options. This passes `-b' to ld when building
9943 a shared library and passes `+Accept TypeMismatch' to ld on all
9944 links. It is the default when GCC is configured, explicitly or
9945 implicitly, with the HP linker. This option does not have any
9946 affect on which ld is called, it only changes what parameters are
9947 passed to that ld. The ld that is called is determined by the
9948 `--with-ld' configure option, GCC's program search path, and
9949 finally by the user's `PATH'. The linker used by GCC can be
9950 printed using `which `gcc -print-prog-name=ld`'. This option is
9951 only available on the 64 bit HP-UX GCC, i.e. configured with
9955 Generate code that uses long call sequences. This ensures that a
9956 call is always able to reach linker generated stubs. The default
9957 is to generate long calls only when the distance from the call
9958 site to the beginning of the function or translation unit, as the
9959 case may be, exceeds a predefined limit set by the branch type
9960 being used. The limits for normal calls are 7,600,000 and 240,000
9961 bytes, respectively for the PA 2.0 and PA 1.X architectures.
9962 Sibcalls are always limited at 240,000 bytes.
9964 Distances are measured from the beginning of functions when using
9965 the `-ffunction-sections' option, or when using the `-mgas' and
9966 `-mno-portable-runtime' options together under HP-UX with the SOM
9969 It is normally not desirable to use this option as it will degrade
9970 performance. However, it may be useful in large applications,
9971 particularly when partial linking is used to build the application.
9973 The types of long calls used depends on the capabilities of the
9974 assembler and linker, and the type of code being generated. The
9975 impact on systems that support long absolute calls, and long pic
9976 symbol-difference or pc-relative calls should be relatively small.
9977 However, an indirect call is used on 32-bit ELF systems in pic code
9978 and it is quite long.
9981 Generate compiler predefines and select a startfile for the
9982 specified UNIX standard. The choices for UNIX-STD are `93', `95'
9983 and `98'. `93' is supported on all HP-UX versions. `95' is
9984 available on HP-UX 10.10 and later. `98' is available on HP-UX
9985 11.11 and later. The default values are `93' for HP-UX 10.00,
9986 `95' for HP-UX 10.10 though to 11.00, and `98' for HP-UX 11.11 and
9989 `-munix=93' provides the same predefines as GCC 3.3 and 3.4.
9990 `-munix=95' provides additional predefines for `XOPEN_UNIX' and
9991 `_XOPEN_SOURCE_EXTENDED', and the startfile `unix95.o'.
9992 `-munix=98' provides additional predefines for `_XOPEN_UNIX',
9993 `_XOPEN_SOURCE_EXTENDED', `_INCLUDE__STDC_A1_SOURCE' and
9994 `_INCLUDE_XOPEN_SOURCE_500', and the startfile `unix98.o'.
9996 It is _important_ to note that this option changes the interfaces
9997 for various library routines. It also affects the operational
9998 behavior of the C library. Thus, _extreme_ care is needed in
10001 Library code that is intended to operate with more than one UNIX
10002 standard must test, set and restore the variable
10003 __XPG4_EXTENDED_MASK as appropriate. Most GNU software doesn't
10004 provide this capability.
10007 Suppress the generation of link options to search libdld.sl when
10008 the `-static' option is specified on HP-UX 10 and later.
10011 The HP-UX implementation of setlocale in libc has a dependency on
10012 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10013 when the `-static' option is specified, special link options are
10014 needed to resolve this dependency.
10016 On HP-UX 10 and later, the GCC driver adds the necessary options to
10017 link with libdld.sl when the `-static' option is specified. This
10018 causes the resulting binary to be dynamic. On the 64-bit port,
10019 the linkers generate dynamic binaries by default in any case. The
10020 `-nolibdld' option can be used to prevent the GCC driver from
10021 adding these link options.
10024 Add support for multithreading with the "dce thread" library under
10025 HP-UX. This option sets flags for both the preprocessor and
10029 File: gcc.info, Node: i386 and x86-64 Options, Next: IA-64 Options, Prev: HPPA Options, Up: Submodel Options
10031 3.17.14 Intel 386 and AMD x86-64 Options
10032 ----------------------------------------
10034 These `-m' options are defined for the i386 and x86-64 family of
10038 Tune to CPU-TYPE everything applicable about the generated code,
10039 except for the ABI and the set of available instructions. The
10040 choices for CPU-TYPE are:
10042 Produce code optimized for the most common IA32/AMD64/EM64T
10043 processors. If you know the CPU on which your code will run,
10044 then you should use the corresponding `-mtune' option instead
10045 of `-mtune=generic'. But, if you do not know exactly what
10046 CPU users of your application will have, then you should use
10049 As new processors are deployed in the marketplace, the
10050 behavior of this option will change. Therefore, if you
10051 upgrade to a newer version of GCC, the code generated option
10052 will change to reflect the processors that were most common
10053 when that version of GCC was released.
10055 There is no `-march=generic' option because `-march'
10056 indicates the instruction set the compiler can use, and there
10057 is no generic instruction set applicable to all processors.
10058 In contrast, `-mtune' indicates the processor (or, in this
10059 case, collection of processors) for which the code is
10063 This selects the CPU to tune for at compilation time by
10064 determining the processor type of the compiling machine.
10065 Using `-mtune=native' will produce code optimized for the
10066 local machine under the constraints of the selected
10067 instruction set. Using `-march=native' will enable all
10068 instruction subsets supported by the local machine (hence the
10069 result might not run on different machines).
10072 Original Intel's i386 CPU.
10075 Intel's i486 CPU. (No scheduling is implemented for this
10079 Intel Pentium CPU with no MMX support.
10082 Intel PentiumMMX CPU based on Pentium core with MMX
10083 instruction set support.
10086 Intel PentiumPro CPU.
10089 Same as `generic', but when used as `march' option, PentiumPro
10090 instruction set will be used, so the code will run on all
10094 Intel Pentium2 CPU based on PentiumPro core with MMX
10095 instruction set support.
10097 _pentium3, pentium3m_
10098 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE
10099 instruction set support.
10102 Low power version of Intel Pentium3 CPU with MMX, SSE and
10103 SSE2 instruction set support. Used by Centrino notebooks.
10105 _pentium4, pentium4m_
10106 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set
10110 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2
10111 and SSE3 instruction set support.
10114 Improved version of Intel Pentium4 CPU with 64-bit
10115 extensions, MMX, SSE, SSE2 and SSE3 instruction set support.
10118 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3
10119 and SSSE3 instruction set support.
10122 AMD K6 CPU with MMX instruction set support.
10125 Improved versions of AMD K6 CPU with MMX and 3dNOW!
10126 instruction set support.
10128 _athlon, athlon-tbird_
10129 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE
10130 prefetch instructions support.
10132 _athlon-4, athlon-xp, athlon-mp_
10133 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and
10134 full SSE instruction set support.
10136 _k8, opteron, athlon64, athlon-fx_
10137 AMD K8 core based CPUs with x86-64 instruction set support.
10138 (This supersets MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and
10139 64-bit instruction set extensions.)
10141 _k8-sse3, opteron-sse3, athlon64-sse3_
10142 Improved versions of k8, opteron and athlon64 with SSE3
10143 instruction set support.
10145 _amdfam10, barcelona_
10146 AMD Family 10h core based CPUs with x86-64 instruction set
10147 support. (This supersets MMX, SSE, SSE2, SSE3, SSE4A,
10148 3dNOW!, enhanced 3dNOW!, ABM and 64-bit instruction set
10152 IDT Winchip C6 CPU, dealt in same way as i486 with additional
10153 MMX instruction set support.
10156 IDT Winchip2 CPU, dealt in same way as i486 with additional
10157 MMX and 3dNOW! instruction set support.
10160 Via C3 CPU with MMX and 3dNOW! instruction set support. (No
10161 scheduling is implemented for this chip.)
10164 Via C3-2 CPU with MMX and SSE instruction set support. (No
10165 scheduling is implemented for this chip.)
10168 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10170 While picking a specific CPU-TYPE will schedule things
10171 appropriately for that particular chip, the compiler will not
10172 generate any code that does not run on the i386 without the
10173 `-march=CPU-TYPE' option being used.
10176 Generate instructions for the machine type CPU-TYPE. The choices
10177 for CPU-TYPE are the same as for `-mtune'. Moreover, specifying
10178 `-march=CPU-TYPE' implies `-mtune=CPU-TYPE'.
10181 A deprecated synonym for `-mtune'.
10184 Generate floating point arithmetics for selected unit UNIT. The
10185 choices for UNIT are:
10188 Use the standard 387 floating point coprocessor present
10189 majority of chips and emulated otherwise. Code compiled with
10190 this option will run almost everywhere. The temporary
10191 results are computed in 80bit precision instead of precision
10192 specified by the type resulting in slightly different results
10193 compared to most of other chips. See `-ffloat-store' for
10194 more detailed description.
10196 This is the default choice for i386 compiler.
10199 Use scalar floating point instructions present in the SSE
10200 instruction set. This instruction set is supported by
10201 Pentium3 and newer chips, in the AMD line by Athlon-4,
10202 Athlon-xp and Athlon-mp chips. The earlier version of SSE
10203 instruction set supports only single precision arithmetics,
10204 thus the double and extended precision arithmetics is still
10205 done using 387. Later version, present only in Pentium4 and
10206 the future AMD x86-64 chips supports double precision
10209 For the i386 compiler, you need to use `-march=CPU-TYPE',
10210 `-msse' or `-msse2' switches to enable SSE extensions and
10211 make this option effective. For the x86-64 compiler, these
10212 extensions are enabled by default.
10214 The resulting code should be considerably faster in the
10215 majority of cases and avoid the numerical instability
10216 problems of 387 code, but may break some existing code that
10217 expects temporaries to be 80bit.
10219 This is the default choice for the x86-64 compiler.
10222 Attempt to utilize both instruction sets at once. This
10223 effectively double the amount of available registers and on
10224 chips with separate execution units for 387 and SSE the
10225 execution resources too. Use this option with care, as it is
10226 still experimental, because the GCC register allocator does
10227 not model separate functional units well resulting in
10228 instable performance.
10231 Output asm instructions using selected DIALECT. Supported choices
10232 are `intel' or `att' (the default one). Darwin does not support
10237 Control whether or not the compiler uses IEEE floating point
10238 comparisons. These handle correctly the case where the result of a
10239 comparison is unordered.
10242 Generate output containing library calls for floating point.
10243 *Warning:* the requisite libraries are not part of GCC. Normally
10244 the facilities of the machine's usual C compiler are used, but
10245 this can't be done directly in cross-compilation. You must make
10246 your own arrangements to provide suitable library functions for
10249 On machines where a function returns floating point results in the
10250 80387 register stack, some floating point opcodes may be emitted
10251 even if `-msoft-float' is used.
10253 `-mno-fp-ret-in-387'
10254 Do not use the FPU registers for return values of functions.
10256 The usual calling convention has functions return values of types
10257 `float' and `double' in an FPU register, even if there is no FPU.
10258 The idea is that the operating system should emulate an FPU.
10260 The option `-mno-fp-ret-in-387' causes such values to be returned
10261 in ordinary CPU registers instead.
10263 `-mno-fancy-math-387'
10264 Some 387 emulators do not support the `sin', `cos' and `sqrt'
10265 instructions for the 387. Specify this option to avoid generating
10266 those instructions. This option is the default on FreeBSD,
10267 OpenBSD and NetBSD. This option is overridden when `-march'
10268 indicates that the target cpu will always have an FPU and so the
10269 instruction will not need emulation. As of revision 2.6.1, these
10270 instructions are not generated unless you also use the
10271 `-funsafe-math-optimizations' switch.
10274 `-mno-align-double'
10275 Control whether GCC aligns `double', `long double', and `long
10276 long' variables on a two word boundary or a one word boundary.
10277 Aligning `double' variables on a two word boundary will produce
10278 code that runs somewhat faster on a `Pentium' at the expense of
10281 On x86-64, `-malign-double' is enabled by default.
10283 *Warning:* if you use the `-malign-double' switch, structures
10284 containing the above types will be aligned differently than the
10285 published application binary interface specifications for the 386
10286 and will not be binary compatible with structures in code compiled
10287 without that switch.
10289 `-m96bit-long-double'
10290 `-m128bit-long-double'
10291 These switches control the size of `long double' type. The i386
10292 application binary interface specifies the size to be 96 bits, so
10293 `-m96bit-long-double' is the default in 32 bit mode.
10295 Modern architectures (Pentium and newer) would prefer `long double'
10296 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10297 conforming to the ABI, this would not be possible. So specifying a
10298 `-m128bit-long-double' will align `long double' to a 16 byte
10299 boundary by padding the `long double' with an additional 32 bit
10302 In the x86-64 compiler, `-m128bit-long-double' is the default
10303 choice as its ABI specifies that `long double' is to be aligned on
10306 Notice that neither of these options enable any extra precision
10307 over the x87 standard of 80 bits for a `long double'.
10309 *Warning:* if you override the default value for your target ABI,
10310 the structures and arrays containing `long double' variables will
10311 change their size as well as function calling convention for
10312 function taking `long double' will be modified. Hence they will
10313 not be binary compatible with arrays or structures in code
10314 compiled without that switch.
10316 `-mmlarge-data-threshold=NUMBER'
10317 When `-mcmodel=medium' is specified, the data greater than
10318 THRESHOLD are placed in large data section. This value must be the
10319 same across all object linked into the binary and defaults to
10323 Use a different function-calling convention, in which functions
10324 that take a fixed number of arguments return with the `ret' NUM
10325 instruction, which pops their arguments while returning. This
10326 saves one instruction in the caller since there is no need to pop
10327 the arguments there.
10329 You can specify that an individual function is called with this
10330 calling sequence with the function attribute `stdcall'. You can
10331 also override the `-mrtd' option by using the function attribute
10332 `cdecl'. *Note Function Attributes::.
10334 *Warning:* this calling convention is incompatible with the one
10335 normally used on Unix, so you cannot use it if you need to call
10336 libraries compiled with the Unix compiler.
10338 Also, you must provide function prototypes for all functions that
10339 take variable numbers of arguments (including `printf'); otherwise
10340 incorrect code will be generated for calls to those functions.
10342 In addition, seriously incorrect code will result if you call a
10343 function with too many arguments. (Normally, extra arguments are
10344 harmlessly ignored.)
10347 Control how many registers are used to pass integer arguments. By
10348 default, no registers are used to pass arguments, and at most 3
10349 registers can be used. You can control this behavior for a
10350 specific function by using the function attribute `regparm'.
10351 *Note Function Attributes::.
10353 *Warning:* if you use this switch, and NUM is nonzero, then you
10354 must build all modules with the same value, including any
10355 libraries. This includes the system libraries and startup modules.
10358 Use SSE register passing conventions for float and double arguments
10359 and return values. You can control this behavior for a specific
10360 function by using the function attribute `sseregparm'. *Note
10361 Function Attributes::.
10363 *Warning:* if you use this switch then you must build all modules
10364 with the same value, including any libraries. This includes the
10365 system libraries and startup modules.
10370 Set 80387 floating-point precision to 32, 64 or 80 bits. When
10371 `-mpc32' is specified, the significands of results of
10372 floating-point operations are rounded to 24 bits (single
10373 precision); `-mpc64' rounds the the significands of results of
10374 floating-point operations to 53 bits (double precision) and
10375 `-mpc80' rounds the significands of results of floating-point
10376 operations to 64 bits (extended double precision), which is the
10377 default. When this option is used, floating-point operations in
10378 higher precisions are not available to the programmer without
10379 setting the FPU control word explicitly.
10381 Setting the rounding of floating-point operations to less than the
10382 default 80 bits can speed some programs by 2% or more. Note that
10383 some mathematical libraries assume that extended precision (80
10384 bit) floating-point operations are enabled by default; routines in
10385 such libraries could suffer significant loss of accuracy,
10386 typically through so-called "catastrophic cancellation", when this
10387 option is used to set the precision to less than extended
10391 Realign the stack at entry. On the Intel x86, the
10392 `-mstackrealign' option will generate an alternate prologue and
10393 epilogue that realigns the runtime stack. This supports mixing
10394 legacy codes that keep a 4-byte aligned stack with modern codes
10395 that keep a 16-byte stack for SSE compatibility. The alternate
10396 prologue and epilogue are slower and bigger than the regular ones,
10397 and the alternate prologue requires an extra scratch register;
10398 this lowers the number of registers available if used in
10399 conjunction with the `regparm' attribute. The `-mstackrealign'
10400 option is incompatible with the nested function prologue; this is
10401 considered a hard error. See also the attribute
10402 `force_align_arg_pointer', applicable to individual functions.
10404 `-mpreferred-stack-boundary=NUM'
10405 Attempt to keep the stack boundary aligned to a 2 raised to NUM
10406 byte boundary. If `-mpreferred-stack-boundary' is not specified,
10407 the default is 4 (16 bytes or 128 bits).
10409 On Pentium and PentiumPro, `double' and `long double' values
10410 should be aligned to an 8 byte boundary (see `-malign-double') or
10411 suffer significant run time performance penalties. On Pentium
10412 III, the Streaming SIMD Extension (SSE) data type `__m128' may not
10413 work properly if it is not 16 byte aligned.
10415 To ensure proper alignment of this values on the stack, the stack
10416 boundary must be as aligned as that required by any value stored
10417 on the stack. Further, every function must be generated such that
10418 it keeps the stack aligned. Thus calling a function compiled with
10419 a higher preferred stack boundary from a function compiled with a
10420 lower preferred stack boundary will most likely misalign the
10421 stack. It is recommended that libraries that use callbacks always
10422 use the default setting.
10424 This extra alignment does consume extra stack space, and generally
10425 increases code size. Code that is sensitive to stack space usage,
10426 such as embedded systems and operating system kernels, may want to
10427 reduce the preferred alignment to `-mpreferred-stack-boundary=2'.
10468 These switches enable or disable the use of instructions in the
10469 MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4A, SSE5, ABM or 3DNow!
10470 extended instruction sets. These extensions are also available as
10471 built-in functions: see *Note X86 Built-in Functions::, for
10472 details of the functions enabled and disabled by these switches.
10474 To have SSE/SSE2 instructions generated automatically from
10475 floating-point code (as opposed to 387 instructions), see
10478 These options will enable GCC to use these extended instructions in
10479 generated code, even without `-mfpmath=sse'. Applications which
10480 perform runtime CPU detection must compile separate files for each
10481 supported architecture, using the appropriate flags. In
10482 particular, the file containing the CPU detection code should be
10483 compiled without these options.
10486 This option instructs GCC to emit a `cld' instruction in the
10487 prologue of functions that use string instructions. String
10488 instructions depend on the DF flag to select between autoincrement
10489 or autodecrement mode. While the ABI specifies the DF flag to be
10490 cleared on function entry, some operating systems violate this
10491 specification by not clearing the DF flag in their exception
10492 dispatchers. The exception handler can be invoked with the DF flag
10493 set which leads to wrong direction mode, when string instructions
10494 are used. This option can be enabled by default on 32-bit x86
10495 targets by configuring GCC with the `--enable-cld' configure
10496 option. Generation of `cld' instructions can be suppressed with
10497 the `-mno-cld' compiler option in this case.
10500 This option will enable GCC to use CMPXCHG16B instruction in
10501 generated code. CMPXCHG16B allows for atomic operations on
10502 128-bit double quadword (or oword) data types. This is useful for
10503 high resolution counters that could be updated by multiple
10504 processors (or cores). This instruction is generated as part of
10505 atomic built-in functions: see *Note Atomic Builtins:: for details.
10508 This option will enable GCC to use SAHF instruction in generated
10509 64-bit code. Early Intel CPUs with Intel 64 lacked LAHF and SAHF
10510 instructions supported by AMD64 until introduction of Pentium 4 G1
10511 step in December 2005. LAHF and SAHF are load and store
10512 instructions, respectively, for certain status flags. In 64-bit
10513 mode, SAHF instruction is used to optimize `fmod', `drem' or
10514 `remainder' built-in functions: see *Note Other Builtins:: for
10518 This option will enable GCC to use RCPSS and RSQRTSS instructions
10519 (and their vectorized variants RCPPS and RSQRTPS) with an
10520 additional Newton-Rhapson step to increase precision instead of
10521 DIVSS and SQRTSS (and their vectorized variants) for single
10522 precision floating point arguments. These instructions are
10523 generated only when `-funsafe-math-optimizations' is enabled
10524 together with `-finite-math-only' and `-fno-trapping-math'. Note
10525 that while the throughput of the sequence is higher than the
10526 throughput of the non-reciprocal instruction, the precision of the
10527 sequence can be decreased by up to 2 ulp (i.e. the inverse of 1.0
10528 equals 0.99999994).
10531 Specifies the ABI type to use for vectorizing intrinsics using an
10532 external library. Supported types are `acml' for the AMD math
10533 core library style of interfacing. GCC will currently emit calls
10534 to `__vrd2_sin', `__vrd2_cos', `__vrd2_exp', `__vrd2_log',
10535 `__vrd2_log2', `__vrd2_log10', `__vrs4_sinf', `__vrs4_cosf',
10536 `__vrs4_expf', `__vrs4_logf', `__vrs4_log2f', `__vrs4_log10f' and
10537 `__vrs4_powf' when using this type and `-ftree-vectorize' is
10538 enabled. A ACML ABI compatible library will have to be specified
10543 Use PUSH operations to store outgoing parameters. This method is
10544 shorter and usually equally fast as method using SUB/MOV
10545 operations and is enabled by default. In some cases disabling it
10546 may improve performance because of improved scheduling and reduced
10549 `-maccumulate-outgoing-args'
10550 If enabled, the maximum amount of space required for outgoing
10551 arguments will be computed in the function prologue. This is
10552 faster on most modern CPUs because of reduced dependencies,
10553 improved scheduling and reduced stack usage when preferred stack
10554 boundary is not equal to 2. The drawback is a notable increase in
10555 code size. This switch implies `-mno-push-args'.
10558 Support thread-safe exception handling on `Mingw32'. Code that
10559 relies on thread-safe exception handling must compile and link all
10560 code with the `-mthreads' option. When compiling, `-mthreads'
10561 defines `-D_MT'; when linking, it links in a special thread helper
10562 library `-lmingwthrd' which cleans up per thread exception
10565 `-mno-align-stringops'
10566 Do not align destination of inlined string operations. This
10567 switch reduces code size and improves performance in case the
10568 destination is already aligned, but GCC doesn't know about it.
10570 `-minline-all-stringops'
10571 By default GCC inlines string operations only when destination is
10572 known to be aligned at least to 4 byte boundary. This enables
10573 more inlining, increase code size, but may improve performance of
10574 code that depends on fast memcpy, strlen and memset for short
10577 `-minline-stringops-dynamically'
10578 For string operation of unknown size, inline runtime checks so for
10579 small blocks inline code is used, while for large blocks library
10582 `-mstringop-strategy=ALG'
10583 Overwrite internal decision heuristic about particular algorithm
10584 to inline string operation with. The allowed values are
10585 `rep_byte', `rep_4byte', `rep_8byte' for expanding using i386
10586 `rep' prefix of specified size, `byte_loop', `loop',
10587 `unrolled_loop' for expanding inline loop, `libcall' for always
10588 expanding library call.
10590 `-momit-leaf-frame-pointer'
10591 Don't keep the frame pointer in a register for leaf functions.
10592 This avoids the instructions to save, set up and restore frame
10593 pointers and makes an extra register available in leaf functions.
10594 The option `-fomit-frame-pointer' removes the frame pointer for
10595 all functions which might make debugging harder.
10597 `-mtls-direct-seg-refs'
10598 `-mno-tls-direct-seg-refs'
10599 Controls whether TLS variables may be accessed with offsets from
10600 the TLS segment register (`%gs' for 32-bit, `%fs' for 64-bit), or
10601 whether the thread base pointer must be added. Whether or not this
10602 is legal depends on the operating system, and whether it maps the
10603 segment to cover the entire TLS area.
10605 For systems that use GNU libc, the default is on.
10609 Enable automatic generation of fused floating point multiply-add
10610 instructions if the ISA supports such instructions. The
10611 -mfused-madd option is on by default. The fused multiply-add
10612 instructions have a different rounding behavior compared to
10613 executing a multiply followed by an add.
10615 These `-m' switches are supported in addition to the above on AMD
10616 x86-64 processors in 64-bit environments.
10620 Generate code for a 32-bit or 64-bit environment. The 32-bit
10621 environment sets int, long and pointer to 32 bits and generates
10622 code that runs on any i386 system. The 64-bit environment sets
10623 int to 32 bits and long and pointer to 64 bits and generates code
10624 for AMD's x86-64 architecture. For darwin only the -m64 option
10625 turns off the `-fno-pic' and `-mdynamic-no-pic' options.
10628 Do not use a so called red zone for x86-64 code. The red zone is
10629 mandated by the x86-64 ABI, it is a 128-byte area beyond the
10630 location of the stack pointer that will not be modified by signal
10631 or interrupt handlers and therefore can be used for temporary data
10632 without adjusting the stack pointer. The flag `-mno-red-zone'
10633 disables this red zone.
10636 Generate code for the small code model: the program and its
10637 symbols must be linked in the lower 2 GB of the address space.
10638 Pointers are 64 bits. Programs can be statically or dynamically
10639 linked. This is the default code model.
10642 Generate code for the kernel code model. The kernel runs in the
10643 negative 2 GB of the address space. This model has to be used for
10647 Generate code for the medium model: The program is linked in the
10648 lower 2 GB of the address space but symbols can be located
10649 anywhere in the address space. Programs can be statically or
10650 dynamically linked, but building of shared libraries are not
10651 supported with the medium model.
10654 Generate code for the large model: This model makes no assumptions
10655 about addresses and sizes of sections.
10658 File: gcc.info, Node: IA-64 Options, Next: M32C Options, Prev: i386 and x86-64 Options, Up: Submodel Options
10660 3.17.15 IA-64 Options
10661 ---------------------
10663 These are the `-m' options defined for the Intel IA-64 architecture.
10666 Generate code for a big endian target. This is the default for
10670 Generate code for a little endian target. This is the default for
10671 AIX5 and GNU/Linux.
10675 Generate (or don't) code for the GNU assembler. This is the
10680 Generate (or don't) code for the GNU linker. This is the default.
10683 Generate code that does not use a global pointer register. The
10684 result is not position independent code, and violates the IA-64
10687 `-mvolatile-asm-stop'
10688 `-mno-volatile-asm-stop'
10689 Generate (or don't) a stop bit immediately before and after
10690 volatile asm statements.
10693 `-mno-register-names'
10694 Generate (or don't) `in', `loc', and `out' register names for the
10695 stacked registers. This may make assembler output more readable.
10699 Disable (or enable) optimizations that use the small data section.
10700 This may be useful for working around optimizer bugs.
10703 Generate code that uses a single constant global pointer value.
10704 This is useful when compiling kernel code.
10707 Generate code that is self-relocatable. This implies
10708 `-mconstant-gp'. This is useful when compiling firmware code.
10710 `-minline-float-divide-min-latency'
10711 Generate code for inline divides of floating point values using
10712 the minimum latency algorithm.
10714 `-minline-float-divide-max-throughput'
10715 Generate code for inline divides of floating point values using
10716 the maximum throughput algorithm.
10718 `-minline-int-divide-min-latency'
10719 Generate code for inline divides of integer values using the
10720 minimum latency algorithm.
10722 `-minline-int-divide-max-throughput'
10723 Generate code for inline divides of integer values using the
10724 maximum throughput algorithm.
10726 `-minline-sqrt-min-latency'
10727 Generate code for inline square roots using the minimum latency
10730 `-minline-sqrt-max-throughput'
10731 Generate code for inline square roots using the maximum throughput
10736 Don't (or do) generate assembler code for the DWARF2 line number
10737 debugging info. This may be useful when not using the GNU
10740 `-mearly-stop-bits'
10741 `-mno-early-stop-bits'
10742 Allow stop bits to be placed earlier than immediately preceding the
10743 instruction that triggered the stop bit. This can improve
10744 instruction scheduling, but does not always do so.
10746 `-mfixed-range=REGISTER-RANGE'
10747 Generate code treating the given register range as fixed registers.
10748 A fixed register is one that the register allocator can not use.
10749 This is useful when compiling kernel code. A register range is
10750 specified as two registers separated by a dash. Multiple register
10751 ranges can be specified separated by a comma.
10753 `-mtls-size=TLS-SIZE'
10754 Specify bit size of immediate TLS offsets. Valid values are 14,
10758 Tune the instruction scheduling for a particular CPU, Valid values
10759 are itanium, itanium1, merced, itanium2, and mckinley.
10763 Add support for multithreading using the POSIX threads library.
10764 This option sets flags for both the preprocessor and linker. It
10765 does not affect the thread safety of object code produced by the
10766 compiler or that of libraries supplied with it. These are HP-UX
10771 Generate code for a 32-bit or 64-bit environment. The 32-bit
10772 environment sets int, long and pointer to 32 bits. The 64-bit
10773 environment sets int to 32 bits and long and pointer to 64 bits.
10774 These are HP-UX specific flags.
10776 `-mno-sched-br-data-spec'
10777 `-msched-br-data-spec'
10778 (Dis/En)able data speculative scheduling before reload. This will
10779 result in generation of the ld.a instructions and the
10780 corresponding check instructions (ld.c / chk.a). The default is
10783 `-msched-ar-data-spec'
10784 `-mno-sched-ar-data-spec'
10785 (En/Dis)able data speculative scheduling after reload. This will
10786 result in generation of the ld.a instructions and the
10787 corresponding check instructions (ld.c / chk.a). The default is
10790 `-mno-sched-control-spec'
10791 `-msched-control-spec'
10792 (Dis/En)able control speculative scheduling. This feature is
10793 available only during region scheduling (i.e. before reload).
10794 This will result in generation of the ld.s instructions and the
10795 corresponding check instructions chk.s . The default is 'disable'.
10797 `-msched-br-in-data-spec'
10798 `-mno-sched-br-in-data-spec'
10799 (En/Dis)able speculative scheduling of the instructions that are
10800 dependent on the data speculative loads before reload. This is
10801 effective only with `-msched-br-data-spec' enabled. The default
10804 `-msched-ar-in-data-spec'
10805 `-mno-sched-ar-in-data-spec'
10806 (En/Dis)able speculative scheduling of the instructions that are
10807 dependent on the data speculative loads after reload. This is
10808 effective only with `-msched-ar-data-spec' enabled. The default
10811 `-msched-in-control-spec'
10812 `-mno-sched-in-control-spec'
10813 (En/Dis)able speculative scheduling of the instructions that are
10814 dependent on the control speculative loads. This is effective
10815 only with `-msched-control-spec' enabled. The default is 'enable'.
10819 (En/Dis)able use of simple data speculation checks ld.c . If
10820 disabled, only chk.a instructions will be emitted to check data
10821 speculative loads. The default is 'enable'.
10823 `-mno-sched-control-ldc'
10824 `-msched-control-ldc'
10825 (Dis/En)able use of ld.c instructions to check control speculative
10826 loads. If enabled, in case of control speculative load with no
10827 speculatively scheduled dependent instructions this load will be
10828 emitted as ld.sa and ld.c will be used to check it. The default
10831 `-mno-sched-spec-verbose'
10832 `-msched-spec-verbose'
10833 (Dis/En)able printing of the information about speculative motions.
10835 `-mno-sched-prefer-non-data-spec-insns'
10836 `-msched-prefer-non-data-spec-insns'
10837 If enabled, data speculative instructions will be chosen for
10838 schedule only if there are no other choices at the moment. This
10839 will make the use of the data speculation much more conservative.
10840 The default is 'disable'.
10842 `-mno-sched-prefer-non-control-spec-insns'
10843 `-msched-prefer-non-control-spec-insns'
10844 If enabled, control speculative instructions will be chosen for
10845 schedule only if there are no other choices at the moment. This
10846 will make the use of the control speculation much more
10847 conservative. The default is 'disable'.
10849 `-mno-sched-count-spec-in-critical-path'
10850 `-msched-count-spec-in-critical-path'
10851 If enabled, speculative dependencies will be considered during
10852 computation of the instructions priorities. This will make the
10853 use of the speculation a bit more conservative. The default is
10858 File: gcc.info, Node: M32C Options, Next: M32R/D Options, Prev: IA-64 Options, Up: Submodel Options
10860 3.17.16 M32C Options
10861 --------------------
10864 Select the CPU for which code is generated. NAME may be one of
10865 `r8c' for the R8C/Tiny series, `m16c' for the M16C (up to /60)
10866 series, `m32cm' for the M16C/80 series, or `m32c' for the M32C/80
10870 Specifies that the program will be run on the simulator. This
10871 causes an alternate runtime library to be linked in which
10872 supports, for example, file I/O. You must not use this option
10873 when generating programs that will run on real hardware; you must
10874 provide your own runtime library for whatever I/O functions are
10878 Specifies the number of memory-based pseudo-registers GCC will use
10879 during code generation. These pseudo-registers will be used like
10880 real registers, so there is a tradeoff between GCC's ability to
10881 fit the code into available registers, and the performance penalty
10882 of using memory instead of registers. Note that all modules in a
10883 program must be compiled with the same value for this option.
10884 Because of that, you must not use this option with the default
10885 runtime libraries gcc builds.
10889 File: gcc.info, Node: M32R/D Options, Next: M680x0 Options, Prev: M32C Options, Up: Submodel Options
10891 3.17.17 M32R/D Options
10892 ----------------------
10894 These `-m' options are defined for Renesas M32R/D architectures:
10897 Generate code for the M32R/2.
10900 Generate code for the M32R/X.
10903 Generate code for the M32R. This is the default.
10906 Assume all objects live in the lower 16MB of memory (so that their
10907 addresses can be loaded with the `ld24' instruction), and assume
10908 all subroutines are reachable with the `bl' instruction. This is
10911 The addressability of a particular object can be set with the
10915 Assume objects may be anywhere in the 32-bit address space (the
10916 compiler will generate `seth/add3' instructions to load their
10917 addresses), and assume all subroutines are reachable with the `bl'
10921 Assume objects may be anywhere in the 32-bit address space (the
10922 compiler will generate `seth/add3' instructions to load their
10923 addresses), and assume subroutines may not be reachable with the
10924 `bl' instruction (the compiler will generate the much slower
10925 `seth/add3/jl' instruction sequence).
10928 Disable use of the small data area. Variables will be put into
10929 one of `.data', `bss', or `.rodata' (unless the `section'
10930 attribute has been specified). This is the default.
10932 The small data area consists of sections `.sdata' and `.sbss'.
10933 Objects may be explicitly put in the small data area with the
10934 `section' attribute using one of these sections.
10937 Put small global and static data in the small data area, but do not
10938 generate special code to reference them.
10941 Put small global and static data in the small data area, and
10942 generate special instructions to reference them.
10945 Put global and static objects less than or equal to NUM bytes into
10946 the small data or bss sections instead of the normal data or bss
10947 sections. The default value of NUM is 8. The `-msdata' option
10948 must be set to one of `sdata' or `use' for this option to have any
10951 All modules should be compiled with the same `-G NUM' value.
10952 Compiling with different values of NUM may or may not work; if it
10953 doesn't the linker will give an error message--incorrect code will
10957 Makes the M32R specific code in the compiler display some
10958 statistics that might help in debugging programs.
10961 Align all loops to a 32-byte boundary.
10964 Do not enforce a 32-byte alignment for loops. This is the default.
10966 `-missue-rate=NUMBER'
10967 Issue NUMBER instructions per cycle. NUMBER can only be 1 or 2.
10969 `-mbranch-cost=NUMBER'
10970 NUMBER can only be 1 or 2. If it is 1 then branches will be
10971 preferred over conditional code, if it is 2, then the opposite will
10974 `-mflush-trap=NUMBER'
10975 Specifies the trap number to use to flush the cache. The default
10976 is 12. Valid numbers are between 0 and 15 inclusive.
10979 Specifies that the cache cannot be flushed by using a trap.
10981 `-mflush-func=NAME'
10982 Specifies the name of the operating system function to call to
10983 flush the cache. The default is __flush_cache_, but a function
10984 call will only be used if a trap is not available.
10987 Indicates that there is no OS function for flushing the cache.
10991 File: gcc.info, Node: M680x0 Options, Next: M68hc1x Options, Prev: M32R/D Options, Up: Submodel Options
10993 3.17.18 M680x0 Options
10994 ----------------------
10996 These are the `-m' options defined for M680x0 and ColdFire processors.
10997 The default settings depend on which architecture was selected when the
10998 compiler was configured; the defaults for the most common choices are
11002 Generate code for a specific M680x0 or ColdFire instruction set
11003 architecture. Permissible values of ARCH for M680x0 architectures
11004 are: `68000', `68010', `68020', `68030', `68040', `68060' and
11005 `cpu32'. ColdFire architectures are selected according to
11006 Freescale's ISA classification and the permissible values are:
11007 `isaa', `isaaplus', `isab' and `isac'.
11009 gcc defines a macro `__mcfARCH__' whenever it is generating code
11010 for a ColdFire target. The ARCH in this macro is one of the
11011 `-march' arguments given above.
11013 When used together, `-march' and `-mtune' select code that runs on
11014 a family of similar processors but that is optimized for a
11015 particular microarchitecture.
11018 Generate code for a specific M680x0 or ColdFire processor. The
11019 M680x0 CPUs are: `68000', `68010', `68020', `68030', `68040',
11020 `68060', `68302', `68332' and `cpu32'. The ColdFire CPUs are
11021 given by the table below, which also classifies the CPUs into
11024 *Family* *`-mcpu' arguments*
11026 `5206' `5202' `5204' `5206'
11028 `5208' `5207' `5208'
11029 `5211a' `5210a' `5211a'
11030 `5213' `5211' `5212' `5213'
11031 `5216' `5214' `5216'
11032 `52235' `52230' `52231' `52232' `52233' `52234' `52235'
11033 `5225' `5224' `5225'
11034 `5235' `5232' `5233' `5234' `5235' `523x'
11037 `5271' `5270' `5271'
11039 `5275' `5274' `5275'
11040 `5282' `5280' `5281' `5282' `528x'
11042 `5329' `5327' `5328' `5329' `532x'
11043 `5373' `5372' `5373' `537x'
11045 `5475' `5470' `5471' `5472' `5473' `5474' `5475' `547x'
11046 `5480' `5481' `5482' `5483' `5484' `5485'
11048 `-mcpu=CPU' overrides `-march=ARCH' if ARCH is compatible with
11049 CPU. Other combinations of `-mcpu' and `-march' are rejected.
11051 gcc defines the macro `__mcf_cpu_CPU' when ColdFire target CPU is
11052 selected. It also defines `__mcf_family_FAMILY', where the value
11053 of FAMILY is given by the table above.
11056 Tune the code for a particular microarchitecture, within the
11057 constraints set by `-march' and `-mcpu'. The M680x0
11058 microarchitectures are: `68000', `68010', `68020', `68030',
11059 `68040', `68060' and `cpu32'. The ColdFire microarchitectures
11060 are: `cfv1', `cfv2', `cfv3', `cfv4' and `cfv4e'.
11062 You can also use `-mtune=68020-40' for code that needs to run
11063 relatively well on 68020, 68030 and 68040 targets.
11064 `-mtune=68020-60' is similar but includes 68060 targets as well.
11065 These two options select the same tuning decisions as `-m68020-40'
11066 and `-m68020-60' respectively.
11068 gcc defines the macros `__mcARCH' and `__mcARCH__' when tuning for
11069 680x0 architecture ARCH. It also defines `mcARCH' unless either
11070 `-ansi' or a non-GNU `-std' option is used. If gcc is tuning for
11071 a range of architectures, as selected by `-mtune=68020-40' or
11072 `-mtune=68020-60', it defines the macros for every architecture in
11075 gcc also defines the macro `__mUARCH__' when tuning for ColdFire
11076 microarchitecture UARCH, where UARCH is one of the arguments given
11081 Generate output for a 68000. This is the default when the
11082 compiler is configured for 68000-based systems. It is equivalent
11085 Use this option for microcontrollers with a 68000 or EC000 core,
11086 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11089 Generate output for a 68010. This is the default when the
11090 compiler is configured for 68010-based systems. It is equivalent
11095 Generate output for a 68020. This is the default when the
11096 compiler is configured for 68020-based systems. It is equivalent
11100 Generate output for a 68030. This is the default when the
11101 compiler is configured for 68030-based systems. It is equivalent
11105 Generate output for a 68040. This is the default when the
11106 compiler is configured for 68040-based systems. It is equivalent
11109 This option inhibits the use of 68881/68882 instructions that have
11110 to be emulated by software on the 68040. Use this option if your
11111 68040 does not have code to emulate those instructions.
11114 Generate output for a 68060. This is the default when the
11115 compiler is configured for 68060-based systems. It is equivalent
11118 This option inhibits the use of 68020 and 68881/68882 instructions
11119 that have to be emulated by software on the 68060. Use this
11120 option if your 68060 does not have code to emulate those
11124 Generate output for a CPU32. This is the default when the
11125 compiler is configured for CPU32-based systems. It is equivalent
11128 Use this option for microcontrollers with a CPU32 or CPU32+ core,
11129 including the 68330, 68331, 68332, 68333, 68334, 68336, 68340,
11130 68341, 68349 and 68360.
11133 Generate output for a 520X ColdFire CPU. This is the default when
11134 the compiler is configured for 520X-based systems. It is
11135 equivalent to `-mcpu=5206', and is now deprecated in favor of that
11138 Use this option for microcontroller with a 5200 core, including
11139 the MCF5202, MCF5203, MCF5204 and MCF5206.
11142 Generate output for a 5206e ColdFire CPU. The option is now
11143 deprecated in favor of the equivalent `-mcpu=5206e'.
11146 Generate output for a member of the ColdFire 528X family. The
11147 option is now deprecated in favor of the equivalent `-mcpu=528x'.
11150 Generate output for a ColdFire 5307 CPU. The option is now
11151 deprecated in favor of the equivalent `-mcpu=5307'.
11154 Generate output for a ColdFire 5407 CPU. The option is now
11155 deprecated in favor of the equivalent `-mcpu=5407'.
11158 Generate output for a ColdFire V4e family CPU (e.g. 547x/548x).
11159 This includes use of hardware floating point instructions. The
11160 option is equivalent to `-mcpu=547x', and is now deprecated in
11161 favor of that option.
11164 Generate output for a 68040, without using any of the new
11165 instructions. This results in code which can run relatively
11166 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11167 generated code does use the 68881 instructions that are emulated
11170 The option is equivalent to `-march=68020' `-mtune=68020-40'.
11173 Generate output for a 68060, without using any of the new
11174 instructions. This results in code which can run relatively
11175 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11176 generated code does use the 68881 instructions that are emulated
11179 The option is equivalent to `-march=68020' `-mtune=68020-60'.
11183 Generate floating-point instructions. This is the default for
11184 68020 and above, and for ColdFire devices that have an FPU. It
11185 defines the macro `__HAVE_68881__' on M680x0 targets and
11186 `__mcffpu__' on ColdFire targets.
11189 Do not generate floating-point instructions; use library calls
11190 instead. This is the default for 68000, 68010, and 68832 targets.
11191 It is also the default for ColdFire devices that have no FPU.
11195 Generate (do not generate) ColdFire hardware divide and remainder
11196 instructions. If `-march' is used without `-mcpu', the default is
11197 "on" for ColdFire architectures and "off" for M680x0
11198 architectures. Otherwise, the default is taken from the target CPU
11199 (either the default CPU, or the one specified by `-mcpu'). For
11200 example, the default is "off" for `-mcpu=5206' and "on" for
11203 gcc defines the macro `__mcfhwdiv__' when this option is enabled.
11206 Consider type `int' to be 16 bits wide, like `short int'.
11207 Additionally, parameters passed on the stack are also aligned to a
11208 16-bit boundary even on targets whose API mandates promotion to
11212 Do not consider type `int' to be 16 bits wide. This is the
11217 Do not use the bit-field instructions. The `-m68000', `-mcpu32'
11218 and `-m5200' options imply `-mnobitfield'.
11221 Do use the bit-field instructions. The `-m68020' option implies
11222 `-mbitfield'. This is the default if you use a configuration
11223 designed for a 68020.
11226 Use a different function-calling convention, in which functions
11227 that take a fixed number of arguments return with the `rtd'
11228 instruction, which pops their arguments while returning. This
11229 saves one instruction in the caller since there is no need to pop
11230 the arguments there.
11232 This calling convention is incompatible with the one normally used
11233 on Unix, so you cannot use it if you need to call libraries
11234 compiled with the Unix compiler.
11236 Also, you must provide function prototypes for all functions that
11237 take variable numbers of arguments (including `printf'); otherwise
11238 incorrect code will be generated for calls to those functions.
11240 In addition, seriously incorrect code will result if you call a
11241 function with too many arguments. (Normally, extra arguments are
11242 harmlessly ignored.)
11244 The `rtd' instruction is supported by the 68010, 68020, 68030,
11245 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11248 Do not use the calling conventions selected by `-mrtd'. This is
11253 Control whether GCC aligns `int', `long', `long long', `float',
11254 `double', and `long double' variables on a 32-bit boundary
11255 (`-malign-int') or a 16-bit boundary (`-mno-align-int'). Aligning
11256 variables on 32-bit boundaries produces code that runs somewhat
11257 faster on processors with 32-bit busses at the expense of more
11260 *Warning:* if you use the `-malign-int' switch, GCC will align
11261 structures containing the above types differently than most
11262 published application binary interface specifications for the m68k.
11265 Use the pc-relative addressing mode of the 68000 directly, instead
11266 of using a global offset table. At present, this option implies
11267 `-fpic', allowing at most a 16-bit offset for pc-relative
11268 addressing. `-fPIC' is not presently supported with `-mpcrel',
11269 though this could be supported for 68020 and higher processors.
11271 `-mno-strict-align'
11273 Do not (do) assume that unaligned memory references will be
11274 handled by the system.
11277 Generate code that allows the data segment to be located in a
11278 different area of memory from the text segment. This allows for
11279 execute in place in an environment without virtual memory
11280 management. This option implies `-fPIC'.
11283 Generate code that assumes that the data segment follows the text
11284 segment. This is the default.
11286 `-mid-shared-library'
11287 Generate code that supports shared libraries via the library ID
11288 method. This allows for execute in place and shared libraries in
11289 an environment without virtual memory management. This option
11292 `-mno-id-shared-library'
11293 Generate code that doesn't assume ID based shared libraries are
11294 being used. This is the default.
11296 `-mshared-library-id=n'
11297 Specified the identification number of the ID based shared library
11298 being compiled. Specifying a value of 0 will generate more
11299 compact code, specifying other values will force the allocation of
11300 that number to the current library but is no more space or time
11301 efficient than omitting this option.
11305 File: gcc.info, Node: M68hc1x Options, Next: MCore Options, Prev: M680x0 Options, Up: Submodel Options
11307 3.17.19 M68hc1x Options
11308 -----------------------
11310 These are the `-m' options defined for the 68hc11 and 68hc12
11311 microcontrollers. The default values for these options depends on
11312 which style of microcontroller was selected when the compiler was
11313 configured; the defaults for the most common choices are given below.
11317 Generate output for a 68HC11. This is the default when the
11318 compiler is configured for 68HC11-based systems.
11322 Generate output for a 68HC12. This is the default when the
11323 compiler is configured for 68HC12-based systems.
11327 Generate output for a 68HCS12.
11330 Enable the use of 68HC12 pre and post auto-increment and
11331 auto-decrement addressing modes.
11335 Enable the use of 68HC12 min and max instructions.
11339 Treat all calls as being far away (near). If calls are assumed to
11340 be far away, the compiler will use the `call' instruction to call
11341 a function and the `rtc' instruction for returning.
11344 Consider type `int' to be 16 bits wide, like `short int'.
11346 `-msoft-reg-count=COUNT'
11347 Specify the number of pseudo-soft registers which are used for the
11348 code generation. The maximum number is 32. Using more pseudo-soft
11349 register may or may not result in better code depending on the
11350 program. The default is 4 for 68HC11 and 2 for 68HC12.
11354 File: gcc.info, Node: MCore Options, Next: MIPS Options, Prev: M68hc1x Options, Up: Submodel Options
11356 3.17.20 MCore Options
11357 ---------------------
11359 These are the `-m' options defined for the Motorola M*Core processors.
11363 Inline constants into the code stream if it can be done in two
11364 instructions or less.
11368 Use the divide instruction. (Enabled by default).
11370 `-mrelax-immediate'
11371 `-mno-relax-immediate'
11372 Allow arbitrary sized immediates in bit operations.
11375 `-mno-wide-bitfields'
11376 Always treat bit-fields as int-sized.
11378 `-m4byte-functions'
11379 `-mno-4byte-functions'
11380 Force all functions to be aligned to a four byte boundary.
11383 `-mno-callgraph-data'
11384 Emit callgraph information.
11388 Prefer word access when reading byte quantities.
11392 Generate code for a little endian target.
11396 Generate code for the 210 processor.
11399 File: gcc.info, Node: MIPS Options, Next: MMIX Options, Prev: MCore Options, Up: Submodel Options
11401 3.17.21 MIPS Options
11402 --------------------
11405 Generate big-endian code.
11408 Generate little-endian code. This is the default for `mips*el-*-*'
11412 Generate code that will run on ARCH, which can be the name of a
11413 generic MIPS ISA, or the name of a particular processor. The ISA
11414 names are: `mips1', `mips2', `mips3', `mips4', `mips32',
11415 `mips32r2', and `mips64'. The processor names are: `4kc', `4km',
11416 `4kp', `4ksc', `4kec', `4kem', `4kep', `4ksd', `5kc', `5kf',
11417 `20kc', `24kc', `24kf2_1', `24kf1_1', `24kec', `24kef2_1',
11418 `24kef1_1', `34kc', `34kf2_1', `34kf1_1', `74kc', `74kf2_1',
11419 `74kf1_1', `74kf3_2', `m4k', `orion', `r2000', `r3000', `r3900',
11420 `r4000', `r4400', `r4600', `r4650', `r6000', `r8000', `rm7000',
11421 `rm9000', `sb1', `sr71000', `vr4100', `vr4111', `vr4120',
11422 `vr4130', `vr4300', `vr5000', `vr5400' and `vr5500'. The special
11423 value `from-abi' selects the most compatible architecture for the
11424 selected ABI (that is, `mips1' for 32-bit ABIs and `mips3' for
11427 In processor names, a final `000' can be abbreviated as `k' (for
11428 example, `-march=r2k'). Prefixes are optional, and `vr' may be
11431 Names of the form `Nf2_1' refer to processors with FPUs clocked at
11432 half the rate of the core, names of the form `Nf1_1' refer to
11433 processors with FPUs clocked at the same rate as the core, and
11434 names of the form `Nf3_2' refer to processors with FPUs clocked a
11435 ratio of 3:2 with respect to the core. For compatibility reasons,
11436 `Nf' is accepted as a synonym for `Nf2_1' while `Nx' and `Bfx' are
11437 accepted as synonyms for `Nf1_1'.
11439 GCC defines two macros based on the value of this option. The
11440 first is `_MIPS_ARCH', which gives the name of target
11441 architecture, as a string. The second has the form
11442 `_MIPS_ARCH_FOO', where FOO is the capitalized value of
11443 `_MIPS_ARCH'. For example, `-march=r2000' will set `_MIPS_ARCH'
11444 to `"r2000"' and define the macro `_MIPS_ARCH_R2000'.
11446 Note that the `_MIPS_ARCH' macro uses the processor names given
11447 above. In other words, it will have the full prefix and will not
11448 abbreviate `000' as `k'. In the case of `from-abi', the macro
11449 names the resolved architecture (either `"mips1"' or `"mips3"').
11450 It names the default architecture when no `-march' option is given.
11453 Optimize for ARCH. Among other things, this option controls the
11454 way instructions are scheduled, and the perceived cost of
11455 arithmetic operations. The list of ARCH values is the same as for
11458 When this option is not used, GCC will optimize for the processor
11459 specified by `-march'. By using `-march' and `-mtune' together,
11460 it is possible to generate code that will run on a family of
11461 processors, but optimize the code for one particular member of
11464 `-mtune' defines the macros `_MIPS_TUNE' and `_MIPS_TUNE_FOO',
11465 which work in the same way as the `-march' ones described above.
11468 Equivalent to `-march=mips1'.
11471 Equivalent to `-march=mips2'.
11474 Equivalent to `-march=mips3'.
11477 Equivalent to `-march=mips4'.
11480 Equivalent to `-march=mips32'.
11483 Equivalent to `-march=mips32r2'.
11486 Equivalent to `-march=mips64'.
11490 Generate (do not generate) MIPS16 code. If GCC is targetting a
11491 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE.
11493 MIPS16 code generation can also be controlled on a per-function
11494 basis by means of `mips16' and `nomips16' attributes. *Note
11495 Function Attributes::, for more information.
11498 Generate MIPS16 code on alternating functions. This option is
11499 provided for regression testing of mixed MIPS16/non-MIPS16 code
11500 generation, and is not intended for ordinary use in compiling user
11503 `-minterlink-mips16'
11504 `-mno-interlink-mips16'
11505 Require (do not require) that non-MIPS16 code be link-compatible
11508 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
11509 it must either use a call or an indirect jump.
11510 `-minterlink-mips16' therefore disables direct jumps unless GCC
11511 knows that the target of the jump is not MIPS16.
11518 Generate code for the given ABI.
11520 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11521 generates 64-bit code when you select a 64-bit architecture, but
11522 you can use `-mgp32' to get 32-bit code instead.
11524 For information about the O64 ABI, see
11525 `http://gcc.gnu.org/projects/mipso64-abi.html'.
11527 GCC supports a variant of the o32 ABI in which floating-point
11528 registers are 64 rather than 32 bits wide. You can select this
11529 combination with `-mabi=32' `-mfp64'. This ABI relies on the
11530 `mthc1' and `mfhc1' instructions and is therefore only supported
11531 for MIPS32R2 processors.
11533 The register assignments for arguments and return values remain the
11534 same, but each scalar value is passed in a single 64-bit register
11535 rather than a pair of 32-bit registers. For example, scalar
11536 floating-point values are returned in `$f0' only, not a
11537 `$f0'/`$f1' pair. The set of call-saved registers also remains
11538 the same, but all 64 bits are saved.
11542 Generate (do not generate) code that is suitable for SVR4-style
11543 dynamic objects. `-mabicalls' is the default for SVR4-based
11548 Generate (do not generate) code that is fully position-independent,
11549 and that can therefore be linked into shared libraries. This
11550 option only affects `-mabicalls'.
11552 All `-mabicalls' code has traditionally been position-independent,
11553 regardless of options like `-fPIC' and `-fpic'. However, as an
11554 extension, the GNU toolchain allows executables to use absolute
11555 accesses for locally-binding symbols. It can also use shorter GP
11556 initialization sequences and generate direct calls to
11557 locally-defined functions. This mode is selected by `-mno-shared'.
11559 `-mno-shared' depends on binutils 2.16 or higher and generates
11560 objects that can only be linked by the GNU linker. However, the
11561 option does not affect the ABI of the final executable; it only
11562 affects the ABI of relocatable objects. Using `-mno-shared' will
11563 generally make executables both smaller and quicker.
11565 `-mshared' is the default.
11569 Lift (do not lift) the usual restrictions on the size of the global
11572 GCC normally uses a single instruction to load values from the GOT.
11573 While this is relatively efficient, it will only work if the GOT
11574 is smaller than about 64k. Anything larger will cause the linker
11575 to report an error such as:
11577 relocation truncated to fit: R_MIPS_GOT16 foobar
11579 If this happens, you should recompile your code with `-mxgot'. It
11580 should then work with very large GOTs, although it will also be
11581 less efficient, since it will take three instructions to fetch the
11582 value of a global symbol.
11584 Note that some linkers can create multiple GOTs. If you have such
11585 a linker, you should only need to use `-mxgot' when a single object
11586 file accesses more than 64k's worth of GOT entries. Very few do.
11588 These options have no effect unless GCC is generating position
11592 Assume that general-purpose registers are 32 bits wide.
11595 Assume that general-purpose registers are 64 bits wide.
11598 Assume that floating-point registers are 32 bits wide.
11601 Assume that floating-point registers are 64 bits wide.
11604 Use floating-point coprocessor instructions.
11607 Do not use floating-point coprocessor instructions. Implement
11608 floating-point calculations using library calls instead.
11611 Assume that the floating-point coprocessor only supports
11612 single-precision operations.
11615 Assume that the floating-point coprocessor supports
11616 double-precision operations. This is the default.
11620 Use (do not use) `ll', `sc', and `sync' instructions to implement
11621 atomic memory built-in functions. When neither option is
11622 specified, GCC will use the instructions if the target architecture
11625 `-mllsc' is useful if the runtime environment can emulate the
11626 instructions and `-mno-llsc' can be useful when compiling for
11627 nonstandard ISAs. You can make either option the default by
11628 configuring GCC with `--with-llsc' and `--without-llsc'
11629 respectively. `--with-llsc' is the default for some
11630 configurations; see the installation documentation for details.
11634 Use (do not use) revision 1 of the MIPS DSP ASE. *Note MIPS DSP
11635 Built-in Functions::. This option defines the preprocessor macro
11636 `__mips_dsp'. It also defines `__mips_dsp_rev' to 1.
11640 Use (do not use) revision 2 of the MIPS DSP ASE. *Note MIPS DSP
11641 Built-in Functions::. This option defines the preprocessor macros
11642 `__mips_dsp' and `__mips_dspr2'. It also defines `__mips_dsp_rev'
11647 Use (do not use) the MIPS SmartMIPS ASE.
11650 `-mno-paired-single'
11651 Use (do not use) paired-single floating-point instructions. *Note
11652 MIPS Paired-Single Support::. This option requires hardware
11653 floating-point support to be enabled.
11657 Use (do not use) MIPS Digital Media Extension instructions. This
11658 option can only be used when generating 64-bit code and requires
11659 hardware floating-point support to be enabled.
11663 Use (do not use) the MIPS-3D ASE. *Note MIPS-3D Built-in
11664 Functions::. The option `-mips3d' implies `-mpaired-single'.
11668 Use (do not use) MT Multithreading instructions.
11671 Force `long' types to be 64 bits wide. See `-mlong32' for an
11672 explanation of the default and the way that the pointer size is
11676 Force `long', `int', and pointer types to be 32 bits wide.
11678 The default size of `int's, `long's and pointers depends on the
11679 ABI. All the supported ABIs use 32-bit `int's. The n64 ABI uses
11680 64-bit `long's, as does the 64-bit EABI; the others use 32-bit
11681 `long's. Pointers are the same size as `long's, or the same size
11682 as integer registers, whichever is smaller.
11686 Assume (do not assume) that all symbols have 32-bit values,
11687 regardless of the selected ABI. This option is useful in
11688 combination with `-mabi=64' and `-mno-abicalls' because it allows
11689 GCC to generate shorter and faster references to symbolic
11693 Put definitions of externally-visible data in a small data section
11694 if that data is no bigger than NUM bytes. GCC can then access the
11695 data more efficiently; see `-mgpopt' for details.
11697 The default `-G' option depends on the configuration.
11701 Extend (do not extend) the `-G' behavior to local data too, such
11702 as to static variables in C. `-mlocal-sdata' is the default for
11703 all configurations.
11705 If the linker complains that an application is using too much
11706 small data, you might want to try rebuilding the less
11707 performance-critical parts with `-mno-local-sdata'. You might
11708 also want to build large libraries with `-mno-local-sdata', so
11709 that the libraries leave more room for the main program.
11712 `-mno-extern-sdata'
11713 Assume (do not assume) that externally-defined data will be in a
11714 small data section if that data is within the `-G' limit.
11715 `-mextern-sdata' is the default for all configurations.
11717 If you compile a module MOD with `-mextern-sdata' `-G NUM'
11718 `-mgpopt', and MOD references a variable VAR that is no bigger
11719 than NUM bytes, you must make sure that VAR is placed in a small
11720 data section. If VAR is defined by another module, you must
11721 either compile that module with a high-enough `-G' setting or
11722 attach a `section' attribute to VAR's definition. If VAR is
11723 common, you must link the application with a high-enough `-G'
11726 The easiest way of satisfying these restrictions is to compile and
11727 link every module with the same `-G' option. However, you may
11728 wish to build a library that supports several different small data
11729 limits. You can do this by compiling the library with the highest
11730 supported `-G' setting and additionally using `-mno-extern-sdata'
11731 to stop the library from making assumptions about
11732 externally-defined data.
11736 Use (do not use) GP-relative accesses for symbols that are known
11737 to be in a small data section; see `-G', `-mlocal-sdata' and
11738 `-mextern-sdata'. `-mgpopt' is the default for all configurations.
11740 `-mno-gpopt' is useful for cases where the `$gp' register might
11741 not hold the value of `_gp'. For example, if the code is part of
11742 a library that might be used in a boot monitor, programs that call
11743 boot monitor routines will pass an unknown value in `$gp'. (In
11744 such situations, the boot monitor itself would usually be compiled
11747 `-mno-gpopt' implies `-mno-local-sdata' and `-mno-extern-sdata'.
11750 `-mno-embedded-data'
11751 Allocate variables to the read-only data section first if
11752 possible, then next in the small data section if possible,
11753 otherwise in data. This gives slightly slower code than the
11754 default, but reduces the amount of RAM required when executing,
11755 and thus may be preferred for some embedded systems.
11757 `-muninit-const-in-rodata'
11758 `-mno-uninit-const-in-rodata'
11759 Put uninitialized `const' variables in the read-only data section.
11760 This option is only meaningful in conjunction with
11763 `-mcode-readable=SETTING'
11764 Specify whether GCC may generate code that reads from executable
11765 sections. There are three possible settings:
11767 `-mcode-readable=yes'
11768 Instructions may freely access executable sections. This is
11769 the default setting.
11771 `-mcode-readable=pcrel'
11772 MIPS16 PC-relative load instructions can access executable
11773 sections, but other instructions must not do so. This option
11774 is useful on 4KSc and 4KSd processors when the code TLBs have
11775 the Read Inhibit bit set. It is also useful on processors
11776 that can be configured to have a dual instruction/data SRAM
11777 interface and that, like the M4K, automatically redirect
11778 PC-relative loads to the instruction RAM.
11780 `-mcode-readable=no'
11781 Instructions must not access executable sections. This
11782 option can be useful on targets that are configured to have a
11783 dual instruction/data SRAM interface but that (unlike the
11784 M4K) do not automatically redirect PC-relative loads to the
11787 `-msplit-addresses'
11788 `-mno-split-addresses'
11789 Enable (disable) use of the `%hi()' and `%lo()' assembler
11790 relocation operators. This option has been superseded by
11791 `-mexplicit-relocs' but is retained for backwards compatibility.
11793 `-mexplicit-relocs'
11794 `-mno-explicit-relocs'
11795 Use (do not use) assembler relocation operators when dealing with
11796 symbolic addresses. The alternative, selected by
11797 `-mno-explicit-relocs', is to use assembler macros instead.
11799 `-mexplicit-relocs' is the default if GCC was configured to use an
11800 assembler that supports relocation operators.
11802 `-mcheck-zero-division'
11803 `-mno-check-zero-division'
11804 Trap (do not trap) on integer division by zero.
11806 The default is `-mcheck-zero-division'.
11810 MIPS systems check for division by zero by generating either a
11811 conditional trap or a break instruction. Using traps results in
11812 smaller code, but is only supported on MIPS II and later. Also,
11813 some versions of the Linux kernel have a bug that prevents trap
11814 from generating the proper signal (`SIGFPE'). Use
11815 `-mdivide-traps' to allow conditional traps on architectures that
11816 support them and `-mdivide-breaks' to force the use of breaks.
11818 The default is usually `-mdivide-traps', but this can be
11819 overridden at configure time using `--with-divide=breaks'.
11820 Divide-by-zero checks can be completely disabled using
11821 `-mno-check-zero-division'.
11825 Force (do not force) the use of `memcpy()' for non-trivial block
11826 moves. The default is `-mno-memcpy', which allows GCC to inline
11827 most constant-sized copies.
11831 Disable (do not disable) use of the `jal' instruction. Calling
11832 functions using `jal' is more efficient but requires the caller
11833 and callee to be in the same 256 megabyte segment.
11835 This option has no effect on abicalls code. The default is
11840 Enable (disable) use of the `mad', `madu' and `mul' instructions,
11841 as provided by the R4650 ISA.
11845 Enable (disable) use of the floating point multiply-accumulate
11846 instructions, when they are available. The default is
11849 When multiply-accumulate instructions are used, the intermediate
11850 product is calculated to infinite precision and is not subject to
11851 the FCSR Flush to Zero bit. This may be undesirable in some
11855 Tell the MIPS assembler to not run its preprocessor over user
11856 assembler files (with a `.s' suffix) when assembling them.
11860 Work around certain R4000 CPU errata:
11861 - A double-word or a variable shift may give an incorrect
11862 result if executed immediately after starting an integer
11865 - A double-word or a variable shift may give an incorrect
11866 result if executed while an integer multiplication is in
11869 - An integer division may give an incorrect result if started
11870 in a delay slot of a taken branch or a jump.
11874 Work around certain R4400 CPU errata:
11875 - A double-word or a variable shift may give an incorrect
11876 result if executed immediately after starting an integer
11881 Work around certain VR4120 errata:
11882 - `dmultu' does not always produce the correct result.
11884 - `div' and `ddiv' do not always produce the correct result if
11885 one of the operands is negative.
11886 The workarounds for the division errata rely on special functions
11887 in `libgcc.a'. At present, these functions are only provided by
11888 the `mips64vr*-elf' configurations.
11890 Other VR4120 errata require a nop to be inserted between certain
11891 pairs of instructions. These errata are handled by the assembler,
11895 Work around the VR4130 `mflo'/`mfhi' errata. The workarounds are
11896 implemented by the assembler rather than by GCC, although GCC will
11897 avoid using `mflo' and `mfhi' if the VR4130 `macc', `macchi',
11898 `dmacc' and `dmacchi' instructions are available instead.
11902 Work around certain SB-1 CPU core errata. (This flag currently
11903 works around the SB-1 revision 2 "F1" and "F2" floating point
11906 `-mflush-func=FUNC'
11908 Specifies the function to call to flush the I and D caches, or to
11909 not call any such function. If called, the function must take the
11910 same arguments as the common `_flush_func()', that is, the address
11911 of the memory range for which the cache is being flushed, the size
11912 of the memory range, and the number 3 (to flush both caches). The
11913 default depends on the target GCC was configured for, but commonly
11914 is either `_flush_func' or `__cpu_flush'.
11917 Set the cost of branches to roughly NUM "simple" instructions.
11918 This cost is only a heuristic and is not guaranteed to produce
11919 consistent results across releases. A zero cost redundantly
11920 selects the default, which is based on the `-mtune' setting.
11923 `-mno-branch-likely'
11924 Enable or disable use of Branch Likely instructions, regardless of
11925 the default for the selected architecture. By default, Branch
11926 Likely instructions may be generated if they are supported by the
11927 selected architecture. An exception is for the MIPS32 and MIPS64
11928 architectures and processors which implement those architectures;
11929 for those, Branch Likely instructions will not be generated by
11930 default because the MIPS32 and MIPS64 architectures specifically
11931 deprecate their use.
11934 `-mno-fp-exceptions'
11935 Specifies whether FP exceptions are enabled. This affects how we
11936 schedule FP instructions for some processors. The default is that
11937 FP exceptions are enabled.
11939 For instance, on the SB-1, if FP exceptions are disabled, and we
11940 are emitting 64-bit code, then we can use both FP pipes.
11941 Otherwise, we can only use one FP pipe.
11944 `-mno-vr4130-align'
11945 The VR4130 pipeline is two-way superscalar, but can only issue two
11946 instructions together if the first one is 8-byte aligned. When
11947 this option is enabled, GCC will align pairs of instructions that
11948 it thinks should execute in parallel.
11950 This option only has an effect when optimizing for the VR4130. It
11951 normally makes code faster, but at the expense of making it bigger.
11952 It is enabled by default at optimization level `-O3'.
11955 File: gcc.info, Node: MMIX Options, Next: MN10300 Options, Prev: MIPS Options, Up: Submodel Options
11957 3.17.22 MMIX Options
11958 --------------------
11960 These options are defined for the MMIX:
11964 Specify that intrinsic library functions are being compiled,
11965 passing all values in registers, no matter the size.
11969 Generate floating-point comparison instructions that compare with
11970 respect to the `rE' epsilon register.
11974 Generate code that passes function parameters and return values
11975 that (in the called function) are seen as registers `$0' and up,
11976 as opposed to the GNU ABI which uses global registers `$231' and
11981 When reading data from memory in sizes shorter than 64 bits, use
11982 (do not use) zero-extending load instructions by default, rather
11983 than sign-extending ones.
11987 Make the result of a division yielding a remainder have the same
11988 sign as the divisor. With the default, `-mno-knuthdiv', the sign
11989 of the remainder follows the sign of the dividend. Both methods
11990 are arithmetically valid, the latter being almost exclusively used.
11992 `-mtoplevel-symbols'
11993 `-mno-toplevel-symbols'
11994 Prepend (do not prepend) a `:' to all global symbols, so the
11995 assembly code can be used with the `PREFIX' assembly directive.
11998 Generate an executable in the ELF format, rather than the default
11999 `mmo' format used by the `mmix' simulator.
12002 `-mno-branch-predict'
12003 Use (do not use) the probable-branch instructions, when static
12004 branch prediction indicates a probable branch.
12007 `-mno-base-addresses'
12008 Generate (do not generate) code that uses _base addresses_. Using
12009 a base address automatically generates a request (handled by the
12010 assembler and the linker) for a constant to be set up in a global
12011 register. The register is used for one or more base address
12012 requests within the range 0 to 255 from the value held in the
12013 register. The generally leads to short and fast code, but the
12014 number of different data items that can be addressed is limited.
12015 This means that a program that uses lots of static data may
12016 require `-mno-base-addresses'.
12020 Force (do not force) generated code to have a single exit point in
12024 File: gcc.info, Node: MN10300 Options, Next: MT Options, Prev: MMIX Options, Up: Submodel Options
12026 3.17.23 MN10300 Options
12027 -----------------------
12029 These `-m' options are defined for Matsushita MN10300 architectures:
12032 Generate code to avoid bugs in the multiply instructions for the
12033 MN10300 processors. This is the default.
12036 Do not generate code to avoid bugs in the multiply instructions
12037 for the MN10300 processors.
12040 Generate code which uses features specific to the AM33 processor.
12043 Do not generate code which uses features specific to the AM33
12044 processor. This is the default.
12046 `-mreturn-pointer-on-d0'
12047 When generating a function which returns a pointer, return the
12048 pointer in both `a0' and `d0'. Otherwise, the pointer is returned
12049 only in a0, and attempts to call such functions without a prototype
12050 would result in errors. Note that this option is on by default;
12051 use `-mno-return-pointer-on-d0' to disable it.
12054 Do not link in the C run-time initialization object file.
12057 Indicate to the linker that it should perform a relaxation
12058 optimization pass to shorten branches, calls and absolute memory
12059 addresses. This option only has an effect when used on the
12060 command line for the final link step.
12062 This option makes symbolic debugging impossible.
12065 File: gcc.info, Node: MT Options, Next: PDP-11 Options, Prev: MN10300 Options, Up: Submodel Options
12070 These `-m' options are defined for Morpho MT architectures:
12073 Generate code that will run on CPU-TYPE, which is the name of a
12074 system representing a certain processor type. Possible values for
12075 CPU-TYPE are `ms1-64-001', `ms1-16-002', `ms1-16-003' and `ms2'.
12077 When this option is not used, the default is `-march=ms1-16-002'.
12080 Use byte loads and stores when generating code.
12083 Do not use byte loads and stores when generating code.
12086 Use simulator runtime
12089 Do not link in the C run-time initialization object file `crti.o'.
12090 Other run-time initialization and termination files such as
12091 `startup.o' and `exit.o' are still included on the linker command
12096 File: gcc.info, Node: PDP-11 Options, Next: PowerPC Options, Prev: MT Options, Up: Submodel Options
12098 3.17.25 PDP-11 Options
12099 ----------------------
12101 These options are defined for the PDP-11:
12104 Use hardware FPP floating point. This is the default. (FIS
12105 floating point on the PDP-11/40 is not supported.)
12108 Do not use hardware floating point.
12111 Return floating-point results in ac0 (fr0 in Unix assembler
12115 Return floating-point results in memory. This is the default.
12118 Generate code for a PDP-11/40.
12121 Generate code for a PDP-11/45. This is the default.
12124 Generate code for a PDP-11/10.
12127 Use inline `movmemhi' patterns for copying memory. This is the
12131 Do not use inline `movmemhi' patterns for copying memory.
12135 Use 16-bit `int'. This is the default.
12143 Use 64-bit `float'. This is the default.
12147 Use 32-bit `float'.
12150 Use `abshi2' pattern. This is the default.
12153 Do not use `abshi2' pattern.
12155 `-mbranch-expensive'
12156 Pretend that branches are expensive. This is for experimenting
12157 with code generation only.
12160 Do not pretend that branches are expensive. This is the default.
12163 Generate code for a system with split I&D.
12166 Generate code for a system without split I&D. This is the default.
12169 Use Unix assembler syntax. This is the default when configured for
12173 Use DEC assembler syntax. This is the default when configured for
12174 any PDP-11 target other than `pdp11-*-bsd'.
12177 File: gcc.info, Node: PowerPC Options, Next: RS/6000 and PowerPC Options, Prev: PDP-11 Options, Up: Submodel Options
12179 3.17.26 PowerPC Options
12180 -----------------------
12182 These are listed under *Note RS/6000 and PowerPC Options::.
12185 File: gcc.info, Node: RS/6000 and PowerPC Options, Next: S/390 and zSeries Options, Prev: PowerPC Options, Up: Submodel Options
12187 3.17.27 IBM RS/6000 and PowerPC Options
12188 ---------------------------------------
12190 These `-m' options are defined for the IBM RS/6000 and PowerPC:
12198 `-mno-powerpc-gpopt'
12200 `-mno-powerpc-gfxopt'
12215 GCC supports two related instruction set architectures for the
12216 RS/6000 and PowerPC. The "POWER" instruction set are those
12217 instructions supported by the `rios' chip set used in the original
12218 RS/6000 systems and the "PowerPC" instruction set is the
12219 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx
12220 microprocessors, and the IBM 4xx, 6xx, and follow-on
12223 Neither architecture is a subset of the other. However there is a
12224 large common subset of instructions supported by both. An MQ
12225 register is included in processors supporting the POWER
12228 You use these options to specify which instructions are available
12229 on the processor you are using. The default value of these
12230 options is determined when configuring GCC. Specifying the
12231 `-mcpu=CPU_TYPE' overrides the specification of these options. We
12232 recommend you use the `-mcpu=CPU_TYPE' option rather than the
12233 options listed above.
12235 The `-mpower' option allows GCC to generate instructions that are
12236 found only in the POWER architecture and to use the MQ register.
12237 Specifying `-mpower2' implies `-power' and also allows GCC to
12238 generate instructions that are present in the POWER2 architecture
12239 but not the original POWER architecture.
12241 The `-mpowerpc' option allows GCC to generate instructions that
12242 are found only in the 32-bit subset of the PowerPC architecture.
12243 Specifying `-mpowerpc-gpopt' implies `-mpowerpc' and also allows
12244 GCC to use the optional PowerPC architecture instructions in the
12245 General Purpose group, including floating-point square root.
12246 Specifying `-mpowerpc-gfxopt' implies `-mpowerpc' and also allows
12247 GCC to use the optional PowerPC architecture instructions in the
12248 Graphics group, including floating-point select.
12250 The `-mmfcrf' option allows GCC to generate the move from
12251 condition register field instruction implemented on the POWER4
12252 processor and other processors that support the PowerPC V2.01
12253 architecture. The `-mpopcntb' option allows GCC to generate the
12254 popcount and double precision FP reciprocal estimate instruction
12255 implemented on the POWER5 processor and other processors that
12256 support the PowerPC V2.02 architecture. The `-mfprnd' option
12257 allows GCC to generate the FP round to integer instructions
12258 implemented on the POWER5+ processor and other processors that
12259 support the PowerPC V2.03 architecture. The `-mcmpb' option
12260 allows GCC to generate the compare bytes instruction implemented
12261 on the POWER6 processor and other processors that support the
12262 PowerPC V2.05 architecture. The `-mmfpgpr' option allows GCC to
12263 generate the FP move to/from general purpose register instructions
12264 implemented on the POWER6X processor and other processors that
12265 support the extended PowerPC V2.05 architecture. The `-mhard-dfp'
12266 option allows GCC to generate the decimal floating point
12267 instructions implemented on some POWER processors.
12269 The `-mpowerpc64' option allows GCC to generate the additional
12270 64-bit instructions that are found in the full PowerPC64
12271 architecture and to treat GPRs as 64-bit, doubleword quantities.
12272 GCC defaults to `-mno-powerpc64'.
12274 If you specify both `-mno-power' and `-mno-powerpc', GCC will use
12275 only the instructions in the common subset of both architectures
12276 plus some special AIX common-mode calls, and will not use the MQ
12277 register. Specifying both `-mpower' and `-mpowerpc' permits GCC
12278 to use any instruction from either architecture and to allow use
12279 of the MQ register; specify this for the Motorola MPC601.
12283 Select which mnemonics to use in the generated assembler code.
12284 With `-mnew-mnemonics', GCC uses the assembler mnemonics defined
12285 for the PowerPC architecture. With `-mold-mnemonics' it uses the
12286 assembler mnemonics defined for the POWER architecture.
12287 Instructions defined in only one architecture have only one
12288 mnemonic; GCC uses that mnemonic irrespective of which of these
12289 options is specified.
12291 GCC defaults to the mnemonics appropriate for the architecture in
12292 use. Specifying `-mcpu=CPU_TYPE' sometimes overrides the value of
12293 these option. Unless you are building a cross-compiler, you
12294 should normally not specify either `-mnew-mnemonics' or
12295 `-mold-mnemonics', but should instead accept the default.
12298 Set architecture type, register usage, choice of mnemonics, and
12299 instruction scheduling parameters for machine type CPU_TYPE.
12300 Supported values for CPU_TYPE are `401', `403', `405', `405fp',
12301 `440', `440fp', `505', `601', `602', `603', `603e', `604', `604e',
12302 `620', `630', `740', `7400', `7450', `750', `801', `821', `823',
12303 `860', `970', `8540', `ec603e', `G3', `G4', `G5', `power',
12304 `power2', `power3', `power4', `power5', `power5+', `power6',
12305 `power6x', `common', `powerpc', `powerpc64', `rios', `rios1',
12306 `rios2', `rsc', and `rs64'.
12308 `-mcpu=common' selects a completely generic processor. Code
12309 generated under this option will run on any POWER or PowerPC
12310 processor. GCC will use only the instructions in the common
12311 subset of both architectures, and will not use the MQ register.
12312 GCC assumes a generic processor model for scheduling purposes.
12314 `-mcpu=power', `-mcpu=power2', `-mcpu=powerpc', and
12315 `-mcpu=powerpc64' specify generic POWER, POWER2, pure 32-bit
12316 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
12317 types, with an appropriate, generic processor model assumed for
12318 scheduling purposes.
12320 The other options specify a specific processor. Code generated
12321 under those options will run best on that processor, and may not
12322 run at all on others.
12324 The `-mcpu' options automatically enable or disable the following
12327 -maltivec -mfprnd -mhard-float -mmfcrf -mmultiple
12328 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64
12329 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr
12331 The particular options set for any particular CPU will vary between
12332 compiler versions, depending on what setting seems to produce
12333 optimal code for that CPU; it doesn't necessarily reflect the
12334 actual hardware's capabilities. If you wish to set an individual
12335 option to a particular value, you may specify it after the `-mcpu'
12336 option, like `-mcpu=970 -mno-altivec'.
12338 On AIX, the `-maltivec' and `-mpowerpc64' options are not enabled
12339 or disabled by the `-mcpu' option at present because AIX does not
12340 have full support for these options. You may still enable or
12341 disable them individually if you're sure it'll work in your
12345 Set the instruction scheduling parameters for machine type
12346 CPU_TYPE, but do not set the architecture type, register usage, or
12347 choice of mnemonics, as `-mcpu=CPU_TYPE' would. The same values
12348 for CPU_TYPE are used for `-mtune' as for `-mcpu'. If both are
12349 specified, the code generated will use the architecture,
12350 registers, and mnemonics set by `-mcpu', but the scheduling
12351 parameters set by `-mtune'.
12355 Generate code to compute division as reciprocal estimate and
12356 iterative refinement, creating opportunities for increased
12357 throughput. This feature requires: optional PowerPC Graphics
12358 instruction set for single precision and FRE instruction for
12359 double precision, assuming divides cannot generate user-visible
12360 traps, and the domain values not include Infinities, denormals or
12365 Generate code that uses (does not use) AltiVec instructions, and
12366 also enable the use of built-in functions that allow more direct
12367 access to the AltiVec instruction set. You may also need to set
12368 `-mabi=altivec' to adjust the current ABI with AltiVec ABI
12374 Generate VRSAVE instructions when generating AltiVec code.
12377 Generate code that allows ld and ld.so to build executables and
12378 shared libraries with non-exec .plt and .got sections. This is a
12379 PowerPC 32-bit SYSV ABI option.
12382 Generate code that uses a BSS .plt section that ld.so fills in, and
12383 requires .plt and .got sections that are both writable and
12384 executable. This is a PowerPC 32-bit SYSV ABI option.
12388 This switch enables or disables the generation of ISEL
12392 This switch has been deprecated. Use `-misel' and `-mno-isel'
12397 This switch enables or disables the generation of SPE simd
12402 This switch enables or disables the generation of PAIRED simd
12406 This option has been deprecated. Use `-mspe' and `-mno-spe'
12409 `-mfloat-gprs=YES/SINGLE/DOUBLE/NO'
12411 This switch enables or disables the generation of floating point
12412 operations on the general purpose registers for architectures that
12415 The argument YES or SINGLE enables the use of single-precision
12416 floating point operations.
12418 The argument DOUBLE enables the use of single and double-precision
12419 floating point operations.
12421 The argument NO disables floating point operations on the general
12424 This option is currently only available on the MPC854x.
12428 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12429 targets (including GNU/Linux). The 32-bit environment sets int,
12430 long and pointer to 32 bits and generates code that runs on any
12431 PowerPC variant. The 64-bit environment sets int to 32 bits and
12432 long and pointer to 64 bits, and generates code for PowerPC64, as
12439 Modify generation of the TOC (Table Of Contents), which is created
12440 for every executable file. The `-mfull-toc' option is selected by
12441 default. In that case, GCC will allocate at least one TOC entry
12442 for each unique non-automatic variable reference in your program.
12443 GCC will also place floating-point constants in the TOC. However,
12444 only 16,384 entries are available in the TOC.
12446 If you receive a linker error message that saying you have
12447 overflowed the available TOC space, you can reduce the amount of
12448 TOC space used with the `-mno-fp-in-toc' and `-mno-sum-in-toc'
12449 options. `-mno-fp-in-toc' prevents GCC from putting floating-point
12450 constants in the TOC and `-mno-sum-in-toc' forces GCC to generate
12451 code to calculate the sum of an address and a constant at run-time
12452 instead of putting that sum into the TOC. You may specify one or
12453 both of these options. Each causes GCC to produce very slightly
12454 slower and larger code at the expense of conserving TOC space.
12456 If you still run out of space in the TOC even when you specify
12457 both of these options, specify `-mminimal-toc' instead. This
12458 option causes GCC to make only one TOC entry for every file. When
12459 you specify this option, GCC will produce code that is slower and
12460 larger but which uses extremely little TOC space. You may wish to
12461 use this option only on files that contain less frequently
12466 Enable 64-bit AIX ABI and calling convention: 64-bit pointers,
12467 64-bit `long' type, and the infrastructure needed to support them.
12468 Specifying `-maix64' implies `-mpowerpc64' and `-mpowerpc', while
12469 `-maix32' disables the 64-bit ABI and implies `-mno-powerpc64'.
12470 GCC defaults to `-maix32'.
12474 Produce code that conforms more closely to IBM XL compiler
12475 semantics when using AIX-compatible ABI. Pass floating-point
12476 arguments to prototyped functions beyond the register save area
12477 (RSA) on the stack in addition to argument FPRs. Do not assume
12478 that most significant double in 128-bit long double value is
12479 properly rounded when comparing values and converting to double.
12480 Use XL symbol names for long double support routines.
12482 The AIX calling convention was extended but not initially
12483 documented to handle an obscure K&R C case of calling a function
12484 that takes the address of its arguments with fewer arguments than
12485 declared. IBM XL compilers access floating point arguments which
12486 do not fit in the RSA from the stack when a subroutine is compiled
12487 without optimization. Because always storing floating-point
12488 arguments on the stack is inefficient and rarely needed, this
12489 option is not enabled by default and only is necessary when
12490 calling subroutines compiled by IBM XL compilers without
12494 Support "IBM RS/6000 SP" "Parallel Environment" (PE). Link an
12495 application written to use message passing with special startup
12496 code to enable the application to run. The system must have PE
12497 installed in the standard location (`/usr/lpp/ppe.poe/'), or the
12498 `specs' file must be overridden with the `-specs=' option to
12499 specify the appropriate directory location. The Parallel
12500 Environment does not support threads, so the `-mpe' option and the
12501 `-pthread' option are incompatible.
12505 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12506 `-malign-natural' overrides the ABI-defined alignment of larger
12507 types, such as floating-point doubles, on their natural size-based
12508 boundary. The option `-malign-power' instructs GCC to follow the
12509 ABI-specified alignment rules. GCC defaults to the standard
12510 alignment defined in the ABI.
12512 On 64-bit Darwin, natural alignment is the default, and
12513 `-malign-power' is not supported.
12517 Generate code that does not use (uses) the floating-point register
12518 set. Software floating point emulation is provided if you use the
12519 `-msoft-float' option, and pass the option to GCC when linking.
12523 Generate code that uses (does not use) the load multiple word
12524 instructions and the store multiple word instructions. These
12525 instructions are generated by default on POWER systems, and not
12526 generated on PowerPC systems. Do not use `-mmultiple' on little
12527 endian PowerPC systems, since those instructions do not work when
12528 the processor is in little endian mode. The exceptions are PPC740
12529 and PPC750 which permit the instructions usage in little endian
12534 Generate code that uses (does not use) the load string instructions
12535 and the store string word instructions to save multiple registers
12536 and do small block moves. These instructions are generated by
12537 default on POWER systems, and not generated on PowerPC systems.
12538 Do not use `-mstring' on little endian PowerPC systems, since those
12539 instructions do not work when the processor is in little endian
12540 mode. The exceptions are PPC740 and PPC750 which permit the
12541 instructions usage in little endian mode.
12545 Generate code that uses (does not use) the load or store
12546 instructions that update the base register to the address of the
12547 calculated memory location. These instructions are generated by
12548 default. If you use `-mno-update', there is a small window
12549 between the time that the stack pointer is updated and the address
12550 of the previous frame is stored, which means code that walks the
12551 stack frame across interrupts or signals may get corrupted data.
12555 Generate code that uses (does not use) the floating point multiply
12556 and accumulate instructions. These instructions are generated by
12557 default if hardware floating is used.
12561 Generate code that uses (does not use) the half-word multiply and
12562 multiply-accumulate instructions on the IBM 405 and 440 processors.
12563 These instructions are generated by default when targetting those
12568 Generate code that uses (does not use) the string-search `dlmzb'
12569 instruction on the IBM 405 and 440 processors. This instruction is
12570 generated by default when targetting those processors.
12574 On System V.4 and embedded PowerPC systems do not (do) force
12575 structures and unions that contain bit-fields to be aligned to the
12576 base type of the bit-field.
12578 For example, by default a structure containing nothing but 8
12579 `unsigned' bit-fields of length 1 would be aligned to a 4 byte
12580 boundary and have a size of 4 bytes. By using `-mno-bit-align',
12581 the structure would be aligned to a 1 byte boundary and be one
12584 `-mno-strict-align'
12586 On System V.4 and embedded PowerPC systems do not (do) assume that
12587 unaligned memory references will be handled by the system.
12591 On embedded PowerPC systems generate code that allows (does not
12592 allow) the program to be relocated to a different address at
12593 runtime. If you use `-mrelocatable' on any module, all objects
12594 linked together must be compiled with `-mrelocatable' or
12595 `-mrelocatable-lib'.
12597 `-mrelocatable-lib'
12598 `-mno-relocatable-lib'
12599 On embedded PowerPC systems generate code that allows (does not
12600 allow) the program to be relocated to a different address at
12601 runtime. Modules compiled with `-mrelocatable-lib' can be linked
12602 with either modules compiled without `-mrelocatable' and
12603 `-mrelocatable-lib' or with modules compiled with the
12604 `-mrelocatable' options.
12608 On System V.4 and embedded PowerPC systems do not (do) assume that
12609 register 2 contains a pointer to a global area pointing to the
12610 addresses used in the program.
12614 On System V.4 and embedded PowerPC systems compile code for the
12615 processor in little endian mode. The `-mlittle-endian' option is
12616 the same as `-mlittle'.
12620 On System V.4 and embedded PowerPC systems compile code for the
12621 processor in big endian mode. The `-mbig-endian' option is the
12625 On Darwin and Mac OS X systems, compile code so that it is not
12626 relocatable, but that its external references are relocatable. The
12627 resulting code is suitable for applications, but not shared
12630 `-mprioritize-restricted-insns=PRIORITY'
12631 This option controls the priority that is assigned to
12632 dispatch-slot restricted instructions during the second scheduling
12633 pass. The argument PRIORITY takes the value 0/1/2 to assign
12634 NO/HIGHEST/SECOND-HIGHEST priority to dispatch slot restricted
12637 `-msched-costly-dep=DEPENDENCE_TYPE'
12638 This option controls which dependences are considered costly by
12639 the target during instruction scheduling. The argument
12640 DEPENDENCE_TYPE takes one of the following values: NO: no
12641 dependence is costly, ALL: all dependences are costly,
12642 TRUE_STORE_TO_LOAD: a true dependence from store to load is costly,
12643 STORE_TO_LOAD: any dependence from store to load is costly,
12644 NUMBER: any dependence which latency >= NUMBER is costly.
12646 `-minsert-sched-nops=SCHEME'
12647 This option controls which nop insertion scheme will be used during
12648 the second scheduling pass. The argument SCHEME takes one of the
12649 following values: NO: Don't insert nops. PAD: Pad with nops any
12650 dispatch group which has vacant issue slots, according to the
12651 scheduler's grouping. REGROUP_EXACT: Insert nops to force costly
12652 dependent insns into separate groups. Insert exactly as many nops
12653 as needed to force an insn to a new group, according to the
12654 estimated processor grouping. NUMBER: Insert nops to force costly
12655 dependent insns into separate groups. Insert NUMBER nops to force
12656 an insn to a new group.
12659 On System V.4 and embedded PowerPC systems compile code using
12660 calling conventions that adheres to the March 1995 draft of the
12661 System V Application Binary Interface, PowerPC processor
12662 supplement. This is the default unless you configured GCC using
12663 `powerpc-*-eabiaix'.
12666 Specify both `-mcall-sysv' and `-meabi' options.
12668 `-mcall-sysv-noeabi'
12669 Specify both `-mcall-sysv' and `-mno-eabi' options.
12672 On System V.4 and embedded PowerPC systems compile code for the
12673 Solaris operating system.
12676 On System V.4 and embedded PowerPC systems compile code for the
12677 Linux-based GNU system.
12680 On System V.4 and embedded PowerPC systems compile code for the
12681 Hurd-based GNU system.
12684 On System V.4 and embedded PowerPC systems compile code for the
12685 NetBSD operating system.
12687 `-maix-struct-return'
12688 Return all structures in memory (as specified by the AIX ABI).
12690 `-msvr4-struct-return'
12691 Return structures smaller than 8 bytes in registers (as specified
12695 Extend the current ABI with a particular extension, or remove such
12696 extension. Valid values are ALTIVEC, NO-ALTIVEC, SPE, NO-SPE,
12697 IBMLONGDOUBLE, IEEELONGDOUBLE.
12700 Extend the current ABI with SPE ABI extensions. This does not
12701 change the default ABI, instead it adds the SPE ABI extensions to
12705 Disable Booke SPE ABI extensions for the current ABI.
12707 `-mabi=ibmlongdouble'
12708 Change the current ABI to use IBM extended precision long double.
12709 This is a PowerPC 32-bit SYSV ABI option.
12711 `-mabi=ieeelongdouble'
12712 Change the current ABI to use IEEE extended precision long double.
12713 This is a PowerPC 32-bit Linux ABI option.
12717 On System V.4 and embedded PowerPC systems assume that all calls to
12718 variable argument functions are properly prototyped. Otherwise,
12719 the compiler must insert an instruction before every non
12720 prototyped call to set or clear bit 6 of the condition code
12721 register (CR) to indicate whether floating point values were
12722 passed in the floating point registers in case the function takes
12723 a variable arguments. With `-mprototype', only calls to
12724 prototyped variable argument functions will set or clear the bit.
12727 On embedded PowerPC systems, assume that the startup module is
12728 called `sim-crt0.o' and that the standard C libraries are
12729 `libsim.a' and `libc.a'. This is the default for
12730 `powerpc-*-eabisim' configurations.
12733 On embedded PowerPC systems, assume that the startup module is
12734 called `crt0.o' and the standard C libraries are `libmvme.a' and
12738 On embedded PowerPC systems, assume that the startup module is
12739 called `crt0.o' and the standard C libraries are `libads.a' and
12743 On embedded PowerPC systems, assume that the startup module is
12744 called `crt0.o' and the standard C libraries are `libyk.a' and
12748 On System V.4 and embedded PowerPC systems, specify that you are
12749 compiling for a VxWorks system.
12752 Specify that you are compiling for the WindISS simulation
12756 On embedded PowerPC systems, set the PPC_EMB bit in the ELF flags
12757 header to indicate that `eabi' extended relocations are used.
12761 On System V.4 and embedded PowerPC systems do (do not) adhere to
12762 the Embedded Applications Binary Interface (eabi) which is a set of
12763 modifications to the System V.4 specifications. Selecting `-meabi'
12764 means that the stack is aligned to an 8 byte boundary, a function
12765 `__eabi' is called to from `main' to set up the eabi environment,
12766 and the `-msdata' option can use both `r2' and `r13' to point to
12767 two separate small data areas. Selecting `-mno-eabi' means that
12768 the stack is aligned to a 16 byte boundary, do not call an
12769 initialization function from `main', and the `-msdata' option will
12770 only use `r13' to point to a single small data area. The `-meabi'
12771 option is on by default if you configured GCC using one of the
12772 `powerpc*-*-eabi*' options.
12775 On System V.4 and embedded PowerPC systems, put small initialized
12776 `const' global and static data in the `.sdata2' section, which is
12777 pointed to by register `r2'. Put small initialized non-`const'
12778 global and static data in the `.sdata' section, which is pointed
12779 to by register `r13'. Put small uninitialized global and static
12780 data in the `.sbss' section, which is adjacent to the `.sdata'
12781 section. The `-msdata=eabi' option is incompatible with the
12782 `-mrelocatable' option. The `-msdata=eabi' option also sets the
12786 On System V.4 and embedded PowerPC systems, put small global and
12787 static data in the `.sdata' section, which is pointed to by
12788 register `r13'. Put small uninitialized global and static data in
12789 the `.sbss' section, which is adjacent to the `.sdata' section.
12790 The `-msdata=sysv' option is incompatible with the `-mrelocatable'
12795 On System V.4 and embedded PowerPC systems, if `-meabi' is used,
12796 compile code the same as `-msdata=eabi', otherwise compile code the
12797 same as `-msdata=sysv'.
12800 On System V.4 and embedded PowerPC systems, put small global data
12801 in the `.sdata' section. Put small uninitialized global data in
12802 the `.sbss' section. Do not use register `r13' to address small
12803 data however. This is the default behavior unless other `-msdata'
12808 On embedded PowerPC systems, put all initialized global and static
12809 data in the `.data' section, and all uninitialized data in the
12813 On embedded PowerPC systems, put global and static items less than
12814 or equal to NUM bytes into the small data or bss sections instead
12815 of the normal data or bss section. By default, NUM is 8. The `-G
12816 NUM' switch is also passed to the linker. All modules should be
12817 compiled with the same `-G NUM' value.
12821 On System V.4 and embedded PowerPC systems do (do not) emit
12822 register names in the assembly language output using symbolic
12827 By default assume that all calls are far away so that a longer more
12828 expensive calling sequence is required. This is required for calls
12829 further than 32 megabytes (33,554,432 bytes) from the current
12830 location. A short call will be generated if the compiler knows
12831 the call cannot be that far away. This setting can be overridden
12832 by the `shortcall' function attribute, or by `#pragma longcall(0)'.
12834 Some linkers are capable of detecting out-of-range calls and
12835 generating glue code on the fly. On these systems, long calls are
12836 unnecessary and generate slower code. As of this writing, the AIX
12837 linker can do this, as can the GNU linker for PowerPC/64. It is
12838 planned to add this feature to the GNU linker for 32-bit PowerPC
12841 On Darwin/PPC systems, `#pragma longcall' will generate "jbsr
12842 callee, L42", plus a "branch island" (glue code). The two target
12843 addresses represent the callee and the "branch island". The
12844 Darwin/PPC linker will prefer the first address and generate a "bl
12845 callee" if the PPC "bl" instruction will reach the callee directly;
12846 otherwise, the linker will generate "bl L42" to call the "branch
12847 island". The "branch island" is appended to the body of the
12848 calling function; it computes the full 32-bit address of the callee
12851 On Mach-O (Darwin) systems, this option directs the compiler emit
12852 to the glue for every direct call, and the Darwin linker decides
12853 whether to use or discard it.
12855 In the future, we may cause GCC to ignore all longcall
12856 specifications when the linker is known to generate glue.
12859 Adds support for multithreading with the "pthreads" library. This
12860 option sets flags for both the preprocessor and linker.
12864 File: gcc.info, Node: S/390 and zSeries Options, Next: Score Options, Prev: RS/6000 and PowerPC Options, Up: Submodel Options
12866 3.17.28 S/390 and zSeries Options
12867 ---------------------------------
12869 These are the `-m' options defined for the S/390 and zSeries
12874 Use (do not use) the hardware floating-point instructions and
12875 registers for floating-point operations. When `-msoft-float' is
12876 specified, functions in `libgcc.a' will be used to perform
12877 floating-point operations. When `-mhard-float' is specified, the
12878 compiler generates IEEE floating-point instructions. This is the
12882 `-mlong-double-128'
12883 These switches control the size of `long double' type. A size of
12884 64bit makes the `long double' type equivalent to the `double'
12885 type. This is the default.
12889 Store (do not store) the address of the caller's frame as
12890 backchain pointer into the callee's stack frame. A backchain may
12891 be needed to allow debugging using tools that do not understand
12892 DWARF-2 call frame information. When `-mno-packed-stack' is in
12893 effect, the backchain pointer is stored at the bottom of the stack
12894 frame; when `-mpacked-stack' is in effect, the backchain is placed
12895 into the topmost word of the 96/160 byte register save area.
12897 In general, code compiled with `-mbackchain' is call-compatible
12898 with code compiled with `-mmo-backchain'; however, use of the
12899 backchain for debugging purposes usually requires that the whole
12900 binary is built with `-mbackchain'. Note that the combination of
12901 `-mbackchain', `-mpacked-stack' and `-mhard-float' is not
12902 supported. In order to build a linux kernel use `-msoft-float'.
12904 The default is to not maintain the backchain.
12908 `-mno-packed-stack'
12909 Use (do not use) the packed stack layout. When
12910 `-mno-packed-stack' is specified, the compiler uses the all fields
12911 of the 96/160 byte register save area only for their default
12912 purpose; unused fields still take up stack space. When
12913 `-mpacked-stack' is specified, register save slots are densely
12914 packed at the top of the register save area; unused space is
12915 reused for other purposes, allowing for more efficient use of the
12916 available stack space. However, when `-mbackchain' is also in
12917 effect, the topmost word of the save area is always used to store
12918 the backchain, and the return address register is always saved two
12919 words below the backchain.
12921 As long as the stack frame backchain is not used, code generated
12922 with `-mpacked-stack' is call-compatible with code generated with
12923 `-mno-packed-stack'. Note that some non-FSF releases of GCC 2.95
12924 for S/390 or zSeries generated code that uses the stack frame
12925 backchain at run time, not just for debugging purposes. Such code
12926 is not call-compatible with code compiled with `-mpacked-stack'.
12927 Also, note that the combination of `-mbackchain', `-mpacked-stack'
12928 and `-mhard-float' is not supported. In order to build a linux
12929 kernel use `-msoft-float'.
12931 The default is to not use the packed stack layout.
12935 Generate (or do not generate) code using the `bras' instruction to
12936 do subroutine calls. This only works reliably if the total
12937 executable size does not exceed 64k. The default is to use the
12938 `basr' instruction instead, which does not have this limitation.
12942 When `-m31' is specified, generate code compliant to the GNU/Linux
12943 for S/390 ABI. When `-m64' is specified, generate code compliant
12944 to the GNU/Linux for zSeries ABI. This allows GCC in particular
12945 to generate 64-bit instructions. For the `s390' targets, the
12946 default is `-m31', while the `s390x' targets default to `-m64'.
12950 When `-mzarch' is specified, generate code using the instructions
12951 available on z/Architecture. When `-mesa' is specified, generate
12952 code using the instructions available on ESA/390. Note that
12953 `-mesa' is not possible with `-m64'. When generating code
12954 compliant to the GNU/Linux for S/390 ABI, the default is `-mesa'.
12955 When generating code compliant to the GNU/Linux for zSeries ABI,
12956 the default is `-mzarch'.
12960 Generate (or do not generate) code using the `mvcle' instruction
12961 to perform block moves. When `-mno-mvcle' is specified, use a
12962 `mvc' loop instead. This is the default unless optimizing for
12967 Print (or do not print) additional debug information when
12968 compiling. The default is to not print debug information.
12971 Generate code that will run on CPU-TYPE, which is the name of a
12972 system representing a certain processor type. Possible values for
12973 CPU-TYPE are `g5', `g6', `z900', and `z990'. When generating code
12974 using the instructions available on z/Architecture, the default is
12975 `-march=z900'. Otherwise, the default is `-march=g5'.
12978 Tune to CPU-TYPE everything applicable about the generated code,
12979 except for the ABI and the set of available instructions. The
12980 list of CPU-TYPE values is the same as for `-march'. The default
12981 is the value used for `-march'.
12985 Generate code that adds (does not add) in TPF OS specific branches
12986 to trace routines in the operating system. This option is off by
12987 default, even when compiling for the TPF OS.
12991 Generate code that uses (does not use) the floating point multiply
12992 and accumulate instructions. These instructions are generated by
12993 default if hardware floating point is used.
12995 `-mwarn-framesize=FRAMESIZE'
12996 Emit a warning if the current function exceeds the given frame
12997 size. Because this is a compile time check it doesn't need to be
12998 a real problem when the program runs. It is intended to identify
12999 functions which most probably cause a stack overflow. It is
13000 useful to be used in an environment with limited stack size e.g.
13003 `-mwarn-dynamicstack'
13004 Emit a warning if the function calls alloca or uses dynamically
13005 sized arrays. This is generally a bad idea with a limited stack
13008 `-mstack-guard=STACK-GUARD'
13010 `-mstack-size=STACK-SIZE'
13011 If these options are provided the s390 back end emits additional
13012 instructions in the function prologue which trigger a trap if the
13013 stack size is STACK-GUARD bytes above the STACK-SIZE (remember
13014 that the stack on s390 grows downward). If the STACK-GUARD option
13015 is omitted the smallest power of 2 larger than the frame size of
13016 the compiled function is chosen. These options are intended to be
13017 used to help debugging stack overflow problems. The additionally
13018 emitted code causes only little overhead and hence can also be
13019 used in production like systems without greater performance
13020 degradation. The given values have to be exact powers of 2 and
13021 STACK-SIZE has to be greater than STACK-GUARD without exceeding
13022 64k. In order to be efficient the extra code makes the assumption
13023 that the stack starts at an address aligned to the value given by
13024 STACK-SIZE. The STACK-GUARD option can only be used in
13025 conjunction with STACK-SIZE.
13028 File: gcc.info, Node: Score Options, Next: SH Options, Prev: S/390 and zSeries Options, Up: Submodel Options
13030 3.17.29 Score Options
13031 ---------------------
13033 These options are defined for Score implementations:
13036 Compile code for big endian mode. This is the default.
13039 Compile code for little endian mode.
13042 Disable generate bcnz instruction.
13045 Enable generate unaligned load and store instruction.
13048 Enable the use of multiply-accumulate instructions. Disabled by
13052 Specify the SCORE5 as the target architecture.
13055 Specify the SCORE5U of the target architecture.
13058 Specify the SCORE7 as the target architecture. This is the default.
13061 Specify the SCORE7D as the target architecture.
13064 File: gcc.info, Node: SH Options, Next: SPARC Options, Prev: Score Options, Up: Submodel Options
13069 These `-m' options are defined for the SH implementations:
13072 Generate code for the SH1.
13075 Generate code for the SH2.
13078 Generate code for the SH2e.
13081 Generate code for the SH3.
13084 Generate code for the SH3e.
13087 Generate code for the SH4 without a floating-point unit.
13090 Generate code for the SH4 with a floating-point unit that only
13091 supports single-precision arithmetic.
13094 Generate code for the SH4 assuming the floating-point unit is in
13095 single-precision mode by default.
13098 Generate code for the SH4.
13101 Generate code for the SH4al-dsp, or for a SH4a in such a way that
13102 the floating-point unit is not used.
13105 Generate code for the SH4a, in such a way that no double-precision
13106 floating point operations are used.
13109 Generate code for the SH4a assuming the floating-point unit is in
13110 single-precision mode by default.
13113 Generate code for the SH4a.
13116 Same as `-m4a-nofpu', except that it implicitly passes `-dsp' to
13117 the assembler. GCC doesn't generate any DSP instructions at the
13121 Compile code for the processor in big endian mode.
13124 Compile code for the processor in little endian mode.
13127 Align doubles at 64-bit boundaries. Note that this changes the
13128 calling conventions, and thus some functions from the standard C
13129 library will not work unless you recompile it first with
13133 Shorten some address references at link time, when possible; uses
13134 the linker option `-relax'.
13137 Use 32-bit offsets in `switch' tables. The default is to use
13141 Enable the use of the instruction `fmovd'.
13144 Comply with the calling conventions defined by Renesas.
13147 Comply with the calling conventions defined by Renesas.
13150 Comply with the calling conventions defined for GCC before the
13151 Renesas conventions were available. This option is the default
13152 for all targets of the SH toolchain except for `sh-symbianelf'.
13155 Mark the `MAC' register as call-clobbered, even if `-mhitachi' is
13159 Increase IEEE-compliance of floating-point code. At the moment,
13160 this is equivalent to `-fno-finite-math-only'. When generating 16
13161 bit SH opcodes, getting IEEE-conforming results for comparisons of
13162 NANs / infinities incurs extra overhead in every floating point
13163 comparison, therefore the default is set to `-ffinite-math-only'.
13165 `-minline-ic_invalidate'
13166 Inline code to invalidate instruction cache entries after setting
13167 up nested function trampolines. This option has no effect if
13168 -musermode is in effect and the selected code generation option
13169 (e.g. -m4) does not allow the use of the icbi instruction. If the
13170 selected code generation option does not allow the use of the icbi
13171 instruction, and -musermode is not in effect, the inlined code will
13172 manipulate the instruction cache address array directly with an
13173 associative write. This not only requires privileged mode, but it
13174 will also fail if the cache line had been mapped via the TLB and
13175 has become unmapped.
13178 Dump instruction size and location in the assembly code.
13181 This option is deprecated. It pads structures to multiple of 4
13182 bytes, which is incompatible with the SH ABI.
13185 Optimize for space instead of speed. Implied by `-Os'.
13188 When generating position-independent code, emit function calls
13189 using the Global Offset Table instead of the Procedure Linkage
13193 Don't generate privileged mode only code; implies
13194 -mno-inline-ic_invalidate if the inlined code would not work in
13195 user mode. This is the default when the target is `sh-*-linux*'.
13198 Set the cost to assume for a multiply insn.
13201 Set the division strategy to use for SHmedia code. STRATEGY must
13202 be one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l,
13203 inv:call, inv:call2, inv:fp . "fp" performs the operation in
13204 floating point. This has a very high latency, but needs only a
13205 few instructions, so it might be a good choice if your code has
13206 enough easily exploitable ILP to allow the compiler to schedule
13207 the floating point instructions together with other instructions.
13208 Division by zero causes a floating point exception. "inv" uses
13209 integer operations to calculate the inverse of the divisor, and
13210 then multiplies the dividend with the inverse. This strategy
13211 allows cse and hoisting of the inverse calculation. Division by
13212 zero calculates an unspecified result, but does not trap.
13213 "inv:minlat" is a variant of "inv" where if no cse / hoisting
13214 opportunities have been found, or if the entire operation has been
13215 hoisted to the same place, the last stages of the inverse
13216 calculation are intertwined with the final multiply to reduce the
13217 overall latency, at the expense of using a few more instructions,
13218 and thus offering fewer scheduling opportunities with other code.
13219 "call" calls a library function that usually implements the
13220 inv:minlat strategy. This gives high code density for
13221 m5-*media-nofpu compilations. "call2" uses a different entry
13222 point of the same library function, where it assumes that a
13223 pointer to a lookup table has already been set up, which exposes
13224 the pointer load to cse / code hoisting optimizations.
13225 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm
13226 for initial code generation, but if the code stays unoptimized,
13227 revert to the "call", "call2", or "fp" strategies, respectively.
13228 Note that the potentially-trapping side effect of division by zero
13229 is carried by a separate instruction, so it is possible that all
13230 the integer instructions are hoisted out, but the marker for the
13231 side effect stays where it is. A recombination to fp operations
13232 or a call is not possible in that case. "inv20u" and "inv20l" are
13233 variants of the "inv:minlat" strategy. In the case that the
13234 inverse calculation was nor separated from the multiply, they speed
13235 up division where the dividend fits into 20 bits (plus sign where
13236 applicable), by inserting a test to skip a number of operations in
13237 this case; this test slows down the case of larger dividends.
13238 inv20u assumes the case of a such a small dividend to be unlikely,
13239 and inv20l assumes it to be likely.
13241 `-mdivsi3_libfunc=NAME'
13242 Set the name of the library function used for 32 bit signed
13243 division to NAME. This only affect the name used in the call and
13244 inv:call division strategies, and the compiler will still expect
13245 the same sets of input/output/clobbered registers as if this
13246 option was not present.
13249 Throttle unrolling to avoid thrashing target registers. This
13250 option only has an effect if the gcc code base supports the
13251 TARGET_ADJUST_UNROLL_MAX target hook.
13253 `-mindexed-addressing'
13254 Enable the use of the indexed addressing mode for
13255 SHmedia32/SHcompact. This is only safe if the hardware and/or OS
13256 implement 32 bit wrap-around semantics for the indexed addressing
13257 mode. The architecture allows the implementation of processors
13258 with 64 bit MMU, which the OS could use to get 32 bit addressing,
13259 but since no current hardware implementation supports this or any
13260 other way to make the indexed addressing mode safe to use in the
13261 32 bit ABI, the default is -mno-indexed-addressing.
13263 `-mgettrcost=NUMBER'
13264 Set the cost assumed for the gettr instruction to NUMBER. The
13265 default is 2 if `-mpt-fixed' is in effect, 100 otherwise.
13268 Assume pt* instructions won't trap. This will generally generate
13269 better scheduled code, but is unsafe on current hardware. The
13270 current architecture definition says that ptabs and ptrel trap
13271 when the target anded with 3 is 3. This has the unintentional
13272 effect of making it unsafe to schedule ptabs / ptrel before a
13273 branch, or hoist it out of a loop. For example,
13274 __do_global_ctors, a part of libgcc that runs constructors at
13275 program startup, calls functions in a list which is delimited by
13276 -1. With the -mpt-fixed option, the ptabs will be done before
13277 testing against -1. That means that all the constructors will be
13278 run a bit quicker, but when the loop comes to the end of the list,
13279 the program crashes because ptabs loads -1 into a target register.
13280 Since this option is unsafe for any hardware implementing the
13281 current architecture specification, the default is -mno-pt-fixed.
13282 Unless the user specifies a specific cost with `-mgettrcost',
13283 -mno-pt-fixed also implies `-mgettrcost=100'; this deters register
13284 allocation using target registers for storing ordinary integers.
13286 `-minvalid-symbols'
13287 Assume symbols might be invalid. Ordinary function symbols
13288 generated by the compiler will always be valid to load with
13289 movi/shori/ptabs or movi/shori/ptrel, but with assembler and/or
13290 linker tricks it is possible to generate symbols that will cause
13291 ptabs / ptrel to trap. This option is only meaningful when
13292 `-mno-pt-fixed' is in effect. It will then prevent
13293 cross-basic-block cse, hoisting and most scheduling of symbol
13294 loads. The default is `-mno-invalid-symbols'.
13297 File: gcc.info, Node: SPARC Options, Next: SPU Options, Prev: SH Options, Up: Submodel Options
13299 3.17.31 SPARC Options
13300 ---------------------
13302 These `-m' options are supported on the SPARC:
13306 Specify `-mapp-regs' to generate output using the global registers
13307 2 through 4, which the SPARC SVR4 ABI reserves for applications.
13308 This is the default.
13310 To be fully SVR4 ABI compliant at the cost of some performance
13311 loss, specify `-mno-app-regs'. You should compile libraries and
13312 system software with this option.
13316 Generate output containing floating point instructions. This is
13321 Generate output containing library calls for floating point.
13322 *Warning:* the requisite libraries are not available for all SPARC
13323 targets. Normally the facilities of the machine's usual C
13324 compiler are used, but this cannot be done directly in
13325 cross-compilation. You must make your own arrangements to provide
13326 suitable library functions for cross-compilation. The embedded
13327 targets `sparc-*-aout' and `sparclite-*-*' do provide software
13328 floating point support.
13330 `-msoft-float' changes the calling convention in the output file;
13331 therefore, it is only useful if you compile _all_ of a program with
13332 this option. In particular, you need to compile `libgcc.a', the
13333 library that comes with GCC, with `-msoft-float' in order for this
13336 `-mhard-quad-float'
13337 Generate output containing quad-word (long double) floating point
13340 `-msoft-quad-float'
13341 Generate output containing library calls for quad-word (long
13342 double) floating point instructions. The functions called are
13343 those specified in the SPARC ABI. This is the default.
13345 As of this writing, there are no SPARC implementations that have
13346 hardware support for the quad-word floating point instructions.
13347 They all invoke a trap handler for one of these instructions, and
13348 then the trap handler emulates the effect of the instruction.
13349 Because of the trap handler overhead, this is much slower than
13350 calling the ABI library routines. Thus the `-msoft-quad-float'
13351 option is the default.
13353 `-mno-unaligned-doubles'
13354 `-munaligned-doubles'
13355 Assume that doubles have 8 byte alignment. This is the default.
13357 With `-munaligned-doubles', GCC assumes that doubles have 8 byte
13358 alignment only if they are contained in another type, or if they
13359 have an absolute address. Otherwise, it assumes they have 4 byte
13360 alignment. Specifying this option avoids some rare compatibility
13361 problems with code generated by other compilers. It is not the
13362 default because it results in a performance loss, especially for
13363 floating point code.
13365 `-mno-faster-structs'
13367 With `-mfaster-structs', the compiler assumes that structures
13368 should have 8 byte alignment. This enables the use of pairs of
13369 `ldd' and `std' instructions for copies in structure assignment,
13370 in place of twice as many `ld' and `st' pairs. However, the use
13371 of this changed alignment directly violates the SPARC ABI. Thus,
13372 it's intended only for use on targets where the developer
13373 acknowledges that their resulting code will not be directly in
13374 line with the rules of the ABI.
13377 `-mimpure-text', used in addition to `-shared', tells the compiler
13378 to not pass `-z text' to the linker when linking a shared object.
13379 Using this option, you can link position-dependent code into a
13382 `-mimpure-text' suppresses the "relocations remain against
13383 allocatable but non-writable sections" linker error message.
13384 However, the necessary relocations will trigger copy-on-write, and
13385 the shared object is not actually shared across processes.
13386 Instead of using `-mimpure-text', you should compile all source
13387 code with `-fpic' or `-fPIC'.
13389 This option is only available on SunOS and Solaris.
13392 Set the instruction set, register set, and instruction scheduling
13393 parameters for machine type CPU_TYPE. Supported values for
13394 CPU_TYPE are `v7', `cypress', `v8', `supersparc', `sparclite',
13395 `f930', `f934', `hypersparc', `sparclite86x', `sparclet',
13396 `tsc701', `v9', `ultrasparc', `ultrasparc3', `niagara' and
13399 Default instruction scheduling parameters are used for values that
13400 select an architecture and not an implementation. These are `v7',
13401 `v8', `sparclite', `sparclet', `v9'.
13403 Here is a list of each supported architecture and their supported
13407 v8: supersparc, hypersparc
13408 sparclite: f930, f934, sparclite86x
13410 v9: ultrasparc, ultrasparc3, niagara, niagara2
13412 By default (unless configured otherwise), GCC generates code for
13413 the V7 variant of the SPARC architecture. With `-mcpu=cypress',
13414 the compiler additionally optimizes it for the Cypress CY7C602
13415 chip, as used in the SPARCStation/SPARCServer 3xx series. This is
13416 also appropriate for the older SPARCStation 1, 2, IPX etc.
13418 With `-mcpu=v8', GCC generates code for the V8 variant of the SPARC
13419 architecture. The only difference from V7 code is that the
13420 compiler emits the integer multiply and integer divide
13421 instructions which exist in SPARC-V8 but not in SPARC-V7. With
13422 `-mcpu=supersparc', the compiler additionally optimizes it for the
13423 SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000
13426 With `-mcpu=sparclite', GCC generates code for the SPARClite
13427 variant of the SPARC architecture. This adds the integer
13428 multiply, integer divide step and scan (`ffs') instructions which
13429 exist in SPARClite but not in SPARC-V7. With `-mcpu=f930', the
13430 compiler additionally optimizes it for the Fujitsu MB86930 chip,
13431 which is the original SPARClite, with no FPU. With `-mcpu=f934',
13432 the compiler additionally optimizes it for the Fujitsu MB86934
13433 chip, which is the more recent SPARClite with FPU.
13435 With `-mcpu=sparclet', GCC generates code for the SPARClet variant
13436 of the SPARC architecture. This adds the integer multiply,
13437 multiply/accumulate, integer divide step and scan (`ffs')
13438 instructions which exist in SPARClet but not in SPARC-V7. With
13439 `-mcpu=tsc701', the compiler additionally optimizes it for the
13440 TEMIC SPARClet chip.
13442 With `-mcpu=v9', GCC generates code for the V9 variant of the SPARC
13443 architecture. This adds 64-bit integer and floating-point move
13444 instructions, 3 additional floating-point condition code registers
13445 and conditional move instructions. With `-mcpu=ultrasparc', the
13446 compiler additionally optimizes it for the Sun UltraSPARC I/II/IIi
13447 chips. With `-mcpu=ultrasparc3', the compiler additionally
13448 optimizes it for the Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+
13449 chips. With `-mcpu=niagara', the compiler additionally optimizes
13450 it for Sun UltraSPARC T1 chips. With `-mcpu=niagara2', the
13451 compiler additionally optimizes it for Sun UltraSPARC T2 chips.
13454 Set the instruction scheduling parameters for machine type
13455 CPU_TYPE, but do not set the instruction set or register set that
13456 the option `-mcpu=CPU_TYPE' would.
13458 The same values for `-mcpu=CPU_TYPE' can be used for
13459 `-mtune=CPU_TYPE', but the only useful values are those that
13460 select a particular cpu implementation. Those are `cypress',
13461 `supersparc', `hypersparc', `f930', `f934', `sparclite86x',
13462 `tsc701', `ultrasparc', `ultrasparc3', `niagara', and `niagara2'.
13466 With `-mv8plus', GCC generates code for the SPARC-V8+ ABI. The
13467 difference from the V8 ABI is that the global and out registers are
13468 considered 64-bit wide. This is enabled by default on Solaris in
13469 32-bit mode for all SPARC-V9 processors.
13473 With `-mvis', GCC generates code that takes advantage of the
13474 UltraSPARC Visual Instruction Set extensions. The default is
13477 These `-m' options are supported in addition to the above on SPARC-V9
13478 processors in 64-bit environments:
13481 Generate code for a processor running in little-endian mode. It
13482 is only available for a few configurations and most notably not on
13487 Generate code for a 32-bit or 64-bit environment. The 32-bit
13488 environment sets int, long and pointer to 32 bits. The 64-bit
13489 environment sets int to 32 bits and long and pointer to 64 bits.
13492 Generate code for the Medium/Low code model: 64-bit addresses,
13493 programs must be linked in the low 32 bits of memory. Programs
13494 can be statically or dynamically linked.
13497 Generate code for the Medium/Middle code model: 64-bit addresses,
13498 programs must be linked in the low 44 bits of memory, the text and
13499 data segments must be less than 2GB in size and the data segment
13500 must be located within 2GB of the text segment.
13503 Generate code for the Medium/Anywhere code model: 64-bit
13504 addresses, programs may be linked anywhere in memory, the text and
13505 data segments must be less than 2GB in size and the data segment
13506 must be located within 2GB of the text segment.
13508 `-mcmodel=embmedany'
13509 Generate code for the Medium/Anywhere code model for embedded
13510 systems: 64-bit addresses, the text and data segments must be less
13511 than 2GB in size, both starting anywhere in memory (determined at
13512 link time). The global register %g4 points to the base of the
13513 data segment. Programs are statically linked and PIC is not
13518 With `-mstack-bias', GCC assumes that the stack pointer, and frame
13519 pointer if present, are offset by -2047 which must be added back
13520 when making stack frame references. This is the default in 64-bit
13521 mode. Otherwise, assume no such offset is present.
13523 These switches are supported in addition to the above on Solaris:
13526 Add support for multithreading using the Solaris threads library.
13527 This option sets flags for both the preprocessor and linker. This
13528 option does not affect the thread safety of object code produced
13529 by the compiler or that of libraries supplied with it.
13532 Add support for multithreading using the POSIX threads library.
13533 This option sets flags for both the preprocessor and linker. This
13534 option does not affect the thread safety of object code produced
13535 by the compiler or that of libraries supplied with it.
13538 This is a synonym for `-pthreads'.
13541 File: gcc.info, Node: SPU Options, Next: System V Options, Prev: SPARC Options, Up: Submodel Options
13543 3.17.32 SPU Options
13544 -------------------
13546 These `-m' options are supported on the SPU:
13550 The loader for SPU does not handle dynamic relocations. By
13551 default, GCC will give an error when it generates code that
13552 requires a dynamic relocation. `-mno-error-reloc' disables the
13553 error, `-mwarn-reloc' will generate a warning instead.
13557 Instructions which initiate or test completion of DMA must not be
13558 reordered with respect to loads and stores of the memory which is
13559 being accessed. Users typically address this problem using the
13560 volatile keyword, but that can lead to inefficient code in places
13561 where the memory is known to not change. Rather than mark the
13562 memory as volatile we treat the DMA instructions as potentially
13563 effecting all memory. With `-munsafe-dma' users must use the
13564 volatile keyword to protect memory accesses.
13567 By default, GCC will generate a branch hint instruction to avoid
13568 pipeline stalls for always taken or probably taken branches. A
13569 hint will not be generated closer than 8 instructions away from
13570 its branch. There is little reason to disable them, except for
13571 debugging purposes, or to make an object a little bit smaller.
13575 By default, GCC generates code assuming that addresses are never
13576 larger than 18 bits. With `-mlarge-mem' code is generated that
13577 assumes a full 32 bit address.
13580 By default, GCC links against startup code that assumes the
13581 SPU-style main function interface (which has an unconventional
13582 parameter list). With `-mstdmain', GCC will link your program
13583 against startup code that assumes a C99-style interface to `main',
13584 including a local copy of `argv' strings.
13586 `-mfixed-range=REGISTER-RANGE'
13587 Generate code treating the given register range as fixed registers.
13588 A fixed register is one that the register allocator can not use.
13589 This is useful when compiling kernel code. A register range is
13590 specified as two registers separated by a dash. Multiple register
13591 ranges can be specified separated by a comma.
13595 File: gcc.info, Node: System V Options, Next: V850 Options, Prev: SPU Options, Up: Submodel Options
13597 3.17.33 Options for System V
13598 ----------------------------
13600 These additional options are available on System V Release 4 for
13601 compatibility with other compilers on those systems:
13604 Create a shared object. It is recommended that `-symbolic' or
13605 `-shared' be used instead.
13608 Identify the versions of each tool used by the compiler, in a
13609 `.ident' assembler directive in the output.
13612 Refrain from adding `.ident' directives to the output file (this is
13616 Search the directories DIRS, and no others, for libraries
13617 specified with `-l'.
13620 Look in the directory DIR to find the M4 preprocessor. The
13621 assembler uses this option.
13624 File: gcc.info, Node: V850 Options, Next: VAX Options, Prev: System V Options, Up: Submodel Options
13626 3.17.34 V850 Options
13627 --------------------
13629 These `-m' options are defined for V850 implementations:
13633 Treat all calls as being far away (near). If calls are assumed to
13634 be far away, the compiler will always load the functions address
13635 up into a register, and call indirect through the pointer.
13639 Do not optimize (do optimize) basic blocks that use the same index
13640 pointer 4 or more times to copy pointer into the `ep' register, and
13641 use the shorter `sld' and `sst' instructions. The `-mep' option
13642 is on by default if you optimize.
13644 `-mno-prolog-function'
13645 `-mprolog-function'
13646 Do not use (do use) external functions to save and restore
13647 registers at the prologue and epilogue of a function. The
13648 external functions are slower, but use less code space if more
13649 than one function saves the same number of registers. The
13650 `-mprolog-function' option is on by default if you optimize.
13653 Try to make the code as small as possible. At present, this just
13654 turns on the `-mep' and `-mprolog-function' options.
13657 Put static or global variables whose size is N bytes or less into
13658 the tiny data area that register `ep' points to. The tiny data
13659 area can hold up to 256 bytes in total (128 bytes for byte
13663 Put static or global variables whose size is N bytes or less into
13664 the small data area that register `gp' points to. The small data
13665 area can hold up to 64 kilobytes.
13668 Put static or global variables whose size is N bytes or less into
13669 the first 32 kilobytes of memory.
13672 Specify that the target processor is the V850.
13675 Generate code suitable for big switch tables. Use this option
13676 only if the assembler/linker complain about out of range branches
13677 within a switch table.
13680 This option will cause r2 and r5 to be used in the code generated
13681 by the compiler. This setting is the default.
13684 This option will cause r2 and r5 to be treated as fixed registers.
13687 Specify that the target processor is the V850E1. The preprocessor
13688 constants `__v850e1__' and `__v850e__' will be defined if this
13692 Specify that the target processor is the V850E. The preprocessor
13693 constant `__v850e__' will be defined if this option is used.
13695 If neither `-mv850' nor `-mv850e' nor `-mv850e1' are defined then
13696 a default target processor will be chosen and the relevant
13697 `__v850*__' preprocessor constant will be defined.
13699 The preprocessor constants `__v850' and `__v851__' are always
13700 defined, regardless of which processor variant is the target.
13703 This option will suppress generation of the CALLT instruction for
13704 the v850e and v850e1 flavors of the v850 architecture. The
13705 default is `-mno-disable-callt' which allows the CALLT instruction
13710 File: gcc.info, Node: VAX Options, Next: VxWorks Options, Prev: V850 Options, Up: Submodel Options
13712 3.17.35 VAX Options
13713 -------------------
13715 These `-m' options are defined for the VAX:
13718 Do not output certain jump instructions (`aobleq' and so on) that
13719 the Unix assembler for the VAX cannot handle across long ranges.
13722 Do output those jump instructions, on the assumption that you will
13723 assemble with the GNU assembler.
13726 Output code for g-format floating point numbers instead of
13730 File: gcc.info, Node: VxWorks Options, Next: x86-64 Options, Prev: VAX Options, Up: Submodel Options
13732 3.17.36 VxWorks Options
13733 -----------------------
13735 The options in this section are defined for all VxWorks targets.
13736 Options specific to the target hardware are listed with the other
13737 options for that target.
13740 GCC can generate code for both VxWorks kernels and real time
13741 processes (RTPs). This option switches from the former to the
13742 latter. It also defines the preprocessor macro `__RTP__'.
13745 Link an RTP executable against shared libraries rather than static
13746 libraries. The options `-static' and `-shared' can also be used
13747 for RTPs (*note Link Options::); `-static' is the default.
13751 These options are passed down to the linker. They are defined for
13752 compatibility with Diab.
13755 Enable lazy binding of function calls. This option is equivalent
13756 to `-Wl,-z,now' and is defined for compatibility with Diab.
13759 Disable lazy binding of function calls. This option is the
13760 default and is defined for compatibility with Diab.
13763 File: gcc.info, Node: x86-64 Options, Next: Xstormy16 Options, Prev: VxWorks Options, Up: Submodel Options
13765 3.17.37 x86-64 Options
13766 ----------------------
13768 These are listed under *Note i386 and x86-64 Options::.
13771 File: gcc.info, Node: Xstormy16 Options, Next: Xtensa Options, Prev: x86-64 Options, Up: Submodel Options
13773 3.17.38 Xstormy16 Options
13774 -------------------------
13776 These options are defined for Xstormy16:
13779 Choose startup files and linker script suitable for the simulator.
13782 File: gcc.info, Node: Xtensa Options, Next: zSeries Options, Prev: Xstormy16 Options, Up: Submodel Options
13784 3.17.39 Xtensa Options
13785 ----------------------
13787 These options are supported for Xtensa targets:
13791 Enable or disable use of `CONST16' instructions for loading
13792 constant values. The `CONST16' instruction is currently not a
13793 standard option from Tensilica. When enabled, `CONST16'
13794 instructions are always used in place of the standard `L32R'
13795 instructions. The use of `CONST16' is enabled by default only if
13796 the `L32R' instruction is not available.
13800 Enable or disable use of fused multiply/add and multiply/subtract
13801 instructions in the floating-point option. This has no effect if
13802 the floating-point option is not also enabled. Disabling fused
13803 multiply/add and multiply/subtract instructions forces the
13804 compiler to use separate instructions for the multiply and
13805 add/subtract operations. This may be desirable in some cases
13806 where strict IEEE 754-compliant results are required: the fused
13807 multiply add/subtract instructions do not round the intermediate
13808 result, thereby producing results with _more_ bits of precision
13809 than specified by the IEEE standard. Disabling fused multiply
13810 add/subtract instructions also ensures that the program output is
13811 not sensitive to the compiler's ability to combine multiply and
13812 add/subtract operations.
13814 `-mtext-section-literals'
13815 `-mno-text-section-literals'
13816 Control the treatment of literal pools. The default is
13817 `-mno-text-section-literals', which places literals in a separate
13818 section in the output file. This allows the literal pool to be
13819 placed in a data RAM/ROM, and it also allows the linker to combine
13820 literal pools from separate object files to remove redundant
13821 literals and improve code size. With `-mtext-section-literals',
13822 the literals are interspersed in the text section in order to keep
13823 them as close as possible to their references. This may be
13824 necessary for large assembly files.
13827 `-mno-target-align'
13828 When this option is enabled, GCC instructs the assembler to
13829 automatically align instructions to reduce branch penalties at the
13830 expense of some code density. The assembler attempts to widen
13831 density instructions to align branch targets and the instructions
13832 following call instructions. If there are not enough preceding
13833 safe density instructions to align a target, no widening will be
13834 performed. The default is `-mtarget-align'. These options do not
13835 affect the treatment of auto-aligned instructions like `LOOP',
13836 which the assembler will always align, either by widening density
13837 instructions or by inserting no-op instructions.
13841 When this option is enabled, GCC instructs the assembler to
13842 translate direct calls to indirect calls unless it can determine
13843 that the target of a direct call is in the range allowed by the
13844 call instruction. This translation typically occurs for calls to
13845 functions in other source files. Specifically, the assembler
13846 translates a direct `CALL' instruction into an `L32R' followed by
13847 a `CALLX' instruction. The default is `-mno-longcalls'. This
13848 option should be used in programs where the call target can
13849 potentially be out of range. This option is implemented in the
13850 assembler, not the compiler, so the assembly code generated by GCC
13851 will still show direct call instructions--look at the disassembled
13852 object code to see the actual instructions. Note that the
13853 assembler will use an indirect call for every cross-file call, not
13854 just those that really will be out of range.
13857 File: gcc.info, Node: zSeries Options, Prev: Xtensa Options, Up: Submodel Options
13859 3.17.40 zSeries Options
13860 -----------------------
13862 These are listed under *Note S/390 and zSeries Options::.
13865 File: gcc.info, Node: Code Gen Options, Next: Environment Variables, Prev: Submodel Options, Up: Invoking GCC
13867 3.18 Options for Code Generation Conventions
13868 ============================================
13870 These machine-independent options control the interface conventions
13871 used in code generation.
13873 Most of them have both positive and negative forms; the negative form
13874 of `-ffoo' would be `-fno-foo'. In the table below, only one of the
13875 forms is listed--the one which is not the default. You can figure out
13876 the other form by either removing `no-' or adding it.
13879 For front-ends that support it, generate additional code to check
13880 that indices used to access arrays are within the declared range.
13881 This is currently only supported by the Java and Fortran
13882 front-ends, where this option defaults to true and false
13886 This option generates traps for signed overflow on addition,
13887 subtraction, multiplication operations.
13890 This option instructs the compiler to assume that signed arithmetic
13891 overflow of addition, subtraction and multiplication wraps around
13892 using twos-complement representation. This flag enables some
13893 optimizations and disables others. This option is enabled by
13894 default for the Java front-end, as required by the Java language
13898 Enable exception handling. Generates extra code needed to
13899 propagate exceptions. For some targets, this implies GCC will
13900 generate frame unwind information for all functions, which can
13901 produce significant data size overhead, although it does not
13902 affect execution. If you do not specify this option, GCC will
13903 enable it by default for languages like C++ which normally require
13904 exception handling, and disable it for languages like C that do
13905 not normally require it. However, you may need to enable this
13906 option when compiling C code that needs to interoperate properly
13907 with exception handlers written in C++. You may also wish to
13908 disable this option if you are compiling older C++ programs that
13909 don't use exception handling.
13911 `-fnon-call-exceptions'
13912 Generate code that allows trapping instructions to throw
13913 exceptions. Note that this requires platform-specific runtime
13914 support that does not exist everywhere. Moreover, it only allows
13915 _trapping_ instructions to throw exceptions, i.e. memory
13916 references or floating point instructions. It does not allow
13917 exceptions to be thrown from arbitrary signal handlers such as
13921 Similar to `-fexceptions', except that it will just generate any
13922 needed static data, but will not affect the generated code in any
13923 other way. You will normally not enable this option; instead, a
13924 language processor that needs this handling would enable it on
13927 `-fasynchronous-unwind-tables'
13928 Generate unwind table in dwarf2 format, if supported by target
13929 machine. The table is exact at each instruction boundary, so it
13930 can be used for stack unwinding from asynchronous events (such as
13931 debugger or garbage collector).
13933 `-fpcc-struct-return'
13934 Return "short" `struct' and `union' values in memory like longer
13935 ones, rather than in registers. This convention is less
13936 efficient, but it has the advantage of allowing intercallability
13937 between GCC-compiled files and files compiled with other
13938 compilers, particularly the Portable C Compiler (pcc).
13940 The precise convention for returning structures in memory depends
13941 on the target configuration macros.
13943 Short structures and unions are those whose size and alignment
13944 match that of some integer type.
13946 *Warning:* code compiled with the `-fpcc-struct-return' switch is
13947 not binary compatible with code compiled with the
13948 `-freg-struct-return' switch. Use it to conform to a non-default
13949 application binary interface.
13951 `-freg-struct-return'
13952 Return `struct' and `union' values in registers when possible.
13953 This is more efficient for small structures than
13954 `-fpcc-struct-return'.
13956 If you specify neither `-fpcc-struct-return' nor
13957 `-freg-struct-return', GCC defaults to whichever convention is
13958 standard for the target. If there is no standard convention, GCC
13959 defaults to `-fpcc-struct-return', except on targets where GCC is
13960 the principal compiler. In those cases, we can choose the
13961 standard, and we chose the more efficient register return
13964 *Warning:* code compiled with the `-freg-struct-return' switch is
13965 not binary compatible with code compiled with the
13966 `-fpcc-struct-return' switch. Use it to conform to a non-default
13967 application binary interface.
13970 Allocate to an `enum' type only as many bytes as it needs for the
13971 declared range of possible values. Specifically, the `enum' type
13972 will be equivalent to the smallest integer type which has enough
13975 *Warning:* the `-fshort-enums' switch causes GCC to generate code
13976 that is not binary compatible with code generated without that
13977 switch. Use it to conform to a non-default application binary
13981 Use the same size for `double' as for `float'.
13983 *Warning:* the `-fshort-double' switch causes GCC to generate code
13984 that is not binary compatible with code generated without that
13985 switch. Use it to conform to a non-default application binary
13989 Override the underlying type for `wchar_t' to be `short unsigned
13990 int' instead of the default for the target. This option is useful
13991 for building programs to run under WINE.
13993 *Warning:* the `-fshort-wchar' switch causes GCC to generate code
13994 that is not binary compatible with code generated without that
13995 switch. Use it to conform to a non-default application binary
13999 In C, allocate even uninitialized global variables in the data
14000 section of the object file, rather than generating them as common
14001 blocks. This has the effect that if the same variable is declared
14002 (without `extern') in two different compilations, you will get an
14003 error when you link them. The only reason this might be useful is
14004 if you wish to verify that the program will work on other systems
14005 which always work this way.
14008 Ignore the `#ident' directive.
14010 `-finhibit-size-directive'
14011 Don't output a `.size' assembler directive, or anything else that
14012 would cause trouble if the function is split in the middle, and the
14013 two halves are placed at locations far apart in memory. This
14014 option is used when compiling `crtstuff.c'; you should not need to
14015 use it for anything else.
14018 Put extra commentary information in the generated assembly code to
14019 make it more readable. This option is generally only of use to
14020 those who actually need to read the generated assembly code
14021 (perhaps while debugging the compiler itself).
14023 `-fno-verbose-asm', the default, causes the extra information to
14024 be omitted and is useful when comparing two assembler files.
14026 `-frecord-gcc-switches'
14027 This switch causes the command line that was used to invoke the
14028 compiler to be recorded into the object file that is being created.
14029 This switch is only implemented on some targets and the exact
14030 format of the recording is target and binary file format
14031 dependent, but it usually takes the form of a section containing
14032 ASCII text. This switch is related to the `-fverbose-asm' switch,
14033 but that switch only records information in the assembler output
14034 file as comments, so it never reaches the object file.
14037 Generate position-independent code (PIC) suitable for use in a
14038 shared library, if supported for the target machine. Such code
14039 accesses all constant addresses through a global offset table
14040 (GOT). The dynamic loader resolves the GOT entries when the
14041 program starts (the dynamic loader is not part of GCC; it is part
14042 of the operating system). If the GOT size for the linked
14043 executable exceeds a machine-specific maximum size, you get an
14044 error message from the linker indicating that `-fpic' does not
14045 work; in that case, recompile with `-fPIC' instead. (These
14046 maximums are 8k on the SPARC and 32k on the m68k and RS/6000. The
14047 386 has no such limit.)
14049 Position-independent code requires special support, and therefore
14050 works only on certain machines. For the 386, GCC supports PIC for
14051 System V but not for the Sun 386i. Code generated for the IBM
14052 RS/6000 is always position-independent.
14054 When this flag is set, the macros `__pic__' and `__PIC__' are
14058 If supported for the target machine, emit position-independent
14059 code, suitable for dynamic linking and avoiding any limit on the
14060 size of the global offset table. This option makes a difference
14061 on the m68k, PowerPC and SPARC.
14063 Position-independent code requires special support, and therefore
14064 works only on certain machines.
14066 When this flag is set, the macros `__pic__' and `__PIC__' are
14071 These options are similar to `-fpic' and `-fPIC', but generated
14072 position independent code can be only linked into executables.
14073 Usually these options are used when `-pie' GCC option will be used
14076 `-fpie' and `-fPIE' both define the macros `__pie__' and
14077 `__PIE__'. The macros have the value 1 for `-fpie' and 2 for
14081 Do not use jump tables for switch statements even where it would be
14082 more efficient than other code generation strategies. This option
14083 is of use in conjunction with `-fpic' or `-fPIC' for building code
14084 which forms part of a dynamic linker and cannot reference the
14085 address of a jump table. On some targets, jump tables do not
14086 require a GOT and this option is not needed.
14089 Treat the register named REG as a fixed register; generated code
14090 should never refer to it (except perhaps as a stack pointer, frame
14091 pointer or in some other fixed role).
14093 REG must be the name of a register. The register names accepted
14094 are machine-specific and are defined in the `REGISTER_NAMES' macro
14095 in the machine description macro file.
14097 This flag does not have a negative form, because it specifies a
14101 Treat the register named REG as an allocable register that is
14102 clobbered by function calls. It may be allocated for temporaries
14103 or variables that do not live across a call. Functions compiled
14104 this way will not save and restore the register REG.
14106 It is an error to used this flag with the frame pointer or stack
14107 pointer. Use of this flag for other registers that have fixed
14108 pervasive roles in the machine's execution model will produce
14109 disastrous results.
14111 This flag does not have a negative form, because it specifies a
14115 Treat the register named REG as an allocable register saved by
14116 functions. It may be allocated even for temporaries or variables
14117 that live across a call. Functions compiled this way will save
14118 and restore the register REG if they use it.
14120 It is an error to used this flag with the frame pointer or stack
14121 pointer. Use of this flag for other registers that have fixed
14122 pervasive roles in the machine's execution model will produce
14123 disastrous results.
14125 A different sort of disaster will result from the use of this flag
14126 for a register in which function values may be returned.
14128 This flag does not have a negative form, because it specifies a
14131 `-fpack-struct[=N]'
14132 Without a value specified, pack all structure members together
14133 without holes. When a value is specified (which must be a small
14134 power of two), pack structure members according to this value,
14135 representing the maximum alignment (that is, objects with default
14136 alignment requirements larger than this will be output potentially
14137 unaligned at the next fitting location.
14139 *Warning:* the `-fpack-struct' switch causes GCC to generate code
14140 that is not binary compatible with code generated without that
14141 switch. Additionally, it makes the code suboptimal. Use it to
14142 conform to a non-default application binary interface.
14144 `-finstrument-functions'
14145 Generate instrumentation calls for entry and exit to functions.
14146 Just after function entry and just before function exit, the
14147 following profiling functions will be called with the address of
14148 the current function and its call site. (On some platforms,
14149 `__builtin_return_address' does not work beyond the current
14150 function, so the call site information may not be available to the
14151 profiling functions otherwise.)
14153 void __cyg_profile_func_enter (void *this_fn,
14155 void __cyg_profile_func_exit (void *this_fn,
14158 The first argument is the address of the start of the current
14159 function, which may be looked up exactly in the symbol table.
14161 This instrumentation is also done for functions expanded inline in
14162 other functions. The profiling calls will indicate where,
14163 conceptually, the inline function is entered and exited. This
14164 means that addressable versions of such functions must be
14165 available. If all your uses of a function are expanded inline,
14166 this may mean an additional expansion of code size. If you use
14167 `extern inline' in your C code, an addressable version of such
14168 functions must be provided. (This is normally the case anyways,
14169 but if you get lucky and the optimizer always expands the
14170 functions inline, you might have gotten away without providing
14173 A function may be given the attribute `no_instrument_function', in
14174 which case this instrumentation will not be done. This can be
14175 used, for example, for the profiling functions listed above,
14176 high-priority interrupt routines, and any functions from which the
14177 profiling functions cannot safely be called (perhaps signal
14178 handlers, if the profiling routines generate output or allocate
14181 `-finstrument-functions-exclude-file-list=FILE,FILE,...'
14182 Set the list of functions that are excluded from instrumentation
14183 (see the description of `-finstrument-functions'). If the file
14184 that contains a function definition matches with one of FILE, then
14185 that function is not instrumented. The match is done on
14186 substrings: if the FILE parameter is a substring of the file name,
14187 it is considered to be a match.
14190 `-finstrument-functions-exclude-file-list=/bits/stl,include/sys'
14191 will exclude any inline function defined in files whose pathnames
14192 contain `/bits/stl' or `include/sys'.
14194 If, for some reason, you want to include letter `','' in one of
14195 SYM, write `'\,''. For example,
14196 `-finstrument-functions-exclude-file-list='\,\,tmp'' (note the
14197 single quote surrounding the option).
14199 `-finstrument-functions-exclude-function-list=SYM,SYM,...'
14200 This is similar to `-finstrument-functions-exclude-file-list', but
14201 this option sets the list of function names to be excluded from
14202 instrumentation. The function name to be matched is its
14203 user-visible name, such as `vector<int> blah(const vector<int>
14204 &)', not the internal mangled name (e.g.,
14205 `_Z4blahRSt6vectorIiSaIiEE'). The match is done on substrings: if
14206 the SYM parameter is a substring of the function name, it is
14207 considered to be a match.
14210 Generate code to verify that you do not go beyond the boundary of
14211 the stack. You should specify this flag if you are running in an
14212 environment with multiple threads, but only rarely need to specify
14213 it in a single-threaded environment since stack overflow is
14214 automatically detected on nearly all systems if there is only one
14217 Note that this switch does not actually cause checking to be done;
14218 the operating system must do that. The switch causes generation
14219 of code to ensure that the operating system sees the stack being
14222 `-fstack-limit-register=REG'
14223 `-fstack-limit-symbol=SYM'
14225 Generate code to ensure that the stack does not grow beyond a
14226 certain value, either the value of a register or the address of a
14227 symbol. If the stack would grow beyond the value, a signal is
14228 raised. For most targets, the signal is raised before the stack
14229 overruns the boundary, so it is possible to catch the signal
14230 without taking special precautions.
14232 For instance, if the stack starts at absolute address `0x80000000'
14233 and grows downwards, you can use the flags
14234 `-fstack-limit-symbol=__stack_limit' and
14235 `-Wl,--defsym,__stack_limit=0x7ffe0000' to enforce a stack limit
14236 of 128KB. Note that this may only work with the GNU linker.
14239 `-fargument-noalias'
14240 `-fargument-noalias-global'
14241 `-fargument-noalias-anything'
14242 Specify the possible relationships among parameters and between
14243 parameters and global data.
14245 `-fargument-alias' specifies that arguments (parameters) may alias
14246 each other and may alias global storage.
14247 `-fargument-noalias' specifies that arguments do not alias each
14248 other, but may alias global storage.
14249 `-fargument-noalias-global' specifies that arguments do not alias
14250 each other and do not alias global storage.
14251 `-fargument-noalias-anything' specifies that arguments do not
14252 alias any other storage.
14254 Each language will automatically use whatever option is required by
14255 the language standard. You should not need to use these options
14258 `-fleading-underscore'
14259 This option and its counterpart, `-fno-leading-underscore',
14260 forcibly change the way C symbols are represented in the object
14261 file. One use is to help link with legacy assembly code.
14263 *Warning:* the `-fleading-underscore' switch causes GCC to
14264 generate code that is not binary compatible with code generated
14265 without that switch. Use it to conform to a non-default
14266 application binary interface. Not all targets provide complete
14267 support for this switch.
14269 `-ftls-model=MODEL'
14270 Alter the thread-local storage model to be used (*note
14271 Thread-Local::). The MODEL argument should be one of
14272 `global-dynamic', `local-dynamic', `initial-exec' or `local-exec'.
14274 The default without `-fpic' is `initial-exec'; with `-fpic' the
14275 default is `global-dynamic'.
14277 `-fvisibility=DEFAULT|INTERNAL|HIDDEN|PROTECTED'
14278 Set the default ELF image symbol visibility to the specified
14279 option--all symbols will be marked with this unless overridden
14280 within the code. Using this feature can very substantially
14281 improve linking and load times of shared object libraries, produce
14282 more optimized code, provide near-perfect API export and prevent
14283 symbol clashes. It is *strongly* recommended that you use this in
14284 any shared objects you distribute.
14286 Despite the nomenclature, `default' always means public ie;
14287 available to be linked against from outside the shared object.
14288 `protected' and `internal' are pretty useless in real-world usage
14289 so the only other commonly used option will be `hidden'. The
14290 default if `-fvisibility' isn't specified is `default', i.e., make
14291 every symbol public--this causes the same behavior as previous
14294 A good explanation of the benefits offered by ensuring ELF symbols
14295 have the correct visibility is given by "How To Write Shared
14296 Libraries" by Ulrich Drepper (which can be found at
14297 `http://people.redhat.com/~drepper/')--however a superior solution
14298 made possible by this option to marking things hidden when the
14299 default is public is to make the default hidden and mark things
14300 public. This is the norm with DLL's on Windows and with
14301 `-fvisibility=hidden' and `__attribute__
14302 ((visibility("default")))' instead of `__declspec(dllexport)' you
14303 get almost identical semantics with identical syntax. This is a
14304 great boon to those working with cross-platform projects.
14306 For those adding visibility support to existing code, you may find
14307 `#pragma GCC visibility' of use. This works by you enclosing the
14308 declarations you wish to set visibility for with (for example)
14309 `#pragma GCC visibility push(hidden)' and `#pragma GCC visibility
14310 pop'. Bear in mind that symbol visibility should be viewed *as
14311 part of the API interface contract* and thus all new code should
14312 always specify visibility when it is not the default ie;
14313 declarations only for use within the local DSO should *always* be
14314 marked explicitly as hidden as so to avoid PLT indirection
14315 overheads--making this abundantly clear also aids readability and
14316 self-documentation of the code. Note that due to ISO C++
14317 specification requirements, operator new and operator delete must
14318 always be of default visibility.
14320 Be aware that headers from outside your project, in particular
14321 system headers and headers from any other library you use, may not
14322 be expecting to be compiled with visibility other than the
14323 default. You may need to explicitly say `#pragma GCC visibility
14324 push(default)' before including any such headers.
14326 `extern' declarations are not affected by `-fvisibility', so a lot
14327 of code can be recompiled with `-fvisibility=hidden' with no
14328 modifications. However, this means that calls to `extern'
14329 functions with no explicit visibility will use the PLT, so it is
14330 more effective to use `__attribute ((visibility))' and/or `#pragma
14331 GCC visibility' to tell the compiler which `extern' declarations
14332 should be treated as hidden.
14334 Note that `-fvisibility' does affect C++ vague linkage entities.
14335 This means that, for instance, an exception class that will be
14336 thrown between DSOs must be explicitly marked with default
14337 visibility so that the `type_info' nodes will be unified between
14340 An overview of these techniques, their benefits and how to use them
14341 is at `http://gcc.gnu.org/wiki/Visibility'.
14345 File: gcc.info, Node: Environment Variables, Next: Precompiled Headers, Prev: Code Gen Options, Up: Invoking GCC
14347 3.19 Environment Variables Affecting GCC
14348 ========================================
14350 This section describes several environment variables that affect how GCC
14351 operates. Some of them work by specifying directories or prefixes to
14352 use when searching for various kinds of files. Some are used to
14353 specify other aspects of the compilation environment.
14355 Note that you can also specify places to search using options such as
14356 `-B', `-I' and `-L' (*note Directory Options::). These take precedence
14357 over places specified using environment variables, which in turn take
14358 precedence over those specified by the configuration of GCC. *Note
14359 Controlling the Compilation Driver `gcc': (gccint)Driver.
14365 These environment variables control the way that GCC uses
14366 localization information that allow GCC to work with different
14367 national conventions. GCC inspects the locale categories
14368 `LC_CTYPE' and `LC_MESSAGES' if it has been configured to do so.
14369 These locale categories can be set to any value supported by your
14370 installation. A typical value is `en_GB.UTF-8' for English in the
14371 United Kingdom encoded in UTF-8.
14373 The `LC_CTYPE' environment variable specifies character
14374 classification. GCC uses it to determine the character boundaries
14375 in a string; this is needed for some multibyte encodings that
14376 contain quote and escape characters that would otherwise be
14377 interpreted as a string end or escape.
14379 The `LC_MESSAGES' environment variable specifies the language to
14380 use in diagnostic messages.
14382 If the `LC_ALL' environment variable is set, it overrides the value
14383 of `LC_CTYPE' and `LC_MESSAGES'; otherwise, `LC_CTYPE' and
14384 `LC_MESSAGES' default to the value of the `LANG' environment
14385 variable. If none of these variables are set, GCC defaults to
14386 traditional C English behavior.
14389 If `TMPDIR' is set, it specifies the directory to use for temporary
14390 files. GCC uses temporary files to hold the output of one stage of
14391 compilation which is to be used as input to the next stage: for
14392 example, the output of the preprocessor, which is the input to the
14396 If `GCC_EXEC_PREFIX' is set, it specifies a prefix to use in the
14397 names of the subprograms executed by the compiler. No slash is
14398 added when this prefix is combined with the name of a subprogram,
14399 but you can specify a prefix that ends with a slash if you wish.
14401 If `GCC_EXEC_PREFIX' is not set, GCC will attempt to figure out an
14402 appropriate prefix to use based on the pathname it was invoked
14405 If GCC cannot find the subprogram using the specified prefix, it
14406 tries looking in the usual places for the subprogram.
14408 The default value of `GCC_EXEC_PREFIX' is `PREFIX/lib/gcc/' where
14409 PREFIX is the prefix to the installed compiler. In many cases
14410 PREFIX is the value of `prefix' when you ran the `configure'
14413 Other prefixes specified with `-B' take precedence over this
14416 This prefix is also used for finding files such as `crt0.o' that
14417 are used for linking.
14419 In addition, the prefix is used in an unusual way in finding the
14420 directories to search for header files. For each of the standard
14421 directories whose name normally begins with `/usr/local/lib/gcc'
14422 (more precisely, with the value of `GCC_INCLUDE_DIR'), GCC tries
14423 replacing that beginning with the specified prefix to produce an
14424 alternate directory name. Thus, with `-Bfoo/', GCC will search
14425 `foo/bar' where it would normally search `/usr/local/lib/bar'.
14426 These alternate directories are searched first; the standard
14427 directories come next. If a standard directory begins with the
14428 configured PREFIX then the value of PREFIX is replaced by
14429 `GCC_EXEC_PREFIX' when looking for header files.
14432 The value of `COMPILER_PATH' is a colon-separated list of
14433 directories, much like `PATH'. GCC tries the directories thus
14434 specified when searching for subprograms, if it can't find the
14435 subprograms using `GCC_EXEC_PREFIX'.
14438 The value of `LIBRARY_PATH' is a colon-separated list of
14439 directories, much like `PATH'. When configured as a native
14440 compiler, GCC tries the directories thus specified when searching
14441 for special linker files, if it can't find them using
14442 `GCC_EXEC_PREFIX'. Linking using GCC also uses these directories
14443 when searching for ordinary libraries for the `-l' option (but
14444 directories specified with `-L' come first).
14447 This variable is used to pass locale information to the compiler.
14448 One way in which this information is used is to determine the
14449 character set to be used when character literals, string literals
14450 and comments are parsed in C and C++. When the compiler is
14451 configured to allow multibyte characters, the following values for
14452 `LANG' are recognized:
14455 Recognize JIS characters.
14458 Recognize SJIS characters.
14461 Recognize EUCJP characters.
14463 If `LANG' is not defined, or if it has some other value, then the
14464 compiler will use mblen and mbtowc as defined by the default
14465 locale to recognize and translate multibyte characters.
14467 Some additional environments variables affect the behavior of the
14472 `CPLUS_INCLUDE_PATH'
14473 `OBJC_INCLUDE_PATH'
14474 Each variable's value is a list of directories separated by a
14475 special character, much like `PATH', in which to look for header
14476 files. The special character, `PATH_SEPARATOR', is
14477 target-dependent and determined at GCC build time. For Microsoft
14478 Windows-based targets it is a semicolon, and for almost all other
14479 targets it is a colon.
14481 `CPATH' specifies a list of directories to be searched as if
14482 specified with `-I', but after any paths given with `-I' options
14483 on the command line. This environment variable is used regardless
14484 of which language is being preprocessed.
14486 The remaining environment variables apply only when preprocessing
14487 the particular language indicated. Each specifies a list of
14488 directories to be searched as if specified with `-isystem', but
14489 after any paths given with `-isystem' options on the command line.
14491 In all these variables, an empty element instructs the compiler to
14492 search its current working directory. Empty elements can appear
14493 at the beginning or end of a path. For instance, if the value of
14494 `CPATH' is `:/special/include', that has the same effect as
14495 `-I. -I/special/include'.
14497 `DEPENDENCIES_OUTPUT'
14498 If this variable is set, its value specifies how to output
14499 dependencies for Make based on the non-system header files
14500 processed by the compiler. System header files are ignored in the
14503 The value of `DEPENDENCIES_OUTPUT' can be just a file name, in
14504 which case the Make rules are written to that file, guessing the
14505 target name from the source file name. Or the value can have the
14506 form `FILE TARGET', in which case the rules are written to file
14507 FILE using TARGET as the target name.
14509 In other words, this environment variable is equivalent to
14510 combining the options `-MM' and `-MF' (*note Preprocessor
14511 Options::), with an optional `-MT' switch too.
14513 `SUNPRO_DEPENDENCIES'
14514 This variable is the same as `DEPENDENCIES_OUTPUT' (see above),
14515 except that system header files are not ignored, so it implies
14516 `-M' rather than `-MM'. However, the dependence on the main input
14517 file is omitted. *Note Preprocessor Options::.
14520 File: gcc.info, Node: Precompiled Headers, Next: Running Protoize, Prev: Environment Variables, Up: Invoking GCC
14522 3.20 Using Precompiled Headers
14523 ==============================
14525 Often large projects have many header files that are included in every
14526 source file. The time the compiler takes to process these header files
14527 over and over again can account for nearly all of the time required to
14528 build the project. To make builds faster, GCC allows users to
14529 `precompile' a header file; then, if builds can use the precompiled
14530 header file they will be much faster.
14532 To create a precompiled header file, simply compile it as you would any
14533 other file, if necessary using the `-x' option to make the driver treat
14534 it as a C or C++ header file. You will probably want to use a tool
14535 like `make' to keep the precompiled header up-to-date when the headers
14536 it contains change.
14538 A precompiled header file will be searched for when `#include' is seen
14539 in the compilation. As it searches for the included file (*note Search
14540 Path: (cpp)Search Path.) the compiler looks for a precompiled header in
14541 each directory just before it looks for the include file in that
14542 directory. The name searched for is the name specified in the
14543 `#include' with `.gch' appended. If the precompiled header file can't
14544 be used, it is ignored.
14546 For instance, if you have `#include "all.h"', and you have `all.h.gch'
14547 in the same directory as `all.h', then the precompiled header file will
14548 be used if possible, and the original header will be used otherwise.
14550 Alternatively, you might decide to put the precompiled header file in a
14551 directory and use `-I' to ensure that directory is searched before (or
14552 instead of) the directory containing the original header. Then, if you
14553 want to check that the precompiled header file is always used, you can
14554 put a file of the same name as the original header in this directory
14555 containing an `#error' command.
14557 This also works with `-include'. So yet another way to use
14558 precompiled headers, good for projects not designed with precompiled
14559 header files in mind, is to simply take most of the header files used by
14560 a project, include them from another header file, precompile that header
14561 file, and `-include' the precompiled header. If the header files have
14562 guards against multiple inclusion, they will be skipped because they've
14563 already been included (in the precompiled header).
14565 If you need to precompile the same header file for different
14566 languages, targets, or compiler options, you can instead make a
14567 _directory_ named like `all.h.gch', and put each precompiled header in
14568 the directory, perhaps using `-o'. It doesn't matter what you call the
14569 files in the directory, every precompiled header in the directory will
14570 be considered. The first precompiled header encountered in the
14571 directory that is valid for this compilation will be used; they're
14572 searched in no particular order.
14574 There are many other possibilities, limited only by your imagination,
14575 good sense, and the constraints of your build system.
14577 A precompiled header file can be used only when these conditions apply:
14579 * Only one precompiled header can be used in a particular
14582 * A precompiled header can't be used once the first C token is seen.
14583 You can have preprocessor directives before a precompiled header;
14584 you can even include a precompiled header from inside another
14585 header, so long as there are no C tokens before the `#include'.
14587 * The precompiled header file must be produced for the same language
14588 as the current compilation. You can't use a C precompiled header
14589 for a C++ compilation.
14591 * The precompiled header file must have been produced by the same
14592 compiler binary as the current compilation is using.
14594 * Any macros defined before the precompiled header is included must
14595 either be defined in the same way as when the precompiled header
14596 was generated, or must not affect the precompiled header, which
14597 usually means that they don't appear in the precompiled header at
14600 The `-D' option is one way to define a macro before a precompiled
14601 header is included; using a `#define' can also do it. There are
14602 also some options that define macros implicitly, like `-O' and
14603 `-Wdeprecated'; the same rule applies to macros defined this way.
14605 * If debugging information is output when using the precompiled
14606 header, using `-g' or similar, the same kind of debugging
14607 information must have been output when building the precompiled
14608 header. However, a precompiled header built using `-g' can be
14609 used in a compilation when no debugging information is being
14612 * The same `-m' options must generally be used when building and
14613 using the precompiled header. *Note Submodel Options::, for any
14614 cases where this rule is relaxed.
14616 * Each of the following options must be the same when building and
14617 using the precompiled header:
14619 -fexceptions -funit-at-a-time
14621 * Some other command-line options starting with `-f', `-p', or `-O'
14622 must be defined in the same way as when the precompiled header was
14623 generated. At present, it's not clear which options are safe to
14624 change and which are not; the safest choice is to use exactly the
14625 same options when generating and using the precompiled header.
14626 The following are known to be safe:
14628 -fmessage-length= -fpreprocessed -fsched-interblock
14629 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
14630 -fsched-verbose=<number> -fschedule-insns -fvisibility=
14634 For all of these except the last, the compiler will automatically
14635 ignore the precompiled header if the conditions aren't met. If you
14636 find an option combination that doesn't work and doesn't cause the
14637 precompiled header to be ignored, please consider filing a bug report,
14640 If you do use differing options when generating and using the
14641 precompiled header, the actual behavior will be a mixture of the
14642 behavior for the options. For instance, if you use `-g' to generate
14643 the precompiled header but not when using it, you may or may not get
14644 debugging information for routines in the precompiled header.
14647 File: gcc.info, Node: Running Protoize, Prev: Precompiled Headers, Up: Invoking GCC
14649 3.21 Running Protoize
14650 =====================
14652 The program `protoize' is an optional part of GCC. You can use it to
14653 add prototypes to a program, thus converting the program to ISO C in
14654 one respect. The companion program `unprotoize' does the reverse: it
14655 removes argument types from any prototypes that are found.
14657 When you run these programs, you must specify a set of source files as
14658 command line arguments. The conversion programs start out by compiling
14659 these files to see what functions they define. The information gathered
14660 about a file FOO is saved in a file named `FOO.X'.
14662 After scanning comes actual conversion. The specified files are all
14663 eligible to be converted; any files they include (whether sources or
14664 just headers) are eligible as well.
14666 But not all the eligible files are converted. By default, `protoize'
14667 and `unprotoize' convert only source and header files in the current
14668 directory. You can specify additional directories whose files should
14669 be converted with the `-d DIRECTORY' option. You can also specify
14670 particular files to exclude with the `-x FILE' option. A file is
14671 converted if it is eligible, its directory name matches one of the
14672 specified directory names, and its name within the directory has not
14675 Basic conversion with `protoize' consists of rewriting most function
14676 definitions and function declarations to specify the types of the
14677 arguments. The only ones not rewritten are those for varargs functions.
14679 `protoize' optionally inserts prototype declarations at the beginning
14680 of the source file, to make them available for any calls that precede
14681 the function's definition. Or it can insert prototype declarations
14682 with block scope in the blocks where undeclared functions are called.
14684 Basic conversion with `unprotoize' consists of rewriting most function
14685 declarations to remove any argument types, and rewriting function
14686 definitions to the old-style pre-ISO form.
14688 Both conversion programs print a warning for any function declaration
14689 or definition that they can't convert. You can suppress these warnings
14692 The output from `protoize' or `unprotoize' replaces the original
14693 source file. The original file is renamed to a name ending with
14694 `.save' (for DOS, the saved filename ends in `.sav' without the
14695 original `.c' suffix). If the `.save' (`.sav' for DOS) file already
14696 exists, then the source file is simply discarded.
14698 `protoize' and `unprotoize' both depend on GCC itself to scan the
14699 program and collect information about the functions it uses. So
14700 neither of these programs will work until GCC is installed.
14702 Here is a table of the options you can use with `protoize' and
14703 `unprotoize'. Each option works with both programs unless otherwise
14707 Look for the file `SYSCALLS.c.X' in DIRECTORY, instead of the
14708 usual directory (normally `/usr/local/lib'). This file contains
14709 prototype information about standard system functions. This option
14710 applies only to `protoize'.
14712 `-c COMPILATION-OPTIONS'
14713 Use COMPILATION-OPTIONS as the options when running `gcc' to
14714 produce the `.X' files. The special option `-aux-info' is always
14715 passed in addition, to tell `gcc' to write a `.X' file.
14717 Note that the compilation options must be given as a single
14718 argument to `protoize' or `unprotoize'. If you want to specify
14719 several `gcc' options, you must quote the entire set of
14720 compilation options to make them a single word in the shell.
14722 There are certain `gcc' arguments that you cannot use, because they
14723 would produce the wrong kind of output. These include `-g', `-O',
14724 `-c', `-S', and `-o' If you include these in the
14725 COMPILATION-OPTIONS, they are ignored.
14728 Rename files to end in `.C' (`.cc' for DOS-based file systems)
14729 instead of `.c'. This is convenient if you are converting a C
14730 program to C++. This option applies only to `protoize'.
14733 Add explicit global declarations. This means inserting explicit
14734 declarations at the beginning of each source file for each function
14735 that is called in the file and was not declared. These
14736 declarations precede the first function definition that contains a
14737 call to an undeclared function. This option applies only to
14741 Indent old-style parameter declarations with the string STRING.
14742 This option applies only to `protoize'.
14744 `unprotoize' converts prototyped function definitions to old-style
14745 function definitions, where the arguments are declared between the
14746 argument list and the initial `{'. By default, `unprotoize' uses
14747 five spaces as the indentation. If you want to indent with just
14748 one space instead, use `-i " "'.
14751 Keep the `.X' files. Normally, they are deleted after conversion
14755 Add explicit local declarations. `protoize' with `-l' inserts a
14756 prototype declaration for each function in each block which calls
14757 the function without any declaration. This option applies only to
14761 Make no real changes. This mode just prints information about the
14762 conversions that would have been done without `-n'.
14765 Make no `.save' files. The original files are simply deleted.
14766 Use this option with caution.
14769 Use the program PROGRAM as the compiler. Normally, the name `gcc'
14773 Work quietly. Most warnings are suppressed.
14776 Print the version number, just like `-v' for `gcc'.
14778 If you need special compiler options to compile one of your program's
14779 source files, then you should generate that file's `.X' file specially,
14780 by running `gcc' on that source file with the appropriate options and
14781 the option `-aux-info'. Then run `protoize' on the entire set of
14782 files. `protoize' will use the existing `.X' file because it is newer
14783 than the source file. For example:
14785 gcc -Dfoo=bar file1.c -aux-info file1.X
14788 You need to include the special files along with the rest in the
14789 `protoize' command, even though their `.X' files already exist, because
14790 otherwise they won't get converted.
14792 *Note Protoize Caveats::, for more information on how to use
14793 `protoize' successfully.
14796 File: gcc.info, Node: C Implementation, Next: C Extensions, Prev: Invoking GCC, Up: Top
14798 4 C Implementation-defined behavior
14799 ***********************************
14801 A conforming implementation of ISO C is required to document its choice
14802 of behavior in each of the areas that are designated "implementation
14803 defined". The following lists all such areas, along with the section
14804 numbers from the ISO/IEC 9899:1990 and ISO/IEC 9899:1999 standards.
14805 Some areas are only implementation-defined in one version of the
14808 Some choices depend on the externally determined ABI for the platform
14809 (including standard character encodings) which GCC follows; these are
14810 listed as "determined by ABI" below. *Note Binary Compatibility:
14811 Compatibility, and `http://gcc.gnu.org/readings.html'. Some choices
14812 are documented in the preprocessor manual. *Note
14813 Implementation-defined behavior: (cpp)Implementation-defined behavior.
14814 Some choices are made by the library and operating system (or other
14815 environment when compiling for a freestanding environment); refer to
14816 their documentation for details.
14820 * Translation implementation::
14821 * Environment implementation::
14822 * Identifiers implementation::
14823 * Characters implementation::
14824 * Integers implementation::
14825 * Floating point implementation::
14826 * Arrays and pointers implementation::
14827 * Hints implementation::
14828 * Structures unions enumerations and bit-fields implementation::
14829 * Qualifiers implementation::
14830 * Declarators implementation::
14831 * Statements implementation::
14832 * Preprocessing directives implementation::
14833 * Library functions implementation::
14834 * Architecture implementation::
14835 * Locale-specific behavior implementation::
14838 File: gcc.info, Node: Translation implementation, Next: Environment implementation, Up: C Implementation
14843 * `How a diagnostic is identified (C90 3.7, C99 3.10, C90 and C99
14846 Diagnostics consist of all the output sent to stderr by GCC.
14848 * `Whether each nonempty sequence of white-space characters other
14849 than new-line is retained or replaced by one space character in
14850 translation phase 3 (C90 and C99 5.1.1.2).'
14852 *Note Implementation-defined behavior: (cpp)Implementation-defined
14857 File: gcc.info, Node: Environment implementation, Next: Identifiers implementation, Prev: Translation implementation, Up: C Implementation
14862 The behavior of most of these points are dependent on the implementation
14863 of the C library, and are not defined by GCC itself.
14865 * `The mapping between physical source file multibyte characters and
14866 the source character set in translation phase 1 (C90 and C99
14869 *Note Implementation-defined behavior: (cpp)Implementation-defined
14874 File: gcc.info, Node: Identifiers implementation, Next: Characters implementation, Prev: Environment implementation, Up: C Implementation
14879 * `Which additional multibyte characters may appear in identifiers
14880 and their correspondence to universal character names (C99 6.4.2).'
14882 *Note Implementation-defined behavior: (cpp)Implementation-defined
14885 * `The number of significant initial characters in an identifier
14886 (C90 6.1.2, C90 and C99 5.2.4.1, C99 6.4.2).'
14888 For internal names, all characters are significant. For external
14889 names, the number of significant characters are defined by the
14890 linker; for almost all targets, all characters are significant.
14892 * `Whether case distinctions are significant in an identifier with
14893 external linkage (C90 6.1.2).'
14895 This is a property of the linker. C99 requires that case
14896 distinctions are always significant in identifiers with external
14897 linkage and systems without this property are not supported by GCC.
14901 File: gcc.info, Node: Characters implementation, Next: Integers implementation, Prev: Identifiers implementation, Up: C Implementation
14906 * `The number of bits in a byte (C90 3.4, C99 3.6).'
14910 * `The values of the members of the execution character set (C90 and
14915 * `The unique value of the member of the execution character set
14916 produced for each of the standard alphabetic escape sequences (C90
14921 * `The value of a `char' object into which has been stored any
14922 character other than a member of the basic execution character set
14923 (C90 6.1.2.5, C99 6.2.5).'
14927 * `Which of `signed char' or `unsigned char' has the same range,
14928 representation, and behavior as "plain" `char' (C90 6.1.2.5, C90
14929 6.2.1.1, C99 6.2.5, C99 6.3.1.1).'
14931 Determined by ABI. The options `-funsigned-char' and
14932 `-fsigned-char' change the default. *Note Options Controlling C
14933 Dialect: C Dialect Options.
14935 * `The mapping of members of the source character set (in character
14936 constants and string literals) to members of the execution
14937 character set (C90 6.1.3.4, C99 6.4.4.4, C90 and C99 5.1.1.2).'
14941 * `The value of an integer character constant containing more than
14942 one character or containing a character or escape sequence that
14943 does not map to a single-byte execution character (C90 6.1.3.4,
14946 *Note Implementation-defined behavior: (cpp)Implementation-defined
14949 * `The value of a wide character constant containing more than one
14950 multibyte character, or containing a multibyte character or escape
14951 sequence not represented in the extended execution character set
14952 (C90 6.1.3.4, C99 6.4.4.4).'
14954 *Note Implementation-defined behavior: (cpp)Implementation-defined
14957 * `The current locale used to convert a wide character constant
14958 consisting of a single multibyte character that maps to a member
14959 of the extended execution character set into a corresponding wide
14960 character code (C90 6.1.3.4, C99 6.4.4.4).'
14962 *Note Implementation-defined behavior: (cpp)Implementation-defined
14965 * `The current locale used to convert a wide string literal into
14966 corresponding wide character codes (C90 6.1.4, C99 6.4.5).'
14968 *Note Implementation-defined behavior: (cpp)Implementation-defined
14971 * `The value of a string literal containing a multibyte character or
14972 escape sequence not represented in the execution character set
14973 (C90 6.1.4, C99 6.4.5).'
14975 *Note Implementation-defined behavior: (cpp)Implementation-defined
14979 File: gcc.info, Node: Integers implementation, Next: Floating point implementation, Prev: Characters implementation, Up: C Implementation
14984 * `Any extended integer types that exist in the implementation (C99
14987 GCC does not support any extended integer types.
14989 * `Whether signed integer types are represented using sign and
14990 magnitude, two's complement, or one's complement, and whether the
14991 extraordinary value is a trap representation or an ordinary value
14994 GCC supports only two's complement integer types, and all bit
14995 patterns are ordinary values.
14997 * `The rank of any extended integer type relative to another extended
14998 integer type with the same precision (C99 6.3.1.1).'
15000 GCC does not support any extended integer types.
15002 * `The result of, or the signal raised by, converting an integer to a
15003 signed integer type when the value cannot be represented in an
15004 object of that type (C90 6.2.1.2, C99 6.3.1.3).'
15006 For conversion to a type of width N, the value is reduced modulo
15007 2^N to be within range of the type; no signal is raised.
15009 * `The results of some bitwise operations on signed integers (C90
15012 Bitwise operators act on the representation of the value including
15013 both the sign and value bits, where the sign bit is considered
15014 immediately above the highest-value value bit. Signed `>>' acts
15015 on negative numbers by sign extension.
15017 GCC does not use the latitude given in C99 only to treat certain
15018 aspects of signed `<<' as undefined, but this is subject to change.
15020 * `The sign of the remainder on integer division (C90 6.3.5).'
15022 GCC always follows the C99 requirement that the result of division
15023 is truncated towards zero.
15027 File: gcc.info, Node: Floating point implementation, Next: Arrays and pointers implementation, Prev: Integers implementation, Up: C Implementation
15032 * `The accuracy of the floating-point operations and of the library
15033 functions in `<math.h>' and `<complex.h>' that return
15034 floating-point results (C90 and C99 5.2.4.2.2).'
15036 The accuracy is unknown.
15038 * `The rounding behaviors characterized by non-standard values of
15039 `FLT_ROUNDS' (C90 and C99 5.2.4.2.2).'
15041 GCC does not use such values.
15043 * `The evaluation methods characterized by non-standard negative
15044 values of `FLT_EVAL_METHOD' (C99 5.2.4.2.2).'
15046 GCC does not use such values.
15048 * `The direction of rounding when an integer is converted to a
15049 floating-point number that cannot exactly represent the original
15050 value (C90 6.2.1.3, C99 6.3.1.4).'
15052 C99 Annex F is followed.
15054 * `The direction of rounding when a floating-point number is
15055 converted to a narrower floating-point number (C90 6.2.1.4, C99
15058 C99 Annex F is followed.
15060 * `How the nearest representable value or the larger or smaller
15061 representable value immediately adjacent to the nearest
15062 representable value is chosen for certain floating constants (C90
15063 6.1.3.1, C99 6.4.4.2).'
15065 C99 Annex F is followed.
15067 * `Whether and how floating expressions are contracted when not
15068 disallowed by the `FP_CONTRACT' pragma (C99 6.5).'
15070 Expressions are currently only contracted if
15071 `-funsafe-math-optimizations' or `-ffast-math' are used. This is
15074 * `The default state for the `FENV_ACCESS' pragma (C99 7.6.1).'
15076 This pragma is not implemented, but the default is to "off" unless
15077 `-frounding-math' is used in which case it is "on".
15079 * `Additional floating-point exceptions, rounding modes,
15080 environments, and classifications, and their macro names (C99 7.6,
15083 This is dependent on the implementation of the C library, and is
15084 not defined by GCC itself.
15086 * `The default state for the `FP_CONTRACT' pragma (C99 7.12.2).'
15088 This pragma is not implemented. Expressions are currently only
15089 contracted if `-funsafe-math-optimizations' or `-ffast-math' are
15090 used. This is subject to change.
15092 * `Whether the "inexact" floating-point exception can be raised when
15093 the rounded result actually does equal the mathematical result in
15094 an IEC 60559 conformant implementation (C99 F.9).'
15096 This is dependent on the implementation of the C library, and is
15097 not defined by GCC itself.
15099 * `Whether the "underflow" (and "inexact") floating-point exception
15100 can be raised when a result is tiny but not inexact in an IEC
15101 60559 conformant implementation (C99 F.9).'
15103 This is dependent on the implementation of the C library, and is
15104 not defined by GCC itself.
15108 File: gcc.info, Node: Arrays and pointers implementation, Next: Hints implementation, Prev: Floating point implementation, Up: C Implementation
15110 4.7 Arrays and pointers
15111 =======================
15113 * `The result of converting a pointer to an integer or vice versa
15114 (C90 6.3.4, C99 6.3.2.3).'
15116 A cast from pointer to integer discards most-significant bits if
15117 the pointer representation is larger than the integer type,
15118 sign-extends(1) if the pointer representation is smaller than the
15119 integer type, otherwise the bits are unchanged.
15121 A cast from integer to pointer discards most-significant bits if
15122 the pointer representation is smaller than the integer type,
15123 extends according to the signedness of the integer type if the
15124 pointer representation is larger than the integer type, otherwise
15125 the bits are unchanged.
15127 When casting from pointer to integer and back again, the resulting
15128 pointer must reference the same object as the original pointer,
15129 otherwise the behavior is undefined. That is, one may not use
15130 integer arithmetic to avoid the undefined behavior of pointer
15131 arithmetic as proscribed in C99 6.5.6/8.
15133 * `The size of the result of subtracting two pointers to elements of
15134 the same array (C90 6.3.6, C99 6.5.6).'
15136 The value is as specified in the standard and the type is
15137 determined by the ABI.
15140 ---------- Footnotes ----------
15142 (1) Future versions of GCC may zero-extend, or use a target-defined
15143 `ptr_extend' pattern. Do not rely on sign extension.
15146 File: gcc.info, Node: Hints implementation, Next: Structures unions enumerations and bit-fields implementation, Prev: Arrays and pointers implementation, Up: C Implementation
15151 * `The extent to which suggestions made by using the `register'
15152 storage-class specifier are effective (C90 6.5.1, C99 6.7.1).'
15154 The `register' specifier affects code generation only in these
15157 * When used as part of the register variable extension, see
15158 *Note Explicit Reg Vars::.
15160 * When `-O0' is in use, the compiler allocates distinct stack
15161 memory for all variables that do not have the `register'
15162 storage-class specifier; if `register' is specified, the
15163 variable may have a shorter lifespan than the code would
15164 indicate and may never be placed in memory.
15166 * On some rare x86 targets, `setjmp' doesn't save the registers
15167 in all circumstances. In those cases, GCC doesn't allocate
15168 any variables in registers unless they are marked `register'.
15171 * `The extent to which suggestions made by using the inline function
15172 specifier are effective (C99 6.7.4).'
15174 GCC will not inline any functions if the `-fno-inline' option is
15175 used or if `-O0' is used. Otherwise, GCC may still be unable to
15176 inline a function for many reasons; the `-Winline' option may be
15177 used to determine if a function has not been inlined and why not.
15181 File: gcc.info, Node: Structures unions enumerations and bit-fields implementation, Next: Qualifiers implementation, Prev: Hints implementation, Up: C Implementation
15183 4.9 Structures, unions, enumerations, and bit-fields
15184 ====================================================
15186 * `A member of a union object is accessed using a member of a
15187 different type (C90 6.3.2.3).'
15189 The relevant bytes of the representation of the object are treated
15190 as an object of the type used for the access. This may be a trap
15193 * `Whether a "plain" `int' bit-field is treated as a `signed int'
15194 bit-field or as an `unsigned int' bit-field (C90 6.5.2, C90
15195 6.5.2.1, C99 6.7.2, C99 6.7.2.1).'
15197 By default it is treated as `signed int' but this may be changed
15198 by the `-funsigned-bitfields' option.
15200 * `Allowable bit-field types other than `_Bool', `signed int', and
15201 `unsigned int' (C99 6.7.2.1).'
15203 No other types are permitted in strictly conforming mode.
15205 * `Whether a bit-field can straddle a storage-unit boundary (C90
15206 6.5.2.1, C99 6.7.2.1).'
15210 * `The order of allocation of bit-fields within a unit (C90 6.5.2.1,
15215 * `The alignment of non-bit-field members of structures (C90
15216 6.5.2.1, C99 6.7.2.1).'
15220 * `The integer type compatible with each enumerated type (C90
15221 6.5.2.2, C99 6.7.2.2).'
15223 Normally, the type is `unsigned int' if there are no negative
15224 values in the enumeration, otherwise `int'. If `-fshort-enums' is
15225 specified, then if there are negative values it is the first of
15226 `signed char', `short' and `int' that can represent all the
15227 values, otherwise it is the first of `unsigned char', `unsigned
15228 short' and `unsigned int' that can represent all the values.
15230 On some targets, `-fshort-enums' is the default; this is
15231 determined by the ABI.
15235 File: gcc.info, Node: Qualifiers implementation, Next: Declarators implementation, Prev: Structures unions enumerations and bit-fields implementation, Up: C Implementation
15240 * `What constitutes an access to an object that has
15241 volatile-qualified type (C90 6.5.3, C99 6.7.3).'
15243 Such an object is normally accessed by pointers and used for
15244 accessing hardware. In most expressions, it is intuitively
15245 obvious what is a read and what is a write. For example
15247 volatile int *dst = SOMEVALUE;
15248 volatile int *src = SOMEOTHERVALUE;
15251 will cause a read of the volatile object pointed to by SRC and
15252 store the value into the volatile object pointed to by DST. There
15253 is no guarantee that these reads and writes are atomic, especially
15254 for objects larger than `int'.
15256 However, if the volatile storage is not being modified, and the
15257 value of the volatile storage is not used, then the situation is
15258 less obvious. For example
15260 volatile int *src = SOMEVALUE;
15263 According to the C standard, such an expression is an rvalue whose
15264 type is the unqualified version of its original type, i.e. `int'.
15265 Whether GCC interprets this as a read of the volatile object being
15266 pointed to or only as a request to evaluate the expression for its
15267 side-effects depends on this type.
15269 If it is a scalar type, or on most targets an aggregate type whose
15270 only member object is of a scalar type, or a union type whose
15271 member objects are of scalar types, the expression is interpreted
15272 by GCC as a read of the volatile object; in the other cases, the
15273 expression is only evaluated for its side-effects.
15277 File: gcc.info, Node: Declarators implementation, Next: Statements implementation, Prev: Qualifiers implementation, Up: C Implementation
15282 * `The maximum number of declarators that may modify an arithmetic,
15283 structure or union type (C90 6.5.4).'
15285 GCC is only limited by available memory.
15289 File: gcc.info, Node: Statements implementation, Next: Preprocessing directives implementation, Prev: Declarators implementation, Up: C Implementation
15294 * `The maximum number of `case' values in a `switch' statement (C90
15297 GCC is only limited by available memory.
15301 File: gcc.info, Node: Preprocessing directives implementation, Next: Library functions implementation, Prev: Statements implementation, Up: C Implementation
15303 4.13 Preprocessing directives
15304 =============================
15306 *Note Implementation-defined behavior: (cpp)Implementation-defined
15307 behavior, for details of these aspects of implementation-defined
15310 * `How sequences in both forms of header names are mapped to headers
15311 or external source file names (C90 6.1.7, C99 6.4.7).'
15313 * `Whether the value of a character constant in a constant expression
15314 that controls conditional inclusion matches the value of the same
15315 character constant in the execution character set (C90 6.8.1, C99
15318 * `Whether the value of a single-character character constant in a
15319 constant expression that controls conditional inclusion may have a
15320 negative value (C90 6.8.1, C99 6.10.1).'
15322 * `The places that are searched for an included `<>' delimited
15323 header, and how the places are specified or the header is
15324 identified (C90 6.8.2, C99 6.10.2).'
15326 * `How the named source file is searched for in an included `""'
15327 delimited header (C90 6.8.2, C99 6.10.2).'
15329 * `The method by which preprocessing tokens (possibly resulting from
15330 macro expansion) in a `#include' directive are combined into a
15331 header name (C90 6.8.2, C99 6.10.2).'
15333 * `The nesting limit for `#include' processing (C90 6.8.2, C99
15336 * `Whether the `#' operator inserts a `\' character before the `\'
15337 character that begins a universal character name in a character
15338 constant or string literal (C99 6.10.3.2).'
15340 * `The behavior on each recognized non-`STDC #pragma' directive (C90
15341 6.8.6, C99 6.10.6).'
15343 *Note Pragmas: (cpp)Pragmas, for details of pragmas accepted by
15344 GCC on all targets. *Note Pragmas Accepted by GCC: Pragmas, for
15345 details of target-specific pragmas.
15347 * `The definitions for `__DATE__' and `__TIME__' when respectively,
15348 the date and time of translation are not available (C90 6.8.8, C99
15353 File: gcc.info, Node: Library functions implementation, Next: Architecture implementation, Prev: Preprocessing directives implementation, Up: C Implementation
15355 4.14 Library functions
15356 ======================
15358 The behavior of most of these points are dependent on the implementation
15359 of the C library, and are not defined by GCC itself.
15361 * `The null pointer constant to which the macro `NULL' expands (C90
15364 In `<stddef.h>', `NULL' expands to `((void *)0)'. GCC does not
15365 provide the other headers which define `NULL' and some library
15366 implementations may use other definitions in those headers.
15370 File: gcc.info, Node: Architecture implementation, Next: Locale-specific behavior implementation, Prev: Library functions implementation, Up: C Implementation
15375 * `The values or expressions assigned to the macros specified in the
15376 headers `<float.h>', `<limits.h>', and `<stdint.h>' (C90 and C99
15377 5.2.4.2, C99 7.18.2, C99 7.18.3).'
15381 * `The number, order, and encoding of bytes in any object (when not
15382 explicitly specified in this International Standard) (C99
15387 * `The value of the result of the `sizeof' operator (C90 6.3.3.4,
15394 File: gcc.info, Node: Locale-specific behavior implementation, Prev: Architecture implementation, Up: C Implementation
15396 4.16 Locale-specific behavior
15397 =============================
15399 The behavior of these points are dependent on the implementation of the
15400 C library, and are not defined by GCC itself.
15403 File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: C Implementation, Up: Top
15405 5 Extensions to the C Language Family
15406 *************************************
15408 GNU C provides several language features not found in ISO standard C.
15409 (The `-pedantic' option directs GCC to print a warning message if any
15410 of these features is used.) To test for the availability of these
15411 features in conditional compilation, check for a predefined macro
15412 `__GNUC__', which is always defined under GCC.
15414 These extensions are available in C and Objective-C. Most of them are
15415 also available in C++. *Note Extensions to the C++ Language: C++
15416 Extensions, for extensions that apply _only_ to C++.
15418 Some features that are in ISO C99 but not C89 or C++ are also, as
15419 extensions, accepted by GCC in C89 mode and in C++.
15423 * Statement Exprs:: Putting statements and declarations inside expressions.
15424 * Local Labels:: Labels local to a block.
15425 * Labels as Values:: Getting pointers to labels, and computed gotos.
15426 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
15427 * Constructing Calls:: Dispatching a call to another function.
15428 * Typeof:: `typeof': referring to the type of an expression.
15429 * Conditionals:: Omitting the middle operand of a `?:' expression.
15430 * Long Long:: Double-word integers---`long long int'.
15431 * Complex:: Data types for complex numbers.
15432 * Floating Types:: Additional Floating Types.
15433 * Decimal Float:: Decimal Floating Types.
15434 * Hex Floats:: Hexadecimal floating-point constants.
15435 * Fixed-Point:: Fixed-Point Types.
15436 * Zero Length:: Zero-length arrays.
15437 * Variable Length:: Arrays whose length is computed at run time.
15438 * Empty Structures:: Structures with no members.
15439 * Variadic Macros:: Macros with a variable number of arguments.
15440 * Escaped Newlines:: Slightly looser rules for escaped newlines.
15441 * Subscripting:: Any array can be subscripted, even if not an lvalue.
15442 * Pointer Arith:: Arithmetic on `void'-pointers and function pointers.
15443 * Initializers:: Non-constant initializers.
15444 * Compound Literals:: Compound literals give structures, unions
15445 or arrays as values.
15446 * Designated Inits:: Labeling elements of initializers.
15447 * Cast to Union:: Casting to union type from any member of the union.
15448 * Case Ranges:: `case 1 ... 9' and such.
15449 * Mixed Declarations:: Mixing declarations and code.
15450 * Function Attributes:: Declaring that functions have no side effects,
15451 or that they can never return.
15452 * Attribute Syntax:: Formal syntax for attributes.
15453 * Function Prototypes:: Prototype declarations and old-style definitions.
15454 * C++ Comments:: C++ comments are recognized.
15455 * Dollar Signs:: Dollar sign is allowed in identifiers.
15456 * Character Escapes:: `\e' stands for the character <ESC>.
15457 * Variable Attributes:: Specifying attributes of variables.
15458 * Type Attributes:: Specifying attributes of types.
15459 * Alignment:: Inquiring about the alignment of a type or variable.
15460 * Inline:: Defining inline functions (as fast as macros).
15461 * Extended Asm:: Assembler instructions with C expressions as operands.
15462 (With them you can define ``built-in'' functions.)
15463 * Constraints:: Constraints for asm operands
15464 * Asm Labels:: Specifying the assembler name to use for a C symbol.
15465 * Explicit Reg Vars:: Defining variables residing in specified registers.
15466 * Alternate Keywords:: `__const__', `__asm__', etc., for header files.
15467 * Incomplete Enums:: `enum foo;', with details to follow.
15468 * Function Names:: Printable strings which are the name of the current
15470 * Return Address:: Getting the return or frame address of a function.
15471 * Vector Extensions:: Using vector instructions through built-in functions.
15472 * Offsetof:: Special syntax for implementing `offsetof'.
15473 * Atomic Builtins:: Built-in functions for atomic memory access.
15474 * Object Size Checking:: Built-in functions for limited buffer overflow
15476 * Other Builtins:: Other built-in functions.
15477 * Target Builtins:: Built-in functions specific to particular targets.
15478 * Target Format Checks:: Format checks specific to particular targets.
15479 * Pragmas:: Pragmas accepted by GCC.
15480 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
15481 * Thread-Local:: Per-thread variables.
15482 * Binary constants:: Binary constants using the `0b' prefix.
15485 File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions
15487 5.1 Statements and Declarations in Expressions
15488 ==============================================
15490 A compound statement enclosed in parentheses may appear as an expression
15491 in GNU C. This allows you to use loops, switches, and local variables
15492 within an expression.
15494 Recall that a compound statement is a sequence of statements surrounded
15495 by braces; in this construct, parentheses go around the braces. For
15498 ({ int y = foo (); int z;
15503 is a valid (though slightly more complex than necessary) expression for
15504 the absolute value of `foo ()'.
15506 The last thing in the compound statement should be an expression
15507 followed by a semicolon; the value of this subexpression serves as the
15508 value of the entire construct. (If you use some other kind of statement
15509 last within the braces, the construct has type `void', and thus
15510 effectively no value.)
15512 This feature is especially useful in making macro definitions "safe"
15513 (so that they evaluate each operand exactly once). For example, the
15514 "maximum" function is commonly defined as a macro in standard C as
15517 #define max(a,b) ((a) > (b) ? (a) : (b))
15519 But this definition computes either A or B twice, with bad results if
15520 the operand has side effects. In GNU C, if you know the type of the
15521 operands (here taken as `int'), you can define the macro safely as
15524 #define maxint(a,b) \
15525 ({int _a = (a), _b = (b); _a > _b ? _a : _b; })
15527 Embedded statements are not allowed in constant expressions, such as
15528 the value of an enumeration constant, the width of a bit-field, or the
15529 initial value of a static variable.
15531 If you don't know the type of the operand, you can still do this, but
15532 you must use `typeof' (*note Typeof::).
15534 In G++, the result value of a statement expression undergoes array and
15535 function pointer decay, and is returned by value to the enclosing
15536 expression. For instance, if `A' is a class, then
15542 will construct a temporary `A' object to hold the result of the
15543 statement expression, and that will be used to invoke `Foo'. Therefore
15544 the `this' pointer observed by `Foo' will not be the address of `a'.
15546 Any temporaries created within a statement within a statement
15547 expression will be destroyed at the statement's end. This makes
15548 statement expressions inside macros slightly different from function
15549 calls. In the latter case temporaries introduced during argument
15550 evaluation will be destroyed at the end of the statement that includes
15551 the function call. In the statement expression case they will be
15552 destroyed during the statement expression. For instance,
15554 #define macro(a) ({__typeof__(a) b = (a); b + 3; })
15555 template<typename T> T function(T a) { T b = a; return b + 3; }
15563 will have different places where temporaries are destroyed. For the
15564 `macro' case, the temporary `X' will be destroyed just after the
15565 initialization of `b'. In the `function' case that temporary will be
15566 destroyed when the function returns.
15568 These considerations mean that it is probably a bad idea to use
15569 statement-expressions of this form in header files that are designed to
15570 work with C++. (Note that some versions of the GNU C Library contained
15571 header files using statement-expression that lead to precisely this
15574 Jumping into a statement expression with `goto' or using a `switch'
15575 statement outside the statement expression with a `case' or `default'
15576 label inside the statement expression is not permitted. Jumping into a
15577 statement expression with a computed `goto' (*note Labels as Values::)
15578 yields undefined behavior. Jumping out of a statement expression is
15579 permitted, but if the statement expression is part of a larger
15580 expression then it is unspecified which other subexpressions of that
15581 expression have been evaluated except where the language definition
15582 requires certain subexpressions to be evaluated before or after the
15583 statement expression. In any case, as with a function call the
15584 evaluation of a statement expression is not interleaved with the
15585 evaluation of other parts of the containing expression. For example,
15587 foo (), (({ bar1 (); goto a; 0; }) + bar2 ()), baz();
15589 will call `foo' and `bar1' and will not call `baz' but may or may not
15590 call `bar2'. If `bar2' is called, it will be called after `foo' and
15594 File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions
15596 5.2 Locally Declared Labels
15597 ===========================
15599 GCC allows you to declare "local labels" in any nested block scope. A
15600 local label is just like an ordinary label, but you can only reference
15601 it (with a `goto' statement, or by taking its address) within the block
15602 in which it was declared.
15604 A local label declaration looks like this:
15610 __label__ LABEL1, LABEL2, /* ... */;
15612 Local label declarations must come at the beginning of the block,
15613 before any ordinary declarations or statements.
15615 The label declaration defines the label _name_, but does not define
15616 the label itself. You must do this in the usual way, with `LABEL:',
15617 within the statements of the statement expression.
15619 The local label feature is useful for complex macros. If a macro
15620 contains nested loops, a `goto' can be useful for breaking out of them.
15621 However, an ordinary label whose scope is the whole function cannot be
15622 used: if the macro can be expanded several times in one function, the
15623 label will be multiply defined in that function. A local label avoids
15624 this problem. For example:
15626 #define SEARCH(value, array, target) \
15629 typeof (target) _SEARCH_target = (target); \
15630 typeof (*(array)) *_SEARCH_array = (array); \
15633 for (i = 0; i < max; i++) \
15634 for (j = 0; j < max; j++) \
15635 if (_SEARCH_array[i][j] == _SEARCH_target) \
15636 { (value) = i; goto found; } \
15641 This could also be written using a statement-expression:
15643 #define SEARCH(array, target) \
15646 typeof (target) _SEARCH_target = (target); \
15647 typeof (*(array)) *_SEARCH_array = (array); \
15650 for (i = 0; i < max; i++) \
15651 for (j = 0; j < max; j++) \
15652 if (_SEARCH_array[i][j] == _SEARCH_target) \
15653 { value = i; goto found; } \
15659 Local label declarations also make the labels they declare visible to
15660 nested functions, if there are any. *Note Nested Functions::, for
15664 File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions
15666 5.3 Labels as Values
15667 ====================
15669 You can get the address of a label defined in the current function (or
15670 a containing function) with the unary operator `&&'. The value has
15671 type `void *'. This value is a constant and can be used wherever a
15672 constant of that type is valid. For example:
15678 To use these values, you need to be able to jump to one. This is done
15679 with the computed goto statement(1), `goto *EXP;'. For example,
15683 Any expression of type `void *' is allowed.
15685 One way of using these constants is in initializing a static array that
15686 will serve as a jump table:
15688 static void *array[] = { &&foo, &&bar, &&hack };
15690 Then you can select a label with indexing, like this:
15694 Note that this does not check whether the subscript is in bounds--array
15695 indexing in C never does that.
15697 Such an array of label values serves a purpose much like that of the
15698 `switch' statement. The `switch' statement is cleaner, so use that
15699 rather than an array unless the problem does not fit a `switch'
15700 statement very well.
15702 Another use of label values is in an interpreter for threaded code.
15703 The labels within the interpreter function can be stored in the
15704 threaded code for super-fast dispatching.
15706 You may not use this mechanism to jump to code in a different function.
15707 If you do that, totally unpredictable things will happen. The best way
15708 to avoid this is to store the label address only in automatic variables
15709 and never pass it as an argument.
15711 An alternate way to write the above example is
15713 static const int array[] = { &&foo - &&foo, &&bar - &&foo,
15715 goto *(&&foo + array[i]);
15717 This is more friendly to code living in shared libraries, as it reduces
15718 the number of dynamic relocations that are needed, and by consequence,
15719 allows the data to be read-only.
15721 The `&&foo' expressions for the same label might have different values
15722 if the containing function is inlined or cloned. If a program relies on
15723 them being always the same, `__attribute__((__noinline__))' should be
15724 used to prevent inlining. If `&&foo' is used in a static variable
15725 initializer, inlining is forbidden.
15727 ---------- Footnotes ----------
15729 (1) The analogous feature in Fortran is called an assigned goto, but
15730 that name seems inappropriate in C, where one can do more than simply
15731 store label addresses in label variables.
15734 File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions
15736 5.4 Nested Functions
15737 ====================
15739 A "nested function" is a function defined inside another function.
15740 (Nested functions are not supported for GNU C++.) The nested function's
15741 name is local to the block where it is defined. For example, here we
15742 define a nested function named `square', and call it twice:
15744 foo (double a, double b)
15746 double square (double z) { return z * z; }
15748 return square (a) + square (b);
15751 The nested function can access all the variables of the containing
15752 function that are visible at the point of its definition. This is
15753 called "lexical scoping". For example, here we show a nested function
15754 which uses an inherited variable named `offset':
15756 bar (int *array, int offset, int size)
15758 int access (int *array, int index)
15759 { return array[index + offset]; }
15762 for (i = 0; i < size; i++)
15763 /* ... */ access (array, i) /* ... */
15766 Nested function definitions are permitted within functions in the
15767 places where variable definitions are allowed; that is, in any block,
15768 mixed with the other declarations and statements in the block.
15770 It is possible to call the nested function from outside the scope of
15771 its name by storing its address or passing the address to another
15774 hack (int *array, int size)
15776 void store (int index, int value)
15777 { array[index] = value; }
15779 intermediate (store, size);
15782 Here, the function `intermediate' receives the address of `store' as
15783 an argument. If `intermediate' calls `store', the arguments given to
15784 `store' are used to store into `array'. But this technique works only
15785 so long as the containing function (`hack', in this example) does not
15788 If you try to call the nested function through its address after the
15789 containing function has exited, all hell will break loose. If you try
15790 to call it after a containing scope level has exited, and if it refers
15791 to some of the variables that are no longer in scope, you may be lucky,
15792 but it's not wise to take the risk. If, however, the nested function
15793 does not refer to anything that has gone out of scope, you should be
15796 GCC implements taking the address of a nested function using a
15797 technique called "trampolines". A paper describing them is available as
15799 `http://people.debian.org/~aaronl/Usenix88-lexic.pdf'.
15801 A nested function can jump to a label inherited from a containing
15802 function, provided the label was explicitly declared in the containing
15803 function (*note Local Labels::). Such a jump returns instantly to the
15804 containing function, exiting the nested function which did the `goto'
15805 and any intermediate functions as well. Here is an example:
15807 bar (int *array, int offset, int size)
15810 int access (int *array, int index)
15814 return array[index + offset];
15818 for (i = 0; i < size; i++)
15819 /* ... */ access (array, i) /* ... */
15823 /* Control comes here from `access'
15824 if it detects an error. */
15829 A nested function always has no linkage. Declaring one with `extern'
15830 or `static' is erroneous. If you need to declare the nested function
15831 before its definition, use `auto' (which is otherwise meaningless for
15832 function declarations).
15834 bar (int *array, int offset, int size)
15837 auto int access (int *, int);
15839 int access (int *array, int index)
15843 return array[index + offset];
15849 File: gcc.info, Node: Constructing Calls, Next: Typeof, Prev: Nested Functions, Up: C Extensions
15851 5.5 Constructing Function Calls
15852 ===============================
15854 Using the built-in functions described below, you can record the
15855 arguments a function received, and call another function with the same
15856 arguments, without knowing the number or types of the arguments.
15858 You can also record the return value of that function call, and later
15859 return that value, without knowing what data type the function tried to
15860 return (as long as your caller expects that data type).
15862 However, these built-in functions may interact badly with some
15863 sophisticated features or other extensions of the language. It is,
15864 therefore, not recommended to use them outside very simple functions
15865 acting as mere forwarders for their arguments.
15867 -- Built-in Function: void * __builtin_apply_args ()
15868 This built-in function returns a pointer to data describing how to
15869 perform a call with the same arguments as were passed to the
15872 The function saves the arg pointer register, structure value
15873 address, and all registers that might be used to pass arguments to
15874 a function into a block of memory allocated on the stack. Then it
15875 returns the address of that block.
15877 -- Built-in Function: void * __builtin_apply (void (*FUNCTION)(), void
15878 *ARGUMENTS, size_t SIZE)
15879 This built-in function invokes FUNCTION with a copy of the
15880 parameters described by ARGUMENTS and SIZE.
15882 The value of ARGUMENTS should be the value returned by
15883 `__builtin_apply_args'. The argument SIZE specifies the size of
15884 the stack argument data, in bytes.
15886 This function returns a pointer to data describing how to return
15887 whatever value was returned by FUNCTION. The data is saved in a
15888 block of memory allocated on the stack.
15890 It is not always simple to compute the proper value for SIZE. The
15891 value is used by `__builtin_apply' to compute the amount of data
15892 that should be pushed on the stack and copied from the incoming
15895 -- Built-in Function: void __builtin_return (void *RESULT)
15896 This built-in function returns the value described by RESULT from
15897 the containing function. You should specify, for RESULT, a value
15898 returned by `__builtin_apply'.
15900 -- Built-in Function: __builtin_va_arg_pack ()
15901 This built-in function represents all anonymous arguments of an
15902 inline function. It can be used only in inline functions which
15903 will be always inlined, never compiled as a separate function,
15904 such as those using `__attribute__ ((__always_inline__))' or
15905 `__attribute__ ((__gnu_inline__))' extern inline functions. It
15906 must be only passed as last argument to some other function with
15907 variable arguments. This is useful for writing small wrapper
15908 inlines for variable argument functions, when using preprocessor
15909 macros is undesirable. For example:
15910 extern int myprintf (FILE *f, const char *format, ...);
15911 extern inline __attribute__ ((__gnu_inline__)) int
15912 myprintf (FILE *f, const char *format, ...)
15914 int r = fprintf (f, "myprintf: ");
15917 int s = fprintf (f, format, __builtin_va_arg_pack ());
15923 -- Built-in Function: __builtin_va_arg_pack_len ()
15924 This built-in function returns the number of anonymous arguments of
15925 an inline function. It can be used only in inline functions which
15926 will be always inlined, never compiled as a separate function, such
15927 as those using `__attribute__ ((__always_inline__))' or
15928 `__attribute__ ((__gnu_inline__))' extern inline functions. For
15929 example following will do link or runtime checking of open
15930 arguments for optimized code:
15931 #ifdef __OPTIMIZE__
15932 extern inline __attribute__((__gnu_inline__)) int
15933 myopen (const char *path, int oflag, ...)
15935 if (__builtin_va_arg_pack_len () > 1)
15936 warn_open_too_many_arguments ();
15938 if (__builtin_constant_p (oflag))
15940 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
15942 warn_open_missing_mode ();
15943 return __open_2 (path, oflag);
15945 return open (path, oflag, __builtin_va_arg_pack ());
15948 if (__builtin_va_arg_pack_len () < 1)
15949 return __open_2 (path, oflag);
15951 return open (path, oflag, __builtin_va_arg_pack ());
15956 File: gcc.info, Node: Typeof, Next: Conditionals, Prev: Constructing Calls, Up: C Extensions
15958 5.6 Referring to a Type with `typeof'
15959 =====================================
15961 Another way to refer to the type of an expression is with `typeof'.
15962 The syntax of using of this keyword looks like `sizeof', but the
15963 construct acts semantically like a type name defined with `typedef'.
15965 There are two ways of writing the argument to `typeof': with an
15966 expression or with a type. Here is an example with an expression:
15970 This assumes that `x' is an array of pointers to functions; the type
15971 described is that of the values of the functions.
15973 Here is an example with a typename as the argument:
15977 Here the type described is that of pointers to `int'.
15979 If you are writing a header file that must work when included in ISO C
15980 programs, write `__typeof__' instead of `typeof'. *Note Alternate
15983 A `typeof'-construct can be used anywhere a typedef name could be
15984 used. For example, you can use it in a declaration, in a cast, or
15985 inside of `sizeof' or `typeof'.
15987 `typeof' is often useful in conjunction with the
15988 statements-within-expressions feature. Here is how the two together can
15989 be used to define a safe "maximum" macro that operates on any
15990 arithmetic type and evaluates each of its arguments exactly once:
15993 ({ typeof (a) _a = (a); \
15994 typeof (b) _b = (b); \
15995 _a > _b ? _a : _b; })
15997 The reason for using names that start with underscores for the local
15998 variables is to avoid conflicts with variable names that occur within
15999 the expressions that are substituted for `a' and `b'. Eventually we
16000 hope to design a new form of declaration syntax that allows you to
16001 declare variables whose scopes start only after their initializers;
16002 this will be a more reliable way to prevent such conflicts.
16004 Some more examples of the use of `typeof':
16006 * This declares `y' with the type of what `x' points to.
16010 * This declares `y' as an array of such values.
16014 * This declares `y' as an array of pointers to characters:
16016 typeof (typeof (char *)[4]) y;
16018 It is equivalent to the following traditional C declaration:
16022 To see the meaning of the declaration using `typeof', and why it
16023 might be a useful way to write, rewrite it with these macros:
16025 #define pointer(T) typeof(T *)
16026 #define array(T, N) typeof(T [N])
16028 Now the declaration can be rewritten this way:
16030 array (pointer (char), 4) y;
16032 Thus, `array (pointer (char), 4)' is the type of arrays of 4
16033 pointers to `char'.
16035 _Compatibility Note:_ In addition to `typeof', GCC 2 supported a more
16036 limited extension which permitted one to write
16040 with the effect of declaring T to have the type of the expression EXPR.
16041 This extension does not work with GCC 3 (versions between 3.0 and 3.2
16042 will crash; 3.2.1 and later give an error). Code which relies on it
16043 should be rewritten to use `typeof':
16045 typedef typeof(EXPR) T;
16047 This will work with all versions of GCC.
16050 File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Typeof, Up: C Extensions
16052 5.7 Conditionals with Omitted Operands
16053 ======================================
16055 The middle operand in a conditional expression may be omitted. Then if
16056 the first operand is nonzero, its value is the value of the conditional
16059 Therefore, the expression
16063 has the value of `x' if that is nonzero; otherwise, the value of `y'.
16065 This example is perfectly equivalent to
16069 In this simple case, the ability to omit the middle operand is not
16070 especially useful. When it becomes useful is when the first operand
16071 does, or may (if it is a macro argument), contain a side effect. Then
16072 repeating the operand in the middle would perform the side effect
16073 twice. Omitting the middle operand uses the value already computed
16074 without the undesirable effects of recomputing it.
16077 File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions
16079 5.8 Double-Word Integers
16080 ========================
16082 ISO C99 supports data types for integers that are at least 64 bits wide,
16083 and as an extension GCC supports them in C89 mode and in C++. Simply
16084 write `long long int' for a signed integer, or `unsigned long long int'
16085 for an unsigned integer. To make an integer constant of type `long
16086 long int', add the suffix `LL' to the integer. To make an integer
16087 constant of type `unsigned long long int', add the suffix `ULL' to the
16090 You can use these types in arithmetic like any other integer types.
16091 Addition, subtraction, and bitwise boolean operations on these types
16092 are open-coded on all types of machines. Multiplication is open-coded
16093 if the machine supports fullword-to-doubleword a widening multiply
16094 instruction. Division and shifts are open-coded only on machines that
16095 provide special support. The operations that are not open-coded use
16096 special library routines that come with GCC.
16098 There may be pitfalls when you use `long long' types for function
16099 arguments, unless you declare function prototypes. If a function
16100 expects type `int' for its argument, and you pass a value of type `long
16101 long int', confusion will result because the caller and the subroutine
16102 will disagree about the number of bytes for the argument. Likewise, if
16103 the function expects `long long int' and you pass `int'. The best way
16104 to avoid such problems is to use prototypes.
16107 File: gcc.info, Node: Complex, Next: Floating Types, Prev: Long Long, Up: C Extensions
16109 5.9 Complex Numbers
16110 ===================
16112 ISO C99 supports complex floating data types, and as an extension GCC
16113 supports them in C89 mode and in C++, and supports complex integer data
16114 types which are not part of ISO C99. You can declare complex types
16115 using the keyword `_Complex'. As an extension, the older GNU keyword
16116 `__complex__' is also supported.
16118 For example, `_Complex double x;' declares `x' as a variable whose
16119 real part and imaginary part are both of type `double'. `_Complex
16120 short int y;' declares `y' to have real and imaginary parts of type
16121 `short int'; this is not likely to be useful, but it shows that the set
16122 of complex types is complete.
16124 To write a constant with a complex data type, use the suffix `i' or
16125 `j' (either one; they are equivalent). For example, `2.5fi' has type
16126 `_Complex float' and `3i' has type `_Complex int'. Such a constant
16127 always has a pure imaginary value, but you can form any complex value
16128 you like by adding one to a real constant. This is a GNU extension; if
16129 you have an ISO C99 conforming C library (such as GNU libc), and want
16130 to construct complex constants of floating type, you should include
16131 `<complex.h>' and use the macros `I' or `_Complex_I' instead.
16133 To extract the real part of a complex-valued expression EXP, write
16134 `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary
16135 part. This is a GNU extension; for values of floating type, you should
16136 use the ISO C99 functions `crealf', `creal', `creall', `cimagf',
16137 `cimag' and `cimagl', declared in `<complex.h>' and also provided as
16138 built-in functions by GCC.
16140 The operator `~' performs complex conjugation when used on a value
16141 with a complex type. This is a GNU extension; for values of floating
16142 type, you should use the ISO C99 functions `conjf', `conj' and `conjl',
16143 declared in `<complex.h>' and also provided as built-in functions by
16146 GCC can allocate complex automatic variables in a noncontiguous
16147 fashion; it's even possible for the real part to be in a register while
16148 the imaginary part is on the stack (or vice-versa). Only the DWARF2
16149 debug info format can represent this, so use of DWARF2 is recommended.
16150 If you are using the stabs debug info format, GCC describes a
16151 noncontiguous complex variable as if it were two separate variables of
16152 noncomplex type. If the variable's actual name is `foo', the two
16153 fictitious variables are named `foo$real' and `foo$imag'. You can
16154 examine and set these two fictitious variables with your debugger.
16157 File: gcc.info, Node: Floating Types, Next: Decimal Float, Prev: Complex, Up: C Extensions
16159 5.10 Additional Floating Types
16160 ==============================
16162 As an extension, the GNU C compiler supports additional floating types,
16163 `__float80' and `__float128' to support 80bit (`XFmode') and 128 bit
16164 (`TFmode') floating types. Support for additional types includes the
16165 arithmetic operators: add, subtract, multiply, divide; unary arithmetic
16166 operators; relational operators; equality operators; and conversions to
16167 and from integer and other floating types. Use a suffix `w' or `W' in
16168 a literal constant of type `__float80' and `q' or `Q' for `_float128'.
16169 You can declare complex types using the corresponding internal complex
16170 type, `XCmode' for `__float80' type and `TCmode' for `__float128' type:
16172 typedef _Complex float __attribute__((mode(TC))) _Complex128;
16173 typedef _Complex float __attribute__((mode(XC))) _Complex80;
16175 Not all targets support additional floating point types. `__float80'
16176 is supported on i386, x86_64 and ia64 targets and target `__float128'
16177 is supported on x86_64 and ia64 targets.
16180 File: gcc.info, Node: Decimal Float, Next: Hex Floats, Prev: Floating Types, Up: C Extensions
16182 5.11 Decimal Floating Types
16183 ===========================
16185 As an extension, the GNU C compiler supports decimal floating types as
16186 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
16187 floating types in GCC will evolve as the draft technical report changes.
16188 Calling conventions for any target might also change. Not all targets
16189 support decimal floating types.
16191 The decimal floating types are `_Decimal32', `_Decimal64', and
16192 `_Decimal128'. They use a radix of ten, unlike the floating types
16193 `float', `double', and `long double' whose radix is not specified by
16194 the C standard but is usually two.
16196 Support for decimal floating types includes the arithmetic operators
16197 add, subtract, multiply, divide; unary arithmetic operators; relational
16198 operators; equality operators; and conversions to and from integer and
16199 other floating types. Use a suffix `df' or `DF' in a literal constant
16200 of type `_Decimal32', `dd' or `DD' for `_Decimal64', and `dl' or `DL'
16203 GCC support of decimal float as specified by the draft technical report
16206 * Translation time data type (TTDT) is not supported.
16208 * When the value of a decimal floating type cannot be represented in
16209 the integer type to which it is being converted, the result is
16210 undefined rather than the result value specified by the draft
16213 Types `_Decimal32', `_Decimal64', and `_Decimal128' are supported by
16214 the DWARF2 debug information format.
16217 File: gcc.info, Node: Hex Floats, Next: Fixed-Point, Prev: Decimal Float, Up: C Extensions
16222 ISO C99 supports floating-point numbers written not only in the usual
16223 decimal notation, such as `1.55e1', but also numbers such as `0x1.fp3'
16224 written in hexadecimal format. As a GNU extension, GCC supports this
16225 in C89 mode (except in some cases when strictly conforming) and in C++.
16226 In that format the `0x' hex introducer and the `p' or `P' exponent
16227 field are mandatory. The exponent is a decimal number that indicates
16228 the power of 2 by which the significant part will be multiplied. Thus
16229 `0x1.f' is 1 15/16, `p3' multiplies it by 8, and the value of `0x1.fp3'
16230 is the same as `1.55e1'.
16232 Unlike for floating-point numbers in the decimal notation the exponent
16233 is always required in the hexadecimal notation. Otherwise the compiler
16234 would not be able to resolve the ambiguity of, e.g., `0x1.f'. This
16235 could mean `1.0f' or `1.9375' since `f' is also the extension for
16236 floating-point constants of type `float'.
16239 File: gcc.info, Node: Fixed-Point, Next: Zero Length, Prev: Hex Floats, Up: C Extensions
16241 5.13 Fixed-Point Types
16242 ======================
16244 As an extension, the GNU C compiler supports fixed-point types as
16245 defined in the N1169 draft of ISO/IEC DTR 18037. Support for
16246 fixed-point types in GCC will evolve as the draft technical report
16247 changes. Calling conventions for any target might also change. Not
16248 all targets support fixed-point types.
16250 The fixed-point types are `short _Fract', `_Fract', `long _Fract',
16251 `long long _Fract', `unsigned short _Fract', `unsigned _Fract',
16252 `unsigned long _Fract', `unsigned long long _Fract', `_Sat short
16253 _Fract', `_Sat _Fract', `_Sat long _Fract', `_Sat long long _Fract',
16254 `_Sat unsigned short _Fract', `_Sat unsigned _Fract', `_Sat unsigned
16255 long _Fract', `_Sat unsigned long long _Fract', `short _Accum',
16256 `_Accum', `long _Accum', `long long _Accum', `unsigned short _Accum',
16257 `unsigned _Accum', `unsigned long _Accum', `unsigned long long _Accum',
16258 `_Sat short _Accum', `_Sat _Accum', `_Sat long _Accum', `_Sat long long
16259 _Accum', `_Sat unsigned short _Accum', `_Sat unsigned _Accum', `_Sat
16260 unsigned long _Accum', `_Sat unsigned long long _Accum'. Fixed-point
16261 data values contain fractional and optional integral parts. The format
16262 of fixed-point data varies and depends on the target machine.
16264 Support for fixed-point types includes prefix and postfix increment
16265 and decrement operators (`++', `--'); unary arithmetic operators (`+',
16266 `-', `!'); binary arithmetic operators (`+', `-', `*', `/'); binary
16267 shift operators (`<<', `>>'); relational operators (`<', `<=', `>=',
16268 `>'); equality operators (`==', `!='); assignment operators (`+=',
16269 `-=', `*=', `/=', `<<=', `>>='); and conversions to and from integer,
16270 floating-point, or fixed-point types.
16272 Use a suffix `hr' or `HR' in a literal constant of type `short _Fract'
16273 and `_Sat short _Fract', `r' or `R' for `_Fract' and `_Sat _Fract',
16274 `lr' or `LR' for `long _Fract' and `_Sat long _Fract', `llr' or `LLR'
16275 for `long long _Fract' and `_Sat long long _Fract', `uhr' or `UHR' for
16276 `unsigned short _Fract' and `_Sat unsigned short _Fract', `ur' or `UR'
16277 for `unsigned _Fract' and `_Sat unsigned _Fract', `ulr' or `ULR' for
16278 `unsigned long _Fract' and `_Sat unsigned long _Fract', `ullr' or
16279 `ULLR' for `unsigned long long _Fract' and `_Sat unsigned long long
16280 _Fract', `hk' or `HK' for `short _Accum' and `_Sat short _Accum', `k'
16281 or `K' for `_Accum' and `_Sat _Accum', `lk' or `LK' for `long _Accum'
16282 and `_Sat long _Accum', `llk' or `LLK' for `long long _Accum' and `_Sat
16283 long long _Accum', `uhk' or `UHK' for `unsigned short _Accum' and `_Sat
16284 unsigned short _Accum', `uk' or `UK' for `unsigned _Accum' and `_Sat
16285 unsigned _Accum', `ulk' or `ULK' for `unsigned long _Accum' and `_Sat
16286 unsigned long _Accum', and `ullk' or `ULLK' for `unsigned long long
16287 _Accum' and `_Sat unsigned long long _Accum'.
16289 GCC support of fixed-point types as specified by the draft technical
16290 report is incomplete:
16292 * Pragmas to control overflow and rounding behaviors are not
16295 Fixed-point types are supported by the DWARF2 debug information format.
16298 File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Fixed-Point, Up: C Extensions
16300 5.14 Arrays of Length Zero
16301 ==========================
16303 Zero-length arrays are allowed in GNU C. They are very useful as the
16304 last element of a structure which is really a header for a
16305 variable-length object:
16312 struct line *thisline = (struct line *)
16313 malloc (sizeof (struct line) + this_length);
16314 thisline->length = this_length;
16316 In ISO C90, you would have to give `contents' a length of 1, which
16317 means either you waste space or complicate the argument to `malloc'.
16319 In ISO C99, you would use a "flexible array member", which is slightly
16320 different in syntax and semantics:
16322 * Flexible array members are written as `contents[]' without the `0'.
16324 * Flexible array members have incomplete type, and so the `sizeof'
16325 operator may not be applied. As a quirk of the original
16326 implementation of zero-length arrays, `sizeof' evaluates to zero.
16328 * Flexible array members may only appear as the last member of a
16329 `struct' that is otherwise non-empty.
16331 * A structure containing a flexible array member, or a union
16332 containing such a structure (possibly recursively), may not be a
16333 member of a structure or an element of an array. (However, these
16334 uses are permitted by GCC as extensions.)
16336 GCC versions before 3.0 allowed zero-length arrays to be statically
16337 initialized, as if they were flexible arrays. In addition to those
16338 cases that were useful, it also allowed initializations in situations
16339 that would corrupt later data. Non-empty initialization of zero-length
16340 arrays is now treated like any case where there are more initializer
16341 elements than the array holds, in that a suitable warning about "excess
16342 elements in array" is given, and the excess elements (all of them, in
16343 this case) are ignored.
16345 Instead GCC allows static initialization of flexible array members.
16346 This is equivalent to defining a new structure containing the original
16347 structure followed by an array of sufficient size to contain the data.
16348 I.e. in the following, `f1' is constructed as if it were declared like
16353 } f1 = { 1, { 2, 3, 4 } };
16356 struct f1 f1; int data[3];
16357 } f2 = { { 1 }, { 2, 3, 4 } };
16359 The convenience of this extension is that `f1' has the desired type,
16360 eliminating the need to consistently refer to `f2.f1'.
16362 This has symmetry with normal static arrays, in that an array of
16363 unknown size is also written with `[]'.
16365 Of course, this extension only makes sense if the extra data comes at
16366 the end of a top-level object, as otherwise we would be overwriting
16367 data at subsequent offsets. To avoid undue complication and confusion
16368 with initialization of deeply nested arrays, we simply disallow any
16369 non-empty initialization except when the structure is the top-level
16370 object. For example:
16372 struct foo { int x; int y[]; };
16373 struct bar { struct foo z; };
16375 struct foo a = { 1, { 2, 3, 4 } }; // Valid.
16376 struct bar b = { { 1, { 2, 3, 4 } } }; // Invalid.
16377 struct bar c = { { 1, { } } }; // Valid.
16378 struct foo d[1] = { { 1 { 2, 3, 4 } } }; // Invalid.
16381 File: gcc.info, Node: Empty Structures, Next: Variadic Macros, Prev: Variable Length, Up: C Extensions
16383 5.15 Structures With No Members
16384 ===============================
16386 GCC permits a C structure to have no members:
16391 The structure will have size zero. In C++, empty structures are part
16392 of the language. G++ treats empty structures as if they had a single
16393 member of type `char'.
16396 File: gcc.info, Node: Variable Length, Next: Empty Structures, Prev: Zero Length, Up: C Extensions
16398 5.16 Arrays of Variable Length
16399 ==============================
16401 Variable-length automatic arrays are allowed in ISO C99, and as an
16402 extension GCC accepts them in C89 mode and in C++. (However, GCC's
16403 implementation of variable-length arrays does not yet conform in detail
16404 to the ISO C99 standard.) These arrays are declared like any other
16405 automatic arrays, but with a length that is not a constant expression.
16406 The storage is allocated at the point of declaration and deallocated
16407 when the brace-level is exited. For example:
16410 concat_fopen (char *s1, char *s2, char *mode)
16412 char str[strlen (s1) + strlen (s2) + 1];
16415 return fopen (str, mode);
16418 Jumping or breaking out of the scope of the array name deallocates the
16419 storage. Jumping into the scope is not allowed; you get an error
16422 You can use the function `alloca' to get an effect much like
16423 variable-length arrays. The function `alloca' is available in many
16424 other C implementations (but not in all). On the other hand,
16425 variable-length arrays are more elegant.
16427 There are other differences between these two methods. Space allocated
16428 with `alloca' exists until the containing _function_ returns. The
16429 space for a variable-length array is deallocated as soon as the array
16430 name's scope ends. (If you use both variable-length arrays and
16431 `alloca' in the same function, deallocation of a variable-length array
16432 will also deallocate anything more recently allocated with `alloca'.)
16434 You can also use variable-length arrays as arguments to functions:
16437 tester (int len, char data[len][len])
16442 The length of an array is computed once when the storage is allocated
16443 and is remembered for the scope of the array in case you access it with
16446 If you want to pass the array first and the length afterward, you can
16447 use a forward declaration in the parameter list--another GNU extension.
16450 tester (int len; char data[len][len], int len)
16455 The `int len' before the semicolon is a "parameter forward
16456 declaration", and it serves the purpose of making the name `len' known
16457 when the declaration of `data' is parsed.
16459 You can write any number of such parameter forward declarations in the
16460 parameter list. They can be separated by commas or semicolons, but the
16461 last one must end with a semicolon, which is followed by the "real"
16462 parameter declarations. Each forward declaration must match a "real"
16463 declaration in parameter name and data type. ISO C99 does not support
16464 parameter forward declarations.
16467 File: gcc.info, Node: Variadic Macros, Next: Escaped Newlines, Prev: Empty Structures, Up: C Extensions
16469 5.17 Macros with a Variable Number of Arguments.
16470 ================================================
16472 In the ISO C standard of 1999, a macro can be declared to accept a
16473 variable number of arguments much as a function can. The syntax for
16474 defining the macro is similar to that of a function. Here is an
16477 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
16479 Here `...' is a "variable argument". In the invocation of such a
16480 macro, it represents the zero or more tokens until the closing
16481 parenthesis that ends the invocation, including any commas. This set of
16482 tokens replaces the identifier `__VA_ARGS__' in the macro body wherever
16483 it appears. See the CPP manual for more information.
16485 GCC has long supported variadic macros, and used a different syntax
16486 that allowed you to give a name to the variable arguments just like any
16487 other argument. Here is an example:
16489 #define debug(format, args...) fprintf (stderr, format, args)
16491 This is in all ways equivalent to the ISO C example above, but arguably
16492 more readable and descriptive.
16494 GNU CPP has two further variadic macro extensions, and permits them to
16495 be used with either of the above forms of macro definition.
16497 In standard C, you are not allowed to leave the variable argument out
16498 entirely; but you are allowed to pass an empty argument. For example,
16499 this invocation is invalid in ISO C, because there is no comma after
16502 debug ("A message")
16504 GNU CPP permits you to completely omit the variable arguments in this
16505 way. In the above examples, the compiler would complain, though since
16506 the expansion of the macro still has the extra comma after the format
16509 To help solve this problem, CPP behaves specially for variable
16510 arguments used with the token paste operator, `##'. If instead you
16513 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
16515 and if the variable arguments are omitted or empty, the `##' operator
16516 causes the preprocessor to remove the comma before it. If you do
16517 provide some variable arguments in your macro invocation, GNU CPP does
16518 not complain about the paste operation and instead places the variable
16519 arguments after the comma. Just like any other pasted macro argument,
16520 these arguments are not macro expanded.
16523 File: gcc.info, Node: Escaped Newlines, Next: Subscripting, Prev: Variadic Macros, Up: C Extensions
16525 5.18 Slightly Looser Rules for Escaped Newlines
16526 ===============================================
16528 Recently, the preprocessor has relaxed its treatment of escaped
16529 newlines. Previously, the newline had to immediately follow a
16530 backslash. The current implementation allows whitespace in the form of
16531 spaces, horizontal and vertical tabs, and form feeds between the
16532 backslash and the subsequent newline. The preprocessor issues a
16533 warning, but treats it as a valid escaped newline and combines the two
16534 lines to form a single logical line. This works within comments and
16535 tokens, as well as between tokens. Comments are _not_ treated as
16536 whitespace for the purposes of this relaxation, since they have not yet
16537 been replaced with spaces.
16540 File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Escaped Newlines, Up: C Extensions
16542 5.19 Non-Lvalue Arrays May Have Subscripts
16543 ==========================================
16545 In ISO C99, arrays that are not lvalues still decay to pointers, and
16546 may be subscripted, although they may not be modified or used after the
16547 next sequence point and the unary `&' operator may not be applied to
16548 them. As an extension, GCC allows such arrays to be subscripted in C89
16549 mode, though otherwise they do not decay to pointers outside C99 mode.
16550 For example, this is valid in GNU C though not valid in C89:
16552 struct foo {int a[4];};
16558 return f().a[index];
16562 File: gcc.info, Node: Pointer Arith, Next: Initializers, Prev: Subscripting, Up: C Extensions
16564 5.20 Arithmetic on `void'- and Function-Pointers
16565 ================================================
16567 In GNU C, addition and subtraction operations are supported on pointers
16568 to `void' and on pointers to functions. This is done by treating the
16569 size of a `void' or of a function as 1.
16571 A consequence of this is that `sizeof' is also allowed on `void' and
16572 on function types, and returns 1.
16574 The option `-Wpointer-arith' requests a warning if these extensions
16578 File: gcc.info, Node: Initializers, Next: Compound Literals, Prev: Pointer Arith, Up: C Extensions
16580 5.21 Non-Constant Initializers
16581 ==============================
16583 As in standard C++ and ISO C99, the elements of an aggregate
16584 initializer for an automatic variable are not required to be constant
16585 expressions in GNU C. Here is an example of an initializer with
16586 run-time varying elements:
16588 foo (float f, float g)
16590 float beat_freqs[2] = { f-g, f+g };
16595 File: gcc.info, Node: Compound Literals, Next: Designated Inits, Prev: Initializers, Up: C Extensions
16597 5.22 Compound Literals
16598 ======================
16600 ISO C99 supports compound literals. A compound literal looks like a
16601 cast containing an initializer. Its value is an object of the type
16602 specified in the cast, containing the elements specified in the
16603 initializer; it is an lvalue. As an extension, GCC supports compound
16604 literals in C89 mode and in C++.
16606 Usually, the specified type is a structure. Assume that `struct foo'
16607 and `structure' are declared as shown:
16609 struct foo {int a; char b[2];} structure;
16611 Here is an example of constructing a `struct foo' with a compound
16614 structure = ((struct foo) {x + y, 'a', 0});
16616 This is equivalent to writing the following:
16619 struct foo temp = {x + y, 'a', 0};
16623 You can also construct an array. If all the elements of the compound
16624 literal are (made up of) simple constant expressions, suitable for use
16625 in initializers of objects of static storage duration, then the compound
16626 literal can be coerced to a pointer to its first element and used in
16627 such an initializer, as shown here:
16629 char **foo = (char *[]) { "x", "y", "z" };
16631 Compound literals for scalar types and union types are is also
16632 allowed, but then the compound literal is equivalent to a cast.
16634 As a GNU extension, GCC allows initialization of objects with static
16635 storage duration by compound literals (which is not possible in ISO
16636 C99, because the initializer is not a constant). It is handled as if
16637 the object was initialized only with the bracket enclosed list if the
16638 types of the compound literal and the object match. The initializer
16639 list of the compound literal must be constant. If the object being
16640 initialized has array type of unknown size, the size is determined by
16641 compound literal size.
16643 static struct foo x = (struct foo) {1, 'a', 'b'};
16644 static int y[] = (int []) {1, 2, 3};
16645 static int z[] = (int [3]) {1};
16647 The above lines are equivalent to the following:
16648 static struct foo x = {1, 'a', 'b'};
16649 static int y[] = {1, 2, 3};
16650 static int z[] = {1, 0, 0};
16653 File: gcc.info, Node: Designated Inits, Next: Cast to Union, Prev: Compound Literals, Up: C Extensions
16655 5.23 Designated Initializers
16656 ============================
16658 Standard C89 requires the elements of an initializer to appear in a
16659 fixed order, the same as the order of the elements in the array or
16660 structure being initialized.
16662 In ISO C99 you can give the elements in any order, specifying the array
16663 indices or structure field names they apply to, and GNU C allows this as
16664 an extension in C89 mode as well. This extension is not implemented in
16667 To specify an array index, write `[INDEX] =' before the element value.
16670 int a[6] = { [4] = 29, [2] = 15 };
16674 int a[6] = { 0, 0, 15, 0, 29, 0 };
16676 The index values must be constant expressions, even if the array being
16677 initialized is automatic.
16679 An alternative syntax for this which has been obsolete since GCC 2.5
16680 but GCC still accepts is to write `[INDEX]' before the element value,
16683 To initialize a range of elements to the same value, write `[FIRST ...
16684 LAST] = VALUE'. This is a GNU extension. For example,
16686 int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
16688 If the value in it has side-effects, the side-effects will happen only
16689 once, not for each initialized field by the range initializer.
16691 Note that the length of the array is the highest value specified plus
16694 In a structure initializer, specify the name of a field to initialize
16695 with `.FIELDNAME =' before the element value. For example, given the
16696 following structure,
16698 struct point { int x, y; };
16700 the following initialization
16702 struct point p = { .y = yvalue, .x = xvalue };
16706 struct point p = { xvalue, yvalue };
16708 Another syntax which has the same meaning, obsolete since GCC 2.5, is
16709 `FIELDNAME:', as shown here:
16711 struct point p = { y: yvalue, x: xvalue };
16713 The `[INDEX]' or `.FIELDNAME' is known as a "designator". You can
16714 also use a designator (or the obsolete colon syntax) when initializing
16715 a union, to specify which element of the union should be used. For
16718 union foo { int i; double d; };
16720 union foo f = { .d = 4 };
16722 will convert 4 to a `double' to store it in the union using the second
16723 element. By contrast, casting 4 to type `union foo' would store it
16724 into the union as the integer `i', since it is an integer. (*Note Cast
16727 You can combine this technique of naming elements with ordinary C
16728 initialization of successive elements. Each initializer element that
16729 does not have a designator applies to the next consecutive element of
16730 the array or structure. For example,
16732 int a[6] = { [1] = v1, v2, [4] = v4 };
16736 int a[6] = { 0, v1, v2, 0, v4, 0 };
16738 Labeling the elements of an array initializer is especially useful
16739 when the indices are characters or belong to an `enum' type. For
16742 int whitespace[256]
16743 = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
16744 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
16746 You can also write a series of `.FIELDNAME' and `[INDEX]' designators
16747 before an `=' to specify a nested subobject to initialize; the list is
16748 taken relative to the subobject corresponding to the closest
16749 surrounding brace pair. For example, with the `struct point'
16752 struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
16754 If the same field is initialized multiple times, it will have value from
16755 the last initialization. If any such overridden initialization has
16756 side-effect, it is unspecified whether the side-effect happens or not.
16757 Currently, GCC will discard them and issue a warning.
16760 File: gcc.info, Node: Case Ranges, Next: Mixed Declarations, Prev: Cast to Union, Up: C Extensions
16765 You can specify a range of consecutive values in a single `case' label,
16770 This has the same effect as the proper number of individual `case'
16771 labels, one for each integer value from LOW to HIGH, inclusive.
16773 This feature is especially useful for ranges of ASCII character codes:
16777 *Be careful:* Write spaces around the `...', for otherwise it may be
16778 parsed wrong when you use it with integer values. For example, write
16788 File: gcc.info, Node: Cast to Union, Next: Case Ranges, Prev: Designated Inits, Up: C Extensions
16790 5.25 Cast to a Union Type
16791 =========================
16793 A cast to union type is similar to other casts, except that the type
16794 specified is a union type. You can specify the type either with `union
16795 TAG' or with a typedef name. A cast to union is actually a constructor
16796 though, not a cast, and hence does not yield an lvalue like normal
16797 casts. (*Note Compound Literals::.)
16799 The types that may be cast to the union type are those of the members
16800 of the union. Thus, given the following union and variables:
16802 union foo { int i; double d; };
16806 both `x' and `y' can be cast to type `union foo'.
16808 Using the cast as the right-hand side of an assignment to a variable of
16809 union type is equivalent to storing in a member of the union:
16813 u = (union foo) x == u.i = x
16814 u = (union foo) y == u.d = y
16816 You can also use the union cast as a function argument:
16818 void hack (union foo);
16820 hack ((union foo) x);
16823 File: gcc.info, Node: Mixed Declarations, Next: Function Attributes, Prev: Case Ranges, Up: C Extensions
16825 5.26 Mixed Declarations and Code
16826 ================================
16828 ISO C99 and ISO C++ allow declarations and code to be freely mixed
16829 within compound statements. As an extension, GCC also allows this in
16830 C89 mode. For example, you could do:
16837 Each identifier is visible from where it is declared until the end of
16838 the enclosing block.
16841 File: gcc.info, Node: Function Attributes, Next: Attribute Syntax, Prev: Mixed Declarations, Up: C Extensions
16843 5.27 Declaring Attributes of Functions
16844 ======================================
16846 In GNU C, you declare certain things about functions called in your
16847 program which help the compiler optimize function calls and check your
16848 code more carefully.
16850 The keyword `__attribute__' allows you to specify special attributes
16851 when making a declaration. This keyword is followed by an attribute
16852 specification inside double parentheses. The following attributes are
16853 currently defined for functions on all targets: `aligned',
16854 `alloc_size', `noreturn', `returns_twice', `noinline', `always_inline',
16855 `flatten', `pure', `const', `nothrow', `sentinel', `format',
16856 `format_arg', `no_instrument_function', `section', `constructor',
16857 `destructor', `used', `unused', `deprecated', `weak', `malloc',
16858 `alias', `warn_unused_result', `nonnull', `gnu_inline',
16859 `externally_visible', `hot', `cold', `artificial', `error' and
16860 `warning'. Several other attributes are defined for functions on
16861 particular target systems. Other attributes, including `section' are
16862 supported for variables declarations (*note Variable Attributes::) and
16863 for types (*note Type Attributes::).
16865 You may also specify attributes with `__' preceding and following each
16866 keyword. This allows you to use them in header files without being
16867 concerned about a possible macro of the same name. For example, you
16868 may use `__noreturn__' instead of `noreturn'.
16870 *Note Attribute Syntax::, for details of the exact syntax for using
16874 The `alias' attribute causes the declaration to be emitted as an
16875 alias for another symbol, which must be specified. For instance,
16877 void __f () { /* Do something. */; }
16878 void f () __attribute__ ((weak, alias ("__f")));
16880 defines `f' to be a weak alias for `__f'. In C++, the mangled
16881 name for the target must be used. It is an error if `__f' is not
16882 defined in the same translation unit.
16884 Not all target machines support this attribute.
16886 `aligned (ALIGNMENT)'
16887 This attribute specifies a minimum alignment for the function,
16890 You cannot use this attribute to decrease the alignment of a
16891 function, only to increase it. However, when you explicitly
16892 specify a function alignment this will override the effect of the
16893 `-falign-functions' (*note Optimize Options::) option for this
16896 Note that the effectiveness of `aligned' attributes may be limited
16897 by inherent limitations in your linker. On many systems, the
16898 linker is only able to arrange for functions to be aligned up to a
16899 certain maximum alignment. (For some linkers, the maximum
16900 supported alignment may be very very small.) See your linker
16901 documentation for further information.
16903 The `aligned' attribute can also be used for variables and fields
16904 (*note Variable Attributes::.)
16907 The `alloc_size' attribute is used to tell the compiler that the
16908 function return value points to memory, where the size is given by
16909 one or two of the functions parameters. GCC uses this information
16910 to improve the correctness of `__builtin_object_size'.
16912 The function parameter(s) denoting the allocated size are
16913 specified by one or two integer arguments supplied to the
16914 attribute. The allocated size is either the value of the single
16915 function argument specified or the product of the two function
16916 arguments specified. Argument numbering starts at one.
16920 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
16921 void my_realloc(void* size_t) __attribute__((alloc_size(2)))
16923 declares that my_calloc will return memory of the size given by
16924 the product of parameter 1 and 2 and that my_realloc will return
16925 memory of the size given by parameter 2.
16928 Generally, functions are not inlined unless optimization is
16929 specified. For functions declared inline, this attribute inlines
16930 the function even if no optimization level was specified.
16933 This attribute should be used with a function which is also
16934 declared with the `inline' keyword. It directs GCC to treat the
16935 function as if it were defined in gnu89 mode even when compiling
16936 in C99 or gnu99 mode.
16938 If the function is declared `extern', then this definition of the
16939 function is used only for inlining. In no case is the function
16940 compiled as a standalone function, not even if you take its address
16941 explicitly. Such an address becomes an external reference, as if
16942 you had only declared the function, and had not defined it. This
16943 has almost the effect of a macro. The way to use this is to put a
16944 function definition in a header file with this attribute, and put
16945 another copy of the function, without `extern', in a library file.
16946 The definition in the header file will cause most calls to the
16947 function to be inlined. If any uses of the function remain, they
16948 will refer to the single copy in the library. Note that the two
16949 definitions of the functions need not be precisely the same,
16950 although if they do not have the same effect your program may
16953 In C, if the function is neither `extern' nor `static', then the
16954 function is compiled as a standalone function, as well as being
16955 inlined where possible.
16957 This is how GCC traditionally handled functions declared `inline'.
16958 Since ISO C99 specifies a different semantics for `inline', this
16959 function attribute is provided as a transition measure and as a
16960 useful feature in its own right. This attribute is available in
16961 GCC 4.1.3 and later. It is available if either of the
16962 preprocessor macros `__GNUC_GNU_INLINE__' or
16963 `__GNUC_STDC_INLINE__' are defined. *Note An Inline Function is
16964 As Fast As a Macro: Inline.
16966 In C++, this attribute does not depend on `extern' in any way, but
16967 it still requires the `inline' keyword to enable its special
16971 This attribute is useful for small inline wrappers which if
16972 possible should appear during debugging as a unit, depending on
16973 the debug info format it will either mean marking the function as
16974 artificial or using the caller location for all instructions
16975 within the inlined body.
16978 Generally, inlining into a function is limited. For a function
16979 marked with this attribute, every call inside this function will
16980 be inlined, if possible. Whether the function itself is
16981 considered for inlining depends on its size and the current
16982 inlining parameters. The `flatten' attribute only works reliably
16983 in unit-at-a-time mode.
16985 `error ("MESSAGE")'
16986 If this attribute is used on a function declaration and a call to
16987 such a function is not eliminated through dead code elimination or
16988 other optimizations, an error which will include MESSAGE will be
16989 diagnosed. This is useful for compile time checking, especially
16990 together with `__builtin_constant_p' and inline functions where
16991 checking the inline function arguments is not possible through
16992 `extern char [(condition) ? 1 : -1];' tricks. While it is
16993 possible to leave the function undefined and thus invoke a link
16994 failure, when using this attribute the problem will be diagnosed
16995 earlier and with exact location of the call even in presence of
16996 inline functions or when not emitting debugging information.
16998 `warning ("MESSAGE")'
16999 If this attribute is used on a function declaration and a call to
17000 such a function is not eliminated through dead code elimination or
17001 other optimizations, a warning which will include MESSAGE will be
17002 diagnosed. This is useful for compile time checking, especially
17003 together with `__builtin_constant_p' and inline functions. While
17004 it is possible to define the function with a message in
17005 `.gnu.warning*' section, when using this attribute the problem
17006 will be diagnosed earlier and with exact location of the call even
17007 in presence of inline functions or when not emitting debugging
17011 On the Intel 386, the `cdecl' attribute causes the compiler to
17012 assume that the calling function will pop off the stack space used
17013 to pass arguments. This is useful to override the effects of the
17017 Many functions do not examine any values except their arguments,
17018 and have no effects except the return value. Basically this is
17019 just slightly more strict class than the `pure' attribute below,
17020 since function is not allowed to read global memory.
17022 Note that a function that has pointer arguments and examines the
17023 data pointed to must _not_ be declared `const'. Likewise, a
17024 function that calls a non-`const' function usually must not be
17025 `const'. It does not make sense for a `const' function to return
17028 The attribute `const' is not implemented in GCC versions earlier
17029 than 2.5. An alternative way to declare that a function has no
17030 side effects, which works in the current version and in some older
17031 versions, is as follows:
17033 typedef int intfn ();
17035 extern const intfn square;
17037 This approach does not work in GNU C++ from 2.6.0 on, since the
17038 language specifies that the `const' must be attached to the return
17043 `constructor (PRIORITY)'
17044 `destructor (PRIORITY)'
17045 The `constructor' attribute causes the function to be called
17046 automatically before execution enters `main ()'. Similarly, the
17047 `destructor' attribute causes the function to be called
17048 automatically after `main ()' has completed or `exit ()' has been
17049 called. Functions with these attributes are useful for
17050 initializing data that will be used implicitly during the
17051 execution of the program.
17053 You may provide an optional integer priority to control the order
17054 in which constructor and destructor functions are run. A
17055 constructor with a smaller priority number runs before a
17056 constructor with a larger priority number; the opposite
17057 relationship holds for destructors. So, if you have a constructor
17058 that allocates a resource and a destructor that deallocates the
17059 same resource, both functions typically have the same priority.
17060 The priorities for constructor and destructor functions are the
17061 same as those specified for namespace-scope C++ objects (*note C++
17064 These attributes are not currently implemented for Objective-C.
17067 The `deprecated' attribute results in a warning if the function is
17068 used anywhere in the source file. This is useful when identifying
17069 functions that are expected to be removed in a future version of a
17070 program. The warning also includes the location of the declaration
17071 of the deprecated function, to enable users to easily find further
17072 information about why the function is deprecated, or what they
17073 should do instead. Note that the warnings only occurs for uses:
17075 int old_fn () __attribute__ ((deprecated));
17077 int (*fn_ptr)() = old_fn;
17079 results in a warning on line 3 but not line 2.
17081 The `deprecated' attribute can also be used for variables and
17082 types (*note Variable Attributes::, *note Type Attributes::.)
17085 On Microsoft Windows targets and Symbian OS targets the
17086 `dllexport' attribute causes the compiler to provide a global
17087 pointer to a pointer in a DLL, so that it can be referenced with
17088 the `dllimport' attribute. On Microsoft Windows targets, the
17089 pointer name is formed by combining `_imp__' and the function or
17092 You can use `__declspec(dllexport)' as a synonym for
17093 `__attribute__ ((dllexport))' for compatibility with other
17096 On systems that support the `visibility' attribute, this attribute
17097 also implies "default" visibility. It is an error to explicitly
17098 specify any other visibility.
17100 Currently, the `dllexport' attribute is ignored for inlined
17101 functions, unless the `-fkeep-inline-functions' flag has been
17102 used. The attribute is also ignored for undefined symbols.
17104 When applied to C++ classes, the attribute marks defined
17105 non-inlined member functions and static data members as exports.
17106 Static consts initialized in-class are not marked unless they are
17107 also defined out-of-class.
17109 For Microsoft Windows targets there are alternative methods for
17110 including the symbol in the DLL's export table such as using a
17111 `.def' file with an `EXPORTS' section or, with GNU ld, using the
17112 `--export-all' linker flag.
17115 On Microsoft Windows and Symbian OS targets, the `dllimport'
17116 attribute causes the compiler to reference a function or variable
17117 via a global pointer to a pointer that is set up by the DLL
17118 exporting the symbol. The attribute implies `extern'. On
17119 Microsoft Windows targets, the pointer name is formed by combining
17120 `_imp__' and the function or variable name.
17122 You can use `__declspec(dllimport)' as a synonym for
17123 `__attribute__ ((dllimport))' for compatibility with other
17126 On systems that support the `visibility' attribute, this attribute
17127 also implies "default" visibility. It is an error to explicitly
17128 specify any other visibility.
17130 Currently, the attribute is ignored for inlined functions. If the
17131 attribute is applied to a symbol _definition_, an error is
17132 reported. If a symbol previously declared `dllimport' is later
17133 defined, the attribute is ignored in subsequent references, and a
17134 warning is emitted. The attribute is also overridden by a
17135 subsequent declaration as `dllexport'.
17137 When applied to C++ classes, the attribute marks non-inlined
17138 member functions and static data members as imports. However, the
17139 attribute is ignored for virtual methods to allow creation of
17140 vtables using thunks.
17142 On the SH Symbian OS target the `dllimport' attribute also has
17143 another affect--it can cause the vtable and run-time type
17144 information for a class to be exported. This happens when the
17145 class has a dllimport'ed constructor or a non-inline, non-pure
17146 virtual function and, for either of those two conditions, the
17147 class also has a inline constructor or destructor and has a key
17148 function that is defined in the current translation unit.
17150 For Microsoft Windows based targets the use of the `dllimport'
17151 attribute on functions is not necessary, but provides a small
17152 performance benefit by eliminating a thunk in the DLL. The use of
17153 the `dllimport' attribute on imported variables was required on
17154 older versions of the GNU linker, but can now be avoided by
17155 passing the `--enable-auto-import' switch to the GNU linker. As
17156 with functions, using the attribute for a variable eliminates a
17159 One drawback to using this attribute is that a pointer to a
17160 _variable_ marked as `dllimport' cannot be used as a constant
17161 address. However, a pointer to a _function_ with the `dllimport'
17162 attribute can be used as a constant initializer; in this case, the
17163 address of a stub function in the import lib is referenced. On
17164 Microsoft Windows targets, the attribute can be disabled for
17165 functions by setting the `-mnop-fun-dllimport' flag.
17168 Use this attribute on the H8/300, H8/300H, and H8S to indicate
17169 that the specified variable should be placed into the eight bit
17170 data section. The compiler will generate more efficient code for
17171 certain operations on data in the eight bit data area. Note the
17172 eight bit data area is limited to 256 bytes of data.
17174 You must use GAS and GLD from GNU binutils version 2.7 or later for
17175 this attribute to work correctly.
17177 `exception_handler'
17178 Use this attribute on the Blackfin to indicate that the specified
17179 function is an exception handler. The compiler will generate
17180 function entry and exit sequences suitable for use in an exception
17181 handler when this attribute is present.
17184 On 68HC11 and 68HC12 the `far' attribute causes the compiler to
17185 use a calling convention that takes care of switching memory banks
17186 when entering and leaving a function. This calling convention is
17187 also the default when using the `-mlong-calls' option.
17189 On 68HC12 the compiler will use the `call' and `rtc' instructions
17190 to call and return from a function.
17192 On 68HC11 the compiler will generate a sequence of instructions to
17193 invoke a board-specific routine to switch the memory bank and call
17194 the real function. The board-specific routine simulates a `call'.
17195 At the end of a function, it will jump to a board-specific routine
17196 instead of using `rts'. The board-specific return routine
17197 simulates the `rtc'.
17200 On the Intel 386, the `fastcall' attribute causes the compiler to
17201 pass the first argument (if of integral type) in the register ECX
17202 and the second argument (if of integral type) in the register EDX.
17203 Subsequent and other typed arguments are passed on the stack.
17204 The called function will pop the arguments off the stack. If the
17205 number of arguments is variable all arguments are pushed on the
17208 `format (ARCHETYPE, STRING-INDEX, FIRST-TO-CHECK)'
17209 The `format' attribute specifies that a function takes `printf',
17210 `scanf', `strftime' or `strfmon' style arguments which should be
17211 type-checked against a format string. For example, the
17215 my_printf (void *my_object, const char *my_format, ...)
17216 __attribute__ ((format (printf, 2, 3)));
17218 causes the compiler to check the arguments in calls to `my_printf'
17219 for consistency with the `printf' style format string argument
17222 The parameter ARCHETYPE determines how the format string is
17223 interpreted, and should be `printf', `scanf', `strftime' or
17224 `strfmon'. (You can also use `__printf__', `__scanf__',
17225 `__strftime__' or `__strfmon__'.) The parameter STRING-INDEX
17226 specifies which argument is the format string argument (starting
17227 from 1), while FIRST-TO-CHECK is the number of the first argument
17228 to check against the format string. For functions where the
17229 arguments are not available to be checked (such as `vprintf'),
17230 specify the third parameter as zero. In this case the compiler
17231 only checks the format string for consistency. For `strftime'
17232 formats, the third parameter is required to be zero. Since
17233 non-static C++ methods have an implicit `this' argument, the
17234 arguments of such methods should be counted from two, not one, when
17235 giving values for STRING-INDEX and FIRST-TO-CHECK.
17237 In the example above, the format string (`my_format') is the second
17238 argument of the function `my_print', and the arguments to check
17239 start with the third argument, so the correct parameters for the
17240 format attribute are 2 and 3.
17242 The `format' attribute allows you to identify your own functions
17243 which take format strings as arguments, so that GCC can check the
17244 calls to these functions for errors. The compiler always (unless
17245 `-ffreestanding' or `-fno-builtin' is used) checks formats for the
17246 standard library functions `printf', `fprintf', `sprintf',
17247 `scanf', `fscanf', `sscanf', `strftime', `vprintf', `vfprintf' and
17248 `vsprintf' whenever such warnings are requested (using
17249 `-Wformat'), so there is no need to modify the header file
17250 `stdio.h'. In C99 mode, the functions `snprintf', `vsnprintf',
17251 `vscanf', `vfscanf' and `vsscanf' are also checked. Except in
17252 strictly conforming C standard modes, the X/Open function
17253 `strfmon' is also checked as are `printf_unlocked' and
17254 `fprintf_unlocked'. *Note Options Controlling C Dialect: C
17257 The target may provide additional types of format checks. *Note
17258 Format Checks Specific to Particular Target Machines: Target
17261 `format_arg (STRING-INDEX)'
17262 The `format_arg' attribute specifies that a function takes a format
17263 string for a `printf', `scanf', `strftime' or `strfmon' style
17264 function and modifies it (for example, to translate it into
17265 another language), so the result can be passed to a `printf',
17266 `scanf', `strftime' or `strfmon' style function (with the
17267 remaining arguments to the format function the same as they would
17268 have been for the unmodified string). For example, the
17272 my_dgettext (char *my_domain, const char *my_format)
17273 __attribute__ ((format_arg (2)));
17275 causes the compiler to check the arguments in calls to a `printf',
17276 `scanf', `strftime' or `strfmon' type function, whose format
17277 string argument is a call to the `my_dgettext' function, for
17278 consistency with the format string argument `my_format'. If the
17279 `format_arg' attribute had not been specified, all the compiler
17280 could tell in such calls to format functions would be that the
17281 format string argument is not constant; this would generate a
17282 warning when `-Wformat-nonliteral' is used, but the calls could
17283 not be checked without the attribute.
17285 The parameter STRING-INDEX specifies which argument is the format
17286 string argument (starting from one). Since non-static C++ methods
17287 have an implicit `this' argument, the arguments of such methods
17288 should be counted from two.
17290 The `format-arg' attribute allows you to identify your own
17291 functions which modify format strings, so that GCC can check the
17292 calls to `printf', `scanf', `strftime' or `strfmon' type function
17293 whose operands are a call to one of your own function. The
17294 compiler always treats `gettext', `dgettext', and `dcgettext' in
17295 this manner except when strict ISO C support is requested by
17296 `-ansi' or an appropriate `-std' option, or `-ffreestanding' or
17297 `-fno-builtin' is used. *Note Options Controlling C Dialect: C
17301 Use this attribute on the H8/300, H8/300H, and H8S to indicate
17302 that the specified function should be called through the function
17303 vector. Calling a function through the function vector will
17304 reduce code size, however; the function vector has a limited size
17305 (maximum 128 entries on the H8/300 and 64 entries on the H8/300H
17306 and H8S) and shares space with the interrupt vector.
17308 You must use GAS and GLD from GNU binutils version 2.7 or later for
17309 this attribute to work correctly.
17311 On M16C/M32C targets, the `function_vector' attribute declares a
17312 special page subroutine call function. Use of this attribute
17313 reduces the code size by 2 bytes for each call generated to the
17314 subroutine. The argument to the attribute is the vector number
17315 entry from the special page vector table which contains the 16
17316 low-order bits of the subroutine's entry address. Each vector
17317 table has special page number (18 to 255) which are used in `jsrs'
17318 instruction. Jump addresses of the routines are generated by
17319 adding 0x0F0000 (in case of M16C targets) or 0xFF0000 (in case of
17320 M32C targets), to the 2 byte addresses set in the vector table.
17321 Therefore you need to ensure that all the special page vector
17322 routines should get mapped within the address range 0x0F0000 to
17323 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF (for M32C).
17325 In the following example 2 bytes will be saved for each call to
17328 void foo (void) __attribute__((function_vector(0x18)));
17338 If functions are defined in one file and are called in another
17339 file, then be sure to write this declaration in both files.
17341 This attribute is ignored for R8C target.
17344 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k, MS1,
17345 and Xstormy16 ports to indicate that the specified function is an
17346 interrupt handler. The compiler will generate function entry and
17347 exit sequences suitable for use in an interrupt handler when this
17348 attribute is present.
17350 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S,
17351 and SH processors can be specified via the `interrupt_handler'
17354 Note, on the AVR, interrupts will be enabled inside the function.
17356 Note, for the ARM, you can specify the kind of interrupt to be
17357 handled by adding an optional parameter to the interrupt attribute
17360 void f () __attribute__ ((interrupt ("IRQ")));
17362 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT
17365 On ARMv7-M the interrupt type is ignored, and the attribute means
17366 the function may be called with a word aligned stack pointer.
17368 `interrupt_handler'
17369 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S,
17370 and SH to indicate that the specified function is an interrupt
17371 handler. The compiler will generate function entry and exit
17372 sequences suitable for use in an interrupt handler when this
17373 attribute is present.
17376 Use this attribute on fido, a subarchitecture of the m68k, to
17377 indicate that the specified function is an interrupt handler that
17378 is designed to run as a thread. The compiler omits generate
17379 prologue/epilogue sequences and replaces the return instruction
17380 with a `sleep' instruction. This attribute is available only on
17384 When used together with `interrupt_handler', `exception_handler'
17385 or `nmi_handler', code will be generated to load the stack pointer
17386 from the USP register in the function prologue.
17389 This attribute specifies a function to be placed into L1
17390 Instruction SRAM. The function will be put into a specific section
17391 named `.l1.text'. With `-mfdpic', function calls with a such
17392 function as the callee or caller will use inlined PLT.
17394 `long_call/short_call'
17395 This attribute specifies how a particular function is called on
17396 ARM. Both attributes override the `-mlong-calls' (*note ARM
17397 Options::) command line switch and `#pragma long_calls' settings.
17398 The `long_call' attribute indicates that the function might be far
17399 away from the call site and require a different (more expensive)
17400 calling sequence. The `short_call' attribute always places the
17401 offset to the function from the call site into the `BL'
17402 instruction directly.
17404 `longcall/shortcall'
17405 On the Blackfin, RS/6000 and PowerPC, the `longcall' attribute
17406 indicates that the function might be far away from the call site
17407 and require a different (more expensive) calling sequence. The
17408 `shortcall' attribute indicates that the function is always close
17409 enough for the shorter calling sequence to be used. These
17410 attributes override both the `-mlongcall' switch and, on the
17411 RS/6000 and PowerPC, the `#pragma longcall' setting.
17413 *Note RS/6000 and PowerPC Options::, for more information on
17414 whether long calls are necessary.
17416 `long_call/near/far'
17417 These attributes specify how a particular function is called on
17418 MIPS. The attributes override the `-mlong-calls' (*note MIPS
17419 Options::) command-line switch. The `long_call' and `far'
17420 attributes are synonyms, and cause the compiler to always call the
17421 function by first loading its address into a register, and then
17422 using the contents of that register. The `near' attribute has the
17423 opposite effect; it specifies that non-PIC calls should be made
17424 using the more efficient `jal' instruction.
17427 The `malloc' attribute is used to tell the compiler that a function
17428 may be treated as if any non-`NULL' pointer it returns cannot
17429 alias any other pointer valid when the function returns. This
17430 will often improve optimization. Standard functions with this
17431 property include `malloc' and `calloc'. `realloc'-like functions
17432 have this property as long as the old pointer is never referred to
17433 (including comparing it to the new pointer) after the function
17434 returns a non-`NULL' value.
17437 On MIPS targets, you can use the `mips16' and `nomips16' function
17438 attributes to locally select or turn off MIPS16 code generation.
17439 A function with the `mips16' attribute is emitted as MIPS16 code,
17440 while MIPS16 code generation is disabled for functions with the
17441 `nomips16' attribute. These attributes override the `-mips16' and
17442 `-mno-mips16' options on the command line (*note MIPS Options::).
17444 When compiling files containing mixed MIPS16 and non-MIPS16 code,
17445 the preprocessor symbol `__mips16' reflects the setting on the
17446 command line, not that within individual functions. Mixed MIPS16
17447 and non-MIPS16 code may interact badly with some GCC extensions
17448 such as `__builtin_apply' (*note Constructing Calls::).
17450 `model (MODEL-NAME)'
17451 On the M32R/D, use this attribute to set the addressability of an
17452 object, and of the code generated for a function. The identifier
17453 MODEL-NAME is one of `small', `medium', or `large', representing
17454 each of the code models.
17456 Small model objects live in the lower 16MB of memory (so that their
17457 addresses can be loaded with the `ld24' instruction), and are
17458 callable with the `bl' instruction.
17460 Medium model objects may live anywhere in the 32-bit address space
17461 (the compiler will generate `seth/add3' instructions to load their
17462 addresses), and are callable with the `bl' instruction.
17464 Large model objects may live anywhere in the 32-bit address space
17465 (the compiler will generate `seth/add3' instructions to load their
17466 addresses), and may not be reachable with the `bl' instruction
17467 (the compiler will generate the much slower `seth/add3/jl'
17468 instruction sequence).
17470 On IA-64, use this attribute to set the addressability of an
17471 object. At present, the only supported identifier for MODEL-NAME
17472 is `small', indicating addressability via "small" (22-bit)
17473 addresses (so that their addresses can be loaded with the `addl'
17474 instruction). Caveat: such addressing is by definition not
17475 position independent and hence this attribute must not be used for
17476 objects defined by shared libraries.
17479 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate
17480 that the specified function does not need prologue/epilogue
17481 sequences generated by the compiler. It is up to the programmer
17482 to provide these sequences.
17485 On 68HC11 and 68HC12 the `near' attribute causes the compiler to
17486 use the normal calling convention based on `jsr' and `rts'. This
17487 attribute can be used to cancel the effect of the `-mlong-calls'
17491 Use this attribute together with `interrupt_handler',
17492 `exception_handler' or `nmi_handler' to indicate that the function
17493 entry code should enable nested interrupts or exceptions.
17496 Use this attribute on the Blackfin to indicate that the specified
17497 function is an NMI handler. The compiler will generate function
17498 entry and exit sequences suitable for use in an NMI handler when
17499 this attribute is present.
17501 `no_instrument_function'
17502 If `-finstrument-functions' is given, profiling function calls will
17503 be generated at entry and exit of most user-compiled functions.
17504 Functions with this attribute will not be so instrumented.
17507 This function attribute prevents a function from being considered
17508 for inlining. If the function does not have side-effects, there
17509 are optimizations other than inlining that causes function calls
17510 to be optimized away, although the function call is live. To keep
17511 such calls from being optimized away, put
17513 (*note Extended Asm::) in the called function, to serve as a
17514 special side-effect.
17516 `nonnull (ARG-INDEX, ...)'
17517 The `nonnull' attribute specifies that some function parameters
17518 should be non-null pointers. For instance, the declaration:
17521 my_memcpy (void *dest, const void *src, size_t len)
17522 __attribute__((nonnull (1, 2)));
17524 causes the compiler to check that, in calls to `my_memcpy',
17525 arguments DEST and SRC are non-null. If the compiler determines
17526 that a null pointer is passed in an argument slot marked as
17527 non-null, and the `-Wnonnull' option is enabled, a warning is
17528 issued. The compiler may also choose to make optimizations based
17529 on the knowledge that certain function arguments will not be null.
17531 If no argument index list is given to the `nonnull' attribute, all
17532 pointer arguments are marked as non-null. To illustrate, the
17533 following declaration is equivalent to the previous example:
17536 my_memcpy (void *dest, const void *src, size_t len)
17537 __attribute__((nonnull));
17540 A few standard library functions, such as `abort' and `exit',
17541 cannot return. GCC knows this automatically. Some programs define
17542 their own functions that never return. You can declare them
17543 `noreturn' to tell the compiler this fact. For example,
17545 void fatal () __attribute__ ((noreturn));
17550 /* ... */ /* Print error message. */ /* ... */
17554 The `noreturn' keyword tells the compiler to assume that `fatal'
17555 cannot return. It can then optimize without regard to what would
17556 happen if `fatal' ever did return. This makes slightly better
17557 code. More importantly, it helps avoid spurious warnings of
17558 uninitialized variables.
17560 The `noreturn' keyword does not affect the exceptional path when
17561 that applies: a `noreturn'-marked function may still return to the
17562 caller by throwing an exception or calling `longjmp'.
17564 Do not assume that registers saved by the calling function are
17565 restored before calling the `noreturn' function.
17567 It does not make sense for a `noreturn' function to have a return
17568 type other than `void'.
17570 The attribute `noreturn' is not implemented in GCC versions
17571 earlier than 2.5. An alternative way to declare that a function
17572 does not return, which works in the current version and in some
17573 older versions, is as follows:
17575 typedef void voidfn ();
17577 volatile voidfn fatal;
17579 This approach does not work in GNU C++.
17582 The `nothrow' attribute is used to inform the compiler that a
17583 function cannot throw an exception. For example, most functions in
17584 the standard C library can be guaranteed not to throw an exception
17585 with the notable exceptions of `qsort' and `bsearch' that take
17586 function pointer arguments. The `nothrow' attribute is not
17587 implemented in GCC versions earlier than 3.3.
17590 Many functions have no effects except the return value and their
17591 return value depends only on the parameters and/or global
17592 variables. Such a function can be subject to common subexpression
17593 elimination and loop optimization just as an arithmetic operator
17594 would be. These functions should be declared with the attribute
17595 `pure'. For example,
17597 int square (int) __attribute__ ((pure));
17599 says that the hypothetical function `square' is safe to call fewer
17600 times than the program says.
17602 Some of common examples of pure functions are `strlen' or `memcmp'.
17603 Interesting non-pure functions are functions with infinite loops
17604 or those depending on volatile memory or other system resource,
17605 that may change between two consecutive calls (such as `feof' in a
17606 multithreading environment).
17608 The attribute `pure' is not implemented in GCC versions earlier
17612 The `hot' attribute is used to inform the compiler that a function
17613 is a hot spot of the compiled program. The function is optimized
17614 more aggressively and on many target it is placed into special
17615 subsection of the text section so all hot functions appears close
17616 together improving locality.
17618 When profile feedback is available, via `-fprofile-use', hot
17619 functions are automatically detected and this attribute is ignored.
17621 The `hot' attribute is not implemented in GCC versions earlier
17625 The `cold' attribute is used to inform the compiler that a
17626 function is unlikely executed. The function is optimized for size
17627 rather than speed and on many targets it is placed into special
17628 subsection of the text section so all cold functions appears close
17629 together improving code locality of non-cold parts of program.
17630 The paths leading to call of cold functions within code are marked
17631 as unlikely by the branch prediction mechanism. It is thus useful
17632 to mark functions used to handle unlikely conditions, such as
17633 `perror', as cold to improve optimization of hot functions that do
17634 call marked functions in rare occasions.
17636 When profile feedback is available, via `-fprofile-use', hot
17637 functions are automatically detected and this attribute is ignored.
17639 The `hot' attribute is not implemented in GCC versions earlier
17643 On the Intel 386, the `regparm' attribute causes the compiler to
17644 pass arguments number one to NUMBER if they are of integral type
17645 in registers EAX, EDX, and ECX instead of on the stack. Functions
17646 that take a variable number of arguments will continue to be
17647 passed all of their arguments on the stack.
17649 Beware that on some ELF systems this attribute is unsuitable for
17650 global functions in shared libraries with lazy binding (which is
17651 the default). Lazy binding will send the first call via resolving
17652 code in the loader, which might assume EAX, EDX and ECX can be
17653 clobbered, as per the standard calling conventions. Solaris 8 is
17654 affected by this. GNU systems with GLIBC 2.1 or higher, and
17655 FreeBSD, are believed to be safe since the loaders there save all
17656 registers. (Lazy binding can be disabled with the linker or the
17657 loader if desired, to avoid the problem.)
17660 On the Intel 386 with SSE support, the `sseregparm' attribute
17661 causes the compiler to pass up to 3 floating point arguments in
17662 SSE registers instead of on the stack. Functions that take a
17663 variable number of arguments will continue to pass all of their
17664 floating point arguments on the stack.
17666 `force_align_arg_pointer'
17667 On the Intel x86, the `force_align_arg_pointer' attribute may be
17668 applied to individual function definitions, generating an alternate
17669 prologue and epilogue that realigns the runtime stack. This
17670 supports mixing legacy codes that run with a 4-byte aligned stack
17671 with modern codes that keep a 16-byte stack for SSE compatibility.
17672 The alternate prologue and epilogue are slower and bigger than
17673 the regular ones, and the alternate prologue requires a scratch
17674 register; this lowers the number of registers available if used in
17675 conjunction with the `regparm' attribute. The
17676 `force_align_arg_pointer' attribute is incompatible with nested
17677 functions; this is considered a hard error.
17680 The `returns_twice' attribute tells the compiler that a function
17681 may return more than one time. The compiler will ensure that all
17682 registers are dead before calling such a function and will emit a
17683 warning about the variables that may be clobbered after the second
17684 return from the function. Examples of such functions are `setjmp'
17685 and `vfork'. The `longjmp'-like counterpart of such function, if
17686 any, might need to be marked with the `noreturn' attribute.
17689 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to
17690 indicate that all registers except the stack pointer should be
17691 saved in the prologue regardless of whether they are used or not.
17693 `section ("SECTION-NAME")'
17694 Normally, the compiler places the code it generates in the `text'
17695 section. Sometimes, however, you need additional sections, or you
17696 need certain particular functions to appear in special sections.
17697 The `section' attribute specifies that a function lives in a
17698 particular section. For example, the declaration:
17700 extern void foobar (void) __attribute__ ((section ("bar")));
17702 puts the function `foobar' in the `bar' section.
17704 Some file formats do not support arbitrary sections so the
17705 `section' attribute is not available on all platforms. If you
17706 need to map the entire contents of a module to a particular
17707 section, consider using the facilities of the linker instead.
17710 This function attribute ensures that a parameter in a function
17711 call is an explicit `NULL'. The attribute is only valid on
17712 variadic functions. By default, the sentinel is located at
17713 position zero, the last parameter of the function call. If an
17714 optional integer position argument P is supplied to the attribute,
17715 the sentinel must be located at position P counting backwards from
17716 the end of the argument list.
17718 __attribute__ ((sentinel))
17720 __attribute__ ((sentinel(0)))
17722 The attribute is automatically set with a position of 0 for the
17723 built-in functions `execl' and `execlp'. The built-in function
17724 `execle' has the attribute set with a position of 1.
17726 A valid `NULL' in this context is defined as zero with any pointer
17727 type. If your system defines the `NULL' macro with an integer type
17728 then you need to add an explicit cast. GCC replaces `stddef.h'
17729 with a copy that redefines NULL appropriately.
17731 The warnings for missing or incorrect sentinels are enabled with
17735 See long_call/short_call.
17738 See longcall/shortcall.
17741 Use this attribute on the AVR to indicate that the specified
17742 function is a signal handler. The compiler will generate function
17743 entry and exit sequences suitable for use in a signal handler when
17744 this attribute is present. Interrupts will be disabled inside the
17748 Use this attribute on the SH to indicate an `interrupt_handler'
17749 function should switch to an alternate stack. It expects a string
17750 argument that names a global variable holding the address of the
17754 void f () __attribute__ ((interrupt_handler,
17755 sp_switch ("alt_stack")));
17758 On the Intel 386, the `stdcall' attribute causes the compiler to
17759 assume that the called function will pop off the stack space used
17760 to pass arguments, unless it takes a variable number of arguments.
17763 Use this attribute on the H8/300H and H8S to indicate that the
17764 specified variable should be placed into the tiny data section.
17765 The compiler will generate more efficient code for loads and stores
17766 on data in the tiny data section. Note the tiny data area is
17767 limited to slightly under 32kbytes of data.
17770 Use this attribute on the SH for an `interrupt_handler' to return
17771 using `trapa' instead of `rte'. This attribute expects an integer
17772 argument specifying the trap number to be used.
17775 This attribute, attached to a function, means that the function is
17776 meant to be possibly unused. GCC will not produce a warning for
17780 This attribute, attached to a function, means that code must be
17781 emitted for the function even if it appears that the function is
17782 not referenced. This is useful, for example, when the function is
17783 referenced only in inline assembly.
17786 This attribute, attached to a global variable or function, renames
17787 a symbol to contain a version string, thus allowing for function
17788 level versioning. HP-UX system header files may use version level
17789 functioning for some system calls.
17791 extern int foo () __attribute__((version_id ("20040821")));
17793 Calls to FOO will be mapped to calls to FOO{20040821}.
17795 `visibility ("VISIBILITY_TYPE")'
17796 This attribute affects the linkage of the declaration to which it
17797 is attached. There are four supported VISIBILITY_TYPE values:
17798 default, hidden, protected or internal visibility.
17800 void __attribute__ ((visibility ("protected")))
17801 f () { /* Do something. */; }
17802 int i __attribute__ ((visibility ("hidden")));
17804 The possible values of VISIBILITY_TYPE correspond to the
17805 visibility settings in the ELF gABI.
17808 Default visibility is the normal case for the object file
17809 format. This value is available for the visibility attribute
17810 to override other options that may change the assumed
17811 visibility of entities.
17813 On ELF, default visibility means that the declaration is
17814 visible to other modules and, in shared libraries, means that
17815 the declared entity may be overridden.
17817 On Darwin, default visibility means that the declaration is
17818 visible to other modules.
17820 Default visibility corresponds to "external linkage" in the
17824 Hidden visibility indicates that the entity declared will
17825 have a new form of linkage, which we'll call "hidden
17826 linkage". Two declarations of an object with hidden linkage
17827 refer to the same object if they are in the same shared
17831 Internal visibility is like hidden visibility, but with
17832 additional processor specific semantics. Unless otherwise
17833 specified by the psABI, GCC defines internal visibility to
17834 mean that a function is _never_ called from another module.
17835 Compare this with hidden functions which, while they cannot
17836 be referenced directly by other modules, can be referenced
17837 indirectly via function pointers. By indicating that a
17838 function cannot be called from outside the module, GCC may
17839 for instance omit the load of a PIC register since it is known
17840 that the calling function loaded the correct value.
17843 Protected visibility is like default visibility except that it
17844 indicates that references within the defining module will
17845 bind to the definition in that module. That is, the declared
17846 entity cannot be overridden by another module.
17849 All visibilities are supported on many, but not all, ELF targets
17850 (supported when the assembler supports the `.visibility'
17851 pseudo-op). Default visibility is supported everywhere. Hidden
17852 visibility is supported on Darwin targets.
17854 The visibility attribute should be applied only to declarations
17855 which would otherwise have external linkage. The attribute should
17856 be applied consistently, so that the same entity should not be
17857 declared with different settings of the attribute.
17859 In C++, the visibility attribute applies to types as well as
17860 functions and objects, because in C++ types have linkage. A class
17861 must not have greater visibility than its non-static data member
17862 types and bases, and class members default to the visibility of
17863 their class. Also, a declaration without explicit visibility is
17864 limited to the visibility of its type.
17866 In C++, you can mark member functions and static member variables
17867 of a class with the visibility attribute. This is useful if if
17868 you know a particular method or static member variable should only
17869 be used from one shared object; then you can mark it hidden while
17870 the rest of the class has default visibility. Care must be taken
17871 to avoid breaking the One Definition Rule; for example, it is
17872 usually not useful to mark an inline method as hidden without
17873 marking the whole class as hidden.
17875 A C++ namespace declaration can also have the visibility attribute.
17876 This attribute applies only to the particular namespace body, not
17877 to other definitions of the same namespace; it is equivalent to
17878 using `#pragma GCC visibility' before and after the namespace
17879 definition (*note Visibility Pragmas::).
17881 In C++, if a template argument has limited visibility, this
17882 restriction is implicitly propagated to the template instantiation.
17883 Otherwise, template instantiations and specializations default to
17884 the visibility of their template.
17886 If both the template and enclosing class have explicit visibility,
17887 the visibility from the template is used.
17889 `warn_unused_result'
17890 The `warn_unused_result' attribute causes a warning to be emitted
17891 if a caller of the function with this attribute does not use its
17892 return value. This is useful for functions where not checking the
17893 result is either a security problem or always a bug, such as
17896 int fn () __attribute__ ((warn_unused_result));
17899 if (fn () < 0) return -1;
17904 results in warning on line 5.
17907 The `weak' attribute causes the declaration to be emitted as a weak
17908 symbol rather than a global. This is primarily useful in defining
17909 library functions which can be overridden in user code, though it
17910 can also be used with non-function declarations. Weak symbols are
17911 supported for ELF targets, and also for a.out targets when using
17912 the GNU assembler and linker.
17915 `weakref ("TARGET")'
17916 The `weakref' attribute marks a declaration as a weak reference.
17917 Without arguments, it should be accompanied by an `alias' attribute
17918 naming the target symbol. Optionally, the TARGET may be given as
17919 an argument to `weakref' itself. In either case, `weakref'
17920 implicitly marks the declaration as `weak'. Without a TARGET,
17921 given as an argument to `weakref' or to `alias', `weakref' is
17922 equivalent to `weak'.
17924 static int x() __attribute__ ((weakref ("y")));
17925 /* is equivalent to... */
17926 static int x() __attribute__ ((weak, weakref, alias ("y")));
17928 static int x() __attribute__ ((weakref));
17929 static int x() __attribute__ ((alias ("y")));
17931 A weak reference is an alias that does not by itself require a
17932 definition to be given for the target symbol. If the target
17933 symbol is only referenced through weak references, then the
17934 becomes a `weak' undefined symbol. If it is directly referenced,
17935 however, then such strong references prevail, and a definition
17936 will be required for the symbol, not necessarily in the same
17939 The effect is equivalent to moving all references to the alias to a
17940 separate translation unit, renaming the alias to the aliased
17941 symbol, declaring it as weak, compiling the two separate
17942 translation units and performing a reloadable link on them.
17944 At present, a declaration to which `weakref' is attached can only
17947 `externally_visible'
17948 This attribute, attached to a global variable or function nullify
17949 effect of `-fwhole-program' command line option, so the object
17950 remain visible outside the current compilation unit
17953 You can specify multiple attributes in a declaration by separating them
17954 by commas within the double parentheses or by immediately following an
17955 attribute declaration with another attribute declaration.
17957 Some people object to the `__attribute__' feature, suggesting that ISO
17958 C's `#pragma' should be used instead. At the time `__attribute__' was
17959 designed, there were two reasons for not doing this.
17961 1. It is impossible to generate `#pragma' commands from a macro.
17963 2. There is no telling what the same `#pragma' might mean in another
17966 These two reasons applied to almost any application that might have
17967 been proposed for `#pragma'. It was basically a mistake to use
17968 `#pragma' for _anything_.
17970 The ISO C99 standard includes `_Pragma', which now allows pragmas to
17971 be generated from macros. In addition, a `#pragma GCC' namespace is
17972 now in use for GCC-specific pragmas. However, it has been found
17973 convenient to use `__attribute__' to achieve a natural attachment of
17974 attributes to their corresponding declarations, whereas `#pragma GCC'
17975 is of use for constructs that do not naturally form part of the
17976 grammar. *Note Miscellaneous Preprocessing Directives: (cpp)Other
17980 File: gcc.info, Node: Attribute Syntax, Next: Function Prototypes, Prev: Function Attributes, Up: C Extensions
17982 5.28 Attribute Syntax
17983 =====================
17985 This section describes the syntax with which `__attribute__' may be
17986 used, and the constructs to which attribute specifiers bind, for the C
17987 language. Some details may vary for C++ and Objective-C. Because of
17988 infelicities in the grammar for attributes, some forms described here
17989 may not be successfully parsed in all cases.
17991 There are some problems with the semantics of attributes in C++. For
17992 example, there are no manglings for attributes, although they may affect
17993 code generation, so problems may arise when attributed types are used in
17994 conjunction with templates or overloading. Similarly, `typeid' does
17995 not distinguish between types with different attributes. Support for
17996 attributes in C++ may be restricted in future to attributes on
17997 declarations only, but not on nested declarators.
17999 *Note Function Attributes::, for details of the semantics of attributes
18000 applying to functions. *Note Variable Attributes::, for details of the
18001 semantics of attributes applying to variables. *Note Type Attributes::,
18002 for details of the semantics of attributes applying to structure, union
18003 and enumerated types.
18005 An "attribute specifier" is of the form `__attribute__
18006 ((ATTRIBUTE-LIST))'. An "attribute list" is a possibly empty
18007 comma-separated sequence of "attributes", where each attribute is one
18010 * Empty. Empty attributes are ignored.
18012 * A word (which may be an identifier such as `unused', or a reserved
18013 word such as `const').
18015 * A word, followed by, in parentheses, parameters for the attribute.
18016 These parameters take one of the following forms:
18018 * An identifier. For example, `mode' attributes use this form.
18020 * An identifier followed by a comma and a non-empty
18021 comma-separated list of expressions. For example, `format'
18022 attributes use this form.
18024 * A possibly empty comma-separated list of expressions. For
18025 example, `format_arg' attributes use this form with the list
18026 being a single integer constant expression, and `alias'
18027 attributes use this form with the list being a single string
18030 An "attribute specifier list" is a sequence of one or more attribute
18031 specifiers, not separated by any other tokens.
18033 In GNU C, an attribute specifier list may appear after the colon
18034 following a label, other than a `case' or `default' label. The only
18035 attribute it makes sense to use after a label is `unused'. This
18036 feature is intended for code generated by programs which contains labels
18037 that may be unused but which is compiled with `-Wall'. It would not
18038 normally be appropriate to use in it human-written code, though it
18039 could be useful in cases where the code that jumps to the label is
18040 contained within an `#ifdef' conditional. GNU C++ does not permit such
18041 placement of attribute lists, as it is permissible for a declaration,
18042 which could begin with an attribute list, to be labelled in C++.
18043 Declarations cannot be labelled in C90 or C99, so the ambiguity does
18046 An attribute specifier list may appear as part of a `struct', `union'
18047 or `enum' specifier. It may go either immediately after the `struct',
18048 `union' or `enum' keyword, or after the closing brace. The former
18049 syntax is preferred. Where attribute specifiers follow the closing
18050 brace, they are considered to relate to the structure, union or
18051 enumerated type defined, not to any enclosing declaration the type
18052 specifier appears in, and the type defined is not complete until after
18053 the attribute specifiers.
18055 Otherwise, an attribute specifier appears as part of a declaration,
18056 counting declarations of unnamed parameters and type names, and relates
18057 to that declaration (which may be nested in another declaration, for
18058 example in the case of a parameter declaration), or to a particular
18059 declarator within a declaration. Where an attribute specifier is
18060 applied to a parameter declared as a function or an array, it should
18061 apply to the function or array rather than the pointer to which the
18062 parameter is implicitly converted, but this is not yet correctly
18065 Any list of specifiers and qualifiers at the start of a declaration may
18066 contain attribute specifiers, whether or not such a list may in that
18067 context contain storage class specifiers. (Some attributes, however,
18068 are essentially in the nature of storage class specifiers, and only make
18069 sense where storage class specifiers may be used; for example,
18070 `section'.) There is one necessary limitation to this syntax: the
18071 first old-style parameter declaration in a function definition cannot
18072 begin with an attribute specifier, because such an attribute applies to
18073 the function instead by syntax described below (which, however, is not
18074 yet implemented in this case). In some other cases, attribute
18075 specifiers are permitted by this grammar but not yet supported by the
18076 compiler. All attribute specifiers in this place relate to the
18077 declaration as a whole. In the obsolescent usage where a type of `int'
18078 is implied by the absence of type specifiers, such a list of specifiers
18079 and qualifiers may be an attribute specifier list with no other
18080 specifiers or qualifiers.
18082 At present, the first parameter in a function prototype must have some
18083 type specifier which is not an attribute specifier; this resolves an
18084 ambiguity in the interpretation of `void f(int (__attribute__((foo))
18085 x))', but is subject to change. At present, if the parentheses of a
18086 function declarator contain only attributes then those attributes are
18087 ignored, rather than yielding an error or warning or implying a single
18088 parameter of type int, but this is subject to change.
18090 An attribute specifier list may appear immediately before a declarator
18091 (other than the first) in a comma-separated list of declarators in a
18092 declaration of more than one identifier using a single list of
18093 specifiers and qualifiers. Such attribute specifiers apply only to the
18094 identifier before whose declarator they appear. For example, in
18096 __attribute__((noreturn)) void d0 (void),
18097 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
18100 the `noreturn' attribute applies to all the functions declared; the
18101 `format' attribute only applies to `d1'.
18103 An attribute specifier list may appear immediately before the comma,
18104 `=' or semicolon terminating the declaration of an identifier other
18105 than a function definition. Such attribute specifiers apply to the
18106 declared object or function. Where an assembler name for an object or
18107 function is specified (*note Asm Labels::), the attribute must follow
18108 the `asm' specification.
18110 An attribute specifier list may, in future, be permitted to appear
18111 after the declarator in a function definition (before any old-style
18112 parameter declarations or the function body).
18114 Attribute specifiers may be mixed with type qualifiers appearing inside
18115 the `[]' of a parameter array declarator, in the C99 construct by which
18116 such qualifiers are applied to the pointer to which the array is
18117 implicitly converted. Such attribute specifiers apply to the pointer,
18118 not to the array, but at present this is not implemented and they are
18121 An attribute specifier list may appear at the start of a nested
18122 declarator. At present, there are some limitations in this usage: the
18123 attributes correctly apply to the declarator, but for most individual
18124 attributes the semantics this implies are not implemented. When
18125 attribute specifiers follow the `*' of a pointer declarator, they may
18126 be mixed with any type qualifiers present. The following describes the
18127 formal semantics of this syntax. It will make the most sense if you
18128 are familiar with the formal specification of declarators in the ISO C
18131 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration `T D1',
18132 where `T' contains declaration specifiers that specify a type TYPE
18133 (such as `int') and `D1' is a declarator that contains an identifier
18134 IDENT. The type specified for IDENT for derived declarators whose type
18135 does not include an attribute specifier is as in the ISO C standard.
18137 If `D1' has the form `( ATTRIBUTE-SPECIFIER-LIST D )', and the
18138 declaration `T D' specifies the type "DERIVED-DECLARATOR-TYPE-LIST
18139 TYPE" for IDENT, then `T D1' specifies the type
18140 "DERIVED-DECLARATOR-TYPE-LIST ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
18142 If `D1' has the form `* TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST
18143 D', and the declaration `T D' specifies the type
18144 "DERIVED-DECLARATOR-TYPE-LIST TYPE" for IDENT, then `T D1' specifies
18145 the type "DERIVED-DECLARATOR-TYPE-LIST
18146 TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
18150 void (__attribute__((noreturn)) ****f) (void);
18152 specifies the type "pointer to pointer to pointer to pointer to
18153 non-returning function returning `void'". As another example,
18155 char *__attribute__((aligned(8))) *f;
18157 specifies the type "pointer to 8-byte-aligned pointer to `char'". Note
18158 again that this does not work with most attributes; for example, the
18159 usage of `aligned' and `noreturn' attributes given above is not yet
18162 For compatibility with existing code written for compiler versions that
18163 did not implement attributes on nested declarators, some laxity is
18164 allowed in the placing of attributes. If an attribute that only applies
18165 to types is applied to a declaration, it will be treated as applying to
18166 the type of that declaration. If an attribute that only applies to
18167 declarations is applied to the type of a declaration, it will be treated
18168 as applying to that declaration; and, for compatibility with code
18169 placing the attributes immediately before the identifier declared, such
18170 an attribute applied to a function return type will be treated as
18171 applying to the function type, and such an attribute applied to an array
18172 element type will be treated as applying to the array type. If an
18173 attribute that only applies to function types is applied to a
18174 pointer-to-function type, it will be treated as applying to the pointer
18175 target type; if such an attribute is applied to a function return type
18176 that is not a pointer-to-function type, it will be treated as applying
18177 to the function type.
18180 File: gcc.info, Node: Function Prototypes, Next: C++ Comments, Prev: Attribute Syntax, Up: C Extensions
18182 5.29 Prototypes and Old-Style Function Definitions
18183 ==================================================
18185 GNU C extends ISO C to allow a function prototype to override a later
18186 old-style non-prototype definition. Consider the following example:
18188 /* Use prototypes unless the compiler is old-fashioned. */
18195 /* Prototype function declaration. */
18196 int isroot P((uid_t));
18198 /* Old-style function definition. */
18200 isroot (x) /* ??? lossage here ??? */
18206 Suppose the type `uid_t' happens to be `short'. ISO C does not allow
18207 this example, because subword arguments in old-style non-prototype
18208 definitions are promoted. Therefore in this example the function
18209 definition's argument is really an `int', which does not match the
18210 prototype argument type of `short'.
18212 This restriction of ISO C makes it hard to write code that is portable
18213 to traditional C compilers, because the programmer does not know
18214 whether the `uid_t' type is `short', `int', or `long'. Therefore, in
18215 cases like these GNU C allows a prototype to override a later old-style
18216 definition. More precisely, in GNU C, a function prototype argument
18217 type overrides the argument type specified by a later old-style
18218 definition if the former type is the same as the latter type before
18219 promotion. Thus in GNU C the above example is equivalent to the
18222 int isroot (uid_t);
18230 GNU C++ does not support old-style function definitions, so this
18231 extension is irrelevant.
18234 File: gcc.info, Node: C++ Comments, Next: Dollar Signs, Prev: Function Prototypes, Up: C Extensions
18236 5.30 C++ Style Comments
18237 =======================
18239 In GNU C, you may use C++ style comments, which start with `//' and
18240 continue until the end of the line. Many other C implementations allow
18241 such comments, and they are included in the 1999 C standard. However,
18242 C++ style comments are not recognized if you specify an `-std' option
18243 specifying a version of ISO C before C99, or `-ansi' (equivalent to
18247 File: gcc.info, Node: Dollar Signs, Next: Character Escapes, Prev: C++ Comments, Up: C Extensions
18249 5.31 Dollar Signs in Identifier Names
18250 =====================================
18252 In GNU C, you may normally use dollar signs in identifier names. This
18253 is because many traditional C implementations allow such identifiers.
18254 However, dollar signs in identifiers are not supported on a few target
18255 machines, typically because the target assembler does not allow them.
18258 File: gcc.info, Node: Character Escapes, Next: Variable Attributes, Prev: Dollar Signs, Up: C Extensions
18260 5.32 The Character <ESC> in Constants
18261 =====================================
18263 You can use the sequence `\e' in a string or character constant to
18264 stand for the ASCII character <ESC>.
18267 File: gcc.info, Node: Alignment, Next: Inline, Prev: Type Attributes, Up: C Extensions
18269 5.33 Inquiring on Alignment of Types or Variables
18270 =================================================
18272 The keyword `__alignof__' allows you to inquire about how an object is
18273 aligned, or the minimum alignment usually required by a type. Its
18274 syntax is just like `sizeof'.
18276 For example, if the target machine requires a `double' value to be
18277 aligned on an 8-byte boundary, then `__alignof__ (double)' is 8. This
18278 is true on many RISC machines. On more traditional machine designs,
18279 `__alignof__ (double)' is 4 or even 2.
18281 Some machines never actually require alignment; they allow reference
18282 to any data type even at an odd address. For these machines,
18283 `__alignof__' reports the smallest alignment that GCC will give the
18284 data type, usually as mandated by the target ABI.
18286 If the operand of `__alignof__' is an lvalue rather than a type, its
18287 value is the required alignment for its type, taking into account any
18288 minimum alignment specified with GCC's `__attribute__' extension (*note
18289 Variable Attributes::). For example, after this declaration:
18291 struct foo { int x; char y; } foo1;
18293 the value of `__alignof__ (foo1.y)' is 1, even though its actual
18294 alignment is probably 2 or 4, the same as `__alignof__ (int)'.
18296 It is an error to ask for the alignment of an incomplete type.
18299 File: gcc.info, Node: Variable Attributes, Next: Type Attributes, Prev: Character Escapes, Up: C Extensions
18301 5.34 Specifying Attributes of Variables
18302 =======================================
18304 The keyword `__attribute__' allows you to specify special attributes of
18305 variables or structure fields. This keyword is followed by an
18306 attribute specification inside double parentheses. Some attributes are
18307 currently defined generically for variables. Other attributes are
18308 defined for variables on particular target systems. Other attributes
18309 are available for functions (*note Function Attributes::) and for types
18310 (*note Type Attributes::). Other front ends might define more
18311 attributes (*note Extensions to the C++ Language: C++ Extensions.).
18313 You may also specify attributes with `__' preceding and following each
18314 keyword. This allows you to use them in header files without being
18315 concerned about a possible macro of the same name. For example, you
18316 may use `__aligned__' instead of `aligned'.
18318 *Note Attribute Syntax::, for details of the exact syntax for using
18321 `aligned (ALIGNMENT)'
18322 This attribute specifies a minimum alignment for the variable or
18323 structure field, measured in bytes. For example, the declaration:
18325 int x __attribute__ ((aligned (16))) = 0;
18327 causes the compiler to allocate the global variable `x' on a
18328 16-byte boundary. On a 68040, this could be used in conjunction
18329 with an `asm' expression to access the `move16' instruction which
18330 requires 16-byte aligned operands.
18332 You can also specify the alignment of structure fields. For
18333 example, to create a double-word aligned `int' pair, you could
18336 struct foo { int x[2] __attribute__ ((aligned (8))); };
18338 This is an alternative to creating a union with a `double' member
18339 that forces the union to be double-word aligned.
18341 As in the preceding examples, you can explicitly specify the
18342 alignment (in bytes) that you wish the compiler to use for a given
18343 variable or structure field. Alternatively, you can leave out the
18344 alignment factor and just ask the compiler to align a variable or
18345 field to the maximum useful alignment for the target machine you
18346 are compiling for. For example, you could write:
18348 short array[3] __attribute__ ((aligned));
18350 Whenever you leave out the alignment factor in an `aligned'
18351 attribute specification, the compiler automatically sets the
18352 alignment for the declared variable or field to the largest
18353 alignment which is ever used for any data type on the target
18354 machine you are compiling for. Doing this can often make copy
18355 operations more efficient, because the compiler can use whatever
18356 instructions copy the biggest chunks of memory when performing
18357 copies to or from the variables or fields that you have aligned
18360 When used on a struct, or struct member, the `aligned' attribute
18361 can only increase the alignment; in order to decrease it, the
18362 `packed' attribute must be specified as well. When used as part
18363 of a typedef, the `aligned' attribute can both increase and
18364 decrease alignment, and specifying the `packed' attribute will
18365 generate a warning.
18367 Note that the effectiveness of `aligned' attributes may be limited
18368 by inherent limitations in your linker. On many systems, the
18369 linker is only able to arrange for variables to be aligned up to a
18370 certain maximum alignment. (For some linkers, the maximum
18371 supported alignment may be very very small.) If your linker is
18372 only able to align variables up to a maximum of 8 byte alignment,
18373 then specifying `aligned(16)' in an `__attribute__' will still
18374 only provide you with 8 byte alignment. See your linker
18375 documentation for further information.
18377 The `aligned' attribute can also be used for functions (*note
18378 Function Attributes::.)
18380 `cleanup (CLEANUP_FUNCTION)'
18381 The `cleanup' attribute runs a function when the variable goes out
18382 of scope. This attribute can only be applied to auto function
18383 scope variables; it may not be applied to parameters or variables
18384 with static storage duration. The function must take one
18385 parameter, a pointer to a type compatible with the variable. The
18386 return value of the function (if any) is ignored.
18388 If `-fexceptions' is enabled, then CLEANUP_FUNCTION will be run
18389 during the stack unwinding that happens during the processing of
18390 the exception. Note that the `cleanup' attribute does not allow
18391 the exception to be caught, only to perform an action. It is
18392 undefined what happens if CLEANUP_FUNCTION does not return
18397 The `common' attribute requests GCC to place a variable in
18398 "common" storage. The `nocommon' attribute requests the
18399 opposite--to allocate space for it directly.
18401 These attributes override the default chosen by the `-fno-common'
18402 and `-fcommon' flags respectively.
18405 The `deprecated' attribute results in a warning if the variable is
18406 used anywhere in the source file. This is useful when identifying
18407 variables that are expected to be removed in a future version of a
18408 program. The warning also includes the location of the declaration
18409 of the deprecated variable, to enable users to easily find further
18410 information about why the variable is deprecated, or what they
18411 should do instead. Note that the warning only occurs for uses:
18413 extern int old_var __attribute__ ((deprecated));
18414 extern int old_var;
18415 int new_fn () { return old_var; }
18417 results in a warning on line 3 but not line 2.
18419 The `deprecated' attribute can also be used for functions and
18420 types (*note Function Attributes::, *note Type Attributes::.)
18423 This attribute specifies the data type for the
18424 declaration--whichever type corresponds to the mode MODE. This in
18425 effect lets you request an integer or floating point type
18426 according to its width.
18428 You may also specify a mode of `byte' or `__byte__' to indicate
18429 the mode corresponding to a one-byte integer, `word' or `__word__'
18430 for the mode of a one-word integer, and `pointer' or `__pointer__'
18431 for the mode used to represent pointers.
18434 The `packed' attribute specifies that a variable or structure field
18435 should have the smallest possible alignment--one byte for a
18436 variable, and one bit for a field, unless you specify a larger
18437 value with the `aligned' attribute.
18439 Here is a structure in which the field `x' is packed, so that it
18440 immediately follows `a':
18445 int x[2] __attribute__ ((packed));
18448 `section ("SECTION-NAME")'
18449 Normally, the compiler places the objects it generates in sections
18450 like `data' and `bss'. Sometimes, however, you need additional
18451 sections, or you need certain particular variables to appear in
18452 special sections, for example to map to special hardware. The
18453 `section' attribute specifies that a variable (or function) lives
18454 in a particular section. For example, this small program uses
18455 several specific section names:
18457 struct duart a __attribute__ ((section ("DUART_A"))) = { 0 };
18458 struct duart b __attribute__ ((section ("DUART_B"))) = { 0 };
18459 char stack[10000] __attribute__ ((section ("STACK"))) = { 0 };
18460 int init_data __attribute__ ((section ("INITDATA"))) = 0;
18464 /* Initialize stack pointer */
18465 init_sp (stack + sizeof (stack));
18467 /* Initialize initialized data */
18468 memcpy (&init_data, &data, &edata - &data);
18470 /* Turn on the serial ports */
18475 Use the `section' attribute with an _initialized_ definition of a
18476 _global_ variable, as shown in the example. GCC issues a warning
18477 and otherwise ignores the `section' attribute in uninitialized
18478 variable declarations.
18480 You may only use the `section' attribute with a fully initialized
18481 global definition because of the way linkers work. The linker
18482 requires each object be defined once, with the exception that
18483 uninitialized variables tentatively go in the `common' (or `bss')
18484 section and can be multiply "defined". You can force a variable
18485 to be initialized with the `-fno-common' flag or the `nocommon'
18488 Some file formats do not support arbitrary sections so the
18489 `section' attribute is not available on all platforms. If you
18490 need to map the entire contents of a module to a particular
18491 section, consider using the facilities of the linker instead.
18494 On Microsoft Windows, in addition to putting variable definitions
18495 in a named section, the section can also be shared among all
18496 running copies of an executable or DLL. For example, this small
18497 program defines shared data by putting it in a named section
18498 `shared' and marking the section shareable:
18500 int foo __attribute__((section ("shared"), shared)) = 0;
18505 /* Read and write foo. All running
18506 copies see the same value. */
18510 You may only use the `shared' attribute along with `section'
18511 attribute with a fully initialized global definition because of
18512 the way linkers work. See `section' attribute for more
18515 The `shared' attribute is only available on Microsoft Windows.
18517 `tls_model ("TLS_MODEL")'
18518 The `tls_model' attribute sets thread-local storage model (*note
18519 Thread-Local::) of a particular `__thread' variable, overriding
18520 `-ftls-model=' command line switch on a per-variable basis. The
18521 TLS_MODEL argument should be one of `global-dynamic',
18522 `local-dynamic', `initial-exec' or `local-exec'.
18524 Not all targets support this attribute.
18527 This attribute, attached to a variable, means that the variable is
18528 meant to be possibly unused. GCC will not produce a warning for
18532 This attribute, attached to a variable, means that the variable
18533 must be emitted even if it appears that the variable is not
18536 `vector_size (BYTES)'
18537 This attribute specifies the vector size for the variable,
18538 measured in bytes. For example, the declaration:
18540 int foo __attribute__ ((vector_size (16)));
18542 causes the compiler to set the mode for `foo', to be 16 bytes,
18543 divided into `int' sized units. Assuming a 32-bit int (a vector of
18544 4 units of 4 bytes), the corresponding mode of `foo' will be V4SI.
18546 This attribute is only applicable to integral and float scalars,
18547 although arrays, pointers, and function return values are allowed
18548 in conjunction with this construct.
18550 Aggregates with this attribute are invalid, even if they are of
18551 the same size as a corresponding scalar. For example, the
18554 struct S { int a; };
18555 struct S __attribute__ ((vector_size (16))) foo;
18557 is invalid even if the size of the structure is the same as the
18561 The `selectany' attribute causes an initialized global variable to
18562 have link-once semantics. When multiple definitions of the
18563 variable are encountered by the linker, the first is selected and
18564 the remainder are discarded. Following usage by the Microsoft
18565 compiler, the linker is told _not_ to warn about size or content
18566 differences of the multiple definitions.
18568 Although the primary usage of this attribute is for POD types, the
18569 attribute can also be applied to global C++ objects that are
18570 initialized by a constructor. In this case, the static
18571 initialization and destruction code for the object is emitted in
18572 each translation defining the object, but the calls to the
18573 constructor and destructor are protected by a link-once guard
18576 The `selectany' attribute is only available on Microsoft Windows
18577 targets. You can use `__declspec (selectany)' as a synonym for
18578 `__attribute__ ((selectany))' for compatibility with other
18582 The `weak' attribute is described in *Note Function Attributes::.
18585 The `dllimport' attribute is described in *Note Function
18589 The `dllexport' attribute is described in *Note Function
18593 5.34.1 Blackfin Variable Attributes
18594 -----------------------------------
18596 Three attributes are currently defined for the Blackfin.
18603 Use these attributes on the Blackfin to place the variable into L1
18604 Data SRAM. Variables with `l1_data' attribute will be put into
18605 the specific section named `.l1.data'. Those with `l1_data_A'
18606 attribute will be put into the specific section named
18607 `.l1.data.A'. Those with `l1_data_B' attribute will be put into
18608 the specific section named `.l1.data.B'.
18610 5.34.2 M32R/D Variable Attributes
18611 ---------------------------------
18613 One attribute is currently defined for the M32R/D.
18615 `model (MODEL-NAME)'
18616 Use this attribute on the M32R/D to set the addressability of an
18617 object. The identifier MODEL-NAME is one of `small', `medium', or
18618 `large', representing each of the code models.
18620 Small model objects live in the lower 16MB of memory (so that their
18621 addresses can be loaded with the `ld24' instruction).
18623 Medium and large model objects may live anywhere in the 32-bit
18624 address space (the compiler will generate `seth/add3' instructions
18625 to load their addresses).
18627 5.34.3 i386 Variable Attributes
18628 -------------------------------
18630 Two attributes are currently defined for i386 configurations:
18631 `ms_struct' and `gcc_struct'
18635 If `packed' is used on a structure, or if bit-fields are used it
18636 may be that the Microsoft ABI packs them differently than GCC
18637 would normally pack them. Particularly when moving packed data
18638 between functions compiled with GCC and the native Microsoft
18639 compiler (either via function call or as data in a file), it may
18640 be necessary to access either format.
18642 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
18643 Windows X86 compilers to match the native Microsoft compiler.
18645 The Microsoft structure layout algorithm is fairly simple with the
18646 exception of the bitfield packing:
18648 The padding and alignment of members of structures and whether a
18649 bit field can straddle a storage-unit boundary
18651 1. Structure members are stored sequentially in the order in
18652 which they are declared: the first member has the lowest
18653 memory address and the last member the highest.
18655 2. Every data object has an alignment-requirement. The
18656 alignment-requirement for all data except structures, unions,
18657 and arrays is either the size of the object or the current
18658 packing size (specified with either the aligned attribute or
18659 the pack pragma), whichever is less. For structures, unions,
18660 and arrays, the alignment-requirement is the largest
18661 alignment-requirement of its members. Every object is
18662 allocated an offset so that:
18664 offset % alignment-requirement == 0
18666 3. Adjacent bit fields are packed into the same 1-, 2-, or
18667 4-byte allocation unit if the integral types are the same
18668 size and if the next bit field fits into the current
18669 allocation unit without crossing the boundary imposed by the
18670 common alignment requirements of the bit fields.
18672 Handling of zero-length bitfields:
18674 MSVC interprets zero-length bitfields in the following ways:
18676 1. If a zero-length bitfield is inserted between two bitfields
18677 that would normally be coalesced, the bitfields will not be
18684 unsigned long bf_1 : 12;
18686 unsigned long bf_2 : 12;
18689 The size of `t1' would be 8 bytes with the zero-length
18690 bitfield. If the zero-length bitfield were removed, `t1''s
18691 size would be 4 bytes.
18693 2. If a zero-length bitfield is inserted after a bitfield,
18694 `foo', and the alignment of the zero-length bitfield is
18695 greater than the member that follows it, `bar', `bar' will be
18696 aligned as the type of the zero-length bitfield.
18714 For `t2', `bar' will be placed at offset 2, rather than
18715 offset 1. Accordingly, the size of `t2' will be 4. For
18716 `t3', the zero-length bitfield will not affect the alignment
18717 of `bar' or, as a result, the size of the structure.
18719 Taking this into account, it is important to note the
18722 1. If a zero-length bitfield follows a normal bitfield, the
18723 type of the zero-length bitfield may affect the
18724 alignment of the structure as whole. For example, `t2'
18725 has a size of 4 bytes, since the zero-length bitfield
18726 follows a normal bitfield, and is of type short.
18728 2. Even if a zero-length bitfield is not followed by a
18729 normal bitfield, it may still affect the alignment of
18738 Here, `t4' will take up 4 bytes.
18740 3. Zero-length bitfields following non-bitfield members are
18750 Here, `t5' will take up 2 bytes.
18752 5.34.4 PowerPC Variable Attributes
18753 ----------------------------------
18755 Three attributes currently are defined for PowerPC configurations:
18756 `altivec', `ms_struct' and `gcc_struct'.
18758 For full documentation of the struct attributes please see the
18759 documentation in the *Note i386 Variable Attributes::, section.
18761 For documentation of `altivec' attribute please see the documentation
18762 in the *Note PowerPC Type Attributes::, section.
18764 5.34.5 SPU Variable Attributes
18765 ------------------------------
18767 The SPU supports the `spu_vector' attribute for variables. For
18768 documentation of this attribute please see the documentation in the
18769 *Note SPU Type Attributes::, section.
18771 5.34.6 Xstormy16 Variable Attributes
18772 ------------------------------------
18774 One attribute is currently defined for xstormy16 configurations:
18778 If a variable has the `below100' attribute (`BELOW100' is allowed
18779 also), GCC will place the variable in the first 0x100 bytes of
18780 memory and use special opcodes to access it. Such variables will
18781 be placed in either the `.bss_below100' section or the
18782 `.data_below100' section.
18785 5.34.7 AVR Variable Attributes
18786 ------------------------------
18789 The `progmem' attribute is used on the AVR to place data in the
18790 Program Memory address space. The AVR is a Harvard Architecture
18791 processor and data normally resides in the Data Memory address
18795 File: gcc.info, Node: Type Attributes, Next: Alignment, Prev: Variable Attributes, Up: C Extensions
18797 5.35 Specifying Attributes of Types
18798 ===================================
18800 The keyword `__attribute__' allows you to specify special attributes of
18801 `struct' and `union' types when you define such types. This keyword is
18802 followed by an attribute specification inside double parentheses.
18803 Seven attributes are currently defined for types: `aligned', `packed',
18804 `transparent_union', `unused', `deprecated', `visibility', and
18805 `may_alias'. Other attributes are defined for functions (*note
18806 Function Attributes::) and for variables (*note Variable Attributes::).
18808 You may also specify any one of these attributes with `__' preceding
18809 and following its keyword. This allows you to use these attributes in
18810 header files without being concerned about a possible macro of the same
18811 name. For example, you may use `__aligned__' instead of `aligned'.
18813 You may specify type attributes in an enum, struct or union type
18814 declaration or definition, or for other types in a `typedef'
18817 For an enum, struct or union type, you may specify attributes either
18818 between the enum, struct or union tag and the name of the type, or just
18819 past the closing curly brace of the _definition_. The former syntax is
18822 *Note Attribute Syntax::, for details of the exact syntax for using
18825 `aligned (ALIGNMENT)'
18826 This attribute specifies a minimum alignment (in bytes) for
18827 variables of the specified type. For example, the declarations:
18829 struct S { short f[3]; } __attribute__ ((aligned (8)));
18830 typedef int more_aligned_int __attribute__ ((aligned (8)));
18832 force the compiler to insure (as far as it can) that each variable
18833 whose type is `struct S' or `more_aligned_int' will be allocated
18834 and aligned _at least_ on a 8-byte boundary. On a SPARC, having
18835 all variables of type `struct S' aligned to 8-byte boundaries
18836 allows the compiler to use the `ldd' and `std' (doubleword load and
18837 store) instructions when copying one variable of type `struct S' to
18838 another, thus improving run-time efficiency.
18840 Note that the alignment of any given `struct' or `union' type is
18841 required by the ISO C standard to be at least a perfect multiple of
18842 the lowest common multiple of the alignments of all of the members
18843 of the `struct' or `union' in question. This means that you _can_
18844 effectively adjust the alignment of a `struct' or `union' type by
18845 attaching an `aligned' attribute to any one of the members of such
18846 a type, but the notation illustrated in the example above is a
18847 more obvious, intuitive, and readable way to request the compiler
18848 to adjust the alignment of an entire `struct' or `union' type.
18850 As in the preceding example, you can explicitly specify the
18851 alignment (in bytes) that you wish the compiler to use for a given
18852 `struct' or `union' type. Alternatively, you can leave out the
18853 alignment factor and just ask the compiler to align a type to the
18854 maximum useful alignment for the target machine you are compiling
18855 for. For example, you could write:
18857 struct S { short f[3]; } __attribute__ ((aligned));
18859 Whenever you leave out the alignment factor in an `aligned'
18860 attribute specification, the compiler automatically sets the
18861 alignment for the type to the largest alignment which is ever used
18862 for any data type on the target machine you are compiling for.
18863 Doing this can often make copy operations more efficient, because
18864 the compiler can use whatever instructions copy the biggest chunks
18865 of memory when performing copies to or from the variables which
18866 have types that you have aligned this way.
18868 In the example above, if the size of each `short' is 2 bytes, then
18869 the size of the entire `struct S' type is 6 bytes. The smallest
18870 power of two which is greater than or equal to that is 8, so the
18871 compiler sets the alignment for the entire `struct S' type to 8
18874 Note that although you can ask the compiler to select a
18875 time-efficient alignment for a given type and then declare only
18876 individual stand-alone objects of that type, the compiler's
18877 ability to select a time-efficient alignment is primarily useful
18878 only when you plan to create arrays of variables having the
18879 relevant (efficiently aligned) type. If you declare or use arrays
18880 of variables of an efficiently-aligned type, then it is likely
18881 that your program will also be doing pointer arithmetic (or
18882 subscripting, which amounts to the same thing) on pointers to the
18883 relevant type, and the code that the compiler generates for these
18884 pointer arithmetic operations will often be more efficient for
18885 efficiently-aligned types than for other types.
18887 The `aligned' attribute can only increase the alignment; but you
18888 can decrease it by specifying `packed' as well. See below.
18890 Note that the effectiveness of `aligned' attributes may be limited
18891 by inherent limitations in your linker. On many systems, the
18892 linker is only able to arrange for variables to be aligned up to a
18893 certain maximum alignment. (For some linkers, the maximum
18894 supported alignment may be very very small.) If your linker is
18895 only able to align variables up to a maximum of 8 byte alignment,
18896 then specifying `aligned(16)' in an `__attribute__' will still
18897 only provide you with 8 byte alignment. See your linker
18898 documentation for further information.
18901 This attribute, attached to `struct' or `union' type definition,
18902 specifies that each member (other than zero-width bitfields) of
18903 the structure or union is placed to minimize the memory required.
18904 When attached to an `enum' definition, it indicates that the
18905 smallest integral type should be used.
18907 Specifying this attribute for `struct' and `union' types is
18908 equivalent to specifying the `packed' attribute on each of the
18909 structure or union members. Specifying the `-fshort-enums' flag
18910 on the line is equivalent to specifying the `packed' attribute on
18911 all `enum' definitions.
18913 In the following example `struct my_packed_struct''s members are
18914 packed closely together, but the internal layout of its `s' member
18915 is not packed--to do that, `struct my_unpacked_struct' would need
18918 struct my_unpacked_struct
18924 struct __attribute__ ((__packed__)) my_packed_struct
18928 struct my_unpacked_struct s;
18931 You may only specify this attribute on the definition of a `enum',
18932 `struct' or `union', not on a `typedef' which does not also define
18933 the enumerated type, structure or union.
18935 `transparent_union'
18936 This attribute, attached to a `union' type definition, indicates
18937 that any function parameter having that union type causes calls to
18938 that function to be treated in a special way.
18940 First, the argument corresponding to a transparent union type can
18941 be of any type in the union; no cast is required. Also, if the
18942 union contains a pointer type, the corresponding argument can be a
18943 null pointer constant or a void pointer expression; and if the
18944 union contains a void pointer type, the corresponding argument can
18945 be any pointer expression. If the union member type is a pointer,
18946 qualifiers like `const' on the referenced type must be respected,
18947 just as with normal pointer conversions.
18949 Second, the argument is passed to the function using the calling
18950 conventions of the first member of the transparent union, not the
18951 calling conventions of the union itself. All members of the union
18952 must have the same machine representation; this is necessary for
18953 this argument passing to work properly.
18955 Transparent unions are designed for library functions that have
18956 multiple interfaces for compatibility reasons. For example,
18957 suppose the `wait' function must accept either a value of type
18958 `int *' to comply with Posix, or a value of type `union wait *' to
18959 comply with the 4.1BSD interface. If `wait''s parameter were
18960 `void *', `wait' would accept both kinds of arguments, but it
18961 would also accept any other pointer type and this would make
18962 argument type checking less useful. Instead, `<sys/wait.h>' might
18963 define the interface as follows:
18965 typedef union __attribute__ ((__transparent_union__))
18969 } wait_status_ptr_t;
18971 pid_t wait (wait_status_ptr_t);
18973 This interface allows either `int *' or `union wait *' arguments
18974 to be passed, using the `int *' calling convention. The program
18975 can call `wait' with arguments of either type:
18977 int w1 () { int w; return wait (&w); }
18978 int w2 () { union wait w; return wait (&w); }
18980 With this interface, `wait''s implementation might look like this:
18982 pid_t wait (wait_status_ptr_t p)
18984 return waitpid (-1, p.__ip, 0);
18988 When attached to a type (including a `union' or a `struct'), this
18989 attribute means that variables of that type are meant to appear
18990 possibly unused. GCC will not produce a warning for any variables
18991 of that type, even if the variable appears to do nothing. This is
18992 often the case with lock or thread classes, which are usually
18993 defined and then not referenced, but contain constructors and
18994 destructors that have nontrivial bookkeeping functions.
18997 The `deprecated' attribute results in a warning if the type is
18998 used anywhere in the source file. This is useful when identifying
18999 types that are expected to be removed in a future version of a
19000 program. If possible, the warning also includes the location of
19001 the declaration of the deprecated type, to enable users to easily
19002 find further information about why the type is deprecated, or what
19003 they should do instead. Note that the warnings only occur for
19004 uses and then only if the type is being applied to an identifier
19005 that itself is not being declared as deprecated.
19007 typedef int T1 __attribute__ ((deprecated));
19011 typedef T1 T3 __attribute__ ((deprecated));
19012 T3 z __attribute__ ((deprecated));
19014 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
19015 warning is issued for line 4 because T2 is not explicitly
19016 deprecated. Line 5 has no warning because T3 is explicitly
19017 deprecated. Similarly for line 6.
19019 The `deprecated' attribute can also be used for functions and
19020 variables (*note Function Attributes::, *note Variable
19024 Accesses to objects with types with this attribute are not
19025 subjected to type-based alias analysis, but are instead assumed to
19026 be able to alias any other type of objects, just like the `char'
19027 type. See `-fstrict-aliasing' for more information on aliasing
19032 typedef short __attribute__((__may_alias__)) short_a;
19037 int a = 0x12345678;
19038 short_a *b = (short_a *) &a;
19042 if (a == 0x12345678)
19048 If you replaced `short_a' with `short' in the variable
19049 declaration, the above program would abort when compiled with
19050 `-fstrict-aliasing', which is on by default at `-O2' or above in
19051 recent GCC versions.
19054 In C++, attribute visibility (*note Function Attributes::) can
19055 also be applied to class, struct, union and enum types. Unlike
19056 other type attributes, the attribute must appear between the
19057 initial keyword and the name of the type; it cannot appear after
19058 the body of the type.
19060 Note that the type visibility is applied to vague linkage entities
19061 associated with the class (vtable, typeinfo node, etc.). In
19062 particular, if a class is thrown as an exception in one shared
19063 object and caught in another, the class must have default
19064 visibility. Otherwise the two shared objects will be unable to
19065 use the same typeinfo node and exception handling will break.
19067 5.35.1 ARM Type Attributes
19068 --------------------------
19070 On those ARM targets that support `dllimport' (such as Symbian
19071 OS), you can use the `notshared' attribute to indicate that the virtual
19072 table and other similar data for a class should not be exported from a
19075 class __declspec(notshared) C {
19077 __declspec(dllimport) C();
19081 __declspec(dllexport)
19084 In this code, `C::C' is exported from the current DLL, but the
19085 virtual table for `C' is not exported. (You can use `__attribute__'
19086 instead of `__declspec' if you prefer, but most Symbian OS code uses
19089 5.35.2 i386 Type Attributes
19090 ---------------------------
19092 Two attributes are currently defined for i386 configurations:
19093 `ms_struct' and `gcc_struct'
19097 If `packed' is used on a structure, or if bit-fields are used it
19098 may be that the Microsoft ABI packs them differently than GCC
19099 would normally pack them. Particularly when moving packed data
19100 between functions compiled with GCC and the native Microsoft
19101 compiler (either via function call or as data in a file), it may
19102 be necessary to access either format.
19104 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
19105 Windows X86 compilers to match the native Microsoft compiler.
19107 To specify multiple attributes, separate them by commas within the
19108 double parentheses: for example, `__attribute__ ((aligned (16),
19111 5.35.3 PowerPC Type Attributes
19112 ------------------------------
19114 Three attributes currently are defined for PowerPC configurations:
19115 `altivec', `ms_struct' and `gcc_struct'.
19117 For full documentation of the struct attributes please see the
19118 documentation in the *Note i386 Type Attributes::, section.
19120 The `altivec' attribute allows one to declare AltiVec vector data
19121 types supported by the AltiVec Programming Interface Manual. The
19122 attribute requires an argument to specify one of three vector types:
19123 `vector__', `pixel__' (always followed by unsigned short), and `bool__'
19124 (always followed by unsigned).
19126 __attribute__((altivec(vector__)))
19127 __attribute__((altivec(pixel__))) unsigned short
19128 __attribute__((altivec(bool__))) unsigned
19130 These attributes mainly are intended to support the `__vector',
19131 `__pixel', and `__bool' AltiVec keywords.
19133 5.35.4 SPU Type Attributes
19134 --------------------------
19136 The SPU supports the `spu_vector' attribute for types. This attribute
19137 allows one to declare vector data types supported by the
19138 Sony/Toshiba/IBM SPU Language Extensions Specification. It is intended
19139 to support the `__vector' keyword.
19142 File: gcc.info, Node: Inline, Next: Extended Asm, Prev: Alignment, Up: C Extensions
19144 5.36 An Inline Function is As Fast As a Macro
19145 =============================================
19147 By declaring a function inline, you can direct GCC to make calls to
19148 that function faster. One way GCC can achieve this is to integrate
19149 that function's code into the code for its callers. This makes
19150 execution faster by eliminating the function-call overhead; in
19151 addition, if any of the actual argument values are constant, their
19152 known values may permit simplifications at compile time so that not all
19153 of the inline function's code needs to be included. The effect on code
19154 size is less predictable; object code may be larger or smaller with
19155 function inlining, depending on the particular case. You can also
19156 direct GCC to try to integrate all "simple enough" functions into their
19157 callers with the option `-finline-functions'.
19159 GCC implements three different semantics of declaring a function
19160 inline. One is available with `-std=gnu89' or `-fgnu89-inline' or when
19161 `gnu_inline' attribute is present on all inline declarations, another
19162 when `-std=c99' or `-std=gnu99' (without `-fgnu89-inline'), and the
19163 third is used when compiling C++.
19165 To declare a function inline, use the `inline' keyword in its
19166 declaration, like this:
19174 If you are writing a header file to be included in ISO C89 programs,
19175 write `__inline__' instead of `inline'. *Note Alternate Keywords::.
19177 The three types of inlining behave similarly in two important cases:
19178 when the `inline' keyword is used on a `static' function, like the
19179 example above, and when a function is first declared without using the
19180 `inline' keyword and then is defined with `inline', like this:
19182 extern int inc (int *a);
19189 In both of these common cases, the program behaves the same as if you
19190 had not used the `inline' keyword, except for its speed.
19192 When a function is both inline and `static', if all calls to the
19193 function are integrated into the caller, and the function's address is
19194 never used, then the function's own assembler code is never referenced.
19195 In this case, GCC does not actually output assembler code for the
19196 function, unless you specify the option `-fkeep-inline-functions'.
19197 Some calls cannot be integrated for various reasons (in particular,
19198 calls that precede the function's definition cannot be integrated, and
19199 neither can recursive calls within the definition). If there is a
19200 nonintegrated call, then the function is compiled to assembler code as
19201 usual. The function must also be compiled as usual if the program
19202 refers to its address, because that can't be inlined.
19204 Note that certain usages in a function definition can make it
19205 unsuitable for inline substitution. Among these usages are: use of
19206 varargs, use of alloca, use of variable sized data types (*note
19207 Variable Length::), use of computed goto (*note Labels as Values::),
19208 use of nonlocal goto, and nested functions (*note Nested Functions::).
19209 Using `-Winline' will warn when a function marked `inline' could not be
19210 substituted, and will give the reason for the failure.
19212 As required by ISO C++, GCC considers member functions defined within
19213 the body of a class to be marked inline even if they are not explicitly
19214 declared with the `inline' keyword. You can override this with
19215 `-fno-default-inline'; *note Options Controlling C++ Dialect: C++
19218 GCC does not inline any functions when not optimizing unless you
19219 specify the `always_inline' attribute for the function, like this:
19222 inline void foo (const char) __attribute__((always_inline));
19224 The remainder of this section is specific to GNU C89 inlining.
19226 When an inline function is not `static', then the compiler must assume
19227 that there may be calls from other source files; since a global symbol
19228 can be defined only once in any program, the function must not be
19229 defined in the other source files, so the calls therein cannot be
19230 integrated. Therefore, a non-`static' inline function is always
19231 compiled on its own in the usual fashion.
19233 If you specify both `inline' and `extern' in the function definition,
19234 then the definition is used only for inlining. In no case is the
19235 function compiled on its own, not even if you refer to its address
19236 explicitly. Such an address becomes an external reference, as if you
19237 had only declared the function, and had not defined it.
19239 This combination of `inline' and `extern' has almost the effect of a
19240 macro. The way to use it is to put a function definition in a header
19241 file with these keywords, and put another copy of the definition
19242 (lacking `inline' and `extern') in a library file. The definition in
19243 the header file will cause most calls to the function to be inlined.
19244 If any uses of the function remain, they will refer to the single copy
19248 File: gcc.info, Node: Extended Asm, Next: Constraints, Prev: Inline, Up: C Extensions
19250 5.37 Assembler Instructions with C Expression Operands
19251 ======================================================
19253 In an assembler instruction using `asm', you can specify the operands
19254 of the instruction using C expressions. This means you need not guess
19255 which registers or memory locations will contain the data you want to
19258 You must specify an assembler instruction template much like what
19259 appears in a machine description, plus an operand constraint string for
19262 For example, here is how to use the 68881's `fsinx' instruction:
19264 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
19266 Here `angle' is the C expression for the input operand while `result'
19267 is that of the output operand. Each has `"f"' as its operand
19268 constraint, saying that a floating point register is required. The `='
19269 in `=f' indicates that the operand is an output; all output operands'
19270 constraints must use `='. The constraints use the same language used
19271 in the machine description (*note Constraints::).
19273 Each operand is described by an operand-constraint string followed by
19274 the C expression in parentheses. A colon separates the assembler
19275 template from the first output operand and another separates the last
19276 output operand from the first input, if any. Commas separate the
19277 operands within each group. The total number of operands is currently
19278 limited to 30; this limitation may be lifted in some future version of
19281 If there are no output operands but there are input operands, you must
19282 place two consecutive colons surrounding the place where the output
19285 As of GCC version 3.1, it is also possible to specify input and output
19286 operands using symbolic names which can be referenced within the
19287 assembler code. These names are specified inside square brackets
19288 preceding the constraint string, and can be referenced inside the
19289 assembler code using `%[NAME]' instead of a percentage sign followed by
19290 the operand number. Using named operands the above example could look
19293 asm ("fsinx %[angle],%[output]"
19294 : [output] "=f" (result)
19295 : [angle] "f" (angle));
19297 Note that the symbolic operand names have no relation whatsoever to
19298 other C identifiers. You may use any name you like, even those of
19299 existing C symbols, but you must ensure that no two operands within the
19300 same assembler construct use the same symbolic name.
19302 Output operand expressions must be lvalues; the compiler can check
19303 this. The input operands need not be lvalues. The compiler cannot
19304 check whether the operands have data types that are reasonable for the
19305 instruction being executed. It does not parse the assembler instruction
19306 template and does not know what it means or even whether it is valid
19307 assembler input. The extended `asm' feature is most often used for
19308 machine instructions the compiler itself does not know exist. If the
19309 output expression cannot be directly addressed (for example, it is a
19310 bit-field), your constraint must allow a register. In that case, GCC
19311 will use the register as the output of the `asm', and then store that
19312 register into the output.
19314 The ordinary output operands must be write-only; GCC will assume that
19315 the values in these operands before the instruction are dead and need
19316 not be generated. Extended asm supports input-output or read-write
19317 operands. Use the constraint character `+' to indicate such an operand
19318 and list it with the output operands. You should only use read-write
19319 operands when the constraints for the operand (or the operand in which
19320 only some of the bits are to be changed) allow a register.
19322 You may, as an alternative, logically split its function into two
19323 separate operands, one input operand and one write-only output operand.
19324 The connection between them is expressed by constraints which say they
19325 need to be in the same location when the instruction executes. You can
19326 use the same C expression for both operands, or different expressions.
19327 For example, here we write the (fictitious) `combine' instruction with
19328 `bar' as its read-only source operand and `foo' as its read-write
19331 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
19333 The constraint `"0"' for operand 1 says that it must occupy the same
19334 location as operand 0. A number in constraint is allowed only in an
19335 input operand and it must refer to an output operand.
19337 Only a number in the constraint can guarantee that one operand will be
19338 in the same place as another. The mere fact that `foo' is the value of
19339 both operands is not enough to guarantee that they will be in the same
19340 place in the generated assembler code. The following would not work
19343 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
19345 Various optimizations or reloading could cause operands 0 and 1 to be
19346 in different registers; GCC knows no reason not to do so. For example,
19347 the compiler might find a copy of the value of `foo' in one register and
19348 use it for operand 1, but generate the output operand 0 in a different
19349 register (copying it afterward to `foo''s own address). Of course,
19350 since the register for operand 1 is not even mentioned in the assembler
19351 code, the result will not work, but GCC can't tell that.
19353 As of GCC version 3.1, one may write `[NAME]' instead of the operand
19354 number for a matching constraint. For example:
19356 asm ("cmoveq %1,%2,%[result]"
19357 : [result] "=r"(result)
19358 : "r" (test), "r"(new), "[result]"(old));
19360 Sometimes you need to make an `asm' operand be a specific register,
19361 but there's no matching constraint letter for that register _by
19362 itself_. To force the operand into that register, use a local variable
19363 for the operand and specify the register in the variable declaration.
19364 *Note Explicit Reg Vars::. Then for the `asm' operand, use any
19365 register constraint letter that matches the register:
19367 register int *p1 asm ("r0") = ...;
19368 register int *p2 asm ("r1") = ...;
19369 register int *result asm ("r0");
19370 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
19372 In the above example, beware that a register that is call-clobbered by
19373 the target ABI will be overwritten by any function call in the
19374 assignment, including library calls for arithmetic operators. Assuming
19375 it is a call-clobbered register, this may happen to `r0' above by the
19376 assignment to `p2'. If you have to use such a register, use temporary
19377 variables for expressions between the register assignment and use:
19380 register int *p1 asm ("r0") = ...;
19381 register int *p2 asm ("r1") = t1;
19382 register int *result asm ("r0");
19383 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
19385 Some instructions clobber specific hard registers. To describe this,
19386 write a third colon after the input operands, followed by the names of
19387 the clobbered hard registers (given as strings). Here is a realistic
19388 example for the VAX:
19390 asm volatile ("movc3 %0,%1,%2"
19392 : "g" (from), "g" (to), "g" (count)
19393 : "r0", "r1", "r2", "r3", "r4", "r5");
19395 You may not write a clobber description in a way that overlaps with an
19396 input or output operand. For example, you may not have an operand
19397 describing a register class with one member if you mention that register
19398 in the clobber list. Variables declared to live in specific registers
19399 (*note Explicit Reg Vars::), and used as asm input or output operands
19400 must have no part mentioned in the clobber description. There is no
19401 way for you to specify that an input operand is modified without also
19402 specifying it as an output operand. Note that if all the output
19403 operands you specify are for this purpose (and hence unused), you will
19404 then also need to specify `volatile' for the `asm' construct, as
19405 described below, to prevent GCC from deleting the `asm' statement as
19408 If you refer to a particular hardware register from the assembler code,
19409 you will probably have to list the register after the third colon to
19410 tell the compiler the register's value is modified. In some assemblers,
19411 the register names begin with `%'; to produce one `%' in the assembler
19412 code, you must write `%%' in the input.
19414 If your assembler instruction can alter the condition code register,
19415 add `cc' to the list of clobbered registers. GCC on some machines
19416 represents the condition codes as a specific hardware register; `cc'
19417 serves to name this register. On other machines, the condition code is
19418 handled differently, and specifying `cc' has no effect. But it is
19419 valid no matter what the machine.
19421 If your assembler instructions access memory in an unpredictable
19422 fashion, add `memory' to the list of clobbered registers. This will
19423 cause GCC to not keep memory values cached in registers across the
19424 assembler instruction and not optimize stores or loads to that memory.
19425 You will also want to add the `volatile' keyword if the memory affected
19426 is not listed in the inputs or outputs of the `asm', as the `memory'
19427 clobber does not count as a side-effect of the `asm'. If you know how
19428 large the accessed memory is, you can add it as input or output but if
19429 this is not known, you should add `memory'. As an example, if you
19430 access ten bytes of a string, you can use a memory input like:
19432 {"m"( ({ struct { char x[10]; } *p = (void *)ptr ; *p; }) )}.
19434 Note that in the following example the memory input is necessary,
19435 otherwise GCC might optimize the store to `x' away:
19441 asm ("magic stuff accessing an 'int' pointed to by '%1'"
19442 "=&d" (r) : "a" (y), "m" (*y));
19446 You can put multiple assembler instructions together in a single `asm'
19447 template, separated by the characters normally used in assembly code
19448 for the system. A combination that works in most places is a newline
19449 to break the line, plus a tab character to move to the instruction field
19450 (written as `\n\t'). Sometimes semicolons can be used, if the
19451 assembler allows semicolons as a line-breaking character. Note that
19452 some assembler dialects use semicolons to start a comment. The input
19453 operands are guaranteed not to use any of the clobbered registers, and
19454 neither will the output operands' addresses, so you can read and write
19455 the clobbered registers as many times as you like. Here is an example
19456 of multiple instructions in a template; it assumes the subroutine
19457 `_foo' accepts arguments in registers 9 and 10:
19459 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
19461 : "g" (from), "g" (to)
19464 Unless an output operand has the `&' constraint modifier, GCC may
19465 allocate it in the same register as an unrelated input operand, on the
19466 assumption the inputs are consumed before the outputs are produced.
19467 This assumption may be false if the assembler code actually consists of
19468 more than one instruction. In such a case, use `&' for each output
19469 operand that may not overlap an input. *Note Modifiers::.
19471 If you want to test the condition code produced by an assembler
19472 instruction, you must include a branch and a label in the `asm'
19473 construct, as follows:
19475 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
19479 This assumes your assembler supports local labels, as the GNU assembler
19480 and most Unix assemblers do.
19482 Speaking of labels, jumps from one `asm' to another are not supported.
19483 The compiler's optimizers do not know about these jumps, and therefore
19484 they cannot take account of them when deciding how to optimize.
19486 Usually the most convenient way to use these `asm' instructions is to
19487 encapsulate them in macros that look like functions. For example,
19490 ({ double __value, __arg = (x); \
19491 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
19494 Here the variable `__arg' is used to make sure that the instruction
19495 operates on a proper `double' value, and to accept only those arguments
19496 `x' which can convert automatically to a `double'.
19498 Another way to make sure the instruction operates on the correct data
19499 type is to use a cast in the `asm'. This is different from using a
19500 variable `__arg' in that it converts more different types. For
19501 example, if the desired type were `int', casting the argument to `int'
19502 would accept a pointer with no complaint, while assigning the argument
19503 to an `int' variable named `__arg' would warn about using a pointer
19504 unless the caller explicitly casts it.
19506 If an `asm' has output operands, GCC assumes for optimization purposes
19507 the instruction has no side effects except to change the output
19508 operands. This does not mean instructions with a side effect cannot be
19509 used, but you must be careful, because the compiler may eliminate them
19510 if the output operands aren't used, or move them out of loops, or
19511 replace two with one if they constitute a common subexpression. Also,
19512 if your instruction does have a side effect on a variable that otherwise
19513 appears not to change, the old value of the variable may be reused later
19514 if it happens to be found in a register.
19516 You can prevent an `asm' instruction from being deleted by writing the
19517 keyword `volatile' after the `asm'. For example:
19519 #define get_and_set_priority(new) \
19521 asm volatile ("get_and_set_priority %0, %1" \
19522 : "=g" (__old) : "g" (new)); \
19525 The `volatile' keyword indicates that the instruction has important
19526 side-effects. GCC will not delete a volatile `asm' if it is reachable.
19527 (The instruction can still be deleted if GCC can prove that
19528 control-flow will never reach the location of the instruction.) Note
19529 that even a volatile `asm' instruction can be moved relative to other
19530 code, including across jump instructions. For example, on many targets
19531 there is a system register which can be set to control the rounding
19532 mode of floating point operations. You might try setting it with a
19533 volatile `asm', like this PowerPC example:
19535 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
19538 This will not work reliably, as the compiler may move the addition back
19539 before the volatile `asm'. To make it work you need to add an
19540 artificial dependency to the `asm' referencing a variable in the code
19541 you don't want moved, for example:
19543 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
19546 Similarly, you can't expect a sequence of volatile `asm' instructions
19547 to remain perfectly consecutive. If you want consecutive output, use a
19548 single `asm'. Also, GCC will perform some optimizations across a
19549 volatile `asm' instruction; GCC does not "forget everything" when it
19550 encounters a volatile `asm' instruction the way some other compilers do.
19552 An `asm' instruction without any output operands will be treated
19553 identically to a volatile `asm' instruction.
19555 It is a natural idea to look for a way to give access to the condition
19556 code left by the assembler instruction. However, when we attempted to
19557 implement this, we found no way to make it work reliably. The problem
19558 is that output operands might need reloading, which would result in
19559 additional following "store" instructions. On most machines, these
19560 instructions would alter the condition code before there was time to
19561 test it. This problem doesn't arise for ordinary "test" and "compare"
19562 instructions because they don't have any output operands.
19564 For reasons similar to those described above, it is not possible to
19565 give an assembler instruction access to the condition code left by
19566 previous instructions.
19568 If you are writing a header file that should be includable in ISO C
19569 programs, write `__asm__' instead of `asm'. *Note Alternate Keywords::.
19571 5.37.1 Size of an `asm'
19572 -----------------------
19574 Some targets require that GCC track the size of each instruction used in
19575 order to generate correct code. Because the final length of an `asm'
19576 is only known by the assembler, GCC must make an estimate as to how big
19577 it will be. The estimate is formed by counting the number of
19578 statements in the pattern of the `asm' and multiplying that by the
19579 length of the longest instruction on that processor. Statements in the
19580 `asm' are identified by newline characters and whatever statement
19581 separator characters are supported by the assembler; on most processors
19582 this is the ``;'' character.
19584 Normally, GCC's estimate is perfectly adequate to ensure that correct
19585 code is generated, but it is possible to confuse the compiler if you use
19586 pseudo instructions or assembler macros that expand into multiple real
19587 instructions or if you use assembler directives that expand to more
19588 space in the object file than would be needed for a single instruction.
19589 If this happens then the assembler will produce a diagnostic saying that
19590 a label is unreachable.
19592 5.37.2 i386 floating point asm operands
19593 ---------------------------------------
19595 There are several rules on the usage of stack-like regs in asm_operands
19596 insns. These rules apply only to the operands that are stack-like regs:
19598 1. Given a set of input regs that die in an asm_operands, it is
19599 necessary to know which are implicitly popped by the asm, and
19600 which must be explicitly popped by gcc.
19602 An input reg that is implicitly popped by the asm must be
19603 explicitly clobbered, unless it is constrained to match an output
19606 2. For any input reg that is implicitly popped by an asm, it is
19607 necessary to know how to adjust the stack to compensate for the
19608 pop. If any non-popped input is closer to the top of the
19609 reg-stack than the implicitly popped reg, it would not be possible
19610 to know what the stack looked like--it's not clear how the rest of
19611 the stack "slides up".
19613 All implicitly popped input regs must be closer to the top of the
19614 reg-stack than any input that is not implicitly popped.
19616 It is possible that if an input dies in an insn, reload might use
19617 the input reg for an output reload. Consider this example:
19619 asm ("foo" : "=t" (a) : "f" (b));
19621 This asm says that input B is not popped by the asm, and that the
19622 asm pushes a result onto the reg-stack, i.e., the stack is one
19623 deeper after the asm than it was before. But, it is possible that
19624 reload will think that it can use the same reg for both the input
19625 and the output, if input B dies in this insn.
19627 If any input operand uses the `f' constraint, all output reg
19628 constraints must use the `&' earlyclobber.
19630 The asm above would be written as
19632 asm ("foo" : "=&t" (a) : "f" (b));
19634 3. Some operands need to be in particular places on the stack. All
19635 output operands fall in this category--there is no other way to
19636 know which regs the outputs appear in unless the user indicates
19637 this in the constraints.
19639 Output operands must specifically indicate which reg an output
19640 appears in after an asm. `=f' is not allowed: the operand
19641 constraints must select a class with a single reg.
19643 4. Output operands may not be "inserted" between existing stack regs.
19644 Since no 387 opcode uses a read/write operand, all output operands
19645 are dead before the asm_operands, and are pushed by the
19646 asm_operands. It makes no sense to push anywhere but the top of
19649 Output operands must start at the top of the reg-stack: output
19650 operands may not "skip" a reg.
19652 5. Some asm statements may need extra stack space for internal
19653 calculations. This can be guaranteed by clobbering stack registers
19654 unrelated to the inputs and outputs.
19657 Here are a couple of reasonable asms to want to write. This asm takes
19658 one input, which is internally popped, and produces two outputs.
19660 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
19662 This asm takes two inputs, which are popped by the `fyl2xp1' opcode,
19663 and replaces them with one output. The user must code the `st(1)'
19664 clobber for reg-stack.c to know that `fyl2xp1' pops both inputs.
19666 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
19669 File: gcc.info, Node: Constraints, Next: Asm Labels, Prev: Extended Asm, Up: C Extensions
19671 5.38 Constraints for `asm' Operands
19672 ===================================
19674 Here are specific details on what constraint letters you can use with
19675 `asm' operands. Constraints can say whether an operand may be in a
19676 register, and which kinds of register; whether the operand can be a
19677 memory reference, and which kinds of address; whether the operand may
19678 be an immediate constant, and which possible values it may have.
19679 Constraints can also require two operands to match.
19683 * Simple Constraints:: Basic use of constraints.
19684 * Multi-Alternative:: When an insn has two alternative constraint-patterns.
19685 * Modifiers:: More precise control over effects of constraints.
19686 * Machine Constraints:: Special constraints for some particular machines.
19689 File: gcc.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints
19691 5.38.1 Simple Constraints
19692 -------------------------
19694 The simplest kind of constraint is a string full of letters, each of
19695 which describes one kind of operand that is permitted. Here are the
19696 letters that are allowed:
19699 Whitespace characters are ignored and can be inserted at any
19700 position except the first. This enables each alternative for
19701 different operands to be visually aligned in the machine
19702 description even if they have different number of constraints and
19706 A memory operand is allowed, with any kind of address that the
19707 machine supports in general.
19710 A memory operand is allowed, but only if the address is
19711 "offsettable". This means that adding a small integer (actually,
19712 the width in bytes of the operand, as determined by its machine
19713 mode) may be added to the address and the result is also a valid
19716 For example, an address which is constant is offsettable; so is an
19717 address that is the sum of a register and a constant (as long as a
19718 slightly larger constant is also within the range of
19719 address-offsets supported by the machine); but an autoincrement or
19720 autodecrement address is not offsettable. More complicated
19721 indirect/indexed addresses may or may not be offsettable depending
19722 on the other addressing modes that the machine supports.
19724 Note that in an output operand which can be matched by another
19725 operand, the constraint letter `o' is valid only when accompanied
19726 by both `<' (if the target machine has predecrement addressing)
19727 and `>' (if the target machine has preincrement addressing).
19730 A memory operand that is not offsettable. In other words,
19731 anything that would fit the `m' constraint but not the `o'
19735 A memory operand with autodecrement addressing (either
19736 predecrement or postdecrement) is allowed.
19739 A memory operand with autoincrement addressing (either
19740 preincrement or postincrement) is allowed.
19743 A register operand is allowed provided that it is in a general
19747 An immediate integer operand (one with constant value) is allowed.
19748 This includes symbolic constants whose values will be known only at
19749 assembly time or later.
19752 An immediate integer operand with a known numeric value is allowed.
19753 Many systems cannot support assembly-time constants for operands
19754 less than a word wide. Constraints for these operands should use
19755 `n' rather than `i'.
19757 `I', `J', `K', ... `P'
19758 Other letters in the range `I' through `P' may be defined in a
19759 machine-dependent fashion to permit immediate integer operands with
19760 explicit integer values in specified ranges. For example, on the
19761 68000, `I' is defined to stand for the range of values 1 to 8.
19762 This is the range permitted as a shift count in the shift
19766 An immediate floating operand (expression code `const_double') is
19767 allowed, but only if the target floating point format is the same
19768 as that of the host machine (on which the compiler is running).
19771 An immediate floating operand (expression code `const_double' or
19772 `const_vector') is allowed.
19775 `G' and `H' may be defined in a machine-dependent fashion to
19776 permit immediate floating operands in particular ranges of values.
19779 An immediate integer operand whose value is not an explicit
19780 integer is allowed.
19782 This might appear strange; if an insn allows a constant operand
19783 with a value not known at compile time, it certainly must allow
19784 any known value. So why use `s' instead of `i'? Sometimes it
19785 allows better code to be generated.
19787 For example, on the 68000 in a fullword instruction it is possible
19788 to use an immediate operand; but if the immediate value is between
19789 -128 and 127, better code results from loading the value into a
19790 register and using the register. This is because the load into
19791 the register can be done with a `moveq' instruction. We arrange
19792 for this to happen by defining the letter `K' to mean "any integer
19793 outside the range -128 to 127", and then specifying `Ks' in the
19794 operand constraints.
19797 Any register, memory or immediate integer operand is allowed,
19798 except for registers that are not general registers.
19801 Any operand whatsoever is allowed.
19803 `0', `1', `2', ... `9'
19804 An operand that matches the specified operand number is allowed.
19805 If a digit is used together with letters within the same
19806 alternative, the digit should come last.
19808 This number is allowed to be more than a single digit. If multiple
19809 digits are encountered consecutively, they are interpreted as a
19810 single decimal integer. There is scant chance for ambiguity,
19811 since to-date it has never been desirable that `10' be interpreted
19812 as matching either operand 1 _or_ operand 0. Should this be
19813 desired, one can use multiple alternatives instead.
19815 This is called a "matching constraint" and what it really means is
19816 that the assembler has only a single operand that fills two roles
19817 which `asm' distinguishes. For example, an add instruction uses
19818 two input operands and an output operand, but on most CISC
19819 machines an add instruction really has only two operands, one of
19820 them an input-output operand:
19824 Matching constraints are used in these circumstances. More
19825 precisely, the two operands that match must include one input-only
19826 operand and one output-only operand. Moreover, the digit must be a
19827 smaller number than the number of the operand that uses it in the
19831 An operand that is a valid memory address is allowed. This is for
19832 "load address" and "push address" instructions.
19834 `p' in the constraint must be accompanied by `address_operand' as
19835 the predicate in the `match_operand'. This predicate interprets
19836 the mode specified in the `match_operand' as the mode of the memory
19837 reference for which the address would be valid.
19840 Other letters can be defined in machine-dependent fashion to stand
19841 for particular classes of registers or other arbitrary operand
19842 types. `d', `a' and `f' are defined on the 68000/68020 to stand
19843 for data, address and floating point registers.
19846 File: gcc.info, Node: Multi-Alternative, Next: Modifiers, Prev: Simple Constraints, Up: Constraints
19848 5.38.2 Multiple Alternative Constraints
19849 ---------------------------------------
19851 Sometimes a single instruction has multiple alternative sets of possible
19852 operands. For example, on the 68000, a logical-or instruction can
19853 combine register or an immediate value into memory, or it can combine
19854 any kind of operand into a register; but it cannot combine one memory
19855 location into another.
19857 These constraints are represented as multiple alternatives. An
19858 alternative can be described by a series of letters for each operand.
19859 The overall constraint for an operand is made from the letters for this
19860 operand from the first alternative, a comma, the letters for this
19861 operand from the second alternative, a comma, and so on until the last
19864 If all the operands fit any one alternative, the instruction is valid.
19865 Otherwise, for each alternative, the compiler counts how many
19866 instructions must be added to copy the operands so that that
19867 alternative applies. The alternative requiring the least copying is
19868 chosen. If two alternatives need the same amount of copying, the one
19869 that comes first is chosen. These choices can be altered with the `?'
19870 and `!' characters:
19873 Disparage slightly the alternative that the `?' appears in, as a
19874 choice when no alternative applies exactly. The compiler regards
19875 this alternative as one unit more costly for each `?' that appears
19879 Disparage severely the alternative that the `!' appears in. This
19880 alternative can still be used if it fits without reloading, but if
19881 reloading is needed, some other alternative will be used.
19884 File: gcc.info, Node: Modifiers, Next: Machine Constraints, Prev: Multi-Alternative, Up: Constraints
19886 5.38.3 Constraint Modifier Characters
19887 -------------------------------------
19889 Here are constraint modifier characters.
19892 Means that this operand is write-only for this instruction: the
19893 previous value is discarded and replaced by output data.
19896 Means that this operand is both read and written by the
19899 When the compiler fixes up the operands to satisfy the constraints,
19900 it needs to know which operands are inputs to the instruction and
19901 which are outputs from it. `=' identifies an output; `+'
19902 identifies an operand that is both input and output; all other
19903 operands are assumed to be input only.
19905 If you specify `=' or `+' in a constraint, you put it in the first
19906 character of the constraint string.
19909 Means (in a particular alternative) that this operand is an
19910 "earlyclobber" operand, which is modified before the instruction is
19911 finished using the input operands. Therefore, this operand may
19912 not lie in a register that is used as an input operand or as part
19913 of any memory address.
19915 `&' applies only to the alternative in which it is written. In
19916 constraints with multiple alternatives, sometimes one alternative
19917 requires `&' while others do not. See, for example, the `movdf'
19920 An input operand can be tied to an earlyclobber operand if its only
19921 use as an input occurs before the early result is written. Adding
19922 alternatives of this form often allows GCC to produce better code
19923 when only some of the inputs can be affected by the earlyclobber.
19924 See, for example, the `mulsi3' insn of the ARM.
19926 `&' does not obviate the need to write `='.
19929 Declares the instruction to be commutative for this operand and the
19930 following operand. This means that the compiler may interchange
19931 the two operands if that is the cheapest way to make all operands
19932 fit the constraints. GCC can only handle one commutative pair in
19933 an asm; if you use more, the compiler may fail. Note that you
19934 need not use the modifier if the two alternatives are strictly
19935 identical; this would only waste time in the reload pass. The
19936 modifier is not operational after register allocation, so the
19937 result of `define_peephole2' and `define_split's performed after
19938 reload cannot rely on `%' to make the intended insn match.
19941 Says that all following characters, up to the next comma, are to be
19942 ignored as a constraint. They are significant only for choosing
19943 register preferences.
19946 Says that the following character should be ignored when choosing
19947 register preferences. `*' has no effect on the meaning of the
19948 constraint as a constraint, and no effect on reloading.
19952 File: gcc.info, Node: Machine Constraints, Prev: Modifiers, Up: Constraints
19954 5.38.4 Constraints for Particular Machines
19955 ------------------------------------------
19957 Whenever possible, you should use the general-purpose constraint letters
19958 in `asm' arguments, since they will convey meaning more readily to
19959 people reading your code. Failing that, use the constraint letters
19960 that usually have very similar meanings across architectures. The most
19961 commonly used constraints are `m' and `r' (for memory and
19962 general-purpose registers respectively; *note Simple Constraints::), and
19963 `I', usually the letter indicating the most common immediate-constant
19966 Each architecture defines additional constraints. These constraints
19967 are used by the compiler itself for instruction generation, as well as
19968 for `asm' statements; therefore, some of the constraints are not
19969 particularly useful for `asm'. Here is a summary of some of the
19970 machine-dependent constraints available on some particular machines; it
19971 includes both constraints that are useful for `asm' and constraints
19972 that aren't. The compiler source file mentioned in the table heading
19973 for each architecture is the definitive reference for the meanings of
19974 that architecture's constraints.
19976 _ARM family--`config/arm/arm.h'_
19979 Floating-point register
19982 VFP floating-point register
19985 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0,
19989 Floating-point constant that would satisfy the constraint `F'
19993 Integer that is valid as an immediate operand in a data
19994 processing instruction. That is, an integer in the range 0
19995 to 255 rotated by a multiple of 2
19998 Integer in the range -4095 to 4095
20001 Integer that satisfies constraint `I' when inverted (ones
20005 Integer that satisfies constraint `I' when negated (twos
20009 Integer in the range 0 to 32
20012 A memory reference where the exact address is in a single
20013 register (``m'' is preferable for `asm' statements)
20016 An item in the constant pool
20019 A symbol in the text segment of the current file
20022 A memory reference suitable for VFP load/store insns
20023 (reg+constant offset)
20026 A memory reference suitable for iWMMXt load/store
20030 A memory reference suitable for the ARMv4 ldrsb instruction.
20032 _AVR family--`config/avr/constraints.md'_
20035 Registers from r0 to r15
20038 Registers from r16 to r23
20041 Registers from r16 to r31
20044 Registers from r24 to r31. These registers can be used in
20048 Pointer register (r26-r31)
20051 Base pointer register (r28-r31)
20054 Stack pointer register (SPH:SPL)
20057 Temporary register r0
20060 Register pair X (r27:r26)
20063 Register pair Y (r29:r28)
20066 Register pair Z (r31:r30)
20069 Constant greater than -1, less than 64
20072 Constant greater than -64, less than 1
20081 Constant that fits in 8 bits
20084 Constant integer -1
20087 Constant integer 8, 16, or 24
20093 A floating point constant 0.0
20096 Integer constant in the range -6 ... 5.
20099 A memory address based on Y or Z pointer with displacement.
20101 _CRX Architecture--`config/crx/crx.h'_
20104 Registers from r0 to r14 (registers without stack pointer)
20107 Register r16 (64-bit accumulator lo register)
20110 Register r17 (64-bit accumulator hi register)
20113 Register pair r16-r17. (64-bit accumulator lo-hi pair)
20116 Constant that fits in 3 bits
20119 Constant that fits in 4 bits
20122 Constant that fits in 5 bits
20125 Constant that is one of -1, 4, -4, 7, 8, 12, 16, 20, 32, 48
20128 Floating point constant that is legal for store immediate
20130 _Hewlett-Packard PA-RISC--`config/pa/pa.h'_
20136 Floating point register
20139 Shift amount register
20142 Floating point register (deprecated)
20145 Upper floating point register (32-bit), floating point
20152 Signed 11-bit integer constant
20155 Signed 14-bit integer constant
20158 Integer constant that can be deposited with a `zdepi'
20162 Signed 5-bit integer constant
20168 Integer constant that can be loaded with a `ldil' instruction
20171 Integer constant whose value plus one is a power of 2
20174 Integer constant that can be used for `and' operations in
20175 `depi' and `extru' instructions
20178 Integer constant 31
20181 Integer constant 63
20184 Floating-point constant 0.0
20187 A `lo_sum' data-linkage-table memory operand
20190 A memory operand that can be used as the destination operand
20191 of an integer store instruction
20194 A scaled or unscaled indexed memory operand
20197 A memory operand for floating-point loads and stores
20200 A register indirect memory operand
20202 _PowerPC and IBM RS6000--`config/rs6000/rs6000.h'_
20205 Address base register
20208 Floating point register
20214 `MQ', `CTR', or `LINK' register
20226 `CR' register (condition register) number 0
20229 `CR' register (condition register)
20232 `FPMEM' stack memory for FPR-GPR transfers
20235 Signed 16-bit constant
20238 Unsigned 16-bit constant shifted left 16 bits (use `L'
20239 instead for `SImode' constants)
20242 Unsigned 16-bit constant
20245 Signed 16-bit constant shifted left 16 bits
20248 Constant larger than 31
20257 Constant whose negation is a signed 16-bit constant
20260 Floating point constant that can be loaded into a register
20261 with one instruction per word
20264 Integer/Floating point constant that can be loaded into a
20265 register using three instructions
20268 Memory operand that is an offset from a register (`m' is
20269 preferable for `asm' statements)
20272 Memory operand that is an indexed or indirect from a register
20273 (`m' is preferable for `asm' statements)
20279 Address operand that is an indexed or indirect from a
20280 register (`p' is preferable for `asm' statements)
20283 Constant suitable as a 64-bit mask operand
20286 Constant suitable as a 32-bit mask operand
20289 System V Release 4 small data area reference
20292 AND masks that can be performed by two rldic{l, r}
20296 Vector constant that does not require memory
20299 _MorphoTech family--`config/mt/mt.h'_
20302 Constant for an arithmetic insn (16-bit signed integer).
20308 Constant for a logical insn (16-bit zero-extended integer).
20311 A constant that can be loaded with `lui' (i.e. the bottom 16
20315 A constant that takes two words to load (i.e. not matched by
20319 Negative 16-bit constants other than -65536.
20322 A 15-bit signed integer constant.
20325 A positive 16-bit constant.
20327 _Intel 386--`config/i386/constraints.md'_
20330 Legacy register--the eight integer registers available on all
20331 i386 processors (`a', `b', `c', `d', `si', `di', `bp', `sp').
20334 Any register accessible as `Rl'. In 32-bit mode, `a', `b',
20335 `c', and `d'; in 64-bit mode, any integer register.
20338 Any register accessible as `Rh': `a', `b', `c', and `d'.
20359 The `a' and `d' registers, as a pair (for instructions that
20360 return half the result in one and half in the other).
20363 Any 80387 floating-point (stack) register.
20366 Top of 80387 floating-point stack (`%st(0)').
20369 Second from top of 80387 floating-point stack (`%st(1)').
20378 Integer constant in the range 0 ... 31, for 32-bit shifts.
20381 Integer constant in the range 0 ... 63, for 64-bit shifts.
20384 Signed 8-bit integer constant.
20387 `0xFF' or `0xFFFF', for andsi as a zero-extending move.
20390 0, 1, 2, or 3 (shifts for the `lea' instruction).
20393 Unsigned 8-bit integer constant (for `in' and `out'
20397 Standard 80387 floating point constant.
20400 Standard SSE floating point constant.
20403 32-bit signed integer constant, or a symbolic reference known
20404 to fit that range (for immediate operands in sign-extending
20405 x86-64 instructions).
20408 32-bit unsigned integer constant, or a symbolic reference
20409 known to fit that range (for immediate operands in
20410 zero-extending x86-64 instructions).
20413 _Intel IA-64--`config/ia64/ia64.h'_
20416 General register `r0' to `r3' for `addl' instruction
20422 Predicate register (`c' as in "conditional")
20425 Application register residing in M-unit
20428 Application register residing in I-unit
20431 Floating-point register
20434 Memory operand. Remember that `m' allows postincrement and
20435 postdecrement which require printing with `%Pn' on IA-64.
20436 Use `S' to disallow postincrement and postdecrement.
20439 Floating-point constant 0.0 or 1.0
20442 14-bit signed integer constant
20445 22-bit signed integer constant
20448 8-bit signed integer constant for logical instructions
20451 8-bit adjusted signed integer constant for compare pseudo-ops
20454 6-bit unsigned integer constant for shift counts
20457 9-bit signed integer constant for load and store
20464 0 or -1 for `dep' instruction
20467 Non-volatile memory for floating-point loads and stores
20470 Integer constant in the range 1 to 4 for `shladd' instruction
20473 Memory operand except postincrement and postdecrement
20475 _FRV--`config/frv/frv.h'_
20478 Register in the class `ACC_REGS' (`acc0' to `acc7').
20481 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7').
20484 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0'
20488 Register in the class `GPR_REGS' (`gr0' to `gr63').
20491 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd
20492 registers are excluded not in the class but through the use
20493 of a machine mode larger than 4 bytes.
20496 Register in the class `FPR_REGS' (`fr0' to `fr63').
20499 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd
20500 registers are excluded not in the class but through the use
20501 of a machine mode larger than 4 bytes.
20504 Register in the class `LR_REG' (the `lr' register).
20507 Register in the class `QUAD_REGS' (`gr2' to `gr63').
20508 Register numbers not divisible by 4 are excluded not in the
20509 class but through the use of a machine mode larger than 8
20513 Register in the class `ICC_REGS' (`icc0' to `icc3').
20516 Register in the class `FCC_REGS' (`fcc0' to `fcc3').
20519 Register in the class `ICR_REGS' (`cc4' to `cc7').
20522 Register in the class `FCR_REGS' (`cc0' to `cc3').
20525 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63').
20526 Register numbers not divisible by 4 are excluded not in the
20527 class but through the use of a machine mode larger than 8
20531 Register in the class `SPR_REGS' (`lcr' and `lr').
20534 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7').
20537 Register in the class `ACCG_REGS' (`accg0' to `accg7').
20540 Register in the class `CR_REGS' (`cc0' to `cc7').
20543 Floating point constant zero
20546 6-bit signed integer constant
20549 10-bit signed integer constant
20552 16-bit signed integer constant
20555 16-bit unsigned integer constant
20558 12-bit signed integer constant that is negative--i.e. in the
20559 range of -2048 to -1
20565 12-bit signed integer constant that is greater than
20566 zero--i.e. in the range of 1 to 2047.
20569 _Blackfin family--`config/bfin/bfin.h'_
20578 A call clobbered P register.
20581 A single register. If N is in the range 0 to 7, the
20582 corresponding D register. If it is `A', then the register P0.
20585 Even-numbered D register
20588 Odd-numbered D register
20591 Accumulator register.
20594 Even-numbered accumulator register.
20597 Odd-numbered accumulator register.
20609 Registers used for circular buffering, i.e. I, B, or L
20625 Any D, P, B, M, I or L register.
20628 Additional registers typically used only in prologues and
20629 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and
20633 Any register except accumulators or CC.
20636 Signed 16 bit integer (in the range -32768 to 32767)
20639 Unsigned 16 bit integer (in the range 0 to 65535)
20642 Signed 7 bit integer (in the range -64 to 63)
20645 Unsigned 7 bit integer (in the range 0 to 127)
20648 Unsigned 5 bit integer (in the range 0 to 31)
20651 Signed 4 bit integer (in the range -8 to 7)
20654 Signed 3 bit integer (in the range -3 to 4)
20657 Unsigned 3 bit integer (in the range 0 to 7)
20660 Constant N, where N is a single-digit constant in the range 0
20664 An integer equal to one of the MACFLAG_XXX constants that is
20665 suitable for use with either accumulator.
20668 An integer equal to one of the MACFLAG_XXX constants that is
20669 suitable for use only with accumulator A1.
20678 An integer constant with exactly a single bit set.
20681 An integer constant with all bits set except exactly one.
20688 _M32C--`config/m32c/m32c.c'_
20693 `$sp', `$fb', `$sb'.
20696 Any control register, when they're 16 bits wide (nothing if
20697 control registers are 24 bits wide)
20700 Any control register, when they're 24 bits wide.
20706 $r0, $r1, $r2, $r3.
20709 $r0 or $r2, or $r2r0 for 32 bit values.
20712 $r1 or $r3, or $r3r1 for 32 bit values.
20715 A register that can hold a 64 bit value.
20718 $r0 or $r1 (registers with addressable high/low bytes)
20727 Address registers when they're 16 bits wide.
20730 Address registers when they're 24 bits wide.
20733 Registers that can hold QI values.
20736 Registers that can be used with displacements ($a0, $a1, $sb).
20739 Registers that can hold 32 bit values.
20742 Registers that can hold 16 bit values.
20745 Registers chat can hold 16 bit values, including all control
20749 $r0 through R1, plus $a0 and $a1.
20752 The flags register.
20755 The memory-based pseudo-registers $mem0 through $mem15.
20758 Registers that can hold pointers (16 bit registers for r8c,
20759 m16c; 24 bit registers for m32cm, m32c).
20762 Matches multiple registers in a PARALLEL to form a larger
20763 register. Used to match function return values.
20778 -8 ... -1 or 1 ... 8
20781 -16 ... -1 or 1 ... 16
20784 -32 ... -1 or 1 ... 32
20790 An 8 bit value with exactly one bit set.
20793 A 16 bit value with exactly one bit set.
20796 The common src/dest memory addressing modes.
20799 Memory addressed using $a0 or $a1.
20802 Memory addressed with immediate addresses.
20805 Memory addressed using the stack pointer ($sp).
20808 Memory addressed using the frame base register ($fb).
20811 Memory addressed using the small base register ($sb).
20816 _MIPS--`config/mips/constraints.md'_
20819 An address register. This is equivalent to `r' unless
20820 generating MIPS16 code.
20823 A floating-point register (if available).
20832 The `hi' and `lo' registers.
20835 A register suitable for use in an indirect jump. This will
20836 always be `$25' for `-mabicalls'.
20839 Equivalent to `r'; retained for backwards compatibility.
20842 A floating-point condition code register.
20845 A signed 16-bit constant (for arithmetic instructions).
20851 An unsigned 16-bit constant (for logic instructions).
20854 A signed 32-bit constant in which the lower 16 bits are zero.
20855 Such constants can be loaded using `lui'.
20858 A constant that cannot be loaded using `lui', `addiu' or
20862 A constant in the range -65535 to -1 (inclusive).
20865 A signed 15-bit constant.
20868 A constant in the range 1 to 65535 (inclusive).
20871 Floating-point zero.
20874 An address that can be used in a non-macro load or store.
20876 _Motorola 680x0--`config/m68k/constraints.md'_
20885 68881 floating-point register, if available
20888 Integer in the range 1 to 8
20891 16-bit signed number
20894 Signed number whose magnitude is greater than 0x80
20897 Integer in the range -8 to -1
20900 Signed number whose magnitude is greater than 0x100
20903 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate
20906 16 (for rotate using swap)
20909 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate
20912 Numbers that mov3q can handle
20915 Floating point constant that is not a 68881 constant
20918 Operands that satisfy 'm' when -mpcrel is in effect
20921 Operands that satisfy 's' when -mpcrel is not in effect
20924 Address register indirect addressing mode
20927 Register offset addressing
20933 symbol_ref or const
20942 Range of signed numbers that don't fit in 16 bits
20945 Integers valid for mvq
20948 Integers valid for a moveq followed by a swap
20951 Integers valid for mvz
20954 Integers valid for mvs
20960 Non-register operands allowed in clr
20963 _Motorola 68HC11 & 68HC12 families--`config/m68hc11/m68hc11.h'_
20978 Temporary soft register _.tmp
20981 A soft register _.d1 to _.d31
20984 Stack pointer register
20993 Pseudo register `z' (replaced by `x' or `y' at the end)
20996 An address register: x, y or z
20999 An address register: x or y
21002 Register pair (x:d) to form a 32-bit value
21005 Constants in the range -65536 to 65535
21008 Constants whose 16-bit low part is zero
21011 Constant integer 1 or -1
21014 Constant integer 16
21017 Constants in the range -8 to 2
21020 _SPARC--`config/sparc/sparc.h'_
21023 Floating-point register on the SPARC-V8 architecture and
21024 lower floating-point register on the SPARC-V9 architecture.
21027 Floating-point register. It is equivalent to `f' on the
21028 SPARC-V8 architecture and contains both lower and upper
21029 floating-point registers on the SPARC-V9 architecture.
21032 Floating-point condition code register.
21035 Lower floating-point register. It is only valid on the
21036 SPARC-V9 architecture when the Visual Instruction Set is
21040 Floating-point register. It is only valid on the SPARC-V9
21041 architecture when the Visual Instruction Set is available.
21044 64-bit global or out register for the SPARC-V8+ architecture.
21047 Signed 13-bit constant
21053 32-bit constant with the low 12 bits clear (a constant that
21054 can be loaded with the `sethi' instruction)
21057 A constant in the range supported by `movcc' instructions
21060 A constant in the range supported by `movrcc' instructions
21063 Same as `K', except that it verifies that bits that are not
21064 in the lower 32-bit range are all zero. Must be used instead
21065 of `K' for modes wider than `SImode'
21071 Floating-point zero
21074 Signed 13-bit constant, sign-extended to 32 or 64 bits
21077 Floating-point constant whose integral representation can be
21078 moved into an integer register using a single sethi
21082 Floating-point constant whose integral representation can be
21083 moved into an integer register using a single mov instruction
21086 Floating-point constant whose integral representation can be
21087 moved into an integer register using a high/lo_sum
21088 instruction sequence
21091 Memory address aligned to an 8-byte boundary
21097 Memory address for `e' constraint registers
21103 _SPU--`config/spu/spu.h'_
21106 An immediate which can be loaded with the il/ila/ilh/ilhu
21107 instructions. const_int is treated as a 64 bit value.
21110 An immediate for and/xor/or instructions. const_int is
21111 treated as a 64 bit value.
21114 An immediate for the `iohl' instruction. const_int is
21115 treated as a 64 bit value.
21118 An immediate which can be loaded with `fsmbi'.
21121 An immediate which can be loaded with the il/ila/ilh/ilhu
21122 instructions. const_int is treated as a 32 bit value.
21125 An immediate for most arithmetic instructions. const_int is
21126 treated as a 32 bit value.
21129 An immediate for and/xor/or instructions. const_int is
21130 treated as a 32 bit value.
21133 An immediate for the `iohl' instruction. const_int is
21134 treated as a 32 bit value.
21137 A constant in the range [-64, 63] for shift/rotate
21141 An unsigned 7-bit constant for conversion/nop/channel
21145 A signed 10-bit constant for most arithmetic instructions.
21148 A signed 16 bit immediate for `stop'.
21151 An unsigned 16-bit constant for `iohl' and `fsmbi'.
21154 An unsigned 7-bit constant whose 3 least significant bits are
21158 An unsigned 3-bit constant for 16-byte rotates and shifts
21161 Call operand, reg, for indirect calls
21164 Call operand, symbol, for relative calls.
21167 Call operand, const_int, for absolute calls.
21170 An immediate which can be loaded with the il/ila/ilh/ilhu
21171 instructions. const_int is sign extended to 128 bit.
21174 An immediate for shift and rotate instructions. const_int is
21175 treated as a 32 bit value.
21178 An immediate for and/xor/or instructions. const_int is sign
21179 extended as a 128 bit.
21182 An immediate for the `iohl' instruction. const_int is sign
21183 extended to 128 bit.
21186 _S/390 and zSeries--`config/s390/s390.h'_
21189 Address register (general purpose register except r0)
21192 Condition code register
21195 Data register (arbitrary general purpose register)
21198 Floating-point register
21201 Unsigned 8-bit constant (0-255)
21204 Unsigned 12-bit constant (0-4095)
21207 Signed 16-bit constant (-32768-32767)
21210 Value appropriate as displacement.
21212 for short displacement
21214 `(-524288..524287)'
21215 for long displacement
21218 Constant integer with a value of 0x7fffffff.
21221 Multiple letter constraint followed by 4 parameter letters.
21223 number of the part counting from most to least
21230 mode of the containing operand
21233 value of the other parts (F--all bits set)
21234 The constraint matches if the specified part of a constant
21235 has a value different from its other parts.
21238 Memory reference without index register and with short
21242 Memory reference with index register and short displacement.
21245 Memory reference without index register but with long
21249 Memory reference with index register and long displacement.
21252 Pointer with short displacement.
21255 Pointer with long displacement.
21258 Shift count operand.
21261 _Score family--`config/score/score.h'_
21264 Registers from r0 to r32.
21267 Registers from r0 to r16.
21270 r8--r11 or r22--r27 registers.
21291 cnt + lcb + scb register.
21294 cr0--cr15 register.
21306 cp1 + cp2 + cp3 registers.
21309 High 16-bit constant (32-bit constant with 16 LSBs zero).
21312 Unsigned 5 bit integer (in the range 0 to 31).
21315 Unsigned 16 bit integer (in the range 0 to 65535).
21318 Signed 16 bit integer (in the range -32768 to 32767).
21321 Unsigned 14 bit integer (in the range 0 to 16383).
21324 Signed 14 bit integer (in the range -8192 to 8191).
21329 _Xstormy16--`config/stormy16/stormy16.h'_
21344 Registers r0 through r7.
21347 Registers r0 and r1.
21350 The carry register.
21353 Registers r8 and r9.
21356 A constant between 0 and 3 inclusive.
21359 A constant that has exactly one bit set.
21362 A constant that has exactly one bit clear.
21365 A constant between 0 and 255 inclusive.
21368 A constant between -255 and 0 inclusive.
21371 A constant between -3 and 0 inclusive.
21374 A constant between 1 and 4 inclusive.
21377 A constant between -4 and -1 inclusive.
21380 A memory reference that is a stack push.
21383 A memory reference that is a stack pop.
21386 A memory reference that refers to a constant address of known
21390 The register indicated by Rx (not implemented yet).
21393 A constant that is not between 2 and 15 inclusive.
21399 _Xtensa--`config/xtensa/constraints.md'_
21402 General-purpose 32-bit register
21405 One-bit boolean register
21408 MAC16 40-bit accumulator register
21411 Signed 12-bit integer constant, for use in MOVI instructions
21414 Signed 8-bit integer constant, for use in ADDI instructions
21417 Integer constant valid for BccI instructions
21420 Unsigned constant valid for BccUI instructions
21425 File: gcc.info, Node: Asm Labels, Next: Explicit Reg Vars, Prev: Constraints, Up: C Extensions
21427 5.39 Controlling Names Used in Assembler Code
21428 =============================================
21430 You can specify the name to be used in the assembler code for a C
21431 function or variable by writing the `asm' (or `__asm__') keyword after
21432 the declarator as follows:
21434 int foo asm ("myfoo") = 2;
21436 This specifies that the name to be used for the variable `foo' in the
21437 assembler code should be `myfoo' rather than the usual `_foo'.
21439 On systems where an underscore is normally prepended to the name of a C
21440 function or variable, this feature allows you to define names for the
21441 linker that do not start with an underscore.
21443 It does not make sense to use this feature with a non-static local
21444 variable since such variables do not have assembler names. If you are
21445 trying to put the variable in a particular register, see *Note Explicit
21446 Reg Vars::. GCC presently accepts such code with a warning, but will
21447 probably be changed to issue an error, rather than a warning, in the
21450 You cannot use `asm' in this way in a function _definition_; but you
21451 can get the same effect by writing a declaration for the function
21452 before its definition and putting `asm' there, like this:
21454 extern func () asm ("FUNC");
21460 It is up to you to make sure that the assembler names you choose do not
21461 conflict with any other assembler symbols. Also, you must not use a
21462 register name; that would produce completely invalid assembler code.
21463 GCC does not as yet have the ability to store static variables in
21464 registers. Perhaps that will be added.
21467 File: gcc.info, Node: Explicit Reg Vars, Next: Alternate Keywords, Prev: Asm Labels, Up: C Extensions
21469 5.40 Variables in Specified Registers
21470 =====================================
21472 GNU C allows you to put a few global variables into specified hardware
21473 registers. You can also specify the register in which an ordinary
21474 register variable should be allocated.
21476 * Global register variables reserve registers throughout the program.
21477 This may be useful in programs such as programming language
21478 interpreters which have a couple of global variables that are
21479 accessed very often.
21481 * Local register variables in specific registers do not reserve the
21482 registers, except at the point where they are used as input or
21483 output operands in an `asm' statement and the `asm' statement
21484 itself is not deleted. The compiler's data flow analysis is
21485 capable of determining where the specified registers contain live
21486 values, and where they are available for other uses. Stores into
21487 local register variables may be deleted when they appear to be
21488 dead according to dataflow analysis. References to local register
21489 variables may be deleted or moved or simplified.
21491 These local variables are sometimes convenient for use with the
21492 extended `asm' feature (*note Extended Asm::), if you want to
21493 write one output of the assembler instruction directly into a
21494 particular register. (This will work provided the register you
21495 specify fits the constraints specified for that operand in the
21500 * Global Reg Vars::
21504 File: gcc.info, Node: Global Reg Vars, Next: Local Reg Vars, Up: Explicit Reg Vars
21506 5.40.1 Defining Global Register Variables
21507 -----------------------------------------
21509 You can define a global register variable in GNU C like this:
21511 register int *foo asm ("a5");
21513 Here `a5' is the name of the register which should be used. Choose a
21514 register which is normally saved and restored by function calls on your
21515 machine, so that library routines will not clobber it.
21517 Naturally the register name is cpu-dependent, so you would need to
21518 conditionalize your program according to cpu type. The register `a5'
21519 would be a good choice on a 68000 for a variable of pointer type. On
21520 machines with register windows, be sure to choose a "global" register
21521 that is not affected magically by the function call mechanism.
21523 In addition, operating systems on one type of cpu may differ in how
21524 they name the registers; then you would need additional conditionals.
21525 For example, some 68000 operating systems call this register `%a5'.
21527 Eventually there may be a way of asking the compiler to choose a
21528 register automatically, but first we need to figure out how it should
21529 choose and how to enable you to guide the choice. No solution is
21532 Defining a global register variable in a certain register reserves that
21533 register entirely for this use, at least within the current compilation.
21534 The register will not be allocated for any other purpose in the
21535 functions in the current compilation. The register will not be saved
21536 and restored by these functions. Stores into this register are never
21537 deleted even if they would appear to be dead, but references may be
21538 deleted or moved or simplified.
21540 It is not safe to access the global register variables from signal
21541 handlers, or from more than one thread of control, because the system
21542 library routines may temporarily use the register for other things
21543 (unless you recompile them specially for the task at hand).
21545 It is not safe for one function that uses a global register variable to
21546 call another such function `foo' by way of a third function `lose' that
21547 was compiled without knowledge of this variable (i.e. in a different
21548 source file in which the variable wasn't declared). This is because
21549 `lose' might save the register and put some other value there. For
21550 example, you can't expect a global register variable to be available in
21551 the comparison-function that you pass to `qsort', since `qsort' might
21552 have put something else in that register. (If you are prepared to
21553 recompile `qsort' with the same global register variable, you can solve
21556 If you want to recompile `qsort' or other source files which do not
21557 actually use your global register variable, so that they will not use
21558 that register for any other purpose, then it suffices to specify the
21559 compiler option `-ffixed-REG'. You need not actually add a global
21560 register declaration to their source code.
21562 A function which can alter the value of a global register variable
21563 cannot safely be called from a function compiled without this variable,
21564 because it could clobber the value the caller expects to find there on
21565 return. Therefore, the function which is the entry point into the part
21566 of the program that uses the global register variable must explicitly
21567 save and restore the value which belongs to its caller.
21569 On most machines, `longjmp' will restore to each global register
21570 variable the value it had at the time of the `setjmp'. On some
21571 machines, however, `longjmp' will not change the value of global
21572 register variables. To be portable, the function that called `setjmp'
21573 should make other arrangements to save the values of the global register
21574 variables, and to restore them in a `longjmp'. This way, the same
21575 thing will happen regardless of what `longjmp' does.
21577 All global register variable declarations must precede all function
21578 definitions. If such a declaration could appear after function
21579 definitions, the declaration would be too late to prevent the register
21580 from being used for other purposes in the preceding functions.
21582 Global register variables may not have initial values, because an
21583 executable file has no means to supply initial contents for a register.
21585 On the SPARC, there are reports that g3 ... g7 are suitable registers,
21586 but certain library functions, such as `getwd', as well as the
21587 subroutines for division and remainder, modify g3 and g4. g1 and g2
21588 are local temporaries.
21590 On the 68000, a2 ... a5 should be suitable, as should d2 ... d7. Of
21591 course, it will not do to use more than a few of those.
21594 File: gcc.info, Node: Local Reg Vars, Prev: Global Reg Vars, Up: Explicit Reg Vars
21596 5.40.2 Specifying Registers for Local Variables
21597 -----------------------------------------------
21599 You can define a local register variable with a specified register like
21602 register int *foo asm ("a5");
21604 Here `a5' is the name of the register which should be used. Note that
21605 this is the same syntax used for defining global register variables,
21606 but for a local variable it would appear within a function.
21608 Naturally the register name is cpu-dependent, but this is not a
21609 problem, since specific registers are most often useful with explicit
21610 assembler instructions (*note Extended Asm::). Both of these things
21611 generally require that you conditionalize your program according to cpu
21614 In addition, operating systems on one type of cpu may differ in how
21615 they name the registers; then you would need additional conditionals.
21616 For example, some 68000 operating systems call this register `%a5'.
21618 Defining such a register variable does not reserve the register; it
21619 remains available for other uses in places where flow control determines
21620 the variable's value is not live.
21622 This option does not guarantee that GCC will generate code that has
21623 this variable in the register you specify at all times. You may not
21624 code an explicit reference to this register in the _assembler
21625 instruction template_ part of an `asm' statement and assume it will
21626 always refer to this variable. However, using the variable as an `asm'
21627 _operand_ guarantees that the specified register is used for the
21630 Stores into local register variables may be deleted when they appear
21631 to be dead according to dataflow analysis. References to local
21632 register variables may be deleted or moved or simplified.
21634 As for global register variables, it's recommended that you choose a
21635 register which is normally saved and restored by function calls on your
21636 machine, so that library routines will not clobber it. A common
21637 pitfall is to initialize multiple call-clobbered registers with
21638 arbitrary expressions, where a function call or library call for an
21639 arithmetic operator will overwrite a register value from a previous
21640 assignment, for example `r0' below:
21641 register int *p1 asm ("r0") = ...;
21642 register int *p2 asm ("r1") = ...;
21643 In those cases, a solution is to use a temporary variable for each
21644 arbitrary expression. *Note Example of asm with clobbered asm reg::.
21647 File: gcc.info, Node: Alternate Keywords, Next: Incomplete Enums, Prev: Explicit Reg Vars, Up: C Extensions
21649 5.41 Alternate Keywords
21650 =======================
21652 `-ansi' and the various `-std' options disable certain keywords. This
21653 causes trouble when you want to use GNU C extensions, or a
21654 general-purpose header file that should be usable by all programs,
21655 including ISO C programs. The keywords `asm', `typeof' and `inline'
21656 are not available in programs compiled with `-ansi' or `-std' (although
21657 `inline' can be used in a program compiled with `-std=c99'). The ISO
21658 C99 keyword `restrict' is only available when `-std=gnu99' (which will
21659 eventually be the default) or `-std=c99' (or the equivalent
21660 `-std=iso9899:1999') is used.
21662 The way to solve these problems is to put `__' at the beginning and
21663 end of each problematical keyword. For example, use `__asm__' instead
21664 of `asm', and `__inline__' instead of `inline'.
21666 Other C compilers won't accept these alternative keywords; if you want
21667 to compile with another compiler, you can define the alternate keywords
21668 as macros to replace them with the customary keywords. It looks like
21672 #define __asm__ asm
21675 `-pedantic' and other options cause warnings for many GNU C extensions.
21676 You can prevent such warnings within one expression by writing
21677 `__extension__' before the expression. `__extension__' has no effect
21681 File: gcc.info, Node: Incomplete Enums, Next: Function Names, Prev: Alternate Keywords, Up: C Extensions
21683 5.42 Incomplete `enum' Types
21684 ============================
21686 You can define an `enum' tag without specifying its possible values.
21687 This results in an incomplete type, much like what you get if you write
21688 `struct foo' without describing the elements. A later declaration
21689 which does specify the possible values completes the type.
21691 You can't allocate variables or storage using the type while it is
21692 incomplete. However, you can work with pointers to that type.
21694 This extension may not be very useful, but it makes the handling of
21695 `enum' more consistent with the way `struct' and `union' are handled.
21697 This extension is not supported by GNU C++.
21700 File: gcc.info, Node: Function Names, Next: Return Address, Prev: Incomplete Enums, Up: C Extensions
21702 5.43 Function Names as Strings
21703 ==============================
21705 GCC provides three magic variables which hold the name of the current
21706 function, as a string. The first of these is `__func__', which is part
21707 of the C99 standard:
21709 The identifier `__func__' is implicitly declared by the translator
21710 as if, immediately following the opening brace of each function
21711 definition, the declaration
21712 static const char __func__[] = "function-name";
21714 appeared, where function-name is the name of the lexically-enclosing
21715 function. This name is the unadorned name of the function.
21717 `__FUNCTION__' is another name for `__func__'. Older versions of GCC
21718 recognize only this name. However, it is not standardized. For
21719 maximum portability, we recommend you use `__func__', but provide a
21720 fallback definition with the preprocessor:
21722 #if __STDC_VERSION__ < 199901L
21724 # define __func__ __FUNCTION__
21726 # define __func__ "<unknown>"
21730 In C, `__PRETTY_FUNCTION__' is yet another name for `__func__'.
21731 However, in C++, `__PRETTY_FUNCTION__' contains the type signature of
21732 the function as well as its bare name. For example, this program:
21735 extern int printf (char *, ...);
21742 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
21743 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
21758 __PRETTY_FUNCTION__ = void a::sub(int)
21760 These identifiers are not preprocessor macros. In GCC 3.3 and
21761 earlier, in C only, `__FUNCTION__' and `__PRETTY_FUNCTION__' were
21762 treated as string literals; they could be used to initialize `char'
21763 arrays, and they could be concatenated with other string literals. GCC
21764 3.4 and later treat them as variables, like `__func__'. In C++,
21765 `__FUNCTION__' and `__PRETTY_FUNCTION__' have always been variables.
21768 File: gcc.info, Node: Return Address, Next: Vector Extensions, Prev: Function Names, Up: C Extensions
21770 5.44 Getting the Return or Frame Address of a Function
21771 ======================================================
21773 These functions may be used to get information about the callers of a
21776 -- Built-in Function: void * __builtin_return_address (unsigned int
21778 This function returns the return address of the current function,
21779 or of one of its callers. The LEVEL argument is number of frames
21780 to scan up the call stack. A value of `0' yields the return
21781 address of the current function, a value of `1' yields the return
21782 address of the caller of the current function, and so forth. When
21783 inlining the expected behavior is that the function will return
21784 the address of the function that will be returned to. To work
21785 around this behavior use the `noinline' function attribute.
21787 The LEVEL argument must be a constant integer.
21789 On some machines it may be impossible to determine the return
21790 address of any function other than the current one; in such cases,
21791 or when the top of the stack has been reached, this function will
21792 return `0' or a random value. In addition,
21793 `__builtin_frame_address' may be used to determine if the top of
21794 the stack has been reached.
21796 This function should only be used with a nonzero argument for
21797 debugging purposes.
21799 -- Built-in Function: void * __builtin_frame_address (unsigned int
21801 This function is similar to `__builtin_return_address', but it
21802 returns the address of the function frame rather than the return
21803 address of the function. Calling `__builtin_frame_address' with a
21804 value of `0' yields the frame address of the current function, a
21805 value of `1' yields the frame address of the caller of the current
21806 function, and so forth.
21808 The frame is the area on the stack which holds local variables and
21809 saved registers. The frame address is normally the address of the
21810 first word pushed on to the stack by the function. However, the
21811 exact definition depends upon the processor and the calling
21812 convention. If the processor has a dedicated frame pointer
21813 register, and the function has a frame, then
21814 `__builtin_frame_address' will return the value of the frame
21817 On some machines it may be impossible to determine the frame
21818 address of any function other than the current one; in such cases,
21819 or when the top of the stack has been reached, this function will
21820 return `0' if the first frame pointer is properly initialized by
21823 This function should only be used with a nonzero argument for
21824 debugging purposes.
21827 File: gcc.info, Node: Vector Extensions, Next: Offsetof, Prev: Return Address, Up: C Extensions
21829 5.45 Using vector instructions through built-in functions
21830 =========================================================
21832 On some targets, the instruction set contains SIMD vector instructions
21833 that operate on multiple values contained in one large register at the
21834 same time. For example, on the i386 the MMX, 3Dnow! and SSE extensions
21835 can be used this way.
21837 The first step in using these extensions is to provide the necessary
21838 data types. This should be done using an appropriate `typedef':
21840 typedef int v4si __attribute__ ((vector_size (16)));
21842 The `int' type specifies the base type, while the attribute specifies
21843 the vector size for the variable, measured in bytes. For example, the
21844 declaration above causes the compiler to set the mode for the `v4si'
21845 type to be 16 bytes wide and divided into `int' sized units. For a
21846 32-bit `int' this means a vector of 4 units of 4 bytes, and the
21847 corresponding mode of `foo' will be V4SI.
21849 The `vector_size' attribute is only applicable to integral and float
21850 scalars, although arrays, pointers, and function return values are
21851 allowed in conjunction with this construct.
21853 All the basic integer types can be used as base types, both as signed
21854 and as unsigned: `char', `short', `int', `long', `long long'. In
21855 addition, `float' and `double' can be used to build floating-point
21858 Specifying a combination that is not valid for the current architecture
21859 will cause GCC to synthesize the instructions using a narrower mode.
21860 For example, if you specify a variable of type `V4SI' and your
21861 architecture does not allow for this specific SIMD type, GCC will
21862 produce code that uses 4 `SIs'.
21864 The types defined in this manner can be used with a subset of normal C
21865 operations. Currently, GCC will allow using the following operators on
21866 these types: `+, -, *, /, unary minus, ^, |, &, ~'.
21868 The operations behave like C++ `valarrays'. Addition is defined as
21869 the addition of the corresponding elements of the operands. For
21870 example, in the code below, each of the 4 elements in A will be added
21871 to the corresponding 4 elements in B and the resulting vector will be
21874 typedef int v4si __attribute__ ((vector_size (16)));
21880 Subtraction, multiplication, division, and the logical operations
21881 operate in a similar manner. Likewise, the result of using the unary
21882 minus or complement operators on a vector type is a vector whose
21883 elements are the negative or complemented values of the corresponding
21884 elements in the operand.
21886 You can declare variables and use them in function calls and returns,
21887 as well as in assignments and some casts. You can specify a vector
21888 type as a return type for a function. Vector types can also be used as
21889 function arguments. It is possible to cast from one vector type to
21890 another, provided they are of the same size (in fact, you can also cast
21891 vectors to and from other datatypes of the same size).
21893 You cannot operate between vectors of different lengths or different
21894 signedness without a cast.
21896 A port that supports hardware vector operations, usually provides a set
21897 of built-in functions that can be used to operate on vectors. For
21898 example, a function to add two vectors and multiply the result by a
21899 third could look like this:
21901 v4si f (v4si a, v4si b, v4si c)
21903 v4si tmp = __builtin_addv4si (a, b);
21904 return __builtin_mulv4si (tmp, c);
21908 File: gcc.info, Node: Offsetof, Next: Atomic Builtins, Prev: Vector Extensions, Up: C Extensions
21913 GCC implements for both C and C++ a syntactic extension to implement
21914 the `offsetof' macro.
21917 "__builtin_offsetof" "(" `typename' "," offsetof_member_designator ")"
21919 offsetof_member_designator:
21921 | offsetof_member_designator "." `identifier'
21922 | offsetof_member_designator "[" `expr' "]"
21924 This extension is sufficient such that
21926 #define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER)
21928 is a suitable definition of the `offsetof' macro. In C++, TYPE may be
21929 dependent. In either case, MEMBER may consist of a single identifier,
21930 or a sequence of member accesses and array references.
21933 File: gcc.info, Node: Atomic Builtins, Next: Object Size Checking, Prev: Offsetof, Up: C Extensions
21935 5.47 Built-in functions for atomic memory access
21936 ================================================
21938 The following builtins are intended to be compatible with those
21939 described in the `Intel Itanium Processor-specific Application Binary
21940 Interface', section 7.4. As such, they depart from the normal GCC
21941 practice of using the "__builtin_" prefix, and further that they are
21942 overloaded such that they work on multiple types.
21944 The definition given in the Intel documentation allows only for the
21945 use of the types `int', `long', `long long' as well as their unsigned
21946 counterparts. GCC will allow any integral scalar or pointer type that
21947 is 1, 2, 4 or 8 bytes in length.
21949 Not all operations are supported by all target processors. If a
21950 particular operation cannot be implemented on the target processor, a
21951 warning will be generated and a call an external function will be
21952 generated. The external function will carry the same name as the
21953 builtin, with an additional suffix `_N' where N is the size of the data
21956 In most cases, these builtins are considered a "full barrier". That
21957 is, no memory operand will be moved across the operation, either
21958 forward or backward. Further, instructions will be issued as necessary
21959 to prevent the processor from speculating loads across the operation
21960 and from queuing stores after the operation.
21962 All of the routines are are described in the Intel documentation to
21963 take "an optional list of variables protected by the memory barrier".
21964 It's not clear what is meant by that; it could mean that _only_ the
21965 following variables are protected, or it could mean that these variables
21966 should in addition be protected. At present GCC ignores this list and
21967 protects all variables which are globally accessible. If in the future
21968 we make some use of this list, an empty list will continue to mean all
21969 globally accessible variables.
21971 `TYPE __sync_fetch_and_add (TYPE *ptr, TYPE value, ...)'
21972 `TYPE __sync_fetch_and_sub (TYPE *ptr, TYPE value, ...)'
21973 `TYPE __sync_fetch_and_or (TYPE *ptr, TYPE value, ...)'
21974 `TYPE __sync_fetch_and_and (TYPE *ptr, TYPE value, ...)'
21975 `TYPE __sync_fetch_and_xor (TYPE *ptr, TYPE value, ...)'
21976 `TYPE __sync_fetch_and_nand (TYPE *ptr, TYPE value, ...)'
21977 These builtins perform the operation suggested by the name, and
21978 returns the value that had previously been in memory. That is,
21980 { tmp = *ptr; *ptr OP= value; return tmp; }
21981 { tmp = *ptr; *ptr = ~tmp & value; return tmp; } // nand
21983 `TYPE __sync_add_and_fetch (TYPE *ptr, TYPE value, ...)'
21984 `TYPE __sync_sub_and_fetch (TYPE *ptr, TYPE value, ...)'
21985 `TYPE __sync_or_and_fetch (TYPE *ptr, TYPE value, ...)'
21986 `TYPE __sync_and_and_fetch (TYPE *ptr, TYPE value, ...)'
21987 `TYPE __sync_xor_and_fetch (TYPE *ptr, TYPE value, ...)'
21988 `TYPE __sync_nand_and_fetch (TYPE *ptr, TYPE value, ...)'
21989 These builtins perform the operation suggested by the name, and
21990 return the new value. That is,
21992 { *ptr OP= value; return *ptr; }
21993 { *ptr = ~*ptr & value; return *ptr; } // nand
21995 `bool __sync_bool_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
21996 `TYPE __sync_val_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
21997 These builtins perform an atomic compare and swap. That is, if
21998 the current value of `*PTR' is OLDVAL, then write NEWVAL into
22001 The "bool" version returns true if the comparison is successful and
22002 NEWVAL was written. The "val" version returns the contents of
22003 `*PTR' before the operation.
22005 `__sync_synchronize (...)'
22006 This builtin issues a full memory barrier.
22008 `TYPE __sync_lock_test_and_set (TYPE *ptr, TYPE value, ...)'
22009 This builtin, as described by Intel, is not a traditional
22010 test-and-set operation, but rather an atomic exchange operation.
22011 It writes VALUE into `*PTR', and returns the previous contents of
22014 Many targets have only minimal support for such locks, and do not
22015 support a full exchange operation. In this case, a target may
22016 support reduced functionality here by which the _only_ valid value
22017 to store is the immediate constant 1. The exact value actually
22018 stored in `*PTR' is implementation defined.
22020 This builtin is not a full barrier, but rather an "acquire
22021 barrier". This means that references after the builtin cannot
22022 move to (or be speculated to) before the builtin, but previous
22023 memory stores may not be globally visible yet, and previous memory
22024 loads may not yet be satisfied.
22026 `void __sync_lock_release (TYPE *ptr, ...)'
22027 This builtin releases the lock acquired by
22028 `__sync_lock_test_and_set'. Normally this means writing the
22029 constant 0 to `*PTR'.
22031 This builtin is not a full barrier, but rather a "release barrier".
22032 This means that all previous memory stores are globally visible,
22033 and all previous memory loads have been satisfied, but following
22034 memory reads are not prevented from being speculated to before the
22038 File: gcc.info, Node: Object Size Checking, Next: Other Builtins, Prev: Atomic Builtins, Up: C Extensions
22040 5.48 Object Size Checking Builtins
22041 ==================================
22043 GCC implements a limited buffer overflow protection mechanism that can
22044 prevent some buffer overflow attacks.
22046 -- Built-in Function: size_t __builtin_object_size (void * PTR, int
22048 is a built-in construct that returns a constant number of bytes
22049 from PTR to the end of the object PTR pointer points to (if known
22050 at compile time). `__builtin_object_size' never evaluates its
22051 arguments for side-effects. If there are any side-effects in
22052 them, it returns `(size_t) -1' for TYPE 0 or 1 and `(size_t) 0'
22053 for TYPE 2 or 3. If there are multiple objects PTR can point to
22054 and all of them are known at compile time, the returned number is
22055 the maximum of remaining byte counts in those objects if TYPE & 2
22056 is 0 and minimum if nonzero. If it is not possible to determine
22057 which objects PTR points to at compile time,
22058 `__builtin_object_size' should return `(size_t) -1' for TYPE 0 or
22059 1 and `(size_t) 0' for TYPE 2 or 3.
22061 TYPE is an integer constant from 0 to 3. If the least significant
22062 bit is clear, objects are whole variables, if it is set, a closest
22063 surrounding subobject is considered the object a pointer points to.
22064 The second bit determines if maximum or minimum of remaining bytes
22067 struct V { char buf1[10]; int b; char buf2[10]; } var;
22068 char *p = &var.buf1[1], *q = &var.b;
22070 /* Here the object p points to is var. */
22071 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
22072 /* The subobject p points to is var.buf1. */
22073 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
22074 /* The object q points to is var. */
22075 assert (__builtin_object_size (q, 0)
22076 == (char *) (&var + 1) - (char *) &var.b);
22077 /* The subobject q points to is var.b. */
22078 assert (__builtin_object_size (q, 1) == sizeof (var.b));
22080 There are built-in functions added for many common string operation
22081 functions, e.g., for `memcpy' `__builtin___memcpy_chk' built-in is
22082 provided. This built-in has an additional last argument, which is the
22083 number of bytes remaining in object the DEST argument points to or
22084 `(size_t) -1' if the size is not known.
22086 The built-in functions are optimized into the normal string functions
22087 like `memcpy' if the last argument is `(size_t) -1' or if it is known
22088 at compile time that the destination object will not be overflown. If
22089 the compiler can determine at compile time the object will be always
22090 overflown, it issues a warning.
22092 The intended use can be e.g.
22095 #define bos0(dest) __builtin_object_size (dest, 0)
22096 #define memcpy(dest, src, n) \
22097 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
22101 /* It is unknown what object p points to, so this is optimized
22102 into plain memcpy - no checking is possible. */
22103 memcpy (p, "abcde", n);
22104 /* Destination is known and length too. It is known at compile
22105 time there will be no overflow. */
22106 memcpy (&buf[5], "abcde", 5);
22107 /* Destination is known, but the length is not known at compile time.
22108 This will result in __memcpy_chk call that can check for overflow
22110 memcpy (&buf[5], "abcde", n);
22111 /* Destination is known and it is known at compile time there will
22112 be overflow. There will be a warning and __memcpy_chk call that
22113 will abort the program at runtime. */
22114 memcpy (&buf[6], "abcde", 5);
22116 Such built-in functions are provided for `memcpy', `mempcpy',
22117 `memmove', `memset', `strcpy', `stpcpy', `strncpy', `strcat' and
22120 There are also checking built-in functions for formatted output
22122 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
22123 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
22124 const char *fmt, ...);
22125 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
22127 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
22128 const char *fmt, va_list ap);
22130 The added FLAG argument is passed unchanged to `__sprintf_chk' etc.
22131 functions and can contain implementation specific flags on what
22132 additional security measures the checking function might take, such as
22133 handling `%n' differently.
22135 The OS argument is the object size S points to, like in the other
22136 built-in functions. There is a small difference in the behavior
22137 though, if OS is `(size_t) -1', the built-in functions are optimized
22138 into the non-checking functions only if FLAG is 0, otherwise the
22139 checking function is called with OS argument set to `(size_t) -1'.
22141 In addition to this, there are checking built-in functions
22142 `__builtin___printf_chk', `__builtin___vprintf_chk',
22143 `__builtin___fprintf_chk' and `__builtin___vfprintf_chk'. These have
22144 just one additional argument, FLAG, right before format string FMT. If
22145 the compiler is able to optimize them to `fputc' etc. functions, it
22146 will, otherwise the checking function should be called and the FLAG
22147 argument passed to it.
22150 File: gcc.info, Node: Other Builtins, Next: Target Builtins, Prev: Object Size Checking, Up: C Extensions
22152 5.49 Other built-in functions provided by GCC
22153 =============================================
22155 GCC provides a large number of built-in functions other than the ones
22156 mentioned above. Some of these are for internal use in the processing
22157 of exceptions or variable-length argument lists and will not be
22158 documented here because they may change from time to time; we do not
22159 recommend general use of these functions.
22161 The remaining functions are provided for optimization purposes.
22163 GCC includes built-in versions of many of the functions in the standard
22164 C library. The versions prefixed with `__builtin_' will always be
22165 treated as having the same meaning as the C library function even if you
22166 specify the `-fno-builtin' option. (*note C Dialect Options::) Many of
22167 these functions are only optimized in certain cases; if they are not
22168 optimized in a particular case, a call to the library function will be
22171 Outside strict ISO C mode (`-ansi', `-std=c89' or `-std=c99'), the
22172 functions `_exit', `alloca', `bcmp', `bzero', `dcgettext', `dgettext',
22173 `dremf', `dreml', `drem', `exp10f', `exp10l', `exp10', `ffsll', `ffsl',
22174 `ffs', `fprintf_unlocked', `fputs_unlocked', `gammaf', `gammal',
22175 `gamma', `gammaf_r', `gammal_r', `gamma_r', `gettext', `index',
22176 `isascii', `j0f', `j0l', `j0', `j1f', `j1l', `j1', `jnf', `jnl', `jn',
22177 `lgammaf_r', `lgammal_r', `lgamma_r', `mempcpy', `pow10f', `pow10l',
22178 `pow10', `printf_unlocked', `rindex', `scalbf', `scalbl', `scalb',
22179 `signbit', `signbitf', `signbitl', `signbitd32', `signbitd64',
22180 `signbitd128', `significandf', `significandl', `significand', `sincosf',
22181 `sincosl', `sincos', `stpcpy', `stpncpy', `strcasecmp', `strdup',
22182 `strfmon', `strncasecmp', `strndup', `toascii', `y0f', `y0l', `y0',
22183 `y1f', `y1l', `y1', `ynf', `ynl' and `yn' may be handled as built-in
22184 functions. All these functions have corresponding versions prefixed
22185 with `__builtin_', which may be used even in strict C89 mode.
22187 The ISO C99 functions `_Exit', `acoshf', `acoshl', `acosh', `asinhf',
22188 `asinhl', `asinh', `atanhf', `atanhl', `atanh', `cabsf', `cabsl',
22189 `cabs', `cacosf', `cacoshf', `cacoshl', `cacosh', `cacosl', `cacos',
22190 `cargf', `cargl', `carg', `casinf', `casinhf', `casinhl', `casinh',
22191 `casinl', `casin', `catanf', `catanhf', `catanhl', `catanh', `catanl',
22192 `catan', `cbrtf', `cbrtl', `cbrt', `ccosf', `ccoshf', `ccoshl',
22193 `ccosh', `ccosl', `ccos', `cexpf', `cexpl', `cexp', `cimagf', `cimagl',
22194 `cimag', `clogf', `clogl', `clog', `conjf', `conjl', `conj',
22195 `copysignf', `copysignl', `copysign', `cpowf', `cpowl', `cpow',
22196 `cprojf', `cprojl', `cproj', `crealf', `creall', `creal', `csinf',
22197 `csinhf', `csinhl', `csinh', `csinl', `csin', `csqrtf', `csqrtl',
22198 `csqrt', `ctanf', `ctanhf', `ctanhl', `ctanh', `ctanl', `ctan',
22199 `erfcf', `erfcl', `erfc', `erff', `erfl', `erf', `exp2f', `exp2l',
22200 `exp2', `expm1f', `expm1l', `expm1', `fdimf', `fdiml', `fdim', `fmaf',
22201 `fmal', `fmaxf', `fmaxl', `fmax', `fma', `fminf', `fminl', `fmin',
22202 `hypotf', `hypotl', `hypot', `ilogbf', `ilogbl', `ilogb', `imaxabs',
22203 `isblank', `iswblank', `lgammaf', `lgammal', `lgamma', `llabs',
22204 `llrintf', `llrintl', `llrint', `llroundf', `llroundl', `llround',
22205 `log1pf', `log1pl', `log1p', `log2f', `log2l', `log2', `logbf',
22206 `logbl', `logb', `lrintf', `lrintl', `lrint', `lroundf', `lroundl',
22207 `lround', `nearbyintf', `nearbyintl', `nearbyint', `nextafterf',
22208 `nextafterl', `nextafter', `nexttowardf', `nexttowardl', `nexttoward',
22209 `remainderf', `remainderl', `remainder', `remquof', `remquol',
22210 `remquo', `rintf', `rintl', `rint', `roundf', `roundl', `round',
22211 `scalblnf', `scalblnl', `scalbln', `scalbnf', `scalbnl', `scalbn',
22212 `snprintf', `tgammaf', `tgammal', `tgamma', `truncf', `truncl', `trunc',
22213 `vfscanf', `vscanf', `vsnprintf' and `vsscanf' are handled as built-in
22214 functions except in strict ISO C90 mode (`-ansi' or `-std=c89').
22216 There are also built-in versions of the ISO C99 functions `acosf',
22217 `acosl', `asinf', `asinl', `atan2f', `atan2l', `atanf', `atanl',
22218 `ceilf', `ceill', `cosf', `coshf', `coshl', `cosl', `expf', `expl',
22219 `fabsf', `fabsl', `floorf', `floorl', `fmodf', `fmodl', `frexpf',
22220 `frexpl', `ldexpf', `ldexpl', `log10f', `log10l', `logf', `logl',
22221 `modfl', `modf', `powf', `powl', `sinf', `sinhf', `sinhl', `sinl',
22222 `sqrtf', `sqrtl', `tanf', `tanhf', `tanhl' and `tanl' that are
22223 recognized in any mode since ISO C90 reserves these names for the
22224 purpose to which ISO C99 puts them. All these functions have
22225 corresponding versions prefixed with `__builtin_'.
22227 The ISO C94 functions `iswalnum', `iswalpha', `iswcntrl', `iswdigit',
22228 `iswgraph', `iswlower', `iswprint', `iswpunct', `iswspace', `iswupper',
22229 `iswxdigit', `towlower' and `towupper' are handled as built-in functions
22230 except in strict ISO C90 mode (`-ansi' or `-std=c89').
22232 The ISO C90 functions `abort', `abs', `acos', `asin', `atan2', `atan',
22233 `calloc', `ceil', `cosh', `cos', `exit', `exp', `fabs', `floor', `fmod',
22234 `fprintf', `fputs', `frexp', `fscanf', `isalnum', `isalpha', `iscntrl',
22235 `isdigit', `isgraph', `islower', `isprint', `ispunct', `isspace',
22236 `isupper', `isxdigit', `tolower', `toupper', `labs', `ldexp', `log10',
22237 `log', `malloc', `memchr', `memcmp', `memcpy', `memset', `modf', `pow',
22238 `printf', `putchar', `puts', `scanf', `sinh', `sin', `snprintf',
22239 `sprintf', `sqrt', `sscanf', `strcat', `strchr', `strcmp', `strcpy',
22240 `strcspn', `strlen', `strncat', `strncmp', `strncpy', `strpbrk',
22241 `strrchr', `strspn', `strstr', `tanh', `tan', `vfprintf', `vprintf' and
22242 `vsprintf' are all recognized as built-in functions unless
22243 `-fno-builtin' is specified (or `-fno-builtin-FUNCTION' is specified
22244 for an individual function). All of these functions have corresponding
22245 versions prefixed with `__builtin_'.
22247 GCC provides built-in versions of the ISO C99 floating point comparison
22248 macros that avoid raising exceptions for unordered operands. They have
22249 the same names as the standard macros ( `isgreater', `isgreaterequal',
22250 `isless', `islessequal', `islessgreater', and `isunordered') , with
22251 `__builtin_' prefixed. We intend for a library implementor to be able
22252 to simply `#define' each standard macro to its built-in equivalent. In
22253 the same fashion, GCC provides `isfinite' and `isnormal' built-ins used
22254 with `__builtin_' prefixed.
22256 -- Built-in Function: int __builtin_types_compatible_p (TYPE1, TYPE2)
22257 You can use the built-in function `__builtin_types_compatible_p' to
22258 determine whether two types are the same.
22260 This built-in function returns 1 if the unqualified versions of the
22261 types TYPE1 and TYPE2 (which are types, not expressions) are
22262 compatible, 0 otherwise. The result of this built-in function can
22263 be used in integer constant expressions.
22265 This built-in function ignores top level qualifiers (e.g., `const',
22266 `volatile'). For example, `int' is equivalent to `const int'.
22268 The type `int[]' and `int[5]' are compatible. On the other hand,
22269 `int' and `char *' are not compatible, even if the size of their
22270 types, on the particular architecture are the same. Also, the
22271 amount of pointer indirection is taken into account when
22272 determining similarity. Consequently, `short *' is not similar to
22273 `short **'. Furthermore, two types that are typedefed are
22274 considered compatible if their underlying types are compatible.
22276 An `enum' type is not considered to be compatible with another
22277 `enum' type even if both are compatible with the same integer
22278 type; this is what the C standard specifies. For example, `enum
22279 {foo, bar}' is not similar to `enum {hot, dog}'.
22281 You would typically use this function in code whose execution
22282 varies depending on the arguments' types. For example:
22286 typeof (x) tmp = (x); \
22287 if (__builtin_types_compatible_p (typeof (x), long double)) \
22288 tmp = foo_long_double (tmp); \
22289 else if (__builtin_types_compatible_p (typeof (x), double)) \
22290 tmp = foo_double (tmp); \
22291 else if (__builtin_types_compatible_p (typeof (x), float)) \
22292 tmp = foo_float (tmp); \
22298 _Note:_ This construct is only available for C.
22301 -- Built-in Function: TYPE __builtin_choose_expr (CONST_EXP, EXP1,
22303 You can use the built-in function `__builtin_choose_expr' to
22304 evaluate code depending on the value of a constant expression.
22305 This built-in function returns EXP1 if CONST_EXP, which is a
22306 constant expression that must be able to be determined at compile
22307 time, is nonzero. Otherwise it returns 0.
22309 This built-in function is analogous to the `? :' operator in C,
22310 except that the expression returned has its type unaltered by
22311 promotion rules. Also, the built-in function does not evaluate
22312 the expression that was not chosen. For example, if CONST_EXP
22313 evaluates to true, EXP2 is not evaluated even if it has
22316 This built-in function can return an lvalue if the chosen argument
22319 If EXP1 is returned, the return type is the same as EXP1's type.
22320 Similarly, if EXP2 is returned, its return type is the same as
22326 __builtin_choose_expr ( \
22327 __builtin_types_compatible_p (typeof (x), double), \
22329 __builtin_choose_expr ( \
22330 __builtin_types_compatible_p (typeof (x), float), \
22332 /* The void expression results in a compile-time error \
22333 when assigning the result to something. */ \
22336 _Note:_ This construct is only available for C. Furthermore, the
22337 unused expression (EXP1 or EXP2 depending on the value of
22338 CONST_EXP) may still generate syntax errors. This may change in
22342 -- Built-in Function: int __builtin_constant_p (EXP)
22343 You can use the built-in function `__builtin_constant_p' to
22344 determine if a value is known to be constant at compile-time and
22345 hence that GCC can perform constant-folding on expressions
22346 involving that value. The argument of the function is the value
22347 to test. The function returns the integer 1 if the argument is
22348 known to be a compile-time constant and 0 if it is not known to be
22349 a compile-time constant. A return of 0 does not indicate that the
22350 value is _not_ a constant, but merely that GCC cannot prove it is
22351 a constant with the specified value of the `-O' option.
22353 You would typically use this function in an embedded application
22354 where memory was a critical resource. If you have some complex
22355 calculation, you may want it to be folded if it involves
22356 constants, but need to call a function if it does not. For
22359 #define Scale_Value(X) \
22360 (__builtin_constant_p (X) \
22361 ? ((X) * SCALE + OFFSET) : Scale (X))
22363 You may use this built-in function in either a macro or an inline
22364 function. However, if you use it in an inlined function and pass
22365 an argument of the function as the argument to the built-in, GCC
22366 will never return 1 when you call the inline function with a
22367 string constant or compound literal (*note Compound Literals::)
22368 and will not return 1 when you pass a constant numeric value to
22369 the inline function unless you specify the `-O' option.
22371 You may also use `__builtin_constant_p' in initializers for static
22372 data. For instance, you can write
22374 static const int table[] = {
22375 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
22379 This is an acceptable initializer even if EXPRESSION is not a
22380 constant expression. GCC must be more conservative about
22381 evaluating the built-in in this case, because it has no
22382 opportunity to perform optimization.
22384 Previous versions of GCC did not accept this built-in in data
22385 initializers. The earliest version where it is completely safe is
22388 -- Built-in Function: long __builtin_expect (long EXP, long C)
22389 You may use `__builtin_expect' to provide the compiler with branch
22390 prediction information. In general, you should prefer to use
22391 actual profile feedback for this (`-fprofile-arcs'), as
22392 programmers are notoriously bad at predicting how their programs
22393 actually perform. However, there are applications in which this
22394 data is hard to collect.
22396 The return value is the value of EXP, which should be an integral
22397 expression. The semantics of the built-in are that it is expected
22398 that EXP == C. For example:
22400 if (__builtin_expect (x, 0))
22403 would indicate that we do not expect to call `foo', since we
22404 expect `x' to be zero. Since you are limited to integral
22405 expressions for EXP, you should use constructions such as
22407 if (__builtin_expect (ptr != NULL, 1))
22410 when testing pointer or floating-point values.
22412 -- Built-in Function: void __builtin_trap (void)
22413 This function causes the program to exit abnormally. GCC
22414 implements this function by using a target-dependent mechanism
22415 (such as intentionally executing an illegal instruction) or by
22416 calling `abort'. The mechanism used may vary from release to
22417 release so you should not rely on any particular implementation.
22419 -- Built-in Function: void __builtin___clear_cache (char *BEGIN, char
22421 This function is used to flush the processor's instruction cache
22422 for the region of memory between BEGIN inclusive and END
22423 exclusive. Some targets require that the instruction cache be
22424 flushed, after modifying memory containing code, in order to obtain
22425 deterministic behavior.
22427 If the target does not require instruction cache flushes,
22428 `__builtin___clear_cache' has no effect. Otherwise either
22429 instructions are emitted in-line to clear the instruction cache or
22430 a call to the `__clear_cache' function in libgcc is made.
22432 -- Built-in Function: void __builtin_prefetch (const void *ADDR, ...)
22433 This function is used to minimize cache-miss latency by moving
22434 data into a cache before it is accessed. You can insert calls to
22435 `__builtin_prefetch' into code for which you know addresses of
22436 data in memory that is likely to be accessed soon. If the target
22437 supports them, data prefetch instructions will be generated. If
22438 the prefetch is done early enough before the access then the data
22439 will be in the cache by the time it is accessed.
22441 The value of ADDR is the address of the memory to prefetch. There
22442 are two optional arguments, RW and LOCALITY. The value of RW is a
22443 compile-time constant one or zero; one means that the prefetch is
22444 preparing for a write to the memory address and zero, the default,
22445 means that the prefetch is preparing for a read. The value
22446 LOCALITY must be a compile-time constant integer between zero and
22447 three. A value of zero means that the data has no temporal
22448 locality, so it need not be left in the cache after the access. A
22449 value of three means that the data has a high degree of temporal
22450 locality and should be left in all levels of cache possible.
22451 Values of one and two mean, respectively, a low or moderate degree
22452 of temporal locality. The default is three.
22454 for (i = 0; i < n; i++)
22456 a[i] = a[i] + b[i];
22457 __builtin_prefetch (&a[i+j], 1, 1);
22458 __builtin_prefetch (&b[i+j], 0, 1);
22462 Data prefetch does not generate faults if ADDR is invalid, but the
22463 address expression itself must be valid. For example, a prefetch
22464 of `p->next' will not fault if `p->next' is not a valid address,
22465 but evaluation will fault if `p' is not a valid address.
22467 If the target does not support data prefetch, the address
22468 expression is evaluated if it includes side effects but no other
22469 code is generated and GCC does not issue a warning.
22471 -- Built-in Function: double __builtin_huge_val (void)
22472 Returns a positive infinity, if supported by the floating-point
22473 format, else `DBL_MAX'. This function is suitable for
22474 implementing the ISO C macro `HUGE_VAL'.
22476 -- Built-in Function: float __builtin_huge_valf (void)
22477 Similar to `__builtin_huge_val', except the return type is `float'.
22479 -- Built-in Function: long double __builtin_huge_vall (void)
22480 Similar to `__builtin_huge_val', except the return type is `long
22483 -- Built-in Function: double __builtin_inf (void)
22484 Similar to `__builtin_huge_val', except a warning is generated if
22485 the target floating-point format does not support infinities.
22487 -- Built-in Function: _Decimal32 __builtin_infd32 (void)
22488 Similar to `__builtin_inf', except the return type is `_Decimal32'.
22490 -- Built-in Function: _Decimal64 __builtin_infd64 (void)
22491 Similar to `__builtin_inf', except the return type is `_Decimal64'.
22493 -- Built-in Function: _Decimal128 __builtin_infd128 (void)
22494 Similar to `__builtin_inf', except the return type is
22497 -- Built-in Function: float __builtin_inff (void)
22498 Similar to `__builtin_inf', except the return type is `float'.
22499 This function is suitable for implementing the ISO C99 macro
22502 -- Built-in Function: long double __builtin_infl (void)
22503 Similar to `__builtin_inf', except the return type is `long
22506 -- Built-in Function: double __builtin_nan (const char *str)
22507 This is an implementation of the ISO C99 function `nan'.
22509 Since ISO C99 defines this function in terms of `strtod', which we
22510 do not implement, a description of the parsing is in order. The
22511 string is parsed as by `strtol'; that is, the base is recognized by
22512 leading `0' or `0x' prefixes. The number parsed is placed in the
22513 significand such that the least significant bit of the number is
22514 at the least significant bit of the significand. The number is
22515 truncated to fit the significand field provided. The significand
22516 is forced to be a quiet NaN.
22518 This function, if given a string literal all of which would have
22519 been consumed by strtol, is evaluated early enough that it is
22520 considered a compile-time constant.
22522 -- Built-in Function: _Decimal32 __builtin_nand32 (const char *str)
22523 Similar to `__builtin_nan', except the return type is `_Decimal32'.
22525 -- Built-in Function: _Decimal64 __builtin_nand64 (const char *str)
22526 Similar to `__builtin_nan', except the return type is `_Decimal64'.
22528 -- Built-in Function: _Decimal128 __builtin_nand128 (const char *str)
22529 Similar to `__builtin_nan', except the return type is
22532 -- Built-in Function: float __builtin_nanf (const char *str)
22533 Similar to `__builtin_nan', except the return type is `float'.
22535 -- Built-in Function: long double __builtin_nanl (const char *str)
22536 Similar to `__builtin_nan', except the return type is `long
22539 -- Built-in Function: double __builtin_nans (const char *str)
22540 Similar to `__builtin_nan', except the significand is forced to be
22541 a signaling NaN. The `nans' function is proposed by WG14 N965.
22543 -- Built-in Function: float __builtin_nansf (const char *str)
22544 Similar to `__builtin_nans', except the return type is `float'.
22546 -- Built-in Function: long double __builtin_nansl (const char *str)
22547 Similar to `__builtin_nans', except the return type is `long
22550 -- Built-in Function: int __builtin_ffs (unsigned int x)
22551 Returns one plus the index of the least significant 1-bit of X, or
22552 if X is zero, returns zero.
22554 -- Built-in Function: int __builtin_clz (unsigned int x)
22555 Returns the number of leading 0-bits in X, starting at the most
22556 significant bit position. If X is 0, the result is undefined.
22558 -- Built-in Function: int __builtin_ctz (unsigned int x)
22559 Returns the number of trailing 0-bits in X, starting at the least
22560 significant bit position. If X is 0, the result is undefined.
22562 -- Built-in Function: int __builtin_popcount (unsigned int x)
22563 Returns the number of 1-bits in X.
22565 -- Built-in Function: int __builtin_parity (unsigned int x)
22566 Returns the parity of X, i.e. the number of 1-bits in X modulo 2.
22568 -- Built-in Function: int __builtin_ffsl (unsigned long)
22569 Similar to `__builtin_ffs', except the argument type is `unsigned
22572 -- Built-in Function: int __builtin_clzl (unsigned long)
22573 Similar to `__builtin_clz', except the argument type is `unsigned
22576 -- Built-in Function: int __builtin_ctzl (unsigned long)
22577 Similar to `__builtin_ctz', except the argument type is `unsigned
22580 -- Built-in Function: int __builtin_popcountl (unsigned long)
22581 Similar to `__builtin_popcount', except the argument type is
22584 -- Built-in Function: int __builtin_parityl (unsigned long)
22585 Similar to `__builtin_parity', except the argument type is
22588 -- Built-in Function: int __builtin_ffsll (unsigned long long)
22589 Similar to `__builtin_ffs', except the argument type is `unsigned
22592 -- Built-in Function: int __builtin_clzll (unsigned long long)
22593 Similar to `__builtin_clz', except the argument type is `unsigned
22596 -- Built-in Function: int __builtin_ctzll (unsigned long long)
22597 Similar to `__builtin_ctz', except the argument type is `unsigned
22600 -- Built-in Function: int __builtin_popcountll (unsigned long long)
22601 Similar to `__builtin_popcount', except the argument type is
22602 `unsigned long long'.
22604 -- Built-in Function: int __builtin_parityll (unsigned long long)
22605 Similar to `__builtin_parity', except the argument type is
22606 `unsigned long long'.
22608 -- Built-in Function: double __builtin_powi (double, int)
22609 Returns the first argument raised to the power of the second.
22610 Unlike the `pow' function no guarantees about precision and
22613 -- Built-in Function: float __builtin_powif (float, int)
22614 Similar to `__builtin_powi', except the argument and return types
22617 -- Built-in Function: long double __builtin_powil (long double, int)
22618 Similar to `__builtin_powi', except the argument and return types
22621 -- Built-in Function: int32_t __builtin_bswap32 (int32_t x)
22622 Returns X with the order of the bytes reversed; for example,
22623 `0xaabbccdd' becomes `0xddccbbaa'. Byte here always means exactly
22626 -- Built-in Function: int64_t __builtin_bswap64 (int64_t x)
22627 Similar to `__builtin_bswap32', except the argument and return
22631 File: gcc.info, Node: Target Builtins, Next: Target Format Checks, Prev: Other Builtins, Up: C Extensions
22633 5.50 Built-in Functions Specific to Particular Target Machines
22634 ==============================================================
22636 On some target machines, GCC supports many built-in functions specific
22637 to those machines. Generally these generate calls to specific machine
22638 instructions, but allow the compiler to schedule those calls.
22642 * Alpha Built-in Functions::
22643 * ARM iWMMXt Built-in Functions::
22644 * ARM NEON Intrinsics::
22645 * Blackfin Built-in Functions::
22646 * FR-V Built-in Functions::
22647 * X86 Built-in Functions::
22648 * MIPS DSP Built-in Functions::
22649 * MIPS Paired-Single Support::
22650 * PowerPC AltiVec Built-in Functions::
22651 * SPARC VIS Built-in Functions::
22652 * SPU Built-in Functions::
22655 File: gcc.info, Node: Alpha Built-in Functions, Next: ARM iWMMXt Built-in Functions, Up: Target Builtins
22657 5.50.1 Alpha Built-in Functions
22658 -------------------------------
22660 These built-in functions are available for the Alpha family of
22661 processors, depending on the command-line switches used.
22663 The following built-in functions are always available. They all
22664 generate the machine instruction that is part of the name.
22666 long __builtin_alpha_implver (void)
22667 long __builtin_alpha_rpcc (void)
22668 long __builtin_alpha_amask (long)
22669 long __builtin_alpha_cmpbge (long, long)
22670 long __builtin_alpha_extbl (long, long)
22671 long __builtin_alpha_extwl (long, long)
22672 long __builtin_alpha_extll (long, long)
22673 long __builtin_alpha_extql (long, long)
22674 long __builtin_alpha_extwh (long, long)
22675 long __builtin_alpha_extlh (long, long)
22676 long __builtin_alpha_extqh (long, long)
22677 long __builtin_alpha_insbl (long, long)
22678 long __builtin_alpha_inswl (long, long)
22679 long __builtin_alpha_insll (long, long)
22680 long __builtin_alpha_insql (long, long)
22681 long __builtin_alpha_inswh (long, long)
22682 long __builtin_alpha_inslh (long, long)
22683 long __builtin_alpha_insqh (long, long)
22684 long __builtin_alpha_mskbl (long, long)
22685 long __builtin_alpha_mskwl (long, long)
22686 long __builtin_alpha_mskll (long, long)
22687 long __builtin_alpha_mskql (long, long)
22688 long __builtin_alpha_mskwh (long, long)
22689 long __builtin_alpha_msklh (long, long)
22690 long __builtin_alpha_mskqh (long, long)
22691 long __builtin_alpha_umulh (long, long)
22692 long __builtin_alpha_zap (long, long)
22693 long __builtin_alpha_zapnot (long, long)
22695 The following built-in functions are always with `-mmax' or
22696 `-mcpu=CPU' where CPU is `pca56' or later. They all generate the
22697 machine instruction that is part of the name.
22699 long __builtin_alpha_pklb (long)
22700 long __builtin_alpha_pkwb (long)
22701 long __builtin_alpha_unpkbl (long)
22702 long __builtin_alpha_unpkbw (long)
22703 long __builtin_alpha_minub8 (long, long)
22704 long __builtin_alpha_minsb8 (long, long)
22705 long __builtin_alpha_minuw4 (long, long)
22706 long __builtin_alpha_minsw4 (long, long)
22707 long __builtin_alpha_maxub8 (long, long)
22708 long __builtin_alpha_maxsb8 (long, long)
22709 long __builtin_alpha_maxuw4 (long, long)
22710 long __builtin_alpha_maxsw4 (long, long)
22711 long __builtin_alpha_perr (long, long)
22713 The following built-in functions are always with `-mcix' or
22714 `-mcpu=CPU' where CPU is `ev67' or later. They all generate the
22715 machine instruction that is part of the name.
22717 long __builtin_alpha_cttz (long)
22718 long __builtin_alpha_ctlz (long)
22719 long __builtin_alpha_ctpop (long)
22721 The following builtins are available on systems that use the OSF/1
22722 PALcode. Normally they invoke the `rduniq' and `wruniq' PAL calls, but
22723 when invoked with `-mtls-kernel', they invoke `rdval' and `wrval'.
22725 void *__builtin_thread_pointer (void)
22726 void __builtin_set_thread_pointer (void *)
22729 File: gcc.info, Node: ARM iWMMXt Built-in Functions, Next: ARM NEON Intrinsics, Prev: Alpha Built-in Functions, Up: Target Builtins
22731 5.50.2 ARM iWMMXt Built-in Functions
22732 ------------------------------------
22734 These built-in functions are available for the ARM family of processors
22735 when the `-mcpu=iwmmxt' switch is used:
22737 typedef int v2si __attribute__ ((vector_size (8)));
22738 typedef short v4hi __attribute__ ((vector_size (8)));
22739 typedef char v8qi __attribute__ ((vector_size (8)));
22741 int __builtin_arm_getwcx (int)
22742 void __builtin_arm_setwcx (int, int)
22743 int __builtin_arm_textrmsb (v8qi, int)
22744 int __builtin_arm_textrmsh (v4hi, int)
22745 int __builtin_arm_textrmsw (v2si, int)
22746 int __builtin_arm_textrmub (v8qi, int)
22747 int __builtin_arm_textrmuh (v4hi, int)
22748 int __builtin_arm_textrmuw (v2si, int)
22749 v8qi __builtin_arm_tinsrb (v8qi, int)
22750 v4hi __builtin_arm_tinsrh (v4hi, int)
22751 v2si __builtin_arm_tinsrw (v2si, int)
22752 long long __builtin_arm_tmia (long long, int, int)
22753 long long __builtin_arm_tmiabb (long long, int, int)
22754 long long __builtin_arm_tmiabt (long long, int, int)
22755 long long __builtin_arm_tmiaph (long long, int, int)
22756 long long __builtin_arm_tmiatb (long long, int, int)
22757 long long __builtin_arm_tmiatt (long long, int, int)
22758 int __builtin_arm_tmovmskb (v8qi)
22759 int __builtin_arm_tmovmskh (v4hi)
22760 int __builtin_arm_tmovmskw (v2si)
22761 long long __builtin_arm_waccb (v8qi)
22762 long long __builtin_arm_wacch (v4hi)
22763 long long __builtin_arm_waccw (v2si)
22764 v8qi __builtin_arm_waddb (v8qi, v8qi)
22765 v8qi __builtin_arm_waddbss (v8qi, v8qi)
22766 v8qi __builtin_arm_waddbus (v8qi, v8qi)
22767 v4hi __builtin_arm_waddh (v4hi, v4hi)
22768 v4hi __builtin_arm_waddhss (v4hi, v4hi)
22769 v4hi __builtin_arm_waddhus (v4hi, v4hi)
22770 v2si __builtin_arm_waddw (v2si, v2si)
22771 v2si __builtin_arm_waddwss (v2si, v2si)
22772 v2si __builtin_arm_waddwus (v2si, v2si)
22773 v8qi __builtin_arm_walign (v8qi, v8qi, int)
22774 long long __builtin_arm_wand(long long, long long)
22775 long long __builtin_arm_wandn (long long, long long)
22776 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
22777 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
22778 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
22779 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
22780 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
22781 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
22782 v2si __builtin_arm_wcmpeqw (v2si, v2si)
22783 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
22784 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
22785 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
22786 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
22787 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
22788 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
22789 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
22790 long long __builtin_arm_wmacsz (v4hi, v4hi)
22791 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
22792 long long __builtin_arm_wmacuz (v4hi, v4hi)
22793 v4hi __builtin_arm_wmadds (v4hi, v4hi)
22794 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
22795 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
22796 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
22797 v2si __builtin_arm_wmaxsw (v2si, v2si)
22798 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
22799 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
22800 v2si __builtin_arm_wmaxuw (v2si, v2si)
22801 v8qi __builtin_arm_wminsb (v8qi, v8qi)
22802 v4hi __builtin_arm_wminsh (v4hi, v4hi)
22803 v2si __builtin_arm_wminsw (v2si, v2si)
22804 v8qi __builtin_arm_wminub (v8qi, v8qi)
22805 v4hi __builtin_arm_wminuh (v4hi, v4hi)
22806 v2si __builtin_arm_wminuw (v2si, v2si)
22807 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
22808 v4hi __builtin_arm_wmulul (v4hi, v4hi)
22809 v4hi __builtin_arm_wmulum (v4hi, v4hi)
22810 long long __builtin_arm_wor (long long, long long)
22811 v2si __builtin_arm_wpackdss (long long, long long)
22812 v2si __builtin_arm_wpackdus (long long, long long)
22813 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
22814 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
22815 v4hi __builtin_arm_wpackwss (v2si, v2si)
22816 v4hi __builtin_arm_wpackwus (v2si, v2si)
22817 long long __builtin_arm_wrord (long long, long long)
22818 long long __builtin_arm_wrordi (long long, int)
22819 v4hi __builtin_arm_wrorh (v4hi, long long)
22820 v4hi __builtin_arm_wrorhi (v4hi, int)
22821 v2si __builtin_arm_wrorw (v2si, long long)
22822 v2si __builtin_arm_wrorwi (v2si, int)
22823 v2si __builtin_arm_wsadb (v8qi, v8qi)
22824 v2si __builtin_arm_wsadbz (v8qi, v8qi)
22825 v2si __builtin_arm_wsadh (v4hi, v4hi)
22826 v2si __builtin_arm_wsadhz (v4hi, v4hi)
22827 v4hi __builtin_arm_wshufh (v4hi, int)
22828 long long __builtin_arm_wslld (long long, long long)
22829 long long __builtin_arm_wslldi (long long, int)
22830 v4hi __builtin_arm_wsllh (v4hi, long long)
22831 v4hi __builtin_arm_wsllhi (v4hi, int)
22832 v2si __builtin_arm_wsllw (v2si, long long)
22833 v2si __builtin_arm_wsllwi (v2si, int)
22834 long long __builtin_arm_wsrad (long long, long long)
22835 long long __builtin_arm_wsradi (long long, int)
22836 v4hi __builtin_arm_wsrah (v4hi, long long)
22837 v4hi __builtin_arm_wsrahi (v4hi, int)
22838 v2si __builtin_arm_wsraw (v2si, long long)
22839 v2si __builtin_arm_wsrawi (v2si, int)
22840 long long __builtin_arm_wsrld (long long, long long)
22841 long long __builtin_arm_wsrldi (long long, int)
22842 v4hi __builtin_arm_wsrlh (v4hi, long long)
22843 v4hi __builtin_arm_wsrlhi (v4hi, int)
22844 v2si __builtin_arm_wsrlw (v2si, long long)
22845 v2si __builtin_arm_wsrlwi (v2si, int)
22846 v8qi __builtin_arm_wsubb (v8qi, v8qi)
22847 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
22848 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
22849 v4hi __builtin_arm_wsubh (v4hi, v4hi)
22850 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
22851 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
22852 v2si __builtin_arm_wsubw (v2si, v2si)
22853 v2si __builtin_arm_wsubwss (v2si, v2si)
22854 v2si __builtin_arm_wsubwus (v2si, v2si)
22855 v4hi __builtin_arm_wunpckehsb (v8qi)
22856 v2si __builtin_arm_wunpckehsh (v4hi)
22857 long long __builtin_arm_wunpckehsw (v2si)
22858 v4hi __builtin_arm_wunpckehub (v8qi)
22859 v2si __builtin_arm_wunpckehuh (v4hi)
22860 long long __builtin_arm_wunpckehuw (v2si)
22861 v4hi __builtin_arm_wunpckelsb (v8qi)
22862 v2si __builtin_arm_wunpckelsh (v4hi)
22863 long long __builtin_arm_wunpckelsw (v2si)
22864 v4hi __builtin_arm_wunpckelub (v8qi)
22865 v2si __builtin_arm_wunpckeluh (v4hi)
22866 long long __builtin_arm_wunpckeluw (v2si)
22867 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
22868 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
22869 v2si __builtin_arm_wunpckihw (v2si, v2si)
22870 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
22871 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
22872 v2si __builtin_arm_wunpckilw (v2si, v2si)
22873 long long __builtin_arm_wxor (long long, long long)
22874 long long __builtin_arm_wzero ()
22877 File: gcc.info, Node: ARM NEON Intrinsics, Next: Blackfin Built-in Functions, Prev: ARM iWMMXt Built-in Functions, Up: Target Builtins
22879 5.50.3 ARM NEON Intrinsics
22880 --------------------------
22882 These built-in intrinsics for the ARM Advanced SIMD extension are
22883 available when the `-mfpu=neon' switch is used:
22888 * uint32x2_t vadd_u32 (uint32x2_t, uint32x2_t)
22889 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
22891 * uint16x4_t vadd_u16 (uint16x4_t, uint16x4_t)
22892 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
22894 * uint8x8_t vadd_u8 (uint8x8_t, uint8x8_t)
22895 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
22897 * int32x2_t vadd_s32 (int32x2_t, int32x2_t)
22898 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
22900 * int16x4_t vadd_s16 (int16x4_t, int16x4_t)
22901 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
22903 * int8x8_t vadd_s8 (int8x8_t, int8x8_t)
22904 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
22906 * uint64x1_t vadd_u64 (uint64x1_t, uint64x1_t)
22907 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
22909 * int64x1_t vadd_s64 (int64x1_t, int64x1_t)
22910 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
22912 * float32x2_t vadd_f32 (float32x2_t, float32x2_t)
22913 _Form of expected instruction(s):_ `vadd.f32 D0, D0, D0'
22915 * uint32x4_t vaddq_u32 (uint32x4_t, uint32x4_t)
22916 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
22918 * uint16x8_t vaddq_u16 (uint16x8_t, uint16x8_t)
22919 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
22921 * uint8x16_t vaddq_u8 (uint8x16_t, uint8x16_t)
22922 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
22924 * int32x4_t vaddq_s32 (int32x4_t, int32x4_t)
22925 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
22927 * int16x8_t vaddq_s16 (int16x8_t, int16x8_t)
22928 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
22930 * int8x16_t vaddq_s8 (int8x16_t, int8x16_t)
22931 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
22933 * uint64x2_t vaddq_u64 (uint64x2_t, uint64x2_t)
22934 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
22936 * int64x2_t vaddq_s64 (int64x2_t, int64x2_t)
22937 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
22939 * float32x4_t vaddq_f32 (float32x4_t, float32x4_t)
22940 _Form of expected instruction(s):_ `vadd.f32 Q0, Q0, Q0'
22942 * uint64x2_t vaddl_u32 (uint32x2_t, uint32x2_t)
22943 _Form of expected instruction(s):_ `vaddl.u32 Q0, D0, D0'
22945 * uint32x4_t vaddl_u16 (uint16x4_t, uint16x4_t)
22946 _Form of expected instruction(s):_ `vaddl.u16 Q0, D0, D0'
22948 * uint16x8_t vaddl_u8 (uint8x8_t, uint8x8_t)
22949 _Form of expected instruction(s):_ `vaddl.u8 Q0, D0, D0'
22951 * int64x2_t vaddl_s32 (int32x2_t, int32x2_t)
22952 _Form of expected instruction(s):_ `vaddl.s32 Q0, D0, D0'
22954 * int32x4_t vaddl_s16 (int16x4_t, int16x4_t)
22955 _Form of expected instruction(s):_ `vaddl.s16 Q0, D0, D0'
22957 * int16x8_t vaddl_s8 (int8x8_t, int8x8_t)
22958 _Form of expected instruction(s):_ `vaddl.s8 Q0, D0, D0'
22960 * uint64x2_t vaddw_u32 (uint64x2_t, uint32x2_t)
22961 _Form of expected instruction(s):_ `vaddw.u32 Q0, Q0, D0'
22963 * uint32x4_t vaddw_u16 (uint32x4_t, uint16x4_t)
22964 _Form of expected instruction(s):_ `vaddw.u16 Q0, Q0, D0'
22966 * uint16x8_t vaddw_u8 (uint16x8_t, uint8x8_t)
22967 _Form of expected instruction(s):_ `vaddw.u8 Q0, Q0, D0'
22969 * int64x2_t vaddw_s32 (int64x2_t, int32x2_t)
22970 _Form of expected instruction(s):_ `vaddw.s32 Q0, Q0, D0'
22972 * int32x4_t vaddw_s16 (int32x4_t, int16x4_t)
22973 _Form of expected instruction(s):_ `vaddw.s16 Q0, Q0, D0'
22975 * int16x8_t vaddw_s8 (int16x8_t, int8x8_t)
22976 _Form of expected instruction(s):_ `vaddw.s8 Q0, Q0, D0'
22978 * uint32x2_t vhadd_u32 (uint32x2_t, uint32x2_t)
22979 _Form of expected instruction(s):_ `vhadd.u32 D0, D0, D0'
22981 * uint16x4_t vhadd_u16 (uint16x4_t, uint16x4_t)
22982 _Form of expected instruction(s):_ `vhadd.u16 D0, D0, D0'
22984 * uint8x8_t vhadd_u8 (uint8x8_t, uint8x8_t)
22985 _Form of expected instruction(s):_ `vhadd.u8 D0, D0, D0'
22987 * int32x2_t vhadd_s32 (int32x2_t, int32x2_t)
22988 _Form of expected instruction(s):_ `vhadd.s32 D0, D0, D0'
22990 * int16x4_t vhadd_s16 (int16x4_t, int16x4_t)
22991 _Form of expected instruction(s):_ `vhadd.s16 D0, D0, D0'
22993 * int8x8_t vhadd_s8 (int8x8_t, int8x8_t)
22994 _Form of expected instruction(s):_ `vhadd.s8 D0, D0, D0'
22996 * uint32x4_t vhaddq_u32 (uint32x4_t, uint32x4_t)
22997 _Form of expected instruction(s):_ `vhadd.u32 Q0, Q0, Q0'
22999 * uint16x8_t vhaddq_u16 (uint16x8_t, uint16x8_t)
23000 _Form of expected instruction(s):_ `vhadd.u16 Q0, Q0, Q0'
23002 * uint8x16_t vhaddq_u8 (uint8x16_t, uint8x16_t)
23003 _Form of expected instruction(s):_ `vhadd.u8 Q0, Q0, Q0'
23005 * int32x4_t vhaddq_s32 (int32x4_t, int32x4_t)
23006 _Form of expected instruction(s):_ `vhadd.s32 Q0, Q0, Q0'
23008 * int16x8_t vhaddq_s16 (int16x8_t, int16x8_t)
23009 _Form of expected instruction(s):_ `vhadd.s16 Q0, Q0, Q0'
23011 * int8x16_t vhaddq_s8 (int8x16_t, int8x16_t)
23012 _Form of expected instruction(s):_ `vhadd.s8 Q0, Q0, Q0'
23014 * uint32x2_t vrhadd_u32 (uint32x2_t, uint32x2_t)
23015 _Form of expected instruction(s):_ `vrhadd.u32 D0, D0, D0'
23017 * uint16x4_t vrhadd_u16 (uint16x4_t, uint16x4_t)
23018 _Form of expected instruction(s):_ `vrhadd.u16 D0, D0, D0'
23020 * uint8x8_t vrhadd_u8 (uint8x8_t, uint8x8_t)
23021 _Form of expected instruction(s):_ `vrhadd.u8 D0, D0, D0'
23023 * int32x2_t vrhadd_s32 (int32x2_t, int32x2_t)
23024 _Form of expected instruction(s):_ `vrhadd.s32 D0, D0, D0'
23026 * int16x4_t vrhadd_s16 (int16x4_t, int16x4_t)
23027 _Form of expected instruction(s):_ `vrhadd.s16 D0, D0, D0'
23029 * int8x8_t vrhadd_s8 (int8x8_t, int8x8_t)
23030 _Form of expected instruction(s):_ `vrhadd.s8 D0, D0, D0'
23032 * uint32x4_t vrhaddq_u32 (uint32x4_t, uint32x4_t)
23033 _Form of expected instruction(s):_ `vrhadd.u32 Q0, Q0, Q0'
23035 * uint16x8_t vrhaddq_u16 (uint16x8_t, uint16x8_t)
23036 _Form of expected instruction(s):_ `vrhadd.u16 Q0, Q0, Q0'
23038 * uint8x16_t vrhaddq_u8 (uint8x16_t, uint8x16_t)
23039 _Form of expected instruction(s):_ `vrhadd.u8 Q0, Q0, Q0'
23041 * int32x4_t vrhaddq_s32 (int32x4_t, int32x4_t)
23042 _Form of expected instruction(s):_ `vrhadd.s32 Q0, Q0, Q0'
23044 * int16x8_t vrhaddq_s16 (int16x8_t, int16x8_t)
23045 _Form of expected instruction(s):_ `vrhadd.s16 Q0, Q0, Q0'
23047 * int8x16_t vrhaddq_s8 (int8x16_t, int8x16_t)
23048 _Form of expected instruction(s):_ `vrhadd.s8 Q0, Q0, Q0'
23050 * uint32x2_t vqadd_u32 (uint32x2_t, uint32x2_t)
23051 _Form of expected instruction(s):_ `vqadd.u32 D0, D0, D0'
23053 * uint16x4_t vqadd_u16 (uint16x4_t, uint16x4_t)
23054 _Form of expected instruction(s):_ `vqadd.u16 D0, D0, D0'
23056 * uint8x8_t vqadd_u8 (uint8x8_t, uint8x8_t)
23057 _Form of expected instruction(s):_ `vqadd.u8 D0, D0, D0'
23059 * int32x2_t vqadd_s32 (int32x2_t, int32x2_t)
23060 _Form of expected instruction(s):_ `vqadd.s32 D0, D0, D0'
23062 * int16x4_t vqadd_s16 (int16x4_t, int16x4_t)
23063 _Form of expected instruction(s):_ `vqadd.s16 D0, D0, D0'
23065 * int8x8_t vqadd_s8 (int8x8_t, int8x8_t)
23066 _Form of expected instruction(s):_ `vqadd.s8 D0, D0, D0'
23068 * uint64x1_t vqadd_u64 (uint64x1_t, uint64x1_t)
23069 _Form of expected instruction(s):_ `vqadd.u64 D0, D0, D0'
23071 * int64x1_t vqadd_s64 (int64x1_t, int64x1_t)
23072 _Form of expected instruction(s):_ `vqadd.s64 D0, D0, D0'
23074 * uint32x4_t vqaddq_u32 (uint32x4_t, uint32x4_t)
23075 _Form of expected instruction(s):_ `vqadd.u32 Q0, Q0, Q0'
23077 * uint16x8_t vqaddq_u16 (uint16x8_t, uint16x8_t)
23078 _Form of expected instruction(s):_ `vqadd.u16 Q0, Q0, Q0'
23080 * uint8x16_t vqaddq_u8 (uint8x16_t, uint8x16_t)
23081 _Form of expected instruction(s):_ `vqadd.u8 Q0, Q0, Q0'
23083 * int32x4_t vqaddq_s32 (int32x4_t, int32x4_t)
23084 _Form of expected instruction(s):_ `vqadd.s32 Q0, Q0, Q0'
23086 * int16x8_t vqaddq_s16 (int16x8_t, int16x8_t)
23087 _Form of expected instruction(s):_ `vqadd.s16 Q0, Q0, Q0'
23089 * int8x16_t vqaddq_s8 (int8x16_t, int8x16_t)
23090 _Form of expected instruction(s):_ `vqadd.s8 Q0, Q0, Q0'
23092 * uint64x2_t vqaddq_u64 (uint64x2_t, uint64x2_t)
23093 _Form of expected instruction(s):_ `vqadd.u64 Q0, Q0, Q0'
23095 * int64x2_t vqaddq_s64 (int64x2_t, int64x2_t)
23096 _Form of expected instruction(s):_ `vqadd.s64 Q0, Q0, Q0'
23098 * uint32x2_t vaddhn_u64 (uint64x2_t, uint64x2_t)
23099 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
23101 * uint16x4_t vaddhn_u32 (uint32x4_t, uint32x4_t)
23102 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
23104 * uint8x8_t vaddhn_u16 (uint16x8_t, uint16x8_t)
23105 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
23107 * int32x2_t vaddhn_s64 (int64x2_t, int64x2_t)
23108 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
23110 * int16x4_t vaddhn_s32 (int32x4_t, int32x4_t)
23111 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
23113 * int8x8_t vaddhn_s16 (int16x8_t, int16x8_t)
23114 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
23116 * uint32x2_t vraddhn_u64 (uint64x2_t, uint64x2_t)
23117 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
23119 * uint16x4_t vraddhn_u32 (uint32x4_t, uint32x4_t)
23120 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
23122 * uint8x8_t vraddhn_u16 (uint16x8_t, uint16x8_t)
23123 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
23125 * int32x2_t vraddhn_s64 (int64x2_t, int64x2_t)
23126 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
23128 * int16x4_t vraddhn_s32 (int32x4_t, int32x4_t)
23129 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
23131 * int8x8_t vraddhn_s16 (int16x8_t, int16x8_t)
23132 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
23134 5.50.3.2 Multiplication
23135 .......................
23137 * uint32x2_t vmul_u32 (uint32x2_t, uint32x2_t)
23138 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
23140 * uint16x4_t vmul_u16 (uint16x4_t, uint16x4_t)
23141 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
23143 * uint8x8_t vmul_u8 (uint8x8_t, uint8x8_t)
23144 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
23146 * int32x2_t vmul_s32 (int32x2_t, int32x2_t)
23147 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
23149 * int16x4_t vmul_s16 (int16x4_t, int16x4_t)
23150 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
23152 * int8x8_t vmul_s8 (int8x8_t, int8x8_t)
23153 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
23155 * float32x2_t vmul_f32 (float32x2_t, float32x2_t)
23156 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0'
23158 * poly8x8_t vmul_p8 (poly8x8_t, poly8x8_t)
23159 _Form of expected instruction(s):_ `vmul.p8 D0, D0, D0'
23161 * uint32x4_t vmulq_u32 (uint32x4_t, uint32x4_t)
23162 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
23164 * uint16x8_t vmulq_u16 (uint16x8_t, uint16x8_t)
23165 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
23167 * uint8x16_t vmulq_u8 (uint8x16_t, uint8x16_t)
23168 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
23170 * int32x4_t vmulq_s32 (int32x4_t, int32x4_t)
23171 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
23173 * int16x8_t vmulq_s16 (int16x8_t, int16x8_t)
23174 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
23176 * int8x16_t vmulq_s8 (int8x16_t, int8x16_t)
23177 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
23179 * float32x4_t vmulq_f32 (float32x4_t, float32x4_t)
23180 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, Q0'
23182 * poly8x16_t vmulq_p8 (poly8x16_t, poly8x16_t)
23183 _Form of expected instruction(s):_ `vmul.p8 Q0, Q0, Q0'
23185 * int32x2_t vqdmulh_s32 (int32x2_t, int32x2_t)
23186 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0'
23188 * int16x4_t vqdmulh_s16 (int16x4_t, int16x4_t)
23189 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0'
23191 * int32x4_t vqdmulhq_s32 (int32x4_t, int32x4_t)
23192 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, Q0'
23194 * int16x8_t vqdmulhq_s16 (int16x8_t, int16x8_t)
23195 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, Q0'
23197 * int32x2_t vqrdmulh_s32 (int32x2_t, int32x2_t)
23198 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0'
23200 * int16x4_t vqrdmulh_s16 (int16x4_t, int16x4_t)
23201 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0'
23203 * int32x4_t vqrdmulhq_s32 (int32x4_t, int32x4_t)
23204 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, Q0'
23206 * int16x8_t vqrdmulhq_s16 (int16x8_t, int16x8_t)
23207 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, Q0'
23209 * uint64x2_t vmull_u32 (uint32x2_t, uint32x2_t)
23210 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0'
23212 * uint32x4_t vmull_u16 (uint16x4_t, uint16x4_t)
23213 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0'
23215 * uint16x8_t vmull_u8 (uint8x8_t, uint8x8_t)
23216 _Form of expected instruction(s):_ `vmull.u8 Q0, D0, D0'
23218 * int64x2_t vmull_s32 (int32x2_t, int32x2_t)
23219 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0'
23221 * int32x4_t vmull_s16 (int16x4_t, int16x4_t)
23222 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0'
23224 * int16x8_t vmull_s8 (int8x8_t, int8x8_t)
23225 _Form of expected instruction(s):_ `vmull.s8 Q0, D0, D0'
23227 * poly16x8_t vmull_p8 (poly8x8_t, poly8x8_t)
23228 _Form of expected instruction(s):_ `vmull.p8 Q0, D0, D0'
23230 * int64x2_t vqdmull_s32 (int32x2_t, int32x2_t)
23231 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0'
23233 * int32x4_t vqdmull_s16 (int16x4_t, int16x4_t)
23234 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0'
23236 5.50.3.3 Multiply-accumulate
23237 ............................
23239 * uint32x2_t vmla_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
23240 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
23242 * uint16x4_t vmla_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
23243 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
23245 * uint8x8_t vmla_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
23246 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
23248 * int32x2_t vmla_s32 (int32x2_t, int32x2_t, int32x2_t)
23249 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
23251 * int16x4_t vmla_s16 (int16x4_t, int16x4_t, int16x4_t)
23252 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
23254 * int8x8_t vmla_s8 (int8x8_t, int8x8_t, int8x8_t)
23255 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
23257 * float32x2_t vmla_f32 (float32x2_t, float32x2_t, float32x2_t)
23258 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0'
23260 * uint32x4_t vmlaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
23261 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
23263 * uint16x8_t vmlaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
23264 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
23266 * uint8x16_t vmlaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
23267 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
23269 * int32x4_t vmlaq_s32 (int32x4_t, int32x4_t, int32x4_t)
23270 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
23272 * int16x8_t vmlaq_s16 (int16x8_t, int16x8_t, int16x8_t)
23273 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
23275 * int8x16_t vmlaq_s8 (int8x16_t, int8x16_t, int8x16_t)
23276 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
23278 * float32x4_t vmlaq_f32 (float32x4_t, float32x4_t, float32x4_t)
23279 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, Q0'
23281 * uint64x2_t vmlal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
23282 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0'
23284 * uint32x4_t vmlal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
23285 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0'
23287 * uint16x8_t vmlal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
23288 _Form of expected instruction(s):_ `vmlal.u8 Q0, D0, D0'
23290 * int64x2_t vmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
23291 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0'
23293 * int32x4_t vmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
23294 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0'
23296 * int16x8_t vmlal_s8 (int16x8_t, int8x8_t, int8x8_t)
23297 _Form of expected instruction(s):_ `vmlal.s8 Q0, D0, D0'
23299 * int64x2_t vqdmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
23300 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0'
23302 * int32x4_t vqdmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
23303 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0'
23305 5.50.3.4 Multiply-subtract
23306 ..........................
23308 * uint32x2_t vmls_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
23309 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
23311 * uint16x4_t vmls_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
23312 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
23314 * uint8x8_t vmls_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
23315 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
23317 * int32x2_t vmls_s32 (int32x2_t, int32x2_t, int32x2_t)
23318 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
23320 * int16x4_t vmls_s16 (int16x4_t, int16x4_t, int16x4_t)
23321 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
23323 * int8x8_t vmls_s8 (int8x8_t, int8x8_t, int8x8_t)
23324 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
23326 * float32x2_t vmls_f32 (float32x2_t, float32x2_t, float32x2_t)
23327 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0'
23329 * uint32x4_t vmlsq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
23330 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
23332 * uint16x8_t vmlsq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
23333 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
23335 * uint8x16_t vmlsq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
23336 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
23338 * int32x4_t vmlsq_s32 (int32x4_t, int32x4_t, int32x4_t)
23339 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
23341 * int16x8_t vmlsq_s16 (int16x8_t, int16x8_t, int16x8_t)
23342 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
23344 * int8x16_t vmlsq_s8 (int8x16_t, int8x16_t, int8x16_t)
23345 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
23347 * float32x4_t vmlsq_f32 (float32x4_t, float32x4_t, float32x4_t)
23348 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, Q0'
23350 * uint64x2_t vmlsl_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
23351 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0'
23353 * uint32x4_t vmlsl_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
23354 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0'
23356 * uint16x8_t vmlsl_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
23357 _Form of expected instruction(s):_ `vmlsl.u8 Q0, D0, D0'
23359 * int64x2_t vmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
23360 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0'
23362 * int32x4_t vmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
23363 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0'
23365 * int16x8_t vmlsl_s8 (int16x8_t, int8x8_t, int8x8_t)
23366 _Form of expected instruction(s):_ `vmlsl.s8 Q0, D0, D0'
23368 * int64x2_t vqdmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
23369 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0'
23371 * int32x4_t vqdmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
23372 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0'
23374 5.50.3.5 Subtraction
23375 ....................
23377 * uint32x2_t vsub_u32 (uint32x2_t, uint32x2_t)
23378 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
23380 * uint16x4_t vsub_u16 (uint16x4_t, uint16x4_t)
23381 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
23383 * uint8x8_t vsub_u8 (uint8x8_t, uint8x8_t)
23384 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
23386 * int32x2_t vsub_s32 (int32x2_t, int32x2_t)
23387 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
23389 * int16x4_t vsub_s16 (int16x4_t, int16x4_t)
23390 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
23392 * int8x8_t vsub_s8 (int8x8_t, int8x8_t)
23393 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
23395 * uint64x1_t vsub_u64 (uint64x1_t, uint64x1_t)
23396 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
23398 * int64x1_t vsub_s64 (int64x1_t, int64x1_t)
23399 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
23401 * float32x2_t vsub_f32 (float32x2_t, float32x2_t)
23402 _Form of expected instruction(s):_ `vsub.f32 D0, D0, D0'
23404 * uint32x4_t vsubq_u32 (uint32x4_t, uint32x4_t)
23405 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
23407 * uint16x8_t vsubq_u16 (uint16x8_t, uint16x8_t)
23408 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
23410 * uint8x16_t vsubq_u8 (uint8x16_t, uint8x16_t)
23411 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
23413 * int32x4_t vsubq_s32 (int32x4_t, int32x4_t)
23414 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
23416 * int16x8_t vsubq_s16 (int16x8_t, int16x8_t)
23417 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
23419 * int8x16_t vsubq_s8 (int8x16_t, int8x16_t)
23420 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
23422 * uint64x2_t vsubq_u64 (uint64x2_t, uint64x2_t)
23423 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
23425 * int64x2_t vsubq_s64 (int64x2_t, int64x2_t)
23426 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
23428 * float32x4_t vsubq_f32 (float32x4_t, float32x4_t)
23429 _Form of expected instruction(s):_ `vsub.f32 Q0, Q0, Q0'
23431 * uint64x2_t vsubl_u32 (uint32x2_t, uint32x2_t)
23432 _Form of expected instruction(s):_ `vsubl.u32 Q0, D0, D0'
23434 * uint32x4_t vsubl_u16 (uint16x4_t, uint16x4_t)
23435 _Form of expected instruction(s):_ `vsubl.u16 Q0, D0, D0'
23437 * uint16x8_t vsubl_u8 (uint8x8_t, uint8x8_t)
23438 _Form of expected instruction(s):_ `vsubl.u8 Q0, D0, D0'
23440 * int64x2_t vsubl_s32 (int32x2_t, int32x2_t)
23441 _Form of expected instruction(s):_ `vsubl.s32 Q0, D0, D0'
23443 * int32x4_t vsubl_s16 (int16x4_t, int16x4_t)
23444 _Form of expected instruction(s):_ `vsubl.s16 Q0, D0, D0'
23446 * int16x8_t vsubl_s8 (int8x8_t, int8x8_t)
23447 _Form of expected instruction(s):_ `vsubl.s8 Q0, D0, D0'
23449 * uint64x2_t vsubw_u32 (uint64x2_t, uint32x2_t)
23450 _Form of expected instruction(s):_ `vsubw.u32 Q0, Q0, D0'
23452 * uint32x4_t vsubw_u16 (uint32x4_t, uint16x4_t)
23453 _Form of expected instruction(s):_ `vsubw.u16 Q0, Q0, D0'
23455 * uint16x8_t vsubw_u8 (uint16x8_t, uint8x8_t)
23456 _Form of expected instruction(s):_ `vsubw.u8 Q0, Q0, D0'
23458 * int64x2_t vsubw_s32 (int64x2_t, int32x2_t)
23459 _Form of expected instruction(s):_ `vsubw.s32 Q0, Q0, D0'
23461 * int32x4_t vsubw_s16 (int32x4_t, int16x4_t)
23462 _Form of expected instruction(s):_ `vsubw.s16 Q0, Q0, D0'
23464 * int16x8_t vsubw_s8 (int16x8_t, int8x8_t)
23465 _Form of expected instruction(s):_ `vsubw.s8 Q0, Q0, D0'
23467 * uint32x2_t vhsub_u32 (uint32x2_t, uint32x2_t)
23468 _Form of expected instruction(s):_ `vhsub.u32 D0, D0, D0'
23470 * uint16x4_t vhsub_u16 (uint16x4_t, uint16x4_t)
23471 _Form of expected instruction(s):_ `vhsub.u16 D0, D0, D0'
23473 * uint8x8_t vhsub_u8 (uint8x8_t, uint8x8_t)
23474 _Form of expected instruction(s):_ `vhsub.u8 D0, D0, D0'
23476 * int32x2_t vhsub_s32 (int32x2_t, int32x2_t)
23477 _Form of expected instruction(s):_ `vhsub.s32 D0, D0, D0'
23479 * int16x4_t vhsub_s16 (int16x4_t, int16x4_t)
23480 _Form of expected instruction(s):_ `vhsub.s16 D0, D0, D0'
23482 * int8x8_t vhsub_s8 (int8x8_t, int8x8_t)
23483 _Form of expected instruction(s):_ `vhsub.s8 D0, D0, D0'
23485 * uint32x4_t vhsubq_u32 (uint32x4_t, uint32x4_t)
23486 _Form of expected instruction(s):_ `vhsub.u32 Q0, Q0, Q0'
23488 * uint16x8_t vhsubq_u16 (uint16x8_t, uint16x8_t)
23489 _Form of expected instruction(s):_ `vhsub.u16 Q0, Q0, Q0'
23491 * uint8x16_t vhsubq_u8 (uint8x16_t, uint8x16_t)
23492 _Form of expected instruction(s):_ `vhsub.u8 Q0, Q0, Q0'
23494 * int32x4_t vhsubq_s32 (int32x4_t, int32x4_t)
23495 _Form of expected instruction(s):_ `vhsub.s32 Q0, Q0, Q0'
23497 * int16x8_t vhsubq_s16 (int16x8_t, int16x8_t)
23498 _Form of expected instruction(s):_ `vhsub.s16 Q0, Q0, Q0'
23500 * int8x16_t vhsubq_s8 (int8x16_t, int8x16_t)
23501 _Form of expected instruction(s):_ `vhsub.s8 Q0, Q0, Q0'
23503 * uint32x2_t vqsub_u32 (uint32x2_t, uint32x2_t)
23504 _Form of expected instruction(s):_ `vqsub.u32 D0, D0, D0'
23506 * uint16x4_t vqsub_u16 (uint16x4_t, uint16x4_t)
23507 _Form of expected instruction(s):_ `vqsub.u16 D0, D0, D0'
23509 * uint8x8_t vqsub_u8 (uint8x8_t, uint8x8_t)
23510 _Form of expected instruction(s):_ `vqsub.u8 D0, D0, D0'
23512 * int32x2_t vqsub_s32 (int32x2_t, int32x2_t)
23513 _Form of expected instruction(s):_ `vqsub.s32 D0, D0, D0'
23515 * int16x4_t vqsub_s16 (int16x4_t, int16x4_t)
23516 _Form of expected instruction(s):_ `vqsub.s16 D0, D0, D0'
23518 * int8x8_t vqsub_s8 (int8x8_t, int8x8_t)
23519 _Form of expected instruction(s):_ `vqsub.s8 D0, D0, D0'
23521 * uint64x1_t vqsub_u64 (uint64x1_t, uint64x1_t)
23522 _Form of expected instruction(s):_ `vqsub.u64 D0, D0, D0'
23524 * int64x1_t vqsub_s64 (int64x1_t, int64x1_t)
23525 _Form of expected instruction(s):_ `vqsub.s64 D0, D0, D0'
23527 * uint32x4_t vqsubq_u32 (uint32x4_t, uint32x4_t)
23528 _Form of expected instruction(s):_ `vqsub.u32 Q0, Q0, Q0'
23530 * uint16x8_t vqsubq_u16 (uint16x8_t, uint16x8_t)
23531 _Form of expected instruction(s):_ `vqsub.u16 Q0, Q0, Q0'
23533 * uint8x16_t vqsubq_u8 (uint8x16_t, uint8x16_t)
23534 _Form of expected instruction(s):_ `vqsub.u8 Q0, Q0, Q0'
23536 * int32x4_t vqsubq_s32 (int32x4_t, int32x4_t)
23537 _Form of expected instruction(s):_ `vqsub.s32 Q0, Q0, Q0'
23539 * int16x8_t vqsubq_s16 (int16x8_t, int16x8_t)
23540 _Form of expected instruction(s):_ `vqsub.s16 Q0, Q0, Q0'
23542 * int8x16_t vqsubq_s8 (int8x16_t, int8x16_t)
23543 _Form of expected instruction(s):_ `vqsub.s8 Q0, Q0, Q0'
23545 * uint64x2_t vqsubq_u64 (uint64x2_t, uint64x2_t)
23546 _Form of expected instruction(s):_ `vqsub.u64 Q0, Q0, Q0'
23548 * int64x2_t vqsubq_s64 (int64x2_t, int64x2_t)
23549 _Form of expected instruction(s):_ `vqsub.s64 Q0, Q0, Q0'
23551 * uint32x2_t vsubhn_u64 (uint64x2_t, uint64x2_t)
23552 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
23554 * uint16x4_t vsubhn_u32 (uint32x4_t, uint32x4_t)
23555 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
23557 * uint8x8_t vsubhn_u16 (uint16x8_t, uint16x8_t)
23558 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
23560 * int32x2_t vsubhn_s64 (int64x2_t, int64x2_t)
23561 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
23563 * int16x4_t vsubhn_s32 (int32x4_t, int32x4_t)
23564 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
23566 * int8x8_t vsubhn_s16 (int16x8_t, int16x8_t)
23567 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
23569 * uint32x2_t vrsubhn_u64 (uint64x2_t, uint64x2_t)
23570 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
23572 * uint16x4_t vrsubhn_u32 (uint32x4_t, uint32x4_t)
23573 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
23575 * uint8x8_t vrsubhn_u16 (uint16x8_t, uint16x8_t)
23576 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
23578 * int32x2_t vrsubhn_s64 (int64x2_t, int64x2_t)
23579 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
23581 * int16x4_t vrsubhn_s32 (int32x4_t, int32x4_t)
23582 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
23584 * int8x8_t vrsubhn_s16 (int16x8_t, int16x8_t)
23585 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
23587 5.50.3.6 Comparison (equal-to)
23588 ..............................
23590 * uint32x2_t vceq_u32 (uint32x2_t, uint32x2_t)
23591 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
23593 * uint16x4_t vceq_u16 (uint16x4_t, uint16x4_t)
23594 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
23596 * uint8x8_t vceq_u8 (uint8x8_t, uint8x8_t)
23597 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
23599 * uint32x2_t vceq_s32 (int32x2_t, int32x2_t)
23600 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
23602 * uint16x4_t vceq_s16 (int16x4_t, int16x4_t)
23603 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
23605 * uint8x8_t vceq_s8 (int8x8_t, int8x8_t)
23606 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
23608 * uint32x2_t vceq_f32 (float32x2_t, float32x2_t)
23609 _Form of expected instruction(s):_ `vceq.f32 D0, D0, D0'
23611 * uint8x8_t vceq_p8 (poly8x8_t, poly8x8_t)
23612 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
23614 * uint32x4_t vceqq_u32 (uint32x4_t, uint32x4_t)
23615 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
23617 * uint16x8_t vceqq_u16 (uint16x8_t, uint16x8_t)
23618 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
23620 * uint8x16_t vceqq_u8 (uint8x16_t, uint8x16_t)
23621 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
23623 * uint32x4_t vceqq_s32 (int32x4_t, int32x4_t)
23624 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
23626 * uint16x8_t vceqq_s16 (int16x8_t, int16x8_t)
23627 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
23629 * uint8x16_t vceqq_s8 (int8x16_t, int8x16_t)
23630 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
23632 * uint32x4_t vceqq_f32 (float32x4_t, float32x4_t)
23633 _Form of expected instruction(s):_ `vceq.f32 Q0, Q0, Q0'
23635 * uint8x16_t vceqq_p8 (poly8x16_t, poly8x16_t)
23636 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
23638 5.50.3.7 Comparison (greater-than-or-equal-to)
23639 ..............................................
23641 * uint32x2_t vcge_u32 (uint32x2_t, uint32x2_t)
23642 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
23644 * uint16x4_t vcge_u16 (uint16x4_t, uint16x4_t)
23645 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
23647 * uint8x8_t vcge_u8 (uint8x8_t, uint8x8_t)
23648 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
23650 * uint32x2_t vcge_s32 (int32x2_t, int32x2_t)
23651 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
23653 * uint16x4_t vcge_s16 (int16x4_t, int16x4_t)
23654 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
23656 * uint8x8_t vcge_s8 (int8x8_t, int8x8_t)
23657 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
23659 * uint32x2_t vcge_f32 (float32x2_t, float32x2_t)
23660 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
23662 * uint32x4_t vcgeq_u32 (uint32x4_t, uint32x4_t)
23663 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
23665 * uint16x8_t vcgeq_u16 (uint16x8_t, uint16x8_t)
23666 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
23668 * uint8x16_t vcgeq_u8 (uint8x16_t, uint8x16_t)
23669 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
23671 * uint32x4_t vcgeq_s32 (int32x4_t, int32x4_t)
23672 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
23674 * uint16x8_t vcgeq_s16 (int16x8_t, int16x8_t)
23675 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
23677 * uint8x16_t vcgeq_s8 (int8x16_t, int8x16_t)
23678 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
23680 * uint32x4_t vcgeq_f32 (float32x4_t, float32x4_t)
23681 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
23683 5.50.3.8 Comparison (less-than-or-equal-to)
23684 ...........................................
23686 * uint32x2_t vcle_u32 (uint32x2_t, uint32x2_t)
23687 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
23689 * uint16x4_t vcle_u16 (uint16x4_t, uint16x4_t)
23690 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
23692 * uint8x8_t vcle_u8 (uint8x8_t, uint8x8_t)
23693 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
23695 * uint32x2_t vcle_s32 (int32x2_t, int32x2_t)
23696 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
23698 * uint16x4_t vcle_s16 (int16x4_t, int16x4_t)
23699 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
23701 * uint8x8_t vcle_s8 (int8x8_t, int8x8_t)
23702 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
23704 * uint32x2_t vcle_f32 (float32x2_t, float32x2_t)
23705 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
23707 * uint32x4_t vcleq_u32 (uint32x4_t, uint32x4_t)
23708 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
23710 * uint16x8_t vcleq_u16 (uint16x8_t, uint16x8_t)
23711 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
23713 * uint8x16_t vcleq_u8 (uint8x16_t, uint8x16_t)
23714 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
23716 * uint32x4_t vcleq_s32 (int32x4_t, int32x4_t)
23717 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
23719 * uint16x8_t vcleq_s16 (int16x8_t, int16x8_t)
23720 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
23722 * uint8x16_t vcleq_s8 (int8x16_t, int8x16_t)
23723 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
23725 * uint32x4_t vcleq_f32 (float32x4_t, float32x4_t)
23726 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
23728 5.50.3.9 Comparison (greater-than)
23729 ..................................
23731 * uint32x2_t vcgt_u32 (uint32x2_t, uint32x2_t)
23732 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
23734 * uint16x4_t vcgt_u16 (uint16x4_t, uint16x4_t)
23735 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
23737 * uint8x8_t vcgt_u8 (uint8x8_t, uint8x8_t)
23738 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
23740 * uint32x2_t vcgt_s32 (int32x2_t, int32x2_t)
23741 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
23743 * uint16x4_t vcgt_s16 (int16x4_t, int16x4_t)
23744 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
23746 * uint8x8_t vcgt_s8 (int8x8_t, int8x8_t)
23747 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
23749 * uint32x2_t vcgt_f32 (float32x2_t, float32x2_t)
23750 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
23752 * uint32x4_t vcgtq_u32 (uint32x4_t, uint32x4_t)
23753 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
23755 * uint16x8_t vcgtq_u16 (uint16x8_t, uint16x8_t)
23756 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
23758 * uint8x16_t vcgtq_u8 (uint8x16_t, uint8x16_t)
23759 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
23761 * uint32x4_t vcgtq_s32 (int32x4_t, int32x4_t)
23762 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
23764 * uint16x8_t vcgtq_s16 (int16x8_t, int16x8_t)
23765 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
23767 * uint8x16_t vcgtq_s8 (int8x16_t, int8x16_t)
23768 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
23770 * uint32x4_t vcgtq_f32 (float32x4_t, float32x4_t)
23771 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
23773 5.50.3.10 Comparison (less-than)
23774 ................................
23776 * uint32x2_t vclt_u32 (uint32x2_t, uint32x2_t)
23777 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
23779 * uint16x4_t vclt_u16 (uint16x4_t, uint16x4_t)
23780 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
23782 * uint8x8_t vclt_u8 (uint8x8_t, uint8x8_t)
23783 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
23785 * uint32x2_t vclt_s32 (int32x2_t, int32x2_t)
23786 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
23788 * uint16x4_t vclt_s16 (int16x4_t, int16x4_t)
23789 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
23791 * uint8x8_t vclt_s8 (int8x8_t, int8x8_t)
23792 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
23794 * uint32x2_t vclt_f32 (float32x2_t, float32x2_t)
23795 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
23797 * uint32x4_t vcltq_u32 (uint32x4_t, uint32x4_t)
23798 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
23800 * uint16x8_t vcltq_u16 (uint16x8_t, uint16x8_t)
23801 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
23803 * uint8x16_t vcltq_u8 (uint8x16_t, uint8x16_t)
23804 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
23806 * uint32x4_t vcltq_s32 (int32x4_t, int32x4_t)
23807 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
23809 * uint16x8_t vcltq_s16 (int16x8_t, int16x8_t)
23810 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
23812 * uint8x16_t vcltq_s8 (int8x16_t, int8x16_t)
23813 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
23815 * uint32x4_t vcltq_f32 (float32x4_t, float32x4_t)
23816 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
23818 5.50.3.11 Comparison (absolute greater-than-or-equal-to)
23819 ........................................................
23821 * uint32x2_t vcage_f32 (float32x2_t, float32x2_t)
23822 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
23824 * uint32x4_t vcageq_f32 (float32x4_t, float32x4_t)
23825 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
23827 5.50.3.12 Comparison (absolute less-than-or-equal-to)
23828 .....................................................
23830 * uint32x2_t vcale_f32 (float32x2_t, float32x2_t)
23831 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
23833 * uint32x4_t vcaleq_f32 (float32x4_t, float32x4_t)
23834 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
23836 5.50.3.13 Comparison (absolute greater-than)
23837 ............................................
23839 * uint32x2_t vcagt_f32 (float32x2_t, float32x2_t)
23840 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
23842 * uint32x4_t vcagtq_f32 (float32x4_t, float32x4_t)
23843 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
23845 5.50.3.14 Comparison (absolute less-than)
23846 .........................................
23848 * uint32x2_t vcalt_f32 (float32x2_t, float32x2_t)
23849 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
23851 * uint32x4_t vcaltq_f32 (float32x4_t, float32x4_t)
23852 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
23854 5.50.3.15 Test bits
23855 ...................
23857 * uint32x2_t vtst_u32 (uint32x2_t, uint32x2_t)
23858 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
23860 * uint16x4_t vtst_u16 (uint16x4_t, uint16x4_t)
23861 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
23863 * uint8x8_t vtst_u8 (uint8x8_t, uint8x8_t)
23864 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
23866 * uint32x2_t vtst_s32 (int32x2_t, int32x2_t)
23867 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
23869 * uint16x4_t vtst_s16 (int16x4_t, int16x4_t)
23870 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
23872 * uint8x8_t vtst_s8 (int8x8_t, int8x8_t)
23873 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
23875 * uint8x8_t vtst_p8 (poly8x8_t, poly8x8_t)
23876 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
23878 * uint32x4_t vtstq_u32 (uint32x4_t, uint32x4_t)
23879 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
23881 * uint16x8_t vtstq_u16 (uint16x8_t, uint16x8_t)
23882 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
23884 * uint8x16_t vtstq_u8 (uint8x16_t, uint8x16_t)
23885 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
23887 * uint32x4_t vtstq_s32 (int32x4_t, int32x4_t)
23888 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
23890 * uint16x8_t vtstq_s16 (int16x8_t, int16x8_t)
23891 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
23893 * uint8x16_t vtstq_s8 (int8x16_t, int8x16_t)
23894 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
23896 * uint8x16_t vtstq_p8 (poly8x16_t, poly8x16_t)
23897 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
23899 5.50.3.16 Absolute difference
23900 .............................
23902 * uint32x2_t vabd_u32 (uint32x2_t, uint32x2_t)
23903 _Form of expected instruction(s):_ `vabd.u32 D0, D0, D0'
23905 * uint16x4_t vabd_u16 (uint16x4_t, uint16x4_t)
23906 _Form of expected instruction(s):_ `vabd.u16 D0, D0, D0'
23908 * uint8x8_t vabd_u8 (uint8x8_t, uint8x8_t)
23909 _Form of expected instruction(s):_ `vabd.u8 D0, D0, D0'
23911 * int32x2_t vabd_s32 (int32x2_t, int32x2_t)
23912 _Form of expected instruction(s):_ `vabd.s32 D0, D0, D0'
23914 * int16x4_t vabd_s16 (int16x4_t, int16x4_t)
23915 _Form of expected instruction(s):_ `vabd.s16 D0, D0, D0'
23917 * int8x8_t vabd_s8 (int8x8_t, int8x8_t)
23918 _Form of expected instruction(s):_ `vabd.s8 D0, D0, D0'
23920 * float32x2_t vabd_f32 (float32x2_t, float32x2_t)
23921 _Form of expected instruction(s):_ `vabd.f32 D0, D0, D0'
23923 * uint32x4_t vabdq_u32 (uint32x4_t, uint32x4_t)
23924 _Form of expected instruction(s):_ `vabd.u32 Q0, Q0, Q0'
23926 * uint16x8_t vabdq_u16 (uint16x8_t, uint16x8_t)
23927 _Form of expected instruction(s):_ `vabd.u16 Q0, Q0, Q0'
23929 * uint8x16_t vabdq_u8 (uint8x16_t, uint8x16_t)
23930 _Form of expected instruction(s):_ `vabd.u8 Q0, Q0, Q0'
23932 * int32x4_t vabdq_s32 (int32x4_t, int32x4_t)
23933 _Form of expected instruction(s):_ `vabd.s32 Q0, Q0, Q0'
23935 * int16x8_t vabdq_s16 (int16x8_t, int16x8_t)
23936 _Form of expected instruction(s):_ `vabd.s16 Q0, Q0, Q0'
23938 * int8x16_t vabdq_s8 (int8x16_t, int8x16_t)
23939 _Form of expected instruction(s):_ `vabd.s8 Q0, Q0, Q0'
23941 * float32x4_t vabdq_f32 (float32x4_t, float32x4_t)
23942 _Form of expected instruction(s):_ `vabd.f32 Q0, Q0, Q0'
23944 * uint64x2_t vabdl_u32 (uint32x2_t, uint32x2_t)
23945 _Form of expected instruction(s):_ `vabdl.u32 Q0, D0, D0'
23947 * uint32x4_t vabdl_u16 (uint16x4_t, uint16x4_t)
23948 _Form of expected instruction(s):_ `vabdl.u16 Q0, D0, D0'
23950 * uint16x8_t vabdl_u8 (uint8x8_t, uint8x8_t)
23951 _Form of expected instruction(s):_ `vabdl.u8 Q0, D0, D0'
23953 * int64x2_t vabdl_s32 (int32x2_t, int32x2_t)
23954 _Form of expected instruction(s):_ `vabdl.s32 Q0, D0, D0'
23956 * int32x4_t vabdl_s16 (int16x4_t, int16x4_t)
23957 _Form of expected instruction(s):_ `vabdl.s16 Q0, D0, D0'
23959 * int16x8_t vabdl_s8 (int8x8_t, int8x8_t)
23960 _Form of expected instruction(s):_ `vabdl.s8 Q0, D0, D0'
23962 5.50.3.17 Absolute difference and accumulate
23963 ............................................
23965 * uint32x2_t vaba_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
23966 _Form of expected instruction(s):_ `vaba.u32 D0, D0, D0'
23968 * uint16x4_t vaba_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
23969 _Form of expected instruction(s):_ `vaba.u16 D0, D0, D0'
23971 * uint8x8_t vaba_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
23972 _Form of expected instruction(s):_ `vaba.u8 D0, D0, D0'
23974 * int32x2_t vaba_s32 (int32x2_t, int32x2_t, int32x2_t)
23975 _Form of expected instruction(s):_ `vaba.s32 D0, D0, D0'
23977 * int16x4_t vaba_s16 (int16x4_t, int16x4_t, int16x4_t)
23978 _Form of expected instruction(s):_ `vaba.s16 D0, D0, D0'
23980 * int8x8_t vaba_s8 (int8x8_t, int8x8_t, int8x8_t)
23981 _Form of expected instruction(s):_ `vaba.s8 D0, D0, D0'
23983 * uint32x4_t vabaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
23984 _Form of expected instruction(s):_ `vaba.u32 Q0, Q0, Q0'
23986 * uint16x8_t vabaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
23987 _Form of expected instruction(s):_ `vaba.u16 Q0, Q0, Q0'
23989 * uint8x16_t vabaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
23990 _Form of expected instruction(s):_ `vaba.u8 Q0, Q0, Q0'
23992 * int32x4_t vabaq_s32 (int32x4_t, int32x4_t, int32x4_t)
23993 _Form of expected instruction(s):_ `vaba.s32 Q0, Q0, Q0'
23995 * int16x8_t vabaq_s16 (int16x8_t, int16x8_t, int16x8_t)
23996 _Form of expected instruction(s):_ `vaba.s16 Q0, Q0, Q0'
23998 * int8x16_t vabaq_s8 (int8x16_t, int8x16_t, int8x16_t)
23999 _Form of expected instruction(s):_ `vaba.s8 Q0, Q0, Q0'
24001 * uint64x2_t vabal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
24002 _Form of expected instruction(s):_ `vabal.u32 Q0, D0, D0'
24004 * uint32x4_t vabal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
24005 _Form of expected instruction(s):_ `vabal.u16 Q0, D0, D0'
24007 * uint16x8_t vabal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
24008 _Form of expected instruction(s):_ `vabal.u8 Q0, D0, D0'
24010 * int64x2_t vabal_s32 (int64x2_t, int32x2_t, int32x2_t)
24011 _Form of expected instruction(s):_ `vabal.s32 Q0, D0, D0'
24013 * int32x4_t vabal_s16 (int32x4_t, int16x4_t, int16x4_t)
24014 _Form of expected instruction(s):_ `vabal.s16 Q0, D0, D0'
24016 * int16x8_t vabal_s8 (int16x8_t, int8x8_t, int8x8_t)
24017 _Form of expected instruction(s):_ `vabal.s8 Q0, D0, D0'
24022 * uint32x2_t vmax_u32 (uint32x2_t, uint32x2_t)
24023 _Form of expected instruction(s):_ `vmax.u32 D0, D0, D0'
24025 * uint16x4_t vmax_u16 (uint16x4_t, uint16x4_t)
24026 _Form of expected instruction(s):_ `vmax.u16 D0, D0, D0'
24028 * uint8x8_t vmax_u8 (uint8x8_t, uint8x8_t)
24029 _Form of expected instruction(s):_ `vmax.u8 D0, D0, D0'
24031 * int32x2_t vmax_s32 (int32x2_t, int32x2_t)
24032 _Form of expected instruction(s):_ `vmax.s32 D0, D0, D0'
24034 * int16x4_t vmax_s16 (int16x4_t, int16x4_t)
24035 _Form of expected instruction(s):_ `vmax.s16 D0, D0, D0'
24037 * int8x8_t vmax_s8 (int8x8_t, int8x8_t)
24038 _Form of expected instruction(s):_ `vmax.s8 D0, D0, D0'
24040 * float32x2_t vmax_f32 (float32x2_t, float32x2_t)
24041 _Form of expected instruction(s):_ `vmax.f32 D0, D0, D0'
24043 * uint32x4_t vmaxq_u32 (uint32x4_t, uint32x4_t)
24044 _Form of expected instruction(s):_ `vmax.u32 Q0, Q0, Q0'
24046 * uint16x8_t vmaxq_u16 (uint16x8_t, uint16x8_t)
24047 _Form of expected instruction(s):_ `vmax.u16 Q0, Q0, Q0'
24049 * uint8x16_t vmaxq_u8 (uint8x16_t, uint8x16_t)
24050 _Form of expected instruction(s):_ `vmax.u8 Q0, Q0, Q0'
24052 * int32x4_t vmaxq_s32 (int32x4_t, int32x4_t)
24053 _Form of expected instruction(s):_ `vmax.s32 Q0, Q0, Q0'
24055 * int16x8_t vmaxq_s16 (int16x8_t, int16x8_t)
24056 _Form of expected instruction(s):_ `vmax.s16 Q0, Q0, Q0'
24058 * int8x16_t vmaxq_s8 (int8x16_t, int8x16_t)
24059 _Form of expected instruction(s):_ `vmax.s8 Q0, Q0, Q0'
24061 * float32x4_t vmaxq_f32 (float32x4_t, float32x4_t)
24062 _Form of expected instruction(s):_ `vmax.f32 Q0, Q0, Q0'
24067 * uint32x2_t vmin_u32 (uint32x2_t, uint32x2_t)
24068 _Form of expected instruction(s):_ `vmin.u32 D0, D0, D0'
24070 * uint16x4_t vmin_u16 (uint16x4_t, uint16x4_t)
24071 _Form of expected instruction(s):_ `vmin.u16 D0, D0, D0'
24073 * uint8x8_t vmin_u8 (uint8x8_t, uint8x8_t)
24074 _Form of expected instruction(s):_ `vmin.u8 D0, D0, D0'
24076 * int32x2_t vmin_s32 (int32x2_t, int32x2_t)
24077 _Form of expected instruction(s):_ `vmin.s32 D0, D0, D0'
24079 * int16x4_t vmin_s16 (int16x4_t, int16x4_t)
24080 _Form of expected instruction(s):_ `vmin.s16 D0, D0, D0'
24082 * int8x8_t vmin_s8 (int8x8_t, int8x8_t)
24083 _Form of expected instruction(s):_ `vmin.s8 D0, D0, D0'
24085 * float32x2_t vmin_f32 (float32x2_t, float32x2_t)
24086 _Form of expected instruction(s):_ `vmin.f32 D0, D0, D0'
24088 * uint32x4_t vminq_u32 (uint32x4_t, uint32x4_t)
24089 _Form of expected instruction(s):_ `vmin.u32 Q0, Q0, Q0'
24091 * uint16x8_t vminq_u16 (uint16x8_t, uint16x8_t)
24092 _Form of expected instruction(s):_ `vmin.u16 Q0, Q0, Q0'
24094 * uint8x16_t vminq_u8 (uint8x16_t, uint8x16_t)
24095 _Form of expected instruction(s):_ `vmin.u8 Q0, Q0, Q0'
24097 * int32x4_t vminq_s32 (int32x4_t, int32x4_t)
24098 _Form of expected instruction(s):_ `vmin.s32 Q0, Q0, Q0'
24100 * int16x8_t vminq_s16 (int16x8_t, int16x8_t)
24101 _Form of expected instruction(s):_ `vmin.s16 Q0, Q0, Q0'
24103 * int8x16_t vminq_s8 (int8x16_t, int8x16_t)
24104 _Form of expected instruction(s):_ `vmin.s8 Q0, Q0, Q0'
24106 * float32x4_t vminq_f32 (float32x4_t, float32x4_t)
24107 _Form of expected instruction(s):_ `vmin.f32 Q0, Q0, Q0'
24109 5.50.3.20 Pairwise add
24110 ......................
24112 * uint32x2_t vpadd_u32 (uint32x2_t, uint32x2_t)
24113 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
24115 * uint16x4_t vpadd_u16 (uint16x4_t, uint16x4_t)
24116 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
24118 * uint8x8_t vpadd_u8 (uint8x8_t, uint8x8_t)
24119 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
24121 * int32x2_t vpadd_s32 (int32x2_t, int32x2_t)
24122 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
24124 * int16x4_t vpadd_s16 (int16x4_t, int16x4_t)
24125 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
24127 * int8x8_t vpadd_s8 (int8x8_t, int8x8_t)
24128 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
24130 * float32x2_t vpadd_f32 (float32x2_t, float32x2_t)
24131 _Form of expected instruction(s):_ `vpadd.f32 D0, D0, D0'
24133 * uint64x1_t vpaddl_u32 (uint32x2_t)
24134 _Form of expected instruction(s):_ `vpaddl.u32 D0, D0'
24136 * uint32x2_t vpaddl_u16 (uint16x4_t)
24137 _Form of expected instruction(s):_ `vpaddl.u16 D0, D0'
24139 * uint16x4_t vpaddl_u8 (uint8x8_t)
24140 _Form of expected instruction(s):_ `vpaddl.u8 D0, D0'
24142 * int64x1_t vpaddl_s32 (int32x2_t)
24143 _Form of expected instruction(s):_ `vpaddl.s32 D0, D0'
24145 * int32x2_t vpaddl_s16 (int16x4_t)
24146 _Form of expected instruction(s):_ `vpaddl.s16 D0, D0'
24148 * int16x4_t vpaddl_s8 (int8x8_t)
24149 _Form of expected instruction(s):_ `vpaddl.s8 D0, D0'
24151 * uint64x2_t vpaddlq_u32 (uint32x4_t)
24152 _Form of expected instruction(s):_ `vpaddl.u32 Q0, Q0'
24154 * uint32x4_t vpaddlq_u16 (uint16x8_t)
24155 _Form of expected instruction(s):_ `vpaddl.u16 Q0, Q0'
24157 * uint16x8_t vpaddlq_u8 (uint8x16_t)
24158 _Form of expected instruction(s):_ `vpaddl.u8 Q0, Q0'
24160 * int64x2_t vpaddlq_s32 (int32x4_t)
24161 _Form of expected instruction(s):_ `vpaddl.s32 Q0, Q0'
24163 * int32x4_t vpaddlq_s16 (int16x8_t)
24164 _Form of expected instruction(s):_ `vpaddl.s16 Q0, Q0'
24166 * int16x8_t vpaddlq_s8 (int8x16_t)
24167 _Form of expected instruction(s):_ `vpaddl.s8 Q0, Q0'
24169 5.50.3.21 Pairwise add, single_opcode widen and accumulate
24170 ..........................................................
24172 * uint64x1_t vpadal_u32 (uint64x1_t, uint32x2_t)
24173 _Form of expected instruction(s):_ `vpadal.u32 D0, D0'
24175 * uint32x2_t vpadal_u16 (uint32x2_t, uint16x4_t)
24176 _Form of expected instruction(s):_ `vpadal.u16 D0, D0'
24178 * uint16x4_t vpadal_u8 (uint16x4_t, uint8x8_t)
24179 _Form of expected instruction(s):_ `vpadal.u8 D0, D0'
24181 * int64x1_t vpadal_s32 (int64x1_t, int32x2_t)
24182 _Form of expected instruction(s):_ `vpadal.s32 D0, D0'
24184 * int32x2_t vpadal_s16 (int32x2_t, int16x4_t)
24185 _Form of expected instruction(s):_ `vpadal.s16 D0, D0'
24187 * int16x4_t vpadal_s8 (int16x4_t, int8x8_t)
24188 _Form of expected instruction(s):_ `vpadal.s8 D0, D0'
24190 * uint64x2_t vpadalq_u32 (uint64x2_t, uint32x4_t)
24191 _Form of expected instruction(s):_ `vpadal.u32 Q0, Q0'
24193 * uint32x4_t vpadalq_u16 (uint32x4_t, uint16x8_t)
24194 _Form of expected instruction(s):_ `vpadal.u16 Q0, Q0'
24196 * uint16x8_t vpadalq_u8 (uint16x8_t, uint8x16_t)
24197 _Form of expected instruction(s):_ `vpadal.u8 Q0, Q0'
24199 * int64x2_t vpadalq_s32 (int64x2_t, int32x4_t)
24200 _Form of expected instruction(s):_ `vpadal.s32 Q0, Q0'
24202 * int32x4_t vpadalq_s16 (int32x4_t, int16x8_t)
24203 _Form of expected instruction(s):_ `vpadal.s16 Q0, Q0'
24205 * int16x8_t vpadalq_s8 (int16x8_t, int8x16_t)
24206 _Form of expected instruction(s):_ `vpadal.s8 Q0, Q0'
24208 5.50.3.22 Folding maximum
24209 .........................
24211 * uint32x2_t vpmax_u32 (uint32x2_t, uint32x2_t)
24212 _Form of expected instruction(s):_ `vpmax.u32 D0, D0, D0'
24214 * uint16x4_t vpmax_u16 (uint16x4_t, uint16x4_t)
24215 _Form of expected instruction(s):_ `vpmax.u16 D0, D0, D0'
24217 * uint8x8_t vpmax_u8 (uint8x8_t, uint8x8_t)
24218 _Form of expected instruction(s):_ `vpmax.u8 D0, D0, D0'
24220 * int32x2_t vpmax_s32 (int32x2_t, int32x2_t)
24221 _Form of expected instruction(s):_ `vpmax.s32 D0, D0, D0'
24223 * int16x4_t vpmax_s16 (int16x4_t, int16x4_t)
24224 _Form of expected instruction(s):_ `vpmax.s16 D0, D0, D0'
24226 * int8x8_t vpmax_s8 (int8x8_t, int8x8_t)
24227 _Form of expected instruction(s):_ `vpmax.s8 D0, D0, D0'
24229 * float32x2_t vpmax_f32 (float32x2_t, float32x2_t)
24230 _Form of expected instruction(s):_ `vpmax.f32 D0, D0, D0'
24232 5.50.3.23 Folding minimum
24233 .........................
24235 * uint32x2_t vpmin_u32 (uint32x2_t, uint32x2_t)
24236 _Form of expected instruction(s):_ `vpmin.u32 D0, D0, D0'
24238 * uint16x4_t vpmin_u16 (uint16x4_t, uint16x4_t)
24239 _Form of expected instruction(s):_ `vpmin.u16 D0, D0, D0'
24241 * uint8x8_t vpmin_u8 (uint8x8_t, uint8x8_t)
24242 _Form of expected instruction(s):_ `vpmin.u8 D0, D0, D0'
24244 * int32x2_t vpmin_s32 (int32x2_t, int32x2_t)
24245 _Form of expected instruction(s):_ `vpmin.s32 D0, D0, D0'
24247 * int16x4_t vpmin_s16 (int16x4_t, int16x4_t)
24248 _Form of expected instruction(s):_ `vpmin.s16 D0, D0, D0'
24250 * int8x8_t vpmin_s8 (int8x8_t, int8x8_t)
24251 _Form of expected instruction(s):_ `vpmin.s8 D0, D0, D0'
24253 * float32x2_t vpmin_f32 (float32x2_t, float32x2_t)
24254 _Form of expected instruction(s):_ `vpmin.f32 D0, D0, D0'
24256 5.50.3.24 Reciprocal step
24257 .........................
24259 * float32x2_t vrecps_f32 (float32x2_t, float32x2_t)
24260 _Form of expected instruction(s):_ `vrecps.f32 D0, D0, D0'
24262 * float32x4_t vrecpsq_f32 (float32x4_t, float32x4_t)
24263 _Form of expected instruction(s):_ `vrecps.f32 Q0, Q0, Q0'
24265 * float32x2_t vrsqrts_f32 (float32x2_t, float32x2_t)
24266 _Form of expected instruction(s):_ `vrsqrts.f32 D0, D0, D0'
24268 * float32x4_t vrsqrtsq_f32 (float32x4_t, float32x4_t)
24269 _Form of expected instruction(s):_ `vrsqrts.f32 Q0, Q0, Q0'
24271 5.50.3.25 Vector shift left
24272 ...........................
24274 * uint32x2_t vshl_u32 (uint32x2_t, int32x2_t)
24275 _Form of expected instruction(s):_ `vshl.u32 D0, D0, D0'
24277 * uint16x4_t vshl_u16 (uint16x4_t, int16x4_t)
24278 _Form of expected instruction(s):_ `vshl.u16 D0, D0, D0'
24280 * uint8x8_t vshl_u8 (uint8x8_t, int8x8_t)
24281 _Form of expected instruction(s):_ `vshl.u8 D0, D0, D0'
24283 * int32x2_t vshl_s32 (int32x2_t, int32x2_t)
24284 _Form of expected instruction(s):_ `vshl.s32 D0, D0, D0'
24286 * int16x4_t vshl_s16 (int16x4_t, int16x4_t)
24287 _Form of expected instruction(s):_ `vshl.s16 D0, D0, D0'
24289 * int8x8_t vshl_s8 (int8x8_t, int8x8_t)
24290 _Form of expected instruction(s):_ `vshl.s8 D0, D0, D0'
24292 * uint64x1_t vshl_u64 (uint64x1_t, int64x1_t)
24293 _Form of expected instruction(s):_ `vshl.u64 D0, D0, D0'
24295 * int64x1_t vshl_s64 (int64x1_t, int64x1_t)
24296 _Form of expected instruction(s):_ `vshl.s64 D0, D0, D0'
24298 * uint32x4_t vshlq_u32 (uint32x4_t, int32x4_t)
24299 _Form of expected instruction(s):_ `vshl.u32 Q0, Q0, Q0'
24301 * uint16x8_t vshlq_u16 (uint16x8_t, int16x8_t)
24302 _Form of expected instruction(s):_ `vshl.u16 Q0, Q0, Q0'
24304 * uint8x16_t vshlq_u8 (uint8x16_t, int8x16_t)
24305 _Form of expected instruction(s):_ `vshl.u8 Q0, Q0, Q0'
24307 * int32x4_t vshlq_s32 (int32x4_t, int32x4_t)
24308 _Form of expected instruction(s):_ `vshl.s32 Q0, Q0, Q0'
24310 * int16x8_t vshlq_s16 (int16x8_t, int16x8_t)
24311 _Form of expected instruction(s):_ `vshl.s16 Q0, Q0, Q0'
24313 * int8x16_t vshlq_s8 (int8x16_t, int8x16_t)
24314 _Form of expected instruction(s):_ `vshl.s8 Q0, Q0, Q0'
24316 * uint64x2_t vshlq_u64 (uint64x2_t, int64x2_t)
24317 _Form of expected instruction(s):_ `vshl.u64 Q0, Q0, Q0'
24319 * int64x2_t vshlq_s64 (int64x2_t, int64x2_t)
24320 _Form of expected instruction(s):_ `vshl.s64 Q0, Q0, Q0'
24322 * uint32x2_t vrshl_u32 (uint32x2_t, int32x2_t)
24323 _Form of expected instruction(s):_ `vrshl.u32 D0, D0, D0'
24325 * uint16x4_t vrshl_u16 (uint16x4_t, int16x4_t)
24326 _Form of expected instruction(s):_ `vrshl.u16 D0, D0, D0'
24328 * uint8x8_t vrshl_u8 (uint8x8_t, int8x8_t)
24329 _Form of expected instruction(s):_ `vrshl.u8 D0, D0, D0'
24331 * int32x2_t vrshl_s32 (int32x2_t, int32x2_t)
24332 _Form of expected instruction(s):_ `vrshl.s32 D0, D0, D0'
24334 * int16x4_t vrshl_s16 (int16x4_t, int16x4_t)
24335 _Form of expected instruction(s):_ `vrshl.s16 D0, D0, D0'
24337 * int8x8_t vrshl_s8 (int8x8_t, int8x8_t)
24338 _Form of expected instruction(s):_ `vrshl.s8 D0, D0, D0'
24340 * uint64x1_t vrshl_u64 (uint64x1_t, int64x1_t)
24341 _Form of expected instruction(s):_ `vrshl.u64 D0, D0, D0'
24343 * int64x1_t vrshl_s64 (int64x1_t, int64x1_t)
24344 _Form of expected instruction(s):_ `vrshl.s64 D0, D0, D0'
24346 * uint32x4_t vrshlq_u32 (uint32x4_t, int32x4_t)
24347 _Form of expected instruction(s):_ `vrshl.u32 Q0, Q0, Q0'
24349 * uint16x8_t vrshlq_u16 (uint16x8_t, int16x8_t)
24350 _Form of expected instruction(s):_ `vrshl.u16 Q0, Q0, Q0'
24352 * uint8x16_t vrshlq_u8 (uint8x16_t, int8x16_t)
24353 _Form of expected instruction(s):_ `vrshl.u8 Q0, Q0, Q0'
24355 * int32x4_t vrshlq_s32 (int32x4_t, int32x4_t)
24356 _Form of expected instruction(s):_ `vrshl.s32 Q0, Q0, Q0'
24358 * int16x8_t vrshlq_s16 (int16x8_t, int16x8_t)
24359 _Form of expected instruction(s):_ `vrshl.s16 Q0, Q0, Q0'
24361 * int8x16_t vrshlq_s8 (int8x16_t, int8x16_t)
24362 _Form of expected instruction(s):_ `vrshl.s8 Q0, Q0, Q0'
24364 * uint64x2_t vrshlq_u64 (uint64x2_t, int64x2_t)
24365 _Form of expected instruction(s):_ `vrshl.u64 Q0, Q0, Q0'
24367 * int64x2_t vrshlq_s64 (int64x2_t, int64x2_t)
24368 _Form of expected instruction(s):_ `vrshl.s64 Q0, Q0, Q0'
24370 * uint32x2_t vqshl_u32 (uint32x2_t, int32x2_t)
24371 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, D0'
24373 * uint16x4_t vqshl_u16 (uint16x4_t, int16x4_t)
24374 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, D0'
24376 * uint8x8_t vqshl_u8 (uint8x8_t, int8x8_t)
24377 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, D0'
24379 * int32x2_t vqshl_s32 (int32x2_t, int32x2_t)
24380 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, D0'
24382 * int16x4_t vqshl_s16 (int16x4_t, int16x4_t)
24383 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, D0'
24385 * int8x8_t vqshl_s8 (int8x8_t, int8x8_t)
24386 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, D0'
24388 * uint64x1_t vqshl_u64 (uint64x1_t, int64x1_t)
24389 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, D0'
24391 * int64x1_t vqshl_s64 (int64x1_t, int64x1_t)
24392 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, D0'
24394 * uint32x4_t vqshlq_u32 (uint32x4_t, int32x4_t)
24395 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, Q0'
24397 * uint16x8_t vqshlq_u16 (uint16x8_t, int16x8_t)
24398 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, Q0'
24400 * uint8x16_t vqshlq_u8 (uint8x16_t, int8x16_t)
24401 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, Q0'
24403 * int32x4_t vqshlq_s32 (int32x4_t, int32x4_t)
24404 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, Q0'
24406 * int16x8_t vqshlq_s16 (int16x8_t, int16x8_t)
24407 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, Q0'
24409 * int8x16_t vqshlq_s8 (int8x16_t, int8x16_t)
24410 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, Q0'
24412 * uint64x2_t vqshlq_u64 (uint64x2_t, int64x2_t)
24413 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, Q0'
24415 * int64x2_t vqshlq_s64 (int64x2_t, int64x2_t)
24416 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, Q0'
24418 * uint32x2_t vqrshl_u32 (uint32x2_t, int32x2_t)
24419 _Form of expected instruction(s):_ `vqrshl.u32 D0, D0, D0'
24421 * uint16x4_t vqrshl_u16 (uint16x4_t, int16x4_t)
24422 _Form of expected instruction(s):_ `vqrshl.u16 D0, D0, D0'
24424 * uint8x8_t vqrshl_u8 (uint8x8_t, int8x8_t)
24425 _Form of expected instruction(s):_ `vqrshl.u8 D0, D0, D0'
24427 * int32x2_t vqrshl_s32 (int32x2_t, int32x2_t)
24428 _Form of expected instruction(s):_ `vqrshl.s32 D0, D0, D0'
24430 * int16x4_t vqrshl_s16 (int16x4_t, int16x4_t)
24431 _Form of expected instruction(s):_ `vqrshl.s16 D0, D0, D0'
24433 * int8x8_t vqrshl_s8 (int8x8_t, int8x8_t)
24434 _Form of expected instruction(s):_ `vqrshl.s8 D0, D0, D0'
24436 * uint64x1_t vqrshl_u64 (uint64x1_t, int64x1_t)
24437 _Form of expected instruction(s):_ `vqrshl.u64 D0, D0, D0'
24439 * int64x1_t vqrshl_s64 (int64x1_t, int64x1_t)
24440 _Form of expected instruction(s):_ `vqrshl.s64 D0, D0, D0'
24442 * uint32x4_t vqrshlq_u32 (uint32x4_t, int32x4_t)
24443 _Form of expected instruction(s):_ `vqrshl.u32 Q0, Q0, Q0'
24445 * uint16x8_t vqrshlq_u16 (uint16x8_t, int16x8_t)
24446 _Form of expected instruction(s):_ `vqrshl.u16 Q0, Q0, Q0'
24448 * uint8x16_t vqrshlq_u8 (uint8x16_t, int8x16_t)
24449 _Form of expected instruction(s):_ `vqrshl.u8 Q0, Q0, Q0'
24451 * int32x4_t vqrshlq_s32 (int32x4_t, int32x4_t)
24452 _Form of expected instruction(s):_ `vqrshl.s32 Q0, Q0, Q0'
24454 * int16x8_t vqrshlq_s16 (int16x8_t, int16x8_t)
24455 _Form of expected instruction(s):_ `vqrshl.s16 Q0, Q0, Q0'
24457 * int8x16_t vqrshlq_s8 (int8x16_t, int8x16_t)
24458 _Form of expected instruction(s):_ `vqrshl.s8 Q0, Q0, Q0'
24460 * uint64x2_t vqrshlq_u64 (uint64x2_t, int64x2_t)
24461 _Form of expected instruction(s):_ `vqrshl.u64 Q0, Q0, Q0'
24463 * int64x2_t vqrshlq_s64 (int64x2_t, int64x2_t)
24464 _Form of expected instruction(s):_ `vqrshl.s64 Q0, Q0, Q0'
24466 5.50.3.26 Vector shift left by constant
24467 .......................................
24469 * uint32x2_t vshl_n_u32 (uint32x2_t, const int)
24470 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
24472 * uint16x4_t vshl_n_u16 (uint16x4_t, const int)
24473 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
24475 * uint8x8_t vshl_n_u8 (uint8x8_t, const int)
24476 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
24478 * int32x2_t vshl_n_s32 (int32x2_t, const int)
24479 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
24481 * int16x4_t vshl_n_s16 (int16x4_t, const int)
24482 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
24484 * int8x8_t vshl_n_s8 (int8x8_t, const int)
24485 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
24487 * uint64x1_t vshl_n_u64 (uint64x1_t, const int)
24488 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
24490 * int64x1_t vshl_n_s64 (int64x1_t, const int)
24491 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
24493 * uint32x4_t vshlq_n_u32 (uint32x4_t, const int)
24494 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
24496 * uint16x8_t vshlq_n_u16 (uint16x8_t, const int)
24497 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
24499 * uint8x16_t vshlq_n_u8 (uint8x16_t, const int)
24500 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
24502 * int32x4_t vshlq_n_s32 (int32x4_t, const int)
24503 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
24505 * int16x8_t vshlq_n_s16 (int16x8_t, const int)
24506 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
24508 * int8x16_t vshlq_n_s8 (int8x16_t, const int)
24509 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
24511 * uint64x2_t vshlq_n_u64 (uint64x2_t, const int)
24512 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
24514 * int64x2_t vshlq_n_s64 (int64x2_t, const int)
24515 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
24517 * uint32x2_t vqshl_n_u32 (uint32x2_t, const int)
24518 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, #0'
24520 * uint16x4_t vqshl_n_u16 (uint16x4_t, const int)
24521 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, #0'
24523 * uint8x8_t vqshl_n_u8 (uint8x8_t, const int)
24524 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, #0'
24526 * int32x2_t vqshl_n_s32 (int32x2_t, const int)
24527 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, #0'
24529 * int16x4_t vqshl_n_s16 (int16x4_t, const int)
24530 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, #0'
24532 * int8x8_t vqshl_n_s8 (int8x8_t, const int)
24533 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, #0'
24535 * uint64x1_t vqshl_n_u64 (uint64x1_t, const int)
24536 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, #0'
24538 * int64x1_t vqshl_n_s64 (int64x1_t, const int)
24539 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, #0'
24541 * uint32x4_t vqshlq_n_u32 (uint32x4_t, const int)
24542 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, #0'
24544 * uint16x8_t vqshlq_n_u16 (uint16x8_t, const int)
24545 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, #0'
24547 * uint8x16_t vqshlq_n_u8 (uint8x16_t, const int)
24548 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, #0'
24550 * int32x4_t vqshlq_n_s32 (int32x4_t, const int)
24551 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, #0'
24553 * int16x8_t vqshlq_n_s16 (int16x8_t, const int)
24554 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, #0'
24556 * int8x16_t vqshlq_n_s8 (int8x16_t, const int)
24557 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, #0'
24559 * uint64x2_t vqshlq_n_u64 (uint64x2_t, const int)
24560 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, #0'
24562 * int64x2_t vqshlq_n_s64 (int64x2_t, const int)
24563 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, #0'
24565 * uint64x1_t vqshlu_n_s64 (int64x1_t, const int)
24566 _Form of expected instruction(s):_ `vqshlu.s64 D0, D0, #0'
24568 * uint32x2_t vqshlu_n_s32 (int32x2_t, const int)
24569 _Form of expected instruction(s):_ `vqshlu.s32 D0, D0, #0'
24571 * uint16x4_t vqshlu_n_s16 (int16x4_t, const int)
24572 _Form of expected instruction(s):_ `vqshlu.s16 D0, D0, #0'
24574 * uint8x8_t vqshlu_n_s8 (int8x8_t, const int)
24575 _Form of expected instruction(s):_ `vqshlu.s8 D0, D0, #0'
24577 * uint64x2_t vqshluq_n_s64 (int64x2_t, const int)
24578 _Form of expected instruction(s):_ `vqshlu.s64 Q0, Q0, #0'
24580 * uint32x4_t vqshluq_n_s32 (int32x4_t, const int)
24581 _Form of expected instruction(s):_ `vqshlu.s32 Q0, Q0, #0'
24583 * uint16x8_t vqshluq_n_s16 (int16x8_t, const int)
24584 _Form of expected instruction(s):_ `vqshlu.s16 Q0, Q0, #0'
24586 * uint8x16_t vqshluq_n_s8 (int8x16_t, const int)
24587 _Form of expected instruction(s):_ `vqshlu.s8 Q0, Q0, #0'
24589 * uint64x2_t vshll_n_u32 (uint32x2_t, const int)
24590 _Form of expected instruction(s):_ `vshll.u32 Q0, D0, #0'
24592 * uint32x4_t vshll_n_u16 (uint16x4_t, const int)
24593 _Form of expected instruction(s):_ `vshll.u16 Q0, D0, #0'
24595 * uint16x8_t vshll_n_u8 (uint8x8_t, const int)
24596 _Form of expected instruction(s):_ `vshll.u8 Q0, D0, #0'
24598 * int64x2_t vshll_n_s32 (int32x2_t, const int)
24599 _Form of expected instruction(s):_ `vshll.s32 Q0, D0, #0'
24601 * int32x4_t vshll_n_s16 (int16x4_t, const int)
24602 _Form of expected instruction(s):_ `vshll.s16 Q0, D0, #0'
24604 * int16x8_t vshll_n_s8 (int8x8_t, const int)
24605 _Form of expected instruction(s):_ `vshll.s8 Q0, D0, #0'
24607 5.50.3.27 Vector shift right by constant
24608 ........................................
24610 * uint32x2_t vshr_n_u32 (uint32x2_t, const int)
24611 _Form of expected instruction(s):_ `vshr.u32 D0, D0, #0'
24613 * uint16x4_t vshr_n_u16 (uint16x4_t, const int)
24614 _Form of expected instruction(s):_ `vshr.u16 D0, D0, #0'
24616 * uint8x8_t vshr_n_u8 (uint8x8_t, const int)
24617 _Form of expected instruction(s):_ `vshr.u8 D0, D0, #0'
24619 * int32x2_t vshr_n_s32 (int32x2_t, const int)
24620 _Form of expected instruction(s):_ `vshr.s32 D0, D0, #0'
24622 * int16x4_t vshr_n_s16 (int16x4_t, const int)
24623 _Form of expected instruction(s):_ `vshr.s16 D0, D0, #0'
24625 * int8x8_t vshr_n_s8 (int8x8_t, const int)
24626 _Form of expected instruction(s):_ `vshr.s8 D0, D0, #0'
24628 * uint64x1_t vshr_n_u64 (uint64x1_t, const int)
24629 _Form of expected instruction(s):_ `vshr.u64 D0, D0, #0'
24631 * int64x1_t vshr_n_s64 (int64x1_t, const int)
24632 _Form of expected instruction(s):_ `vshr.s64 D0, D0, #0'
24634 * uint32x4_t vshrq_n_u32 (uint32x4_t, const int)
24635 _Form of expected instruction(s):_ `vshr.u32 Q0, Q0, #0'
24637 * uint16x8_t vshrq_n_u16 (uint16x8_t, const int)
24638 _Form of expected instruction(s):_ `vshr.u16 Q0, Q0, #0'
24640 * uint8x16_t vshrq_n_u8 (uint8x16_t, const int)
24641 _Form of expected instruction(s):_ `vshr.u8 Q0, Q0, #0'
24643 * int32x4_t vshrq_n_s32 (int32x4_t, const int)
24644 _Form of expected instruction(s):_ `vshr.s32 Q0, Q0, #0'
24646 * int16x8_t vshrq_n_s16 (int16x8_t, const int)
24647 _Form of expected instruction(s):_ `vshr.s16 Q0, Q0, #0'
24649 * int8x16_t vshrq_n_s8 (int8x16_t, const int)
24650 _Form of expected instruction(s):_ `vshr.s8 Q0, Q0, #0'
24652 * uint64x2_t vshrq_n_u64 (uint64x2_t, const int)
24653 _Form of expected instruction(s):_ `vshr.u64 Q0, Q0, #0'
24655 * int64x2_t vshrq_n_s64 (int64x2_t, const int)
24656 _Form of expected instruction(s):_ `vshr.s64 Q0, Q0, #0'
24658 * uint32x2_t vrshr_n_u32 (uint32x2_t, const int)
24659 _Form of expected instruction(s):_ `vrshr.u32 D0, D0, #0'
24661 * uint16x4_t vrshr_n_u16 (uint16x4_t, const int)
24662 _Form of expected instruction(s):_ `vrshr.u16 D0, D0, #0'
24664 * uint8x8_t vrshr_n_u8 (uint8x8_t, const int)
24665 _Form of expected instruction(s):_ `vrshr.u8 D0, D0, #0'
24667 * int32x2_t vrshr_n_s32 (int32x2_t, const int)
24668 _Form of expected instruction(s):_ `vrshr.s32 D0, D0, #0'
24670 * int16x4_t vrshr_n_s16 (int16x4_t, const int)
24671 _Form of expected instruction(s):_ `vrshr.s16 D0, D0, #0'
24673 * int8x8_t vrshr_n_s8 (int8x8_t, const int)
24674 _Form of expected instruction(s):_ `vrshr.s8 D0, D0, #0'
24676 * uint64x1_t vrshr_n_u64 (uint64x1_t, const int)
24677 _Form of expected instruction(s):_ `vrshr.u64 D0, D0, #0'
24679 * int64x1_t vrshr_n_s64 (int64x1_t, const int)
24680 _Form of expected instruction(s):_ `vrshr.s64 D0, D0, #0'
24682 * uint32x4_t vrshrq_n_u32 (uint32x4_t, const int)
24683 _Form of expected instruction(s):_ `vrshr.u32 Q0, Q0, #0'
24685 * uint16x8_t vrshrq_n_u16 (uint16x8_t, const int)
24686 _Form of expected instruction(s):_ `vrshr.u16 Q0, Q0, #0'
24688 * uint8x16_t vrshrq_n_u8 (uint8x16_t, const int)
24689 _Form of expected instruction(s):_ `vrshr.u8 Q0, Q0, #0'
24691 * int32x4_t vrshrq_n_s32 (int32x4_t, const int)
24692 _Form of expected instruction(s):_ `vrshr.s32 Q0, Q0, #0'
24694 * int16x8_t vrshrq_n_s16 (int16x8_t, const int)
24695 _Form of expected instruction(s):_ `vrshr.s16 Q0, Q0, #0'
24697 * int8x16_t vrshrq_n_s8 (int8x16_t, const int)
24698 _Form of expected instruction(s):_ `vrshr.s8 Q0, Q0, #0'
24700 * uint64x2_t vrshrq_n_u64 (uint64x2_t, const int)
24701 _Form of expected instruction(s):_ `vrshr.u64 Q0, Q0, #0'
24703 * int64x2_t vrshrq_n_s64 (int64x2_t, const int)
24704 _Form of expected instruction(s):_ `vrshr.s64 Q0, Q0, #0'
24706 * uint32x2_t vshrn_n_u64 (uint64x2_t, const int)
24707 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
24709 * uint16x4_t vshrn_n_u32 (uint32x4_t, const int)
24710 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
24712 * uint8x8_t vshrn_n_u16 (uint16x8_t, const int)
24713 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
24715 * int32x2_t vshrn_n_s64 (int64x2_t, const int)
24716 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
24718 * int16x4_t vshrn_n_s32 (int32x4_t, const int)
24719 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
24721 * int8x8_t vshrn_n_s16 (int16x8_t, const int)
24722 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
24724 * uint32x2_t vrshrn_n_u64 (uint64x2_t, const int)
24725 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
24727 * uint16x4_t vrshrn_n_u32 (uint32x4_t, const int)
24728 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
24730 * uint8x8_t vrshrn_n_u16 (uint16x8_t, const int)
24731 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
24733 * int32x2_t vrshrn_n_s64 (int64x2_t, const int)
24734 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
24736 * int16x4_t vrshrn_n_s32 (int32x4_t, const int)
24737 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
24739 * int8x8_t vrshrn_n_s16 (int16x8_t, const int)
24740 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
24742 * uint32x2_t vqshrn_n_u64 (uint64x2_t, const int)
24743 _Form of expected instruction(s):_ `vqshrn.u64 D0, Q0, #0'
24745 * uint16x4_t vqshrn_n_u32 (uint32x4_t, const int)
24746 _Form of expected instruction(s):_ `vqshrn.u32 D0, Q0, #0'
24748 * uint8x8_t vqshrn_n_u16 (uint16x8_t, const int)
24749 _Form of expected instruction(s):_ `vqshrn.u16 D0, Q0, #0'
24751 * int32x2_t vqshrn_n_s64 (int64x2_t, const int)
24752 _Form of expected instruction(s):_ `vqshrn.s64 D0, Q0, #0'
24754 * int16x4_t vqshrn_n_s32 (int32x4_t, const int)
24755 _Form of expected instruction(s):_ `vqshrn.s32 D0, Q0, #0'
24757 * int8x8_t vqshrn_n_s16 (int16x8_t, const int)
24758 _Form of expected instruction(s):_ `vqshrn.s16 D0, Q0, #0'
24760 * uint32x2_t vqrshrn_n_u64 (uint64x2_t, const int)
24761 _Form of expected instruction(s):_ `vqrshrn.u64 D0, Q0, #0'
24763 * uint16x4_t vqrshrn_n_u32 (uint32x4_t, const int)
24764 _Form of expected instruction(s):_ `vqrshrn.u32 D0, Q0, #0'
24766 * uint8x8_t vqrshrn_n_u16 (uint16x8_t, const int)
24767 _Form of expected instruction(s):_ `vqrshrn.u16 D0, Q0, #0'
24769 * int32x2_t vqrshrn_n_s64 (int64x2_t, const int)
24770 _Form of expected instruction(s):_ `vqrshrn.s64 D0, Q0, #0'
24772 * int16x4_t vqrshrn_n_s32 (int32x4_t, const int)
24773 _Form of expected instruction(s):_ `vqrshrn.s32 D0, Q0, #0'
24775 * int8x8_t vqrshrn_n_s16 (int16x8_t, const int)
24776 _Form of expected instruction(s):_ `vqrshrn.s16 D0, Q0, #0'
24778 * uint32x2_t vqshrun_n_s64 (int64x2_t, const int)
24779 _Form of expected instruction(s):_ `vqshrun.s64 D0, Q0, #0'
24781 * uint16x4_t vqshrun_n_s32 (int32x4_t, const int)
24782 _Form of expected instruction(s):_ `vqshrun.s32 D0, Q0, #0'
24784 * uint8x8_t vqshrun_n_s16 (int16x8_t, const int)
24785 _Form of expected instruction(s):_ `vqshrun.s16 D0, Q0, #0'
24787 * uint32x2_t vqrshrun_n_s64 (int64x2_t, const int)
24788 _Form of expected instruction(s):_ `vqrshrun.s64 D0, Q0, #0'
24790 * uint16x4_t vqrshrun_n_s32 (int32x4_t, const int)
24791 _Form of expected instruction(s):_ `vqrshrun.s32 D0, Q0, #0'
24793 * uint8x8_t vqrshrun_n_s16 (int16x8_t, const int)
24794 _Form of expected instruction(s):_ `vqrshrun.s16 D0, Q0, #0'
24796 5.50.3.28 Vector shift right by constant and accumulate
24797 .......................................................
24799 * uint32x2_t vsra_n_u32 (uint32x2_t, uint32x2_t, const int)
24800 _Form of expected instruction(s):_ `vsra.u32 D0, D0, #0'
24802 * uint16x4_t vsra_n_u16 (uint16x4_t, uint16x4_t, const int)
24803 _Form of expected instruction(s):_ `vsra.u16 D0, D0, #0'
24805 * uint8x8_t vsra_n_u8 (uint8x8_t, uint8x8_t, const int)
24806 _Form of expected instruction(s):_ `vsra.u8 D0, D0, #0'
24808 * int32x2_t vsra_n_s32 (int32x2_t, int32x2_t, const int)
24809 _Form of expected instruction(s):_ `vsra.s32 D0, D0, #0'
24811 * int16x4_t vsra_n_s16 (int16x4_t, int16x4_t, const int)
24812 _Form of expected instruction(s):_ `vsra.s16 D0, D0, #0'
24814 * int8x8_t vsra_n_s8 (int8x8_t, int8x8_t, const int)
24815 _Form of expected instruction(s):_ `vsra.s8 D0, D0, #0'
24817 * uint64x1_t vsra_n_u64 (uint64x1_t, uint64x1_t, const int)
24818 _Form of expected instruction(s):_ `vsra.u64 D0, D0, #0'
24820 * int64x1_t vsra_n_s64 (int64x1_t, int64x1_t, const int)
24821 _Form of expected instruction(s):_ `vsra.s64 D0, D0, #0'
24823 * uint32x4_t vsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
24824 _Form of expected instruction(s):_ `vsra.u32 Q0, Q0, #0'
24826 * uint16x8_t vsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
24827 _Form of expected instruction(s):_ `vsra.u16 Q0, Q0, #0'
24829 * uint8x16_t vsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
24830 _Form of expected instruction(s):_ `vsra.u8 Q0, Q0, #0'
24832 * int32x4_t vsraq_n_s32 (int32x4_t, int32x4_t, const int)
24833 _Form of expected instruction(s):_ `vsra.s32 Q0, Q0, #0'
24835 * int16x8_t vsraq_n_s16 (int16x8_t, int16x8_t, const int)
24836 _Form of expected instruction(s):_ `vsra.s16 Q0, Q0, #0'
24838 * int8x16_t vsraq_n_s8 (int8x16_t, int8x16_t, const int)
24839 _Form of expected instruction(s):_ `vsra.s8 Q0, Q0, #0'
24841 * uint64x2_t vsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
24842 _Form of expected instruction(s):_ `vsra.u64 Q0, Q0, #0'
24844 * int64x2_t vsraq_n_s64 (int64x2_t, int64x2_t, const int)
24845 _Form of expected instruction(s):_ `vsra.s64 Q0, Q0, #0'
24847 * uint32x2_t vrsra_n_u32 (uint32x2_t, uint32x2_t, const int)
24848 _Form of expected instruction(s):_ `vrsra.u32 D0, D0, #0'
24850 * uint16x4_t vrsra_n_u16 (uint16x4_t, uint16x4_t, const int)
24851 _Form of expected instruction(s):_ `vrsra.u16 D0, D0, #0'
24853 * uint8x8_t vrsra_n_u8 (uint8x8_t, uint8x8_t, const int)
24854 _Form of expected instruction(s):_ `vrsra.u8 D0, D0, #0'
24856 * int32x2_t vrsra_n_s32 (int32x2_t, int32x2_t, const int)
24857 _Form of expected instruction(s):_ `vrsra.s32 D0, D0, #0'
24859 * int16x4_t vrsra_n_s16 (int16x4_t, int16x4_t, const int)
24860 _Form of expected instruction(s):_ `vrsra.s16 D0, D0, #0'
24862 * int8x8_t vrsra_n_s8 (int8x8_t, int8x8_t, const int)
24863 _Form of expected instruction(s):_ `vrsra.s8 D0, D0, #0'
24865 * uint64x1_t vrsra_n_u64 (uint64x1_t, uint64x1_t, const int)
24866 _Form of expected instruction(s):_ `vrsra.u64 D0, D0, #0'
24868 * int64x1_t vrsra_n_s64 (int64x1_t, int64x1_t, const int)
24869 _Form of expected instruction(s):_ `vrsra.s64 D0, D0, #0'
24871 * uint32x4_t vrsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
24872 _Form of expected instruction(s):_ `vrsra.u32 Q0, Q0, #0'
24874 * uint16x8_t vrsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
24875 _Form of expected instruction(s):_ `vrsra.u16 Q0, Q0, #0'
24877 * uint8x16_t vrsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
24878 _Form of expected instruction(s):_ `vrsra.u8 Q0, Q0, #0'
24880 * int32x4_t vrsraq_n_s32 (int32x4_t, int32x4_t, const int)
24881 _Form of expected instruction(s):_ `vrsra.s32 Q0, Q0, #0'
24883 * int16x8_t vrsraq_n_s16 (int16x8_t, int16x8_t, const int)
24884 _Form of expected instruction(s):_ `vrsra.s16 Q0, Q0, #0'
24886 * int8x16_t vrsraq_n_s8 (int8x16_t, int8x16_t, const int)
24887 _Form of expected instruction(s):_ `vrsra.s8 Q0, Q0, #0'
24889 * uint64x2_t vrsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
24890 _Form of expected instruction(s):_ `vrsra.u64 Q0, Q0, #0'
24892 * int64x2_t vrsraq_n_s64 (int64x2_t, int64x2_t, const int)
24893 _Form of expected instruction(s):_ `vrsra.s64 Q0, Q0, #0'
24895 5.50.3.29 Vector shift right and insert
24896 .......................................
24898 * uint32x2_t vsri_n_u32 (uint32x2_t, uint32x2_t, const int)
24899 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
24901 * uint16x4_t vsri_n_u16 (uint16x4_t, uint16x4_t, const int)
24902 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
24904 * uint8x8_t vsri_n_u8 (uint8x8_t, uint8x8_t, const int)
24905 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
24907 * int32x2_t vsri_n_s32 (int32x2_t, int32x2_t, const int)
24908 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
24910 * int16x4_t vsri_n_s16 (int16x4_t, int16x4_t, const int)
24911 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
24913 * int8x8_t vsri_n_s8 (int8x8_t, int8x8_t, const int)
24914 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
24916 * uint64x1_t vsri_n_u64 (uint64x1_t, uint64x1_t, const int)
24917 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
24919 * int64x1_t vsri_n_s64 (int64x1_t, int64x1_t, const int)
24920 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
24922 * poly16x4_t vsri_n_p16 (poly16x4_t, poly16x4_t, const int)
24923 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
24925 * poly8x8_t vsri_n_p8 (poly8x8_t, poly8x8_t, const int)
24926 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
24928 * uint32x4_t vsriq_n_u32 (uint32x4_t, uint32x4_t, const int)
24929 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
24931 * uint16x8_t vsriq_n_u16 (uint16x8_t, uint16x8_t, const int)
24932 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
24934 * uint8x16_t vsriq_n_u8 (uint8x16_t, uint8x16_t, const int)
24935 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
24937 * int32x4_t vsriq_n_s32 (int32x4_t, int32x4_t, const int)
24938 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
24940 * int16x8_t vsriq_n_s16 (int16x8_t, int16x8_t, const int)
24941 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
24943 * int8x16_t vsriq_n_s8 (int8x16_t, int8x16_t, const int)
24944 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
24946 * uint64x2_t vsriq_n_u64 (uint64x2_t, uint64x2_t, const int)
24947 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
24949 * int64x2_t vsriq_n_s64 (int64x2_t, int64x2_t, const int)
24950 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
24952 * poly16x8_t vsriq_n_p16 (poly16x8_t, poly16x8_t, const int)
24953 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
24955 * poly8x16_t vsriq_n_p8 (poly8x16_t, poly8x16_t, const int)
24956 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
24958 5.50.3.30 Vector shift left and insert
24959 ......................................
24961 * uint32x2_t vsli_n_u32 (uint32x2_t, uint32x2_t, const int)
24962 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
24964 * uint16x4_t vsli_n_u16 (uint16x4_t, uint16x4_t, const int)
24965 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
24967 * uint8x8_t vsli_n_u8 (uint8x8_t, uint8x8_t, const int)
24968 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
24970 * int32x2_t vsli_n_s32 (int32x2_t, int32x2_t, const int)
24971 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
24973 * int16x4_t vsli_n_s16 (int16x4_t, int16x4_t, const int)
24974 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
24976 * int8x8_t vsli_n_s8 (int8x8_t, int8x8_t, const int)
24977 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
24979 * uint64x1_t vsli_n_u64 (uint64x1_t, uint64x1_t, const int)
24980 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
24982 * int64x1_t vsli_n_s64 (int64x1_t, int64x1_t, const int)
24983 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
24985 * poly16x4_t vsli_n_p16 (poly16x4_t, poly16x4_t, const int)
24986 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
24988 * poly8x8_t vsli_n_p8 (poly8x8_t, poly8x8_t, const int)
24989 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
24991 * uint32x4_t vsliq_n_u32 (uint32x4_t, uint32x4_t, const int)
24992 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
24994 * uint16x8_t vsliq_n_u16 (uint16x8_t, uint16x8_t, const int)
24995 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
24997 * uint8x16_t vsliq_n_u8 (uint8x16_t, uint8x16_t, const int)
24998 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
25000 * int32x4_t vsliq_n_s32 (int32x4_t, int32x4_t, const int)
25001 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
25003 * int16x8_t vsliq_n_s16 (int16x8_t, int16x8_t, const int)
25004 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
25006 * int8x16_t vsliq_n_s8 (int8x16_t, int8x16_t, const int)
25007 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
25009 * uint64x2_t vsliq_n_u64 (uint64x2_t, uint64x2_t, const int)
25010 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
25012 * int64x2_t vsliq_n_s64 (int64x2_t, int64x2_t, const int)
25013 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
25015 * poly16x8_t vsliq_n_p16 (poly16x8_t, poly16x8_t, const int)
25016 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
25018 * poly8x16_t vsliq_n_p8 (poly8x16_t, poly8x16_t, const int)
25019 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
25021 5.50.3.31 Absolute value
25022 ........................
25024 * float32x2_t vabs_f32 (float32x2_t)
25025 _Form of expected instruction(s):_ `vabs.f32 D0, D0'
25027 * int32x2_t vabs_s32 (int32x2_t)
25028 _Form of expected instruction(s):_ `vabs.s32 D0, D0'
25030 * int16x4_t vabs_s16 (int16x4_t)
25031 _Form of expected instruction(s):_ `vabs.s16 D0, D0'
25033 * int8x8_t vabs_s8 (int8x8_t)
25034 _Form of expected instruction(s):_ `vabs.s8 D0, D0'
25036 * float32x4_t vabsq_f32 (float32x4_t)
25037 _Form of expected instruction(s):_ `vabs.f32 Q0, Q0'
25039 * int32x4_t vabsq_s32 (int32x4_t)
25040 _Form of expected instruction(s):_ `vabs.s32 Q0, Q0'
25042 * int16x8_t vabsq_s16 (int16x8_t)
25043 _Form of expected instruction(s):_ `vabs.s16 Q0, Q0'
25045 * int8x16_t vabsq_s8 (int8x16_t)
25046 _Form of expected instruction(s):_ `vabs.s8 Q0, Q0'
25048 * int32x2_t vqabs_s32 (int32x2_t)
25049 _Form of expected instruction(s):_ `vqabs.s32 D0, D0'
25051 * int16x4_t vqabs_s16 (int16x4_t)
25052 _Form of expected instruction(s):_ `vqabs.s16 D0, D0'
25054 * int8x8_t vqabs_s8 (int8x8_t)
25055 _Form of expected instruction(s):_ `vqabs.s8 D0, D0'
25057 * int32x4_t vqabsq_s32 (int32x4_t)
25058 _Form of expected instruction(s):_ `vqabs.s32 Q0, Q0'
25060 * int16x8_t vqabsq_s16 (int16x8_t)
25061 _Form of expected instruction(s):_ `vqabs.s16 Q0, Q0'
25063 * int8x16_t vqabsq_s8 (int8x16_t)
25064 _Form of expected instruction(s):_ `vqabs.s8 Q0, Q0'
25069 * float32x2_t vneg_f32 (float32x2_t)
25070 _Form of expected instruction(s):_ `vneg.f32 D0, D0'
25072 * int32x2_t vneg_s32 (int32x2_t)
25073 _Form of expected instruction(s):_ `vneg.s32 D0, D0'
25075 * int16x4_t vneg_s16 (int16x4_t)
25076 _Form of expected instruction(s):_ `vneg.s16 D0, D0'
25078 * int8x8_t vneg_s8 (int8x8_t)
25079 _Form of expected instruction(s):_ `vneg.s8 D0, D0'
25081 * float32x4_t vnegq_f32 (float32x4_t)
25082 _Form of expected instruction(s):_ `vneg.f32 Q0, Q0'
25084 * int32x4_t vnegq_s32 (int32x4_t)
25085 _Form of expected instruction(s):_ `vneg.s32 Q0, Q0'
25087 * int16x8_t vnegq_s16 (int16x8_t)
25088 _Form of expected instruction(s):_ `vneg.s16 Q0, Q0'
25090 * int8x16_t vnegq_s8 (int8x16_t)
25091 _Form of expected instruction(s):_ `vneg.s8 Q0, Q0'
25093 * int32x2_t vqneg_s32 (int32x2_t)
25094 _Form of expected instruction(s):_ `vqneg.s32 D0, D0'
25096 * int16x4_t vqneg_s16 (int16x4_t)
25097 _Form of expected instruction(s):_ `vqneg.s16 D0, D0'
25099 * int8x8_t vqneg_s8 (int8x8_t)
25100 _Form of expected instruction(s):_ `vqneg.s8 D0, D0'
25102 * int32x4_t vqnegq_s32 (int32x4_t)
25103 _Form of expected instruction(s):_ `vqneg.s32 Q0, Q0'
25105 * int16x8_t vqnegq_s16 (int16x8_t)
25106 _Form of expected instruction(s):_ `vqneg.s16 Q0, Q0'
25108 * int8x16_t vqnegq_s8 (int8x16_t)
25109 _Form of expected instruction(s):_ `vqneg.s8 Q0, Q0'
25111 5.50.3.33 Bitwise not
25112 .....................
25114 * uint32x2_t vmvn_u32 (uint32x2_t)
25115 _Form of expected instruction(s):_ `vmvn D0, D0'
25117 * uint16x4_t vmvn_u16 (uint16x4_t)
25118 _Form of expected instruction(s):_ `vmvn D0, D0'
25120 * uint8x8_t vmvn_u8 (uint8x8_t)
25121 _Form of expected instruction(s):_ `vmvn D0, D0'
25123 * int32x2_t vmvn_s32 (int32x2_t)
25124 _Form of expected instruction(s):_ `vmvn D0, D0'
25126 * int16x4_t vmvn_s16 (int16x4_t)
25127 _Form of expected instruction(s):_ `vmvn D0, D0'
25129 * int8x8_t vmvn_s8 (int8x8_t)
25130 _Form of expected instruction(s):_ `vmvn D0, D0'
25132 * poly8x8_t vmvn_p8 (poly8x8_t)
25133 _Form of expected instruction(s):_ `vmvn D0, D0'
25135 * uint32x4_t vmvnq_u32 (uint32x4_t)
25136 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25138 * uint16x8_t vmvnq_u16 (uint16x8_t)
25139 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25141 * uint8x16_t vmvnq_u8 (uint8x16_t)
25142 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25144 * int32x4_t vmvnq_s32 (int32x4_t)
25145 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25147 * int16x8_t vmvnq_s16 (int16x8_t)
25148 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25150 * int8x16_t vmvnq_s8 (int8x16_t)
25151 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25153 * poly8x16_t vmvnq_p8 (poly8x16_t)
25154 _Form of expected instruction(s):_ `vmvn Q0, Q0'
25156 5.50.3.34 Count leading sign bits
25157 .................................
25159 * int32x2_t vcls_s32 (int32x2_t)
25160 _Form of expected instruction(s):_ `vcls.s32 D0, D0'
25162 * int16x4_t vcls_s16 (int16x4_t)
25163 _Form of expected instruction(s):_ `vcls.s16 D0, D0'
25165 * int8x8_t vcls_s8 (int8x8_t)
25166 _Form of expected instruction(s):_ `vcls.s8 D0, D0'
25168 * int32x4_t vclsq_s32 (int32x4_t)
25169 _Form of expected instruction(s):_ `vcls.s32 Q0, Q0'
25171 * int16x8_t vclsq_s16 (int16x8_t)
25172 _Form of expected instruction(s):_ `vcls.s16 Q0, Q0'
25174 * int8x16_t vclsq_s8 (int8x16_t)
25175 _Form of expected instruction(s):_ `vcls.s8 Q0, Q0'
25177 5.50.3.35 Count leading zeros
25178 .............................
25180 * uint32x2_t vclz_u32 (uint32x2_t)
25181 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
25183 * uint16x4_t vclz_u16 (uint16x4_t)
25184 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
25186 * uint8x8_t vclz_u8 (uint8x8_t)
25187 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
25189 * int32x2_t vclz_s32 (int32x2_t)
25190 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
25192 * int16x4_t vclz_s16 (int16x4_t)
25193 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
25195 * int8x8_t vclz_s8 (int8x8_t)
25196 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
25198 * uint32x4_t vclzq_u32 (uint32x4_t)
25199 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
25201 * uint16x8_t vclzq_u16 (uint16x8_t)
25202 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
25204 * uint8x16_t vclzq_u8 (uint8x16_t)
25205 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
25207 * int32x4_t vclzq_s32 (int32x4_t)
25208 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
25210 * int16x8_t vclzq_s16 (int16x8_t)
25211 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
25213 * int8x16_t vclzq_s8 (int8x16_t)
25214 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
25216 5.50.3.36 Count number of set bits
25217 ..................................
25219 * uint8x8_t vcnt_u8 (uint8x8_t)
25220 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
25222 * int8x8_t vcnt_s8 (int8x8_t)
25223 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
25225 * poly8x8_t vcnt_p8 (poly8x8_t)
25226 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
25228 * uint8x16_t vcntq_u8 (uint8x16_t)
25229 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
25231 * int8x16_t vcntq_s8 (int8x16_t)
25232 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
25234 * poly8x16_t vcntq_p8 (poly8x16_t)
25235 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
25237 5.50.3.37 Reciprocal estimate
25238 .............................
25240 * float32x2_t vrecpe_f32 (float32x2_t)
25241 _Form of expected instruction(s):_ `vrecpe.f32 D0, D0'
25243 * uint32x2_t vrecpe_u32 (uint32x2_t)
25244 _Form of expected instruction(s):_ `vrecpe.u32 D0, D0'
25246 * float32x4_t vrecpeq_f32 (float32x4_t)
25247 _Form of expected instruction(s):_ `vrecpe.f32 Q0, Q0'
25249 * uint32x4_t vrecpeq_u32 (uint32x4_t)
25250 _Form of expected instruction(s):_ `vrecpe.u32 Q0, Q0'
25252 5.50.3.38 Reciprocal square-root estimate
25253 .........................................
25255 * float32x2_t vrsqrte_f32 (float32x2_t)
25256 _Form of expected instruction(s):_ `vrsqrte.f32 D0, D0'
25258 * uint32x2_t vrsqrte_u32 (uint32x2_t)
25259 _Form of expected instruction(s):_ `vrsqrte.u32 D0, D0'
25261 * float32x4_t vrsqrteq_f32 (float32x4_t)
25262 _Form of expected instruction(s):_ `vrsqrte.f32 Q0, Q0'
25264 * uint32x4_t vrsqrteq_u32 (uint32x4_t)
25265 _Form of expected instruction(s):_ `vrsqrte.u32 Q0, Q0'
25267 5.50.3.39 Get lanes from a vector
25268 .................................
25270 * uint32_t vget_lane_u32 (uint32x2_t, const int)
25271 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
25273 * uint16_t vget_lane_u16 (uint16x4_t, const int)
25274 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
25276 * uint8_t vget_lane_u8 (uint8x8_t, const int)
25277 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
25279 * int32_t vget_lane_s32 (int32x2_t, const int)
25280 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
25282 * int16_t vget_lane_s16 (int16x4_t, const int)
25283 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
25285 * int8_t vget_lane_s8 (int8x8_t, const int)
25286 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
25288 * float32_t vget_lane_f32 (float32x2_t, const int)
25289 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
25291 * poly16_t vget_lane_p16 (poly16x4_t, const int)
25292 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
25294 * poly8_t vget_lane_p8 (poly8x8_t, const int)
25295 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
25297 * uint64_t vget_lane_u64 (uint64x1_t, const int)
25298 _Form of expected instruction(s):_ `vmov R0, R0, D0'
25300 * int64_t vget_lane_s64 (int64x1_t, const int)
25301 _Form of expected instruction(s):_ `vmov R0, R0, D0'
25303 * uint32_t vgetq_lane_u32 (uint32x4_t, const int)
25304 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
25306 * uint16_t vgetq_lane_u16 (uint16x8_t, const int)
25307 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
25309 * uint8_t vgetq_lane_u8 (uint8x16_t, const int)
25310 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
25312 * int32_t vgetq_lane_s32 (int32x4_t, const int)
25313 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
25315 * int16_t vgetq_lane_s16 (int16x8_t, const int)
25316 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
25318 * int8_t vgetq_lane_s8 (int8x16_t, const int)
25319 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
25321 * float32_t vgetq_lane_f32 (float32x4_t, const int)
25322 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
25324 * poly16_t vgetq_lane_p16 (poly16x8_t, const int)
25325 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
25327 * poly8_t vgetq_lane_p8 (poly8x16_t, const int)
25328 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
25330 * uint64_t vgetq_lane_u64 (uint64x2_t, const int)
25331 _Form of expected instruction(s):_ `vmov R0, R0, D0'
25333 * int64_t vgetq_lane_s64 (int64x2_t, const int)
25334 _Form of expected instruction(s):_ `vmov R0, R0, D0'
25336 5.50.3.40 Set lanes in a vector
25337 ...............................
25339 * uint32x2_t vset_lane_u32 (uint32_t, uint32x2_t, const int)
25340 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25342 * uint16x4_t vset_lane_u16 (uint16_t, uint16x4_t, const int)
25343 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25345 * uint8x8_t vset_lane_u8 (uint8_t, uint8x8_t, const int)
25346 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25348 * int32x2_t vset_lane_s32 (int32_t, int32x2_t, const int)
25349 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25351 * int16x4_t vset_lane_s16 (int16_t, int16x4_t, const int)
25352 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25354 * int8x8_t vset_lane_s8 (int8_t, int8x8_t, const int)
25355 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25357 * float32x2_t vset_lane_f32 (float32_t, float32x2_t, const int)
25358 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25360 * poly16x4_t vset_lane_p16 (poly16_t, poly16x4_t, const int)
25361 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25363 * poly8x8_t vset_lane_p8 (poly8_t, poly8x8_t, const int)
25364 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25366 * uint64x1_t vset_lane_u64 (uint64_t, uint64x1_t, const int)
25367 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25369 * int64x1_t vset_lane_s64 (int64_t, int64x1_t, const int)
25370 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25372 * uint32x4_t vsetq_lane_u32 (uint32_t, uint32x4_t, const int)
25373 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25375 * uint16x8_t vsetq_lane_u16 (uint16_t, uint16x8_t, const int)
25376 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25378 * uint8x16_t vsetq_lane_u8 (uint8_t, uint8x16_t, const int)
25379 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25381 * int32x4_t vsetq_lane_s32 (int32_t, int32x4_t, const int)
25382 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25384 * int16x8_t vsetq_lane_s16 (int16_t, int16x8_t, const int)
25385 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25387 * int8x16_t vsetq_lane_s8 (int8_t, int8x16_t, const int)
25388 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25390 * float32x4_t vsetq_lane_f32 (float32_t, float32x4_t, const int)
25391 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
25393 * poly16x8_t vsetq_lane_p16 (poly16_t, poly16x8_t, const int)
25394 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
25396 * poly8x16_t vsetq_lane_p8 (poly8_t, poly8x16_t, const int)
25397 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
25399 * uint64x2_t vsetq_lane_u64 (uint64_t, uint64x2_t, const int)
25400 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25402 * int64x2_t vsetq_lane_s64 (int64_t, int64x2_t, const int)
25403 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25405 5.50.3.41 Create vector from literal bit pattern
25406 ................................................
25408 * uint32x2_t vcreate_u32 (uint64_t)
25410 * uint16x4_t vcreate_u16 (uint64_t)
25412 * uint8x8_t vcreate_u8 (uint64_t)
25414 * int32x2_t vcreate_s32 (uint64_t)
25416 * int16x4_t vcreate_s16 (uint64_t)
25418 * int8x8_t vcreate_s8 (uint64_t)
25420 * uint64x1_t vcreate_u64 (uint64_t)
25422 * int64x1_t vcreate_s64 (uint64_t)
25424 * float32x2_t vcreate_f32 (uint64_t)
25426 * poly16x4_t vcreate_p16 (uint64_t)
25428 * poly8x8_t vcreate_p8 (uint64_t)
25430 5.50.3.42 Set all lanes to the same value
25431 .........................................
25433 * uint32x2_t vdup_n_u32 (uint32_t)
25434 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25436 * uint16x4_t vdup_n_u16 (uint16_t)
25437 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25439 * uint8x8_t vdup_n_u8 (uint8_t)
25440 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25442 * int32x2_t vdup_n_s32 (int32_t)
25443 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25445 * int16x4_t vdup_n_s16 (int16_t)
25446 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25448 * int8x8_t vdup_n_s8 (int8_t)
25449 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25451 * float32x2_t vdup_n_f32 (float32_t)
25452 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25454 * poly16x4_t vdup_n_p16 (poly16_t)
25455 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25457 * poly8x8_t vdup_n_p8 (poly8_t)
25458 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25460 * uint64x1_t vdup_n_u64 (uint64_t)
25461 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25463 * int64x1_t vdup_n_s64 (int64_t)
25464 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25466 * uint32x4_t vdupq_n_u32 (uint32_t)
25467 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25469 * uint16x8_t vdupq_n_u16 (uint16_t)
25470 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25472 * uint8x16_t vdupq_n_u8 (uint8_t)
25473 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25475 * int32x4_t vdupq_n_s32 (int32_t)
25476 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25478 * int16x8_t vdupq_n_s16 (int16_t)
25479 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25481 * int8x16_t vdupq_n_s8 (int8_t)
25482 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25484 * float32x4_t vdupq_n_f32 (float32_t)
25485 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25487 * poly16x8_t vdupq_n_p16 (poly16_t)
25488 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25490 * poly8x16_t vdupq_n_p8 (poly8_t)
25491 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25493 * uint64x2_t vdupq_n_u64 (uint64_t)
25494 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25496 * int64x2_t vdupq_n_s64 (int64_t)
25497 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25499 * uint32x2_t vmov_n_u32 (uint32_t)
25500 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25502 * uint16x4_t vmov_n_u16 (uint16_t)
25503 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25505 * uint8x8_t vmov_n_u8 (uint8_t)
25506 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25508 * int32x2_t vmov_n_s32 (int32_t)
25509 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25511 * int16x4_t vmov_n_s16 (int16_t)
25512 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25514 * int8x8_t vmov_n_s8 (int8_t)
25515 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25517 * float32x2_t vmov_n_f32 (float32_t)
25518 _Form of expected instruction(s):_ `vdup.32 D0, R0'
25520 * poly16x4_t vmov_n_p16 (poly16_t)
25521 _Form of expected instruction(s):_ `vdup.16 D0, R0'
25523 * poly8x8_t vmov_n_p8 (poly8_t)
25524 _Form of expected instruction(s):_ `vdup.8 D0, R0'
25526 * uint64x1_t vmov_n_u64 (uint64_t)
25527 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25529 * int64x1_t vmov_n_s64 (int64_t)
25530 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25532 * uint32x4_t vmovq_n_u32 (uint32_t)
25533 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25535 * uint16x8_t vmovq_n_u16 (uint16_t)
25536 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25538 * uint8x16_t vmovq_n_u8 (uint8_t)
25539 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25541 * int32x4_t vmovq_n_s32 (int32_t)
25542 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25544 * int16x8_t vmovq_n_s16 (int16_t)
25545 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25547 * int8x16_t vmovq_n_s8 (int8_t)
25548 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25550 * float32x4_t vmovq_n_f32 (float32_t)
25551 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
25553 * poly16x8_t vmovq_n_p16 (poly16_t)
25554 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
25556 * poly8x16_t vmovq_n_p8 (poly8_t)
25557 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
25559 * uint64x2_t vmovq_n_u64 (uint64_t)
25560 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25562 * int64x2_t vmovq_n_s64 (int64_t)
25563 _Form of expected instruction(s):_ `vmov D0, R0, R0'
25565 * uint32x2_t vdup_lane_u32 (uint32x2_t, const int)
25566 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
25568 * uint16x4_t vdup_lane_u16 (uint16x4_t, const int)
25569 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
25571 * uint8x8_t vdup_lane_u8 (uint8x8_t, const int)
25572 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
25574 * int32x2_t vdup_lane_s32 (int32x2_t, const int)
25575 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
25577 * int16x4_t vdup_lane_s16 (int16x4_t, const int)
25578 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
25580 * int8x8_t vdup_lane_s8 (int8x8_t, const int)
25581 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
25583 * float32x2_t vdup_lane_f32 (float32x2_t, const int)
25584 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
25586 * poly16x4_t vdup_lane_p16 (poly16x4_t, const int)
25587 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
25589 * poly8x8_t vdup_lane_p8 (poly8x8_t, const int)
25590 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
25592 * uint64x1_t vdup_lane_u64 (uint64x1_t, const int)
25594 * int64x1_t vdup_lane_s64 (int64x1_t, const int)
25596 * uint32x4_t vdupq_lane_u32 (uint32x2_t, const int)
25597 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
25599 * uint16x8_t vdupq_lane_u16 (uint16x4_t, const int)
25600 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
25602 * uint8x16_t vdupq_lane_u8 (uint8x8_t, const int)
25603 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
25605 * int32x4_t vdupq_lane_s32 (int32x2_t, const int)
25606 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
25608 * int16x8_t vdupq_lane_s16 (int16x4_t, const int)
25609 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
25611 * int8x16_t vdupq_lane_s8 (int8x8_t, const int)
25612 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
25614 * float32x4_t vdupq_lane_f32 (float32x2_t, const int)
25615 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
25617 * poly16x8_t vdupq_lane_p16 (poly16x4_t, const int)
25618 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
25620 * poly8x16_t vdupq_lane_p8 (poly8x8_t, const int)
25621 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
25623 * uint64x2_t vdupq_lane_u64 (uint64x1_t, const int)
25625 * int64x2_t vdupq_lane_s64 (int64x1_t, const int)
25627 5.50.3.43 Combining vectors
25628 ...........................
25630 * uint32x4_t vcombine_u32 (uint32x2_t, uint32x2_t)
25632 * uint16x8_t vcombine_u16 (uint16x4_t, uint16x4_t)
25634 * uint8x16_t vcombine_u8 (uint8x8_t, uint8x8_t)
25636 * int32x4_t vcombine_s32 (int32x2_t, int32x2_t)
25638 * int16x8_t vcombine_s16 (int16x4_t, int16x4_t)
25640 * int8x16_t vcombine_s8 (int8x8_t, int8x8_t)
25642 * uint64x2_t vcombine_u64 (uint64x1_t, uint64x1_t)
25644 * int64x2_t vcombine_s64 (int64x1_t, int64x1_t)
25646 * float32x4_t vcombine_f32 (float32x2_t, float32x2_t)
25648 * poly16x8_t vcombine_p16 (poly16x4_t, poly16x4_t)
25650 * poly8x16_t vcombine_p8 (poly8x8_t, poly8x8_t)
25652 5.50.3.44 Splitting vectors
25653 ...........................
25655 * uint32x2_t vget_high_u32 (uint32x4_t)
25657 * uint16x4_t vget_high_u16 (uint16x8_t)
25659 * uint8x8_t vget_high_u8 (uint8x16_t)
25661 * int32x2_t vget_high_s32 (int32x4_t)
25663 * int16x4_t vget_high_s16 (int16x8_t)
25665 * int8x8_t vget_high_s8 (int8x16_t)
25667 * uint64x1_t vget_high_u64 (uint64x2_t)
25669 * int64x1_t vget_high_s64 (int64x2_t)
25671 * float32x2_t vget_high_f32 (float32x4_t)
25673 * poly16x4_t vget_high_p16 (poly16x8_t)
25675 * poly8x8_t vget_high_p8 (poly8x16_t)
25677 * uint32x2_t vget_low_u32 (uint32x4_t)
25678 _Form of expected instruction(s):_ `vmov D0, D0'
25680 * uint16x4_t vget_low_u16 (uint16x8_t)
25681 _Form of expected instruction(s):_ `vmov D0, D0'
25683 * uint8x8_t vget_low_u8 (uint8x16_t)
25684 _Form of expected instruction(s):_ `vmov D0, D0'
25686 * int32x2_t vget_low_s32 (int32x4_t)
25687 _Form of expected instruction(s):_ `vmov D0, D0'
25689 * int16x4_t vget_low_s16 (int16x8_t)
25690 _Form of expected instruction(s):_ `vmov D0, D0'
25692 * int8x8_t vget_low_s8 (int8x16_t)
25693 _Form of expected instruction(s):_ `vmov D0, D0'
25695 * uint64x1_t vget_low_u64 (uint64x2_t)
25696 _Form of expected instruction(s):_ `vmov D0, D0'
25698 * int64x1_t vget_low_s64 (int64x2_t)
25699 _Form of expected instruction(s):_ `vmov D0, D0'
25701 * float32x2_t vget_low_f32 (float32x4_t)
25702 _Form of expected instruction(s):_ `vmov D0, D0'
25704 * poly16x4_t vget_low_p16 (poly16x8_t)
25705 _Form of expected instruction(s):_ `vmov D0, D0'
25707 * poly8x8_t vget_low_p8 (poly8x16_t)
25708 _Form of expected instruction(s):_ `vmov D0, D0'
25710 5.50.3.45 Conversions
25711 .....................
25713 * float32x2_t vcvt_f32_u32 (uint32x2_t)
25714 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0'
25716 * float32x2_t vcvt_f32_s32 (int32x2_t)
25717 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0'
25719 * uint32x2_t vcvt_u32_f32 (float32x2_t)
25720 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0'
25722 * int32x2_t vcvt_s32_f32 (float32x2_t)
25723 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0'
25725 * float32x4_t vcvtq_f32_u32 (uint32x4_t)
25726 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0'
25728 * float32x4_t vcvtq_f32_s32 (int32x4_t)
25729 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0'
25731 * uint32x4_t vcvtq_u32_f32 (float32x4_t)
25732 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0'
25734 * int32x4_t vcvtq_s32_f32 (float32x4_t)
25735 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0'
25737 * float32x2_t vcvt_n_f32_u32 (uint32x2_t, const int)
25738 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0, #0'
25740 * float32x2_t vcvt_n_f32_s32 (int32x2_t, const int)
25741 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0, #0'
25743 * uint32x2_t vcvt_n_u32_f32 (float32x2_t, const int)
25744 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0, #0'
25746 * int32x2_t vcvt_n_s32_f32 (float32x2_t, const int)
25747 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0, #0'
25749 * float32x4_t vcvtq_n_f32_u32 (uint32x4_t, const int)
25750 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0, #0'
25752 * float32x4_t vcvtq_n_f32_s32 (int32x4_t, const int)
25753 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0, #0'
25755 * uint32x4_t vcvtq_n_u32_f32 (float32x4_t, const int)
25756 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0, #0'
25758 * int32x4_t vcvtq_n_s32_f32 (float32x4_t, const int)
25759 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0, #0'
25761 5.50.3.46 Move, single_opcode narrowing
25762 .......................................
25764 * uint32x2_t vmovn_u64 (uint64x2_t)
25765 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
25767 * uint16x4_t vmovn_u32 (uint32x4_t)
25768 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
25770 * uint8x8_t vmovn_u16 (uint16x8_t)
25771 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
25773 * int32x2_t vmovn_s64 (int64x2_t)
25774 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
25776 * int16x4_t vmovn_s32 (int32x4_t)
25777 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
25779 * int8x8_t vmovn_s16 (int16x8_t)
25780 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
25782 * uint32x2_t vqmovn_u64 (uint64x2_t)
25783 _Form of expected instruction(s):_ `vqmovn.u64 D0, Q0'
25785 * uint16x4_t vqmovn_u32 (uint32x4_t)
25786 _Form of expected instruction(s):_ `vqmovn.u32 D0, Q0'
25788 * uint8x8_t vqmovn_u16 (uint16x8_t)
25789 _Form of expected instruction(s):_ `vqmovn.u16 D0, Q0'
25791 * int32x2_t vqmovn_s64 (int64x2_t)
25792 _Form of expected instruction(s):_ `vqmovn.s64 D0, Q0'
25794 * int16x4_t vqmovn_s32 (int32x4_t)
25795 _Form of expected instruction(s):_ `vqmovn.s32 D0, Q0'
25797 * int8x8_t vqmovn_s16 (int16x8_t)
25798 _Form of expected instruction(s):_ `vqmovn.s16 D0, Q0'
25800 * uint32x2_t vqmovun_s64 (int64x2_t)
25801 _Form of expected instruction(s):_ `vqmovun.s64 D0, Q0'
25803 * uint16x4_t vqmovun_s32 (int32x4_t)
25804 _Form of expected instruction(s):_ `vqmovun.s32 D0, Q0'
25806 * uint8x8_t vqmovun_s16 (int16x8_t)
25807 _Form of expected instruction(s):_ `vqmovun.s16 D0, Q0'
25809 5.50.3.47 Move, single_opcode long
25810 ..................................
25812 * uint64x2_t vmovl_u32 (uint32x2_t)
25813 _Form of expected instruction(s):_ `vmovl.u32 Q0, D0'
25815 * uint32x4_t vmovl_u16 (uint16x4_t)
25816 _Form of expected instruction(s):_ `vmovl.u16 Q0, D0'
25818 * uint16x8_t vmovl_u8 (uint8x8_t)
25819 _Form of expected instruction(s):_ `vmovl.u8 Q0, D0'
25821 * int64x2_t vmovl_s32 (int32x2_t)
25822 _Form of expected instruction(s):_ `vmovl.s32 Q0, D0'
25824 * int32x4_t vmovl_s16 (int16x4_t)
25825 _Form of expected instruction(s):_ `vmovl.s16 Q0, D0'
25827 * int16x8_t vmovl_s8 (int8x8_t)
25828 _Form of expected instruction(s):_ `vmovl.s8 Q0, D0'
25830 5.50.3.48 Table lookup
25831 ......................
25833 * poly8x8_t vtbl1_p8 (poly8x8_t, uint8x8_t)
25834 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
25836 * int8x8_t vtbl1_s8 (int8x8_t, int8x8_t)
25837 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
25839 * uint8x8_t vtbl1_u8 (uint8x8_t, uint8x8_t)
25840 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
25842 * poly8x8_t vtbl2_p8 (poly8x8x2_t, uint8x8_t)
25843 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
25845 * int8x8_t vtbl2_s8 (int8x8x2_t, int8x8_t)
25846 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
25848 * uint8x8_t vtbl2_u8 (uint8x8x2_t, uint8x8_t)
25849 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
25851 * poly8x8_t vtbl3_p8 (poly8x8x3_t, uint8x8_t)
25852 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
25854 * int8x8_t vtbl3_s8 (int8x8x3_t, int8x8_t)
25855 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
25857 * uint8x8_t vtbl3_u8 (uint8x8x3_t, uint8x8_t)
25858 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
25860 * poly8x8_t vtbl4_p8 (poly8x8x4_t, uint8x8_t)
25861 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
25864 * int8x8_t vtbl4_s8 (int8x8x4_t, int8x8_t)
25865 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
25868 * uint8x8_t vtbl4_u8 (uint8x8x4_t, uint8x8_t)
25869 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
25872 5.50.3.49 Extended table lookup
25873 ...............................
25875 * poly8x8_t vtbx1_p8 (poly8x8_t, poly8x8_t, uint8x8_t)
25876 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
25878 * int8x8_t vtbx1_s8 (int8x8_t, int8x8_t, int8x8_t)
25879 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
25881 * uint8x8_t vtbx1_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
25882 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
25884 * poly8x8_t vtbx2_p8 (poly8x8_t, poly8x8x2_t, uint8x8_t)
25885 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
25887 * int8x8_t vtbx2_s8 (int8x8_t, int8x8x2_t, int8x8_t)
25888 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
25890 * uint8x8_t vtbx2_u8 (uint8x8_t, uint8x8x2_t, uint8x8_t)
25891 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
25893 * poly8x8_t vtbx3_p8 (poly8x8_t, poly8x8x3_t, uint8x8_t)
25894 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
25896 * int8x8_t vtbx3_s8 (int8x8_t, int8x8x3_t, int8x8_t)
25897 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
25899 * uint8x8_t vtbx3_u8 (uint8x8_t, uint8x8x3_t, uint8x8_t)
25900 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
25902 * poly8x8_t vtbx4_p8 (poly8x8_t, poly8x8x4_t, uint8x8_t)
25903 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
25906 * int8x8_t vtbx4_s8 (int8x8_t, int8x8x4_t, int8x8_t)
25907 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
25910 * uint8x8_t vtbx4_u8 (uint8x8_t, uint8x8x4_t, uint8x8_t)
25911 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
25914 5.50.3.50 Multiply, lane
25915 ........................
25917 * float32x2_t vmul_lane_f32 (float32x2_t, float32x2_t, const int)
25918 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
25920 * uint32x2_t vmul_lane_u32 (uint32x2_t, uint32x2_t, const int)
25921 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
25923 * uint16x4_t vmul_lane_u16 (uint16x4_t, uint16x4_t, const int)
25924 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
25926 * int32x2_t vmul_lane_s32 (int32x2_t, int32x2_t, const int)
25927 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
25929 * int16x4_t vmul_lane_s16 (int16x4_t, int16x4_t, const int)
25930 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
25932 * float32x4_t vmulq_lane_f32 (float32x4_t, float32x2_t, const int)
25933 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
25935 * uint32x4_t vmulq_lane_u32 (uint32x4_t, uint32x2_t, const int)
25936 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
25938 * uint16x8_t vmulq_lane_u16 (uint16x8_t, uint16x4_t, const int)
25939 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
25941 * int32x4_t vmulq_lane_s32 (int32x4_t, int32x2_t, const int)
25942 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
25944 * int16x8_t vmulq_lane_s16 (int16x8_t, int16x4_t, const int)
25945 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
25947 5.50.3.51 Long multiply, lane
25948 .............................
25950 * uint64x2_t vmull_lane_u32 (uint32x2_t, uint32x2_t, const int)
25951 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
25953 * uint32x4_t vmull_lane_u16 (uint16x4_t, uint16x4_t, const int)
25954 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
25956 * int64x2_t vmull_lane_s32 (int32x2_t, int32x2_t, const int)
25957 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
25959 * int32x4_t vmull_lane_s16 (int16x4_t, int16x4_t, const int)
25960 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
25962 5.50.3.52 Saturating doubling long multiply, lane
25963 .................................................
25965 * int64x2_t vqdmull_lane_s32 (int32x2_t, int32x2_t, const int)
25966 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
25968 * int32x4_t vqdmull_lane_s16 (int16x4_t, int16x4_t, const int)
25969 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
25971 5.50.3.53 Saturating doubling multiply high, lane
25972 .................................................
25974 * int32x4_t vqdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
25975 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
25977 * int16x8_t vqdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
25978 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
25980 * int32x2_t vqdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
25981 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
25983 * int16x4_t vqdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
25984 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
25986 * int32x4_t vqrdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
25987 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
25989 * int16x8_t vqrdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
25990 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
25992 * int32x2_t vqrdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
25993 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
25995 * int16x4_t vqrdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
25996 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
25998 5.50.3.54 Multiply-accumulate, lane
25999 ...................................
26001 * float32x2_t vmla_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
26003 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
26005 * uint32x2_t vmla_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
26007 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
26009 * uint16x4_t vmla_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
26011 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
26013 * int32x2_t vmla_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
26015 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
26017 * int16x4_t vmla_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
26019 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
26021 * float32x4_t vmlaq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
26023 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
26025 * uint32x4_t vmlaq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
26027 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
26029 * uint16x8_t vmlaq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
26031 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
26033 * int32x4_t vmlaq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
26035 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
26037 * int16x8_t vmlaq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
26039 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
26041 * uint64x2_t vmlal_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
26043 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
26045 * uint32x4_t vmlal_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
26047 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
26049 * int64x2_t vmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
26051 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
26053 * int32x4_t vmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
26055 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
26057 * int64x2_t vqdmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
26059 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
26061 * int32x4_t vqdmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
26063 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
26065 5.50.3.55 Multiply-subtract, lane
26066 .................................
26068 * float32x2_t vmls_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
26070 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
26072 * uint32x2_t vmls_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
26074 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
26076 * uint16x4_t vmls_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
26078 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
26080 * int32x2_t vmls_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
26082 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
26084 * int16x4_t vmls_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
26086 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
26088 * float32x4_t vmlsq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
26090 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
26092 * uint32x4_t vmlsq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
26094 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
26096 * uint16x8_t vmlsq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
26098 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
26100 * int32x4_t vmlsq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
26102 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
26104 * int16x8_t vmlsq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
26106 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
26108 * uint64x2_t vmlsl_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
26110 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
26112 * uint32x4_t vmlsl_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
26114 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
26116 * int64x2_t vmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
26118 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
26120 * int32x4_t vmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
26122 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
26124 * int64x2_t vqdmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
26126 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
26128 * int32x4_t vqdmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
26130 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
26132 5.50.3.56 Vector multiply by scalar
26133 ...................................
26135 * float32x2_t vmul_n_f32 (float32x2_t, float32_t)
26136 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
26138 * uint32x2_t vmul_n_u32 (uint32x2_t, uint32_t)
26139 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
26141 * uint16x4_t vmul_n_u16 (uint16x4_t, uint16_t)
26142 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
26144 * int32x2_t vmul_n_s32 (int32x2_t, int32_t)
26145 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
26147 * int16x4_t vmul_n_s16 (int16x4_t, int16_t)
26148 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
26150 * float32x4_t vmulq_n_f32 (float32x4_t, float32_t)
26151 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
26153 * uint32x4_t vmulq_n_u32 (uint32x4_t, uint32_t)
26154 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
26156 * uint16x8_t vmulq_n_u16 (uint16x8_t, uint16_t)
26157 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
26159 * int32x4_t vmulq_n_s32 (int32x4_t, int32_t)
26160 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
26162 * int16x8_t vmulq_n_s16 (int16x8_t, int16_t)
26163 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
26165 5.50.3.57 Vector long multiply by scalar
26166 ........................................
26168 * uint64x2_t vmull_n_u32 (uint32x2_t, uint32_t)
26169 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
26171 * uint32x4_t vmull_n_u16 (uint16x4_t, uint16_t)
26172 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
26174 * int64x2_t vmull_n_s32 (int32x2_t, int32_t)
26175 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
26177 * int32x4_t vmull_n_s16 (int16x4_t, int16_t)
26178 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
26180 5.50.3.58 Vector saturating doubling long multiply by scalar
26181 ............................................................
26183 * int64x2_t vqdmull_n_s32 (int32x2_t, int32_t)
26184 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
26186 * int32x4_t vqdmull_n_s16 (int16x4_t, int16_t)
26187 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
26189 5.50.3.59 Vector saturating doubling multiply high by scalar
26190 ............................................................
26192 * int32x4_t vqdmulhq_n_s32 (int32x4_t, int32_t)
26193 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
26195 * int16x8_t vqdmulhq_n_s16 (int16x8_t, int16_t)
26196 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
26198 * int32x2_t vqdmulh_n_s32 (int32x2_t, int32_t)
26199 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
26201 * int16x4_t vqdmulh_n_s16 (int16x4_t, int16_t)
26202 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
26204 * int32x4_t vqrdmulhq_n_s32 (int32x4_t, int32_t)
26205 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
26207 * int16x8_t vqrdmulhq_n_s16 (int16x8_t, int16_t)
26208 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
26210 * int32x2_t vqrdmulh_n_s32 (int32x2_t, int32_t)
26211 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
26213 * int16x4_t vqrdmulh_n_s16 (int16x4_t, int16_t)
26214 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
26216 5.50.3.60 Vector multiply-accumulate by scalar
26217 ..............................................
26219 * float32x2_t vmla_n_f32 (float32x2_t, float32x2_t, float32_t)
26220 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
26222 * uint32x2_t vmla_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
26223 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
26225 * uint16x4_t vmla_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
26226 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
26228 * int32x2_t vmla_n_s32 (int32x2_t, int32x2_t, int32_t)
26229 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
26231 * int16x4_t vmla_n_s16 (int16x4_t, int16x4_t, int16_t)
26232 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
26234 * float32x4_t vmlaq_n_f32 (float32x4_t, float32x4_t, float32_t)
26235 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
26237 * uint32x4_t vmlaq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
26238 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
26240 * uint16x8_t vmlaq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
26241 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
26243 * int32x4_t vmlaq_n_s32 (int32x4_t, int32x4_t, int32_t)
26244 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
26246 * int16x8_t vmlaq_n_s16 (int16x8_t, int16x8_t, int16_t)
26247 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
26249 * uint64x2_t vmlal_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
26250 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
26252 * uint32x4_t vmlal_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
26253 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
26255 * int64x2_t vmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
26256 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
26258 * int32x4_t vmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
26259 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
26261 * int64x2_t vqdmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
26262 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
26264 * int32x4_t vqdmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
26265 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
26267 5.50.3.61 Vector multiply-subtract by scalar
26268 ............................................
26270 * float32x2_t vmls_n_f32 (float32x2_t, float32x2_t, float32_t)
26271 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
26273 * uint32x2_t vmls_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
26274 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
26276 * uint16x4_t vmls_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
26277 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
26279 * int32x2_t vmls_n_s32 (int32x2_t, int32x2_t, int32_t)
26280 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
26282 * int16x4_t vmls_n_s16 (int16x4_t, int16x4_t, int16_t)
26283 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
26285 * float32x4_t vmlsq_n_f32 (float32x4_t, float32x4_t, float32_t)
26286 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
26288 * uint32x4_t vmlsq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
26289 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
26291 * uint16x8_t vmlsq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
26292 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
26294 * int32x4_t vmlsq_n_s32 (int32x4_t, int32x4_t, int32_t)
26295 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
26297 * int16x8_t vmlsq_n_s16 (int16x8_t, int16x8_t, int16_t)
26298 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
26300 * uint64x2_t vmlsl_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
26301 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
26303 * uint32x4_t vmlsl_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
26304 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
26306 * int64x2_t vmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
26307 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
26309 * int32x4_t vmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
26310 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
26312 * int64x2_t vqdmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
26313 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
26315 * int32x4_t vqdmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
26316 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
26318 5.50.3.62 Vector extract
26319 ........................
26321 * uint32x2_t vext_u32 (uint32x2_t, uint32x2_t, const int)
26322 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
26324 * uint16x4_t vext_u16 (uint16x4_t, uint16x4_t, const int)
26325 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
26327 * uint8x8_t vext_u8 (uint8x8_t, uint8x8_t, const int)
26328 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
26330 * int32x2_t vext_s32 (int32x2_t, int32x2_t, const int)
26331 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
26333 * int16x4_t vext_s16 (int16x4_t, int16x4_t, const int)
26334 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
26336 * int8x8_t vext_s8 (int8x8_t, int8x8_t, const int)
26337 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
26339 * uint64x1_t vext_u64 (uint64x1_t, uint64x1_t, const int)
26340 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
26342 * int64x1_t vext_s64 (int64x1_t, int64x1_t, const int)
26343 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
26345 * float32x2_t vext_f32 (float32x2_t, float32x2_t, const int)
26346 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
26348 * poly16x4_t vext_p16 (poly16x4_t, poly16x4_t, const int)
26349 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
26351 * poly8x8_t vext_p8 (poly8x8_t, poly8x8_t, const int)
26352 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
26354 * uint32x4_t vextq_u32 (uint32x4_t, uint32x4_t, const int)
26355 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
26357 * uint16x8_t vextq_u16 (uint16x8_t, uint16x8_t, const int)
26358 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
26360 * uint8x16_t vextq_u8 (uint8x16_t, uint8x16_t, const int)
26361 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
26363 * int32x4_t vextq_s32 (int32x4_t, int32x4_t, const int)
26364 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
26366 * int16x8_t vextq_s16 (int16x8_t, int16x8_t, const int)
26367 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
26369 * int8x16_t vextq_s8 (int8x16_t, int8x16_t, const int)
26370 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
26372 * uint64x2_t vextq_u64 (uint64x2_t, uint64x2_t, const int)
26373 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
26375 * int64x2_t vextq_s64 (int64x2_t, int64x2_t, const int)
26376 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
26378 * float32x4_t vextq_f32 (float32x4_t, float32x4_t, const int)
26379 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
26381 * poly16x8_t vextq_p16 (poly16x8_t, poly16x8_t, const int)
26382 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
26384 * poly8x16_t vextq_p8 (poly8x16_t, poly8x16_t, const int)
26385 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
26387 5.50.3.63 Reverse elements
26388 ..........................
26390 * uint32x2_t vrev64_u32 (uint32x2_t)
26391 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
26393 * uint16x4_t vrev64_u16 (uint16x4_t)
26394 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
26396 * uint8x8_t vrev64_u8 (uint8x8_t)
26397 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
26399 * int32x2_t vrev64_s32 (int32x2_t)
26400 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
26402 * int16x4_t vrev64_s16 (int16x4_t)
26403 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
26405 * int8x8_t vrev64_s8 (int8x8_t)
26406 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
26408 * float32x2_t vrev64_f32 (float32x2_t)
26409 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
26411 * poly16x4_t vrev64_p16 (poly16x4_t)
26412 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
26414 * poly8x8_t vrev64_p8 (poly8x8_t)
26415 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
26417 * uint32x4_t vrev64q_u32 (uint32x4_t)
26418 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
26420 * uint16x8_t vrev64q_u16 (uint16x8_t)
26421 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
26423 * uint8x16_t vrev64q_u8 (uint8x16_t)
26424 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
26426 * int32x4_t vrev64q_s32 (int32x4_t)
26427 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
26429 * int16x8_t vrev64q_s16 (int16x8_t)
26430 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
26432 * int8x16_t vrev64q_s8 (int8x16_t)
26433 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
26435 * float32x4_t vrev64q_f32 (float32x4_t)
26436 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
26438 * poly16x8_t vrev64q_p16 (poly16x8_t)
26439 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
26441 * poly8x16_t vrev64q_p8 (poly8x16_t)
26442 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
26444 * uint16x4_t vrev32_u16 (uint16x4_t)
26445 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
26447 * int16x4_t vrev32_s16 (int16x4_t)
26448 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
26450 * uint8x8_t vrev32_u8 (uint8x8_t)
26451 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
26453 * int8x8_t vrev32_s8 (int8x8_t)
26454 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
26456 * poly16x4_t vrev32_p16 (poly16x4_t)
26457 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
26459 * poly8x8_t vrev32_p8 (poly8x8_t)
26460 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
26462 * uint16x8_t vrev32q_u16 (uint16x8_t)
26463 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
26465 * int16x8_t vrev32q_s16 (int16x8_t)
26466 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
26468 * uint8x16_t vrev32q_u8 (uint8x16_t)
26469 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
26471 * int8x16_t vrev32q_s8 (int8x16_t)
26472 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
26474 * poly16x8_t vrev32q_p16 (poly16x8_t)
26475 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
26477 * poly8x16_t vrev32q_p8 (poly8x16_t)
26478 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
26480 * uint8x8_t vrev16_u8 (uint8x8_t)
26481 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
26483 * int8x8_t vrev16_s8 (int8x8_t)
26484 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
26486 * poly8x8_t vrev16_p8 (poly8x8_t)
26487 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
26489 * uint8x16_t vrev16q_u8 (uint8x16_t)
26490 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
26492 * int8x16_t vrev16q_s8 (int8x16_t)
26493 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
26495 * poly8x16_t vrev16q_p8 (poly8x16_t)
26496 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
26498 5.50.3.64 Bit selection
26499 .......................
26501 * uint32x2_t vbsl_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
26502 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26503 D0, D0, D0' _or_ `vbif D0, D0, D0'
26505 * uint16x4_t vbsl_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
26506 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26507 D0, D0, D0' _or_ `vbif D0, D0, D0'
26509 * uint8x8_t vbsl_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
26510 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26511 D0, D0, D0' _or_ `vbif D0, D0, D0'
26513 * int32x2_t vbsl_s32 (uint32x2_t, int32x2_t, int32x2_t)
26514 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26515 D0, D0, D0' _or_ `vbif D0, D0, D0'
26517 * int16x4_t vbsl_s16 (uint16x4_t, int16x4_t, int16x4_t)
26518 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26519 D0, D0, D0' _or_ `vbif D0, D0, D0'
26521 * int8x8_t vbsl_s8 (uint8x8_t, int8x8_t, int8x8_t)
26522 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26523 D0, D0, D0' _or_ `vbif D0, D0, D0'
26525 * uint64x1_t vbsl_u64 (uint64x1_t, uint64x1_t, uint64x1_t)
26526 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26527 D0, D0, D0' _or_ `vbif D0, D0, D0'
26529 * int64x1_t vbsl_s64 (uint64x1_t, int64x1_t, int64x1_t)
26530 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26531 D0, D0, D0' _or_ `vbif D0, D0, D0'
26533 * float32x2_t vbsl_f32 (uint32x2_t, float32x2_t, float32x2_t)
26534 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26535 D0, D0, D0' _or_ `vbif D0, D0, D0'
26537 * poly16x4_t vbsl_p16 (uint16x4_t, poly16x4_t, poly16x4_t)
26538 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26539 D0, D0, D0' _or_ `vbif D0, D0, D0'
26541 * poly8x8_t vbsl_p8 (uint8x8_t, poly8x8_t, poly8x8_t)
26542 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
26543 D0, D0, D0' _or_ `vbif D0, D0, D0'
26545 * uint32x4_t vbslq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
26546 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26547 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26549 * uint16x8_t vbslq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
26550 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26551 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26553 * uint8x16_t vbslq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
26554 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26555 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26557 * int32x4_t vbslq_s32 (uint32x4_t, int32x4_t, int32x4_t)
26558 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26559 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26561 * int16x8_t vbslq_s16 (uint16x8_t, int16x8_t, int16x8_t)
26562 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26563 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26565 * int8x16_t vbslq_s8 (uint8x16_t, int8x16_t, int8x16_t)
26566 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26567 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26569 * uint64x2_t vbslq_u64 (uint64x2_t, uint64x2_t, uint64x2_t)
26570 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26571 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26573 * int64x2_t vbslq_s64 (uint64x2_t, int64x2_t, int64x2_t)
26574 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26575 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26577 * float32x4_t vbslq_f32 (uint32x4_t, float32x4_t, float32x4_t)
26578 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26579 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26581 * poly16x8_t vbslq_p16 (uint16x8_t, poly16x8_t, poly16x8_t)
26582 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26583 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26585 * poly8x16_t vbslq_p8 (uint8x16_t, poly8x16_t, poly8x16_t)
26586 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
26587 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
26589 5.50.3.65 Transpose elements
26590 ............................
26592 * uint32x2x2_t vtrn_u32 (uint32x2_t, uint32x2_t)
26593 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
26595 * uint16x4x2_t vtrn_u16 (uint16x4_t, uint16x4_t)
26596 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
26598 * uint8x8x2_t vtrn_u8 (uint8x8_t, uint8x8_t)
26599 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
26601 * int32x2x2_t vtrn_s32 (int32x2_t, int32x2_t)
26602 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
26604 * int16x4x2_t vtrn_s16 (int16x4_t, int16x4_t)
26605 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
26607 * int8x8x2_t vtrn_s8 (int8x8_t, int8x8_t)
26608 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
26610 * float32x2x2_t vtrn_f32 (float32x2_t, float32x2_t)
26611 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
26613 * poly16x4x2_t vtrn_p16 (poly16x4_t, poly16x4_t)
26614 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
26616 * poly8x8x2_t vtrn_p8 (poly8x8_t, poly8x8_t)
26617 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
26619 * uint32x4x2_t vtrnq_u32 (uint32x4_t, uint32x4_t)
26620 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
26622 * uint16x8x2_t vtrnq_u16 (uint16x8_t, uint16x8_t)
26623 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
26625 * uint8x16x2_t vtrnq_u8 (uint8x16_t, uint8x16_t)
26626 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
26628 * int32x4x2_t vtrnq_s32 (int32x4_t, int32x4_t)
26629 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
26631 * int16x8x2_t vtrnq_s16 (int16x8_t, int16x8_t)
26632 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
26634 * int8x16x2_t vtrnq_s8 (int8x16_t, int8x16_t)
26635 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
26637 * float32x4x2_t vtrnq_f32 (float32x4_t, float32x4_t)
26638 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
26640 * poly16x8x2_t vtrnq_p16 (poly16x8_t, poly16x8_t)
26641 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
26643 * poly8x16x2_t vtrnq_p8 (poly8x16_t, poly8x16_t)
26644 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
26646 5.50.3.66 Zip elements
26647 ......................
26649 * uint32x2x2_t vzip_u32 (uint32x2_t, uint32x2_t)
26650 _Form of expected instruction(s):_ `vzip.32 D0, D1'
26652 * uint16x4x2_t vzip_u16 (uint16x4_t, uint16x4_t)
26653 _Form of expected instruction(s):_ `vzip.16 D0, D1'
26655 * uint8x8x2_t vzip_u8 (uint8x8_t, uint8x8_t)
26656 _Form of expected instruction(s):_ `vzip.8 D0, D1'
26658 * int32x2x2_t vzip_s32 (int32x2_t, int32x2_t)
26659 _Form of expected instruction(s):_ `vzip.32 D0, D1'
26661 * int16x4x2_t vzip_s16 (int16x4_t, int16x4_t)
26662 _Form of expected instruction(s):_ `vzip.16 D0, D1'
26664 * int8x8x2_t vzip_s8 (int8x8_t, int8x8_t)
26665 _Form of expected instruction(s):_ `vzip.8 D0, D1'
26667 * float32x2x2_t vzip_f32 (float32x2_t, float32x2_t)
26668 _Form of expected instruction(s):_ `vzip.32 D0, D1'
26670 * poly16x4x2_t vzip_p16 (poly16x4_t, poly16x4_t)
26671 _Form of expected instruction(s):_ `vzip.16 D0, D1'
26673 * poly8x8x2_t vzip_p8 (poly8x8_t, poly8x8_t)
26674 _Form of expected instruction(s):_ `vzip.8 D0, D1'
26676 * uint32x4x2_t vzipq_u32 (uint32x4_t, uint32x4_t)
26677 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
26679 * uint16x8x2_t vzipq_u16 (uint16x8_t, uint16x8_t)
26680 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
26682 * uint8x16x2_t vzipq_u8 (uint8x16_t, uint8x16_t)
26683 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
26685 * int32x4x2_t vzipq_s32 (int32x4_t, int32x4_t)
26686 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
26688 * int16x8x2_t vzipq_s16 (int16x8_t, int16x8_t)
26689 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
26691 * int8x16x2_t vzipq_s8 (int8x16_t, int8x16_t)
26692 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
26694 * float32x4x2_t vzipq_f32 (float32x4_t, float32x4_t)
26695 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
26697 * poly16x8x2_t vzipq_p16 (poly16x8_t, poly16x8_t)
26698 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
26700 * poly8x16x2_t vzipq_p8 (poly8x16_t, poly8x16_t)
26701 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
26703 5.50.3.67 Unzip elements
26704 ........................
26706 * uint32x2x2_t vuzp_u32 (uint32x2_t, uint32x2_t)
26707 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
26709 * uint16x4x2_t vuzp_u16 (uint16x4_t, uint16x4_t)
26710 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
26712 * uint8x8x2_t vuzp_u8 (uint8x8_t, uint8x8_t)
26713 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
26715 * int32x2x2_t vuzp_s32 (int32x2_t, int32x2_t)
26716 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
26718 * int16x4x2_t vuzp_s16 (int16x4_t, int16x4_t)
26719 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
26721 * int8x8x2_t vuzp_s8 (int8x8_t, int8x8_t)
26722 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
26724 * float32x2x2_t vuzp_f32 (float32x2_t, float32x2_t)
26725 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
26727 * poly16x4x2_t vuzp_p16 (poly16x4_t, poly16x4_t)
26728 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
26730 * poly8x8x2_t vuzp_p8 (poly8x8_t, poly8x8_t)
26731 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
26733 * uint32x4x2_t vuzpq_u32 (uint32x4_t, uint32x4_t)
26734 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
26736 * uint16x8x2_t vuzpq_u16 (uint16x8_t, uint16x8_t)
26737 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
26739 * uint8x16x2_t vuzpq_u8 (uint8x16_t, uint8x16_t)
26740 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
26742 * int32x4x2_t vuzpq_s32 (int32x4_t, int32x4_t)
26743 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
26745 * int16x8x2_t vuzpq_s16 (int16x8_t, int16x8_t)
26746 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
26748 * int8x16x2_t vuzpq_s8 (int8x16_t, int8x16_t)
26749 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
26751 * float32x4x2_t vuzpq_f32 (float32x4_t, float32x4_t)
26752 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
26754 * poly16x8x2_t vuzpq_p16 (poly16x8_t, poly16x8_t)
26755 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
26757 * poly8x16x2_t vuzpq_p8 (poly8x16_t, poly8x16_t)
26758 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
26760 5.50.3.68 Element/structure loads, VLD1 variants
26761 ................................................
26763 * uint32x2_t vld1_u32 (const uint32_t *)
26764 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
26766 * uint16x4_t vld1_u16 (const uint16_t *)
26767 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
26769 * uint8x8_t vld1_u8 (const uint8_t *)
26770 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
26772 * int32x2_t vld1_s32 (const int32_t *)
26773 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
26775 * int16x4_t vld1_s16 (const int16_t *)
26776 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
26778 * int8x8_t vld1_s8 (const int8_t *)
26779 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
26781 * uint64x1_t vld1_u64 (const uint64_t *)
26782 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26784 * int64x1_t vld1_s64 (const int64_t *)
26785 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26787 * float32x2_t vld1_f32 (const float32_t *)
26788 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
26790 * poly16x4_t vld1_p16 (const poly16_t *)
26791 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
26793 * poly8x8_t vld1_p8 (const poly8_t *)
26794 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
26796 * uint32x4_t vld1q_u32 (const uint32_t *)
26797 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
26799 * uint16x8_t vld1q_u16 (const uint16_t *)
26800 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
26802 * uint8x16_t vld1q_u8 (const uint8_t *)
26803 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
26805 * int32x4_t vld1q_s32 (const int32_t *)
26806 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
26808 * int16x8_t vld1q_s16 (const int16_t *)
26809 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
26811 * int8x16_t vld1q_s8 (const int8_t *)
26812 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
26814 * uint64x2_t vld1q_u64 (const uint64_t *)
26815 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
26817 * int64x2_t vld1q_s64 (const int64_t *)
26818 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
26820 * float32x4_t vld1q_f32 (const float32_t *)
26821 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
26823 * poly16x8_t vld1q_p16 (const poly16_t *)
26824 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
26826 * poly8x16_t vld1q_p8 (const poly8_t *)
26827 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
26829 * uint32x2_t vld1_lane_u32 (const uint32_t *, uint32x2_t, const int)
26830 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26832 * uint16x4_t vld1_lane_u16 (const uint16_t *, uint16x4_t, const int)
26833 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26835 * uint8x8_t vld1_lane_u8 (const uint8_t *, uint8x8_t, const int)
26836 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26838 * int32x2_t vld1_lane_s32 (const int32_t *, int32x2_t, const int)
26839 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26841 * int16x4_t vld1_lane_s16 (const int16_t *, int16x4_t, const int)
26842 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26844 * int8x8_t vld1_lane_s8 (const int8_t *, int8x8_t, const int)
26845 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26847 * float32x2_t vld1_lane_f32 (const float32_t *, float32x2_t, const
26849 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26851 * poly16x4_t vld1_lane_p16 (const poly16_t *, poly16x4_t, const int)
26852 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26854 * poly8x8_t vld1_lane_p8 (const poly8_t *, poly8x8_t, const int)
26855 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26857 * uint64x1_t vld1_lane_u64 (const uint64_t *, uint64x1_t, const int)
26858 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26860 * int64x1_t vld1_lane_s64 (const int64_t *, int64x1_t, const int)
26861 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26863 * uint32x4_t vld1q_lane_u32 (const uint32_t *, uint32x4_t, const int)
26864 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26866 * uint16x8_t vld1q_lane_u16 (const uint16_t *, uint16x8_t, const int)
26867 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26869 * uint8x16_t vld1q_lane_u8 (const uint8_t *, uint8x16_t, const int)
26870 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26872 * int32x4_t vld1q_lane_s32 (const int32_t *, int32x4_t, const int)
26873 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26875 * int16x8_t vld1q_lane_s16 (const int16_t *, int16x8_t, const int)
26876 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26878 * int8x16_t vld1q_lane_s8 (const int8_t *, int8x16_t, const int)
26879 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26881 * float32x4_t vld1q_lane_f32 (const float32_t *, float32x4_t, const
26883 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
26885 * poly16x8_t vld1q_lane_p16 (const poly16_t *, poly16x8_t, const int)
26886 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
26888 * poly8x16_t vld1q_lane_p8 (const poly8_t *, poly8x16_t, const int)
26889 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
26891 * uint64x2_t vld1q_lane_u64 (const uint64_t *, uint64x2_t, const int)
26892 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26894 * int64x2_t vld1q_lane_s64 (const int64_t *, int64x2_t, const int)
26895 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26897 * uint32x2_t vld1_dup_u32 (const uint32_t *)
26898 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
26900 * uint16x4_t vld1_dup_u16 (const uint16_t *)
26901 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
26903 * uint8x8_t vld1_dup_u8 (const uint8_t *)
26904 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
26906 * int32x2_t vld1_dup_s32 (const int32_t *)
26907 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
26909 * int16x4_t vld1_dup_s16 (const int16_t *)
26910 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
26912 * int8x8_t vld1_dup_s8 (const int8_t *)
26913 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
26915 * float32x2_t vld1_dup_f32 (const float32_t *)
26916 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
26918 * poly16x4_t vld1_dup_p16 (const poly16_t *)
26919 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
26921 * poly8x8_t vld1_dup_p8 (const poly8_t *)
26922 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
26924 * uint64x1_t vld1_dup_u64 (const uint64_t *)
26925 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26927 * int64x1_t vld1_dup_s64 (const int64_t *)
26928 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
26930 * uint32x4_t vld1q_dup_u32 (const uint32_t *)
26931 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
26933 * uint16x8_t vld1q_dup_u16 (const uint16_t *)
26934 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
26936 * uint8x16_t vld1q_dup_u8 (const uint8_t *)
26937 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
26939 * int32x4_t vld1q_dup_s32 (const int32_t *)
26940 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
26942 * int16x8_t vld1q_dup_s16 (const int16_t *)
26943 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
26945 * int8x16_t vld1q_dup_s8 (const int8_t *)
26946 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
26948 * float32x4_t vld1q_dup_f32 (const float32_t *)
26949 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
26951 * poly16x8_t vld1q_dup_p16 (const poly16_t *)
26952 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
26954 * poly8x16_t vld1q_dup_p8 (const poly8_t *)
26955 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
26957 * uint64x2_t vld1q_dup_u64 (const uint64_t *)
26958 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
26960 * int64x2_t vld1q_dup_s64 (const int64_t *)
26961 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
26963 5.50.3.69 Element/structure stores, VST1 variants
26964 .................................................
26966 * void vst1_u32 (uint32_t *, uint32x2_t)
26967 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
26969 * void vst1_u16 (uint16_t *, uint16x4_t)
26970 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
26972 * void vst1_u8 (uint8_t *, uint8x8_t)
26973 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
26975 * void vst1_s32 (int32_t *, int32x2_t)
26976 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
26978 * void vst1_s16 (int16_t *, int16x4_t)
26979 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
26981 * void vst1_s8 (int8_t *, int8x8_t)
26982 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
26984 * void vst1_u64 (uint64_t *, uint64x1_t)
26985 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
26987 * void vst1_s64 (int64_t *, int64x1_t)
26988 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
26990 * void vst1_f32 (float32_t *, float32x2_t)
26991 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
26993 * void vst1_p16 (poly16_t *, poly16x4_t)
26994 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
26996 * void vst1_p8 (poly8_t *, poly8x8_t)
26997 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
26999 * void vst1q_u32 (uint32_t *, uint32x4_t)
27000 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
27002 * void vst1q_u16 (uint16_t *, uint16x8_t)
27003 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
27005 * void vst1q_u8 (uint8_t *, uint8x16_t)
27006 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
27008 * void vst1q_s32 (int32_t *, int32x4_t)
27009 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
27011 * void vst1q_s16 (int16_t *, int16x8_t)
27012 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
27014 * void vst1q_s8 (int8_t *, int8x16_t)
27015 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
27017 * void vst1q_u64 (uint64_t *, uint64x2_t)
27018 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
27020 * void vst1q_s64 (int64_t *, int64x2_t)
27021 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
27023 * void vst1q_f32 (float32_t *, float32x4_t)
27024 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
27026 * void vst1q_p16 (poly16_t *, poly16x8_t)
27027 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
27029 * void vst1q_p8 (poly8_t *, poly8x16_t)
27030 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
27032 * void vst1_lane_u32 (uint32_t *, uint32x2_t, const int)
27033 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27035 * void vst1_lane_u16 (uint16_t *, uint16x4_t, const int)
27036 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27038 * void vst1_lane_u8 (uint8_t *, uint8x8_t, const int)
27039 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27041 * void vst1_lane_s32 (int32_t *, int32x2_t, const int)
27042 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27044 * void vst1_lane_s16 (int16_t *, int16x4_t, const int)
27045 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27047 * void vst1_lane_s8 (int8_t *, int8x8_t, const int)
27048 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27050 * void vst1_lane_f32 (float32_t *, float32x2_t, const int)
27051 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27053 * void vst1_lane_p16 (poly16_t *, poly16x4_t, const int)
27054 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27056 * void vst1_lane_p8 (poly8_t *, poly8x8_t, const int)
27057 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27059 * void vst1_lane_s64 (int64_t *, int64x1_t, const int)
27060 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
27062 * void vst1_lane_u64 (uint64_t *, uint64x1_t, const int)
27063 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
27065 * void vst1q_lane_u32 (uint32_t *, uint32x4_t, const int)
27066 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27068 * void vst1q_lane_u16 (uint16_t *, uint16x8_t, const int)
27069 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27071 * void vst1q_lane_u8 (uint8_t *, uint8x16_t, const int)
27072 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27074 * void vst1q_lane_s32 (int32_t *, int32x4_t, const int)
27075 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27077 * void vst1q_lane_s16 (int16_t *, int16x8_t, const int)
27078 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27080 * void vst1q_lane_s8 (int8_t *, int8x16_t, const int)
27081 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27083 * void vst1q_lane_f32 (float32_t *, float32x4_t, const int)
27084 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
27086 * void vst1q_lane_p16 (poly16_t *, poly16x8_t, const int)
27087 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
27089 * void vst1q_lane_p8 (poly8_t *, poly8x16_t, const int)
27090 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
27092 * void vst1q_lane_s64 (int64_t *, int64x2_t, const int)
27093 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
27095 * void vst1q_lane_u64 (uint64_t *, uint64x2_t, const int)
27096 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
27098 5.50.3.70 Element/structure loads, VLD2 variants
27099 ................................................
27101 * uint32x2x2_t vld2_u32 (const uint32_t *)
27102 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27104 * uint16x4x2_t vld2_u16 (const uint16_t *)
27105 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27107 * uint8x8x2_t vld2_u8 (const uint8_t *)
27108 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27110 * int32x2x2_t vld2_s32 (const int32_t *)
27111 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27113 * int16x4x2_t vld2_s16 (const int16_t *)
27114 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27116 * int8x8x2_t vld2_s8 (const int8_t *)
27117 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27119 * float32x2x2_t vld2_f32 (const float32_t *)
27120 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27122 * poly16x4x2_t vld2_p16 (const poly16_t *)
27123 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27125 * poly8x8x2_t vld2_p8 (const poly8_t *)
27126 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27128 * uint64x1x2_t vld2_u64 (const uint64_t *)
27129 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
27131 * int64x1x2_t vld2_s64 (const int64_t *)
27132 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
27134 * uint32x4x2_t vld2q_u32 (const uint32_t *)
27135 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27137 * uint16x8x2_t vld2q_u16 (const uint16_t *)
27138 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27140 * uint8x16x2_t vld2q_u8 (const uint8_t *)
27141 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27143 * int32x4x2_t vld2q_s32 (const int32_t *)
27144 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27146 * int16x8x2_t vld2q_s16 (const int16_t *)
27147 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27149 * int8x16x2_t vld2q_s8 (const int8_t *)
27150 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27152 * float32x4x2_t vld2q_f32 (const float32_t *)
27153 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
27155 * poly16x8x2_t vld2q_p16 (const poly16_t *)
27156 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
27158 * poly8x16x2_t vld2q_p8 (const poly8_t *)
27159 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
27161 * uint32x2x2_t vld2_lane_u32 (const uint32_t *, uint32x2x2_t, const
27163 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27165 * uint16x4x2_t vld2_lane_u16 (const uint16_t *, uint16x4x2_t, const
27167 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27169 * uint8x8x2_t vld2_lane_u8 (const uint8_t *, uint8x8x2_t, const int)
27170 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
27172 * int32x2x2_t vld2_lane_s32 (const int32_t *, int32x2x2_t, const int)
27173 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27175 * int16x4x2_t vld2_lane_s16 (const int16_t *, int16x4x2_t, const int)
27176 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27178 * int8x8x2_t vld2_lane_s8 (const int8_t *, int8x8x2_t, const int)
27179 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
27181 * float32x2x2_t vld2_lane_f32 (const float32_t *, float32x2x2_t,
27183 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27185 * poly16x4x2_t vld2_lane_p16 (const poly16_t *, poly16x4x2_t, const
27187 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27189 * poly8x8x2_t vld2_lane_p8 (const poly8_t *, poly8x8x2_t, const int)
27190 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
27192 * int32x4x2_t vld2q_lane_s32 (const int32_t *, int32x4x2_t, const
27194 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27196 * int16x8x2_t vld2q_lane_s16 (const int16_t *, int16x8x2_t, const
27198 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27200 * uint32x4x2_t vld2q_lane_u32 (const uint32_t *, uint32x4x2_t, const
27202 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27204 * uint16x8x2_t vld2q_lane_u16 (const uint16_t *, uint16x8x2_t, const
27206 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27208 * float32x4x2_t vld2q_lane_f32 (const float32_t *, float32x4x2_t,
27210 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
27212 * poly16x8x2_t vld2q_lane_p16 (const poly16_t *, poly16x8x2_t, const
27214 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
27216 * uint32x2x2_t vld2_dup_u32 (const uint32_t *)
27217 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
27219 * uint16x4x2_t vld2_dup_u16 (const uint16_t *)
27220 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
27222 * uint8x8x2_t vld2_dup_u8 (const uint8_t *)
27223 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
27225 * int32x2x2_t vld2_dup_s32 (const int32_t *)
27226 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
27228 * int16x4x2_t vld2_dup_s16 (const int16_t *)
27229 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
27231 * int8x8x2_t vld2_dup_s8 (const int8_t *)
27232 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
27234 * float32x2x2_t vld2_dup_f32 (const float32_t *)
27235 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
27237 * poly16x4x2_t vld2_dup_p16 (const poly16_t *)
27238 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
27240 * poly8x8x2_t vld2_dup_p8 (const poly8_t *)
27241 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
27243 * uint64x1x2_t vld2_dup_u64 (const uint64_t *)
27244 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
27246 * int64x1x2_t vld2_dup_s64 (const int64_t *)
27247 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
27249 5.50.3.71 Element/structure stores, VST2 variants
27250 .................................................
27252 * void vst2_u32 (uint32_t *, uint32x2x2_t)
27253 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27255 * void vst2_u16 (uint16_t *, uint16x4x2_t)
27256 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27258 * void vst2_u8 (uint8_t *, uint8x8x2_t)
27259 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27261 * void vst2_s32 (int32_t *, int32x2x2_t)
27262 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27264 * void vst2_s16 (int16_t *, int16x4x2_t)
27265 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27267 * void vst2_s8 (int8_t *, int8x8x2_t)
27268 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27270 * void vst2_f32 (float32_t *, float32x2x2_t)
27271 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27273 * void vst2_p16 (poly16_t *, poly16x4x2_t)
27274 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27276 * void vst2_p8 (poly8_t *, poly8x8x2_t)
27277 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27279 * void vst2_u64 (uint64_t *, uint64x1x2_t)
27280 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
27282 * void vst2_s64 (int64_t *, int64x1x2_t)
27283 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
27285 * void vst2q_u32 (uint32_t *, uint32x4x2_t)
27286 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27288 * void vst2q_u16 (uint16_t *, uint16x8x2_t)
27289 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27291 * void vst2q_u8 (uint8_t *, uint8x16x2_t)
27292 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27294 * void vst2q_s32 (int32_t *, int32x4x2_t)
27295 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27297 * void vst2q_s16 (int16_t *, int16x8x2_t)
27298 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27300 * void vst2q_s8 (int8_t *, int8x16x2_t)
27301 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27303 * void vst2q_f32 (float32_t *, float32x4x2_t)
27304 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
27306 * void vst2q_p16 (poly16_t *, poly16x8x2_t)
27307 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
27309 * void vst2q_p8 (poly8_t *, poly8x16x2_t)
27310 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
27312 * void vst2_lane_u32 (uint32_t *, uint32x2x2_t, const int)
27313 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27315 * void vst2_lane_u16 (uint16_t *, uint16x4x2_t, const int)
27316 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27318 * void vst2_lane_u8 (uint8_t *, uint8x8x2_t, const int)
27319 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
27321 * void vst2_lane_s32 (int32_t *, int32x2x2_t, const int)
27322 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27324 * void vst2_lane_s16 (int16_t *, int16x4x2_t, const int)
27325 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27327 * void vst2_lane_s8 (int8_t *, int8x8x2_t, const int)
27328 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
27330 * void vst2_lane_f32 (float32_t *, float32x2x2_t, const int)
27331 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27333 * void vst2_lane_p16 (poly16_t *, poly16x4x2_t, const int)
27334 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27336 * void vst2_lane_p8 (poly8_t *, poly8x8x2_t, const int)
27337 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
27339 * void vst2q_lane_s32 (int32_t *, int32x4x2_t, const int)
27340 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27342 * void vst2q_lane_s16 (int16_t *, int16x8x2_t, const int)
27343 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27345 * void vst2q_lane_u32 (uint32_t *, uint32x4x2_t, const int)
27346 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27348 * void vst2q_lane_u16 (uint16_t *, uint16x8x2_t, const int)
27349 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27351 * void vst2q_lane_f32 (float32_t *, float32x4x2_t, const int)
27352 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
27354 * void vst2q_lane_p16 (poly16_t *, poly16x8x2_t, const int)
27355 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
27357 5.50.3.72 Element/structure loads, VLD3 variants
27358 ................................................
27360 * uint32x2x3_t vld3_u32 (const uint32_t *)
27361 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27363 * uint16x4x3_t vld3_u16 (const uint16_t *)
27364 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27366 * uint8x8x3_t vld3_u8 (const uint8_t *)
27367 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27369 * int32x2x3_t vld3_s32 (const int32_t *)
27370 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27372 * int16x4x3_t vld3_s16 (const int16_t *)
27373 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27375 * int8x8x3_t vld3_s8 (const int8_t *)
27376 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27378 * float32x2x3_t vld3_f32 (const float32_t *)
27379 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27381 * poly16x4x3_t vld3_p16 (const poly16_t *)
27382 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27384 * poly8x8x3_t vld3_p8 (const poly8_t *)
27385 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27387 * uint64x1x3_t vld3_u64 (const uint64_t *)
27388 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
27390 * int64x1x3_t vld3_s64 (const int64_t *)
27391 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
27393 * uint32x4x3_t vld3q_u32 (const uint32_t *)
27394 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27396 * uint16x8x3_t vld3q_u16 (const uint16_t *)
27397 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27399 * uint8x16x3_t vld3q_u8 (const uint8_t *)
27400 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27402 * int32x4x3_t vld3q_s32 (const int32_t *)
27403 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27405 * int16x8x3_t vld3q_s16 (const int16_t *)
27406 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27408 * int8x16x3_t vld3q_s8 (const int8_t *)
27409 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27411 * float32x4x3_t vld3q_f32 (const float32_t *)
27412 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
27414 * poly16x8x3_t vld3q_p16 (const poly16_t *)
27415 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
27417 * poly8x16x3_t vld3q_p8 (const poly8_t *)
27418 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
27420 * uint32x2x3_t vld3_lane_u32 (const uint32_t *, uint32x2x3_t, const
27422 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27425 * uint16x4x3_t vld3_lane_u16 (const uint16_t *, uint16x4x3_t, const
27427 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27430 * uint8x8x3_t vld3_lane_u8 (const uint8_t *, uint8x8x3_t, const int)
27431 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
27434 * int32x2x3_t vld3_lane_s32 (const int32_t *, int32x2x3_t, const int)
27435 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27438 * int16x4x3_t vld3_lane_s16 (const int16_t *, int16x4x3_t, const int)
27439 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27442 * int8x8x3_t vld3_lane_s8 (const int8_t *, int8x8x3_t, const int)
27443 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
27446 * float32x2x3_t vld3_lane_f32 (const float32_t *, float32x2x3_t,
27448 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27451 * poly16x4x3_t vld3_lane_p16 (const poly16_t *, poly16x4x3_t, const
27453 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27456 * poly8x8x3_t vld3_lane_p8 (const poly8_t *, poly8x8x3_t, const int)
27457 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
27460 * int32x4x3_t vld3q_lane_s32 (const int32_t *, int32x4x3_t, const
27462 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27465 * int16x8x3_t vld3q_lane_s16 (const int16_t *, int16x8x3_t, const
27467 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27470 * uint32x4x3_t vld3q_lane_u32 (const uint32_t *, uint32x4x3_t, const
27472 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27475 * uint16x8x3_t vld3q_lane_u16 (const uint16_t *, uint16x8x3_t, const
27477 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27480 * float32x4x3_t vld3q_lane_f32 (const float32_t *, float32x4x3_t,
27482 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
27485 * poly16x8x3_t vld3q_lane_p16 (const poly16_t *, poly16x8x3_t, const
27487 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
27490 * uint32x2x3_t vld3_dup_u32 (const uint32_t *)
27491 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
27494 * uint16x4x3_t vld3_dup_u16 (const uint16_t *)
27495 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
27498 * uint8x8x3_t vld3_dup_u8 (const uint8_t *)
27499 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
27502 * int32x2x3_t vld3_dup_s32 (const int32_t *)
27503 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
27506 * int16x4x3_t vld3_dup_s16 (const int16_t *)
27507 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
27510 * int8x8x3_t vld3_dup_s8 (const int8_t *)
27511 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
27514 * float32x2x3_t vld3_dup_f32 (const float32_t *)
27515 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
27518 * poly16x4x3_t vld3_dup_p16 (const poly16_t *)
27519 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
27522 * poly8x8x3_t vld3_dup_p8 (const poly8_t *)
27523 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
27526 * uint64x1x3_t vld3_dup_u64 (const uint64_t *)
27527 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
27529 * int64x1x3_t vld3_dup_s64 (const int64_t *)
27530 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
27532 5.50.3.73 Element/structure stores, VST3 variants
27533 .................................................
27535 * void vst3_u32 (uint32_t *, uint32x2x3_t)
27536 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
27538 * void vst3_u16 (uint16_t *, uint16x4x3_t)
27539 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
27541 * void vst3_u8 (uint8_t *, uint8x8x3_t)
27542 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
27544 * void vst3_s32 (int32_t *, int32x2x3_t)
27545 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
27547 * void vst3_s16 (int16_t *, int16x4x3_t)
27548 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
27550 * void vst3_s8 (int8_t *, int8x8x3_t)
27551 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
27553 * void vst3_f32 (float32_t *, float32x2x3_t)
27554 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
27556 * void vst3_p16 (poly16_t *, poly16x4x3_t)
27557 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
27559 * void vst3_p8 (poly8_t *, poly8x8x3_t)
27560 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
27562 * void vst3_u64 (uint64_t *, uint64x1x3_t)
27563 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
27565 * void vst3_s64 (int64_t *, int64x1x3_t)
27566 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
27568 * void vst3q_u32 (uint32_t *, uint32x4x3_t)
27569 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
27571 * void vst3q_u16 (uint16_t *, uint16x8x3_t)
27572 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
27574 * void vst3q_u8 (uint8_t *, uint8x16x3_t)
27575 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
27577 * void vst3q_s32 (int32_t *, int32x4x3_t)
27578 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
27580 * void vst3q_s16 (int16_t *, int16x8x3_t)
27581 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
27583 * void vst3q_s8 (int8_t *, int8x16x3_t)
27584 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
27586 * void vst3q_f32 (float32_t *, float32x4x3_t)
27587 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
27589 * void vst3q_p16 (poly16_t *, poly16x8x3_t)
27590 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
27592 * void vst3q_p8 (poly8_t *, poly8x16x3_t)
27593 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
27595 * void vst3_lane_u32 (uint32_t *, uint32x2x3_t, const int)
27596 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27599 * void vst3_lane_u16 (uint16_t *, uint16x4x3_t, const int)
27600 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27603 * void vst3_lane_u8 (uint8_t *, uint8x8x3_t, const int)
27604 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
27607 * void vst3_lane_s32 (int32_t *, int32x2x3_t, const int)
27608 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27611 * void vst3_lane_s16 (int16_t *, int16x4x3_t, const int)
27612 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27615 * void vst3_lane_s8 (int8_t *, int8x8x3_t, const int)
27616 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
27619 * void vst3_lane_f32 (float32_t *, float32x2x3_t, const int)
27620 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27623 * void vst3_lane_p16 (poly16_t *, poly16x4x3_t, const int)
27624 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27627 * void vst3_lane_p8 (poly8_t *, poly8x8x3_t, const int)
27628 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
27631 * void vst3q_lane_s32 (int32_t *, int32x4x3_t, const int)
27632 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27635 * void vst3q_lane_s16 (int16_t *, int16x8x3_t, const int)
27636 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27639 * void vst3q_lane_u32 (uint32_t *, uint32x4x3_t, const int)
27640 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27643 * void vst3q_lane_u16 (uint16_t *, uint16x8x3_t, const int)
27644 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27647 * void vst3q_lane_f32 (float32_t *, float32x4x3_t, const int)
27648 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
27651 * void vst3q_lane_p16 (poly16_t *, poly16x8x3_t, const int)
27652 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
27655 5.50.3.74 Element/structure loads, VLD4 variants
27656 ................................................
27658 * uint32x2x4_t vld4_u32 (const uint32_t *)
27659 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27661 * uint16x4x4_t vld4_u16 (const uint16_t *)
27662 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27664 * uint8x8x4_t vld4_u8 (const uint8_t *)
27665 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27667 * int32x2x4_t vld4_s32 (const int32_t *)
27668 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27670 * int16x4x4_t vld4_s16 (const int16_t *)
27671 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27673 * int8x8x4_t vld4_s8 (const int8_t *)
27674 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27676 * float32x2x4_t vld4_f32 (const float32_t *)
27677 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27679 * poly16x4x4_t vld4_p16 (const poly16_t *)
27680 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27682 * poly8x8x4_t vld4_p8 (const poly8_t *)
27683 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27685 * uint64x1x4_t vld4_u64 (const uint64_t *)
27686 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
27688 * int64x1x4_t vld4_s64 (const int64_t *)
27689 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
27691 * uint32x4x4_t vld4q_u32 (const uint32_t *)
27692 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27694 * uint16x8x4_t vld4q_u16 (const uint16_t *)
27695 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27697 * uint8x16x4_t vld4q_u8 (const uint8_t *)
27698 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27700 * int32x4x4_t vld4q_s32 (const int32_t *)
27701 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27703 * int16x8x4_t vld4q_s16 (const int16_t *)
27704 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27706 * int8x16x4_t vld4q_s8 (const int8_t *)
27707 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27709 * float32x4x4_t vld4q_f32 (const float32_t *)
27710 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
27712 * poly16x8x4_t vld4q_p16 (const poly16_t *)
27713 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
27715 * poly8x16x4_t vld4q_p8 (const poly8_t *)
27716 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
27718 * uint32x2x4_t vld4_lane_u32 (const uint32_t *, uint32x2x4_t, const
27720 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27723 * uint16x4x4_t vld4_lane_u16 (const uint16_t *, uint16x4x4_t, const
27725 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27728 * uint8x8x4_t vld4_lane_u8 (const uint8_t *, uint8x8x4_t, const int)
27729 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
27732 * int32x2x4_t vld4_lane_s32 (const int32_t *, int32x2x4_t, const int)
27733 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27736 * int16x4x4_t vld4_lane_s16 (const int16_t *, int16x4x4_t, const int)
27737 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27740 * int8x8x4_t vld4_lane_s8 (const int8_t *, int8x8x4_t, const int)
27741 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
27744 * float32x2x4_t vld4_lane_f32 (const float32_t *, float32x2x4_t,
27746 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27749 * poly16x4x4_t vld4_lane_p16 (const poly16_t *, poly16x4x4_t, const
27751 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27754 * poly8x8x4_t vld4_lane_p8 (const poly8_t *, poly8x8x4_t, const int)
27755 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
27758 * int32x4x4_t vld4q_lane_s32 (const int32_t *, int32x4x4_t, const
27760 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27763 * int16x8x4_t vld4q_lane_s16 (const int16_t *, int16x8x4_t, const
27765 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27768 * uint32x4x4_t vld4q_lane_u32 (const uint32_t *, uint32x4x4_t, const
27770 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27773 * uint16x8x4_t vld4q_lane_u16 (const uint16_t *, uint16x8x4_t, const
27775 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27778 * float32x4x4_t vld4q_lane_f32 (const float32_t *, float32x4x4_t,
27780 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
27783 * poly16x8x4_t vld4q_lane_p16 (const poly16_t *, poly16x8x4_t, const
27785 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
27788 * uint32x2x4_t vld4_dup_u32 (const uint32_t *)
27789 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
27792 * uint16x4x4_t vld4_dup_u16 (const uint16_t *)
27793 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
27796 * uint8x8x4_t vld4_dup_u8 (const uint8_t *)
27797 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
27800 * int32x2x4_t vld4_dup_s32 (const int32_t *)
27801 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
27804 * int16x4x4_t vld4_dup_s16 (const int16_t *)
27805 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
27808 * int8x8x4_t vld4_dup_s8 (const int8_t *)
27809 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
27812 * float32x2x4_t vld4_dup_f32 (const float32_t *)
27813 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
27816 * poly16x4x4_t vld4_dup_p16 (const poly16_t *)
27817 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
27820 * poly8x8x4_t vld4_dup_p8 (const poly8_t *)
27821 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
27824 * uint64x1x4_t vld4_dup_u64 (const uint64_t *)
27825 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
27827 * int64x1x4_t vld4_dup_s64 (const int64_t *)
27828 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
27830 5.50.3.75 Element/structure stores, VST4 variants
27831 .................................................
27833 * void vst4_u32 (uint32_t *, uint32x2x4_t)
27834 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27836 * void vst4_u16 (uint16_t *, uint16x4x4_t)
27837 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27839 * void vst4_u8 (uint8_t *, uint8x8x4_t)
27840 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27842 * void vst4_s32 (int32_t *, int32x2x4_t)
27843 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27845 * void vst4_s16 (int16_t *, int16x4x4_t)
27846 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27848 * void vst4_s8 (int8_t *, int8x8x4_t)
27849 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27851 * void vst4_f32 (float32_t *, float32x2x4_t)
27852 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27854 * void vst4_p16 (poly16_t *, poly16x4x4_t)
27855 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27857 * void vst4_p8 (poly8_t *, poly8x8x4_t)
27858 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27860 * void vst4_u64 (uint64_t *, uint64x1x4_t)
27861 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
27863 * void vst4_s64 (int64_t *, int64x1x4_t)
27864 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
27866 * void vst4q_u32 (uint32_t *, uint32x4x4_t)
27867 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27869 * void vst4q_u16 (uint16_t *, uint16x8x4_t)
27870 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27872 * void vst4q_u8 (uint8_t *, uint8x16x4_t)
27873 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27875 * void vst4q_s32 (int32_t *, int32x4x4_t)
27876 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27878 * void vst4q_s16 (int16_t *, int16x8x4_t)
27879 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27881 * void vst4q_s8 (int8_t *, int8x16x4_t)
27882 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27884 * void vst4q_f32 (float32_t *, float32x4x4_t)
27885 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
27887 * void vst4q_p16 (poly16_t *, poly16x8x4_t)
27888 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
27890 * void vst4q_p8 (poly8_t *, poly8x16x4_t)
27891 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
27893 * void vst4_lane_u32 (uint32_t *, uint32x2x4_t, const int)
27894 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27897 * void vst4_lane_u16 (uint16_t *, uint16x4x4_t, const int)
27898 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27901 * void vst4_lane_u8 (uint8_t *, uint8x8x4_t, const int)
27902 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
27905 * void vst4_lane_s32 (int32_t *, int32x2x4_t, const int)
27906 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27909 * void vst4_lane_s16 (int16_t *, int16x4x4_t, const int)
27910 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27913 * void vst4_lane_s8 (int8_t *, int8x8x4_t, const int)
27914 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
27917 * void vst4_lane_f32 (float32_t *, float32x2x4_t, const int)
27918 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27921 * void vst4_lane_p16 (poly16_t *, poly16x4x4_t, const int)
27922 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27925 * void vst4_lane_p8 (poly8_t *, poly8x8x4_t, const int)
27926 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
27929 * void vst4q_lane_s32 (int32_t *, int32x4x4_t, const int)
27930 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27933 * void vst4q_lane_s16 (int16_t *, int16x8x4_t, const int)
27934 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27937 * void vst4q_lane_u32 (uint32_t *, uint32x4x4_t, const int)
27938 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27941 * void vst4q_lane_u16 (uint16_t *, uint16x8x4_t, const int)
27942 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27945 * void vst4q_lane_f32 (float32_t *, float32x4x4_t, const int)
27946 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
27949 * void vst4q_lane_p16 (poly16_t *, poly16x8x4_t, const int)
27950 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
27953 5.50.3.76 Logical operations (AND)
27954 ..................................
27956 * uint32x2_t vand_u32 (uint32x2_t, uint32x2_t)
27957 _Form of expected instruction(s):_ `vand D0, D0, D0'
27959 * uint16x4_t vand_u16 (uint16x4_t, uint16x4_t)
27960 _Form of expected instruction(s):_ `vand D0, D0, D0'
27962 * uint8x8_t vand_u8 (uint8x8_t, uint8x8_t)
27963 _Form of expected instruction(s):_ `vand D0, D0, D0'
27965 * int32x2_t vand_s32 (int32x2_t, int32x2_t)
27966 _Form of expected instruction(s):_ `vand D0, D0, D0'
27968 * int16x4_t vand_s16 (int16x4_t, int16x4_t)
27969 _Form of expected instruction(s):_ `vand D0, D0, D0'
27971 * int8x8_t vand_s8 (int8x8_t, int8x8_t)
27972 _Form of expected instruction(s):_ `vand D0, D0, D0'
27974 * uint64x1_t vand_u64 (uint64x1_t, uint64x1_t)
27975 _Form of expected instruction(s):_ `vand D0, D0, D0'
27977 * int64x1_t vand_s64 (int64x1_t, int64x1_t)
27978 _Form of expected instruction(s):_ `vand D0, D0, D0'
27980 * uint32x4_t vandq_u32 (uint32x4_t, uint32x4_t)
27981 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27983 * uint16x8_t vandq_u16 (uint16x8_t, uint16x8_t)
27984 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27986 * uint8x16_t vandq_u8 (uint8x16_t, uint8x16_t)
27987 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27989 * int32x4_t vandq_s32 (int32x4_t, int32x4_t)
27990 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27992 * int16x8_t vandq_s16 (int16x8_t, int16x8_t)
27993 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27995 * int8x16_t vandq_s8 (int8x16_t, int8x16_t)
27996 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
27998 * uint64x2_t vandq_u64 (uint64x2_t, uint64x2_t)
27999 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
28001 * int64x2_t vandq_s64 (int64x2_t, int64x2_t)
28002 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
28004 5.50.3.77 Logical operations (OR)
28005 .................................
28007 * uint32x2_t vorr_u32 (uint32x2_t, uint32x2_t)
28008 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28010 * uint16x4_t vorr_u16 (uint16x4_t, uint16x4_t)
28011 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28013 * uint8x8_t vorr_u8 (uint8x8_t, uint8x8_t)
28014 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28016 * int32x2_t vorr_s32 (int32x2_t, int32x2_t)
28017 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28019 * int16x4_t vorr_s16 (int16x4_t, int16x4_t)
28020 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28022 * int8x8_t vorr_s8 (int8x8_t, int8x8_t)
28023 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28025 * uint64x1_t vorr_u64 (uint64x1_t, uint64x1_t)
28026 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28028 * int64x1_t vorr_s64 (int64x1_t, int64x1_t)
28029 _Form of expected instruction(s):_ `vorr D0, D0, D0'
28031 * uint32x4_t vorrq_u32 (uint32x4_t, uint32x4_t)
28032 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28034 * uint16x8_t vorrq_u16 (uint16x8_t, uint16x8_t)
28035 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28037 * uint8x16_t vorrq_u8 (uint8x16_t, uint8x16_t)
28038 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28040 * int32x4_t vorrq_s32 (int32x4_t, int32x4_t)
28041 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28043 * int16x8_t vorrq_s16 (int16x8_t, int16x8_t)
28044 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28046 * int8x16_t vorrq_s8 (int8x16_t, int8x16_t)
28047 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28049 * uint64x2_t vorrq_u64 (uint64x2_t, uint64x2_t)
28050 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28052 * int64x2_t vorrq_s64 (int64x2_t, int64x2_t)
28053 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
28055 5.50.3.78 Logical operations (exclusive OR)
28056 ...........................................
28058 * uint32x2_t veor_u32 (uint32x2_t, uint32x2_t)
28059 _Form of expected instruction(s):_ `veor D0, D0, D0'
28061 * uint16x4_t veor_u16 (uint16x4_t, uint16x4_t)
28062 _Form of expected instruction(s):_ `veor D0, D0, D0'
28064 * uint8x8_t veor_u8 (uint8x8_t, uint8x8_t)
28065 _Form of expected instruction(s):_ `veor D0, D0, D0'
28067 * int32x2_t veor_s32 (int32x2_t, int32x2_t)
28068 _Form of expected instruction(s):_ `veor D0, D0, D0'
28070 * int16x4_t veor_s16 (int16x4_t, int16x4_t)
28071 _Form of expected instruction(s):_ `veor D0, D0, D0'
28073 * int8x8_t veor_s8 (int8x8_t, int8x8_t)
28074 _Form of expected instruction(s):_ `veor D0, D0, D0'
28076 * uint64x1_t veor_u64 (uint64x1_t, uint64x1_t)
28077 _Form of expected instruction(s):_ `veor D0, D0, D0'
28079 * int64x1_t veor_s64 (int64x1_t, int64x1_t)
28080 _Form of expected instruction(s):_ `veor D0, D0, D0'
28082 * uint32x4_t veorq_u32 (uint32x4_t, uint32x4_t)
28083 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28085 * uint16x8_t veorq_u16 (uint16x8_t, uint16x8_t)
28086 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28088 * uint8x16_t veorq_u8 (uint8x16_t, uint8x16_t)
28089 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28091 * int32x4_t veorq_s32 (int32x4_t, int32x4_t)
28092 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28094 * int16x8_t veorq_s16 (int16x8_t, int16x8_t)
28095 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28097 * int8x16_t veorq_s8 (int8x16_t, int8x16_t)
28098 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28100 * uint64x2_t veorq_u64 (uint64x2_t, uint64x2_t)
28101 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28103 * int64x2_t veorq_s64 (int64x2_t, int64x2_t)
28104 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
28106 5.50.3.79 Logical operations (AND-NOT)
28107 ......................................
28109 * uint32x2_t vbic_u32 (uint32x2_t, uint32x2_t)
28110 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28112 * uint16x4_t vbic_u16 (uint16x4_t, uint16x4_t)
28113 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28115 * uint8x8_t vbic_u8 (uint8x8_t, uint8x8_t)
28116 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28118 * int32x2_t vbic_s32 (int32x2_t, int32x2_t)
28119 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28121 * int16x4_t vbic_s16 (int16x4_t, int16x4_t)
28122 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28124 * int8x8_t vbic_s8 (int8x8_t, int8x8_t)
28125 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28127 * uint64x1_t vbic_u64 (uint64x1_t, uint64x1_t)
28128 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28130 * int64x1_t vbic_s64 (int64x1_t, int64x1_t)
28131 _Form of expected instruction(s):_ `vbic D0, D0, D0'
28133 * uint32x4_t vbicq_u32 (uint32x4_t, uint32x4_t)
28134 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28136 * uint16x8_t vbicq_u16 (uint16x8_t, uint16x8_t)
28137 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28139 * uint8x16_t vbicq_u8 (uint8x16_t, uint8x16_t)
28140 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28142 * int32x4_t vbicq_s32 (int32x4_t, int32x4_t)
28143 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28145 * int16x8_t vbicq_s16 (int16x8_t, int16x8_t)
28146 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28148 * int8x16_t vbicq_s8 (int8x16_t, int8x16_t)
28149 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28151 * uint64x2_t vbicq_u64 (uint64x2_t, uint64x2_t)
28152 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28154 * int64x2_t vbicq_s64 (int64x2_t, int64x2_t)
28155 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
28157 5.50.3.80 Logical operations (OR-NOT)
28158 .....................................
28160 * uint32x2_t vorn_u32 (uint32x2_t, uint32x2_t)
28161 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28163 * uint16x4_t vorn_u16 (uint16x4_t, uint16x4_t)
28164 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28166 * uint8x8_t vorn_u8 (uint8x8_t, uint8x8_t)
28167 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28169 * int32x2_t vorn_s32 (int32x2_t, int32x2_t)
28170 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28172 * int16x4_t vorn_s16 (int16x4_t, int16x4_t)
28173 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28175 * int8x8_t vorn_s8 (int8x8_t, int8x8_t)
28176 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28178 * uint64x1_t vorn_u64 (uint64x1_t, uint64x1_t)
28179 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28181 * int64x1_t vorn_s64 (int64x1_t, int64x1_t)
28182 _Form of expected instruction(s):_ `vorn D0, D0, D0'
28184 * uint32x4_t vornq_u32 (uint32x4_t, uint32x4_t)
28185 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28187 * uint16x8_t vornq_u16 (uint16x8_t, uint16x8_t)
28188 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28190 * uint8x16_t vornq_u8 (uint8x16_t, uint8x16_t)
28191 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28193 * int32x4_t vornq_s32 (int32x4_t, int32x4_t)
28194 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28196 * int16x8_t vornq_s16 (int16x8_t, int16x8_t)
28197 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28199 * int8x16_t vornq_s8 (int8x16_t, int8x16_t)
28200 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28202 * uint64x2_t vornq_u64 (uint64x2_t, uint64x2_t)
28203 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28205 * int64x2_t vornq_s64 (int64x2_t, int64x2_t)
28206 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
28208 5.50.3.81 Reinterpret casts
28209 ...........................
28211 * poly8x8_t vreinterpret_p8_u32 (uint32x2_t)
28213 * poly8x8_t vreinterpret_p8_u16 (uint16x4_t)
28215 * poly8x8_t vreinterpret_p8_u8 (uint8x8_t)
28217 * poly8x8_t vreinterpret_p8_s32 (int32x2_t)
28219 * poly8x8_t vreinterpret_p8_s16 (int16x4_t)
28221 * poly8x8_t vreinterpret_p8_s8 (int8x8_t)
28223 * poly8x8_t vreinterpret_p8_u64 (uint64x1_t)
28225 * poly8x8_t vreinterpret_p8_s64 (int64x1_t)
28227 * poly8x8_t vreinterpret_p8_f32 (float32x2_t)
28229 * poly8x8_t vreinterpret_p8_p16 (poly16x4_t)
28231 * poly8x16_t vreinterpretq_p8_u32 (uint32x4_t)
28233 * poly8x16_t vreinterpretq_p8_u16 (uint16x8_t)
28235 * poly8x16_t vreinterpretq_p8_u8 (uint8x16_t)
28237 * poly8x16_t vreinterpretq_p8_s32 (int32x4_t)
28239 * poly8x16_t vreinterpretq_p8_s16 (int16x8_t)
28241 * poly8x16_t vreinterpretq_p8_s8 (int8x16_t)
28243 * poly8x16_t vreinterpretq_p8_u64 (uint64x2_t)
28245 * poly8x16_t vreinterpretq_p8_s64 (int64x2_t)
28247 * poly8x16_t vreinterpretq_p8_f32 (float32x4_t)
28249 * poly8x16_t vreinterpretq_p8_p16 (poly16x8_t)
28251 * poly16x4_t vreinterpret_p16_u32 (uint32x2_t)
28253 * poly16x4_t vreinterpret_p16_u16 (uint16x4_t)
28255 * poly16x4_t vreinterpret_p16_u8 (uint8x8_t)
28257 * poly16x4_t vreinterpret_p16_s32 (int32x2_t)
28259 * poly16x4_t vreinterpret_p16_s16 (int16x4_t)
28261 * poly16x4_t vreinterpret_p16_s8 (int8x8_t)
28263 * poly16x4_t vreinterpret_p16_u64 (uint64x1_t)
28265 * poly16x4_t vreinterpret_p16_s64 (int64x1_t)
28267 * poly16x4_t vreinterpret_p16_f32 (float32x2_t)
28269 * poly16x4_t vreinterpret_p16_p8 (poly8x8_t)
28271 * poly16x8_t vreinterpretq_p16_u32 (uint32x4_t)
28273 * poly16x8_t vreinterpretq_p16_u16 (uint16x8_t)
28275 * poly16x8_t vreinterpretq_p16_u8 (uint8x16_t)
28277 * poly16x8_t vreinterpretq_p16_s32 (int32x4_t)
28279 * poly16x8_t vreinterpretq_p16_s16 (int16x8_t)
28281 * poly16x8_t vreinterpretq_p16_s8 (int8x16_t)
28283 * poly16x8_t vreinterpretq_p16_u64 (uint64x2_t)
28285 * poly16x8_t vreinterpretq_p16_s64 (int64x2_t)
28287 * poly16x8_t vreinterpretq_p16_f32 (float32x4_t)
28289 * poly16x8_t vreinterpretq_p16_p8 (poly8x16_t)
28291 * float32x2_t vreinterpret_f32_u32 (uint32x2_t)
28293 * float32x2_t vreinterpret_f32_u16 (uint16x4_t)
28295 * float32x2_t vreinterpret_f32_u8 (uint8x8_t)
28297 * float32x2_t vreinterpret_f32_s32 (int32x2_t)
28299 * float32x2_t vreinterpret_f32_s16 (int16x4_t)
28301 * float32x2_t vreinterpret_f32_s8 (int8x8_t)
28303 * float32x2_t vreinterpret_f32_u64 (uint64x1_t)
28305 * float32x2_t vreinterpret_f32_s64 (int64x1_t)
28307 * float32x2_t vreinterpret_f32_p16 (poly16x4_t)
28309 * float32x2_t vreinterpret_f32_p8 (poly8x8_t)
28311 * float32x4_t vreinterpretq_f32_u32 (uint32x4_t)
28313 * float32x4_t vreinterpretq_f32_u16 (uint16x8_t)
28315 * float32x4_t vreinterpretq_f32_u8 (uint8x16_t)
28317 * float32x4_t vreinterpretq_f32_s32 (int32x4_t)
28319 * float32x4_t vreinterpretq_f32_s16 (int16x8_t)
28321 * float32x4_t vreinterpretq_f32_s8 (int8x16_t)
28323 * float32x4_t vreinterpretq_f32_u64 (uint64x2_t)
28325 * float32x4_t vreinterpretq_f32_s64 (int64x2_t)
28327 * float32x4_t vreinterpretq_f32_p16 (poly16x8_t)
28329 * float32x4_t vreinterpretq_f32_p8 (poly8x16_t)
28331 * int64x1_t vreinterpret_s64_u32 (uint32x2_t)
28333 * int64x1_t vreinterpret_s64_u16 (uint16x4_t)
28335 * int64x1_t vreinterpret_s64_u8 (uint8x8_t)
28337 * int64x1_t vreinterpret_s64_s32 (int32x2_t)
28339 * int64x1_t vreinterpret_s64_s16 (int16x4_t)
28341 * int64x1_t vreinterpret_s64_s8 (int8x8_t)
28343 * int64x1_t vreinterpret_s64_u64 (uint64x1_t)
28345 * int64x1_t vreinterpret_s64_f32 (float32x2_t)
28347 * int64x1_t vreinterpret_s64_p16 (poly16x4_t)
28349 * int64x1_t vreinterpret_s64_p8 (poly8x8_t)
28351 * int64x2_t vreinterpretq_s64_u32 (uint32x4_t)
28353 * int64x2_t vreinterpretq_s64_u16 (uint16x8_t)
28355 * int64x2_t vreinterpretq_s64_u8 (uint8x16_t)
28357 * int64x2_t vreinterpretq_s64_s32 (int32x4_t)
28359 * int64x2_t vreinterpretq_s64_s16 (int16x8_t)
28361 * int64x2_t vreinterpretq_s64_s8 (int8x16_t)
28363 * int64x2_t vreinterpretq_s64_u64 (uint64x2_t)
28365 * int64x2_t vreinterpretq_s64_f32 (float32x4_t)
28367 * int64x2_t vreinterpretq_s64_p16 (poly16x8_t)
28369 * int64x2_t vreinterpretq_s64_p8 (poly8x16_t)
28371 * uint64x1_t vreinterpret_u64_u32 (uint32x2_t)
28373 * uint64x1_t vreinterpret_u64_u16 (uint16x4_t)
28375 * uint64x1_t vreinterpret_u64_u8 (uint8x8_t)
28377 * uint64x1_t vreinterpret_u64_s32 (int32x2_t)
28379 * uint64x1_t vreinterpret_u64_s16 (int16x4_t)
28381 * uint64x1_t vreinterpret_u64_s8 (int8x8_t)
28383 * uint64x1_t vreinterpret_u64_s64 (int64x1_t)
28385 * uint64x1_t vreinterpret_u64_f32 (float32x2_t)
28387 * uint64x1_t vreinterpret_u64_p16 (poly16x4_t)
28389 * uint64x1_t vreinterpret_u64_p8 (poly8x8_t)
28391 * uint64x2_t vreinterpretq_u64_u32 (uint32x4_t)
28393 * uint64x2_t vreinterpretq_u64_u16 (uint16x8_t)
28395 * uint64x2_t vreinterpretq_u64_u8 (uint8x16_t)
28397 * uint64x2_t vreinterpretq_u64_s32 (int32x4_t)
28399 * uint64x2_t vreinterpretq_u64_s16 (int16x8_t)
28401 * uint64x2_t vreinterpretq_u64_s8 (int8x16_t)
28403 * uint64x2_t vreinterpretq_u64_s64 (int64x2_t)
28405 * uint64x2_t vreinterpretq_u64_f32 (float32x4_t)
28407 * uint64x2_t vreinterpretq_u64_p16 (poly16x8_t)
28409 * uint64x2_t vreinterpretq_u64_p8 (poly8x16_t)
28411 * int8x8_t vreinterpret_s8_u32 (uint32x2_t)
28413 * int8x8_t vreinterpret_s8_u16 (uint16x4_t)
28415 * int8x8_t vreinterpret_s8_u8 (uint8x8_t)
28417 * int8x8_t vreinterpret_s8_s32 (int32x2_t)
28419 * int8x8_t vreinterpret_s8_s16 (int16x4_t)
28421 * int8x8_t vreinterpret_s8_u64 (uint64x1_t)
28423 * int8x8_t vreinterpret_s8_s64 (int64x1_t)
28425 * int8x8_t vreinterpret_s8_f32 (float32x2_t)
28427 * int8x8_t vreinterpret_s8_p16 (poly16x4_t)
28429 * int8x8_t vreinterpret_s8_p8 (poly8x8_t)
28431 * int8x16_t vreinterpretq_s8_u32 (uint32x4_t)
28433 * int8x16_t vreinterpretq_s8_u16 (uint16x8_t)
28435 * int8x16_t vreinterpretq_s8_u8 (uint8x16_t)
28437 * int8x16_t vreinterpretq_s8_s32 (int32x4_t)
28439 * int8x16_t vreinterpretq_s8_s16 (int16x8_t)
28441 * int8x16_t vreinterpretq_s8_u64 (uint64x2_t)
28443 * int8x16_t vreinterpretq_s8_s64 (int64x2_t)
28445 * int8x16_t vreinterpretq_s8_f32 (float32x4_t)
28447 * int8x16_t vreinterpretq_s8_p16 (poly16x8_t)
28449 * int8x16_t vreinterpretq_s8_p8 (poly8x16_t)
28451 * int16x4_t vreinterpret_s16_u32 (uint32x2_t)
28453 * int16x4_t vreinterpret_s16_u16 (uint16x4_t)
28455 * int16x4_t vreinterpret_s16_u8 (uint8x8_t)
28457 * int16x4_t vreinterpret_s16_s32 (int32x2_t)
28459 * int16x4_t vreinterpret_s16_s8 (int8x8_t)
28461 * int16x4_t vreinterpret_s16_u64 (uint64x1_t)
28463 * int16x4_t vreinterpret_s16_s64 (int64x1_t)
28465 * int16x4_t vreinterpret_s16_f32 (float32x2_t)
28467 * int16x4_t vreinterpret_s16_p16 (poly16x4_t)
28469 * int16x4_t vreinterpret_s16_p8 (poly8x8_t)
28471 * int16x8_t vreinterpretq_s16_u32 (uint32x4_t)
28473 * int16x8_t vreinterpretq_s16_u16 (uint16x8_t)
28475 * int16x8_t vreinterpretq_s16_u8 (uint8x16_t)
28477 * int16x8_t vreinterpretq_s16_s32 (int32x4_t)
28479 * int16x8_t vreinterpretq_s16_s8 (int8x16_t)
28481 * int16x8_t vreinterpretq_s16_u64 (uint64x2_t)
28483 * int16x8_t vreinterpretq_s16_s64 (int64x2_t)
28485 * int16x8_t vreinterpretq_s16_f32 (float32x4_t)
28487 * int16x8_t vreinterpretq_s16_p16 (poly16x8_t)
28489 * int16x8_t vreinterpretq_s16_p8 (poly8x16_t)
28491 * int32x2_t vreinterpret_s32_u32 (uint32x2_t)
28493 * int32x2_t vreinterpret_s32_u16 (uint16x4_t)
28495 * int32x2_t vreinterpret_s32_u8 (uint8x8_t)
28497 * int32x2_t vreinterpret_s32_s16 (int16x4_t)
28499 * int32x2_t vreinterpret_s32_s8 (int8x8_t)
28501 * int32x2_t vreinterpret_s32_u64 (uint64x1_t)
28503 * int32x2_t vreinterpret_s32_s64 (int64x1_t)
28505 * int32x2_t vreinterpret_s32_f32 (float32x2_t)
28507 * int32x2_t vreinterpret_s32_p16 (poly16x4_t)
28509 * int32x2_t vreinterpret_s32_p8 (poly8x8_t)
28511 * int32x4_t vreinterpretq_s32_u32 (uint32x4_t)
28513 * int32x4_t vreinterpretq_s32_u16 (uint16x8_t)
28515 * int32x4_t vreinterpretq_s32_u8 (uint8x16_t)
28517 * int32x4_t vreinterpretq_s32_s16 (int16x8_t)
28519 * int32x4_t vreinterpretq_s32_s8 (int8x16_t)
28521 * int32x4_t vreinterpretq_s32_u64 (uint64x2_t)
28523 * int32x4_t vreinterpretq_s32_s64 (int64x2_t)
28525 * int32x4_t vreinterpretq_s32_f32 (float32x4_t)
28527 * int32x4_t vreinterpretq_s32_p16 (poly16x8_t)
28529 * int32x4_t vreinterpretq_s32_p8 (poly8x16_t)
28531 * uint8x8_t vreinterpret_u8_u32 (uint32x2_t)
28533 * uint8x8_t vreinterpret_u8_u16 (uint16x4_t)
28535 * uint8x8_t vreinterpret_u8_s32 (int32x2_t)
28537 * uint8x8_t vreinterpret_u8_s16 (int16x4_t)
28539 * uint8x8_t vreinterpret_u8_s8 (int8x8_t)
28541 * uint8x8_t vreinterpret_u8_u64 (uint64x1_t)
28543 * uint8x8_t vreinterpret_u8_s64 (int64x1_t)
28545 * uint8x8_t vreinterpret_u8_f32 (float32x2_t)
28547 * uint8x8_t vreinterpret_u8_p16 (poly16x4_t)
28549 * uint8x8_t vreinterpret_u8_p8 (poly8x8_t)
28551 * uint8x16_t vreinterpretq_u8_u32 (uint32x4_t)
28553 * uint8x16_t vreinterpretq_u8_u16 (uint16x8_t)
28555 * uint8x16_t vreinterpretq_u8_s32 (int32x4_t)
28557 * uint8x16_t vreinterpretq_u8_s16 (int16x8_t)
28559 * uint8x16_t vreinterpretq_u8_s8 (int8x16_t)
28561 * uint8x16_t vreinterpretq_u8_u64 (uint64x2_t)
28563 * uint8x16_t vreinterpretq_u8_s64 (int64x2_t)
28565 * uint8x16_t vreinterpretq_u8_f32 (float32x4_t)
28567 * uint8x16_t vreinterpretq_u8_p16 (poly16x8_t)
28569 * uint8x16_t vreinterpretq_u8_p8 (poly8x16_t)
28571 * uint16x4_t vreinterpret_u16_u32 (uint32x2_t)
28573 * uint16x4_t vreinterpret_u16_u8 (uint8x8_t)
28575 * uint16x4_t vreinterpret_u16_s32 (int32x2_t)
28577 * uint16x4_t vreinterpret_u16_s16 (int16x4_t)
28579 * uint16x4_t vreinterpret_u16_s8 (int8x8_t)
28581 * uint16x4_t vreinterpret_u16_u64 (uint64x1_t)
28583 * uint16x4_t vreinterpret_u16_s64 (int64x1_t)
28585 * uint16x4_t vreinterpret_u16_f32 (float32x2_t)
28587 * uint16x4_t vreinterpret_u16_p16 (poly16x4_t)
28589 * uint16x4_t vreinterpret_u16_p8 (poly8x8_t)
28591 * uint16x8_t vreinterpretq_u16_u32 (uint32x4_t)
28593 * uint16x8_t vreinterpretq_u16_u8 (uint8x16_t)
28595 * uint16x8_t vreinterpretq_u16_s32 (int32x4_t)
28597 * uint16x8_t vreinterpretq_u16_s16 (int16x8_t)
28599 * uint16x8_t vreinterpretq_u16_s8 (int8x16_t)
28601 * uint16x8_t vreinterpretq_u16_u64 (uint64x2_t)
28603 * uint16x8_t vreinterpretq_u16_s64 (int64x2_t)
28605 * uint16x8_t vreinterpretq_u16_f32 (float32x4_t)
28607 * uint16x8_t vreinterpretq_u16_p16 (poly16x8_t)
28609 * uint16x8_t vreinterpretq_u16_p8 (poly8x16_t)
28611 * uint32x2_t vreinterpret_u32_u16 (uint16x4_t)
28613 * uint32x2_t vreinterpret_u32_u8 (uint8x8_t)
28615 * uint32x2_t vreinterpret_u32_s32 (int32x2_t)
28617 * uint32x2_t vreinterpret_u32_s16 (int16x4_t)
28619 * uint32x2_t vreinterpret_u32_s8 (int8x8_t)
28621 * uint32x2_t vreinterpret_u32_u64 (uint64x1_t)
28623 * uint32x2_t vreinterpret_u32_s64 (int64x1_t)
28625 * uint32x2_t vreinterpret_u32_f32 (float32x2_t)
28627 * uint32x2_t vreinterpret_u32_p16 (poly16x4_t)
28629 * uint32x2_t vreinterpret_u32_p8 (poly8x8_t)
28631 * uint32x4_t vreinterpretq_u32_u16 (uint16x8_t)
28633 * uint32x4_t vreinterpretq_u32_u8 (uint8x16_t)
28635 * uint32x4_t vreinterpretq_u32_s32 (int32x4_t)
28637 * uint32x4_t vreinterpretq_u32_s16 (int16x8_t)
28639 * uint32x4_t vreinterpretq_u32_s8 (int8x16_t)
28641 * uint32x4_t vreinterpretq_u32_u64 (uint64x2_t)
28643 * uint32x4_t vreinterpretq_u32_s64 (int64x2_t)
28645 * uint32x4_t vreinterpretq_u32_f32 (float32x4_t)
28647 * uint32x4_t vreinterpretq_u32_p16 (poly16x8_t)
28649 * uint32x4_t vreinterpretq_u32_p8 (poly8x16_t)
28652 File: gcc.info, Node: Blackfin Built-in Functions, Next: FR-V Built-in Functions, Prev: ARM NEON Intrinsics, Up: Target Builtins
28654 5.50.4 Blackfin Built-in Functions
28655 ----------------------------------
28657 Currently, there are two Blackfin-specific built-in functions. These
28658 are used for generating `CSYNC' and `SSYNC' machine insns without using
28659 inline assembly; by using these built-in functions the compiler can
28660 automatically add workarounds for hardware errata involving these
28661 instructions. These functions are named as follows:
28663 void __builtin_bfin_csync (void)
28664 void __builtin_bfin_ssync (void)
28667 File: gcc.info, Node: FR-V Built-in Functions, Next: X86 Built-in Functions, Prev: Blackfin Built-in Functions, Up: Target Builtins
28669 5.50.5 FR-V Built-in Functions
28670 ------------------------------
28672 GCC provides many FR-V-specific built-in functions. In general, these
28673 functions are intended to be compatible with those described by `FR-V
28674 Family, Softune C/C++ Compiler Manual (V6), Fujitsu Semiconductor'.
28675 The two exceptions are `__MDUNPACKH' and `__MBTOHE', the gcc forms of
28676 which pass 128-bit values by pointer rather than by value.
28678 Most of the functions are named after specific FR-V instructions.
28679 Such functions are said to be "directly mapped" and are summarized here
28685 * Directly-mapped Integer Functions::
28686 * Directly-mapped Media Functions::
28687 * Raw read/write Functions::
28688 * Other Built-in Functions::
28691 File: gcc.info, Node: Argument Types, Next: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
28693 5.50.5.1 Argument Types
28694 .......................
28696 The arguments to the built-in functions can be divided into three
28697 groups: register numbers, compile-time constants and run-time values.
28698 In order to make this classification clear at a glance, the arguments
28699 and return values are given the following pseudo types:
28701 Pseudo type Real C type Constant? Description
28702 `uh' `unsigned short' No an unsigned halfword
28703 `uw1' `unsigned int' No an unsigned word
28704 `sw1' `int' No a signed word
28705 `uw2' `unsigned long long' No an unsigned doubleword
28706 `sw2' `long long' No a signed doubleword
28707 `const' `int' Yes an integer constant
28708 `acc' `int' Yes an ACC register number
28709 `iacc' `int' Yes an IACC register number
28711 These pseudo types are not defined by GCC, they are simply a notational
28712 convenience used in this manual.
28714 Arguments of type `uh', `uw1', `sw1', `uw2' and `sw2' are evaluated at
28715 run time. They correspond to register operands in the underlying FR-V
28718 `const' arguments represent immediate operands in the underlying FR-V
28719 instructions. They must be compile-time constants.
28721 `acc' arguments are evaluated at compile time and specify the number
28722 of an accumulator register. For example, an `acc' argument of 2 will
28723 select the ACC2 register.
28725 `iacc' arguments are similar to `acc' arguments but specify the number
28726 of an IACC register. See *note Other Built-in Functions:: for more
28730 File: gcc.info, Node: Directly-mapped Integer Functions, Next: Directly-mapped Media Functions, Prev: Argument Types, Up: FR-V Built-in Functions
28732 5.50.5.2 Directly-mapped Integer Functions
28733 ..........................................
28735 The functions listed below map directly to FR-V I-type instructions.
28737 Function prototype Example usage Assembly output
28738 `sw1 __ADDSS (sw1, sw1)' `C = __ADDSS (A, B)' `ADDSS A,B,C'
28739 `sw1 __SCAN (sw1, sw1)' `C = __SCAN (A, B)' `SCAN A,B,C'
28740 `sw1 __SCUTSS (sw1)' `B = __SCUTSS (A)' `SCUTSS A,B'
28741 `sw1 __SLASS (sw1, sw1)' `C = __SLASS (A, B)' `SLASS A,B,C'
28742 `void __SMASS (sw1, sw1)' `__SMASS (A, B)' `SMASS A,B'
28743 `void __SMSSS (sw1, sw1)' `__SMSSS (A, B)' `SMSSS A,B'
28744 `void __SMU (sw1, sw1)' `__SMU (A, B)' `SMU A,B'
28745 `sw2 __SMUL (sw1, sw1)' `C = __SMUL (A, B)' `SMUL A,B,C'
28746 `sw1 __SUBSS (sw1, sw1)' `C = __SUBSS (A, B)' `SUBSS A,B,C'
28747 `uw2 __UMUL (uw1, uw1)' `C = __UMUL (A, B)' `UMUL A,B,C'
28750 File: gcc.info, Node: Directly-mapped Media Functions, Next: Raw read/write Functions, Prev: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
28752 5.50.5.3 Directly-mapped Media Functions
28753 ........................................
28755 The functions listed below map directly to FR-V M-type instructions.
28757 Function prototype Example usage Assembly output
28758 `uw1 __MABSHS (sw1)' `B = __MABSHS (A)' `MABSHS A,B'
28759 `void __MADDACCS (acc, acc)' `__MADDACCS (B, A)' `MADDACCS A,B'
28760 `sw1 __MADDHSS (sw1, sw1)' `C = __MADDHSS (A, B)' `MADDHSS A,B,C'
28761 `uw1 __MADDHUS (uw1, uw1)' `C = __MADDHUS (A, B)' `MADDHUS A,B,C'
28762 `uw1 __MAND (uw1, uw1)' `C = __MAND (A, B)' `MAND A,B,C'
28763 `void __MASACCS (acc, acc)' `__MASACCS (B, A)' `MASACCS A,B'
28764 `uw1 __MAVEH (uw1, uw1)' `C = __MAVEH (A, B)' `MAVEH A,B,C'
28765 `uw2 __MBTOH (uw1)' `B = __MBTOH (A)' `MBTOH A,B'
28766 `void __MBTOHE (uw1 *, uw1)' `__MBTOHE (&B, A)' `MBTOHE A,B'
28767 `void __MCLRACC (acc)' `__MCLRACC (A)' `MCLRACC A'
28768 `void __MCLRACCA (void)' `__MCLRACCA ()' `MCLRACCA'
28769 `uw1 __Mcop1 (uw1, uw1)' `C = __Mcop1 (A, B)' `Mcop1 A,B,C'
28770 `uw1 __Mcop2 (uw1, uw1)' `C = __Mcop2 (A, B)' `Mcop2 A,B,C'
28771 `uw1 __MCPLHI (uw2, const)' `C = __MCPLHI (A, B)' `MCPLHI A,#B,C'
28772 `uw1 __MCPLI (uw2, const)' `C = __MCPLI (A, B)' `MCPLI A,#B,C'
28773 `void __MCPXIS (acc, sw1, sw1)' `__MCPXIS (C, A, B)' `MCPXIS A,B,C'
28774 `void __MCPXIU (acc, uw1, uw1)' `__MCPXIU (C, A, B)' `MCPXIU A,B,C'
28775 `void __MCPXRS (acc, sw1, sw1)' `__MCPXRS (C, A, B)' `MCPXRS A,B,C'
28776 `void __MCPXRU (acc, uw1, uw1)' `__MCPXRU (C, A, B)' `MCPXRU A,B,C'
28777 `uw1 __MCUT (acc, uw1)' `C = __MCUT (A, B)' `MCUT A,B,C'
28778 `uw1 __MCUTSS (acc, sw1)' `C = __MCUTSS (A, B)' `MCUTSS A,B,C'
28779 `void __MDADDACCS (acc, acc)' `__MDADDACCS (B, A)' `MDADDACCS A,B'
28780 `void __MDASACCS (acc, acc)' `__MDASACCS (B, A)' `MDASACCS A,B'
28781 `uw2 __MDCUTSSI (acc, const)' `C = __MDCUTSSI (A, B)' `MDCUTSSI A,#B,C'
28782 `uw2 __MDPACKH (uw2, uw2)' `C = __MDPACKH (A, B)' `MDPACKH A,B,C'
28783 `uw2 __MDROTLI (uw2, const)' `C = __MDROTLI (A, B)' `MDROTLI A,#B,C'
28784 `void __MDSUBACCS (acc, acc)' `__MDSUBACCS (B, A)' `MDSUBACCS A,B'
28785 `void __MDUNPACKH (uw1 *, uw2)' `__MDUNPACKH (&B, A)' `MDUNPACKH A,B'
28786 `uw2 __MEXPDHD (uw1, const)' `C = __MEXPDHD (A, B)' `MEXPDHD A,#B,C'
28787 `uw1 __MEXPDHW (uw1, const)' `C = __MEXPDHW (A, B)' `MEXPDHW A,#B,C'
28788 `uw1 __MHDSETH (uw1, const)' `C = __MHDSETH (A, B)' `MHDSETH A,#B,C'
28789 `sw1 __MHDSETS (const)' `B = __MHDSETS (A)' `MHDSETS #A,B'
28790 `uw1 __MHSETHIH (uw1, const)' `B = __MHSETHIH (B, A)' `MHSETHIH #A,B'
28791 `sw1 __MHSETHIS (sw1, const)' `B = __MHSETHIS (B, A)' `MHSETHIS #A,B'
28792 `uw1 __MHSETLOH (uw1, const)' `B = __MHSETLOH (B, A)' `MHSETLOH #A,B'
28793 `sw1 __MHSETLOS (sw1, const)' `B = __MHSETLOS (B, A)' `MHSETLOS #A,B'
28794 `uw1 __MHTOB (uw2)' `B = __MHTOB (A)' `MHTOB A,B'
28795 `void __MMACHS (acc, sw1, sw1)' `__MMACHS (C, A, B)' `MMACHS A,B,C'
28796 `void __MMACHU (acc, uw1, uw1)' `__MMACHU (C, A, B)' `MMACHU A,B,C'
28797 `void __MMRDHS (acc, sw1, sw1)' `__MMRDHS (C, A, B)' `MMRDHS A,B,C'
28798 `void __MMRDHU (acc, uw1, uw1)' `__MMRDHU (C, A, B)' `MMRDHU A,B,C'
28799 `void __MMULHS (acc, sw1, sw1)' `__MMULHS (C, A, B)' `MMULHS A,B,C'
28800 `void __MMULHU (acc, uw1, uw1)' `__MMULHU (C, A, B)' `MMULHU A,B,C'
28801 `void __MMULXHS (acc, sw1, sw1)' `__MMULXHS (C, A, B)' `MMULXHS A,B,C'
28802 `void __MMULXHU (acc, uw1, uw1)' `__MMULXHU (C, A, B)' `MMULXHU A,B,C'
28803 `uw1 __MNOT (uw1)' `B = __MNOT (A)' `MNOT A,B'
28804 `uw1 __MOR (uw1, uw1)' `C = __MOR (A, B)' `MOR A,B,C'
28805 `uw1 __MPACKH (uh, uh)' `C = __MPACKH (A, B)' `MPACKH A,B,C'
28806 `sw2 __MQADDHSS (sw2, sw2)' `C = __MQADDHSS (A, B)' `MQADDHSS A,B,C'
28807 `uw2 __MQADDHUS (uw2, uw2)' `C = __MQADDHUS (A, B)' `MQADDHUS A,B,C'
28808 `void __MQCPXIS (acc, sw2, sw2)' `__MQCPXIS (C, A, B)' `MQCPXIS A,B,C'
28809 `void __MQCPXIU (acc, uw2, uw2)' `__MQCPXIU (C, A, B)' `MQCPXIU A,B,C'
28810 `void __MQCPXRS (acc, sw2, sw2)' `__MQCPXRS (C, A, B)' `MQCPXRS A,B,C'
28811 `void __MQCPXRU (acc, uw2, uw2)' `__MQCPXRU (C, A, B)' `MQCPXRU A,B,C'
28812 `sw2 __MQLCLRHS (sw2, sw2)' `C = __MQLCLRHS (A, B)' `MQLCLRHS A,B,C'
28813 `sw2 __MQLMTHS (sw2, sw2)' `C = __MQLMTHS (A, B)' `MQLMTHS A,B,C'
28814 `void __MQMACHS (acc, sw2, sw2)' `__MQMACHS (C, A, B)' `MQMACHS A,B,C'
28815 `void __MQMACHU (acc, uw2, uw2)' `__MQMACHU (C, A, B)' `MQMACHU A,B,C'
28816 `void __MQMACXHS (acc, sw2, `__MQMACXHS (C, A, B)' `MQMACXHS A,B,C'
28818 `void __MQMULHS (acc, sw2, sw2)' `__MQMULHS (C, A, B)' `MQMULHS A,B,C'
28819 `void __MQMULHU (acc, uw2, uw2)' `__MQMULHU (C, A, B)' `MQMULHU A,B,C'
28820 `void __MQMULXHS (acc, sw2, `__MQMULXHS (C, A, B)' `MQMULXHS A,B,C'
28822 `void __MQMULXHU (acc, uw2, `__MQMULXHU (C, A, B)' `MQMULXHU A,B,C'
28824 `sw2 __MQSATHS (sw2, sw2)' `C = __MQSATHS (A, B)' `MQSATHS A,B,C'
28825 `uw2 __MQSLLHI (uw2, int)' `C = __MQSLLHI (A, B)' `MQSLLHI A,B,C'
28826 `sw2 __MQSRAHI (sw2, int)' `C = __MQSRAHI (A, B)' `MQSRAHI A,B,C'
28827 `sw2 __MQSUBHSS (sw2, sw2)' `C = __MQSUBHSS (A, B)' `MQSUBHSS A,B,C'
28828 `uw2 __MQSUBHUS (uw2, uw2)' `C = __MQSUBHUS (A, B)' `MQSUBHUS A,B,C'
28829 `void __MQXMACHS (acc, sw2, `__MQXMACHS (C, A, B)' `MQXMACHS A,B,C'
28831 `void __MQXMACXHS (acc, sw2, `__MQXMACXHS (C, A, B)' `MQXMACXHS A,B,C'
28833 `uw1 __MRDACC (acc)' `B = __MRDACC (A)' `MRDACC A,B'
28834 `uw1 __MRDACCG (acc)' `B = __MRDACCG (A)' `MRDACCG A,B'
28835 `uw1 __MROTLI (uw1, const)' `C = __MROTLI (A, B)' `MROTLI A,#B,C'
28836 `uw1 __MROTRI (uw1, const)' `C = __MROTRI (A, B)' `MROTRI A,#B,C'
28837 `sw1 __MSATHS (sw1, sw1)' `C = __MSATHS (A, B)' `MSATHS A,B,C'
28838 `uw1 __MSATHU (uw1, uw1)' `C = __MSATHU (A, B)' `MSATHU A,B,C'
28839 `uw1 __MSLLHI (uw1, const)' `C = __MSLLHI (A, B)' `MSLLHI A,#B,C'
28840 `sw1 __MSRAHI (sw1, const)' `C = __MSRAHI (A, B)' `MSRAHI A,#B,C'
28841 `uw1 __MSRLHI (uw1, const)' `C = __MSRLHI (A, B)' `MSRLHI A,#B,C'
28842 `void __MSUBACCS (acc, acc)' `__MSUBACCS (B, A)' `MSUBACCS A,B'
28843 `sw1 __MSUBHSS (sw1, sw1)' `C = __MSUBHSS (A, B)' `MSUBHSS A,B,C'
28844 `uw1 __MSUBHUS (uw1, uw1)' `C = __MSUBHUS (A, B)' `MSUBHUS A,B,C'
28845 `void __MTRAP (void)' `__MTRAP ()' `MTRAP'
28846 `uw2 __MUNPACKH (uw1)' `B = __MUNPACKH (A)' `MUNPACKH A,B'
28847 `uw1 __MWCUT (uw2, uw1)' `C = __MWCUT (A, B)' `MWCUT A,B,C'
28848 `void __MWTACC (acc, uw1)' `__MWTACC (B, A)' `MWTACC A,B'
28849 `void __MWTACCG (acc, uw1)' `__MWTACCG (B, A)' `MWTACCG A,B'
28850 `uw1 __MXOR (uw1, uw1)' `C = __MXOR (A, B)' `MXOR A,B,C'
28853 File: gcc.info, Node: Raw read/write Functions, Next: Other Built-in Functions, Prev: Directly-mapped Media Functions, Up: FR-V Built-in Functions
28855 5.50.5.4 Raw read/write Functions
28856 .................................
28858 This sections describes built-in functions related to read and write
28859 instructions to access memory. These functions generate `membar'
28860 instructions to flush the I/O load and stores where appropriate, as
28861 described in Fujitsu's manual described above.
28863 `unsigned char __builtin_read8 (void *DATA)'
28865 `unsigned short __builtin_read16 (void *DATA)'
28867 `unsigned long __builtin_read32 (void *DATA)'
28869 `unsigned long long __builtin_read64 (void *DATA)'
28871 `void __builtin_write8 (void *DATA, unsigned char DATUM)'
28873 `void __builtin_write16 (void *DATA, unsigned short DATUM)'
28875 `void __builtin_write32 (void *DATA, unsigned long DATUM)'
28877 `void __builtin_write64 (void *DATA, unsigned long long DATUM)'
28880 File: gcc.info, Node: Other Built-in Functions, Prev: Raw read/write Functions, Up: FR-V Built-in Functions
28882 5.50.5.5 Other Built-in Functions
28883 .................................
28885 This section describes built-in functions that are not named after a
28886 specific FR-V instruction.
28888 `sw2 __IACCreadll (iacc REG)'
28889 Return the full 64-bit value of IACC0. The REG argument is
28890 reserved for future expansion and must be 0.
28892 `sw1 __IACCreadl (iacc REG)'
28893 Return the value of IACC0H if REG is 0 and IACC0L if REG is 1.
28894 Other values of REG are rejected as invalid.
28896 `void __IACCsetll (iacc REG, sw2 X)'
28897 Set the full 64-bit value of IACC0 to X. The REG argument is
28898 reserved for future expansion and must be 0.
28900 `void __IACCsetl (iacc REG, sw1 X)'
28901 Set IACC0H to X if REG is 0 and IACC0L to X if REG is 1. Other
28902 values of REG are rejected as invalid.
28904 `void __data_prefetch0 (const void *X)'
28905 Use the `dcpl' instruction to load the contents of address X into
28908 `void __data_prefetch (const void *X)'
28909 Use the `nldub' instruction to load the contents of address X into
28910 the data cache. The instruction will be issued in slot I1.
28913 File: gcc.info, Node: X86 Built-in Functions, Next: MIPS DSP Built-in Functions, Prev: FR-V Built-in Functions, Up: Target Builtins
28915 5.50.6 X86 Built-in Functions
28916 -----------------------------
28918 These built-in functions are available for the i386 and x86-64 family
28919 of computers, depending on the command-line switches used.
28921 Note that, if you specify command-line switches such as `-msse', the
28922 compiler could use the extended instruction sets even if the built-ins
28923 are not used explicitly in the program. For this reason, applications
28924 which perform runtime CPU detection must compile separate files for each
28925 supported architecture, using the appropriate flags. In particular,
28926 the file containing the CPU detection code should be compiled without
28929 The following machine modes are available for use with MMX built-in
28930 functions (*note Vector Extensions::): `V2SI' for a vector of two
28931 32-bit integers, `V4HI' for a vector of four 16-bit integers, and
28932 `V8QI' for a vector of eight 8-bit integers. Some of the built-in
28933 functions operate on MMX registers as a whole 64-bit entity, these use
28934 `DI' as their mode.
28936 If 3Dnow extensions are enabled, `V2SF' is used as a mode for a vector
28937 of two 32-bit floating point values.
28939 If SSE extensions are enabled, `V4SF' is used for a vector of four
28940 32-bit floating point values. Some instructions use a vector of four
28941 32-bit integers, these use `V4SI'. Finally, some instructions operate
28942 on an entire vector register, interpreting it as a 128-bit integer,
28943 these use mode `TI'.
28945 In 64-bit mode, the x86-64 family of processors uses additional
28946 built-in functions for efficient use of `TF' (`__float128') 128-bit
28947 floating point and `TC' 128-bit complex floating point values.
28949 The following floating point built-in functions are available in 64-bit
28950 mode. All of them implement the function that is part of the name.
28952 __float128 __builtin_fabsq (__float128)
28953 __float128 __builtin_copysignq (__float128, __float128)
28955 The following floating point built-in functions are made available in
28958 `__float128 __builtin_infq (void)'
28959 Similar to `__builtin_inf', except the return type is `__float128'.
28961 The following built-in functions are made available by `-mmmx'. All
28962 of them generate the machine instruction that is part of the name.
28964 v8qi __builtin_ia32_paddb (v8qi, v8qi)
28965 v4hi __builtin_ia32_paddw (v4hi, v4hi)
28966 v2si __builtin_ia32_paddd (v2si, v2si)
28967 v8qi __builtin_ia32_psubb (v8qi, v8qi)
28968 v4hi __builtin_ia32_psubw (v4hi, v4hi)
28969 v2si __builtin_ia32_psubd (v2si, v2si)
28970 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
28971 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
28972 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
28973 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
28974 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
28975 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
28976 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
28977 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
28978 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
28979 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
28980 di __builtin_ia32_pand (di, di)
28981 di __builtin_ia32_pandn (di,di)
28982 di __builtin_ia32_por (di, di)
28983 di __builtin_ia32_pxor (di, di)
28984 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
28985 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
28986 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
28987 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
28988 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
28989 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
28990 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
28991 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
28992 v2si __builtin_ia32_punpckhdq (v2si, v2si)
28993 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
28994 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
28995 v2si __builtin_ia32_punpckldq (v2si, v2si)
28996 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
28997 v4hi __builtin_ia32_packssdw (v2si, v2si)
28998 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
29000 The following built-in functions are made available either with
29001 `-msse', or with a combination of `-m3dnow' and `-march=athlon'. All
29002 of them generate the machine instruction that is part of the name.
29004 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
29005 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
29006 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
29007 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
29008 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
29009 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
29010 v8qi __builtin_ia32_pminub (v8qi, v8qi)
29011 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
29012 int __builtin_ia32_pextrw (v4hi, int)
29013 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
29014 int __builtin_ia32_pmovmskb (v8qi)
29015 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
29016 void __builtin_ia32_movntq (di *, di)
29017 void __builtin_ia32_sfence (void)
29019 The following built-in functions are available when `-msse' is used.
29020 All of them generate the machine instruction that is part of the name.
29022 int __builtin_ia32_comieq (v4sf, v4sf)
29023 int __builtin_ia32_comineq (v4sf, v4sf)
29024 int __builtin_ia32_comilt (v4sf, v4sf)
29025 int __builtin_ia32_comile (v4sf, v4sf)
29026 int __builtin_ia32_comigt (v4sf, v4sf)
29027 int __builtin_ia32_comige (v4sf, v4sf)
29028 int __builtin_ia32_ucomieq (v4sf, v4sf)
29029 int __builtin_ia32_ucomineq (v4sf, v4sf)
29030 int __builtin_ia32_ucomilt (v4sf, v4sf)
29031 int __builtin_ia32_ucomile (v4sf, v4sf)
29032 int __builtin_ia32_ucomigt (v4sf, v4sf)
29033 int __builtin_ia32_ucomige (v4sf, v4sf)
29034 v4sf __builtin_ia32_addps (v4sf, v4sf)
29035 v4sf __builtin_ia32_subps (v4sf, v4sf)
29036 v4sf __builtin_ia32_mulps (v4sf, v4sf)
29037 v4sf __builtin_ia32_divps (v4sf, v4sf)
29038 v4sf __builtin_ia32_addss (v4sf, v4sf)
29039 v4sf __builtin_ia32_subss (v4sf, v4sf)
29040 v4sf __builtin_ia32_mulss (v4sf, v4sf)
29041 v4sf __builtin_ia32_divss (v4sf, v4sf)
29042 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
29043 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
29044 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
29045 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
29046 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
29047 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
29048 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
29049 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
29050 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
29051 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
29052 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
29053 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
29054 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
29055 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
29056 v4si __builtin_ia32_cmpless (v4sf, v4sf)
29057 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
29058 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
29059 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
29060 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
29061 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
29062 v4sf __builtin_ia32_maxps (v4sf, v4sf)
29063 v4sf __builtin_ia32_maxss (v4sf, v4sf)
29064 v4sf __builtin_ia32_minps (v4sf, v4sf)
29065 v4sf __builtin_ia32_minss (v4sf, v4sf)
29066 v4sf __builtin_ia32_andps (v4sf, v4sf)
29067 v4sf __builtin_ia32_andnps (v4sf, v4sf)
29068 v4sf __builtin_ia32_orps (v4sf, v4sf)
29069 v4sf __builtin_ia32_xorps (v4sf, v4sf)
29070 v4sf __builtin_ia32_movss (v4sf, v4sf)
29071 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
29072 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
29073 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
29074 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
29075 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
29076 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
29077 v2si __builtin_ia32_cvtps2pi (v4sf)
29078 int __builtin_ia32_cvtss2si (v4sf)
29079 v2si __builtin_ia32_cvttps2pi (v4sf)
29080 int __builtin_ia32_cvttss2si (v4sf)
29081 v4sf __builtin_ia32_rcpps (v4sf)
29082 v4sf __builtin_ia32_rsqrtps (v4sf)
29083 v4sf __builtin_ia32_sqrtps (v4sf)
29084 v4sf __builtin_ia32_rcpss (v4sf)
29085 v4sf __builtin_ia32_rsqrtss (v4sf)
29086 v4sf __builtin_ia32_sqrtss (v4sf)
29087 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
29088 void __builtin_ia32_movntps (float *, v4sf)
29089 int __builtin_ia32_movmskps (v4sf)
29091 The following built-in functions are available when `-msse' is used.
29093 `v4sf __builtin_ia32_loadaps (float *)'
29094 Generates the `movaps' machine instruction as a load from memory.
29096 `void __builtin_ia32_storeaps (float *, v4sf)'
29097 Generates the `movaps' machine instruction as a store to memory.
29099 `v4sf __builtin_ia32_loadups (float *)'
29100 Generates the `movups' machine instruction as a load from memory.
29102 `void __builtin_ia32_storeups (float *, v4sf)'
29103 Generates the `movups' machine instruction as a store to memory.
29105 `v4sf __builtin_ia32_loadsss (float *)'
29106 Generates the `movss' machine instruction as a load from memory.
29108 `void __builtin_ia32_storess (float *, v4sf)'
29109 Generates the `movss' machine instruction as a store to memory.
29111 `v4sf __builtin_ia32_loadhps (v4sf, v2si *)'
29112 Generates the `movhps' machine instruction as a load from memory.
29114 `v4sf __builtin_ia32_loadlps (v4sf, v2si *)'
29115 Generates the `movlps' machine instruction as a load from memory
29117 `void __builtin_ia32_storehps (v4sf, v2si *)'
29118 Generates the `movhps' machine instruction as a store to memory.
29120 `void __builtin_ia32_storelps (v4sf, v2si *)'
29121 Generates the `movlps' machine instruction as a store to memory.
29123 The following built-in functions are available when `-msse2' is used.
29124 All of them generate the machine instruction that is part of the name.
29126 int __builtin_ia32_comisdeq (v2df, v2df)
29127 int __builtin_ia32_comisdlt (v2df, v2df)
29128 int __builtin_ia32_comisdle (v2df, v2df)
29129 int __builtin_ia32_comisdgt (v2df, v2df)
29130 int __builtin_ia32_comisdge (v2df, v2df)
29131 int __builtin_ia32_comisdneq (v2df, v2df)
29132 int __builtin_ia32_ucomisdeq (v2df, v2df)
29133 int __builtin_ia32_ucomisdlt (v2df, v2df)
29134 int __builtin_ia32_ucomisdle (v2df, v2df)
29135 int __builtin_ia32_ucomisdgt (v2df, v2df)
29136 int __builtin_ia32_ucomisdge (v2df, v2df)
29137 int __builtin_ia32_ucomisdneq (v2df, v2df)
29138 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
29139 v2df __builtin_ia32_cmpltpd (v2df, v2df)
29140 v2df __builtin_ia32_cmplepd (v2df, v2df)
29141 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
29142 v2df __builtin_ia32_cmpgepd (v2df, v2df)
29143 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
29144 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
29145 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
29146 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
29147 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
29148 v2df __builtin_ia32_cmpngepd (v2df, v2df)
29149 v2df __builtin_ia32_cmpordpd (v2df, v2df)
29150 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
29151 v2df __builtin_ia32_cmpltsd (v2df, v2df)
29152 v2df __builtin_ia32_cmplesd (v2df, v2df)
29153 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
29154 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
29155 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
29156 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
29157 v2df __builtin_ia32_cmpordsd (v2df, v2df)
29158 v2di __builtin_ia32_paddq (v2di, v2di)
29159 v2di __builtin_ia32_psubq (v2di, v2di)
29160 v2df __builtin_ia32_addpd (v2df, v2df)
29161 v2df __builtin_ia32_subpd (v2df, v2df)
29162 v2df __builtin_ia32_mulpd (v2df, v2df)
29163 v2df __builtin_ia32_divpd (v2df, v2df)
29164 v2df __builtin_ia32_addsd (v2df, v2df)
29165 v2df __builtin_ia32_subsd (v2df, v2df)
29166 v2df __builtin_ia32_mulsd (v2df, v2df)
29167 v2df __builtin_ia32_divsd (v2df, v2df)
29168 v2df __builtin_ia32_minpd (v2df, v2df)
29169 v2df __builtin_ia32_maxpd (v2df, v2df)
29170 v2df __builtin_ia32_minsd (v2df, v2df)
29171 v2df __builtin_ia32_maxsd (v2df, v2df)
29172 v2df __builtin_ia32_andpd (v2df, v2df)
29173 v2df __builtin_ia32_andnpd (v2df, v2df)
29174 v2df __builtin_ia32_orpd (v2df, v2df)
29175 v2df __builtin_ia32_xorpd (v2df, v2df)
29176 v2df __builtin_ia32_movsd (v2df, v2df)
29177 v2df __builtin_ia32_unpckhpd (v2df, v2df)
29178 v2df __builtin_ia32_unpcklpd (v2df, v2df)
29179 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
29180 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
29181 v4si __builtin_ia32_paddd128 (v4si, v4si)
29182 v2di __builtin_ia32_paddq128 (v2di, v2di)
29183 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
29184 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
29185 v4si __builtin_ia32_psubd128 (v4si, v4si)
29186 v2di __builtin_ia32_psubq128 (v2di, v2di)
29187 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
29188 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
29189 v2di __builtin_ia32_pand128 (v2di, v2di)
29190 v2di __builtin_ia32_pandn128 (v2di, v2di)
29191 v2di __builtin_ia32_por128 (v2di, v2di)
29192 v2di __builtin_ia32_pxor128 (v2di, v2di)
29193 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
29194 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
29195 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
29196 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
29197 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
29198 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
29199 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
29200 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
29201 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
29202 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
29203 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
29204 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
29205 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
29206 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
29207 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
29208 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
29209 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
29210 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
29211 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
29212 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
29213 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
29214 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
29215 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
29216 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
29217 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
29218 v2df __builtin_ia32_loadupd (double *)
29219 void __builtin_ia32_storeupd (double *, v2df)
29220 v2df __builtin_ia32_loadhpd (v2df, double *)
29221 v2df __builtin_ia32_loadlpd (v2df, double *)
29222 int __builtin_ia32_movmskpd (v2df)
29223 int __builtin_ia32_pmovmskb128 (v16qi)
29224 void __builtin_ia32_movnti (int *, int)
29225 void __builtin_ia32_movntpd (double *, v2df)
29226 void __builtin_ia32_movntdq (v2df *, v2df)
29227 v4si __builtin_ia32_pshufd (v4si, int)
29228 v8hi __builtin_ia32_pshuflw (v8hi, int)
29229 v8hi __builtin_ia32_pshufhw (v8hi, int)
29230 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
29231 v2df __builtin_ia32_sqrtpd (v2df)
29232 v2df __builtin_ia32_sqrtsd (v2df)
29233 v2df __builtin_ia32_shufpd (v2df, v2df, int)
29234 v2df __builtin_ia32_cvtdq2pd (v4si)
29235 v4sf __builtin_ia32_cvtdq2ps (v4si)
29236 v4si __builtin_ia32_cvtpd2dq (v2df)
29237 v2si __builtin_ia32_cvtpd2pi (v2df)
29238 v4sf __builtin_ia32_cvtpd2ps (v2df)
29239 v4si __builtin_ia32_cvttpd2dq (v2df)
29240 v2si __builtin_ia32_cvttpd2pi (v2df)
29241 v2df __builtin_ia32_cvtpi2pd (v2si)
29242 int __builtin_ia32_cvtsd2si (v2df)
29243 int __builtin_ia32_cvttsd2si (v2df)
29244 long long __builtin_ia32_cvtsd2si64 (v2df)
29245 long long __builtin_ia32_cvttsd2si64 (v2df)
29246 v4si __builtin_ia32_cvtps2dq (v4sf)
29247 v2df __builtin_ia32_cvtps2pd (v4sf)
29248 v4si __builtin_ia32_cvttps2dq (v4sf)
29249 v2df __builtin_ia32_cvtsi2sd (v2df, int)
29250 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
29251 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
29252 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
29253 void __builtin_ia32_clflush (const void *)
29254 void __builtin_ia32_lfence (void)
29255 void __builtin_ia32_mfence (void)
29256 v16qi __builtin_ia32_loaddqu (const char *)
29257 void __builtin_ia32_storedqu (char *, v16qi)
29258 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
29259 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
29260 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
29261 v4si __builtin_ia32_pslld128 (v4si, v2di)
29262 v2di __builtin_ia32_psllq128 (v4si, v2di)
29263 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
29264 v4si __builtin_ia32_psrld128 (v4si, v2di)
29265 v2di __builtin_ia32_psrlq128 (v2di, v2di)
29266 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
29267 v4si __builtin_ia32_psrad128 (v4si, v2di)
29268 v2di __builtin_ia32_pslldqi128 (v2di, int)
29269 v8hi __builtin_ia32_psllwi128 (v8hi, int)
29270 v4si __builtin_ia32_pslldi128 (v4si, int)
29271 v2di __builtin_ia32_psllqi128 (v2di, int)
29272 v2di __builtin_ia32_psrldqi128 (v2di, int)
29273 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
29274 v4si __builtin_ia32_psrldi128 (v4si, int)
29275 v2di __builtin_ia32_psrlqi128 (v2di, int)
29276 v8hi __builtin_ia32_psrawi128 (v8hi, int)
29277 v4si __builtin_ia32_psradi128 (v4si, int)
29278 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
29280 The following built-in functions are available when `-msse3' is used.
29281 All of them generate the machine instruction that is part of the name.
29283 v2df __builtin_ia32_addsubpd (v2df, v2df)
29284 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
29285 v2df __builtin_ia32_haddpd (v2df, v2df)
29286 v4sf __builtin_ia32_haddps (v4sf, v4sf)
29287 v2df __builtin_ia32_hsubpd (v2df, v2df)
29288 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
29289 v16qi __builtin_ia32_lddqu (char const *)
29290 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
29291 v2df __builtin_ia32_movddup (v2df)
29292 v4sf __builtin_ia32_movshdup (v4sf)
29293 v4sf __builtin_ia32_movsldup (v4sf)
29294 void __builtin_ia32_mwait (unsigned int, unsigned int)
29296 The following built-in functions are available when `-msse3' is used.
29298 `v2df __builtin_ia32_loadddup (double const *)'
29299 Generates the `movddup' machine instruction as a load from memory.
29301 The following built-in functions are available when `-mssse3' is used.
29302 All of them generate the machine instruction that is part of the name
29303 with MMX registers.
29305 v2si __builtin_ia32_phaddd (v2si, v2si)
29306 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
29307 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
29308 v2si __builtin_ia32_phsubd (v2si, v2si)
29309 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
29310 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
29311 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
29312 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
29313 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
29314 v8qi __builtin_ia32_psignb (v8qi, v8qi)
29315 v2si __builtin_ia32_psignd (v2si, v2si)
29316 v4hi __builtin_ia32_psignw (v4hi, v4hi)
29317 long long __builtin_ia32_palignr (long long, long long, int)
29318 v8qi __builtin_ia32_pabsb (v8qi)
29319 v2si __builtin_ia32_pabsd (v2si)
29320 v4hi __builtin_ia32_pabsw (v4hi)
29322 The following built-in functions are available when `-mssse3' is used.
29323 All of them generate the machine instruction that is part of the name
29324 with SSE registers.
29326 v4si __builtin_ia32_phaddd128 (v4si, v4si)
29327 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
29328 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
29329 v4si __builtin_ia32_phsubd128 (v4si, v4si)
29330 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
29331 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
29332 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
29333 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
29334 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
29335 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
29336 v4si __builtin_ia32_psignd128 (v4si, v4si)
29337 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
29338 v2di __builtin_ia32_palignr (v2di, v2di, int)
29339 v16qi __builtin_ia32_pabsb128 (v16qi)
29340 v4si __builtin_ia32_pabsd128 (v4si)
29341 v8hi __builtin_ia32_pabsw128 (v8hi)
29343 The following built-in functions are available when `-msse4.1' is
29344 used. All of them generate the machine instruction that is part of the
29347 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
29348 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
29349 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
29350 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
29351 v2df __builtin_ia32_dppd (v2df, v2df, const int)
29352 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
29353 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
29354 v2di __builtin_ia32_movntdqa (v2di *);
29355 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
29356 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
29357 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
29358 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
29359 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
29360 v8hi __builtin_ia32_phminposuw128 (v8hi)
29361 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
29362 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
29363 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
29364 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
29365 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
29366 v4si __builtin_ia32_pminsd128 (v4si, v4si)
29367 v4si __builtin_ia32_pminud128 (v4si, v4si)
29368 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
29369 v4si __builtin_ia32_pmovsxbd128 (v16qi)
29370 v2di __builtin_ia32_pmovsxbq128 (v16qi)
29371 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
29372 v2di __builtin_ia32_pmovsxdq128 (v4si)
29373 v4si __builtin_ia32_pmovsxwd128 (v8hi)
29374 v2di __builtin_ia32_pmovsxwq128 (v8hi)
29375 v4si __builtin_ia32_pmovzxbd128 (v16qi)
29376 v2di __builtin_ia32_pmovzxbq128 (v16qi)
29377 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
29378 v2di __builtin_ia32_pmovzxdq128 (v4si)
29379 v4si __builtin_ia32_pmovzxwd128 (v8hi)
29380 v2di __builtin_ia32_pmovzxwq128 (v8hi)
29381 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
29382 v4si __builtin_ia32_pmulld128 (v4si, v4si)
29383 int __builtin_ia32_ptestc128 (v2di, v2di)
29384 int __builtin_ia32_ptestnzc128 (v2di, v2di)
29385 int __builtin_ia32_ptestz128 (v2di, v2di)
29386 v2df __builtin_ia32_roundpd (v2df, const int)
29387 v4sf __builtin_ia32_roundps (v4sf, const int)
29388 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
29389 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
29391 The following built-in functions are available when `-msse4.1' is used.
29393 `v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)'
29394 Generates the `insertps' machine instruction.
29396 `int __builtin_ia32_vec_ext_v16qi (v16qi, const int)'
29397 Generates the `pextrb' machine instruction.
29399 `v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)'
29400 Generates the `pinsrb' machine instruction.
29402 `v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)'
29403 Generates the `pinsrd' machine instruction.
29405 `v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)'
29406 Generates the `pinsrq' machine instruction in 64bit mode.
29408 The following built-in functions are changed to generate new SSE4.1
29409 instructions when `-msse4.1' is used.
29411 `float __builtin_ia32_vec_ext_v4sf (v4sf, const int)'
29412 Generates the `extractps' machine instruction.
29414 `int __builtin_ia32_vec_ext_v4si (v4si, const int)'
29415 Generates the `pextrd' machine instruction.
29417 `long long __builtin_ia32_vec_ext_v2di (v2di, const int)'
29418 Generates the `pextrq' machine instruction in 64bit mode.
29420 The following built-in functions are available when `-msse4.2' is
29421 used. All of them generate the machine instruction that is part of the
29424 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
29425 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
29426 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
29427 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
29428 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
29429 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
29430 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
29431 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
29432 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
29433 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
29434 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
29435 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
29436 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
29437 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
29438 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
29440 The following built-in functions are available when `-msse4.2' is used.
29442 `unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)'
29443 Generates the `crc32b' machine instruction.
29445 `unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)'
29446 Generates the `crc32w' machine instruction.
29448 `unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)'
29449 Generates the `crc32l' machine instruction.
29451 `unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)'
29453 The following built-in functions are changed to generate new SSE4.2
29454 instructions when `-msse4.2' is used.
29456 `int __builtin_popcount (unsigned int)'
29457 Generates the `popcntl' machine instruction.
29459 `int __builtin_popcountl (unsigned long)'
29460 Generates the `popcntl' or `popcntq' machine instruction,
29461 depending on the size of `unsigned long'.
29463 `int __builtin_popcountll (unsigned long long)'
29464 Generates the `popcntq' machine instruction.
29466 The following built-in functions are available when `-msse4a' is used.
29467 All of them generate the machine instruction that is part of the name.
29469 void __builtin_ia32_movntsd (double *, v2df)
29470 void __builtin_ia32_movntss (float *, v4sf)
29471 v2di __builtin_ia32_extrq (v2di, v16qi)
29472 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
29473 v2di __builtin_ia32_insertq (v2di, v2di)
29474 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
29476 The following built-in functions are available when `-msse5' is used.
29477 All of them generate the machine instruction that is part of the name
29478 with MMX registers.
29480 v2df __builtin_ia32_comeqpd (v2df, v2df)
29481 v2df __builtin_ia32_comeqps (v2df, v2df)
29482 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
29483 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
29484 v2df __builtin_ia32_comfalsepd (v2df, v2df)
29485 v2df __builtin_ia32_comfalseps (v2df, v2df)
29486 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
29487 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
29488 v2df __builtin_ia32_comgepd (v2df, v2df)
29489 v2df __builtin_ia32_comgeps (v2df, v2df)
29490 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
29491 v4sf __builtin_ia32_comgess (v4sf, v4sf)
29492 v2df __builtin_ia32_comgtpd (v2df, v2df)
29493 v2df __builtin_ia32_comgtps (v2df, v2df)
29494 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
29495 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
29496 v2df __builtin_ia32_comlepd (v2df, v2df)
29497 v2df __builtin_ia32_comleps (v2df, v2df)
29498 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
29499 v4sf __builtin_ia32_comless (v4sf, v4sf)
29500 v2df __builtin_ia32_comltpd (v2df, v2df)
29501 v2df __builtin_ia32_comltps (v2df, v2df)
29502 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
29503 v4sf __builtin_ia32_comltss (v4sf, v4sf)
29504 v2df __builtin_ia32_comnepd (v2df, v2df)
29505 v2df __builtin_ia32_comneps (v2df, v2df)
29506 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
29507 v4sf __builtin_ia32_comness (v4sf, v4sf)
29508 v2df __builtin_ia32_comordpd (v2df, v2df)
29509 v2df __builtin_ia32_comordps (v2df, v2df)
29510 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
29511 v4sf __builtin_ia32_comordss (v4sf, v4sf)
29512 v2df __builtin_ia32_comtruepd (v2df, v2df)
29513 v2df __builtin_ia32_comtrueps (v2df, v2df)
29514 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
29515 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
29516 v2df __builtin_ia32_comueqpd (v2df, v2df)
29517 v2df __builtin_ia32_comueqps (v2df, v2df)
29518 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
29519 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
29520 v2df __builtin_ia32_comugepd (v2df, v2df)
29521 v2df __builtin_ia32_comugeps (v2df, v2df)
29522 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
29523 v4sf __builtin_ia32_comugess (v4sf, v4sf)
29524 v2df __builtin_ia32_comugtpd (v2df, v2df)
29525 v2df __builtin_ia32_comugtps (v2df, v2df)
29526 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
29527 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
29528 v2df __builtin_ia32_comulepd (v2df, v2df)
29529 v2df __builtin_ia32_comuleps (v2df, v2df)
29530 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
29531 v4sf __builtin_ia32_comuless (v4sf, v4sf)
29532 v2df __builtin_ia32_comultpd (v2df, v2df)
29533 v2df __builtin_ia32_comultps (v2df, v2df)
29534 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
29535 v4sf __builtin_ia32_comultss (v4sf, v4sf)
29536 v2df __builtin_ia32_comunepd (v2df, v2df)
29537 v2df __builtin_ia32_comuneps (v2df, v2df)
29538 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
29539 v4sf __builtin_ia32_comuness (v4sf, v4sf)
29540 v2df __builtin_ia32_comunordpd (v2df, v2df)
29541 v2df __builtin_ia32_comunordps (v2df, v2df)
29542 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
29543 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
29544 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
29545 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
29546 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
29547 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
29548 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
29549 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
29550 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
29551 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
29552 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
29553 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
29554 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
29555 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
29556 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
29557 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
29558 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
29559 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
29560 v2df __builtin_ia32_frczpd (v2df)
29561 v4sf __builtin_ia32_frczps (v4sf)
29562 v2df __builtin_ia32_frczsd (v2df, v2df)
29563 v4sf __builtin_ia32_frczss (v4sf, v4sf)
29564 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
29565 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
29566 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
29567 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
29568 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
29569 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
29570 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
29571 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
29572 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
29573 v4si __builtin_ia32_pcomeqd (v4si, v4si)
29574 v2di __builtin_ia32_pcomeqq (v2di, v2di)
29575 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
29576 v4si __builtin_ia32_pcomequd (v4si, v4si)
29577 v2di __builtin_ia32_pcomequq (v2di, v2di)
29578 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
29579 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
29580 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
29581 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
29582 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
29583 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
29584 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
29585 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
29586 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
29587 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
29588 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
29589 v4si __builtin_ia32_pcomged (v4si, v4si)
29590 v2di __builtin_ia32_pcomgeq (v2di, v2di)
29591 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
29592 v4si __builtin_ia32_pcomgeud (v4si, v4si)
29593 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
29594 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
29595 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
29596 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
29597 v4si __builtin_ia32_pcomgtd (v4si, v4si)
29598 v2di __builtin_ia32_pcomgtq (v2di, v2di)
29599 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
29600 v4si __builtin_ia32_pcomgtud (v4si, v4si)
29601 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
29602 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
29603 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
29604 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
29605 v4si __builtin_ia32_pcomled (v4si, v4si)
29606 v2di __builtin_ia32_pcomleq (v2di, v2di)
29607 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
29608 v4si __builtin_ia32_pcomleud (v4si, v4si)
29609 v2di __builtin_ia32_pcomleuq (v2di, v2di)
29610 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
29611 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
29612 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
29613 v4si __builtin_ia32_pcomltd (v4si, v4si)
29614 v2di __builtin_ia32_pcomltq (v2di, v2di)
29615 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
29616 v4si __builtin_ia32_pcomltud (v4si, v4si)
29617 v2di __builtin_ia32_pcomltuq (v2di, v2di)
29618 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
29619 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
29620 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
29621 v4si __builtin_ia32_pcomned (v4si, v4si)
29622 v2di __builtin_ia32_pcomneq (v2di, v2di)
29623 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
29624 v4si __builtin_ia32_pcomneud (v4si, v4si)
29625 v2di __builtin_ia32_pcomneuq (v2di, v2di)
29626 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
29627 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
29628 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
29629 v4si __builtin_ia32_pcomtrued (v4si, v4si)
29630 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
29631 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
29632 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
29633 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
29634 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
29635 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
29636 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
29637 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
29638 v4si __builtin_ia32_phaddbd (v16qi)
29639 v2di __builtin_ia32_phaddbq (v16qi)
29640 v8hi __builtin_ia32_phaddbw (v16qi)
29641 v2di __builtin_ia32_phadddq (v4si)
29642 v4si __builtin_ia32_phaddubd (v16qi)
29643 v2di __builtin_ia32_phaddubq (v16qi)
29644 v8hi __builtin_ia32_phaddubw (v16qi)
29645 v2di __builtin_ia32_phaddudq (v4si)
29646 v4si __builtin_ia32_phadduwd (v8hi)
29647 v2di __builtin_ia32_phadduwq (v8hi)
29648 v4si __builtin_ia32_phaddwd (v8hi)
29649 v2di __builtin_ia32_phaddwq (v8hi)
29650 v8hi __builtin_ia32_phsubbw (v16qi)
29651 v2di __builtin_ia32_phsubdq (v4si)
29652 v4si __builtin_ia32_phsubwd (v8hi)
29653 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
29654 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
29655 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
29656 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
29657 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
29658 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
29659 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
29660 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
29661 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
29662 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
29663 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
29664 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
29665 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
29666 v16qi __builtin_ia32_protb (v16qi, v16qi)
29667 v4si __builtin_ia32_protd (v4si, v4si)
29668 v2di __builtin_ia32_protq (v2di, v2di)
29669 v8hi __builtin_ia32_protw (v8hi, v8hi)
29670 v16qi __builtin_ia32_pshab (v16qi, v16qi)
29671 v4si __builtin_ia32_pshad (v4si, v4si)
29672 v2di __builtin_ia32_pshaq (v2di, v2di)
29673 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
29674 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
29675 v4si __builtin_ia32_pshld (v4si, v4si)
29676 v2di __builtin_ia32_pshlq (v2di, v2di)
29677 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
29679 The following builtin-in functions are available when `-msse5' is
29680 used. The second argument must be an integer constant and generate the
29681 machine instruction that is part of the name with the `_imm' suffix
29684 v16qi __builtin_ia32_protb_imm (v16qi, int)
29685 v4si __builtin_ia32_protd_imm (v4si, int)
29686 v2di __builtin_ia32_protq_imm (v2di, int)
29687 v8hi __builtin_ia32_protw_imm (v8hi, int)
29689 The following built-in functions are available when `-m3dnow' is used.
29690 All of them generate the machine instruction that is part of the name.
29692 void __builtin_ia32_femms (void)
29693 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
29694 v2si __builtin_ia32_pf2id (v2sf)
29695 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
29696 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
29697 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
29698 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
29699 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
29700 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
29701 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
29702 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
29703 v2sf __builtin_ia32_pfrcp (v2sf)
29704 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
29705 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
29706 v2sf __builtin_ia32_pfrsqrt (v2sf)
29707 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
29708 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
29709 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
29710 v2sf __builtin_ia32_pi2fd (v2si)
29711 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
29713 The following built-in functions are available when both `-m3dnow' and
29714 `-march=athlon' are used. All of them generate the machine instruction
29715 that is part of the name.
29717 v2si __builtin_ia32_pf2iw (v2sf)
29718 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
29719 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
29720 v2sf __builtin_ia32_pi2fw (v2si)
29721 v2sf __builtin_ia32_pswapdsf (v2sf)
29722 v2si __builtin_ia32_pswapdsi (v2si)
29725 File: gcc.info, Node: MIPS DSP Built-in Functions, Next: MIPS Paired-Single Support, Prev: X86 Built-in Functions, Up: Target Builtins
29727 5.50.7 MIPS DSP Built-in Functions
29728 ----------------------------------
29730 The MIPS DSP Application-Specific Extension (ASE) includes new
29731 instructions that are designed to improve the performance of DSP and
29732 media applications. It provides instructions that operate on packed
29733 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
29735 GCC supports MIPS DSP operations using both the generic vector
29736 extensions (*note Vector Extensions::) and a collection of
29737 MIPS-specific built-in functions. Both kinds of support are enabled by
29738 the `-mdsp' command-line option.
29740 Revision 2 of the ASE was introduced in the second half of 2006. This
29741 revision adds extra instructions to the original ASE, but is otherwise
29742 backwards-compatible with it. You can select revision 2 using the
29743 command-line option `-mdspr2'; this option implies `-mdsp'.
29745 At present, GCC only provides support for operations on 32-bit
29746 vectors. The vector type associated with 8-bit integer data is usually
29747 called `v4i8', the vector type associated with Q7 is usually called
29748 `v4q7', the vector type associated with 16-bit integer data is usually
29749 called `v2i16', and the vector type associated with Q15 is usually
29750 called `v2q15'. They can be defined in C as follows:
29752 typedef signed char v4i8 __attribute__ ((vector_size(4)));
29753 typedef signed char v4q7 __attribute__ ((vector_size(4)));
29754 typedef short v2i16 __attribute__ ((vector_size(4)));
29755 typedef short v2q15 __attribute__ ((vector_size(4)));
29757 `v4i8', `v4q7', `v2i16' and `v2q15' values are initialized in the same
29758 way as aggregates. For example:
29760 v4i8 a = {1, 2, 3, 4};
29762 b = (v4i8) {5, 6, 7, 8};
29764 v2q15 c = {0x0fcb, 0x3a75};
29766 d = (v2q15) {0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15};
29768 _Note:_ The CPU's endianness determines the order in which values are
29769 packed. On little-endian targets, the first value is the least
29770 significant and the last value is the most significant. The opposite
29771 order applies to big-endian targets. For example, the code above will
29772 set the lowest byte of `a' to `1' on little-endian targets and `4' on
29773 big-endian targets.
29775 _Note:_ Q7, Q15 and Q31 values must be initialized with their integer
29776 representation. As shown in this example, the integer representation
29777 of a Q7 value can be obtained by multiplying the fractional value by
29778 `0x1.0p7'. The equivalent for Q15 values is to multiply by `0x1.0p15'.
29779 The equivalent for Q31 values is to multiply by `0x1.0p31'.
29781 The table below lists the `v4i8' and `v2q15' operations for which
29782 hardware support exists. `a' and `b' are `v4i8' values, and `c' and
29783 `d' are `v2q15' values.
29785 C code MIPS instruction
29791 The table below lists the `v2i16' operation for which hardware support
29792 exists for the DSP ASE REV 2. `e' and `f' are `v2i16' values.
29794 C code MIPS instruction
29797 It is easier to describe the DSP built-in functions if we first define
29798 the following types:
29802 typedef unsigned int ui32;
29803 typedef long long a64;
29805 `q31' and `i32' are actually the same as `int', but we use `q31' to
29806 indicate a Q31 fractional value and `i32' to indicate a 32-bit integer
29807 value. Similarly, `a64' is the same as `long long', but we use `a64'
29808 to indicate values that will be placed in one of the four DSP
29809 accumulators (`$ac0', `$ac1', `$ac2' or `$ac3').
29811 Also, some built-in functions prefer or require immediate numbers as
29812 parameters, because the corresponding DSP instructions accept both
29813 immediate numbers and register operands, or accept immediate numbers
29814 only. The immediate parameters are listed as follows.
29821 imm0_255: 0 to 255.
29822 imm_n32_31: -32 to 31.
29823 imm_n512_511: -512 to 511.
29825 The following built-in functions map directly to a particular MIPS DSP
29826 instruction. Please refer to the architecture specification for
29827 details on what each instruction does.
29829 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
29830 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
29831 q31 __builtin_mips_addq_s_w (q31, q31)
29832 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
29833 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
29834 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
29835 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
29836 q31 __builtin_mips_subq_s_w (q31, q31)
29837 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
29838 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
29839 i32 __builtin_mips_addsc (i32, i32)
29840 i32 __builtin_mips_addwc (i32, i32)
29841 i32 __builtin_mips_modsub (i32, i32)
29842 i32 __builtin_mips_raddu_w_qb (v4i8)
29843 v2q15 __builtin_mips_absq_s_ph (v2q15)
29844 q31 __builtin_mips_absq_s_w (q31)
29845 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
29846 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
29847 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
29848 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
29849 q31 __builtin_mips_preceq_w_phl (v2q15)
29850 q31 __builtin_mips_preceq_w_phr (v2q15)
29851 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
29852 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
29853 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
29854 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
29855 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
29856 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
29857 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
29858 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
29859 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
29860 v4i8 __builtin_mips_shll_qb (v4i8, i32)
29861 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
29862 v2q15 __builtin_mips_shll_ph (v2q15, i32)
29863 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
29864 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
29865 q31 __builtin_mips_shll_s_w (q31, imm0_31)
29866 q31 __builtin_mips_shll_s_w (q31, i32)
29867 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
29868 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
29869 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
29870 v2q15 __builtin_mips_shra_ph (v2q15, i32)
29871 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
29872 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
29873 q31 __builtin_mips_shra_r_w (q31, imm0_31)
29874 q31 __builtin_mips_shra_r_w (q31, i32)
29875 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
29876 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
29877 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
29878 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
29879 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
29880 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
29881 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
29882 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
29883 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
29884 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
29885 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
29886 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
29887 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
29888 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
29889 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
29890 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
29891 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
29892 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
29893 i32 __builtin_mips_bitrev (i32)
29894 i32 __builtin_mips_insv (i32, i32)
29895 v4i8 __builtin_mips_repl_qb (imm0_255)
29896 v4i8 __builtin_mips_repl_qb (i32)
29897 v2q15 __builtin_mips_repl_ph (imm_n512_511)
29898 v2q15 __builtin_mips_repl_ph (i32)
29899 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
29900 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
29901 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
29902 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
29903 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
29904 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
29905 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
29906 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
29907 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
29908 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
29909 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
29910 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
29911 i32 __builtin_mips_extr_w (a64, imm0_31)
29912 i32 __builtin_mips_extr_w (a64, i32)
29913 i32 __builtin_mips_extr_r_w (a64, imm0_31)
29914 i32 __builtin_mips_extr_s_h (a64, i32)
29915 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
29916 i32 __builtin_mips_extr_rs_w (a64, i32)
29917 i32 __builtin_mips_extr_s_h (a64, imm0_31)
29918 i32 __builtin_mips_extr_r_w (a64, i32)
29919 i32 __builtin_mips_extp (a64, imm0_31)
29920 i32 __builtin_mips_extp (a64, i32)
29921 i32 __builtin_mips_extpdp (a64, imm0_31)
29922 i32 __builtin_mips_extpdp (a64, i32)
29923 a64 __builtin_mips_shilo (a64, imm_n32_31)
29924 a64 __builtin_mips_shilo (a64, i32)
29925 a64 __builtin_mips_mthlip (a64, i32)
29926 void __builtin_mips_wrdsp (i32, imm0_63)
29927 i32 __builtin_mips_rddsp (imm0_63)
29928 i32 __builtin_mips_lbux (void *, i32)
29929 i32 __builtin_mips_lhx (void *, i32)
29930 i32 __builtin_mips_lwx (void *, i32)
29931 i32 __builtin_mips_bposge32 (void)
29933 The following built-in functions map directly to a particular MIPS DSP
29934 REV 2 instruction. Please refer to the architecture specification for
29935 details on what each instruction does.
29937 v4q7 __builtin_mips_absq_s_qb (v4q7);
29938 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
29939 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
29940 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
29941 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
29942 i32 __builtin_mips_append (i32, i32, imm0_31);
29943 i32 __builtin_mips_balign (i32, i32, imm0_3);
29944 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
29945 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
29946 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
29947 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
29948 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
29949 a64 __builtin_mips_madd (a64, i32, i32);
29950 a64 __builtin_mips_maddu (a64, ui32, ui32);
29951 a64 __builtin_mips_msub (a64, i32, i32);
29952 a64 __builtin_mips_msubu (a64, ui32, ui32);
29953 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
29954 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
29955 q31 __builtin_mips_mulq_rs_w (q31, q31);
29956 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
29957 q31 __builtin_mips_mulq_s_w (q31, q31);
29958 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
29959 a64 __builtin_mips_mult (i32, i32);
29960 a64 __builtin_mips_multu (ui32, ui32);
29961 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
29962 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
29963 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
29964 i32 __builtin_mips_prepend (i32, i32, imm0_31);
29965 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
29966 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
29967 v4i8 __builtin_mips_shra_qb (v4i8, i32);
29968 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
29969 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
29970 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
29971 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
29972 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
29973 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
29974 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
29975 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
29976 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
29977 q31 __builtin_mips_addqh_w (q31, q31);
29978 q31 __builtin_mips_addqh_r_w (q31, q31);
29979 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
29980 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
29981 q31 __builtin_mips_subqh_w (q31, q31);
29982 q31 __builtin_mips_subqh_r_w (q31, q31);
29983 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
29984 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
29985 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
29986 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
29987 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
29988 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
29991 File: gcc.info, Node: MIPS Paired-Single Support, Next: PowerPC AltiVec Built-in Functions, Prev: MIPS DSP Built-in Functions, Up: Target Builtins
29993 5.50.8 MIPS Paired-Single Support
29994 ---------------------------------
29996 The MIPS64 architecture includes a number of instructions that operate
29997 on pairs of single-precision floating-point values. Each pair is
29998 packed into a 64-bit floating-point register, with one element being
29999 designated the "upper half" and the other being designated the "lower
30002 GCC supports paired-single operations using both the generic vector
30003 extensions (*note Vector Extensions::) and a collection of
30004 MIPS-specific built-in functions. Both kinds of support are enabled by
30005 the `-mpaired-single' command-line option.
30007 The vector type associated with paired-single values is usually called
30008 `v2sf'. It can be defined in C as follows:
30010 typedef float v2sf __attribute__ ((vector_size (8)));
30012 `v2sf' values are initialized in the same way as aggregates. For
30015 v2sf a = {1.5, 9.1};
30020 _Note:_ The CPU's endianness determines which value is stored in the
30021 upper half of a register and which value is stored in the lower half.
30022 On little-endian targets, the first value is the lower one and the
30023 second value is the upper one. The opposite order applies to
30024 big-endian targets. For example, the code above will set the lower
30025 half of `a' to `1.5' on little-endian targets and `9.1' on big-endian
30030 * Paired-Single Arithmetic::
30031 * Paired-Single Built-in Functions::
30032 * MIPS-3D Built-in Functions::
30035 File: gcc.info, Node: Paired-Single Arithmetic, Next: Paired-Single Built-in Functions, Up: MIPS Paired-Single Support
30037 5.50.8.1 Paired-Single Arithmetic
30038 .................................
30040 The table below lists the `v2sf' operations for which hardware support
30041 exists. `a', `b' and `c' are `v2sf' values and `x' is an integral
30044 C code MIPS instruction
30049 `a * b + c' `madd.ps'
30050 `a * b - c' `msub.ps'
30051 `-(a * b + c)' `nmadd.ps'
30052 `-(a * b - c)' `nmsub.ps'
30053 `x ? a : b' `movn.ps'/`movz.ps'
30055 Note that the multiply-accumulate instructions can be disabled using
30056 the command-line option `-mno-fused-madd'.
30059 File: gcc.info, Node: Paired-Single Built-in Functions, Next: MIPS-3D Built-in Functions, Prev: Paired-Single Arithmetic, Up: MIPS Paired-Single Support
30061 5.50.8.2 Paired-Single Built-in Functions
30062 .........................................
30064 The following paired-single functions map directly to a particular MIPS
30065 instruction. Please refer to the architecture specification for
30066 details on what each instruction does.
30068 `v2sf __builtin_mips_pll_ps (v2sf, v2sf)'
30069 Pair lower lower (`pll.ps').
30071 `v2sf __builtin_mips_pul_ps (v2sf, v2sf)'
30072 Pair upper lower (`pul.ps').
30074 `v2sf __builtin_mips_plu_ps (v2sf, v2sf)'
30075 Pair lower upper (`plu.ps').
30077 `v2sf __builtin_mips_puu_ps (v2sf, v2sf)'
30078 Pair upper upper (`puu.ps').
30080 `v2sf __builtin_mips_cvt_ps_s (float, float)'
30081 Convert pair to paired single (`cvt.ps.s').
30083 `float __builtin_mips_cvt_s_pl (v2sf)'
30084 Convert pair lower to single (`cvt.s.pl').
30086 `float __builtin_mips_cvt_s_pu (v2sf)'
30087 Convert pair upper to single (`cvt.s.pu').
30089 `v2sf __builtin_mips_abs_ps (v2sf)'
30090 Absolute value (`abs.ps').
30092 `v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)'
30093 Align variable (`alnv.ps').
30095 _Note:_ The value of the third parameter must be 0 or 4 modulo 8,
30096 otherwise the result will be unpredictable. Please read the
30097 instruction description for details.
30099 The following multi-instruction functions are also available. In each
30100 case, COND can be any of the 16 floating-point conditions: `f', `un',
30101 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
30102 `lt', `nge', `le' or `ngt'.
30104 `v2sf __builtin_mips_movt_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
30105 `v2sf __builtin_mips_movf_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
30106 Conditional move based on floating point comparison (`c.COND.ps',
30107 `movt.ps'/`movf.ps').
30109 The `movt' functions return the value X computed by:
30115 The `movf' functions are similar but use `movf.ps' instead of
30118 `int __builtin_mips_upper_c_COND_ps (v2sf A, v2sf B)'
30119 `int __builtin_mips_lower_c_COND_ps (v2sf A, v2sf B)'
30120 Comparison of two paired-single values (`c.COND.ps',
30123 These functions compare A and B using `c.COND.ps' and return
30124 either the upper or lower half of the result. For example:
30127 if (__builtin_mips_upper_c_eq_ps (a, b))
30128 upper_halves_are_equal ();
30130 upper_halves_are_unequal ();
30132 if (__builtin_mips_lower_c_eq_ps (a, b))
30133 lower_halves_are_equal ();
30135 lower_halves_are_unequal ();
30138 File: gcc.info, Node: MIPS-3D Built-in Functions, Prev: Paired-Single Built-in Functions, Up: MIPS Paired-Single Support
30140 5.50.8.3 MIPS-3D Built-in Functions
30141 ...................................
30143 The MIPS-3D Application-Specific Extension (ASE) includes additional
30144 paired-single instructions that are designed to improve the performance
30145 of 3D graphics operations. Support for these instructions is controlled
30146 by the `-mips3d' command-line option.
30148 The functions listed below map directly to a particular MIPS-3D
30149 instruction. Please refer to the architecture specification for more
30150 details on what each instruction does.
30152 `v2sf __builtin_mips_addr_ps (v2sf, v2sf)'
30153 Reduction add (`addr.ps').
30155 `v2sf __builtin_mips_mulr_ps (v2sf, v2sf)'
30156 Reduction multiply (`mulr.ps').
30158 `v2sf __builtin_mips_cvt_pw_ps (v2sf)'
30159 Convert paired single to paired word (`cvt.pw.ps').
30161 `v2sf __builtin_mips_cvt_ps_pw (v2sf)'
30162 Convert paired word to paired single (`cvt.ps.pw').
30164 `float __builtin_mips_recip1_s (float)'
30165 `double __builtin_mips_recip1_d (double)'
30166 `v2sf __builtin_mips_recip1_ps (v2sf)'
30167 Reduced precision reciprocal (sequence step 1) (`recip1.FMT').
30169 `float __builtin_mips_recip2_s (float, float)'
30170 `double __builtin_mips_recip2_d (double, double)'
30171 `v2sf __builtin_mips_recip2_ps (v2sf, v2sf)'
30172 Reduced precision reciprocal (sequence step 2) (`recip2.FMT').
30174 `float __builtin_mips_rsqrt1_s (float)'
30175 `double __builtin_mips_rsqrt1_d (double)'
30176 `v2sf __builtin_mips_rsqrt1_ps (v2sf)'
30177 Reduced precision reciprocal square root (sequence step 1)
30180 `float __builtin_mips_rsqrt2_s (float, float)'
30181 `double __builtin_mips_rsqrt2_d (double, double)'
30182 `v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)'
30183 Reduced precision reciprocal square root (sequence step 2)
30186 The following multi-instruction functions are also available. In each
30187 case, COND can be any of the 16 floating-point conditions: `f', `un',
30188 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
30189 `lt', `nge', `le' or `ngt'.
30191 `int __builtin_mips_cabs_COND_s (float A, float B)'
30192 `int __builtin_mips_cabs_COND_d (double A, double B)'
30193 Absolute comparison of two scalar values (`cabs.COND.FMT',
30196 These functions compare A and B using `cabs.COND.s' or
30197 `cabs.COND.d' and return the result as a boolean value. For
30201 if (__builtin_mips_cabs_eq_s (a, b))
30206 `int __builtin_mips_upper_cabs_COND_ps (v2sf A, v2sf B)'
30207 `int __builtin_mips_lower_cabs_COND_ps (v2sf A, v2sf B)'
30208 Absolute comparison of two paired-single values (`cabs.COND.ps',
30211 These functions compare A and B using `cabs.COND.ps' and return
30212 either the upper or lower half of the result. For example:
30215 if (__builtin_mips_upper_cabs_eq_ps (a, b))
30216 upper_halves_are_equal ();
30218 upper_halves_are_unequal ();
30220 if (__builtin_mips_lower_cabs_eq_ps (a, b))
30221 lower_halves_are_equal ();
30223 lower_halves_are_unequal ();
30225 `v2sf __builtin_mips_movt_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
30226 `v2sf __builtin_mips_movf_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
30227 Conditional move based on absolute comparison (`cabs.COND.ps',
30228 `movt.ps'/`movf.ps').
30230 The `movt' functions return the value X computed by:
30232 cabs.COND.ps CC,A,B
30236 The `movf' functions are similar but use `movf.ps' instead of
30239 `int __builtin_mips_any_c_COND_ps (v2sf A, v2sf B)'
30240 `int __builtin_mips_all_c_COND_ps (v2sf A, v2sf B)'
30241 `int __builtin_mips_any_cabs_COND_ps (v2sf A, v2sf B)'
30242 `int __builtin_mips_all_cabs_COND_ps (v2sf A, v2sf B)'
30243 Comparison of two paired-single values (`c.COND.ps'/`cabs.COND.ps',
30244 `bc1any2t'/`bc1any2f').
30246 These functions compare A and B using `c.COND.ps' or
30247 `cabs.COND.ps'. The `any' forms return true if either result is
30248 true and the `all' forms return true if both results are true.
30252 if (__builtin_mips_any_c_eq_ps (a, b))
30257 if (__builtin_mips_all_c_eq_ps (a, b))
30262 `int __builtin_mips_any_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
30263 `int __builtin_mips_all_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
30264 `int __builtin_mips_any_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
30265 `int __builtin_mips_all_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
30266 Comparison of four paired-single values
30267 (`c.COND.ps'/`cabs.COND.ps', `bc1any4t'/`bc1any4f').
30269 These functions use `c.COND.ps' or `cabs.COND.ps' to compare A
30270 with B and to compare C with D. The `any' forms return true if
30271 any of the four results are true and the `all' forms return true
30272 if all four results are true. For example:
30275 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
30280 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
30286 File: gcc.info, Node: PowerPC AltiVec Built-in Functions, Next: SPARC VIS Built-in Functions, Prev: MIPS Paired-Single Support, Up: Target Builtins
30288 5.50.9 PowerPC AltiVec Built-in Functions
30289 -----------------------------------------
30291 GCC provides an interface for the PowerPC family of processors to access
30292 the AltiVec operations described in Motorola's AltiVec Programming
30293 Interface Manual. The interface is made available by including
30294 `<altivec.h>' and using `-maltivec' and `-mabi=altivec'. The interface
30295 supports the following vector types.
30297 vector unsigned char
30301 vector unsigned short
30302 vector signed short
30306 vector unsigned int
30311 GCC's implementation of the high-level language interface available
30312 from C and C++ code differs from Motorola's documentation in several
30315 * A vector constant is a list of constant expressions within curly
30318 * A vector initializer requires no cast if the vector constant is of
30319 the same type as the variable it is initializing.
30321 * If `signed' or `unsigned' is omitted, the signedness of the vector
30322 type is the default signedness of the base type. The default
30323 varies depending on the operating system, so a portable program
30324 should always specify the signedness.
30326 * Compiling with `-maltivec' adds keywords `__vector', `__pixel',
30327 and `__bool'. Macros `vector', `pixel', and `bool' are defined in
30328 `<altivec.h>' and can be undefined.
30330 * GCC allows using a `typedef' name as the type specifier for a
30333 * For C, overloaded functions are implemented with macros so the
30334 following does not work:
30336 vec_add ((vector signed int){1, 2, 3, 4}, foo);
30338 Since `vec_add' is a macro, the vector constant in the example is
30339 treated as four separate arguments. Wrap the entire argument in
30340 parentheses for this to work.
30342 _Note:_ Only the `<altivec.h>' interface is supported. Internally,
30343 GCC uses built-in functions to achieve the functionality in the
30344 aforementioned header file, but they are not supported and are subject
30345 to change without notice.
30347 The following interfaces are supported for the generic and specific
30348 AltiVec operations and the AltiVec predicates. In cases where there is
30349 a direct mapping between generic and specific operations, only the
30350 generic names are shown here, although the specific operations can also
30353 Arguments that are documented as `const int' require literal integral
30354 values within the range required for that operation.
30356 vector signed char vec_abs (vector signed char);
30357 vector signed short vec_abs (vector signed short);
30358 vector signed int vec_abs (vector signed int);
30359 vector float vec_abs (vector float);
30361 vector signed char vec_abss (vector signed char);
30362 vector signed short vec_abss (vector signed short);
30363 vector signed int vec_abss (vector signed int);
30365 vector signed char vec_add (vector bool char, vector signed char);
30366 vector signed char vec_add (vector signed char, vector bool char);
30367 vector signed char vec_add (vector signed char, vector signed char);
30368 vector unsigned char vec_add (vector bool char, vector unsigned char);
30369 vector unsigned char vec_add (vector unsigned char, vector bool char);
30370 vector unsigned char vec_add (vector unsigned char,
30371 vector unsigned char);
30372 vector signed short vec_add (vector bool short, vector signed short);
30373 vector signed short vec_add (vector signed short, vector bool short);
30374 vector signed short vec_add (vector signed short, vector signed short);
30375 vector unsigned short vec_add (vector bool short,
30376 vector unsigned short);
30377 vector unsigned short vec_add (vector unsigned short,
30378 vector bool short);
30379 vector unsigned short vec_add (vector unsigned short,
30380 vector unsigned short);
30381 vector signed int vec_add (vector bool int, vector signed int);
30382 vector signed int vec_add (vector signed int, vector bool int);
30383 vector signed int vec_add (vector signed int, vector signed int);
30384 vector unsigned int vec_add (vector bool int, vector unsigned int);
30385 vector unsigned int vec_add (vector unsigned int, vector bool int);
30386 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
30387 vector float vec_add (vector float, vector float);
30389 vector float vec_vaddfp (vector float, vector float);
30391 vector signed int vec_vadduwm (vector bool int, vector signed int);
30392 vector signed int vec_vadduwm (vector signed int, vector bool int);
30393 vector signed int vec_vadduwm (vector signed int, vector signed int);
30394 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
30395 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
30396 vector unsigned int vec_vadduwm (vector unsigned int,
30397 vector unsigned int);
30399 vector signed short vec_vadduhm (vector bool short,
30400 vector signed short);
30401 vector signed short vec_vadduhm (vector signed short,
30402 vector bool short);
30403 vector signed short vec_vadduhm (vector signed short,
30404 vector signed short);
30405 vector unsigned short vec_vadduhm (vector bool short,
30406 vector unsigned short);
30407 vector unsigned short vec_vadduhm (vector unsigned short,
30408 vector bool short);
30409 vector unsigned short vec_vadduhm (vector unsigned short,
30410 vector unsigned short);
30412 vector signed char vec_vaddubm (vector bool char, vector signed char);
30413 vector signed char vec_vaddubm (vector signed char, vector bool char);
30414 vector signed char vec_vaddubm (vector signed char, vector signed char);
30415 vector unsigned char vec_vaddubm (vector bool char,
30416 vector unsigned char);
30417 vector unsigned char vec_vaddubm (vector unsigned char,
30419 vector unsigned char vec_vaddubm (vector unsigned char,
30420 vector unsigned char);
30422 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
30424 vector unsigned char vec_adds (vector bool char, vector unsigned char);
30425 vector unsigned char vec_adds (vector unsigned char, vector bool char);
30426 vector unsigned char vec_adds (vector unsigned char,
30427 vector unsigned char);
30428 vector signed char vec_adds (vector bool char, vector signed char);
30429 vector signed char vec_adds (vector signed char, vector bool char);
30430 vector signed char vec_adds (vector signed char, vector signed char);
30431 vector unsigned short vec_adds (vector bool short,
30432 vector unsigned short);
30433 vector unsigned short vec_adds (vector unsigned short,
30434 vector bool short);
30435 vector unsigned short vec_adds (vector unsigned short,
30436 vector unsigned short);
30437 vector signed short vec_adds (vector bool short, vector signed short);
30438 vector signed short vec_adds (vector signed short, vector bool short);
30439 vector signed short vec_adds (vector signed short, vector signed short);
30440 vector unsigned int vec_adds (vector bool int, vector unsigned int);
30441 vector unsigned int vec_adds (vector unsigned int, vector bool int);
30442 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
30443 vector signed int vec_adds (vector bool int, vector signed int);
30444 vector signed int vec_adds (vector signed int, vector bool int);
30445 vector signed int vec_adds (vector signed int, vector signed int);
30447 vector signed int vec_vaddsws (vector bool int, vector signed int);
30448 vector signed int vec_vaddsws (vector signed int, vector bool int);
30449 vector signed int vec_vaddsws (vector signed int, vector signed int);
30451 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
30452 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
30453 vector unsigned int vec_vadduws (vector unsigned int,
30454 vector unsigned int);
30456 vector signed short vec_vaddshs (vector bool short,
30457 vector signed short);
30458 vector signed short vec_vaddshs (vector signed short,
30459 vector bool short);
30460 vector signed short vec_vaddshs (vector signed short,
30461 vector signed short);
30463 vector unsigned short vec_vadduhs (vector bool short,
30464 vector unsigned short);
30465 vector unsigned short vec_vadduhs (vector unsigned short,
30466 vector bool short);
30467 vector unsigned short vec_vadduhs (vector unsigned short,
30468 vector unsigned short);
30470 vector signed char vec_vaddsbs (vector bool char, vector signed char);
30471 vector signed char vec_vaddsbs (vector signed char, vector bool char);
30472 vector signed char vec_vaddsbs (vector signed char, vector signed char);
30474 vector unsigned char vec_vaddubs (vector bool char,
30475 vector unsigned char);
30476 vector unsigned char vec_vaddubs (vector unsigned char,
30478 vector unsigned char vec_vaddubs (vector unsigned char,
30479 vector unsigned char);
30481 vector float vec_and (vector float, vector float);
30482 vector float vec_and (vector float, vector bool int);
30483 vector float vec_and (vector bool int, vector float);
30484 vector bool int vec_and (vector bool int, vector bool int);
30485 vector signed int vec_and (vector bool int, vector signed int);
30486 vector signed int vec_and (vector signed int, vector bool int);
30487 vector signed int vec_and (vector signed int, vector signed int);
30488 vector unsigned int vec_and (vector bool int, vector unsigned int);
30489 vector unsigned int vec_and (vector unsigned int, vector bool int);
30490 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
30491 vector bool short vec_and (vector bool short, vector bool short);
30492 vector signed short vec_and (vector bool short, vector signed short);
30493 vector signed short vec_and (vector signed short, vector bool short);
30494 vector signed short vec_and (vector signed short, vector signed short);
30495 vector unsigned short vec_and (vector bool short,
30496 vector unsigned short);
30497 vector unsigned short vec_and (vector unsigned short,
30498 vector bool short);
30499 vector unsigned short vec_and (vector unsigned short,
30500 vector unsigned short);
30501 vector signed char vec_and (vector bool char, vector signed char);
30502 vector bool char vec_and (vector bool char, vector bool char);
30503 vector signed char vec_and (vector signed char, vector bool char);
30504 vector signed char vec_and (vector signed char, vector signed char);
30505 vector unsigned char vec_and (vector bool char, vector unsigned char);
30506 vector unsigned char vec_and (vector unsigned char, vector bool char);
30507 vector unsigned char vec_and (vector unsigned char,
30508 vector unsigned char);
30510 vector float vec_andc (vector float, vector float);
30511 vector float vec_andc (vector float, vector bool int);
30512 vector float vec_andc (vector bool int, vector float);
30513 vector bool int vec_andc (vector bool int, vector bool int);
30514 vector signed int vec_andc (vector bool int, vector signed int);
30515 vector signed int vec_andc (vector signed int, vector bool int);
30516 vector signed int vec_andc (vector signed int, vector signed int);
30517 vector unsigned int vec_andc (vector bool int, vector unsigned int);
30518 vector unsigned int vec_andc (vector unsigned int, vector bool int);
30519 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
30520 vector bool short vec_andc (vector bool short, vector bool short);
30521 vector signed short vec_andc (vector bool short, vector signed short);
30522 vector signed short vec_andc (vector signed short, vector bool short);
30523 vector signed short vec_andc (vector signed short, vector signed short);
30524 vector unsigned short vec_andc (vector bool short,
30525 vector unsigned short);
30526 vector unsigned short vec_andc (vector unsigned short,
30527 vector bool short);
30528 vector unsigned short vec_andc (vector unsigned short,
30529 vector unsigned short);
30530 vector signed char vec_andc (vector bool char, vector signed char);
30531 vector bool char vec_andc (vector bool char, vector bool char);
30532 vector signed char vec_andc (vector signed char, vector bool char);
30533 vector signed char vec_andc (vector signed char, vector signed char);
30534 vector unsigned char vec_andc (vector bool char, vector unsigned char);
30535 vector unsigned char vec_andc (vector unsigned char, vector bool char);
30536 vector unsigned char vec_andc (vector unsigned char,
30537 vector unsigned char);
30539 vector unsigned char vec_avg (vector unsigned char,
30540 vector unsigned char);
30541 vector signed char vec_avg (vector signed char, vector signed char);
30542 vector unsigned short vec_avg (vector unsigned short,
30543 vector unsigned short);
30544 vector signed short vec_avg (vector signed short, vector signed short);
30545 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
30546 vector signed int vec_avg (vector signed int, vector signed int);
30548 vector signed int vec_vavgsw (vector signed int, vector signed int);
30550 vector unsigned int vec_vavguw (vector unsigned int,
30551 vector unsigned int);
30553 vector signed short vec_vavgsh (vector signed short,
30554 vector signed short);
30556 vector unsigned short vec_vavguh (vector unsigned short,
30557 vector unsigned short);
30559 vector signed char vec_vavgsb (vector signed char, vector signed char);
30561 vector unsigned char vec_vavgub (vector unsigned char,
30562 vector unsigned char);
30564 vector float vec_ceil (vector float);
30566 vector signed int vec_cmpb (vector float, vector float);
30568 vector bool char vec_cmpeq (vector signed char, vector signed char);
30569 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
30570 vector bool short vec_cmpeq (vector signed short, vector signed short);
30571 vector bool short vec_cmpeq (vector unsigned short,
30572 vector unsigned short);
30573 vector bool int vec_cmpeq (vector signed int, vector signed int);
30574 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
30575 vector bool int vec_cmpeq (vector float, vector float);
30577 vector bool int vec_vcmpeqfp (vector float, vector float);
30579 vector bool int vec_vcmpequw (vector signed int, vector signed int);
30580 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
30582 vector bool short vec_vcmpequh (vector signed short,
30583 vector signed short);
30584 vector bool short vec_vcmpequh (vector unsigned short,
30585 vector unsigned short);
30587 vector bool char vec_vcmpequb (vector signed char, vector signed char);
30588 vector bool char vec_vcmpequb (vector unsigned char,
30589 vector unsigned char);
30591 vector bool int vec_cmpge (vector float, vector float);
30593 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
30594 vector bool char vec_cmpgt (vector signed char, vector signed char);
30595 vector bool short vec_cmpgt (vector unsigned short,
30596 vector unsigned short);
30597 vector bool short vec_cmpgt (vector signed short, vector signed short);
30598 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
30599 vector bool int vec_cmpgt (vector signed int, vector signed int);
30600 vector bool int vec_cmpgt (vector float, vector float);
30602 vector bool int vec_vcmpgtfp (vector float, vector float);
30604 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
30606 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
30608 vector bool short vec_vcmpgtsh (vector signed short,
30609 vector signed short);
30611 vector bool short vec_vcmpgtuh (vector unsigned short,
30612 vector unsigned short);
30614 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
30616 vector bool char vec_vcmpgtub (vector unsigned char,
30617 vector unsigned char);
30619 vector bool int vec_cmple (vector float, vector float);
30621 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
30622 vector bool char vec_cmplt (vector signed char, vector signed char);
30623 vector bool short vec_cmplt (vector unsigned short,
30624 vector unsigned short);
30625 vector bool short vec_cmplt (vector signed short, vector signed short);
30626 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
30627 vector bool int vec_cmplt (vector signed int, vector signed int);
30628 vector bool int vec_cmplt (vector float, vector float);
30630 vector float vec_ctf (vector unsigned int, const int);
30631 vector float vec_ctf (vector signed int, const int);
30633 vector float vec_vcfsx (vector signed int, const int);
30635 vector float vec_vcfux (vector unsigned int, const int);
30637 vector signed int vec_cts (vector float, const int);
30639 vector unsigned int vec_ctu (vector float, const int);
30641 void vec_dss (const int);
30643 void vec_dssall (void);
30645 void vec_dst (const vector unsigned char *, int, const int);
30646 void vec_dst (const vector signed char *, int, const int);
30647 void vec_dst (const vector bool char *, int, const int);
30648 void vec_dst (const vector unsigned short *, int, const int);
30649 void vec_dst (const vector signed short *, int, const int);
30650 void vec_dst (const vector bool short *, int, const int);
30651 void vec_dst (const vector pixel *, int, const int);
30652 void vec_dst (const vector unsigned int *, int, const int);
30653 void vec_dst (const vector signed int *, int, const int);
30654 void vec_dst (const vector bool int *, int, const int);
30655 void vec_dst (const vector float *, int, const int);
30656 void vec_dst (const unsigned char *, int, const int);
30657 void vec_dst (const signed char *, int, const int);
30658 void vec_dst (const unsigned short *, int, const int);
30659 void vec_dst (const short *, int, const int);
30660 void vec_dst (const unsigned int *, int, const int);
30661 void vec_dst (const int *, int, const int);
30662 void vec_dst (const unsigned long *, int, const int);
30663 void vec_dst (const long *, int, const int);
30664 void vec_dst (const float *, int, const int);
30666 void vec_dstst (const vector unsigned char *, int, const int);
30667 void vec_dstst (const vector signed char *, int, const int);
30668 void vec_dstst (const vector bool char *, int, const int);
30669 void vec_dstst (const vector unsigned short *, int, const int);
30670 void vec_dstst (const vector signed short *, int, const int);
30671 void vec_dstst (const vector bool short *, int, const int);
30672 void vec_dstst (const vector pixel *, int, const int);
30673 void vec_dstst (const vector unsigned int *, int, const int);
30674 void vec_dstst (const vector signed int *, int, const int);
30675 void vec_dstst (const vector bool int *, int, const int);
30676 void vec_dstst (const vector float *, int, const int);
30677 void vec_dstst (const unsigned char *, int, const int);
30678 void vec_dstst (const signed char *, int, const int);
30679 void vec_dstst (const unsigned short *, int, const int);
30680 void vec_dstst (const short *, int, const int);
30681 void vec_dstst (const unsigned int *, int, const int);
30682 void vec_dstst (const int *, int, const int);
30683 void vec_dstst (const unsigned long *, int, const int);
30684 void vec_dstst (const long *, int, const int);
30685 void vec_dstst (const float *, int, const int);
30687 void vec_dststt (const vector unsigned char *, int, const int);
30688 void vec_dststt (const vector signed char *, int, const int);
30689 void vec_dststt (const vector bool char *, int, const int);
30690 void vec_dststt (const vector unsigned short *, int, const int);
30691 void vec_dststt (const vector signed short *, int, const int);
30692 void vec_dststt (const vector bool short *, int, const int);
30693 void vec_dststt (const vector pixel *, int, const int);
30694 void vec_dststt (const vector unsigned int *, int, const int);
30695 void vec_dststt (const vector signed int *, int, const int);
30696 void vec_dststt (const vector bool int *, int, const int);
30697 void vec_dststt (const vector float *, int, const int);
30698 void vec_dststt (const unsigned char *, int, const int);
30699 void vec_dststt (const signed char *, int, const int);
30700 void vec_dststt (const unsigned short *, int, const int);
30701 void vec_dststt (const short *, int, const int);
30702 void vec_dststt (const unsigned int *, int, const int);
30703 void vec_dststt (const int *, int, const int);
30704 void vec_dststt (const unsigned long *, int, const int);
30705 void vec_dststt (const long *, int, const int);
30706 void vec_dststt (const float *, int, const int);
30708 void vec_dstt (const vector unsigned char *, int, const int);
30709 void vec_dstt (const vector signed char *, int, const int);
30710 void vec_dstt (const vector bool char *, int, const int);
30711 void vec_dstt (const vector unsigned short *, int, const int);
30712 void vec_dstt (const vector signed short *, int, const int);
30713 void vec_dstt (const vector bool short *, int, const int);
30714 void vec_dstt (const vector pixel *, int, const int);
30715 void vec_dstt (const vector unsigned int *, int, const int);
30716 void vec_dstt (const vector signed int *, int, const int);
30717 void vec_dstt (const vector bool int *, int, const int);
30718 void vec_dstt (const vector float *, int, const int);
30719 void vec_dstt (const unsigned char *, int, const int);
30720 void vec_dstt (const signed char *, int, const int);
30721 void vec_dstt (const unsigned short *, int, const int);
30722 void vec_dstt (const short *, int, const int);
30723 void vec_dstt (const unsigned int *, int, const int);
30724 void vec_dstt (const int *, int, const int);
30725 void vec_dstt (const unsigned long *, int, const int);
30726 void vec_dstt (const long *, int, const int);
30727 void vec_dstt (const float *, int, const int);
30729 vector float vec_expte (vector float);
30731 vector float vec_floor (vector float);
30733 vector float vec_ld (int, const vector float *);
30734 vector float vec_ld (int, const float *);
30735 vector bool int vec_ld (int, const vector bool int *);
30736 vector signed int vec_ld (int, const vector signed int *);
30737 vector signed int vec_ld (int, const int *);
30738 vector signed int vec_ld (int, const long *);
30739 vector unsigned int vec_ld (int, const vector unsigned int *);
30740 vector unsigned int vec_ld (int, const unsigned int *);
30741 vector unsigned int vec_ld (int, const unsigned long *);
30742 vector bool short vec_ld (int, const vector bool short *);
30743 vector pixel vec_ld (int, const vector pixel *);
30744 vector signed short vec_ld (int, const vector signed short *);
30745 vector signed short vec_ld (int, const short *);
30746 vector unsigned short vec_ld (int, const vector unsigned short *);
30747 vector unsigned short vec_ld (int, const unsigned short *);
30748 vector bool char vec_ld (int, const vector bool char *);
30749 vector signed char vec_ld (int, const vector signed char *);
30750 vector signed char vec_ld (int, const signed char *);
30751 vector unsigned char vec_ld (int, const vector unsigned char *);
30752 vector unsigned char vec_ld (int, const unsigned char *);
30754 vector signed char vec_lde (int, const signed char *);
30755 vector unsigned char vec_lde (int, const unsigned char *);
30756 vector signed short vec_lde (int, const short *);
30757 vector unsigned short vec_lde (int, const unsigned short *);
30758 vector float vec_lde (int, const float *);
30759 vector signed int vec_lde (int, const int *);
30760 vector unsigned int vec_lde (int, const unsigned int *);
30761 vector signed int vec_lde (int, const long *);
30762 vector unsigned int vec_lde (int, const unsigned long *);
30764 vector float vec_lvewx (int, float *);
30765 vector signed int vec_lvewx (int, int *);
30766 vector unsigned int vec_lvewx (int, unsigned int *);
30767 vector signed int vec_lvewx (int, long *);
30768 vector unsigned int vec_lvewx (int, unsigned long *);
30770 vector signed short vec_lvehx (int, short *);
30771 vector unsigned short vec_lvehx (int, unsigned short *);
30773 vector signed char vec_lvebx (int, char *);
30774 vector unsigned char vec_lvebx (int, unsigned char *);
30776 vector float vec_ldl (int, const vector float *);
30777 vector float vec_ldl (int, const float *);
30778 vector bool int vec_ldl (int, const vector bool int *);
30779 vector signed int vec_ldl (int, const vector signed int *);
30780 vector signed int vec_ldl (int, const int *);
30781 vector signed int vec_ldl (int, const long *);
30782 vector unsigned int vec_ldl (int, const vector unsigned int *);
30783 vector unsigned int vec_ldl (int, const unsigned int *);
30784 vector unsigned int vec_ldl (int, const unsigned long *);
30785 vector bool short vec_ldl (int, const vector bool short *);
30786 vector pixel vec_ldl (int, const vector pixel *);
30787 vector signed short vec_ldl (int, const vector signed short *);
30788 vector signed short vec_ldl (int, const short *);
30789 vector unsigned short vec_ldl (int, const vector unsigned short *);
30790 vector unsigned short vec_ldl (int, const unsigned short *);
30791 vector bool char vec_ldl (int, const vector bool char *);
30792 vector signed char vec_ldl (int, const vector signed char *);
30793 vector signed char vec_ldl (int, const signed char *);
30794 vector unsigned char vec_ldl (int, const vector unsigned char *);
30795 vector unsigned char vec_ldl (int, const unsigned char *);
30797 vector float vec_loge (vector float);
30799 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
30800 vector unsigned char vec_lvsl (int, const volatile signed char *);
30801 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
30802 vector unsigned char vec_lvsl (int, const volatile short *);
30803 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
30804 vector unsigned char vec_lvsl (int, const volatile int *);
30805 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
30806 vector unsigned char vec_lvsl (int, const volatile long *);
30807 vector unsigned char vec_lvsl (int, const volatile float *);
30809 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
30810 vector unsigned char vec_lvsr (int, const volatile signed char *);
30811 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
30812 vector unsigned char vec_lvsr (int, const volatile short *);
30813 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
30814 vector unsigned char vec_lvsr (int, const volatile int *);
30815 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
30816 vector unsigned char vec_lvsr (int, const volatile long *);
30817 vector unsigned char vec_lvsr (int, const volatile float *);
30819 vector float vec_madd (vector float, vector float, vector float);
30821 vector signed short vec_madds (vector signed short,
30822 vector signed short,
30823 vector signed short);
30825 vector unsigned char vec_max (vector bool char, vector unsigned char);
30826 vector unsigned char vec_max (vector unsigned char, vector bool char);
30827 vector unsigned char vec_max (vector unsigned char,
30828 vector unsigned char);
30829 vector signed char vec_max (vector bool char, vector signed char);
30830 vector signed char vec_max (vector signed char, vector bool char);
30831 vector signed char vec_max (vector signed char, vector signed char);
30832 vector unsigned short vec_max (vector bool short,
30833 vector unsigned short);
30834 vector unsigned short vec_max (vector unsigned short,
30835 vector bool short);
30836 vector unsigned short vec_max (vector unsigned short,
30837 vector unsigned short);
30838 vector signed short vec_max (vector bool short, vector signed short);
30839 vector signed short vec_max (vector signed short, vector bool short);
30840 vector signed short vec_max (vector signed short, vector signed short);
30841 vector unsigned int vec_max (vector bool int, vector unsigned int);
30842 vector unsigned int vec_max (vector unsigned int, vector bool int);
30843 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
30844 vector signed int vec_max (vector bool int, vector signed int);
30845 vector signed int vec_max (vector signed int, vector bool int);
30846 vector signed int vec_max (vector signed int, vector signed int);
30847 vector float vec_max (vector float, vector float);
30849 vector float vec_vmaxfp (vector float, vector float);
30851 vector signed int vec_vmaxsw (vector bool int, vector signed int);
30852 vector signed int vec_vmaxsw (vector signed int, vector bool int);
30853 vector signed int vec_vmaxsw (vector signed int, vector signed int);
30855 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
30856 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
30857 vector unsigned int vec_vmaxuw (vector unsigned int,
30858 vector unsigned int);
30860 vector signed short vec_vmaxsh (vector bool short, vector signed short);
30861 vector signed short vec_vmaxsh (vector signed short, vector bool short);
30862 vector signed short vec_vmaxsh (vector signed short,
30863 vector signed short);
30865 vector unsigned short vec_vmaxuh (vector bool short,
30866 vector unsigned short);
30867 vector unsigned short vec_vmaxuh (vector unsigned short,
30868 vector bool short);
30869 vector unsigned short vec_vmaxuh (vector unsigned short,
30870 vector unsigned short);
30872 vector signed char vec_vmaxsb (vector bool char, vector signed char);
30873 vector signed char vec_vmaxsb (vector signed char, vector bool char);
30874 vector signed char vec_vmaxsb (vector signed char, vector signed char);
30876 vector unsigned char vec_vmaxub (vector bool char,
30877 vector unsigned char);
30878 vector unsigned char vec_vmaxub (vector unsigned char,
30880 vector unsigned char vec_vmaxub (vector unsigned char,
30881 vector unsigned char);
30883 vector bool char vec_mergeh (vector bool char, vector bool char);
30884 vector signed char vec_mergeh (vector signed char, vector signed char);
30885 vector unsigned char vec_mergeh (vector unsigned char,
30886 vector unsigned char);
30887 vector bool short vec_mergeh (vector bool short, vector bool short);
30888 vector pixel vec_mergeh (vector pixel, vector pixel);
30889 vector signed short vec_mergeh (vector signed short,
30890 vector signed short);
30891 vector unsigned short vec_mergeh (vector unsigned short,
30892 vector unsigned short);
30893 vector float vec_mergeh (vector float, vector float);
30894 vector bool int vec_mergeh (vector bool int, vector bool int);
30895 vector signed int vec_mergeh (vector signed int, vector signed int);
30896 vector unsigned int vec_mergeh (vector unsigned int,
30897 vector unsigned int);
30899 vector float vec_vmrghw (vector float, vector float);
30900 vector bool int vec_vmrghw (vector bool int, vector bool int);
30901 vector signed int vec_vmrghw (vector signed int, vector signed int);
30902 vector unsigned int vec_vmrghw (vector unsigned int,
30903 vector unsigned int);
30905 vector bool short vec_vmrghh (vector bool short, vector bool short);
30906 vector signed short vec_vmrghh (vector signed short,
30907 vector signed short);
30908 vector unsigned short vec_vmrghh (vector unsigned short,
30909 vector unsigned short);
30910 vector pixel vec_vmrghh (vector pixel, vector pixel);
30912 vector bool char vec_vmrghb (vector bool char, vector bool char);
30913 vector signed char vec_vmrghb (vector signed char, vector signed char);
30914 vector unsigned char vec_vmrghb (vector unsigned char,
30915 vector unsigned char);
30917 vector bool char vec_mergel (vector bool char, vector bool char);
30918 vector signed char vec_mergel (vector signed char, vector signed char);
30919 vector unsigned char vec_mergel (vector unsigned char,
30920 vector unsigned char);
30921 vector bool short vec_mergel (vector bool short, vector bool short);
30922 vector pixel vec_mergel (vector pixel, vector pixel);
30923 vector signed short vec_mergel (vector signed short,
30924 vector signed short);
30925 vector unsigned short vec_mergel (vector unsigned short,
30926 vector unsigned short);
30927 vector float vec_mergel (vector float, vector float);
30928 vector bool int vec_mergel (vector bool int, vector bool int);
30929 vector signed int vec_mergel (vector signed int, vector signed int);
30930 vector unsigned int vec_mergel (vector unsigned int,
30931 vector unsigned int);
30933 vector float vec_vmrglw (vector float, vector float);
30934 vector signed int vec_vmrglw (vector signed int, vector signed int);
30935 vector unsigned int vec_vmrglw (vector unsigned int,
30936 vector unsigned int);
30937 vector bool int vec_vmrglw (vector bool int, vector bool int);
30939 vector bool short vec_vmrglh (vector bool short, vector bool short);
30940 vector signed short vec_vmrglh (vector signed short,
30941 vector signed short);
30942 vector unsigned short vec_vmrglh (vector unsigned short,
30943 vector unsigned short);
30944 vector pixel vec_vmrglh (vector pixel, vector pixel);
30946 vector bool char vec_vmrglb (vector bool char, vector bool char);
30947 vector signed char vec_vmrglb (vector signed char, vector signed char);
30948 vector unsigned char vec_vmrglb (vector unsigned char,
30949 vector unsigned char);
30951 vector unsigned short vec_mfvscr (void);
30953 vector unsigned char vec_min (vector bool char, vector unsigned char);
30954 vector unsigned char vec_min (vector unsigned char, vector bool char);
30955 vector unsigned char vec_min (vector unsigned char,
30956 vector unsigned char);
30957 vector signed char vec_min (vector bool char, vector signed char);
30958 vector signed char vec_min (vector signed char, vector bool char);
30959 vector signed char vec_min (vector signed char, vector signed char);
30960 vector unsigned short vec_min (vector bool short,
30961 vector unsigned short);
30962 vector unsigned short vec_min (vector unsigned short,
30963 vector bool short);
30964 vector unsigned short vec_min (vector unsigned short,
30965 vector unsigned short);
30966 vector signed short vec_min (vector bool short, vector signed short);
30967 vector signed short vec_min (vector signed short, vector bool short);
30968 vector signed short vec_min (vector signed short, vector signed short);
30969 vector unsigned int vec_min (vector bool int, vector unsigned int);
30970 vector unsigned int vec_min (vector unsigned int, vector bool int);
30971 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
30972 vector signed int vec_min (vector bool int, vector signed int);
30973 vector signed int vec_min (vector signed int, vector bool int);
30974 vector signed int vec_min (vector signed int, vector signed int);
30975 vector float vec_min (vector float, vector float);
30977 vector float vec_vminfp (vector float, vector float);
30979 vector signed int vec_vminsw (vector bool int, vector signed int);
30980 vector signed int vec_vminsw (vector signed int, vector bool int);
30981 vector signed int vec_vminsw (vector signed int, vector signed int);
30983 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
30984 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
30985 vector unsigned int vec_vminuw (vector unsigned int,
30986 vector unsigned int);
30988 vector signed short vec_vminsh (vector bool short, vector signed short);
30989 vector signed short vec_vminsh (vector signed short, vector bool short);
30990 vector signed short vec_vminsh (vector signed short,
30991 vector signed short);
30993 vector unsigned short vec_vminuh (vector bool short,
30994 vector unsigned short);
30995 vector unsigned short vec_vminuh (vector unsigned short,
30996 vector bool short);
30997 vector unsigned short vec_vminuh (vector unsigned short,
30998 vector unsigned short);
31000 vector signed char vec_vminsb (vector bool char, vector signed char);
31001 vector signed char vec_vminsb (vector signed char, vector bool char);
31002 vector signed char vec_vminsb (vector signed char, vector signed char);
31004 vector unsigned char vec_vminub (vector bool char,
31005 vector unsigned char);
31006 vector unsigned char vec_vminub (vector unsigned char,
31008 vector unsigned char vec_vminub (vector unsigned char,
31009 vector unsigned char);
31011 vector signed short vec_mladd (vector signed short,
31012 vector signed short,
31013 vector signed short);
31014 vector signed short vec_mladd (vector signed short,
31015 vector unsigned short,
31016 vector unsigned short);
31017 vector signed short vec_mladd (vector unsigned short,
31018 vector signed short,
31019 vector signed short);
31020 vector unsigned short vec_mladd (vector unsigned short,
31021 vector unsigned short,
31022 vector unsigned short);
31024 vector signed short vec_mradds (vector signed short,
31025 vector signed short,
31026 vector signed short);
31028 vector unsigned int vec_msum (vector unsigned char,
31029 vector unsigned char,
31030 vector unsigned int);
31031 vector signed int vec_msum (vector signed char,
31032 vector unsigned char,
31033 vector signed int);
31034 vector unsigned int vec_msum (vector unsigned short,
31035 vector unsigned short,
31036 vector unsigned int);
31037 vector signed int vec_msum (vector signed short,
31038 vector signed short,
31039 vector signed int);
31041 vector signed int vec_vmsumshm (vector signed short,
31042 vector signed short,
31043 vector signed int);
31045 vector unsigned int vec_vmsumuhm (vector unsigned short,
31046 vector unsigned short,
31047 vector unsigned int);
31049 vector signed int vec_vmsummbm (vector signed char,
31050 vector unsigned char,
31051 vector signed int);
31053 vector unsigned int vec_vmsumubm (vector unsigned char,
31054 vector unsigned char,
31055 vector unsigned int);
31057 vector unsigned int vec_msums (vector unsigned short,
31058 vector unsigned short,
31059 vector unsigned int);
31060 vector signed int vec_msums (vector signed short,
31061 vector signed short,
31062 vector signed int);
31064 vector signed int vec_vmsumshs (vector signed short,
31065 vector signed short,
31066 vector signed int);
31068 vector unsigned int vec_vmsumuhs (vector unsigned short,
31069 vector unsigned short,
31070 vector unsigned int);
31072 void vec_mtvscr (vector signed int);
31073 void vec_mtvscr (vector unsigned int);
31074 void vec_mtvscr (vector bool int);
31075 void vec_mtvscr (vector signed short);
31076 void vec_mtvscr (vector unsigned short);
31077 void vec_mtvscr (vector bool short);
31078 void vec_mtvscr (vector pixel);
31079 void vec_mtvscr (vector signed char);
31080 void vec_mtvscr (vector unsigned char);
31081 void vec_mtvscr (vector bool char);
31083 vector unsigned short vec_mule (vector unsigned char,
31084 vector unsigned char);
31085 vector signed short vec_mule (vector signed char,
31086 vector signed char);
31087 vector unsigned int vec_mule (vector unsigned short,
31088 vector unsigned short);
31089 vector signed int vec_mule (vector signed short, vector signed short);
31091 vector signed int vec_vmulesh (vector signed short,
31092 vector signed short);
31094 vector unsigned int vec_vmuleuh (vector unsigned short,
31095 vector unsigned short);
31097 vector signed short vec_vmulesb (vector signed char,
31098 vector signed char);
31100 vector unsigned short vec_vmuleub (vector unsigned char,
31101 vector unsigned char);
31103 vector unsigned short vec_mulo (vector unsigned char,
31104 vector unsigned char);
31105 vector signed short vec_mulo (vector signed char, vector signed char);
31106 vector unsigned int vec_mulo (vector unsigned short,
31107 vector unsigned short);
31108 vector signed int vec_mulo (vector signed short, vector signed short);
31110 vector signed int vec_vmulosh (vector signed short,
31111 vector signed short);
31113 vector unsigned int vec_vmulouh (vector unsigned short,
31114 vector unsigned short);
31116 vector signed short vec_vmulosb (vector signed char,
31117 vector signed char);
31119 vector unsigned short vec_vmuloub (vector unsigned char,
31120 vector unsigned char);
31122 vector float vec_nmsub (vector float, vector float, vector float);
31124 vector float vec_nor (vector float, vector float);
31125 vector signed int vec_nor (vector signed int, vector signed int);
31126 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
31127 vector bool int vec_nor (vector bool int, vector bool int);
31128 vector signed short vec_nor (vector signed short, vector signed short);
31129 vector unsigned short vec_nor (vector unsigned short,
31130 vector unsigned short);
31131 vector bool short vec_nor (vector bool short, vector bool short);
31132 vector signed char vec_nor (vector signed char, vector signed char);
31133 vector unsigned char vec_nor (vector unsigned char,
31134 vector unsigned char);
31135 vector bool char vec_nor (vector bool char, vector bool char);
31137 vector float vec_or (vector float, vector float);
31138 vector float vec_or (vector float, vector bool int);
31139 vector float vec_or (vector bool int, vector float);
31140 vector bool int vec_or (vector bool int, vector bool int);
31141 vector signed int vec_or (vector bool int, vector signed int);
31142 vector signed int vec_or (vector signed int, vector bool int);
31143 vector signed int vec_or (vector signed int, vector signed int);
31144 vector unsigned int vec_or (vector bool int, vector unsigned int);
31145 vector unsigned int vec_or (vector unsigned int, vector bool int);
31146 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
31147 vector bool short vec_or (vector bool short, vector bool short);
31148 vector signed short vec_or (vector bool short, vector signed short);
31149 vector signed short vec_or (vector signed short, vector bool short);
31150 vector signed short vec_or (vector signed short, vector signed short);
31151 vector unsigned short vec_or (vector bool short, vector unsigned short);
31152 vector unsigned short vec_or (vector unsigned short, vector bool short);
31153 vector unsigned short vec_or (vector unsigned short,
31154 vector unsigned short);
31155 vector signed char vec_or (vector bool char, vector signed char);
31156 vector bool char vec_or (vector bool char, vector bool char);
31157 vector signed char vec_or (vector signed char, vector bool char);
31158 vector signed char vec_or (vector signed char, vector signed char);
31159 vector unsigned char vec_or (vector bool char, vector unsigned char);
31160 vector unsigned char vec_or (vector unsigned char, vector bool char);
31161 vector unsigned char vec_or (vector unsigned char,
31162 vector unsigned char);
31164 vector signed char vec_pack (vector signed short, vector signed short);
31165 vector unsigned char vec_pack (vector unsigned short,
31166 vector unsigned short);
31167 vector bool char vec_pack (vector bool short, vector bool short);
31168 vector signed short vec_pack (vector signed int, vector signed int);
31169 vector unsigned short vec_pack (vector unsigned int,
31170 vector unsigned int);
31171 vector bool short vec_pack (vector bool int, vector bool int);
31173 vector bool short vec_vpkuwum (vector bool int, vector bool int);
31174 vector signed short vec_vpkuwum (vector signed int, vector signed int);
31175 vector unsigned short vec_vpkuwum (vector unsigned int,
31176 vector unsigned int);
31178 vector bool char vec_vpkuhum (vector bool short, vector bool short);
31179 vector signed char vec_vpkuhum (vector signed short,
31180 vector signed short);
31181 vector unsigned char vec_vpkuhum (vector unsigned short,
31182 vector unsigned short);
31184 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
31186 vector unsigned char vec_packs (vector unsigned short,
31187 vector unsigned short);
31188 vector signed char vec_packs (vector signed short, vector signed short);
31189 vector unsigned short vec_packs (vector unsigned int,
31190 vector unsigned int);
31191 vector signed short vec_packs (vector signed int, vector signed int);
31193 vector signed short vec_vpkswss (vector signed int, vector signed int);
31195 vector unsigned short vec_vpkuwus (vector unsigned int,
31196 vector unsigned int);
31198 vector signed char vec_vpkshss (vector signed short,
31199 vector signed short);
31201 vector unsigned char vec_vpkuhus (vector unsigned short,
31202 vector unsigned short);
31204 vector unsigned char vec_packsu (vector unsigned short,
31205 vector unsigned short);
31206 vector unsigned char vec_packsu (vector signed short,
31207 vector signed short);
31208 vector unsigned short vec_packsu (vector unsigned int,
31209 vector unsigned int);
31210 vector unsigned short vec_packsu (vector signed int, vector signed int);
31212 vector unsigned short vec_vpkswus (vector signed int,
31213 vector signed int);
31215 vector unsigned char vec_vpkshus (vector signed short,
31216 vector signed short);
31218 vector float vec_perm (vector float,
31220 vector unsigned char);
31221 vector signed int vec_perm (vector signed int,
31223 vector unsigned char);
31224 vector unsigned int vec_perm (vector unsigned int,
31225 vector unsigned int,
31226 vector unsigned char);
31227 vector bool int vec_perm (vector bool int,
31229 vector unsigned char);
31230 vector signed short vec_perm (vector signed short,
31231 vector signed short,
31232 vector unsigned char);
31233 vector unsigned short vec_perm (vector unsigned short,
31234 vector unsigned short,
31235 vector unsigned char);
31236 vector bool short vec_perm (vector bool short,
31238 vector unsigned char);
31239 vector pixel vec_perm (vector pixel,
31241 vector unsigned char);
31242 vector signed char vec_perm (vector signed char,
31243 vector signed char,
31244 vector unsigned char);
31245 vector unsigned char vec_perm (vector unsigned char,
31246 vector unsigned char,
31247 vector unsigned char);
31248 vector bool char vec_perm (vector bool char,
31250 vector unsigned char);
31252 vector float vec_re (vector float);
31254 vector signed char vec_rl (vector signed char,
31255 vector unsigned char);
31256 vector unsigned char vec_rl (vector unsigned char,
31257 vector unsigned char);
31258 vector signed short vec_rl (vector signed short, vector unsigned short);
31259 vector unsigned short vec_rl (vector unsigned short,
31260 vector unsigned short);
31261 vector signed int vec_rl (vector signed int, vector unsigned int);
31262 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
31264 vector signed int vec_vrlw (vector signed int, vector unsigned int);
31265 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
31267 vector signed short vec_vrlh (vector signed short,
31268 vector unsigned short);
31269 vector unsigned short vec_vrlh (vector unsigned short,
31270 vector unsigned short);
31272 vector signed char vec_vrlb (vector signed char, vector unsigned char);
31273 vector unsigned char vec_vrlb (vector unsigned char,
31274 vector unsigned char);
31276 vector float vec_round (vector float);
31278 vector float vec_rsqrte (vector float);
31280 vector float vec_sel (vector float, vector float, vector bool int);
31281 vector float vec_sel (vector float, vector float, vector unsigned int);
31282 vector signed int vec_sel (vector signed int,
31285 vector signed int vec_sel (vector signed int,
31287 vector unsigned int);
31288 vector unsigned int vec_sel (vector unsigned int,
31289 vector unsigned int,
31291 vector unsigned int vec_sel (vector unsigned int,
31292 vector unsigned int,
31293 vector unsigned int);
31294 vector bool int vec_sel (vector bool int,
31297 vector bool int vec_sel (vector bool int,
31299 vector unsigned int);
31300 vector signed short vec_sel (vector signed short,
31301 vector signed short,
31302 vector bool short);
31303 vector signed short vec_sel (vector signed short,
31304 vector signed short,
31305 vector unsigned short);
31306 vector unsigned short vec_sel (vector unsigned short,
31307 vector unsigned short,
31308 vector bool short);
31309 vector unsigned short vec_sel (vector unsigned short,
31310 vector unsigned short,
31311 vector unsigned short);
31312 vector bool short vec_sel (vector bool short,
31314 vector bool short);
31315 vector bool short vec_sel (vector bool short,
31317 vector unsigned short);
31318 vector signed char vec_sel (vector signed char,
31319 vector signed char,
31321 vector signed char vec_sel (vector signed char,
31322 vector signed char,
31323 vector unsigned char);
31324 vector unsigned char vec_sel (vector unsigned char,
31325 vector unsigned char,
31327 vector unsigned char vec_sel (vector unsigned char,
31328 vector unsigned char,
31329 vector unsigned char);
31330 vector bool char vec_sel (vector bool char,
31333 vector bool char vec_sel (vector bool char,
31335 vector unsigned char);
31337 vector signed char vec_sl (vector signed char,
31338 vector unsigned char);
31339 vector unsigned char vec_sl (vector unsigned char,
31340 vector unsigned char);
31341 vector signed short vec_sl (vector signed short, vector unsigned short);
31342 vector unsigned short vec_sl (vector unsigned short,
31343 vector unsigned short);
31344 vector signed int vec_sl (vector signed int, vector unsigned int);
31345 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
31347 vector signed int vec_vslw (vector signed int, vector unsigned int);
31348 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
31350 vector signed short vec_vslh (vector signed short,
31351 vector unsigned short);
31352 vector unsigned short vec_vslh (vector unsigned short,
31353 vector unsigned short);
31355 vector signed char vec_vslb (vector signed char, vector unsigned char);
31356 vector unsigned char vec_vslb (vector unsigned char,
31357 vector unsigned char);
31359 vector float vec_sld (vector float, vector float, const int);
31360 vector signed int vec_sld (vector signed int,
31363 vector unsigned int vec_sld (vector unsigned int,
31364 vector unsigned int,
31366 vector bool int vec_sld (vector bool int,
31369 vector signed short vec_sld (vector signed short,
31370 vector signed short,
31372 vector unsigned short vec_sld (vector unsigned short,
31373 vector unsigned short,
31375 vector bool short vec_sld (vector bool short,
31378 vector pixel vec_sld (vector pixel,
31381 vector signed char vec_sld (vector signed char,
31382 vector signed char,
31384 vector unsigned char vec_sld (vector unsigned char,
31385 vector unsigned char,
31387 vector bool char vec_sld (vector bool char,
31391 vector signed int vec_sll (vector signed int,
31392 vector unsigned int);
31393 vector signed int vec_sll (vector signed int,
31394 vector unsigned short);
31395 vector signed int vec_sll (vector signed int,
31396 vector unsigned char);
31397 vector unsigned int vec_sll (vector unsigned int,
31398 vector unsigned int);
31399 vector unsigned int vec_sll (vector unsigned int,
31400 vector unsigned short);
31401 vector unsigned int vec_sll (vector unsigned int,
31402 vector unsigned char);
31403 vector bool int vec_sll (vector bool int,
31404 vector unsigned int);
31405 vector bool int vec_sll (vector bool int,
31406 vector unsigned short);
31407 vector bool int vec_sll (vector bool int,
31408 vector unsigned char);
31409 vector signed short vec_sll (vector signed short,
31410 vector unsigned int);
31411 vector signed short vec_sll (vector signed short,
31412 vector unsigned short);
31413 vector signed short vec_sll (vector signed short,
31414 vector unsigned char);
31415 vector unsigned short vec_sll (vector unsigned short,
31416 vector unsigned int);
31417 vector unsigned short vec_sll (vector unsigned short,
31418 vector unsigned short);
31419 vector unsigned short vec_sll (vector unsigned short,
31420 vector unsigned char);
31421 vector bool short vec_sll (vector bool short, vector unsigned int);
31422 vector bool short vec_sll (vector bool short, vector unsigned short);
31423 vector bool short vec_sll (vector bool short, vector unsigned char);
31424 vector pixel vec_sll (vector pixel, vector unsigned int);
31425 vector pixel vec_sll (vector pixel, vector unsigned short);
31426 vector pixel vec_sll (vector pixel, vector unsigned char);
31427 vector signed char vec_sll (vector signed char, vector unsigned int);
31428 vector signed char vec_sll (vector signed char, vector unsigned short);
31429 vector signed char vec_sll (vector signed char, vector unsigned char);
31430 vector unsigned char vec_sll (vector unsigned char,
31431 vector unsigned int);
31432 vector unsigned char vec_sll (vector unsigned char,
31433 vector unsigned short);
31434 vector unsigned char vec_sll (vector unsigned char,
31435 vector unsigned char);
31436 vector bool char vec_sll (vector bool char, vector unsigned int);
31437 vector bool char vec_sll (vector bool char, vector unsigned short);
31438 vector bool char vec_sll (vector bool char, vector unsigned char);
31440 vector float vec_slo (vector float, vector signed char);
31441 vector float vec_slo (vector float, vector unsigned char);
31442 vector signed int vec_slo (vector signed int, vector signed char);
31443 vector signed int vec_slo (vector signed int, vector unsigned char);
31444 vector unsigned int vec_slo (vector unsigned int, vector signed char);
31445 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
31446 vector signed short vec_slo (vector signed short, vector signed char);
31447 vector signed short vec_slo (vector signed short, vector unsigned char);
31448 vector unsigned short vec_slo (vector unsigned short,
31449 vector signed char);
31450 vector unsigned short vec_slo (vector unsigned short,
31451 vector unsigned char);
31452 vector pixel vec_slo (vector pixel, vector signed char);
31453 vector pixel vec_slo (vector pixel, vector unsigned char);
31454 vector signed char vec_slo (vector signed char, vector signed char);
31455 vector signed char vec_slo (vector signed char, vector unsigned char);
31456 vector unsigned char vec_slo (vector unsigned char, vector signed char);
31457 vector unsigned char vec_slo (vector unsigned char,
31458 vector unsigned char);
31460 vector signed char vec_splat (vector signed char, const int);
31461 vector unsigned char vec_splat (vector unsigned char, const int);
31462 vector bool char vec_splat (vector bool char, const int);
31463 vector signed short vec_splat (vector signed short, const int);
31464 vector unsigned short vec_splat (vector unsigned short, const int);
31465 vector bool short vec_splat (vector bool short, const int);
31466 vector pixel vec_splat (vector pixel, const int);
31467 vector float vec_splat (vector float, const int);
31468 vector signed int vec_splat (vector signed int, const int);
31469 vector unsigned int vec_splat (vector unsigned int, const int);
31470 vector bool int vec_splat (vector bool int, const int);
31472 vector float vec_vspltw (vector float, const int);
31473 vector signed int vec_vspltw (vector signed int, const int);
31474 vector unsigned int vec_vspltw (vector unsigned int, const int);
31475 vector bool int vec_vspltw (vector bool int, const int);
31477 vector bool short vec_vsplth (vector bool short, const int);
31478 vector signed short vec_vsplth (vector signed short, const int);
31479 vector unsigned short vec_vsplth (vector unsigned short, const int);
31480 vector pixel vec_vsplth (vector pixel, const int);
31482 vector signed char vec_vspltb (vector signed char, const int);
31483 vector unsigned char vec_vspltb (vector unsigned char, const int);
31484 vector bool char vec_vspltb (vector bool char, const int);
31486 vector signed char vec_splat_s8 (const int);
31488 vector signed short vec_splat_s16 (const int);
31490 vector signed int vec_splat_s32 (const int);
31492 vector unsigned char vec_splat_u8 (const int);
31494 vector unsigned short vec_splat_u16 (const int);
31496 vector unsigned int vec_splat_u32 (const int);
31498 vector signed char vec_sr (vector signed char, vector unsigned char);
31499 vector unsigned char vec_sr (vector unsigned char,
31500 vector unsigned char);
31501 vector signed short vec_sr (vector signed short,
31502 vector unsigned short);
31503 vector unsigned short vec_sr (vector unsigned short,
31504 vector unsigned short);
31505 vector signed int vec_sr (vector signed int, vector unsigned int);
31506 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
31508 vector signed int vec_vsrw (vector signed int, vector unsigned int);
31509 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
31511 vector signed short vec_vsrh (vector signed short,
31512 vector unsigned short);
31513 vector unsigned short vec_vsrh (vector unsigned short,
31514 vector unsigned short);
31516 vector signed char vec_vsrb (vector signed char, vector unsigned char);
31517 vector unsigned char vec_vsrb (vector unsigned char,
31518 vector unsigned char);
31520 vector signed char vec_sra (vector signed char, vector unsigned char);
31521 vector unsigned char vec_sra (vector unsigned char,
31522 vector unsigned char);
31523 vector signed short vec_sra (vector signed short,
31524 vector unsigned short);
31525 vector unsigned short vec_sra (vector unsigned short,
31526 vector unsigned short);
31527 vector signed int vec_sra (vector signed int, vector unsigned int);
31528 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
31530 vector signed int vec_vsraw (vector signed int, vector unsigned int);
31531 vector unsigned int vec_vsraw (vector unsigned int,
31532 vector unsigned int);
31534 vector signed short vec_vsrah (vector signed short,
31535 vector unsigned short);
31536 vector unsigned short vec_vsrah (vector unsigned short,
31537 vector unsigned short);
31539 vector signed char vec_vsrab (vector signed char, vector unsigned char);
31540 vector unsigned char vec_vsrab (vector unsigned char,
31541 vector unsigned char);
31543 vector signed int vec_srl (vector signed int, vector unsigned int);
31544 vector signed int vec_srl (vector signed int, vector unsigned short);
31545 vector signed int vec_srl (vector signed int, vector unsigned char);
31546 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
31547 vector unsigned int vec_srl (vector unsigned int,
31548 vector unsigned short);
31549 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
31550 vector bool int vec_srl (vector bool int, vector unsigned int);
31551 vector bool int vec_srl (vector bool int, vector unsigned short);
31552 vector bool int vec_srl (vector bool int, vector unsigned char);
31553 vector signed short vec_srl (vector signed short, vector unsigned int);
31554 vector signed short vec_srl (vector signed short,
31555 vector unsigned short);
31556 vector signed short vec_srl (vector signed short, vector unsigned char);
31557 vector unsigned short vec_srl (vector unsigned short,
31558 vector unsigned int);
31559 vector unsigned short vec_srl (vector unsigned short,
31560 vector unsigned short);
31561 vector unsigned short vec_srl (vector unsigned short,
31562 vector unsigned char);
31563 vector bool short vec_srl (vector bool short, vector unsigned int);
31564 vector bool short vec_srl (vector bool short, vector unsigned short);
31565 vector bool short vec_srl (vector bool short, vector unsigned char);
31566 vector pixel vec_srl (vector pixel, vector unsigned int);
31567 vector pixel vec_srl (vector pixel, vector unsigned short);
31568 vector pixel vec_srl (vector pixel, vector unsigned char);
31569 vector signed char vec_srl (vector signed char, vector unsigned int);
31570 vector signed char vec_srl (vector signed char, vector unsigned short);
31571 vector signed char vec_srl (vector signed char, vector unsigned char);
31572 vector unsigned char vec_srl (vector unsigned char,
31573 vector unsigned int);
31574 vector unsigned char vec_srl (vector unsigned char,
31575 vector unsigned short);
31576 vector unsigned char vec_srl (vector unsigned char,
31577 vector unsigned char);
31578 vector bool char vec_srl (vector bool char, vector unsigned int);
31579 vector bool char vec_srl (vector bool char, vector unsigned short);
31580 vector bool char vec_srl (vector bool char, vector unsigned char);
31582 vector float vec_sro (vector float, vector signed char);
31583 vector float vec_sro (vector float, vector unsigned char);
31584 vector signed int vec_sro (vector signed int, vector signed char);
31585 vector signed int vec_sro (vector signed int, vector unsigned char);
31586 vector unsigned int vec_sro (vector unsigned int, vector signed char);
31587 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
31588 vector signed short vec_sro (vector signed short, vector signed char);
31589 vector signed short vec_sro (vector signed short, vector unsigned char);
31590 vector unsigned short vec_sro (vector unsigned short,
31591 vector signed char);
31592 vector unsigned short vec_sro (vector unsigned short,
31593 vector unsigned char);
31594 vector pixel vec_sro (vector pixel, vector signed char);
31595 vector pixel vec_sro (vector pixel, vector unsigned char);
31596 vector signed char vec_sro (vector signed char, vector signed char);
31597 vector signed char vec_sro (vector signed char, vector unsigned char);
31598 vector unsigned char vec_sro (vector unsigned char, vector signed char);
31599 vector unsigned char vec_sro (vector unsigned char,
31600 vector unsigned char);
31602 void vec_st (vector float, int, vector float *);
31603 void vec_st (vector float, int, float *);
31604 void vec_st (vector signed int, int, vector signed int *);
31605 void vec_st (vector signed int, int, int *);
31606 void vec_st (vector unsigned int, int, vector unsigned int *);
31607 void vec_st (vector unsigned int, int, unsigned int *);
31608 void vec_st (vector bool int, int, vector bool int *);
31609 void vec_st (vector bool int, int, unsigned int *);
31610 void vec_st (vector bool int, int, int *);
31611 void vec_st (vector signed short, int, vector signed short *);
31612 void vec_st (vector signed short, int, short *);
31613 void vec_st (vector unsigned short, int, vector unsigned short *);
31614 void vec_st (vector unsigned short, int, unsigned short *);
31615 void vec_st (vector bool short, int, vector bool short *);
31616 void vec_st (vector bool short, int, unsigned short *);
31617 void vec_st (vector pixel, int, vector pixel *);
31618 void vec_st (vector pixel, int, unsigned short *);
31619 void vec_st (vector pixel, int, short *);
31620 void vec_st (vector bool short, int, short *);
31621 void vec_st (vector signed char, int, vector signed char *);
31622 void vec_st (vector signed char, int, signed char *);
31623 void vec_st (vector unsigned char, int, vector unsigned char *);
31624 void vec_st (vector unsigned char, int, unsigned char *);
31625 void vec_st (vector bool char, int, vector bool char *);
31626 void vec_st (vector bool char, int, unsigned char *);
31627 void vec_st (vector bool char, int, signed char *);
31629 void vec_ste (vector signed char, int, signed char *);
31630 void vec_ste (vector unsigned char, int, unsigned char *);
31631 void vec_ste (vector bool char, int, signed char *);
31632 void vec_ste (vector bool char, int, unsigned char *);
31633 void vec_ste (vector signed short, int, short *);
31634 void vec_ste (vector unsigned short, int, unsigned short *);
31635 void vec_ste (vector bool short, int, short *);
31636 void vec_ste (vector bool short, int, unsigned short *);
31637 void vec_ste (vector pixel, int, short *);
31638 void vec_ste (vector pixel, int, unsigned short *);
31639 void vec_ste (vector float, int, float *);
31640 void vec_ste (vector signed int, int, int *);
31641 void vec_ste (vector unsigned int, int, unsigned int *);
31642 void vec_ste (vector bool int, int, int *);
31643 void vec_ste (vector bool int, int, unsigned int *);
31645 void vec_stvewx (vector float, int, float *);
31646 void vec_stvewx (vector signed int, int, int *);
31647 void vec_stvewx (vector unsigned int, int, unsigned int *);
31648 void vec_stvewx (vector bool int, int, int *);
31649 void vec_stvewx (vector bool int, int, unsigned int *);
31651 void vec_stvehx (vector signed short, int, short *);
31652 void vec_stvehx (vector unsigned short, int, unsigned short *);
31653 void vec_stvehx (vector bool short, int, short *);
31654 void vec_stvehx (vector bool short, int, unsigned short *);
31655 void vec_stvehx (vector pixel, int, short *);
31656 void vec_stvehx (vector pixel, int, unsigned short *);
31658 void vec_stvebx (vector signed char, int, signed char *);
31659 void vec_stvebx (vector unsigned char, int, unsigned char *);
31660 void vec_stvebx (vector bool char, int, signed char *);
31661 void vec_stvebx (vector bool char, int, unsigned char *);
31663 void vec_stl (vector float, int, vector float *);
31664 void vec_stl (vector float, int, float *);
31665 void vec_stl (vector signed int, int, vector signed int *);
31666 void vec_stl (vector signed int, int, int *);
31667 void vec_stl (vector unsigned int, int, vector unsigned int *);
31668 void vec_stl (vector unsigned int, int, unsigned int *);
31669 void vec_stl (vector bool int, int, vector bool int *);
31670 void vec_stl (vector bool int, int, unsigned int *);
31671 void vec_stl (vector bool int, int, int *);
31672 void vec_stl (vector signed short, int, vector signed short *);
31673 void vec_stl (vector signed short, int, short *);
31674 void vec_stl (vector unsigned short, int, vector unsigned short *);
31675 void vec_stl (vector unsigned short, int, unsigned short *);
31676 void vec_stl (vector bool short, int, vector bool short *);
31677 void vec_stl (vector bool short, int, unsigned short *);
31678 void vec_stl (vector bool short, int, short *);
31679 void vec_stl (vector pixel, int, vector pixel *);
31680 void vec_stl (vector pixel, int, unsigned short *);
31681 void vec_stl (vector pixel, int, short *);
31682 void vec_stl (vector signed char, int, vector signed char *);
31683 void vec_stl (vector signed char, int, signed char *);
31684 void vec_stl (vector unsigned char, int, vector unsigned char *);
31685 void vec_stl (vector unsigned char, int, unsigned char *);
31686 void vec_stl (vector bool char, int, vector bool char *);
31687 void vec_stl (vector bool char, int, unsigned char *);
31688 void vec_stl (vector bool char, int, signed char *);
31690 vector signed char vec_sub (vector bool char, vector signed char);
31691 vector signed char vec_sub (vector signed char, vector bool char);
31692 vector signed char vec_sub (vector signed char, vector signed char);
31693 vector unsigned char vec_sub (vector bool char, vector unsigned char);
31694 vector unsigned char vec_sub (vector unsigned char, vector bool char);
31695 vector unsigned char vec_sub (vector unsigned char,
31696 vector unsigned char);
31697 vector signed short vec_sub (vector bool short, vector signed short);
31698 vector signed short vec_sub (vector signed short, vector bool short);
31699 vector signed short vec_sub (vector signed short, vector signed short);
31700 vector unsigned short vec_sub (vector bool short,
31701 vector unsigned short);
31702 vector unsigned short vec_sub (vector unsigned short,
31703 vector bool short);
31704 vector unsigned short vec_sub (vector unsigned short,
31705 vector unsigned short);
31706 vector signed int vec_sub (vector bool int, vector signed int);
31707 vector signed int vec_sub (vector signed int, vector bool int);
31708 vector signed int vec_sub (vector signed int, vector signed int);
31709 vector unsigned int vec_sub (vector bool int, vector unsigned int);
31710 vector unsigned int vec_sub (vector unsigned int, vector bool int);
31711 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
31712 vector float vec_sub (vector float, vector float);
31714 vector float vec_vsubfp (vector float, vector float);
31716 vector signed int vec_vsubuwm (vector bool int, vector signed int);
31717 vector signed int vec_vsubuwm (vector signed int, vector bool int);
31718 vector signed int vec_vsubuwm (vector signed int, vector signed int);
31719 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
31720 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
31721 vector unsigned int vec_vsubuwm (vector unsigned int,
31722 vector unsigned int);
31724 vector signed short vec_vsubuhm (vector bool short,
31725 vector signed short);
31726 vector signed short vec_vsubuhm (vector signed short,
31727 vector bool short);
31728 vector signed short vec_vsubuhm (vector signed short,
31729 vector signed short);
31730 vector unsigned short vec_vsubuhm (vector bool short,
31731 vector unsigned short);
31732 vector unsigned short vec_vsubuhm (vector unsigned short,
31733 vector bool short);
31734 vector unsigned short vec_vsubuhm (vector unsigned short,
31735 vector unsigned short);
31737 vector signed char vec_vsububm (vector bool char, vector signed char);
31738 vector signed char vec_vsububm (vector signed char, vector bool char);
31739 vector signed char vec_vsububm (vector signed char, vector signed char);
31740 vector unsigned char vec_vsububm (vector bool char,
31741 vector unsigned char);
31742 vector unsigned char vec_vsububm (vector unsigned char,
31744 vector unsigned char vec_vsububm (vector unsigned char,
31745 vector unsigned char);
31747 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
31749 vector unsigned char vec_subs (vector bool char, vector unsigned char);
31750 vector unsigned char vec_subs (vector unsigned char, vector bool char);
31751 vector unsigned char vec_subs (vector unsigned char,
31752 vector unsigned char);
31753 vector signed char vec_subs (vector bool char, vector signed char);
31754 vector signed char vec_subs (vector signed char, vector bool char);
31755 vector signed char vec_subs (vector signed char, vector signed char);
31756 vector unsigned short vec_subs (vector bool short,
31757 vector unsigned short);
31758 vector unsigned short vec_subs (vector unsigned short,
31759 vector bool short);
31760 vector unsigned short vec_subs (vector unsigned short,
31761 vector unsigned short);
31762 vector signed short vec_subs (vector bool short, vector signed short);
31763 vector signed short vec_subs (vector signed short, vector bool short);
31764 vector signed short vec_subs (vector signed short, vector signed short);
31765 vector unsigned int vec_subs (vector bool int, vector unsigned int);
31766 vector unsigned int vec_subs (vector unsigned int, vector bool int);
31767 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
31768 vector signed int vec_subs (vector bool int, vector signed int);
31769 vector signed int vec_subs (vector signed int, vector bool int);
31770 vector signed int vec_subs (vector signed int, vector signed int);
31772 vector signed int vec_vsubsws (vector bool int, vector signed int);
31773 vector signed int vec_vsubsws (vector signed int, vector bool int);
31774 vector signed int vec_vsubsws (vector signed int, vector signed int);
31776 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
31777 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
31778 vector unsigned int vec_vsubuws (vector unsigned int,
31779 vector unsigned int);
31781 vector signed short vec_vsubshs (vector bool short,
31782 vector signed short);
31783 vector signed short vec_vsubshs (vector signed short,
31784 vector bool short);
31785 vector signed short vec_vsubshs (vector signed short,
31786 vector signed short);
31788 vector unsigned short vec_vsubuhs (vector bool short,
31789 vector unsigned short);
31790 vector unsigned short vec_vsubuhs (vector unsigned short,
31791 vector bool short);
31792 vector unsigned short vec_vsubuhs (vector unsigned short,
31793 vector unsigned short);
31795 vector signed char vec_vsubsbs (vector bool char, vector signed char);
31796 vector signed char vec_vsubsbs (vector signed char, vector bool char);
31797 vector signed char vec_vsubsbs (vector signed char, vector signed char);
31799 vector unsigned char vec_vsububs (vector bool char,
31800 vector unsigned char);
31801 vector unsigned char vec_vsububs (vector unsigned char,
31803 vector unsigned char vec_vsububs (vector unsigned char,
31804 vector unsigned char);
31806 vector unsigned int vec_sum4s (vector unsigned char,
31807 vector unsigned int);
31808 vector signed int vec_sum4s (vector signed char, vector signed int);
31809 vector signed int vec_sum4s (vector signed short, vector signed int);
31811 vector signed int vec_vsum4shs (vector signed short, vector signed int);
31813 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
31815 vector unsigned int vec_vsum4ubs (vector unsigned char,
31816 vector unsigned int);
31818 vector signed int vec_sum2s (vector signed int, vector signed int);
31820 vector signed int vec_sums (vector signed int, vector signed int);
31822 vector float vec_trunc (vector float);
31824 vector signed short vec_unpackh (vector signed char);
31825 vector bool short vec_unpackh (vector bool char);
31826 vector signed int vec_unpackh (vector signed short);
31827 vector bool int vec_unpackh (vector bool short);
31828 vector unsigned int vec_unpackh (vector pixel);
31830 vector bool int vec_vupkhsh (vector bool short);
31831 vector signed int vec_vupkhsh (vector signed short);
31833 vector unsigned int vec_vupkhpx (vector pixel);
31835 vector bool short vec_vupkhsb (vector bool char);
31836 vector signed short vec_vupkhsb (vector signed char);
31838 vector signed short vec_unpackl (vector signed char);
31839 vector bool short vec_unpackl (vector bool char);
31840 vector unsigned int vec_unpackl (vector pixel);
31841 vector signed int vec_unpackl (vector signed short);
31842 vector bool int vec_unpackl (vector bool short);
31844 vector unsigned int vec_vupklpx (vector pixel);
31846 vector bool int vec_vupklsh (vector bool short);
31847 vector signed int vec_vupklsh (vector signed short);
31849 vector bool short vec_vupklsb (vector bool char);
31850 vector signed short vec_vupklsb (vector signed char);
31852 vector float vec_xor (vector float, vector float);
31853 vector float vec_xor (vector float, vector bool int);
31854 vector float vec_xor (vector bool int, vector float);
31855 vector bool int vec_xor (vector bool int, vector bool int);
31856 vector signed int vec_xor (vector bool int, vector signed int);
31857 vector signed int vec_xor (vector signed int, vector bool int);
31858 vector signed int vec_xor (vector signed int, vector signed int);
31859 vector unsigned int vec_xor (vector bool int, vector unsigned int);
31860 vector unsigned int vec_xor (vector unsigned int, vector bool int);
31861 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
31862 vector bool short vec_xor (vector bool short, vector bool short);
31863 vector signed short vec_xor (vector bool short, vector signed short);
31864 vector signed short vec_xor (vector signed short, vector bool short);
31865 vector signed short vec_xor (vector signed short, vector signed short);
31866 vector unsigned short vec_xor (vector bool short,
31867 vector unsigned short);
31868 vector unsigned short vec_xor (vector unsigned short,
31869 vector bool short);
31870 vector unsigned short vec_xor (vector unsigned short,
31871 vector unsigned short);
31872 vector signed char vec_xor (vector bool char, vector signed char);
31873 vector bool char vec_xor (vector bool char, vector bool char);
31874 vector signed char vec_xor (vector signed char, vector bool char);
31875 vector signed char vec_xor (vector signed char, vector signed char);
31876 vector unsigned char vec_xor (vector bool char, vector unsigned char);
31877 vector unsigned char vec_xor (vector unsigned char, vector bool char);
31878 vector unsigned char vec_xor (vector unsigned char,
31879 vector unsigned char);
31881 int vec_all_eq (vector signed char, vector bool char);
31882 int vec_all_eq (vector signed char, vector signed char);
31883 int vec_all_eq (vector unsigned char, vector bool char);
31884 int vec_all_eq (vector unsigned char, vector unsigned char);
31885 int vec_all_eq (vector bool char, vector bool char);
31886 int vec_all_eq (vector bool char, vector unsigned char);
31887 int vec_all_eq (vector bool char, vector signed char);
31888 int vec_all_eq (vector signed short, vector bool short);
31889 int vec_all_eq (vector signed short, vector signed short);
31890 int vec_all_eq (vector unsigned short, vector bool short);
31891 int vec_all_eq (vector unsigned short, vector unsigned short);
31892 int vec_all_eq (vector bool short, vector bool short);
31893 int vec_all_eq (vector bool short, vector unsigned short);
31894 int vec_all_eq (vector bool short, vector signed short);
31895 int vec_all_eq (vector pixel, vector pixel);
31896 int vec_all_eq (vector signed int, vector bool int);
31897 int vec_all_eq (vector signed int, vector signed int);
31898 int vec_all_eq (vector unsigned int, vector bool int);
31899 int vec_all_eq (vector unsigned int, vector unsigned int);
31900 int vec_all_eq (vector bool int, vector bool int);
31901 int vec_all_eq (vector bool int, vector unsigned int);
31902 int vec_all_eq (vector bool int, vector signed int);
31903 int vec_all_eq (vector float, vector float);
31905 int vec_all_ge (vector bool char, vector unsigned char);
31906 int vec_all_ge (vector unsigned char, vector bool char);
31907 int vec_all_ge (vector unsigned char, vector unsigned char);
31908 int vec_all_ge (vector bool char, vector signed char);
31909 int vec_all_ge (vector signed char, vector bool char);
31910 int vec_all_ge (vector signed char, vector signed char);
31911 int vec_all_ge (vector bool short, vector unsigned short);
31912 int vec_all_ge (vector unsigned short, vector bool short);
31913 int vec_all_ge (vector unsigned short, vector unsigned short);
31914 int vec_all_ge (vector signed short, vector signed short);
31915 int vec_all_ge (vector bool short, vector signed short);
31916 int vec_all_ge (vector signed short, vector bool short);
31917 int vec_all_ge (vector bool int, vector unsigned int);
31918 int vec_all_ge (vector unsigned int, vector bool int);
31919 int vec_all_ge (vector unsigned int, vector unsigned int);
31920 int vec_all_ge (vector bool int, vector signed int);
31921 int vec_all_ge (vector signed int, vector bool int);
31922 int vec_all_ge (vector signed int, vector signed int);
31923 int vec_all_ge (vector float, vector float);
31925 int vec_all_gt (vector bool char, vector unsigned char);
31926 int vec_all_gt (vector unsigned char, vector bool char);
31927 int vec_all_gt (vector unsigned char, vector unsigned char);
31928 int vec_all_gt (vector bool char, vector signed char);
31929 int vec_all_gt (vector signed char, vector bool char);
31930 int vec_all_gt (vector signed char, vector signed char);
31931 int vec_all_gt (vector bool short, vector unsigned short);
31932 int vec_all_gt (vector unsigned short, vector bool short);
31933 int vec_all_gt (vector unsigned short, vector unsigned short);
31934 int vec_all_gt (vector bool short, vector signed short);
31935 int vec_all_gt (vector signed short, vector bool short);
31936 int vec_all_gt (vector signed short, vector signed short);
31937 int vec_all_gt (vector bool int, vector unsigned int);
31938 int vec_all_gt (vector unsigned int, vector bool int);
31939 int vec_all_gt (vector unsigned int, vector unsigned int);
31940 int vec_all_gt (vector bool int, vector signed int);
31941 int vec_all_gt (vector signed int, vector bool int);
31942 int vec_all_gt (vector signed int, vector signed int);
31943 int vec_all_gt (vector float, vector float);
31945 int vec_all_in (vector float, vector float);
31947 int vec_all_le (vector bool char, vector unsigned char);
31948 int vec_all_le (vector unsigned char, vector bool char);
31949 int vec_all_le (vector unsigned char, vector unsigned char);
31950 int vec_all_le (vector bool char, vector signed char);
31951 int vec_all_le (vector signed char, vector bool char);
31952 int vec_all_le (vector signed char, vector signed char);
31953 int vec_all_le (vector bool short, vector unsigned short);
31954 int vec_all_le (vector unsigned short, vector bool short);
31955 int vec_all_le (vector unsigned short, vector unsigned short);
31956 int vec_all_le (vector bool short, vector signed short);
31957 int vec_all_le (vector signed short, vector bool short);
31958 int vec_all_le (vector signed short, vector signed short);
31959 int vec_all_le (vector bool int, vector unsigned int);
31960 int vec_all_le (vector unsigned int, vector bool int);
31961 int vec_all_le (vector unsigned int, vector unsigned int);
31962 int vec_all_le (vector bool int, vector signed int);
31963 int vec_all_le (vector signed int, vector bool int);
31964 int vec_all_le (vector signed int, vector signed int);
31965 int vec_all_le (vector float, vector float);
31967 int vec_all_lt (vector bool char, vector unsigned char);
31968 int vec_all_lt (vector unsigned char, vector bool char);
31969 int vec_all_lt (vector unsigned char, vector unsigned char);
31970 int vec_all_lt (vector bool char, vector signed char);
31971 int vec_all_lt (vector signed char, vector bool char);
31972 int vec_all_lt (vector signed char, vector signed char);
31973 int vec_all_lt (vector bool short, vector unsigned short);
31974 int vec_all_lt (vector unsigned short, vector bool short);
31975 int vec_all_lt (vector unsigned short, vector unsigned short);
31976 int vec_all_lt (vector bool short, vector signed short);
31977 int vec_all_lt (vector signed short, vector bool short);
31978 int vec_all_lt (vector signed short, vector signed short);
31979 int vec_all_lt (vector bool int, vector unsigned int);
31980 int vec_all_lt (vector unsigned int, vector bool int);
31981 int vec_all_lt (vector unsigned int, vector unsigned int);
31982 int vec_all_lt (vector bool int, vector signed int);
31983 int vec_all_lt (vector signed int, vector bool int);
31984 int vec_all_lt (vector signed int, vector signed int);
31985 int vec_all_lt (vector float, vector float);
31987 int vec_all_nan (vector float);
31989 int vec_all_ne (vector signed char, vector bool char);
31990 int vec_all_ne (vector signed char, vector signed char);
31991 int vec_all_ne (vector unsigned char, vector bool char);
31992 int vec_all_ne (vector unsigned char, vector unsigned char);
31993 int vec_all_ne (vector bool char, vector bool char);
31994 int vec_all_ne (vector bool char, vector unsigned char);
31995 int vec_all_ne (vector bool char, vector signed char);
31996 int vec_all_ne (vector signed short, vector bool short);
31997 int vec_all_ne (vector signed short, vector signed short);
31998 int vec_all_ne (vector unsigned short, vector bool short);
31999 int vec_all_ne (vector unsigned short, vector unsigned short);
32000 int vec_all_ne (vector bool short, vector bool short);
32001 int vec_all_ne (vector bool short, vector unsigned short);
32002 int vec_all_ne (vector bool short, vector signed short);
32003 int vec_all_ne (vector pixel, vector pixel);
32004 int vec_all_ne (vector signed int, vector bool int);
32005 int vec_all_ne (vector signed int, vector signed int);
32006 int vec_all_ne (vector unsigned int, vector bool int);
32007 int vec_all_ne (vector unsigned int, vector unsigned int);
32008 int vec_all_ne (vector bool int, vector bool int);
32009 int vec_all_ne (vector bool int, vector unsigned int);
32010 int vec_all_ne (vector bool int, vector signed int);
32011 int vec_all_ne (vector float, vector float);
32013 int vec_all_nge (vector float, vector float);
32015 int vec_all_ngt (vector float, vector float);
32017 int vec_all_nle (vector float, vector float);
32019 int vec_all_nlt (vector float, vector float);
32021 int vec_all_numeric (vector float);
32023 int vec_any_eq (vector signed char, vector bool char);
32024 int vec_any_eq (vector signed char, vector signed char);
32025 int vec_any_eq (vector unsigned char, vector bool char);
32026 int vec_any_eq (vector unsigned char, vector unsigned char);
32027 int vec_any_eq (vector bool char, vector bool char);
32028 int vec_any_eq (vector bool char, vector unsigned char);
32029 int vec_any_eq (vector bool char, vector signed char);
32030 int vec_any_eq (vector signed short, vector bool short);
32031 int vec_any_eq (vector signed short, vector signed short);
32032 int vec_any_eq (vector unsigned short, vector bool short);
32033 int vec_any_eq (vector unsigned short, vector unsigned short);
32034 int vec_any_eq (vector bool short, vector bool short);
32035 int vec_any_eq (vector bool short, vector unsigned short);
32036 int vec_any_eq (vector bool short, vector signed short);
32037 int vec_any_eq (vector pixel, vector pixel);
32038 int vec_any_eq (vector signed int, vector bool int);
32039 int vec_any_eq (vector signed int, vector signed int);
32040 int vec_any_eq (vector unsigned int, vector bool int);
32041 int vec_any_eq (vector unsigned int, vector unsigned int);
32042 int vec_any_eq (vector bool int, vector bool int);
32043 int vec_any_eq (vector bool int, vector unsigned int);
32044 int vec_any_eq (vector bool int, vector signed int);
32045 int vec_any_eq (vector float, vector float);
32047 int vec_any_ge (vector signed char, vector bool char);
32048 int vec_any_ge (vector unsigned char, vector bool char);
32049 int vec_any_ge (vector unsigned char, vector unsigned char);
32050 int vec_any_ge (vector signed char, vector signed char);
32051 int vec_any_ge (vector bool char, vector unsigned char);
32052 int vec_any_ge (vector bool char, vector signed char);
32053 int vec_any_ge (vector unsigned short, vector bool short);
32054 int vec_any_ge (vector unsigned short, vector unsigned short);
32055 int vec_any_ge (vector signed short, vector signed short);
32056 int vec_any_ge (vector signed short, vector bool short);
32057 int vec_any_ge (vector bool short, vector unsigned short);
32058 int vec_any_ge (vector bool short, vector signed short);
32059 int vec_any_ge (vector signed int, vector bool int);
32060 int vec_any_ge (vector unsigned int, vector bool int);
32061 int vec_any_ge (vector unsigned int, vector unsigned int);
32062 int vec_any_ge (vector signed int, vector signed int);
32063 int vec_any_ge (vector bool int, vector unsigned int);
32064 int vec_any_ge (vector bool int, vector signed int);
32065 int vec_any_ge (vector float, vector float);
32067 int vec_any_gt (vector bool char, vector unsigned char);
32068 int vec_any_gt (vector unsigned char, vector bool char);
32069 int vec_any_gt (vector unsigned char, vector unsigned char);
32070 int vec_any_gt (vector bool char, vector signed char);
32071 int vec_any_gt (vector signed char, vector bool char);
32072 int vec_any_gt (vector signed char, vector signed char);
32073 int vec_any_gt (vector bool short, vector unsigned short);
32074 int vec_any_gt (vector unsigned short, vector bool short);
32075 int vec_any_gt (vector unsigned short, vector unsigned short);
32076 int vec_any_gt (vector bool short, vector signed short);
32077 int vec_any_gt (vector signed short, vector bool short);
32078 int vec_any_gt (vector signed short, vector signed short);
32079 int vec_any_gt (vector bool int, vector unsigned int);
32080 int vec_any_gt (vector unsigned int, vector bool int);
32081 int vec_any_gt (vector unsigned int, vector unsigned int);
32082 int vec_any_gt (vector bool int, vector signed int);
32083 int vec_any_gt (vector signed int, vector bool int);
32084 int vec_any_gt (vector signed int, vector signed int);
32085 int vec_any_gt (vector float, vector float);
32087 int vec_any_le (vector bool char, vector unsigned char);
32088 int vec_any_le (vector unsigned char, vector bool char);
32089 int vec_any_le (vector unsigned char, vector unsigned char);
32090 int vec_any_le (vector bool char, vector signed char);
32091 int vec_any_le (vector signed char, vector bool char);
32092 int vec_any_le (vector signed char, vector signed char);
32093 int vec_any_le (vector bool short, vector unsigned short);
32094 int vec_any_le (vector unsigned short, vector bool short);
32095 int vec_any_le (vector unsigned short, vector unsigned short);
32096 int vec_any_le (vector bool short, vector signed short);
32097 int vec_any_le (vector signed short, vector bool short);
32098 int vec_any_le (vector signed short, vector signed short);
32099 int vec_any_le (vector bool int, vector unsigned int);
32100 int vec_any_le (vector unsigned int, vector bool int);
32101 int vec_any_le (vector unsigned int, vector unsigned int);
32102 int vec_any_le (vector bool int, vector signed int);
32103 int vec_any_le (vector signed int, vector bool int);
32104 int vec_any_le (vector signed int, vector signed int);
32105 int vec_any_le (vector float, vector float);
32107 int vec_any_lt (vector bool char, vector unsigned char);
32108 int vec_any_lt (vector unsigned char, vector bool char);
32109 int vec_any_lt (vector unsigned char, vector unsigned char);
32110 int vec_any_lt (vector bool char, vector signed char);
32111 int vec_any_lt (vector signed char, vector bool char);
32112 int vec_any_lt (vector signed char, vector signed char);
32113 int vec_any_lt (vector bool short, vector unsigned short);
32114 int vec_any_lt (vector unsigned short, vector bool short);
32115 int vec_any_lt (vector unsigned short, vector unsigned short);
32116 int vec_any_lt (vector bool short, vector signed short);
32117 int vec_any_lt (vector signed short, vector bool short);
32118 int vec_any_lt (vector signed short, vector signed short);
32119 int vec_any_lt (vector bool int, vector unsigned int);
32120 int vec_any_lt (vector unsigned int, vector bool int);
32121 int vec_any_lt (vector unsigned int, vector unsigned int);
32122 int vec_any_lt (vector bool int, vector signed int);
32123 int vec_any_lt (vector signed int, vector bool int);
32124 int vec_any_lt (vector signed int, vector signed int);
32125 int vec_any_lt (vector float, vector float);
32127 int vec_any_nan (vector float);
32129 int vec_any_ne (vector signed char, vector bool char);
32130 int vec_any_ne (vector signed char, vector signed char);
32131 int vec_any_ne (vector unsigned char, vector bool char);
32132 int vec_any_ne (vector unsigned char, vector unsigned char);
32133 int vec_any_ne (vector bool char, vector bool char);
32134 int vec_any_ne (vector bool char, vector unsigned char);
32135 int vec_any_ne (vector bool char, vector signed char);
32136 int vec_any_ne (vector signed short, vector bool short);
32137 int vec_any_ne (vector signed short, vector signed short);
32138 int vec_any_ne (vector unsigned short, vector bool short);
32139 int vec_any_ne (vector unsigned short, vector unsigned short);
32140 int vec_any_ne (vector bool short, vector bool short);
32141 int vec_any_ne (vector bool short, vector unsigned short);
32142 int vec_any_ne (vector bool short, vector signed short);
32143 int vec_any_ne (vector pixel, vector pixel);
32144 int vec_any_ne (vector signed int, vector bool int);
32145 int vec_any_ne (vector signed int, vector signed int);
32146 int vec_any_ne (vector unsigned int, vector bool int);
32147 int vec_any_ne (vector unsigned int, vector unsigned int);
32148 int vec_any_ne (vector bool int, vector bool int);
32149 int vec_any_ne (vector bool int, vector unsigned int);
32150 int vec_any_ne (vector bool int, vector signed int);
32151 int vec_any_ne (vector float, vector float);
32153 int vec_any_nge (vector float, vector float);
32155 int vec_any_ngt (vector float, vector float);
32157 int vec_any_nle (vector float, vector float);
32159 int vec_any_nlt (vector float, vector float);
32161 int vec_any_numeric (vector float);
32163 int vec_any_out (vector float, vector float);
32166 File: gcc.info, Node: SPARC VIS Built-in Functions, Next: SPU Built-in Functions, Prev: PowerPC AltiVec Built-in Functions, Up: Target Builtins
32168 5.50.10 SPARC VIS Built-in Functions
32169 ------------------------------------
32171 GCC supports SIMD operations on the SPARC using both the generic vector
32172 extensions (*note Vector Extensions::) as well as built-in functions for
32173 the SPARC Visual Instruction Set (VIS). When you use the `-mvis'
32174 switch, the VIS extension is exposed as the following built-in
32177 typedef int v2si __attribute__ ((vector_size (8)));
32178 typedef short v4hi __attribute__ ((vector_size (8)));
32179 typedef short v2hi __attribute__ ((vector_size (4)));
32180 typedef char v8qi __attribute__ ((vector_size (8)));
32181 typedef char v4qi __attribute__ ((vector_size (4)));
32183 void * __builtin_vis_alignaddr (void *, long);
32184 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
32185 v2si __builtin_vis_faligndatav2si (v2si, v2si);
32186 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
32187 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
32189 v4hi __builtin_vis_fexpand (v4qi);
32191 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
32192 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
32193 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
32194 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
32195 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
32196 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
32197 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
32199 v4qi __builtin_vis_fpack16 (v4hi);
32200 v8qi __builtin_vis_fpack32 (v2si, v2si);
32201 v2hi __builtin_vis_fpackfix (v2si);
32202 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
32204 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
32207 File: gcc.info, Node: SPU Built-in Functions, Prev: SPARC VIS Built-in Functions, Up: Target Builtins
32209 5.50.11 SPU Built-in Functions
32210 ------------------------------
32212 GCC provides extensions for the SPU processor as described in the
32213 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
32214 found at `http://cell.scei.co.jp/' or
32215 `http://www.ibm.com/developerworks/power/cell/'. GCC's implementation
32216 differs in several ways.
32218 * The optional extension of specifying vector constants in
32219 parentheses is not supported.
32221 * A vector initializer requires no cast if the vector constant is of
32222 the same type as the variable it is initializing.
32224 * If `signed' or `unsigned' is omitted, the signedness of the vector
32225 type is the default signedness of the base type. The default
32226 varies depending on the operating system, so a portable program
32227 should always specify the signedness.
32229 * By default, the keyword `__vector' is added. The macro `vector' is
32230 defined in `<spu_intrinsics.h>' and can be undefined.
32232 * GCC allows using a `typedef' name as the type specifier for a
32235 * For C, overloaded functions are implemented with macros so the
32236 following does not work:
32238 spu_add ((vector signed int){1, 2, 3, 4}, foo);
32240 Since `spu_add' is a macro, the vector constant in the example is
32241 treated as four separate arguments. Wrap the entire argument in
32242 parentheses for this to work.
32244 * The extended version of `__builtin_expect' is not supported.
32247 _Note:_ Only the interface described in the aforementioned
32248 specification is supported. Internally, GCC uses built-in functions to
32249 implement the required functionality, but these are not supported and
32250 are subject to change without notice.
32253 File: gcc.info, Node: Target Format Checks, Next: Pragmas, Prev: Target Builtins, Up: C Extensions
32255 5.51 Format Checks Specific to Particular Target Machines
32256 =========================================================
32258 For some target machines, GCC supports additional options to the format
32259 attribute (*note Declaring Attributes of Functions: Function
32264 * Solaris Format Checks::
32267 File: gcc.info, Node: Solaris Format Checks, Up: Target Format Checks
32269 5.51.1 Solaris Format Checks
32270 ----------------------------
32272 Solaris targets support the `cmn_err' (or `__cmn_err__') format check.
32273 `cmn_err' accepts a subset of the standard `printf' conversions, and
32274 the two-argument `%b' conversion for displaying bit-fields. See the
32275 Solaris man page for `cmn_err' for more information.
32278 File: gcc.info, Node: Pragmas, Next: Unnamed Fields, Prev: Target Format Checks, Up: C Extensions
32280 5.52 Pragmas Accepted by GCC
32281 ============================
32283 GCC supports several types of pragmas, primarily in order to compile
32284 code originally written for other compilers. Note that in general we
32285 do not recommend the use of pragmas; *Note Function Attributes::, for
32286 further explanation.
32292 * RS/6000 and PowerPC Pragmas::
32294 * Solaris Pragmas::
32295 * Symbol-Renaming Pragmas::
32296 * Structure-Packing Pragmas::
32298 * Diagnostic Pragmas::
32299 * Visibility Pragmas::
32302 File: gcc.info, Node: ARM Pragmas, Next: M32C Pragmas, Up: Pragmas
32307 The ARM target defines pragmas for controlling the default addition of
32308 `long_call' and `short_call' attributes to functions. *Note Function
32309 Attributes::, for information about the effects of these attributes.
32312 Set all subsequent functions to have the `long_call' attribute.
32315 Set all subsequent functions to have the `short_call' attribute.
32318 Do not affect the `long_call' or `short_call' attributes of
32319 subsequent functions.
32322 File: gcc.info, Node: M32C Pragmas, Next: RS/6000 and PowerPC Pragmas, Prev: ARM Pragmas, Up: Pragmas
32324 5.52.2 M32C Pragmas
32325 -------------------
32328 Overrides the command line option `-memregs=' for the current
32329 file. Use with care! This pragma must be before any function in
32330 the file, and mixing different memregs values in different objects
32331 may make them incompatible. This pragma is useful when a
32332 performance-critical function uses a memreg for temporary values,
32333 as it may allow you to reduce the number of memregs used.
32337 File: gcc.info, Node: RS/6000 and PowerPC Pragmas, Next: Darwin Pragmas, Prev: M32C Pragmas, Up: Pragmas
32339 5.52.3 RS/6000 and PowerPC Pragmas
32340 ----------------------------------
32342 The RS/6000 and PowerPC targets define one pragma for controlling
32343 whether or not the `longcall' attribute is added to function
32344 declarations by default. This pragma overrides the `-mlongcall'
32345 option, but not the `longcall' and `shortcall' attributes. *Note
32346 RS/6000 and PowerPC Options::, for more information about when long
32347 calls are and are not necessary.
32350 Apply the `longcall' attribute to all subsequent function
32354 Do not apply the `longcall' attribute to subsequent function
32358 File: gcc.info, Node: Darwin Pragmas, Next: Solaris Pragmas, Prev: RS/6000 and PowerPC Pragmas, Up: Pragmas
32360 5.52.4 Darwin Pragmas
32361 ---------------------
32363 The following pragmas are available for all architectures running the
32364 Darwin operating system. These are useful for compatibility with other
32368 This pragma is accepted, but has no effect.
32370 `options align=ALIGNMENT'
32371 This pragma sets the alignment of fields in structures. The
32372 values of ALIGNMENT may be `mac68k', to emulate m68k alignment, or
32373 `power', to emulate PowerPC alignment. Uses of this pragma nest
32374 properly; to restore the previous setting, use `reset' for the
32377 `segment TOKENS...'
32378 This pragma is accepted, but has no effect.
32380 `unused (VAR [, VAR]...)'
32381 This pragma declares variables to be possibly unused. GCC will not
32382 produce warnings for the listed variables. The effect is similar
32383 to that of the `unused' attribute, except that this pragma may
32384 appear anywhere within the variables' scopes.
32387 File: gcc.info, Node: Solaris Pragmas, Next: Symbol-Renaming Pragmas, Prev: Darwin Pragmas, Up: Pragmas
32389 5.52.5 Solaris Pragmas
32390 ----------------------
32392 The Solaris target supports `#pragma redefine_extname' (*note
32393 Symbol-Renaming Pragmas::). It also supports additional `#pragma'
32394 directives for compatibility with the system compiler.
32396 `align ALIGNMENT (VARIABLE [, VARIABLE]...)'
32397 Increase the minimum alignment of each VARIABLE to ALIGNMENT.
32398 This is the same as GCC's `aligned' attribute *note Variable
32399 Attributes::). Macro expansion occurs on the arguments to this
32400 pragma when compiling C and Objective-C. It does not currently
32401 occur when compiling C++, but this is a bug which may be fixed in
32404 `fini (FUNCTION [, FUNCTION]...)'
32405 This pragma causes each listed FUNCTION to be called after main,
32406 or during shared module unloading, by adding a call to the `.fini'
32409 `init (FUNCTION [, FUNCTION]...)'
32410 This pragma causes each listed FUNCTION to be called during
32411 initialization (before `main') or during shared module loading, by
32412 adding a call to the `.init' section.
32416 File: gcc.info, Node: Symbol-Renaming Pragmas, Next: Structure-Packing Pragmas, Prev: Solaris Pragmas, Up: Pragmas
32418 5.52.6 Symbol-Renaming Pragmas
32419 ------------------------------
32421 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
32422 supports two `#pragma' directives which change the name used in
32423 assembly for a given declaration. These pragmas are only available on
32424 platforms whose system headers need them. To get this effect on all
32425 platforms supported by GCC, use the asm labels extension (*note Asm
32428 `redefine_extname OLDNAME NEWNAME'
32429 This pragma gives the C function OLDNAME the assembly symbol
32430 NEWNAME. The preprocessor macro `__PRAGMA_REDEFINE_EXTNAME' will
32431 be defined if this pragma is available (currently only on Solaris).
32433 `extern_prefix STRING'
32434 This pragma causes all subsequent external function and variable
32435 declarations to have STRING prepended to their assembly symbols.
32436 This effect may be terminated with another `extern_prefix' pragma
32437 whose argument is an empty string. The preprocessor macro
32438 `__PRAGMA_EXTERN_PREFIX' will be defined if this pragma is
32439 available (currently only on Tru64 UNIX).
32441 These pragmas and the asm labels extension interact in a complicated
32442 manner. Here are some corner cases you may want to be aware of.
32444 1. Both pragmas silently apply only to declarations with external
32445 linkage. Asm labels do not have this restriction.
32447 2. In C++, both pragmas silently apply only to declarations with "C"
32448 linkage. Again, asm labels do not have this restriction.
32450 3. If any of the three ways of changing the assembly name of a
32451 declaration is applied to a declaration whose assembly name has
32452 already been determined (either by a previous use of one of these
32453 features, or because the compiler needed the assembly name in
32454 order to generate code), and the new name is different, a warning
32455 issues and the name does not change.
32457 4. The OLDNAME used by `#pragma redefine_extname' is always the
32460 5. If `#pragma extern_prefix' is in effect, and a declaration occurs
32461 with an asm label attached, the prefix is silently ignored for
32464 6. If `#pragma extern_prefix' and `#pragma redefine_extname' apply to
32465 the same declaration, whichever triggered first wins, and a
32466 warning issues if they contradict each other. (We would like to
32467 have `#pragma redefine_extname' always win, for consistency with
32468 asm labels, but if `#pragma extern_prefix' triggers first we have
32469 no way of knowing that that happened.)
32472 File: gcc.info, Node: Structure-Packing Pragmas, Next: Weak Pragmas, Prev: Symbol-Renaming Pragmas, Up: Pragmas
32474 5.52.7 Structure-Packing Pragmas
32475 --------------------------------
32477 For compatibility with Win32, GCC supports a set of `#pragma'
32478 directives which change the maximum alignment of members of structures
32479 (other than zero-width bitfields), unions, and classes subsequently
32480 defined. The N value below always is required to be a small power of
32481 two and specifies the new alignment in bytes.
32483 1. `#pragma pack(N)' simply sets the new alignment.
32485 2. `#pragma pack()' sets the alignment to the one that was in effect
32486 when compilation started (see also command line option
32487 `-fpack-struct[=<n>]' *note Code Gen Options::).
32489 3. `#pragma pack(push[,N])' pushes the current alignment setting on
32490 an internal stack and then optionally sets the new alignment.
32492 4. `#pragma pack(pop)' restores the alignment setting to the one
32493 saved at the top of the internal stack (and removes that stack
32494 entry). Note that `#pragma pack([N])' does not influence this
32495 internal stack; thus it is possible to have `#pragma pack(push)'
32496 followed by multiple `#pragma pack(N)' instances and finalized by
32497 a single `#pragma pack(pop)'.
32499 Some targets, e.g. i386 and powerpc, support the `ms_struct' `#pragma'
32500 which lays out a structure as the documented `__attribute__
32502 1. `#pragma ms_struct on' turns on the layout for structures declared.
32504 2. `#pragma ms_struct off' turns off the layout for structures
32507 3. `#pragma ms_struct reset' goes back to the default layout.
32510 File: gcc.info, Node: Weak Pragmas, Next: Diagnostic Pragmas, Prev: Structure-Packing Pragmas, Up: Pragmas
32512 5.52.8 Weak Pragmas
32513 -------------------
32515 For compatibility with SVR4, GCC supports a set of `#pragma' directives
32516 for declaring symbols to be weak, and defining weak aliases.
32518 `#pragma weak SYMBOL'
32519 This pragma declares SYMBOL to be weak, as if the declaration had
32520 the attribute of the same name. The pragma may appear before or
32521 after the declaration of SYMBOL, but must appear before either its
32522 first use or its definition. It is not an error for SYMBOL to
32523 never be defined at all.
32525 `#pragma weak SYMBOL1 = SYMBOL2'
32526 This pragma declares SYMBOL1 to be a weak alias of SYMBOL2. It is
32527 an error if SYMBOL2 is not defined in the current translation unit.
32530 File: gcc.info, Node: Diagnostic Pragmas, Next: Visibility Pragmas, Prev: Weak Pragmas, Up: Pragmas
32532 5.52.9 Diagnostic Pragmas
32533 -------------------------
32535 GCC allows the user to selectively enable or disable certain types of
32536 diagnostics, and change the kind of the diagnostic. For example, a
32537 project's policy might require that all sources compile with `-Werror'
32538 but certain files might have exceptions allowing specific types of
32539 warnings. Or, a project might selectively enable diagnostics and treat
32540 them as errors depending on which preprocessor macros are defined.
32542 `#pragma GCC diagnostic KIND OPTION'
32543 Modifies the disposition of a diagnostic. Note that not all
32544 diagnostics are modifiable; at the moment only warnings (normally
32545 controlled by `-W...') can be controlled, and not all of them.
32546 Use `-fdiagnostics-show-option' to determine which diagnostics are
32547 controllable and which option controls them.
32549 KIND is `error' to treat this diagnostic as an error, `warning' to
32550 treat it like a warning (even if `-Werror' is in effect), or
32551 `ignored' if the diagnostic is to be ignored. OPTION is a double
32552 quoted string which matches the command line option.
32554 #pragma GCC diagnostic warning "-Wformat"
32555 #pragma GCC diagnostic error "-Wformat"
32556 #pragma GCC diagnostic ignored "-Wformat"
32558 Note that these pragmas override any command line options. Also,
32559 while it is syntactically valid to put these pragmas anywhere in
32560 your sources, the only supported location for them is before any
32561 data or functions are defined. Doing otherwise may result in
32562 unpredictable results depending on how the optimizer manages your
32563 sources. If the same option is listed multiple times, the last
32564 one specified is the one that is in effect. This pragma is not
32565 intended to be a general purpose replacement for command line
32566 options, but for implementing strict control over project policies.
32570 File: gcc.info, Node: Visibility Pragmas, Prev: Diagnostic Pragmas, Up: Pragmas
32572 5.52.10 Visibility Pragmas
32573 --------------------------
32575 `#pragma GCC visibility push(VISIBILITY)'
32576 `#pragma GCC visibility pop'
32577 This pragma allows the user to set the visibility for multiple
32578 declarations without having to give each a visibility attribute
32579 *Note Function Attributes::, for more information about visibility
32580 and the attribute syntax.
32582 In C++, `#pragma GCC visibility' affects only namespace-scope
32583 declarations. Class members and template specializations are not
32584 affected; if you want to override the visibility for a particular
32585 member or instantiation, you must use an attribute.
32589 File: gcc.info, Node: Unnamed Fields, Next: Thread-Local, Prev: Pragmas, Up: C Extensions
32591 5.53 Unnamed struct/union fields within structs/unions
32592 ======================================================
32594 For compatibility with other compilers, GCC allows you to define a
32595 structure or union that contains, as fields, structures and unions
32596 without names. For example:
32607 In this example, the user would be able to access members of the
32608 unnamed union with code like `foo.b'. Note that only unnamed structs
32609 and unions are allowed, you may not have, for example, an unnamed `int'.
32611 You must never create such structures that cause ambiguous field
32612 definitions. For example, this structure:
32621 It is ambiguous which `a' is being referred to with `foo.a'. Such
32622 constructs are not supported and must be avoided. In the future, such
32623 constructs may be detected and treated as compilation errors.
32625 Unless `-fms-extensions' is used, the unnamed field must be a
32626 structure or union definition without a tag (for example, `struct { int
32627 a; };'). If `-fms-extensions' is used, the field may also be a
32628 definition with a tag such as `struct foo { int a; };', a reference to
32629 a previously defined structure or union such as `struct foo;', or a
32630 reference to a `typedef' name for a previously defined structure or
32634 File: gcc.info, Node: Thread-Local, Next: Binary constants, Prev: Unnamed Fields, Up: C Extensions
32636 5.54 Thread-Local Storage
32637 =========================
32639 Thread-local storage (TLS) is a mechanism by which variables are
32640 allocated such that there is one instance of the variable per extant
32641 thread. The run-time model GCC uses to implement this originates in
32642 the IA-64 processor-specific ABI, but has since been migrated to other
32643 processors as well. It requires significant support from the linker
32644 (`ld'), dynamic linker (`ld.so'), and system libraries (`libc.so' and
32645 `libpthread.so'), so it is not available everywhere.
32647 At the user level, the extension is visible with a new storage class
32648 keyword: `__thread'. For example:
32651 extern __thread struct state s;
32652 static __thread char *p;
32654 The `__thread' specifier may be used alone, with the `extern' or
32655 `static' specifiers, but with no other storage class specifier. When
32656 used with `extern' or `static', `__thread' must appear immediately
32657 after the other storage class specifier.
32659 The `__thread' specifier may be applied to any global, file-scoped
32660 static, function-scoped static, or static data member of a class. It
32661 may not be applied to block-scoped automatic or non-static data member.
32663 When the address-of operator is applied to a thread-local variable, it
32664 is evaluated at run-time and returns the address of the current thread's
32665 instance of that variable. An address so obtained may be used by any
32666 thread. When a thread terminates, any pointers to thread-local
32667 variables in that thread become invalid.
32669 No static initialization may refer to the address of a thread-local
32672 In C++, if an initializer is present for a thread-local variable, it
32673 must be a CONSTANT-EXPRESSION, as defined in 5.19.2 of the ANSI/ISO C++
32676 See ELF Handling For Thread-Local Storage
32677 (http://people.redhat.com/drepper/tls.pdf) for a detailed explanation of
32678 the four thread-local storage addressing models, and how the run-time
32679 is expected to function.
32683 * C99 Thread-Local Edits::
32684 * C++98 Thread-Local Edits::
32687 File: gcc.info, Node: C99 Thread-Local Edits, Next: C++98 Thread-Local Edits, Up: Thread-Local
32689 5.54.1 ISO/IEC 9899:1999 Edits for Thread-Local Storage
32690 -------------------------------------------------------
32692 The following are a set of changes to ISO/IEC 9899:1999 (aka C99) that
32693 document the exact semantics of the language extension.
32695 * `5.1.2 Execution environments'
32697 Add new text after paragraph 1
32699 Within either execution environment, a "thread" is a flow of
32700 control within a program. It is implementation defined
32701 whether or not there may be more than one thread associated
32702 with a program. It is implementation defined how threads
32703 beyond the first are created, the name and type of the
32704 function called at thread startup, and how threads may be
32705 terminated. However, objects with thread storage duration
32706 shall be initialized before thread startup.
32708 * `6.2.4 Storage durations of objects'
32710 Add new text before paragraph 3
32712 An object whose identifier is declared with the storage-class
32713 specifier `__thread' has "thread storage duration". Its
32714 lifetime is the entire execution of the thread, and its
32715 stored value is initialized only once, prior to thread
32722 * `6.7.1 Storage-class specifiers'
32724 Add `__thread' to the list of storage class specifiers in
32727 Change paragraph 2 to
32729 With the exception of `__thread', at most one storage-class
32730 specifier may be given [...]. The `__thread' specifier may
32731 be used alone, or immediately following `extern' or `static'.
32733 Add new text after paragraph 6
32735 The declaration of an identifier for a variable that has
32736 block scope that specifies `__thread' shall also specify
32737 either `extern' or `static'.
32739 The `__thread' specifier shall be used only with variables.
32742 File: gcc.info, Node: C++98 Thread-Local Edits, Prev: C99 Thread-Local Edits, Up: Thread-Local
32744 5.54.2 ISO/IEC 14882:1998 Edits for Thread-Local Storage
32745 --------------------------------------------------------
32747 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
32748 that document the exact semantics of the language extension.
32750 * [intro.execution]
32752 New text after paragraph 4
32754 A "thread" is a flow of control within the abstract machine.
32755 It is implementation defined whether or not there may be more
32758 New text after paragraph 7
32760 It is unspecified whether additional action must be taken to
32761 ensure when and whether side effects are visible to other
32768 * [basic.start.main]
32770 Add after paragraph 5
32772 The thread that begins execution at the `main' function is
32773 called the "main thread". It is implementation defined how
32774 functions beginning threads other than the main thread are
32775 designated or typed. A function so designated, as well as
32776 the `main' function, is called a "thread startup function".
32777 It is implementation defined what happens if a thread startup
32778 function returns. It is implementation defined what happens
32779 to other threads when any thread calls `exit'.
32781 * [basic.start.init]
32783 Add after paragraph 4
32785 The storage for an object of thread storage duration shall be
32786 statically initialized before the first statement of the
32787 thread startup function. An object of thread storage
32788 duration shall not require dynamic initialization.
32790 * [basic.start.term]
32792 Add after paragraph 3
32794 The type of an object with thread storage duration shall not
32795 have a non-trivial destructor, nor shall it be an array type
32796 whose elements (directly or indirectly) have non-trivial
32801 Add "thread storage duration" to the list in paragraph 1.
32805 Thread, static, and automatic storage durations are
32806 associated with objects introduced by declarations [...].
32808 Add `__thread' to the list of specifiers in paragraph 3.
32810 * [basic.stc.thread]
32812 New section before [basic.stc.static]
32814 The keyword `__thread' applied to a non-local object gives the
32815 object thread storage duration.
32817 A local variable or class data member declared both `static'
32818 and `__thread' gives the variable or member thread storage
32821 * [basic.stc.static]
32825 All objects which have neither thread storage duration,
32826 dynamic storage duration nor are local [...].
32830 Add `__thread' to the list in paragraph 1.
32834 With the exception of `__thread', at most one
32835 STORAGE-CLASS-SPECIFIER shall appear in a given
32836 DECL-SPECIFIER-SEQ. The `__thread' specifier may be used
32837 alone, or immediately following the `extern' or `static'
32840 Add after paragraph 5
32842 The `__thread' specifier can be applied only to the names of
32843 objects and to anonymous unions.
32847 Add after paragraph 6
32849 Non-`static' members shall not be `__thread'.
32852 File: gcc.info, Node: Binary constants, Prev: Thread-Local, Up: C Extensions
32854 5.55 Binary constants using the `0b' prefix
32855 ===========================================
32857 Integer constants can be written as binary constants, consisting of a
32858 sequence of `0' and `1' digits, prefixed by `0b' or `0B'. This is
32859 particularly useful in environments that operate a lot on the bit-level
32860 (like microcontrollers).
32862 The following statements are identical:
32869 The type of these constants follows the same rules as for octal or
32870 hexadecimal integer constants, so suffixes like `L' or `UL' can be
32874 File: gcc.info, Node: C++ Extensions, Next: Objective-C, Prev: C Extensions, Up: Top
32876 6 Extensions to the C++ Language
32877 ********************************
32879 The GNU compiler provides these extensions to the C++ language (and you
32880 can also use most of the C language extensions in your C++ programs).
32881 If you want to write code that checks whether these features are
32882 available, you can test for the GNU compiler the same way as for C
32883 programs: check for a predefined macro `__GNUC__'. You can also use
32884 `__GNUG__' to test specifically for GNU C++ (*note Predefined Macros:
32885 (cpp)Common Predefined Macros.).
32889 * Volatiles:: What constitutes an access to a volatile object.
32890 * Restricted Pointers:: C99 restricted pointers and references.
32891 * Vague Linkage:: Where G++ puts inlines, vtables and such.
32892 * C++ Interface:: You can use a single C++ header file for both
32893 declarations and definitions.
32894 * Template Instantiation:: Methods for ensuring that exactly one copy of
32895 each needed template instantiation is emitted.
32896 * Bound member functions:: You can extract a function pointer to the
32897 method denoted by a `->*' or `.*' expression.
32898 * C++ Attributes:: Variable, function, and type attributes for C++ only.
32899 * Namespace Association:: Strong using-directives for namespace association.
32900 * Type Traits:: Compiler support for type traits
32901 * Java Exceptions:: Tweaking exception handling to work with Java.
32902 * Deprecated Features:: Things will disappear from g++.
32903 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
32906 File: gcc.info, Node: Volatiles, Next: Restricted Pointers, Up: C++ Extensions
32908 6.1 When is a Volatile Object Accessed?
32909 =======================================
32911 Both the C and C++ standard have the concept of volatile objects. These
32912 are normally accessed by pointers and used for accessing hardware. The
32913 standards encourage compilers to refrain from optimizations concerning
32914 accesses to volatile objects. The C standard leaves it implementation
32915 defined as to what constitutes a volatile access. The C++ standard
32916 omits to specify this, except to say that C++ should behave in a
32917 similar manner to C with respect to volatiles, where possible. The
32918 minimum either standard specifies is that at a sequence point all
32919 previous accesses to volatile objects have stabilized and no subsequent
32920 accesses have occurred. Thus an implementation is free to reorder and
32921 combine volatile accesses which occur between sequence points, but
32922 cannot do so for accesses across a sequence point. The use of
32923 volatiles does not allow you to violate the restriction on updating
32924 objects multiple times within a sequence point.
32926 *Note Volatile qualifier and the C compiler: Qualifiers implementation.
32928 The behavior differs slightly between C and C++ in the non-obvious
32931 volatile int *src = SOMEVALUE;
32934 With C, such expressions are rvalues, and GCC interprets this either
32935 as a read of the volatile object being pointed to or only as request to
32936 evaluate the side-effects. The C++ standard specifies that such
32937 expressions do not undergo lvalue to rvalue conversion, and that the
32938 type of the dereferenced object may be incomplete. The C++ standard
32939 does not specify explicitly that it is this lvalue to rvalue conversion
32940 which may be responsible for causing an access. However, there is
32941 reason to believe that it is, because otherwise certain simple
32942 expressions become undefined. However, because it would surprise most
32943 programmers, G++ treats dereferencing a pointer to volatile object of
32944 complete type when the value is unused as GCC would do for an
32945 equivalent type in C. When the object has incomplete type, G++ issues
32946 a warning; if you wish to force an error, you must force a conversion
32947 to rvalue with, for instance, a static cast.
32949 When using a reference to volatile, G++ does not treat equivalent
32950 expressions as accesses to volatiles, but instead issues a warning that
32951 no volatile is accessed. The rationale for this is that otherwise it
32952 becomes difficult to determine where volatile access occur, and not
32953 possible to ignore the return value from functions returning volatile
32954 references. Again, if you wish to force a read, cast the reference to
32958 File: gcc.info, Node: Restricted Pointers, Next: Vague Linkage, Prev: Volatiles, Up: C++ Extensions
32960 6.2 Restricting Pointer Aliasing
32961 ================================
32963 As with the C front end, G++ understands the C99 feature of restricted
32964 pointers, specified with the `__restrict__', or `__restrict' type
32965 qualifier. Because you cannot compile C++ by specifying the `-std=c99'
32966 language flag, `restrict' is not a keyword in C++.
32968 In addition to allowing restricted pointers, you can specify restricted
32969 references, which indicate that the reference is not aliased in the
32972 void fn (int *__restrict__ rptr, int &__restrict__ rref)
32977 In the body of `fn', RPTR points to an unaliased integer and RREF
32978 refers to a (different) unaliased integer.
32980 You may also specify whether a member function's THIS pointer is
32981 unaliased by using `__restrict__' as a member function qualifier.
32983 void T::fn () __restrict__
32988 Within the body of `T::fn', THIS will have the effective definition `T
32989 *__restrict__ const this'. Notice that the interpretation of a
32990 `__restrict__' member function qualifier is different to that of
32991 `const' or `volatile' qualifier, in that it is applied to the pointer
32992 rather than the object. This is consistent with other compilers which
32993 implement restricted pointers.
32995 As with all outermost parameter qualifiers, `__restrict__' is ignored
32996 in function definition matching. This means you only need to specify
32997 `__restrict__' in a function definition, rather than in a function
33001 File: gcc.info, Node: Vague Linkage, Next: C++ Interface, Prev: Restricted Pointers, Up: C++ Extensions
33006 There are several constructs in C++ which require space in the object
33007 file but are not clearly tied to a single translation unit. We say that
33008 these constructs have "vague linkage". Typically such constructs are
33009 emitted wherever they are needed, though sometimes we can be more
33013 Inline functions are typically defined in a header file which can
33014 be included in many different compilations. Hopefully they can
33015 usually be inlined, but sometimes an out-of-line copy is
33016 necessary, if the address of the function is taken or if inlining
33017 fails. In general, we emit an out-of-line copy in all translation
33018 units where one is needed. As an exception, we only emit inline
33019 virtual functions with the vtable, since it will always require a
33022 Local static variables and string constants used in an inline
33023 function are also considered to have vague linkage, since they
33024 must be shared between all inlined and out-of-line instances of
33028 C++ virtual functions are implemented in most compilers using a
33029 lookup table, known as a vtable. The vtable contains pointers to
33030 the virtual functions provided by a class, and each object of the
33031 class contains a pointer to its vtable (or vtables, in some
33032 multiple-inheritance situations). If the class declares any
33033 non-inline, non-pure virtual functions, the first one is chosen as
33034 the "key method" for the class, and the vtable is only emitted in
33035 the translation unit where the key method is defined.
33037 _Note:_ If the chosen key method is later defined as inline, the
33038 vtable will still be emitted in every translation unit which
33039 defines it. Make sure that any inline virtuals are declared
33040 inline in the class body, even if they are not defined there.
33043 C++ requires information about types to be written out in order to
33044 implement `dynamic_cast', `typeid' and exception handling. For
33045 polymorphic classes (classes with virtual functions), the type_info
33046 object is written out along with the vtable so that `dynamic_cast'
33047 can determine the dynamic type of a class object at runtime. For
33048 all other types, we write out the type_info object when it is
33049 used: when applying `typeid' to an expression, throwing an object,
33050 or referring to a type in a catch clause or exception
33053 Template Instantiations
33054 Most everything in this section also applies to template
33055 instantiations, but there are other options as well. *Note
33056 Where's the Template?: Template Instantiation.
33059 When used with GNU ld version 2.8 or later on an ELF system such as
33060 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
33061 these constructs will be discarded at link time. This is known as
33064 On targets that don't support COMDAT, but do support weak symbols, GCC
33065 will use them. This way one copy will override all the others, but the
33066 unused copies will still take up space in the executable.
33068 For targets which do not support either COMDAT or weak symbols, most
33069 entities with vague linkage will be emitted as local symbols to avoid
33070 duplicate definition errors from the linker. This will not happen for
33071 local statics in inlines, however, as having multiple copies will
33072 almost certainly break things.
33074 *Note Declarations and Definitions in One Header: C++ Interface, for
33075 another way to control placement of these constructs.
33078 File: gcc.info, Node: C++ Interface, Next: Template Instantiation, Prev: Vague Linkage, Up: C++ Extensions
33080 6.4 #pragma interface and implementation
33081 ========================================
33083 `#pragma interface' and `#pragma implementation' provide the user with
33084 a way of explicitly directing the compiler to emit entities with vague
33085 linkage (and debugging information) in a particular translation unit.
33087 _Note:_ As of GCC 2.7.2, these `#pragma's are not useful in most
33088 cases, because of COMDAT support and the "key method" heuristic
33089 mentioned in *Note Vague Linkage::. Using them can actually cause your
33090 program to grow due to unnecessary out-of-line copies of inline
33091 functions. Currently (3.4) the only benefit of these `#pragma's is
33092 reduced duplication of debugging information, and that should be
33093 addressed soon on DWARF 2 targets with the use of COMDAT groups.
33095 `#pragma interface'
33096 `#pragma interface "SUBDIR/OBJECTS.h"'
33097 Use this directive in _header files_ that define object classes,
33098 to save space in most of the object files that use those classes.
33099 Normally, local copies of certain information (backup copies of
33100 inline member functions, debugging information, and the internal
33101 tables that implement virtual functions) must be kept in each
33102 object file that includes class definitions. You can use this
33103 pragma to avoid such duplication. When a header file containing
33104 `#pragma interface' is included in a compilation, this auxiliary
33105 information will not be generated (unless the main input source
33106 file itself uses `#pragma implementation'). Instead, the object
33107 files will contain references to be resolved at link time.
33109 The second form of this directive is useful for the case where you
33110 have multiple headers with the same name in different directories.
33111 If you use this form, you must specify the same string to `#pragma
33114 `#pragma implementation'
33115 `#pragma implementation "OBJECTS.h"'
33116 Use this pragma in a _main input file_, when you want full output
33117 from included header files to be generated (and made globally
33118 visible). The included header file, in turn, should use `#pragma
33119 interface'. Backup copies of inline member functions, debugging
33120 information, and the internal tables used to implement virtual
33121 functions are all generated in implementation files.
33123 If you use `#pragma implementation' with no argument, it applies to
33124 an include file with the same basename(1) as your source file.
33125 For example, in `allclass.cc', giving just `#pragma implementation'
33126 by itself is equivalent to `#pragma implementation "allclass.h"'.
33128 In versions of GNU C++ prior to 2.6.0 `allclass.h' was treated as
33129 an implementation file whenever you would include it from
33130 `allclass.cc' even if you never specified `#pragma
33131 implementation'. This was deemed to be more trouble than it was
33132 worth, however, and disabled.
33134 Use the string argument if you want a single implementation file to
33135 include code from multiple header files. (You must also use
33136 `#include' to include the header file; `#pragma implementation'
33137 only specifies how to use the file--it doesn't actually include
33140 There is no way to split up the contents of a single header file
33141 into multiple implementation files.
33143 `#pragma implementation' and `#pragma interface' also have an effect
33144 on function inlining.
33146 If you define a class in a header file marked with `#pragma
33147 interface', the effect on an inline function defined in that class is
33148 similar to an explicit `extern' declaration--the compiler emits no code
33149 at all to define an independent version of the function. Its
33150 definition is used only for inlining with its callers.
33152 Conversely, when you include the same header file in a main source file
33153 that declares it as `#pragma implementation', the compiler emits code
33154 for the function itself; this defines a version of the function that
33155 can be found via pointers (or by callers compiled without inlining).
33156 If all calls to the function can be inlined, you can avoid emitting the
33157 function by compiling with `-fno-implement-inlines'. If any calls were
33158 not inlined, you will get linker errors.
33160 ---------- Footnotes ----------
33162 (1) A file's "basename" was the name stripped of all leading path
33163 information and of trailing suffixes, such as `.h' or `.C' or `.cc'.
33166 File: gcc.info, Node: Template Instantiation, Next: Bound member functions, Prev: C++ Interface, Up: C++ Extensions
33168 6.5 Where's the Template?
33169 =========================
33171 C++ templates are the first language feature to require more
33172 intelligence from the environment than one usually finds on a UNIX
33173 system. Somehow the compiler and linker have to make sure that each
33174 template instance occurs exactly once in the executable if it is needed,
33175 and not at all otherwise. There are two basic approaches to this
33176 problem, which are referred to as the Borland model and the Cfront
33180 Borland C++ solved the template instantiation problem by adding
33181 the code equivalent of common blocks to their linker; the compiler
33182 emits template instances in each translation unit that uses them,
33183 and the linker collapses them together. The advantage of this
33184 model is that the linker only has to consider the object files
33185 themselves; there is no external complexity to worry about. This
33186 disadvantage is that compilation time is increased because the
33187 template code is being compiled repeatedly. Code written for this
33188 model tends to include definitions of all templates in the header
33189 file, since they must be seen to be instantiated.
33192 The AT&T C++ translator, Cfront, solved the template instantiation
33193 problem by creating the notion of a template repository, an
33194 automatically maintained place where template instances are
33195 stored. A more modern version of the repository works as follows:
33196 As individual object files are built, the compiler places any
33197 template definitions and instantiations encountered in the
33198 repository. At link time, the link wrapper adds in the objects in
33199 the repository and compiles any needed instances that were not
33200 previously emitted. The advantages of this model are more optimal
33201 compilation speed and the ability to use the system linker; to
33202 implement the Borland model a compiler vendor also needs to
33203 replace the linker. The disadvantages are vastly increased
33204 complexity, and thus potential for error; for some code this can be
33205 just as transparent, but in practice it can been very difficult to
33206 build multiple programs in one directory and one program in
33207 multiple directories. Code written for this model tends to
33208 separate definitions of non-inline member templates into a
33209 separate file, which should be compiled separately.
33211 When used with GNU ld version 2.8 or later on an ELF system such as
33212 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
33213 Borland model. On other systems, G++ implements neither automatic
33216 A future version of G++ will support a hybrid model whereby the
33217 compiler will emit any instantiations for which the template definition
33218 is included in the compile, and store template definitions and
33219 instantiation context information into the object file for the rest.
33220 The link wrapper will extract that information as necessary and invoke
33221 the compiler to produce the remaining instantiations. The linker will
33222 then combine duplicate instantiations.
33224 In the mean time, you have the following options for dealing with
33225 template instantiations:
33227 1. Compile your template-using code with `-frepo'. The compiler will
33228 generate files with the extension `.rpo' listing all of the
33229 template instantiations used in the corresponding object files
33230 which could be instantiated there; the link wrapper, `collect2',
33231 will then update the `.rpo' files to tell the compiler where to
33232 place those instantiations and rebuild any affected object files.
33233 The link-time overhead is negligible after the first pass, as the
33234 compiler will continue to place the instantiations in the same
33237 This is your best option for application code written for the
33238 Borland model, as it will just work. Code written for the Cfront
33239 model will need to be modified so that the template definitions
33240 are available at one or more points of instantiation; usually this
33241 is as simple as adding `#include <tmethods.cc>' to the end of each
33244 For library code, if you want the library to provide all of the
33245 template instantiations it needs, just try to link all of its
33246 object files together; the link will fail, but cause the
33247 instantiations to be generated as a side effect. Be warned,
33248 however, that this may cause conflicts if multiple libraries try
33249 to provide the same instantiations. For greater control, use
33250 explicit instantiation as described in the next option.
33252 2. Compile your code with `-fno-implicit-templates' to disable the
33253 implicit generation of template instances, and explicitly
33254 instantiate all the ones you use. This approach requires more
33255 knowledge of exactly which instances you need than do the others,
33256 but it's less mysterious and allows greater control. You can
33257 scatter the explicit instantiations throughout your program,
33258 perhaps putting them in the translation units where the instances
33259 are used or the translation units that define the templates
33260 themselves; you can put all of the explicit instantiations you
33261 need into one big file; or you can create small files like
33266 template class Foo<int>;
33267 template ostream& operator <<
33268 (ostream&, const Foo<int>&);
33270 for each of the instances you need, and create a template
33271 instantiation library from those.
33273 If you are using Cfront-model code, you can probably get away with
33274 not using `-fno-implicit-templates' when compiling files that don't
33275 `#include' the member template definitions.
33277 If you use one big file to do the instantiations, you may want to
33278 compile it without `-fno-implicit-templates' so you get all of the
33279 instances required by your explicit instantiations (but not by any
33280 other files) without having to specify them as well.
33282 G++ has extended the template instantiation syntax given in the ISO
33283 standard to allow forward declaration of explicit instantiations
33284 (with `extern'), instantiation of the compiler support data for a
33285 template class (i.e. the vtable) without instantiating any of its
33286 members (with `inline'), and instantiation of only the static data
33287 members of a template class, without the support data or member
33288 functions (with (`static'):
33290 extern template int max (int, int);
33291 inline template class Foo<int>;
33292 static template class Foo<int>;
33294 3. Do nothing. Pretend G++ does implement automatic instantiation
33295 management. Code written for the Borland model will work fine, but
33296 each translation unit will contain instances of each of the
33297 templates it uses. In a large program, this can lead to an
33298 unacceptable amount of code duplication.
33301 File: gcc.info, Node: Bound member functions, Next: C++ Attributes, Prev: Template Instantiation, Up: C++ Extensions
33303 6.6 Extracting the function pointer from a bound pointer to member function
33304 ===========================================================================
33306 In C++, pointer to member functions (PMFs) are implemented using a wide
33307 pointer of sorts to handle all the possible call mechanisms; the PMF
33308 needs to store information about how to adjust the `this' pointer, and
33309 if the function pointed to is virtual, where to find the vtable, and
33310 where in the vtable to look for the member function. If you are using
33311 PMFs in an inner loop, you should really reconsider that decision. If
33312 that is not an option, you can extract the pointer to the function that
33313 would be called for a given object/PMF pair and call it directly inside
33314 the inner loop, to save a bit of time.
33316 Note that you will still be paying the penalty for the call through a
33317 function pointer; on most modern architectures, such a call defeats the
33318 branch prediction features of the CPU. This is also true of normal
33319 virtual function calls.
33321 The syntax for this extension is
33324 extern int (A::*fp)();
33325 typedef int (*fptr)(A *);
33327 fptr p = (fptr)(a.*fp);
33329 For PMF constants (i.e. expressions of the form `&Klasse::Member'), no
33330 object is needed to obtain the address of the function. They can be
33331 converted to function pointers directly:
33333 fptr p1 = (fptr)(&A::foo);
33335 You must specify `-Wno-pmf-conversions' to use this extension.
33338 File: gcc.info, Node: C++ Attributes, Next: Namespace Association, Prev: Bound member functions, Up: C++ Extensions
33340 6.7 C++-Specific Variable, Function, and Type Attributes
33341 ========================================================
33343 Some attributes only make sense for C++ programs.
33345 `init_priority (PRIORITY)'
33346 In Standard C++, objects defined at namespace scope are guaranteed
33347 to be initialized in an order in strict accordance with that of
33348 their definitions _in a given translation unit_. No guarantee is
33349 made for initializations across translation units. However, GNU
33350 C++ allows users to control the order of initialization of objects
33351 defined at namespace scope with the `init_priority' attribute by
33352 specifying a relative PRIORITY, a constant integral expression
33353 currently bounded between 101 and 65535 inclusive. Lower numbers
33354 indicate a higher priority.
33356 In the following example, `A' would normally be created before
33357 `B', but the `init_priority' attribute has reversed that order:
33359 Some_Class A __attribute__ ((init_priority (2000)));
33360 Some_Class B __attribute__ ((init_priority (543)));
33362 Note that the particular values of PRIORITY do not matter; only
33363 their relative ordering.
33366 This type attribute informs C++ that the class is a Java
33367 interface. It may only be applied to classes declared within an
33368 `extern "Java"' block. Calls to methods declared in this
33369 interface will be dispatched using GCJ's interface table
33370 mechanism, instead of regular virtual table dispatch.
33373 See also *Note Namespace Association::.
33376 File: gcc.info, Node: Namespace Association, Next: Type Traits, Prev: C++ Attributes, Up: C++ Extensions
33378 6.8 Namespace Association
33379 =========================
33381 *Caution:* The semantics of this extension are not fully defined.
33382 Users should refrain from using this extension as its semantics may
33383 change subtly over time. It is possible that this extension will be
33384 removed in future versions of G++.
33386 A using-directive with `__attribute ((strong))' is stronger than a
33387 normal using-directive in two ways:
33389 * Templates from the used namespace can be specialized and explicitly
33390 instantiated as though they were members of the using namespace.
33392 * The using namespace is considered an associated namespace of all
33393 templates in the used namespace for purposes of argument-dependent
33396 The used namespace must be nested within the using namespace so that
33397 normal unqualified lookup works properly.
33399 This is useful for composing a namespace transparently from
33400 implementation namespaces. For example:
33404 template <class T> struct A { };
33406 using namespace debug __attribute ((__strong__));
33407 template <> struct A<int> { }; // ok to specialize
33409 template <class T> void f (A<T>);
33414 f (std::A<float>()); // lookup finds std::f
33419 File: gcc.info, Node: Type Traits, Next: Java Exceptions, Prev: Namespace Association, Up: C++ Extensions
33424 The C++ front-end implements syntactic extensions that allow to
33425 determine at compile time various characteristics of a type (or of a
33428 `__has_nothrow_assign (type)'
33429 If `type' is const qualified or is a reference type then the trait
33430 is false. Otherwise if `__has_trivial_assign (type)' is true then
33431 the trait is true, else if `type' is a cv class or union type with
33432 copy assignment operators that are known not to throw an exception
33433 then the trait is true, else it is false. Requires: `type' shall
33434 be a complete type, an array type of unknown bound, or is a `void'
33437 `__has_nothrow_copy (type)'
33438 If `__has_trivial_copy (type)' is true then the trait is true,
33439 else if `type' is a cv class or union type with copy constructors
33440 that are known not to throw an exception then the trait is true,
33441 else it is false. Requires: `type' shall be a complete type, an
33442 array type of unknown bound, or is a `void' type.
33444 `__has_nothrow_constructor (type)'
33445 If `__has_trivial_constructor (type)' is true then the trait is
33446 true, else if `type' is a cv class or union type (or array
33447 thereof) with a default constructor that is known not to throw an
33448 exception then the trait is true, else it is false. Requires:
33449 `type' shall be a complete type, an array type of unknown bound,
33450 or is a `void' type.
33452 `__has_trivial_assign (type)'
33453 If `type' is const qualified or is a reference type then the trait
33454 is false. Otherwise if `__is_pod (type)' is true then the trait is
33455 true, else if `type' is a cv class or union type with a trivial
33456 copy assignment ([class.copy]) then the trait is true, else it is
33457 false. Requires: `type' shall be a complete type, an array type
33458 of unknown bound, or is a `void' type.
33460 `__has_trivial_copy (type)'
33461 If `__is_pod (type)' is true or `type' is a reference type then
33462 the trait is true, else if `type' is a cv class or union type with
33463 a trivial copy constructor ([class.copy]) then the trait is true,
33464 else it is false. Requires: `type' shall be a complete type, an
33465 array type of unknown bound, or is a `void' type.
33467 `__has_trivial_constructor (type)'
33468 If `__is_pod (type)' is true then the trait is true, else if
33469 `type' is a cv class or union type (or array thereof) with a
33470 trivial default constructor ([class.ctor]) then the trait is true,
33471 else it is false. Requires: `type' shall be a complete type, an
33472 array type of unknown bound, or is a `void' type.
33474 `__has_trivial_destructor (type)'
33475 If `__is_pod (type)' is true or `type' is a reference type then
33476 the trait is true, else if `type' is a cv class or union type (or
33477 array thereof) with a trivial destructor ([class.dtor]) then the
33478 trait is true, else it is false. Requires: `type' shall be a
33479 complete type, an array type of unknown bound, or is a `void' type.
33481 `__has_virtual_destructor (type)'
33482 If `type' is a class type with a virtual destructor ([class.dtor])
33483 then the trait is true, else it is false. Requires: `type' shall
33484 be a complete type, an array type of unknown bound, or is a `void'
33487 `__is_abstract (type)'
33488 If `type' is an abstract class ([class.abstract]) then the trait
33489 is true, else it is false. Requires: `type' shall be a complete
33490 type, an array type of unknown bound, or is a `void' type.
33492 `__is_base_of (base_type, derived_type)'
33493 If `base_type' is a base class of `derived_type' ([class.derived])
33494 then the trait is true, otherwise it is false. Top-level cv
33495 qualifications of `base_type' and `derived_type' are ignored. For
33496 the purposes of this trait, a class type is considered is own
33497 base. Requires: if `__is_class (base_type)' and `__is_class
33498 (derived_type)' are true and `base_type' and `derived_type' are
33499 not the same type (disregarding cv-qualifiers), `derived_type'
33500 shall be a complete type. Diagnostic is produced if this
33501 requirement is not met.
33503 `__is_class (type)'
33504 If `type' is a cv class type, and not a union type
33505 ([basic.compound]) the the trait is true, else it is false.
33507 `__is_empty (type)'
33508 If `__is_class (type)' is false then the trait is false.
33509 Otherwise `type' is considered empty if and only if: `type' has no
33510 non-static data members, or all non-static data members, if any,
33511 are bit-fields of lenght 0, and `type' has no virtual members, and
33512 `type' has no virtual base classes, and `type' has no base classes
33513 `base_type' for which `__is_empty (base_type)' is false.
33514 Requires: `type' shall be a complete type, an array type of
33515 unknown bound, or is a `void' type.
33518 If `type' is a cv enumeration type ([basic.compound]) the the
33519 trait is true, else it is false.
33522 If `type' is a cv POD type ([basic.types]) then the trait is true,
33523 else it is false. Requires: `type' shall be a complete type, an
33524 array type of unknown bound, or is a `void' type.
33526 `__is_polymorphic (type)'
33527 If `type' is a polymorphic class ([class.virtual]) then the trait
33528 is true, else it is false. Requires: `type' shall be a complete
33529 type, an array type of unknown bound, or is a `void' type.
33531 `__is_union (type)'
33532 If `type' is a cv union type ([basic.compound]) the the trait is
33533 true, else it is false.
33537 File: gcc.info, Node: Java Exceptions, Next: Deprecated Features, Prev: Type Traits, Up: C++ Extensions
33539 6.10 Java Exceptions
33540 ====================
33542 The Java language uses a slightly different exception handling model
33543 from C++. Normally, GNU C++ will automatically detect when you are
33544 writing C++ code that uses Java exceptions, and handle them
33545 appropriately. However, if C++ code only needs to execute destructors
33546 when Java exceptions are thrown through it, GCC will guess incorrectly.
33547 Sample problematic code is:
33549 struct S { ~S(); };
33550 extern void bar(); // is written in Java, and may throw exceptions
33557 The usual effect of an incorrect guess is a link failure, complaining of
33558 a missing routine called `__gxx_personality_v0'.
33560 You can inform the compiler that Java exceptions are to be used in a
33561 translation unit, irrespective of what it might think, by writing
33562 `#pragma GCC java_exceptions' at the head of the file. This `#pragma'
33563 must appear before any functions that throw or catch exceptions, or run
33564 destructors when exceptions are thrown through them.
33566 You cannot mix Java and C++ exceptions in the same translation unit.
33567 It is believed to be safe to throw a C++ exception from one file through
33568 another file compiled for the Java exception model, or vice versa, but
33569 there may be bugs in this area.
33572 File: gcc.info, Node: Deprecated Features, Next: Backwards Compatibility, Prev: Java Exceptions, Up: C++ Extensions
33574 6.11 Deprecated Features
33575 ========================
33577 In the past, the GNU C++ compiler was extended to experiment with new
33578 features, at a time when the C++ language was still evolving. Now that
33579 the C++ standard is complete, some of those features are superseded by
33580 superior alternatives. Using the old features might cause a warning in
33581 some cases that the feature will be dropped in the future. In other
33582 cases, the feature might be gone already.
33584 While the list below is not exhaustive, it documents some of the
33585 options that are now deprecated:
33587 `-fexternal-templates'
33588 `-falt-external-templates'
33589 These are two of the many ways for G++ to implement template
33590 instantiation. *Note Template Instantiation::. The C++ standard
33591 clearly defines how template definitions have to be organized
33592 across implementation units. G++ has an implicit instantiation
33593 mechanism that should work just fine for standard-conforming code.
33595 `-fstrict-prototype'
33596 `-fno-strict-prototype'
33597 Previously it was possible to use an empty prototype parameter
33598 list to indicate an unspecified number of parameters (like C),
33599 rather than no parameters, as C++ demands. This feature has been
33600 removed, except where it is required for backwards compatibility
33601 *Note Backwards Compatibility::.
33603 G++ allows a virtual function returning `void *' to be overridden by
33604 one returning a different pointer type. This extension to the
33605 covariant return type rules is now deprecated and will be removed from a
33608 The G++ minimum and maximum operators (`<?' and `>?') and their
33609 compound forms (`<?=') and `>?=') have been deprecated and are now
33610 removed from G++. Code using these operators should be modified to use
33611 `std::min' and `std::max' instead.
33613 The named return value extension has been deprecated, and is now
33616 The use of initializer lists with new expressions has been deprecated,
33617 and is now removed from G++.
33619 Floating and complex non-type template parameters have been deprecated,
33620 and are now removed from G++.
33622 The implicit typename extension has been deprecated and is now removed
33625 The use of default arguments in function pointers, function typedefs
33626 and other places where they are not permitted by the standard is
33627 deprecated and will be removed from a future version of G++.
33629 G++ allows floating-point literals to appear in integral constant
33630 expressions, e.g. ` enum E { e = int(2.2 * 3.7) } ' This extension is
33631 deprecated and will be removed from a future version.
33633 G++ allows static data members of const floating-point type to be
33634 declared with an initializer in a class definition. The standard only
33635 allows initializers for static members of const integral types and const
33636 enumeration types so this extension has been deprecated and will be
33637 removed from a future version.
33640 File: gcc.info, Node: Backwards Compatibility, Prev: Deprecated Features, Up: C++ Extensions
33642 6.12 Backwards Compatibility
33643 ============================
33645 Now that there is a definitive ISO standard C++, G++ has a specification
33646 to adhere to. The C++ language evolved over time, and features that
33647 used to be acceptable in previous drafts of the standard, such as the
33648 ARM [Annotated C++ Reference Manual], are no longer accepted. In order
33649 to allow compilation of C++ written to such drafts, G++ contains some
33650 backwards compatibilities. _All such backwards compatibility features
33651 are liable to disappear in future versions of G++._ They should be
33652 considered deprecated *Note Deprecated Features::.
33655 If a variable is declared at for scope, it used to remain in scope
33656 until the end of the scope which contained the for statement
33657 (rather than just within the for scope). G++ retains this, but
33658 issues a warning, if such a variable is accessed outside the for
33661 `Implicit C language'
33662 Old C system header files did not contain an `extern "C" {...}'
33663 scope to set the language. On such systems, all header files are
33664 implicitly scoped inside a C language scope. Also, an empty
33665 prototype `()' will be treated as an unspecified number of
33666 arguments, rather than no arguments, as C++ demands.
33669 File: gcc.info, Node: Objective-C, Next: Compatibility, Prev: C++ Extensions, Up: Top
33671 7 GNU Objective-C runtime features
33672 **********************************
33674 This document is meant to describe some of the GNU Objective-C runtime
33675 features. It is not intended to teach you Objective-C, there are
33676 several resources on the Internet that present the language. Questions
33677 and comments about this document to Ovidiu Predescu <ovidiu@cup.hp.com>.
33681 * Executing code before main::
33683 * Garbage Collection::
33684 * Constant string objects::
33685 * compatibility_alias::
33688 File: gcc.info, Node: Executing code before main, Next: Type encoding, Prev: Objective-C, Up: Objective-C
33690 7.1 `+load': Executing code before main
33691 =======================================
33693 The GNU Objective-C runtime provides a way that allows you to execute
33694 code before the execution of the program enters the `main' function.
33695 The code is executed on a per-class and a per-category basis, through a
33696 special class method `+load'.
33698 This facility is very useful if you want to initialize global variables
33699 which can be accessed by the program directly, without sending a message
33700 to the class first. The usual way to initialize global variables, in
33701 the `+initialize' method, might not be useful because `+initialize' is
33702 only called when the first message is sent to a class object, which in
33703 some cases could be too late.
33705 Suppose for example you have a `FileStream' class that declares
33706 `Stdin', `Stdout' and `Stderr' as global variables, like below:
33709 FileStream *Stdin = nil;
33710 FileStream *Stdout = nil;
33711 FileStream *Stderr = nil;
33713 @implementation FileStream
33717 Stdin = [[FileStream new] initWithFd:0];
33718 Stdout = [[FileStream new] initWithFd:1];
33719 Stderr = [[FileStream new] initWithFd:2];
33722 /* Other methods here */
33725 In this example, the initialization of `Stdin', `Stdout' and `Stderr'
33726 in `+initialize' occurs too late. The programmer can send a message to
33727 one of these objects before the variables are actually initialized,
33728 thus sending messages to the `nil' object. The `+initialize' method
33729 which actually initializes the global variables is not invoked until
33730 the first message is sent to the class object. The solution would
33731 require these variables to be initialized just before entering `main'.
33733 The correct solution of the above problem is to use the `+load' method
33734 instead of `+initialize':
33737 @implementation FileStream
33741 Stdin = [[FileStream new] initWithFd:0];
33742 Stdout = [[FileStream new] initWithFd:1];
33743 Stderr = [[FileStream new] initWithFd:2];
33746 /* Other methods here */
33749 The `+load' is a method that is not overridden by categories. If a
33750 class and a category of it both implement `+load', both methods are
33751 invoked. This allows some additional initializations to be performed in
33754 This mechanism is not intended to be a replacement for `+initialize'.
33755 You should be aware of its limitations when you decide to use it
33756 instead of `+initialize'.
33760 * What you can and what you cannot do in +load::
33763 File: gcc.info, Node: What you can and what you cannot do in +load, Prev: Executing code before main, Up: Executing code before main
33765 7.1.1 What you can and what you cannot do in `+load'
33766 ----------------------------------------------------
33768 The `+load' implementation in the GNU runtime guarantees you the
33771 * you can write whatever C code you like;
33773 * you can send messages to Objective-C constant strings (`@"this is a
33774 constant string"');
33776 * you can allocate and send messages to objects whose class is
33777 implemented in the same file;
33779 * the `+load' implementation of all super classes of a class are
33780 executed before the `+load' of that class is executed;
33782 * the `+load' implementation of a class is executed before the
33783 `+load' implementation of any category.
33786 In particular, the following things, even if they can work in a
33787 particular case, are not guaranteed:
33789 * allocation of or sending messages to arbitrary objects;
33791 * allocation of or sending messages to objects whose classes have a
33792 category implemented in the same file;
33795 You should make no assumptions about receiving `+load' in sibling
33796 classes when you write `+load' of a class. The order in which sibling
33797 classes receive `+load' is not guaranteed.
33799 The order in which `+load' and `+initialize' are called could be
33800 problematic if this matters. If you don't allocate objects inside
33801 `+load', it is guaranteed that `+load' is called before `+initialize'.
33802 If you create an object inside `+load' the `+initialize' method of
33803 object's class is invoked even if `+load' was not invoked. Note if you
33804 explicitly call `+load' on a class, `+initialize' will be called first.
33805 To avoid possible problems try to implement only one of these methods.
33807 The `+load' method is also invoked when a bundle is dynamically loaded
33808 into your running program. This happens automatically without any
33809 intervening operation from you. When you write bundles and you need to
33810 write `+load' you can safely create and send messages to objects whose
33811 classes already exist in the running program. The same restrictions as
33812 above apply to classes defined in bundle.
33815 File: gcc.info, Node: Type encoding, Next: Garbage Collection, Prev: Executing code before main, Up: Objective-C
33820 The Objective-C compiler generates type encodings for all the types.
33821 These type encodings are used at runtime to find out information about
33822 selectors and methods and about objects and classes.
33824 The types are encoded in the following way:
33828 `unsigned char' `C'
33830 `unsigned short' `S'
33834 `unsigned long' `L'
33846 Complex types `j' followed by the inner type. For example
33847 `_Complex double' is encoded as "jd".
33848 bit-fields `b' followed by the starting position of the
33849 bit-field, the type of the bit-field and the size of
33850 the bit-field (the bit-fields encoding was changed
33851 from the NeXT's compiler encoding, see below)
33853 The encoding of bit-fields has changed to allow bit-fields to be
33854 properly handled by the runtime functions that compute sizes and
33855 alignments of types that contain bit-fields. The previous encoding
33856 contained only the size of the bit-field. Using only this information
33857 it is not possible to reliably compute the size occupied by the
33858 bit-field. This is very important in the presence of the Boehm's
33859 garbage collector because the objects are allocated using the typed
33860 memory facility available in this collector. The typed memory
33861 allocation requires information about where the pointers are located
33864 The position in the bit-field is the position, counting in bits, of the
33865 bit closest to the beginning of the structure.
33867 The non-atomic types are encoded as follows:
33869 pointers `^' followed by the pointed type.
33870 arrays `[' followed by the number of elements in the array
33871 followed by the type of the elements followed by `]'
33872 structures `{' followed by the name of the structure (or `?' if the
33873 structure is unnamed), the `=' sign, the type of the
33875 unions `(' followed by the name of the structure (or `?' if the
33876 union is unnamed), the `=' sign, the type of the members
33879 Here are some types and their encodings, as they are generated by the
33880 compiler on an i386 machine:
33883 Objective-C type Compiler encoding
33885 struct { `{?=i[3f]b128i3b131i2c}'
33894 In addition to the types the compiler also encodes the type
33895 specifiers. The table below describes the encoding of the current
33896 Objective-C type specifiers:
33908 The type specifiers are encoded just before the type. Unlike types
33909 however, the type specifiers are only encoded when they appear in method
33913 File: gcc.info, Node: Garbage Collection, Next: Constant string objects, Prev: Type encoding, Up: Objective-C
33915 7.3 Garbage Collection
33916 ======================
33918 Support for a new memory management policy has been added by using a
33919 powerful conservative garbage collector, known as the
33920 Boehm-Demers-Weiser conservative garbage collector. It is available
33921 from `http://www.hpl.hp.com/personal/Hans_Boehm/gc/'.
33923 To enable the support for it you have to configure the compiler using
33924 an additional argument, `--enable-objc-gc'. You need to have garbage
33925 collector installed before building the compiler. This will build an
33926 additional runtime library which has several enhancements to support
33927 the garbage collector. The new library has a new name, `libobjc_gc.a'
33928 to not conflict with the non-garbage-collected library.
33930 When the garbage collector is used, the objects are allocated using the
33931 so-called typed memory allocation mechanism available in the
33932 Boehm-Demers-Weiser collector. This mode requires precise information
33933 on where pointers are located inside objects. This information is
33934 computed once per class, immediately after the class has been
33937 There is a new runtime function `class_ivar_set_gcinvisible()' which
33938 can be used to declare a so-called "weak pointer" reference. Such a
33939 pointer is basically hidden for the garbage collector; this can be
33940 useful in certain situations, especially when you want to keep track of
33941 the allocated objects, yet allow them to be collected. This kind of
33942 pointers can only be members of objects, you cannot declare a global
33943 pointer as a weak reference. Every type which is a pointer type can be
33944 declared a weak pointer, including `id', `Class' and `SEL'.
33946 Here is an example of how to use this feature. Suppose you want to
33947 implement a class whose instances hold a weak pointer reference; the
33948 following class does this:
33951 @interface WeakPointer : Object
33953 const void* weakPointer;
33956 - initWithPointer:(const void*)p;
33957 - (const void*)weakPointer;
33961 @implementation WeakPointer
33965 class_ivar_set_gcinvisible (self, "weakPointer", YES);
33968 - initWithPointer:(const void*)p
33974 - (const void*)weakPointer
33976 return weakPointer;
33981 Weak pointers are supported through a new type character specifier
33982 represented by the `!' character. The `class_ivar_set_gcinvisible()'
33983 function adds or removes this specifier to the string type description
33984 of the instance variable named as argument.
33987 File: gcc.info, Node: Constant string objects, Next: compatibility_alias, Prev: Garbage Collection, Up: Objective-C
33989 7.4 Constant string objects
33990 ===========================
33992 GNU Objective-C provides constant string objects that are generated
33993 directly by the compiler. You declare a constant string object by
33994 prefixing a C constant string with the character `@':
33996 id myString = @"this is a constant string object";
33998 The constant string objects are by default instances of the
33999 `NXConstantString' class which is provided by the GNU Objective-C
34000 runtime. To get the definition of this class you must include the
34001 `objc/NXConstStr.h' header file.
34003 User defined libraries may want to implement their own constant string
34004 class. To be able to support them, the GNU Objective-C compiler
34005 provides a new command line options
34006 `-fconstant-string-class=CLASS-NAME'. The provided class should adhere
34007 to a strict structure, the same as `NXConstantString''s structure:
34010 @interface MyConstantStringClass
34018 `NXConstantString' inherits from `Object'; user class libraries may
34019 choose to inherit the customized constant string class from a different
34020 class than `Object'. There is no requirement in the methods the
34021 constant string class has to implement, but the final ivar layout of
34022 the class must be the compatible with the given structure.
34024 When the compiler creates the statically allocated constant string
34025 object, the `c_string' field will be filled by the compiler with the
34026 string; the `length' field will be filled by the compiler with the
34027 string length; the `isa' pointer will be filled with `NULL' by the
34028 compiler, and it will later be fixed up automatically at runtime by the
34029 GNU Objective-C runtime library to point to the class which was set by
34030 the `-fconstant-string-class' option when the object file is loaded (if
34031 you wonder how it works behind the scenes, the name of the class to
34032 use, and the list of static objects to fixup, are stored by the
34033 compiler in the object file in a place where the GNU runtime library
34034 will find them at runtime).
34036 As a result, when a file is compiled with the
34037 `-fconstant-string-class' option, all the constant string objects will
34038 be instances of the class specified as argument to this option. It is
34039 possible to have multiple compilation units referring to different
34040 constant string classes, neither the compiler nor the linker impose any
34041 restrictions in doing this.
34044 File: gcc.info, Node: compatibility_alias, Prev: Constant string objects, Up: Objective-C
34046 7.5 compatibility_alias
34047 =======================
34049 This is a feature of the Objective-C compiler rather than of the
34050 runtime, anyway since it is documented nowhere and its existence was
34051 forgotten, we are documenting it here.
34053 The keyword `@compatibility_alias' allows you to define a class name
34054 as equivalent to another class name. For example:
34056 @compatibility_alias WOApplication GSWApplication;
34058 tells the compiler that each time it encounters `WOApplication' as a
34059 class name, it should replace it with `GSWApplication' (that is,
34060 `WOApplication' is just an alias for `GSWApplication').
34062 There are some constraints on how this can be used--
34064 * `WOApplication' (the alias) must not be an existing class;
34066 * `GSWApplication' (the real class) must be an existing class.
34070 File: gcc.info, Node: Compatibility, Next: Gcov, Prev: Objective-C, Up: Top
34072 8 Binary Compatibility
34073 **********************
34075 Binary compatibility encompasses several related concepts:
34077 "application binary interface (ABI)"
34078 The set of runtime conventions followed by all of the tools that
34079 deal with binary representations of a program, including
34080 compilers, assemblers, linkers, and language runtime support.
34081 Some ABIs are formal with a written specification, possibly
34082 designed by multiple interested parties. Others are simply the
34083 way things are actually done by a particular set of tools.
34086 A compiler conforms to an ABI if it generates code that follows
34087 all of the specifications enumerated by that ABI. A library
34088 conforms to an ABI if it is implemented according to that ABI. An
34089 application conforms to an ABI if it is built using tools that
34090 conform to that ABI and does not contain source code that
34091 specifically changes behavior specified by the ABI.
34093 "calling conventions"
34094 Calling conventions are a subset of an ABI that specify of how
34095 arguments are passed and function results are returned.
34098 Different sets of tools are interoperable if they generate files
34099 that can be used in the same program. The set of tools includes
34100 compilers, assemblers, linkers, libraries, header files, startup
34101 files, and debuggers. Binaries produced by different sets of
34102 tools are not interoperable unless they implement the same ABI.
34103 This applies to different versions of the same tools as well as
34104 tools from different vendors.
34107 Whether a function in a binary built by one set of tools can call a
34108 function in a binary built by a different set of tools is a subset
34109 of interoperability.
34111 "implementation-defined features"
34112 Language standards include lists of implementation-defined
34113 features whose behavior can vary from one implementation to
34114 another. Some of these features are normally covered by a
34115 platform's ABI and others are not. The features that are not
34116 covered by an ABI generally affect how a program behaves, but not
34120 Conformance to the same ABI and the same behavior of
34121 implementation-defined features are both relevant for
34124 The application binary interface implemented by a C or C++ compiler
34125 affects code generation and runtime support for:
34127 * size and alignment of data types
34129 * layout of structured types
34131 * calling conventions
34133 * register usage conventions
34135 * interfaces for runtime arithmetic support
34137 * object file formats
34139 In addition, the application binary interface implemented by a C++
34140 compiler affects code generation and runtime support for:
34143 * exception handling
34145 * invoking constructors and destructors
34147 * layout, alignment, and padding of classes
34149 * layout and alignment of virtual tables
34151 Some GCC compilation options cause the compiler to generate code that
34152 does not conform to the platform's default ABI. Other options cause
34153 different program behavior for implementation-defined features that are
34154 not covered by an ABI. These options are provided for consistency with
34155 other compilers that do not follow the platform's default ABI or the
34156 usual behavior of implementation-defined features for the platform. Be
34157 very careful about using such options.
34159 Most platforms have a well-defined ABI that covers C code, but ABIs
34160 that cover C++ functionality are not yet common.
34162 Starting with GCC 3.2, GCC binary conventions for C++ are based on a
34163 written, vendor-neutral C++ ABI that was designed to be specific to
34164 64-bit Itanium but also includes generic specifications that apply to
34165 any platform. This C++ ABI is also implemented by other compiler
34166 vendors on some platforms, notably GNU/Linux and BSD systems. We have
34167 tried hard to provide a stable ABI that will be compatible with future
34168 GCC releases, but it is possible that we will encounter problems that
34169 make this difficult. Such problems could include different
34170 interpretations of the C++ ABI by different vendors, bugs in the ABI, or
34171 bugs in the implementation of the ABI in different compilers. GCC's
34172 `-Wabi' switch warns when G++ generates code that is probably not
34173 compatible with the C++ ABI.
34175 The C++ library used with a C++ compiler includes the Standard C++
34176 Library, with functionality defined in the C++ Standard, plus language
34177 runtime support. The runtime support is included in a C++ ABI, but
34178 there is no formal ABI for the Standard C++ Library. Two
34179 implementations of that library are interoperable if one follows the
34180 de-facto ABI of the other and if they are both built with the same
34181 compiler, or with compilers that conform to the same ABI for C++
34182 compiler and runtime support.
34184 When G++ and another C++ compiler conform to the same C++ ABI, but the
34185 implementations of the Standard C++ Library that they normally use do
34186 not follow the same ABI for the Standard C++ Library, object files
34187 built with those compilers can be used in the same program only if they
34188 use the same C++ library. This requires specifying the location of the
34189 C++ library header files when invoking the compiler whose usual library
34190 is not being used. The location of GCC's C++ header files depends on
34191 how the GCC build was configured, but can be seen by using the G++ `-v'
34192 option. With default configuration options for G++ 3.3 the compile
34193 line for a different C++ compiler needs to include
34195 -IGCC_INSTALL_DIRECTORY/include/c++/3.3
34197 Similarly, compiling code with G++ that must use a C++ library other
34198 than the GNU C++ library requires specifying the location of the header
34199 files for that other library.
34201 The most straightforward way to link a program to use a particular C++
34202 library is to use a C++ driver that specifies that C++ library by
34203 default. The `g++' driver, for example, tells the linker where to find
34204 GCC's C++ library (`libstdc++') plus the other libraries and startup
34205 files it needs, in the proper order.
34207 If a program must use a different C++ library and it's not possible to
34208 do the final link using a C++ driver that uses that library by default,
34209 it is necessary to tell `g++' the location and name of that library.
34210 It might also be necessary to specify different startup files and other
34211 runtime support libraries, and to suppress the use of GCC's support
34212 libraries with one or more of the options `-nostdlib', `-nostartfiles',
34213 and `-nodefaultlibs'.
34216 File: gcc.info, Node: Gcov, Next: Trouble, Prev: Compatibility, Up: Top
34218 9 `gcov'--a Test Coverage Program
34219 *********************************
34221 `gcov' is a tool you can use in conjunction with GCC to test code
34222 coverage in your programs.
34226 * Gcov Intro:: Introduction to gcov.
34227 * Invoking Gcov:: How to use gcov.
34228 * Gcov and Optimization:: Using gcov with GCC optimization.
34229 * Gcov Data Files:: The files used by gcov.
34230 * Cross-profiling:: Data file relocation.
34233 File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov
34235 9.1 Introduction to `gcov'
34236 ==========================
34238 `gcov' is a test coverage program. Use it in concert with GCC to
34239 analyze your programs to help create more efficient, faster running
34240 code and to discover untested parts of your program. You can use
34241 `gcov' as a profiling tool to help discover where your optimization
34242 efforts will best affect your code. You can also use `gcov' along with
34243 the other profiling tool, `gprof', to assess which parts of your code
34244 use the greatest amount of computing time.
34246 Profiling tools help you analyze your code's performance. Using a
34247 profiler such as `gcov' or `gprof', you can find out some basic
34248 performance statistics, such as:
34250 * how often each line of code executes
34252 * what lines of code are actually executed
34254 * how much computing time each section of code uses
34256 Once you know these things about how your code works when compiled, you
34257 can look at each module to see which modules should be optimized.
34258 `gcov' helps you determine where to work on optimization.
34260 Software developers also use coverage testing in concert with
34261 testsuites, to make sure software is actually good enough for a release.
34262 Testsuites can verify that a program works as expected; a coverage
34263 program tests to see how much of the program is exercised by the
34264 testsuite. Developers can then determine what kinds of test cases need
34265 to be added to the testsuites to create both better testing and a better
34268 You should compile your code without optimization if you plan to use
34269 `gcov' because the optimization, by combining some lines of code into
34270 one function, may not give you as much information as you need to look
34271 for `hot spots' where the code is using a great deal of computer time.
34272 Likewise, because `gcov' accumulates statistics by line (at the lowest
34273 resolution), it works best with a programming style that places only
34274 one statement on each line. If you use complicated macros that expand
34275 to loops or to other control structures, the statistics are less
34276 helpful--they only report on the line where the macro call appears. If
34277 your complex macros behave like functions, you can replace them with
34278 inline functions to solve this problem.
34280 `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates how
34281 many times each line of a source file `SOURCEFILE.c' has executed. You
34282 can use these logfiles along with `gprof' to aid in fine-tuning the
34283 performance of your programs. `gprof' gives timing information you can
34284 use along with the information you get from `gcov'.
34286 `gcov' works only on code compiled with GCC. It is not compatible
34287 with any other profiling or test coverage mechanism.
34290 File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov
34292 9.2 Invoking `gcov'
34293 ===================
34295 gcov [OPTIONS] SOURCEFILES
34297 `gcov' accepts the following options:
34301 Display help about using `gcov' (on the standard output), and exit
34302 without doing any further processing.
34306 Display the `gcov' version number (on the standard output), and
34307 exit without doing any further processing.
34311 Write individual execution counts for every basic block. Normally
34312 gcov outputs execution counts only for the main blocks of a line.
34313 With this option you can determine if blocks within a single line
34314 are not being executed.
34317 `--branch-probabilities'
34318 Write branch frequencies to the output file, and write branch
34319 summary info to the standard output. This option allows you to
34320 see how often each branch in your program was taken.
34321 Unconditional branches will not be shown, unless the `-u' option
34326 Write branch frequencies as the number of branches taken, rather
34327 than the percentage of branches taken.
34331 Do not create the `gcov' output file.
34334 `--long-file-names'
34335 Create long file names for included source files. For example, if
34336 the header file `x.h' contains code, and was included in the file
34337 `a.c', then running `gcov' on the file `a.c' will produce an
34338 output file called `a.c##x.h.gcov' instead of `x.h.gcov'. This
34339 can be useful if `x.h' is included in multiple source files. If
34340 you use the `-p' option, both the including and included file
34341 names will be complete path names.
34345 Preserve complete path information in the names of generated
34346 `.gcov' files. Without this option, just the filename component is
34347 used. With this option, all directories are used, with `/'
34348 characters translated to `#' characters, `.' directory components
34349 removed and `..' components renamed to `^'. This is useful if
34350 sourcefiles are in several different directories. It also affects
34354 `--function-summaries'
34355 Output summaries for each function in addition to the file level
34358 `-o DIRECTORY|FILE'
34359 `--object-directory DIRECTORY'
34360 `--object-file FILE'
34361 Specify either the directory containing the gcov data files, or the
34362 object path name. The `.gcno', and `.gcda' data files are
34363 searched for using this option. If a directory is specified, the
34364 data files are in that directory and named after the source file
34365 name, without its extension. If a file is specified here, the
34366 data files are named after that file, without its extension. If
34367 this option is not supplied, it defaults to the current directory.
34370 `--unconditional-branches'
34371 When branch probabilities are given, include those of
34372 unconditional branches. Unconditional branches are normally not
34376 `gcov' should be run with the current directory the same as that when
34377 you invoked the compiler. Otherwise it will not be able to locate the
34378 source files. `gcov' produces files called `MANGLEDNAME.gcov' in the
34379 current directory. These contain the coverage information of the
34380 source file they correspond to. One `.gcov' file is produced for each
34381 source file containing code, which was compiled to produce the data
34382 files. The MANGLEDNAME part of the output file name is usually simply
34383 the source file name, but can be something more complicated if the `-l'
34384 or `-p' options are given. Refer to those options for details.
34386 The `.gcov' files contain the `:' separated fields along with program
34387 source code. The format is
34389 EXECUTION_COUNT:LINE_NUMBER:SOURCE LINE TEXT
34391 Additional block information may succeed each line, when requested by
34392 command line option. The EXECUTION_COUNT is `-' for lines containing
34393 no code and `#####' for lines which were never executed. Some lines of
34394 information at the start have LINE_NUMBER of zero.
34396 The preamble lines are of the form
34400 The ordering and number of these preamble lines will be augmented as
34401 `gcov' development progresses -- do not rely on them remaining
34402 unchanged. Use TAG to locate a particular preamble line.
34404 The additional block information is of the form
34408 The INFORMATION is human readable, but designed to be simple enough
34409 for machine parsing too.
34411 When printing percentages, 0% and 100% are only printed when the values
34412 are _exactly_ 0% and 100% respectively. Other values which would
34413 conventionally be rounded to 0% or 100% are instead printed as the
34414 nearest non-boundary value.
34416 When using `gcov', you must first compile your program with two
34417 special GCC options: `-fprofile-arcs -ftest-coverage'. This tells the
34418 compiler to generate additional information needed by gcov (basically a
34419 flow graph of the program) and also includes additional code in the
34420 object files for generating the extra profiling information needed by
34421 gcov. These additional files are placed in the directory where the
34422 object file is located.
34424 Running the program will cause profile output to be generated. For
34425 each source file compiled with `-fprofile-arcs', an accompanying
34426 `.gcda' file will be placed in the object file directory.
34428 Running `gcov' with your program's source file names as arguments will
34429 now produce a listing of the code along with frequency of execution for
34430 each line. For example, if your program is called `tmp.c', this is
34431 what you see when you use the basic `gcov' facility:
34433 $ gcc -fprofile-arcs -ftest-coverage tmp.c
34436 90.00% of 10 source lines executed in file tmp.c
34437 Creating tmp.c.gcov.
34439 The file `tmp.c.gcov' contains output from `gcov'. Here is a sample:
34442 -: 0:Graph:tmp.gcno
34446 -: 1:#include <stdio.h>
34448 -: 3:int main (void)
34450 1: 5: int i, total;
34454 11: 9: for (i = 0; i < 10; i++)
34455 10: 10: total += i;
34457 1: 12: if (total != 45)
34458 #####: 13: printf ("Failure\n");
34460 1: 15: printf ("Success\n");
34464 When you use the `-a' option, you will get individual block counts,
34465 and the output looks like this:
34468 -: 0:Graph:tmp.gcno
34472 -: 1:#include <stdio.h>
34474 -: 3:int main (void)
34477 1: 5: int i, total;
34481 11: 9: for (i = 0; i < 10; i++)
34483 10: 10: total += i;
34486 1: 12: if (total != 45)
34488 #####: 13: printf ("Failure\n");
34491 1: 15: printf ("Success\n");
34497 In this mode, each basic block is only shown on one line - the last
34498 line of the block. A multi-line block will only contribute to the
34499 execution count of that last line, and other lines will not be shown to
34500 contain code, unless previous blocks end on those lines. The total
34501 execution count of a line is shown and subsequent lines show the
34502 execution counts for individual blocks that end on that line. After
34503 each block, the branch and call counts of the block will be shown, if
34504 the `-b' option is given.
34506 Because of the way GCC instruments calls, a call count can be shown
34507 after a line with no individual blocks. As you can see, line 13
34508 contains a basic block that was not executed.
34510 When you use the `-b' option, your output looks like this:
34513 90.00% of 10 source lines executed in file tmp.c
34514 80.00% of 5 branches executed in file tmp.c
34515 80.00% of 5 branches taken at least once in file tmp.c
34516 50.00% of 2 calls executed in file tmp.c
34517 Creating tmp.c.gcov.
34519 Here is a sample of a resulting `tmp.c.gcov' file:
34522 -: 0:Graph:tmp.gcno
34526 -: 1:#include <stdio.h>
34528 -: 3:int main (void)
34529 function main called 1 returned 1 blocks executed 75%
34531 1: 5: int i, total;
34535 11: 9: for (i = 0; i < 10; i++)
34536 branch 0 taken 91% (fallthrough)
34538 10: 10: total += i;
34540 1: 12: if (total != 45)
34541 branch 0 taken 0% (fallthrough)
34542 branch 1 taken 100%
34543 #####: 13: printf ("Failure\n");
34544 call 0 never executed
34546 1: 15: printf ("Success\n");
34547 call 0 called 1 returned 100%
34551 For each function, a line is printed showing how many times the
34552 function is called, how many times it returns and what percentage of the
34553 function's blocks were executed.
34555 For each basic block, a line is printed after the last line of the
34556 basic block describing the branch or call that ends the basic block.
34557 There can be multiple branches and calls listed for a single source
34558 line if there are multiple basic blocks that end on that line. In this
34559 case, the branches and calls are each given a number. There is no
34560 simple way to map these branches and calls back to source constructs.
34561 In general, though, the lowest numbered branch or call will correspond
34562 to the leftmost construct on the source line.
34564 For a branch, if it was executed at least once, then a percentage
34565 indicating the number of times the branch was taken divided by the
34566 number of times the branch was executed will be printed. Otherwise, the
34567 message "never executed" is printed.
34569 For a call, if it was executed at least once, then a percentage
34570 indicating the number of times the call returned divided by the number
34571 of times the call was executed will be printed. This will usually be
34572 100%, but may be less for functions that call `exit' or `longjmp', and
34573 thus may not return every time they are called.
34575 The execution counts are cumulative. If the example program were
34576 executed again without removing the `.gcda' file, the count for the
34577 number of times each line in the source was executed would be added to
34578 the results of the previous run(s). This is potentially useful in
34579 several ways. For example, it could be used to accumulate data over a
34580 number of program runs as part of a test verification suite, or to
34581 provide more accurate long-term information over a large number of
34584 The data in the `.gcda' files is saved immediately before the program
34585 exits. For each source file compiled with `-fprofile-arcs', the
34586 profiling code first attempts to read in an existing `.gcda' file; if
34587 the file doesn't match the executable (differing number of basic block
34588 counts) it will ignore the contents of the file. It then adds in the
34589 new execution counts and finally writes the data to the file.
34592 File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov
34594 9.3 Using `gcov' with GCC Optimization
34595 ======================================
34597 If you plan to use `gcov' to help optimize your code, you must first
34598 compile your program with two special GCC options: `-fprofile-arcs
34599 -ftest-coverage'. Aside from that, you can use any other GCC options;
34600 but if you want to prove that every single line in your program was
34601 executed, you should not compile with optimization at the same time.
34602 On some machines the optimizer can eliminate some simple code lines by
34603 combining them with other lines. For example, code like this:
34610 can be compiled into one instruction on some machines. In this case,
34611 there is no way for `gcov' to calculate separate execution counts for
34612 each line because there isn't separate code for each line. Hence the
34613 `gcov' output looks like this if you compiled the program with
34616 100: 12:if (a != b)
34621 The output shows that this block of code, combined by optimization,
34622 executed 100 times. In one sense this result is correct, because there
34623 was only one instruction representing all four of these lines. However,
34624 the output does not indicate how many times the result was 0 and how
34625 many times the result was 1.
34627 Inlineable functions can create unexpected line counts. Line counts
34628 are shown for the source code of the inlineable function, but what is
34629 shown depends on where the function is inlined, or if it is not inlined
34632 If the function is not inlined, the compiler must emit an out of line
34633 copy of the function, in any object file that needs it. If `fileA.o'
34634 and `fileB.o' both contain out of line bodies of a particular
34635 inlineable function, they will also both contain coverage counts for
34636 that function. When `fileA.o' and `fileB.o' are linked together, the
34637 linker will, on many systems, select one of those out of line bodies
34638 for all calls to that function, and remove or ignore the other.
34639 Unfortunately, it will not remove the coverage counters for the unused
34640 function body. Hence when instrumented, all but one use of that
34641 function will show zero counts.
34643 If the function is inlined in several places, the block structure in
34644 each location might not be the same. For instance, a condition might
34645 now be calculable at compile time in some instances. Because the
34646 coverage of all the uses of the inline function will be shown for the
34647 same source lines, the line counts themselves might seem inconsistent.
34650 File: gcc.info, Node: Gcov Data Files, Next: Cross-profiling, Prev: Gcov and Optimization, Up: Gcov
34652 9.4 Brief description of `gcov' data files
34653 ==========================================
34655 `gcov' uses two files for profiling. The names of these files are
34656 derived from the original _object_ file by substituting the file suffix
34657 with either `.gcno', or `.gcda'. All of these files are placed in the
34658 same directory as the object file, and contain data stored in a
34659 platform-independent format.
34661 The `.gcno' file is generated when the source file is compiled with
34662 the GCC `-ftest-coverage' option. It contains information to
34663 reconstruct the basic block graphs and assign source line numbers to
34666 The `.gcda' file is generated when a program containing object files
34667 built with the GCC `-fprofile-arcs' option is executed. A separate
34668 `.gcda' file is created for each object file compiled with this option.
34669 It contains arc transition counts, and some summary information.
34671 The full details of the file format is specified in `gcov-io.h', and
34672 functions provided in that header file should be used to access the
34676 File: gcc.info, Node: Cross-profiling, Prev: Gcov Data Files, Up: Gcov
34678 9.5 Data file relocation to support cross-profiling
34679 ===================================================
34681 Running the program will cause profile output to be generated. For each
34682 source file compiled with `-fprofile-arcs', an accompanying `.gcda'
34683 file will be placed in the object file directory. That implicitly
34684 requires running the program on the same system as it was built or
34685 having the same absolute directory structure on the target system. The
34686 program will try to create the needed directory structure, if it is not
34689 To support cross-profiling, a program compiled with `-fprofile-arcs'
34690 can relocate the data files based on two environment variables:
34692 * GCOV_PREFIX contains the prefix to add to the absolute paths in
34693 the object file. Prefix must be absolute as well, otherwise its
34694 value is ignored. The default is no prefix.
34696 * GCOV_PREFIX_STRIP indicates the how many initial directory names
34697 to strip off the hardwired absolute paths. Default value is 0.
34699 _Note:_ GCOV_PREFIX_STRIP has no effect if GCOV_PREFIX is
34700 undefined, empty or non-absolute.
34702 For example, if the object file `/user/build/foo.o' was built with
34703 `-fprofile-arcs', the final executable will try to create the data file
34704 `/user/build/foo.gcda' when running on the target system. This will
34705 fail if the corresponding directory does not exist and it is unable to
34706 create it. This can be overcome by, for example, setting the
34707 environment as `GCOV_PREFIX=/target/run' and `GCOV_PREFIX_STRIP=1'.
34708 Such a setting will name the data file `/target/run/build/foo.gcda'.
34710 You must move the data files to the expected directory tree in order to
34711 use them for profile directed optimizations (`--use-profile'), or to
34712 use the `gcov' tool.
34715 File: gcc.info, Node: Trouble, Next: Bugs, Prev: Gcov, Up: Top
34717 10 Known Causes of Trouble with GCC
34718 ***********************************
34720 This section describes known problems that affect users of GCC. Most
34721 of these are not GCC bugs per se--if they were, we would fix them. But
34722 the result for a user may be like the result of a bug.
34724 Some of these problems are due to bugs in other software, some are
34725 missing features that are too much work to add, and some are places
34726 where people's opinions differ as to what is best.
34730 * Actual Bugs:: Bugs we will fix later.
34731 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
34732 * Interoperation:: Problems using GCC with other compilers,
34733 and with certain linkers, assemblers and debuggers.
34734 * Incompatibilities:: GCC is incompatible with traditional C.
34735 * Fixed Headers:: GCC uses corrected versions of system header files.
34736 This is necessary, but doesn't always work smoothly.
34737 * Standard Libraries:: GCC uses the system C library, which might not be
34738 compliant with the ISO C standard.
34739 * Disappointments:: Regrettable things we can't change, but not quite bugs.
34740 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
34741 * Protoize Caveats:: Things to watch out for when using `protoize'.
34742 * Non-bugs:: Things we think are right, but some others disagree.
34743 * Warnings and Errors:: Which problems in your code get warnings,
34744 and which get errors.
34747 File: gcc.info, Node: Actual Bugs, Next: Cross-Compiler Problems, Up: Trouble
34749 10.1 Actual Bugs We Haven't Fixed Yet
34750 =====================================
34752 * The `fixincludes' script interacts badly with automounters; if the
34753 directory of system header files is automounted, it tends to be
34754 unmounted while `fixincludes' is running. This would seem to be a
34755 bug in the automounter. We don't know any good way to work around
34758 * The `fixproto' script will sometimes add prototypes for the
34759 `sigsetjmp' and `siglongjmp' functions that reference the
34760 `jmp_buf' type before that type is defined. To work around this,
34761 edit the offending file and place the typedef in front of the
34765 File: gcc.info, Node: Cross-Compiler Problems, Next: Interoperation, Prev: Actual Bugs, Up: Trouble
34767 10.2 Cross-Compiler Problems
34768 ============================
34770 You may run into problems with cross compilation on certain machines,
34771 for several reasons.
34773 * At present, the program `mips-tfile' which adds debug support to
34774 object files on MIPS systems does not work in a cross compile
34778 File: gcc.info, Node: Interoperation, Next: Incompatibilities, Prev: Cross-Compiler Problems, Up: Trouble
34780 10.3 Interoperation
34781 ===================
34783 This section lists various difficulties encountered in using GCC
34784 together with other compilers or with the assemblers, linkers,
34785 libraries and debuggers on certain systems.
34787 * On many platforms, GCC supports a different ABI for C++ than do
34788 other compilers, so the object files compiled by GCC cannot be
34789 used with object files generated by another C++ compiler.
34791 An area where the difference is most apparent is name mangling.
34792 The use of different name mangling is intentional, to protect you
34793 from more subtle problems. Compilers differ as to many internal
34794 details of C++ implementation, including: how class instances are
34795 laid out, how multiple inheritance is implemented, and how virtual
34796 function calls are handled. If the name encoding were made the
34797 same, your programs would link against libraries provided from
34798 other compilers--but the programs would then crash when run.
34799 Incompatible libraries are then detected at link time, rather than
34802 * On some BSD systems, including some versions of Ultrix, use of
34803 profiling causes static variable destructors (currently used only
34804 in C++) not to be run.
34806 * On some SGI systems, when you use `-lgl_s' as an option, it gets
34807 translated magically to `-lgl_s -lX11_s -lc_s'. Naturally, this
34808 does not happen when you use GCC. You must specify all three
34809 options explicitly.
34811 * On a SPARC, GCC aligns all values of type `double' on an 8-byte
34812 boundary, and it expects every `double' to be so aligned. The Sun
34813 compiler usually gives `double' values 8-byte alignment, with one
34814 exception: function arguments of type `double' may not be aligned.
34816 As a result, if a function compiled with Sun CC takes the address
34817 of an argument of type `double' and passes this pointer of type
34818 `double *' to a function compiled with GCC, dereferencing the
34819 pointer may cause a fatal signal.
34821 One way to solve this problem is to compile your entire program
34822 with GCC. Another solution is to modify the function that is
34823 compiled with Sun CC to copy the argument into a local variable;
34824 local variables are always properly aligned. A third solution is
34825 to modify the function that uses the pointer to dereference it via
34826 the following function `access_double' instead of directly with
34830 access_double (double *unaligned_ptr)
34832 union d2i { double d; int i[2]; };
34834 union d2i *p = (union d2i *) unaligned_ptr;
34843 Storing into the pointer can be done likewise with the same union.
34845 * On Solaris, the `malloc' function in the `libmalloc.a' library may
34846 allocate memory that is only 4 byte aligned. Since GCC on the
34847 SPARC assumes that doubles are 8 byte aligned, this may result in a
34848 fatal signal if doubles are stored in memory allocated by the
34849 `libmalloc.a' library.
34851 The solution is to not use the `libmalloc.a' library. Use instead
34852 `malloc' and related functions from `libc.a'; they do not have
34855 * On the HP PA machine, ADB sometimes fails to work on functions
34856 compiled with GCC. Specifically, it fails to work on functions
34857 that use `alloca' or variable-size arrays. This is because GCC
34858 doesn't generate HP-UX unwind descriptors for such functions. It
34859 may even be impossible to generate them.
34861 * Debugging (`-g') is not supported on the HP PA machine, unless you
34862 use the preliminary GNU tools.
34864 * Taking the address of a label may generate errors from the HP-UX
34865 PA assembler. GAS for the PA does not have this problem.
34867 * Using floating point parameters for indirect calls to static
34868 functions will not work when using the HP assembler. There simply
34869 is no way for GCC to specify what registers hold arguments for
34870 static functions when using the HP assembler. GAS for the PA does
34871 not have this problem.
34873 * In extremely rare cases involving some very large functions you may
34874 receive errors from the HP linker complaining about an out of
34875 bounds unconditional branch offset. This used to occur more often
34876 in previous versions of GCC, but is now exceptionally rare. If
34877 you should run into it, you can work around by making your
34880 * GCC compiled code sometimes emits warnings from the HP-UX
34881 assembler of the form:
34883 (warning) Use of GR3 when
34884 frame >= 8192 may cause conflict.
34886 These warnings are harmless and can be safely ignored.
34888 * In extremely rare cases involving some very large functions you may
34889 receive errors from the AIX Assembler complaining about a
34890 displacement that is too large. If you should run into it, you
34891 can work around by making your function smaller.
34893 * The `libstdc++.a' library in GCC relies on the SVR4 dynamic linker
34894 semantics which merges global symbols between libraries and
34895 applications, especially necessary for C++ streams functionality.
34896 This is not the default behavior of AIX shared libraries and
34897 dynamic linking. `libstdc++.a' is built on AIX with
34898 "runtime-linking" enabled so that symbol merging can occur. To
34899 utilize this feature, the application linked with `libstdc++.a'
34900 must include the `-Wl,-brtl' flag on the link line. G++ cannot
34901 impose this because this option may interfere with the semantics
34902 of the user program and users may not always use `g++' to link his
34903 or her application. Applications are not required to use the
34904 `-Wl,-brtl' flag on the link line--the rest of the `libstdc++.a'
34905 library which is not dependent on the symbol merging semantics
34906 will continue to function correctly.
34908 * An application can interpose its own definition of functions for
34909 functions invoked by `libstdc++.a' with "runtime-linking" enabled
34910 on AIX. To accomplish this the application must be linked with
34911 "runtime-linking" option and the functions explicitly must be
34912 exported by the application (`-Wl,-brtl,-bE:exportfile').
34914 * AIX on the RS/6000 provides support (NLS) for environments outside
34915 of the United States. Compilers and assemblers use NLS to support
34916 locale-specific representations of various objects including
34917 floating-point numbers (`.' vs `,' for separating decimal
34918 fractions). There have been problems reported where the library
34919 linked with GCC does not produce the same floating-point formats
34920 that the assembler accepts. If you have this problem, set the
34921 `LANG' environment variable to `C' or `En_US'.
34923 * Even if you specify `-fdollars-in-identifiers', you cannot
34924 successfully use `$' in identifiers on the RS/6000 due to a
34925 restriction in the IBM assembler. GAS supports these identifiers.
34929 File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: Interoperation, Up: Trouble
34931 10.4 Incompatibilities of GCC
34932 =============================
34934 There are several noteworthy incompatibilities between GNU C and K&R
34935 (non-ISO) versions of C.
34937 * GCC normally makes string constants read-only. If several
34938 identical-looking string constants are used, GCC stores only one
34939 copy of the string.
34941 One consequence is that you cannot call `mktemp' with a string
34942 constant argument. The function `mktemp' always alters the string
34943 its argument points to.
34945 Another consequence is that `sscanf' does not work on some very
34946 old systems when passed a string constant as its format control
34947 string or input. This is because `sscanf' incorrectly tries to
34948 write into the string constant. Likewise `fscanf' and `scanf'.
34950 The solution to these problems is to change the program to use
34951 `char'-array variables with initialization strings for these
34952 purposes instead of string constants.
34954 * `-2147483648' is positive.
34956 This is because 2147483648 cannot fit in the type `int', so
34957 (following the ISO C rules) its data type is `unsigned long int'.
34958 Negating this value yields 2147483648 again.
34960 * GCC does not substitute macro arguments when they appear inside of
34961 string constants. For example, the following macro in GCC
34965 will produce output `"a"' regardless of what the argument A is.
34967 * When you use `setjmp' and `longjmp', the only automatic variables
34968 guaranteed to remain valid are those declared `volatile'. This is
34969 a consequence of automatic register allocation. Consider this
34983 /* `longjmp (j)' may occur in `fun3'. */
34984 return a + fun3 ();
34987 Here `a' may or may not be restored to its first value when the
34988 `longjmp' occurs. If `a' is allocated in a register, then its
34989 first value is restored; otherwise, it keeps the last value stored
34992 If you use the `-W' option with the `-O' option, you will get a
34993 warning when GCC thinks such a problem might be possible.
34995 * Programs that use preprocessing directives in the middle of macro
34996 arguments do not work with GCC. For example, a program like this
35003 ISO C does not permit such a construct.
35005 * K&R compilers allow comments to cross over an inclusion boundary
35006 (i.e. started in an include file and ended in the including file).
35008 * Declarations of external variables and functions within a block
35009 apply only to the block containing the declaration. In other
35010 words, they have the same scope as any other declaration in the
35013 In some other C compilers, a `extern' declaration affects all the
35014 rest of the file even if it happens within a block.
35016 * In traditional C, you can combine `long', etc., with a typedef
35017 name, as shown here:
35020 typedef long foo bar;
35022 In ISO C, this is not allowed: `long' and other type modifiers
35023 require an explicit `int'.
35025 * PCC allows typedef names to be used as function parameters.
35027 * Traditional C allows the following erroneous pair of declarations
35028 to appear together in a given scope:
35033 * GCC treats all characters of identifiers as significant.
35034 According to K&R-1 (2.2), "No more than the first eight characters
35035 are significant, although more may be used.". Also according to
35036 K&R-1 (2.2), "An identifier is a sequence of letters and digits;
35037 the first character must be a letter. The underscore _ counts as
35038 a letter.", but GCC also allows dollar signs in identifiers.
35040 * PCC allows whitespace in the middle of compound assignment
35041 operators such as `+='. GCC, following the ISO standard, does not
35044 * GCC complains about unterminated character constants inside of
35045 preprocessing conditionals that fail. Some programs have English
35046 comments enclosed in conditionals that are guaranteed to fail; if
35047 these comments contain apostrophes, GCC will probably report an
35048 error. For example, this code would produce an error:
35051 You can't expect this to work.
35054 The best solution to such a problem is to put the text into an
35055 actual C comment delimited by `/*...*/'.
35057 * Many user programs contain the declaration `long time ();'. In the
35058 past, the system header files on many systems did not actually
35059 declare `time', so it did not matter what type your program
35060 declared it to return. But in systems with ISO C headers, `time'
35061 is declared to return `time_t', and if that is not the same as
35062 `long', then `long time ();' is erroneous.
35064 The solution is to change your program to use appropriate system
35065 headers (`<time.h>' on systems with ISO C headers) and not to
35066 declare `time' if the system header files declare it, or failing
35067 that to use `time_t' as the return type of `time'.
35069 * When compiling functions that return `float', PCC converts it to a
35070 double. GCC actually returns a `float'. If you are concerned
35071 with PCC compatibility, you should declare your functions to return
35072 `double'; you might as well say what you mean.
35074 * When compiling functions that return structures or unions, GCC
35075 output code normally uses a method different from that used on most
35076 versions of Unix. As a result, code compiled with GCC cannot call
35077 a structure-returning function compiled with PCC, and vice versa.
35079 The method used by GCC is as follows: a structure or union which is
35080 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or
35081 union with any other size is stored into an address supplied by
35082 the caller (usually in a special, fixed register, but on some
35083 machines it is passed on the stack). The target hook
35084 `TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address.
35086 By contrast, PCC on most target machines returns structures and
35087 unions of any size by copying the data into an area of static
35088 storage, and then returning the address of that storage as if it
35089 were a pointer value. The caller must copy the data from that
35090 memory area to the place where the value is wanted. GCC does not
35091 use this method because it is slower and nonreentrant.
35093 On some newer machines, PCC uses a reentrant convention for all
35094 structure and union returning. GCC on most of these machines uses
35095 a compatible convention when returning structures and unions in
35096 memory, but still returns small structures and unions in registers.
35098 You can tell GCC to use a compatible convention for all structure
35099 and union returning with the option `-fpcc-struct-return'.
35101 * GCC complains about program fragments such as `0x74ae-0x4000'
35102 which appear to be two hexadecimal constants separated by the minus
35103 operator. Actually, this string is a single "preprocessing token".
35104 Each such token must correspond to one token in C. Since this
35105 does not, GCC prints an error message. Although it may appear
35106 obvious that what is meant is an operator and two values, the ISO
35107 C standard specifically requires that this be treated as erroneous.
35109 A "preprocessing token" is a "preprocessing number" if it begins
35110 with a digit and is followed by letters, underscores, digits,
35111 periods and `e+', `e-', `E+', `E-', `p+', `p-', `P+', or `P-'
35112 character sequences. (In strict C89 mode, the sequences `p+',
35113 `p-', `P+' and `P-' cannot appear in preprocessing numbers.)
35115 To make the above program fragment valid, place whitespace in
35116 front of the minus sign. This whitespace will end the
35117 preprocessing number.
35120 File: gcc.info, Node: Fixed Headers, Next: Standard Libraries, Prev: Incompatibilities, Up: Trouble
35122 10.5 Fixed Header Files
35123 =======================
35125 GCC needs to install corrected versions of some system header files.
35126 This is because most target systems have some header files that won't
35127 work with GCC unless they are changed. Some have bugs, some are
35128 incompatible with ISO C, and some depend on special features of other
35131 Installing GCC automatically creates and installs the fixed header
35132 files, by running a program called `fixincludes'. Normally, you don't
35133 need to pay attention to this. But there are cases where it doesn't do
35134 the right thing automatically.
35136 * If you update the system's header files, such as by installing a
35137 new system version, the fixed header files of GCC are not
35138 automatically updated. They can be updated using the `mkheaders'
35139 script installed in `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
35141 * On some systems, header file directories contain machine-specific
35142 symbolic links in certain places. This makes it possible to share
35143 most of the header files among hosts running the same version of
35144 the system on different machine models.
35146 The programs that fix the header files do not understand this
35147 special way of using symbolic links; therefore, the directory of
35148 fixed header files is good only for the machine model used to
35151 It is possible to make separate sets of fixed header files for the
35152 different machine models, and arrange a structure of symbolic
35153 links so as to use the proper set, but you'll have to do this by
35157 File: gcc.info, Node: Standard Libraries, Next: Disappointments, Prev: Fixed Headers, Up: Trouble
35159 10.6 Standard Libraries
35160 =======================
35162 GCC by itself attempts to be a conforming freestanding implementation.
35163 *Note Language Standards Supported by GCC: Standards, for details of
35164 what this means. Beyond the library facilities required of such an
35165 implementation, the rest of the C library is supplied by the vendor of
35166 the operating system. If that C library doesn't conform to the C
35167 standards, then your programs might get warnings (especially when using
35168 `-Wall') that you don't expect.
35170 For example, the `sprintf' function on SunOS 4.1.3 returns `char *'
35171 while the C standard says that `sprintf' returns an `int'. The
35172 `fixincludes' program could make the prototype for this function match
35173 the Standard, but that would be wrong, since the function will still
35176 If you need a Standard compliant library, then you need to find one, as
35177 GCC does not provide one. The GNU C library (called `glibc') provides
35178 ISO C, POSIX, BSD, SystemV and X/Open compatibility for GNU/Linux and
35179 HURD-based GNU systems; no recent version of it supports other systems,
35180 though some very old versions did. Version 2.2 of the GNU C library
35181 includes nearly complete C99 support. You could also ask your
35182 operating system vendor if newer libraries are available.
35185 File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Standard Libraries, Up: Trouble
35187 10.7 Disappointments and Misunderstandings
35188 ==========================================
35190 These problems are perhaps regrettable, but we don't know any practical
35193 * Certain local variables aren't recognized by debuggers when you
35194 compile with optimization.
35196 This occurs because sometimes GCC optimizes the variable out of
35197 existence. There is no way to tell the debugger how to compute the
35198 value such a variable "would have had", and it is not clear that
35199 would be desirable anyway. So GCC simply does not mention the
35200 eliminated variable when it writes debugging information.
35202 You have to expect a certain amount of disagreement between the
35203 executable and your source code, when you use optimization.
35205 * Users often think it is a bug when GCC reports an error for code
35208 int foo (struct mumble *);
35210 struct mumble { ... };
35212 int foo (struct mumble *x)
35215 This code really is erroneous, because the scope of `struct
35216 mumble' in the prototype is limited to the argument list
35217 containing it. It does not refer to the `struct mumble' defined
35218 with file scope immediately below--they are two unrelated types
35219 with similar names in different scopes.
35221 But in the definition of `foo', the file-scope type is used
35222 because that is available to be inherited. Thus, the definition
35223 and the prototype do not match, and you get an error.
35225 This behavior may seem silly, but it's what the ISO standard
35226 specifies. It is easy enough for you to make your code work by
35227 moving the definition of `struct mumble' above the prototype.
35228 It's not worth being incompatible with ISO C just to avoid an
35229 error for the example shown above.
35231 * Accesses to bit-fields even in volatile objects works by accessing
35232 larger objects, such as a byte or a word. You cannot rely on what
35233 size of object is accessed in order to read or write the
35234 bit-field; it may even vary for a given bit-field according to the
35237 If you care about controlling the amount of memory that is
35238 accessed, use volatile but do not use bit-fields.
35240 * GCC comes with shell scripts to fix certain known problems in
35241 system header files. They install corrected copies of various
35242 header files in a special directory where only GCC will normally
35243 look for them. The scripts adapt to various systems by searching
35244 all the system header files for the problem cases that we know
35247 If new system header files are installed, nothing automatically
35248 arranges to update the corrected header files. They can be
35249 updated using the `mkheaders' script installed in
35250 `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
35252 * On 68000 and x86 systems, for instance, you can get paradoxical
35253 results if you test the precise values of floating point numbers.
35254 For example, you can find that a floating point value which is not
35255 a NaN is not equal to itself. This results from the fact that the
35256 floating point registers hold a few more bits of precision than
35257 fit in a `double' in memory. Compiled code moves values between
35258 memory and floating point registers at its convenience, and moving
35259 them into memory truncates them.
35261 You can partially avoid this problem by using the `-ffloat-store'
35262 option (*note Optimize Options::).
35264 * On AIX and other platforms without weak symbol support, templates
35265 need to be instantiated explicitly and symbols for static members
35266 of templates will not be generated.
35268 * On AIX, GCC scans object files and library archives for static
35269 constructors and destructors when linking an application before the
35270 linker prunes unreferenced symbols. This is necessary to prevent
35271 the AIX linker from mistakenly assuming that static constructor or
35272 destructor are unused and removing them before the scanning can
35273 occur. All static constructors and destructors found will be
35274 referenced even though the modules in which they occur may not be
35275 used by the program. This may lead to both increased executable
35276 size and unexpected symbol references.
35279 File: gcc.info, Node: C++ Misunderstandings, Next: Protoize Caveats, Prev: Disappointments, Up: Trouble
35281 10.8 Common Misunderstandings with GNU C++
35282 ==========================================
35284 C++ is a complex language and an evolving one, and its standard
35285 definition (the ISO C++ standard) was only recently completed. As a
35286 result, your C++ compiler may occasionally surprise you, even when its
35287 behavior is correct. This section discusses some areas that frequently
35288 give rise to questions of this sort.
35292 * Static Definitions:: Static member declarations are not definitions
35293 * Name lookup:: Name lookup, templates, and accessing members of base classes
35294 * Temporaries:: Temporaries may vanish before you expect
35295 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
35298 File: gcc.info, Node: Static Definitions, Next: Name lookup, Up: C++ Misunderstandings
35300 10.8.1 Declare _and_ Define Static Members
35301 ------------------------------------------
35303 When a class has static data members, it is not enough to _declare_ the
35304 static member; you must also _define_ it. For example:
35313 This declaration only establishes that the class `Foo' has an `int'
35314 named `Foo::bar', and a member function named `Foo::method'. But you
35315 still need to define _both_ `method' and `bar' elsewhere. According to
35316 the ISO standard, you must supply an initializer in one (and only one)
35317 source file, such as:
35321 Other C++ compilers may not correctly implement the standard behavior.
35322 As a result, when you switch to `g++' from one of these compilers, you
35323 may discover that a program that appeared to work correctly in fact
35324 does not conform to the standard: `g++' reports as undefined symbols
35325 any static data members that lack definitions.
35328 File: gcc.info, Node: Name lookup, Next: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings
35330 10.8.2 Name lookup, templates, and accessing members of base classes
35331 --------------------------------------------------------------------
35333 The C++ standard prescribes that all names that are not dependent on
35334 template parameters are bound to their present definitions when parsing
35335 a template function or class.(1) Only names that are dependent are
35336 looked up at the point of instantiation. For example, consider
35341 template <typename T>
35350 static const int N;
35353 Here, the names `foo' and `N' appear in a context that does not depend
35354 on the type of `T'. The compiler will thus require that they are
35355 defined in the context of use in the template, not only before the
35356 point of instantiation, and will here use `::foo(double)' and `A::N',
35357 respectively. In particular, it will convert the integer value to a
35358 `double' when passing it to `::foo(double)'.
35360 Conversely, `bar' and the call to `foo' in the fourth marked line are
35361 used in contexts that do depend on the type of `T', so they are only
35362 looked up at the point of instantiation, and you can provide
35363 declarations for them after declaring the template, but before
35364 instantiating it. In particular, if you instantiate `A::f<int>', the
35365 last line will call an overloaded `::foo(int)' if one was provided,
35366 even if after the declaration of `struct A'.
35368 This distinction between lookup of dependent and non-dependent names is
35369 called two-stage (or dependent) name lookup. G++ implements it since
35372 Two-stage name lookup sometimes leads to situations with behavior
35373 different from non-template codes. The most common is probably this:
35375 template <typename T> struct Base {
35379 template <typename T> struct Derived : public Base<T> {
35380 int get_i() { return i; }
35383 In `get_i()', `i' is not used in a dependent context, so the compiler
35384 will look for a name declared at the enclosing namespace scope (which
35385 is the global scope here). It will not look into the base class, since
35386 that is dependent and you may declare specializations of `Base' even
35387 after declaring `Derived', so the compiler can't really know what `i'
35388 would refer to. If there is no global variable `i', then you will get
35391 In order to make it clear that you want the member of the base class,
35392 you need to defer lookup until instantiation time, at which the base
35393 class is known. For this, you need to access `i' in a dependent
35394 context, by either using `this->i' (remember that `this' is of type
35395 `Derived<T>*', so is obviously dependent), or using `Base<T>::i'.
35396 Alternatively, `Base<T>::i' might be brought into scope by a
35397 `using'-declaration.
35399 Another, similar example involves calling member functions of a base
35402 template <typename T> struct Base {
35406 template <typename T> struct Derived : Base<T> {
35407 int g() { return f(); };
35410 Again, the call to `f()' is not dependent on template arguments (there
35411 are no arguments that depend on the type `T', and it is also not
35412 otherwise specified that the call should be in a dependent context).
35413 Thus a global declaration of such a function must be available, since
35414 the one in the base class is not visible until instantiation time. The
35415 compiler will consequently produce the following error message:
35417 x.cc: In member function `int Derived<T>::g()':
35418 x.cc:6: error: there are no arguments to `f' that depend on a template
35419 parameter, so a declaration of `f' must be available
35420 x.cc:6: error: (if you use `-fpermissive', G++ will accept your code, but
35421 allowing the use of an undeclared name is deprecated)
35423 To make the code valid either use `this->f()', or `Base<T>::f()'.
35424 Using the `-fpermissive' flag will also let the compiler accept the
35425 code, by marking all function calls for which no declaration is visible
35426 at the time of definition of the template for later lookup at
35427 instantiation time, as if it were a dependent call. We do not
35428 recommend using `-fpermissive' to work around invalid code, and it will
35429 also only catch cases where functions in base classes are called, not
35430 where variables in base classes are used (as in the example above).
35432 Note that some compilers (including G++ versions prior to 3.4) get
35433 these examples wrong and accept above code without an error. Those
35434 compilers do not implement two-stage name lookup correctly.
35436 ---------- Footnotes ----------
35438 (1) The C++ standard just uses the term "dependent" for names that
35439 depend on the type or value of template parameters. This shorter term
35440 will also be used in the rest of this section.
35443 File: gcc.info, Node: Temporaries, Next: Copy Assignment, Prev: Name lookup, Up: C++ Misunderstandings
35445 10.8.3 Temporaries May Vanish Before You Expect
35446 -----------------------------------------------
35448 It is dangerous to use pointers or references to _portions_ of a
35449 temporary object. The compiler may very well delete the object before
35450 you expect it to, leaving a pointer to garbage. The most common place
35451 where this problem crops up is in classes like string classes,
35452 especially ones that define a conversion function to type `char *' or
35453 `const char *'--which is one reason why the standard `string' class
35454 requires you to call the `c_str' member function. However, any class
35455 that returns a pointer to some internal structure is potentially
35456 subject to this problem.
35458 For example, a program may use a function `strfunc' that returns
35459 `string' objects, and another function `charfunc' that operates on
35460 pointers to `char':
35463 void charfunc (const char *);
35468 const char *p = strfunc().c_str();
35475 In this situation, it may seem reasonable to save a pointer to the C
35476 string returned by the `c_str' member function and use that rather than
35477 call `c_str' repeatedly. However, the temporary string created by the
35478 call to `strfunc' is destroyed after `p' is initialized, at which point
35479 `p' is left pointing to freed memory.
35481 Code like this may run successfully under some other compilers,
35482 particularly obsolete cfront-based compilers that delete temporaries
35483 along with normal local variables. However, the GNU C++ behavior is
35484 standard-conforming, so if your program depends on late destruction of
35485 temporaries it is not portable.
35487 The safe way to write such code is to give the temporary a name, which
35488 forces it to remain until the end of the scope of the name. For
35491 const string& tmp = strfunc ();
35492 charfunc (tmp.c_str ());
35495 File: gcc.info, Node: Copy Assignment, Prev: Temporaries, Up: C++ Misunderstandings
35497 10.8.4 Implicit Copy-Assignment for Virtual Bases
35498 -------------------------------------------------
35500 When a base class is virtual, only one subobject of the base class
35501 belongs to each full object. Also, the constructors and destructors are
35502 invoked only once, and called from the most-derived class. However,
35503 such objects behave unspecified when being assigned. For example:
35507 Base(char *n) : name(strdup(n)){}
35508 Base& operator= (const Base& other){
35510 name = strdup (other.name);
35514 struct A:virtual Base{
35519 struct B:virtual Base{
35524 struct Derived:public A, public B{
35525 Derived():Base("Derived"){}
35528 void func(Derived &d1, Derived &d2)
35533 The C++ standard specifies that `Base::Base' is only called once when
35534 constructing or copy-constructing a Derived object. It is unspecified
35535 whether `Base::operator=' is called more than once when the implicit
35536 copy-assignment for Derived objects is invoked (as it is inside `func'
35539 G++ implements the "intuitive" algorithm for copy-assignment: assign
35540 all direct bases, then assign all members. In that algorithm, the
35541 virtual base subobject can be encountered more than once. In the
35542 example, copying proceeds in the following order: `val', `name' (via
35543 `strdup'), `bval', and `name' again.
35545 If application code relies on copy-assignment, a user-defined
35546 copy-assignment operator removes any uncertainties. With such an
35547 operator, the application can define whether and how the virtual base
35548 subobject is assigned.
35551 File: gcc.info, Node: Protoize Caveats, Next: Non-bugs, Prev: C++ Misunderstandings, Up: Trouble
35553 10.9 Caveats of using `protoize'
35554 ================================
35556 The conversion programs `protoize' and `unprotoize' can sometimes
35557 change a source file in a way that won't work unless you rearrange it.
35559 * `protoize' can insert references to a type name or type tag before
35560 the definition, or in a file where they are not defined.
35562 If this happens, compiler error messages should show you where the
35563 new references are, so fixing the file by hand is straightforward.
35565 * There are some C constructs which `protoize' cannot figure out.
35566 For example, it can't determine argument types for declaring a
35567 pointer-to-function variable; this you must do by hand. `protoize'
35568 inserts a comment containing `???' each time it finds such a
35569 variable; so you can find all such variables by searching for this
35570 string. ISO C does not require declaring the argument types of
35571 pointer-to-function types.
35573 * Using `unprotoize' can easily introduce bugs. If the program
35574 relied on prototypes to bring about conversion of arguments, these
35575 conversions will not take place in the program without prototypes.
35576 One case in which you can be sure `unprotoize' is safe is when you
35577 are removing prototypes that were made with `protoize'; if the
35578 program worked before without any prototypes, it will work again
35581 You can find all the places where this problem might occur by
35582 compiling the program with the `-Wtraditional-conversion' option.
35583 It prints a warning whenever an argument is converted.
35585 * Both conversion programs can be confused if there are macro calls
35586 in and around the text to be converted. In other words, the
35587 standard syntax for a declaration or definition must not result
35588 from expanding a macro. This problem is inherent in the design of
35589 C and cannot be fixed. If only a few functions have confusing
35590 macro calls, you can easily convert them manually.
35592 * `protoize' cannot get the argument types for a function whose
35593 definition was not actually compiled due to preprocessing
35594 conditionals. When this happens, `protoize' changes nothing in
35595 regard to such a function. `protoize' tries to detect such
35596 instances and warn about them.
35598 You can generally work around this problem by using `protoize' step
35599 by step, each time specifying a different set of `-D' options for
35600 compilation, until all of the functions have been converted.
35601 There is no automatic way to verify that you have got them all,
35604 * Confusion may result if there is an occasion to convert a function
35605 declaration or definition in a region of source code where there
35606 is more than one formal parameter list present. Thus, attempts to
35607 convert code containing multiple (conditionally compiled) versions
35608 of a single function header (in the same vicinity) may not produce
35609 the desired (or expected) results.
35611 If you plan on converting source files which contain such code, it
35612 is recommended that you first make sure that each conditionally
35613 compiled region of source code which contains an alternative
35614 function header also contains at least one additional follower
35615 token (past the final right parenthesis of the function header).
35616 This should circumvent the problem.
35618 * `unprotoize' can become confused when trying to convert a function
35619 definition or declaration which contains a declaration for a
35620 pointer-to-function formal argument which has the same name as the
35621 function being defined or declared. We recommend you avoid such
35622 choices of formal parameter names.
35624 * You might also want to correct some of the indentation by hand and
35625 break long lines. (The conversion programs don't write lines
35626 longer than eighty characters in any case.)
35629 File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: Protoize Caveats, Up: Trouble
35631 10.10 Certain Changes We Don't Want to Make
35632 ===========================================
35634 This section lists changes that people frequently request, but which we
35635 do not make because we think GCC is better without them.
35637 * Checking the number and type of arguments to a function which has
35638 an old-fashioned definition and no prototype.
35640 Such a feature would work only occasionally--only for calls that
35641 appear in the same file as the called function, following the
35642 definition. The only way to check all calls reliably is to add a
35643 prototype for the function. But adding a prototype eliminates the
35644 motivation for this feature. So the feature is not worthwhile.
35646 * Warning about using an expression whose type is signed as a shift
35649 Shift count operands are probably signed more often than unsigned.
35650 Warning about this would cause far more annoyance than good.
35652 * Warning about assigning a signed value to an unsigned variable.
35654 Such assignments must be very common; warning about them would
35655 cause more annoyance than good.
35657 * Warning when a non-void function value is ignored.
35659 C contains many standard functions that return a value that most
35660 programs choose to ignore. One obvious example is `printf'.
35661 Warning about this practice only leads the defensive programmer to
35662 clutter programs with dozens of casts to `void'. Such casts are
35663 required so frequently that they become visual noise. Writing
35664 those casts becomes so automatic that they no longer convey useful
35665 information about the intentions of the programmer. For functions
35666 where the return value should never be ignored, use the
35667 `warn_unused_result' function attribute (*note Function
35670 * Making `-fshort-enums' the default.
35672 This would cause storage layout to be incompatible with most other
35673 C compilers. And it doesn't seem very important, given that you
35674 can get the same result in other ways. The case where it matters
35675 most is when the enumeration-valued object is inside a structure,
35676 and in that case you can specify a field width explicitly.
35678 * Making bit-fields unsigned by default on particular machines where
35679 "the ABI standard" says to do so.
35681 The ISO C standard leaves it up to the implementation whether a
35682 bit-field declared plain `int' is signed or not. This in effect
35683 creates two alternative dialects of C.
35685 The GNU C compiler supports both dialects; you can specify the
35686 signed dialect with `-fsigned-bitfields' and the unsigned dialect
35687 with `-funsigned-bitfields'. However, this leaves open the
35688 question of which dialect to use by default.
35690 Currently, the preferred dialect makes plain bit-fields signed,
35691 because this is simplest. Since `int' is the same as `signed int'
35692 in every other context, it is cleanest for them to be the same in
35693 bit-fields as well.
35695 Some computer manufacturers have published Application Binary
35696 Interface standards which specify that plain bit-fields should be
35697 unsigned. It is a mistake, however, to say anything about this
35698 issue in an ABI. This is because the handling of plain bit-fields
35699 distinguishes two dialects of C. Both dialects are meaningful on
35700 every type of machine. Whether a particular object file was
35701 compiled using signed bit-fields or unsigned is of no concern to
35702 other object files, even if they access the same bit-fields in the
35703 same data structures.
35705 A given program is written in one or the other of these two
35706 dialects. The program stands a chance to work on most any machine
35707 if it is compiled with the proper dialect. It is unlikely to work
35708 at all if compiled with the wrong dialect.
35710 Many users appreciate the GNU C compiler because it provides an
35711 environment that is uniform across machines. These users would be
35712 inconvenienced if the compiler treated plain bit-fields
35713 differently on certain machines.
35715 Occasionally users write programs intended only for a particular
35716 machine type. On these occasions, the users would benefit if the
35717 GNU C compiler were to support by default the same dialect as the
35718 other compilers on that machine. But such applications are rare.
35719 And users writing a program to run on more than one type of
35720 machine cannot possibly benefit from this kind of compatibility.
35722 This is why GCC does and will treat plain bit-fields in the same
35723 fashion on all types of machines (by default).
35725 There are some arguments for making bit-fields unsigned by default
35726 on all machines. If, for example, this becomes a universal de
35727 facto standard, it would make sense for GCC to go along with it.
35728 This is something to be considered in the future.
35730 (Of course, users strongly concerned about portability should
35731 indicate explicitly in each bit-field whether it is signed or not.
35732 In this way, they write programs which have the same meaning in
35735 * Undefining `__STDC__' when `-ansi' is not used.
35737 Currently, GCC defines `__STDC__' unconditionally. This provides
35738 good results in practice.
35740 Programmers normally use conditionals on `__STDC__' to ask whether
35741 it is safe to use certain features of ISO C, such as function
35742 prototypes or ISO token concatenation. Since plain `gcc' supports
35743 all the features of ISO C, the correct answer to these questions is
35746 Some users try to use `__STDC__' to check for the availability of
35747 certain library facilities. This is actually incorrect usage in
35748 an ISO C program, because the ISO C standard says that a conforming
35749 freestanding implementation should define `__STDC__' even though it
35750 does not have the library facilities. `gcc -ansi -pedantic' is a
35751 conforming freestanding implementation, and it is therefore
35752 required to define `__STDC__', even though it does not come with
35755 Sometimes people say that defining `__STDC__' in a compiler that
35756 does not completely conform to the ISO C standard somehow violates
35757 the standard. This is illogical. The standard is a standard for
35758 compilers that claim to support ISO C, such as `gcc -ansi'--not
35759 for other compilers such as plain `gcc'. Whatever the ISO C
35760 standard says is relevant to the design of plain `gcc' without
35761 `-ansi' only for pragmatic reasons, not as a requirement.
35763 GCC normally defines `__STDC__' to be 1, and in addition defines
35764 `__STRICT_ANSI__' if you specify the `-ansi' option, or a `-std'
35765 option for strict conformance to some version of ISO C. On some
35766 hosts, system include files use a different convention, where
35767 `__STDC__' is normally 0, but is 1 if the user specifies strict
35768 conformance to the C Standard. GCC follows the host convention
35769 when processing system include files, but when processing user
35770 files it follows the usual GNU C convention.
35772 * Undefining `__STDC__' in C++.
35774 Programs written to compile with C++-to-C translators get the
35775 value of `__STDC__' that goes with the C compiler that is
35776 subsequently used. These programs must test `__STDC__' to
35777 determine what kind of C preprocessor that compiler uses: whether
35778 they should concatenate tokens in the ISO C fashion or in the
35779 traditional fashion.
35781 These programs work properly with GNU C++ if `__STDC__' is defined.
35782 They would not work otherwise.
35784 In addition, many header files are written to provide prototypes
35785 in ISO C but not in traditional C. Many of these header files can
35786 work without change in C++ provided `__STDC__' is defined. If
35787 `__STDC__' is not defined, they will all fail, and will all need
35788 to be changed to test explicitly for C++ as well.
35790 * Deleting "empty" loops.
35792 Historically, GCC has not deleted "empty" loops under the
35793 assumption that the most likely reason you would put one in a
35794 program is to have a delay, so deleting them will not make real
35795 programs run any faster.
35797 However, the rationale here is that optimization of a nonempty loop
35798 cannot produce an empty one. This held for carefully written C
35799 compiled with less powerful optimizers but is not always the case
35800 for carefully written C++ or with more powerful optimizers. Thus
35801 GCC will remove operations from loops whenever it can determine
35802 those operations are not externally visible (apart from the time
35803 taken to execute them, of course). In case the loop can be proved
35804 to be finite, GCC will also remove the loop itself.
35806 Be aware of this when performing timing tests, for instance the
35807 following loop can be completely removed, provided
35808 `some_expression' can provably not change any global state.
35814 for (ix = 0; ix != 10000; ix++)
35815 sum += some_expression;
35818 Even though `sum' is accumulated in the loop, no use is made of
35819 that summation, so the accumulation can be removed.
35821 * Making side effects happen in the same order as in some other
35824 It is never safe to depend on the order of evaluation of side
35825 effects. For example, a function call like this may very well
35826 behave differently from one compiler to another:
35828 void func (int, int);
35833 There is no guarantee (in either the C or the C++ standard language
35834 definitions) that the increments will be evaluated in any
35835 particular order. Either increment might happen first. `func'
35836 might get the arguments `2, 3', or it might get `3, 2', or even
35839 * Making certain warnings into errors by default.
35841 Some ISO C testsuites report failure when the compiler does not
35842 produce an error message for a certain program.
35844 ISO C requires a "diagnostic" message for certain kinds of invalid
35845 programs, but a warning is defined by GCC to count as a
35846 diagnostic. If GCC produces a warning but not an error, that is
35847 correct ISO C support. If testsuites call this "failure", they
35848 should be run with the GCC option `-pedantic-errors', which will
35849 turn these warnings into errors.
35853 File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble
35855 10.11 Warning Messages and Error Messages
35856 =========================================
35858 The GNU compiler can produce two kinds of diagnostics: errors and
35859 warnings. Each kind has a different purpose:
35861 "Errors" report problems that make it impossible to compile your
35862 program. GCC reports errors with the source file name and line
35863 number where the problem is apparent.
35865 "Warnings" report other unusual conditions in your code that _may_
35866 indicate a problem, although compilation can (and does) proceed.
35867 Warning messages also report the source file name and line number,
35868 but include the text `warning:' to distinguish them from error
35871 Warnings may indicate danger points where you should check to make sure
35872 that your program really does what you intend; or the use of obsolete
35873 features; or the use of nonstandard features of GNU C or C++. Many
35874 warnings are issued only if you ask for them, with one of the `-W'
35875 options (for instance, `-Wall' requests a variety of useful warnings).
35877 GCC always tries to compile your program if possible; it never
35878 gratuitously rejects a program whose meaning is clear merely because
35879 (for instance) it fails to conform to a standard. In some cases,
35880 however, the C and C++ standards specify that certain extensions are
35881 forbidden, and a diagnostic _must_ be issued by a conforming compiler.
35882 The `-pedantic' option tells GCC to issue warnings in such cases;
35883 `-pedantic-errors' says to make them errors instead. This does not
35884 mean that _all_ non-ISO constructs get warnings or errors.
35886 *Note Options to Request or Suppress Warnings: Warning Options, for
35887 more detail on these and related command-line options.
35890 File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top
35895 Your bug reports play an essential role in making GCC reliable.
35897 When you encounter a problem, the first thing to do is to see if it is
35898 already known. *Note Trouble::. If it isn't known, then you should
35899 report the problem.
35903 * Criteria: Bug Criteria. Have you really found a bug?
35904 * Reporting: Bug Reporting. How to report a bug effectively.
35905 * Known: Trouble. Known problems.
35906 * Help: Service. Where to ask for help.
35909 File: gcc.info, Node: Bug Criteria, Next: Bug Reporting, Up: Bugs
35911 11.1 Have You Found a Bug?
35912 ==========================
35914 If you are not sure whether you have found a bug, here are some
35917 * If the compiler gets a fatal signal, for any input whatever, that
35918 is a compiler bug. Reliable compilers never crash.
35920 * If the compiler produces invalid assembly code, for any input
35921 whatever (except an `asm' statement), that is a compiler bug,
35922 unless the compiler reports errors (not just warnings) which would
35923 ordinarily prevent the assembler from being run.
35925 * If the compiler produces valid assembly code that does not
35926 correctly execute the input source code, that is a compiler bug.
35928 However, you must double-check to make sure, because you may have a
35929 program whose behavior is undefined, which happened by chance to
35930 give the desired results with another C or C++ compiler.
35932 For example, in many nonoptimizing compilers, you can write `x;'
35933 at the end of a function instead of `return x;', with the same
35934 results. But the value of the function is undefined if `return'
35935 is omitted; it is not a bug when GCC produces different results.
35937 Problems often result from expressions with two increment
35938 operators, as in `f (*p++, *p++)'. Your previous compiler might
35939 have interpreted that expression the way you intended; GCC might
35940 interpret it another way. Neither compiler is wrong. The bug is
35943 After you have localized the error to a single source line, it
35944 should be easy to check for these things. If your program is
35945 correct and well defined, you have found a compiler bug.
35947 * If the compiler produces an error message for valid input, that is
35950 * If the compiler does not produce an error message for invalid
35951 input, that is a compiler bug. However, you should note that your
35952 idea of "invalid input" might be someone else's idea of "an
35953 extension" or "support for traditional practice".
35955 * If you are an experienced user of one of the languages GCC
35956 supports, your suggestions for improvement of GCC are welcome in
35960 File: gcc.info, Node: Bug Reporting, Prev: Bug Criteria, Up: Bugs
35962 11.2 How and where to Report Bugs
35963 =================================
35965 Bugs should be reported to the bug database at
35966 `http://gcc.gnu.org/bugs.html'.
35969 File: gcc.info, Node: Service, Next: Contributing, Prev: Bugs, Up: Top
35971 12 How To Get Help with GCC
35972 ***************************
35974 If you need help installing, using or changing GCC, there are two ways
35977 * Send a message to a suitable network mailing list. First try
35978 <gcc-help@gcc.gnu.org> (for help installing or using GCC), and if
35979 that brings no response, try <gcc@gcc.gnu.org>. For help changing
35980 GCC, ask <gcc@gcc.gnu.org>. If you think you have found a bug in
35981 GCC, please report it following the instructions at *note Bug
35984 * Look in the service directory for someone who might help you for a
35985 fee. The service directory is found at
35986 `http://www.gnu.org/prep/service.html'.
35988 For further information, see `http://gcc.gnu.org/faq.html#support'.
35991 File: gcc.info, Node: Contributing, Next: Funding, Prev: Service, Up: Top
35993 13 Contributing to GCC Development
35994 **********************************
35996 If you would like to help pretest GCC releases to assure they work well,
35997 current development sources are available by SVN (see
35998 `http://gcc.gnu.org/svn.html'). Source and binary snapshots are also
35999 available for FTP; see `http://gcc.gnu.org/snapshots.html'.
36001 If you would like to work on improvements to GCC, please read the
36002 advice at these URLs:
36004 `http://gcc.gnu.org/contribute.html'
36005 `http://gcc.gnu.org/contributewhy.html'
36007 for information on how to make useful contributions and avoid
36008 duplication of effort. Suggested projects are listed at
36009 `http://gcc.gnu.org/projects/'.
36012 File: gcc.info, Node: Funding, Next: GNU Project, Prev: Contributing, Up: Top
36014 Funding Free Software
36015 *********************
36017 If you want to have more free software a few years from now, it makes
36018 sense for you to help encourage people to contribute funds for its
36019 development. The most effective approach known is to encourage
36020 commercial redistributors to donate.
36022 Users of free software systems can boost the pace of development by
36023 encouraging for-a-fee distributors to donate part of their selling price
36024 to free software developers--the Free Software Foundation, and others.
36026 The way to convince distributors to do this is to demand it and expect
36027 it from them. So when you compare distributors, judge them partly by
36028 how much they give to free software development. Show distributors
36029 they must compete to be the one who gives the most.
36031 To make this approach work, you must insist on numbers that you can
36032 compare, such as, "We will donate ten dollars to the Frobnitz project
36033 for each disk sold." Don't be satisfied with a vague promise, such as
36034 "A portion of the profits are donated," since it doesn't give a basis
36037 Even a precise fraction "of the profits from this disk" is not very
36038 meaningful, since creative accounting and unrelated business decisions
36039 can greatly alter what fraction of the sales price counts as profit.
36040 If the price you pay is $50, ten percent of the profit is probably less
36041 than a dollar; it might be a few cents, or nothing at all.
36043 Some redistributors do development work themselves. This is useful
36044 too; but to keep everyone honest, you need to inquire how much they do,
36045 and what kind. Some kinds of development make much more long-term
36046 difference than others. For example, maintaining a separate version of
36047 a program contributes very little; maintaining the standard version of a
36048 program for the whole community contributes much. Easy new ports
36049 contribute little, since someone else would surely do them; difficult
36050 ports such as adding a new CPU to the GNU Compiler Collection
36051 contribute more; major new features or packages contribute the most.
36053 By establishing the idea that supporting further development is "the
36054 proper thing to do" when distributing free software for a fee, we can
36055 assure a steady flow of resources into making more free software.
36057 Copyright (C) 1994 Free Software Foundation, Inc.
36058 Verbatim copying and redistribution of this section is permitted
36059 without royalty; alteration is not permitted.
36062 File: gcc.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top
36064 The GNU Project and GNU/Linux
36065 *****************************
36067 The GNU Project was launched in 1984 to develop a complete Unix-like
36068 operating system which is free software: the GNU system. (GNU is a
36069 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".)
36070 Variants of the GNU operating system, which use the kernel Linux, are
36071 now widely used; though these systems are often referred to as "Linux",
36072 they are more accurately called GNU/Linux systems.
36074 For more information, see:
36075 `http://www.gnu.org/'
36076 `http://www.gnu.org/gnu/linux-and-gnu.html'
36079 File: gcc.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top
36081 GNU General Public License
36082 **************************
36084 Version 3, 29 June 2007
36086 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/'
36088 Everyone is permitted to copy and distribute verbatim copies of this
36089 license document, but changing it is not allowed.
36094 The GNU General Public License is a free, copyleft license for software
36095 and other kinds of works.
36097 The licenses for most software and other practical works are designed
36098 to take away your freedom to share and change the works. By contrast,
36099 the GNU General Public License is intended to guarantee your freedom to
36100 share and change all versions of a program-to make sure it remains free
36101 software for all its users. We, the Free Software Foundation, use the
36102 GNU General Public License for most of our software; it applies also to
36103 any other work released this way by its authors. You can apply it to
36104 your programs, too.
36106 When we speak of free software, we are referring to freedom, not
36107 price. Our General Public Licenses are designed to make sure that you
36108 have the freedom to distribute copies of free software (and charge for
36109 them if you wish), that you receive source code or can get it if you
36110 want it, that you can change the software or use pieces of it in new
36111 free programs, and that you know you can do these things.
36113 To protect your rights, we need to prevent others from denying you
36114 these rights or asking you to surrender the rights. Therefore, you
36115 have certain responsibilities if you distribute copies of the software,
36116 or if you modify it: responsibilities to respect the freedom of others.
36118 For example, if you distribute copies of such a program, whether
36119 gratis or for a fee, you must pass on to the recipients the same
36120 freedoms that you received. You must make sure that they, too, receive
36121 or can get the source code. And you must show them these terms so they
36124 Developers that use the GNU GPL protect your rights with two steps:
36125 (1) assert copyright on the software, and (2) offer you this License
36126 giving you legal permission to copy, distribute and/or modify it.
36128 For the developers' and authors' protection, the GPL clearly explains
36129 that there is no warranty for this free software. For both users' and
36130 authors' sake, the GPL requires that modified versions be marked as
36131 changed, so that their problems will not be attributed erroneously to
36132 authors of previous versions.
36134 Some devices are designed to deny users access to install or run
36135 modified versions of the software inside them, although the
36136 manufacturer can do so. This is fundamentally incompatible with the
36137 aim of protecting users' freedom to change the software. The
36138 systematic pattern of such abuse occurs in the area of products for
36139 individuals to use, which is precisely where it is most unacceptable.
36140 Therefore, we have designed this version of the GPL to prohibit the
36141 practice for those products. If such problems arise substantially in
36142 other domains, we stand ready to extend this provision to those domains
36143 in future versions of the GPL, as needed to protect the freedom of
36146 Finally, every program is threatened constantly by software patents.
36147 States should not allow patents to restrict development and use of
36148 software on general-purpose computers, but in those that do, we wish to
36149 avoid the special danger that patents applied to a free program could
36150 make it effectively proprietary. To prevent this, the GPL assures that
36151 patents cannot be used to render the program non-free.
36153 The precise terms and conditions for copying, distribution and
36154 modification follow.
36156 TERMS AND CONDITIONS
36157 ====================
36161 "This License" refers to version 3 of the GNU General Public
36164 "Copyright" also means copyright-like laws that apply to other
36165 kinds of works, such as semiconductor masks.
36167 "The Program" refers to any copyrightable work licensed under this
36168 License. Each licensee is addressed as "you". "Licensees" and
36169 "recipients" may be individuals or organizations.
36171 To "modify" a work means to copy from or adapt all or part of the
36172 work in a fashion requiring copyright permission, other than the
36173 making of an exact copy. The resulting work is called a "modified
36174 version" of the earlier work or a work "based on" the earlier work.
36176 A "covered work" means either the unmodified Program or a work
36177 based on the Program.
36179 To "propagate" a work means to do anything with it that, without
36180 permission, would make you directly or secondarily liable for
36181 infringement under applicable copyright law, except executing it
36182 on a computer or modifying a private copy. Propagation includes
36183 copying, distribution (with or without modification), making
36184 available to the public, and in some countries other activities as
36187 To "convey" a work means any kind of propagation that enables other
36188 parties to make or receive copies. Mere interaction with a user
36189 through a computer network, with no transfer of a copy, is not
36192 An interactive user interface displays "Appropriate Legal Notices"
36193 to the extent that it includes a convenient and prominently visible
36194 feature that (1) displays an appropriate copyright notice, and (2)
36195 tells the user that there is no warranty for the work (except to
36196 the extent that warranties are provided), that licensees may
36197 convey the work under this License, and how to view a copy of this
36198 License. If the interface presents a list of user commands or
36199 options, such as a menu, a prominent item in the list meets this
36204 The "source code" for a work means the preferred form of the work
36205 for making modifications to it. "Object code" means any
36206 non-source form of a work.
36208 A "Standard Interface" means an interface that either is an
36209 official standard defined by a recognized standards body, or, in
36210 the case of interfaces specified for a particular programming
36211 language, one that is widely used among developers working in that
36214 The "System Libraries" of an executable work include anything,
36215 other than the work as a whole, that (a) is included in the normal
36216 form of packaging a Major Component, but which is not part of that
36217 Major Component, and (b) serves only to enable use of the work
36218 with that Major Component, or to implement a Standard Interface
36219 for which an implementation is available to the public in source
36220 code form. A "Major Component", in this context, means a major
36221 essential component (kernel, window system, and so on) of the
36222 specific operating system (if any) on which the executable work
36223 runs, or a compiler used to produce the work, or an object code
36224 interpreter used to run it.
36226 The "Corresponding Source" for a work in object code form means all
36227 the source code needed to generate, install, and (for an executable
36228 work) run the object code and to modify the work, including
36229 scripts to control those activities. However, it does not include
36230 the work's System Libraries, or general-purpose tools or generally
36231 available free programs which are used unmodified in performing
36232 those activities but which are not part of the work. For example,
36233 Corresponding Source includes interface definition files
36234 associated with source files for the work, and the source code for
36235 shared libraries and dynamically linked subprograms that the work
36236 is specifically designed to require, such as by intimate data
36237 communication or control flow between those subprograms and other
36240 The Corresponding Source need not include anything that users can
36241 regenerate automatically from other parts of the Corresponding
36244 The Corresponding Source for a work in source code form is that
36247 2. Basic Permissions.
36249 All rights granted under this License are granted for the term of
36250 copyright on the Program, and are irrevocable provided the stated
36251 conditions are met. This License explicitly affirms your unlimited
36252 permission to run the unmodified Program. The output from running
36253 a covered work is covered by this License only if the output,
36254 given its content, constitutes a covered work. This License
36255 acknowledges your rights of fair use or other equivalent, as
36256 provided by copyright law.
36258 You may make, run and propagate covered works that you do not
36259 convey, without conditions so long as your license otherwise
36260 remains in force. You may convey covered works to others for the
36261 sole purpose of having them make modifications exclusively for
36262 you, or provide you with facilities for running those works,
36263 provided that you comply with the terms of this License in
36264 conveying all material for which you do not control copyright.
36265 Those thus making or running the covered works for you must do so
36266 exclusively on your behalf, under your direction and control, on
36267 terms that prohibit them from making any copies of your
36268 copyrighted material outside their relationship with you.
36270 Conveying under any other circumstances is permitted solely under
36271 the conditions stated below. Sublicensing is not allowed; section
36272 10 makes it unnecessary.
36274 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
36276 No covered work shall be deemed part of an effective technological
36277 measure under any applicable law fulfilling obligations under
36278 article 11 of the WIPO copyright treaty adopted on 20 December
36279 1996, or similar laws prohibiting or restricting circumvention of
36282 When you convey a covered work, you waive any legal power to forbid
36283 circumvention of technological measures to the extent such
36284 circumvention is effected by exercising rights under this License
36285 with respect to the covered work, and you disclaim any intention
36286 to limit operation or modification of the work as a means of
36287 enforcing, against the work's users, your or third parties' legal
36288 rights to forbid circumvention of technological measures.
36290 4. Conveying Verbatim Copies.
36292 You may convey verbatim copies of the Program's source code as you
36293 receive it, in any medium, provided that you conspicuously and
36294 appropriately publish on each copy an appropriate copyright notice;
36295 keep intact all notices stating that this License and any
36296 non-permissive terms added in accord with section 7 apply to the
36297 code; keep intact all notices of the absence of any warranty; and
36298 give all recipients a copy of this License along with the Program.
36300 You may charge any price or no price for each copy that you convey,
36301 and you may offer support or warranty protection for a fee.
36303 5. Conveying Modified Source Versions.
36305 You may convey a work based on the Program, or the modifications to
36306 produce it from the Program, in the form of source code under the
36307 terms of section 4, provided that you also meet all of these
36310 a. The work must carry prominent notices stating that you
36311 modified it, and giving a relevant date.
36313 b. The work must carry prominent notices stating that it is
36314 released under this License and any conditions added under
36315 section 7. This requirement modifies the requirement in
36316 section 4 to "keep intact all notices".
36318 c. You must license the entire work, as a whole, under this
36319 License to anyone who comes into possession of a copy. This
36320 License will therefore apply, along with any applicable
36321 section 7 additional terms, to the whole of the work, and all
36322 its parts, regardless of how they are packaged. This License
36323 gives no permission to license the work in any other way, but
36324 it does not invalidate such permission if you have separately
36327 d. If the work has interactive user interfaces, each must display
36328 Appropriate Legal Notices; however, if the Program has
36329 interactive interfaces that do not display Appropriate Legal
36330 Notices, your work need not make them do so.
36332 A compilation of a covered work with other separate and independent
36333 works, which are not by their nature extensions of the covered
36334 work, and which are not combined with it such as to form a larger
36335 program, in or on a volume of a storage or distribution medium, is
36336 called an "aggregate" if the compilation and its resulting
36337 copyright are not used to limit the access or legal rights of the
36338 compilation's users beyond what the individual works permit.
36339 Inclusion of a covered work in an aggregate does not cause this
36340 License to apply to the other parts of the aggregate.
36342 6. Conveying Non-Source Forms.
36344 You may convey a covered work in object code form under the terms
36345 of sections 4 and 5, provided that you also convey the
36346 machine-readable Corresponding Source under the terms of this
36347 License, in one of these ways:
36349 a. Convey the object code in, or embodied in, a physical product
36350 (including a physical distribution medium), accompanied by the
36351 Corresponding Source fixed on a durable physical medium
36352 customarily used for software interchange.
36354 b. Convey the object code in, or embodied in, a physical product
36355 (including a physical distribution medium), accompanied by a
36356 written offer, valid for at least three years and valid for
36357 as long as you offer spare parts or customer support for that
36358 product model, to give anyone who possesses the object code
36359 either (1) a copy of the Corresponding Source for all the
36360 software in the product that is covered by this License, on a
36361 durable physical medium customarily used for software
36362 interchange, for a price no more than your reasonable cost of
36363 physically performing this conveying of source, or (2) access
36364 to copy the Corresponding Source from a network server at no
36367 c. Convey individual copies of the object code with a copy of
36368 the written offer to provide the Corresponding Source. This
36369 alternative is allowed only occasionally and noncommercially,
36370 and only if you received the object code with such an offer,
36371 in accord with subsection 6b.
36373 d. Convey the object code by offering access from a designated
36374 place (gratis or for a charge), and offer equivalent access
36375 to the Corresponding Source in the same way through the same
36376 place at no further charge. You need not require recipients
36377 to copy the Corresponding Source along with the object code.
36378 If the place to copy the object code is a network server, the
36379 Corresponding Source may be on a different server (operated
36380 by you or a third party) that supports equivalent copying
36381 facilities, provided you maintain clear directions next to
36382 the object code saying where to find the Corresponding Source.
36383 Regardless of what server hosts the Corresponding Source, you
36384 remain obligated to ensure that it is available for as long
36385 as needed to satisfy these requirements.
36387 e. Convey the object code using peer-to-peer transmission,
36388 provided you inform other peers where the object code and
36389 Corresponding Source of the work are being offered to the
36390 general public at no charge under subsection 6d.
36393 A separable portion of the object code, whose source code is
36394 excluded from the Corresponding Source as a System Library, need
36395 not be included in conveying the object code work.
36397 A "User Product" is either (1) a "consumer product", which means
36398 any tangible personal property which is normally used for personal,
36399 family, or household purposes, or (2) anything designed or sold for
36400 incorporation into a dwelling. In determining whether a product
36401 is a consumer product, doubtful cases shall be resolved in favor of
36402 coverage. For a particular product received by a particular user,
36403 "normally used" refers to a typical or common use of that class of
36404 product, regardless of the status of the particular user or of the
36405 way in which the particular user actually uses, or expects or is
36406 expected to use, the product. A product is a consumer product
36407 regardless of whether the product has substantial commercial,
36408 industrial or non-consumer uses, unless such uses represent the
36409 only significant mode of use of the product.
36411 "Installation Information" for a User Product means any methods,
36412 procedures, authorization keys, or other information required to
36413 install and execute modified versions of a covered work in that
36414 User Product from a modified version of its Corresponding Source.
36415 The information must suffice to ensure that the continued
36416 functioning of the modified object code is in no case prevented or
36417 interfered with solely because modification has been made.
36419 If you convey an object code work under this section in, or with,
36420 or specifically for use in, a User Product, and the conveying
36421 occurs as part of a transaction in which the right of possession
36422 and use of the User Product is transferred to the recipient in
36423 perpetuity or for a fixed term (regardless of how the transaction
36424 is characterized), the Corresponding Source conveyed under this
36425 section must be accompanied by the Installation Information. But
36426 this requirement does not apply if neither you nor any third party
36427 retains the ability to install modified object code on the User
36428 Product (for example, the work has been installed in ROM).
36430 The requirement to provide Installation Information does not
36431 include a requirement to continue to provide support service,
36432 warranty, or updates for a work that has been modified or
36433 installed by the recipient, or for the User Product in which it
36434 has been modified or installed. Access to a network may be denied
36435 when the modification itself materially and adversely affects the
36436 operation of the network or violates the rules and protocols for
36437 communication across the network.
36439 Corresponding Source conveyed, and Installation Information
36440 provided, in accord with this section must be in a format that is
36441 publicly documented (and with an implementation available to the
36442 public in source code form), and must require no special password
36443 or key for unpacking, reading or copying.
36445 7. Additional Terms.
36447 "Additional permissions" are terms that supplement the terms of
36448 this License by making exceptions from one or more of its
36449 conditions. Additional permissions that are applicable to the
36450 entire Program shall be treated as though they were included in
36451 this License, to the extent that they are valid under applicable
36452 law. If additional permissions apply only to part of the Program,
36453 that part may be used separately under those permissions, but the
36454 entire Program remains governed by this License without regard to
36455 the additional permissions.
36457 When you convey a copy of a covered work, you may at your option
36458 remove any additional permissions from that copy, or from any part
36459 of it. (Additional permissions may be written to require their own
36460 removal in certain cases when you modify the work.) You may place
36461 additional permissions on material, added by you to a covered work,
36462 for which you have or can give appropriate copyright permission.
36464 Notwithstanding any other provision of this License, for material
36465 you add to a covered work, you may (if authorized by the copyright
36466 holders of that material) supplement the terms of this License
36469 a. Disclaiming warranty or limiting liability differently from
36470 the terms of sections 15 and 16 of this License; or
36472 b. Requiring preservation of specified reasonable legal notices
36473 or author attributions in that material or in the Appropriate
36474 Legal Notices displayed by works containing it; or
36476 c. Prohibiting misrepresentation of the origin of that material,
36477 or requiring that modified versions of such material be
36478 marked in reasonable ways as different from the original
36481 d. Limiting the use for publicity purposes of names of licensors
36482 or authors of the material; or
36484 e. Declining to grant rights under trademark law for use of some
36485 trade names, trademarks, or service marks; or
36487 f. Requiring indemnification of licensors and authors of that
36488 material by anyone who conveys the material (or modified
36489 versions of it) with contractual assumptions of liability to
36490 the recipient, for any liability that these contractual
36491 assumptions directly impose on those licensors and authors.
36493 All other non-permissive additional terms are considered "further
36494 restrictions" within the meaning of section 10. If the Program as
36495 you received it, or any part of it, contains a notice stating that
36496 it is governed by this License along with a term that is a further
36497 restriction, you may remove that term. If a license document
36498 contains a further restriction but permits relicensing or
36499 conveying under this License, you may add to a covered work
36500 material governed by the terms of that license document, provided
36501 that the further restriction does not survive such relicensing or
36504 If you add terms to a covered work in accord with this section, you
36505 must place, in the relevant source files, a statement of the
36506 additional terms that apply to those files, or a notice indicating
36507 where to find the applicable terms.
36509 Additional terms, permissive or non-permissive, may be stated in
36510 the form of a separately written license, or stated as exceptions;
36511 the above requirements apply either way.
36515 You may not propagate or modify a covered work except as expressly
36516 provided under this License. Any attempt otherwise to propagate or
36517 modify it is void, and will automatically terminate your rights
36518 under this License (including any patent licenses granted under
36519 the third paragraph of section 11).
36521 However, if you cease all violation of this License, then your
36522 license from a particular copyright holder is reinstated (a)
36523 provisionally, unless and until the copyright holder explicitly
36524 and finally terminates your license, and (b) permanently, if the
36525 copyright holder fails to notify you of the violation by some
36526 reasonable means prior to 60 days after the cessation.
36528 Moreover, your license from a particular copyright holder is
36529 reinstated permanently if the copyright holder notifies you of the
36530 violation by some reasonable means, this is the first time you have
36531 received notice of violation of this License (for any work) from
36532 that copyright holder, and you cure the violation prior to 30 days
36533 after your receipt of the notice.
36535 Termination of your rights under this section does not terminate
36536 the licenses of parties who have received copies or rights from
36537 you under this License. If your rights have been terminated and
36538 not permanently reinstated, you do not qualify to receive new
36539 licenses for the same material under section 10.
36541 9. Acceptance Not Required for Having Copies.
36543 You are not required to accept this License in order to receive or
36544 run a copy of the Program. Ancillary propagation of a covered work
36545 occurring solely as a consequence of using peer-to-peer
36546 transmission to receive a copy likewise does not require
36547 acceptance. However, nothing other than this License grants you
36548 permission to propagate or modify any covered work. These actions
36549 infringe copyright if you do not accept this License. Therefore,
36550 by modifying or propagating a covered work, you indicate your
36551 acceptance of this License to do so.
36553 10. Automatic Licensing of Downstream Recipients.
36555 Each time you convey a covered work, the recipient automatically
36556 receives a license from the original licensors, to run, modify and
36557 propagate that work, subject to this License. You are not
36558 responsible for enforcing compliance by third parties with this
36561 An "entity transaction" is a transaction transferring control of an
36562 organization, or substantially all assets of one, or subdividing an
36563 organization, or merging organizations. If propagation of a
36564 covered work results from an entity transaction, each party to that
36565 transaction who receives a copy of the work also receives whatever
36566 licenses to the work the party's predecessor in interest had or
36567 could give under the previous paragraph, plus a right to
36568 possession of the Corresponding Source of the work from the
36569 predecessor in interest, if the predecessor has it or can get it
36570 with reasonable efforts.
36572 You may not impose any further restrictions on the exercise of the
36573 rights granted or affirmed under this License. For example, you
36574 may not impose a license fee, royalty, or other charge for
36575 exercise of rights granted under this License, and you may not
36576 initiate litigation (including a cross-claim or counterclaim in a
36577 lawsuit) alleging that any patent claim is infringed by making,
36578 using, selling, offering for sale, or importing the Program or any
36583 A "contributor" is a copyright holder who authorizes use under this
36584 License of the Program or a work on which the Program is based.
36585 The work thus licensed is called the contributor's "contributor
36588 A contributor's "essential patent claims" are all patent claims
36589 owned or controlled by the contributor, whether already acquired or
36590 hereafter acquired, that would be infringed by some manner,
36591 permitted by this License, of making, using, or selling its
36592 contributor version, but do not include claims that would be
36593 infringed only as a consequence of further modification of the
36594 contributor version. For purposes of this definition, "control"
36595 includes the right to grant patent sublicenses in a manner
36596 consistent with the requirements of this License.
36598 Each contributor grants you a non-exclusive, worldwide,
36599 royalty-free patent license under the contributor's essential
36600 patent claims, to make, use, sell, offer for sale, import and
36601 otherwise run, modify and propagate the contents of its
36602 contributor version.
36604 In the following three paragraphs, a "patent license" is any
36605 express agreement or commitment, however denominated, not to
36606 enforce a patent (such as an express permission to practice a
36607 patent or covenant not to sue for patent infringement). To
36608 "grant" such a patent license to a party means to make such an
36609 agreement or commitment not to enforce a patent against the party.
36611 If you convey a covered work, knowingly relying on a patent
36612 license, and the Corresponding Source of the work is not available
36613 for anyone to copy, free of charge and under the terms of this
36614 License, through a publicly available network server or other
36615 readily accessible means, then you must either (1) cause the
36616 Corresponding Source to be so available, or (2) arrange to deprive
36617 yourself of the benefit of the patent license for this particular
36618 work, or (3) arrange, in a manner consistent with the requirements
36619 of this License, to extend the patent license to downstream
36620 recipients. "Knowingly relying" means you have actual knowledge
36621 that, but for the patent license, your conveying the covered work
36622 in a country, or your recipient's use of the covered work in a
36623 country, would infringe one or more identifiable patents in that
36624 country that you have reason to believe are valid.
36626 If, pursuant to or in connection with a single transaction or
36627 arrangement, you convey, or propagate by procuring conveyance of, a
36628 covered work, and grant a patent license to some of the parties
36629 receiving the covered work authorizing them to use, propagate,
36630 modify or convey a specific copy of the covered work, then the
36631 patent license you grant is automatically extended to all
36632 recipients of the covered work and works based on it.
36634 A patent license is "discriminatory" if it does not include within
36635 the scope of its coverage, prohibits the exercise of, or is
36636 conditioned on the non-exercise of one or more of the rights that
36637 are specifically granted under this License. You may not convey a
36638 covered work if you are a party to an arrangement with a third
36639 party that is in the business of distributing software, under
36640 which you make payment to the third party based on the extent of
36641 your activity of conveying the work, and under which the third
36642 party grants, to any of the parties who would receive the covered
36643 work from you, a discriminatory patent license (a) in connection
36644 with copies of the covered work conveyed by you (or copies made
36645 from those copies), or (b) primarily for and in connection with
36646 specific products or compilations that contain the covered work,
36647 unless you entered into that arrangement, or that patent license
36648 was granted, prior to 28 March 2007.
36650 Nothing in this License shall be construed as excluding or limiting
36651 any implied license or other defenses to infringement that may
36652 otherwise be available to you under applicable patent law.
36654 12. No Surrender of Others' Freedom.
36656 If conditions are imposed on you (whether by court order,
36657 agreement or otherwise) that contradict the conditions of this
36658 License, they do not excuse you from the conditions of this
36659 License. If you cannot convey a covered work so as to satisfy
36660 simultaneously your obligations under this License and any other
36661 pertinent obligations, then as a consequence you may not convey it
36662 at all. For example, if you agree to terms that obligate you to
36663 collect a royalty for further conveying from those to whom you
36664 convey the Program, the only way you could satisfy both those
36665 terms and this License would be to refrain entirely from conveying
36668 13. Use with the GNU Affero General Public License.
36670 Notwithstanding any other provision of this License, you have
36671 permission to link or combine any covered work with a work licensed
36672 under version 3 of the GNU Affero General Public License into a
36673 single combined work, and to convey the resulting work. The terms
36674 of this License will continue to apply to the part which is the
36675 covered work, but the special requirements of the GNU Affero
36676 General Public License, section 13, concerning interaction through
36677 a network will apply to the combination as such.
36679 14. Revised Versions of this License.
36681 The Free Software Foundation may publish revised and/or new
36682 versions of the GNU General Public License from time to time.
36683 Such new versions will be similar in spirit to the present
36684 version, but may differ in detail to address new problems or
36687 Each version is given a distinguishing version number. If the
36688 Program specifies that a certain numbered version of the GNU
36689 General Public License "or any later version" applies to it, you
36690 have the option of following the terms and conditions either of
36691 that numbered version or of any later version published by the
36692 Free Software Foundation. If the Program does not specify a
36693 version number of the GNU General Public License, you may choose
36694 any version ever published by the Free Software Foundation.
36696 If the Program specifies that a proxy can decide which future
36697 versions of the GNU General Public License can be used, that
36698 proxy's public statement of acceptance of a version permanently
36699 authorizes you to choose that version for the Program.
36701 Later license versions may give you additional or different
36702 permissions. However, no additional obligations are imposed on any
36703 author or copyright holder as a result of your choosing to follow a
36706 15. Disclaimer of Warranty.
36708 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
36709 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE
36710 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS"
36711 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED,
36712 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
36713 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE
36714 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.
36715 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
36716 NECESSARY SERVICING, REPAIR OR CORRECTION.
36718 16. Limitation of Liability.
36720 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
36721 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES
36722 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU
36723 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
36724 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
36725 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA
36726 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
36727 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
36728 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF
36729 THE POSSIBILITY OF SUCH DAMAGES.
36731 17. Interpretation of Sections 15 and 16.
36733 If the disclaimer of warranty and limitation of liability provided
36734 above cannot be given local legal effect according to their terms,
36735 reviewing courts shall apply local law that most closely
36736 approximates an absolute waiver of all civil liability in
36737 connection with the Program, unless a warranty or assumption of
36738 liability accompanies a copy of the Program in return for a fee.
36741 END OF TERMS AND CONDITIONS
36742 ===========================
36744 How to Apply These Terms to Your New Programs
36745 =============================================
36747 If you develop a new program, and you want it to be of the greatest
36748 possible use to the public, the best way to achieve this is to make it
36749 free software which everyone can redistribute and change under these
36752 To do so, attach the following notices to the program. It is safest
36753 to attach them to the start of each source file to most effectively
36754 state the exclusion of warranty; and each file should have at least the
36755 "copyright" line and a pointer to where the full notice is found.
36757 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES.
36758 Copyright (C) YEAR NAME OF AUTHOR
36760 This program is free software: you can redistribute it and/or modify
36761 it under the terms of the GNU General Public License as published by
36762 the Free Software Foundation, either version 3 of the License, or (at
36763 your option) any later version.
36765 This program is distributed in the hope that it will be useful, but
36766 WITHOUT ANY WARRANTY; without even the implied warranty of
36767 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
36768 General Public License for more details.
36770 You should have received a copy of the GNU General Public License
36771 along with this program. If not, see `http://www.gnu.org/licenses/'.
36773 Also add information on how to contact you by electronic and paper
36776 If the program does terminal interaction, make it output a short
36777 notice like this when it starts in an interactive mode:
36779 PROGRAM Copyright (C) YEAR NAME OF AUTHOR
36780 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
36781 This is free software, and you are welcome to redistribute it
36782 under certain conditions; type `show c' for details.
36784 The hypothetical commands `show w' and `show c' should show the
36785 appropriate parts of the General Public License. Of course, your
36786 program's commands might be different; for a GUI interface, you would
36787 use an "about box".
36789 You should also get your employer (if you work as a programmer) or
36790 school, if any, to sign a "copyright disclaimer" for the program, if
36791 necessary. For more information on this, and how to apply and follow
36792 the GNU GPL, see `http://www.gnu.org/licenses/'.
36794 The GNU General Public License does not permit incorporating your
36795 program into proprietary programs. If your program is a subroutine
36796 library, you may consider it more useful to permit linking proprietary
36797 applications with the library. If this is what you want to do, use the
36798 GNU Lesser General Public License instead of this License. But first,
36799 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'.
36802 File: gcc.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top
36804 GNU Free Documentation License
36805 ******************************
36807 Version 1.2, November 2002
36809 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc.
36810 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
36812 Everyone is permitted to copy and distribute verbatim copies
36813 of this license document, but changing it is not allowed.
36817 The purpose of this License is to make a manual, textbook, or other
36818 functional and useful document "free" in the sense of freedom: to
36819 assure everyone the effective freedom to copy and redistribute it,
36820 with or without modifying it, either commercially or
36821 noncommercially. Secondarily, this License preserves for the
36822 author and publisher a way to get credit for their work, while not
36823 being considered responsible for modifications made by others.
36825 This License is a kind of "copyleft", which means that derivative
36826 works of the document must themselves be free in the same sense.
36827 It complements the GNU General Public License, which is a copyleft
36828 license designed for free software.
36830 We have designed this License in order to use it for manuals for
36831 free software, because free software needs free documentation: a
36832 free program should come with manuals providing the same freedoms
36833 that the software does. But this License is not limited to
36834 software manuals; it can be used for any textual work, regardless
36835 of subject matter or whether it is published as a printed book.
36836 We recommend this License principally for works whose purpose is
36837 instruction or reference.
36839 1. APPLICABILITY AND DEFINITIONS
36841 This License applies to any manual or other work, in any medium,
36842 that contains a notice placed by the copyright holder saying it
36843 can be distributed under the terms of this License. Such a notice
36844 grants a world-wide, royalty-free license, unlimited in duration,
36845 to use that work under the conditions stated herein. The
36846 "Document", below, refers to any such manual or work. Any member
36847 of the public is a licensee, and is addressed as "you". You
36848 accept the license if you copy, modify or distribute the work in a
36849 way requiring permission under copyright law.
36851 A "Modified Version" of the Document means any work containing the
36852 Document or a portion of it, either copied verbatim, or with
36853 modifications and/or translated into another language.
36855 A "Secondary Section" is a named appendix or a front-matter section
36856 of the Document that deals exclusively with the relationship of the
36857 publishers or authors of the Document to the Document's overall
36858 subject (or to related matters) and contains nothing that could
36859 fall directly within that overall subject. (Thus, if the Document
36860 is in part a textbook of mathematics, a Secondary Section may not
36861 explain any mathematics.) The relationship could be a matter of
36862 historical connection with the subject or with related matters, or
36863 of legal, commercial, philosophical, ethical or political position
36866 The "Invariant Sections" are certain Secondary Sections whose
36867 titles are designated, as being those of Invariant Sections, in
36868 the notice that says that the Document is released under this
36869 License. If a section does not fit the above definition of
36870 Secondary then it is not allowed to be designated as Invariant.
36871 The Document may contain zero Invariant Sections. If the Document
36872 does not identify any Invariant Sections then there are none.
36874 The "Cover Texts" are certain short passages of text that are
36875 listed, as Front-Cover Texts or Back-Cover Texts, in the notice
36876 that says that the Document is released under this License. A
36877 Front-Cover Text may be at most 5 words, and a Back-Cover Text may
36878 be at most 25 words.
36880 A "Transparent" copy of the Document means a machine-readable copy,
36881 represented in a format whose specification is available to the
36882 general public, that is suitable for revising the document
36883 straightforwardly with generic text editors or (for images
36884 composed of pixels) generic paint programs or (for drawings) some
36885 widely available drawing editor, and that is suitable for input to
36886 text formatters or for automatic translation to a variety of
36887 formats suitable for input to text formatters. A copy made in an
36888 otherwise Transparent file format whose markup, or absence of
36889 markup, has been arranged to thwart or discourage subsequent
36890 modification by readers is not Transparent. An image format is
36891 not Transparent if used for any substantial amount of text. A
36892 copy that is not "Transparent" is called "Opaque".
36894 Examples of suitable formats for Transparent copies include plain
36895 ASCII without markup, Texinfo input format, LaTeX input format,
36896 SGML or XML using a publicly available DTD, and
36897 standard-conforming simple HTML, PostScript or PDF designed for
36898 human modification. Examples of transparent image formats include
36899 PNG, XCF and JPG. Opaque formats include proprietary formats that
36900 can be read and edited only by proprietary word processors, SGML or
36901 XML for which the DTD and/or processing tools are not generally
36902 available, and the machine-generated HTML, PostScript or PDF
36903 produced by some word processors for output purposes only.
36905 The "Title Page" means, for a printed book, the title page itself,
36906 plus such following pages as are needed to hold, legibly, the
36907 material this License requires to appear in the title page. For
36908 works in formats which do not have any title page as such, "Title
36909 Page" means the text near the most prominent appearance of the
36910 work's title, preceding the beginning of the body of the text.
36912 A section "Entitled XYZ" means a named subunit of the Document
36913 whose title either is precisely XYZ or contains XYZ in parentheses
36914 following text that translates XYZ in another language. (Here XYZ
36915 stands for a specific section name mentioned below, such as
36916 "Acknowledgements", "Dedications", "Endorsements", or "History".)
36917 To "Preserve the Title" of such a section when you modify the
36918 Document means that it remains a section "Entitled XYZ" according
36919 to this definition.
36921 The Document may include Warranty Disclaimers next to the notice
36922 which states that this License applies to the Document. These
36923 Warranty Disclaimers are considered to be included by reference in
36924 this License, but only as regards disclaiming warranties: any other
36925 implication that these Warranty Disclaimers may have is void and
36926 has no effect on the meaning of this License.
36928 2. VERBATIM COPYING
36930 You may copy and distribute the Document in any medium, either
36931 commercially or noncommercially, provided that this License, the
36932 copyright notices, and the license notice saying this License
36933 applies to the Document are reproduced in all copies, and that you
36934 add no other conditions whatsoever to those of this License. You
36935 may not use technical measures to obstruct or control the reading
36936 or further copying of the copies you make or distribute. However,
36937 you may accept compensation in exchange for copies. If you
36938 distribute a large enough number of copies you must also follow
36939 the conditions in section 3.
36941 You may also lend copies, under the same conditions stated above,
36942 and you may publicly display copies.
36944 3. COPYING IN QUANTITY
36946 If you publish printed copies (or copies in media that commonly
36947 have printed covers) of the Document, numbering more than 100, and
36948 the Document's license notice requires Cover Texts, you must
36949 enclose the copies in covers that carry, clearly and legibly, all
36950 these Cover Texts: Front-Cover Texts on the front cover, and
36951 Back-Cover Texts on the back cover. Both covers must also clearly
36952 and legibly identify you as the publisher of these copies. The
36953 front cover must present the full title with all words of the
36954 title equally prominent and visible. You may add other material
36955 on the covers in addition. Copying with changes limited to the
36956 covers, as long as they preserve the title of the Document and
36957 satisfy these conditions, can be treated as verbatim copying in
36960 If the required texts for either cover are too voluminous to fit
36961 legibly, you should put the first ones listed (as many as fit
36962 reasonably) on the actual cover, and continue the rest onto
36965 If you publish or distribute Opaque copies of the Document
36966 numbering more than 100, you must either include a
36967 machine-readable Transparent copy along with each Opaque copy, or
36968 state in or with each Opaque copy a computer-network location from
36969 which the general network-using public has access to download
36970 using public-standard network protocols a complete Transparent
36971 copy of the Document, free of added material. If you use the
36972 latter option, you must take reasonably prudent steps, when you
36973 begin distribution of Opaque copies in quantity, to ensure that
36974 this Transparent copy will remain thus accessible at the stated
36975 location until at least one year after the last time you
36976 distribute an Opaque copy (directly or through your agents or
36977 retailers) of that edition to the public.
36979 It is requested, but not required, that you contact the authors of
36980 the Document well before redistributing any large number of
36981 copies, to give them a chance to provide you with an updated
36982 version of the Document.
36986 You may copy and distribute a Modified Version of the Document
36987 under the conditions of sections 2 and 3 above, provided that you
36988 release the Modified Version under precisely this License, with
36989 the Modified Version filling the role of the Document, thus
36990 licensing distribution and modification of the Modified Version to
36991 whoever possesses a copy of it. In addition, you must do these
36992 things in the Modified Version:
36994 A. Use in the Title Page (and on the covers, if any) a title
36995 distinct from that of the Document, and from those of
36996 previous versions (which should, if there were any, be listed
36997 in the History section of the Document). You may use the
36998 same title as a previous version if the original publisher of
36999 that version gives permission.
37001 B. List on the Title Page, as authors, one or more persons or
37002 entities responsible for authorship of the modifications in
37003 the Modified Version, together with at least five of the
37004 principal authors of the Document (all of its principal
37005 authors, if it has fewer than five), unless they release you
37006 from this requirement.
37008 C. State on the Title page the name of the publisher of the
37009 Modified Version, as the publisher.
37011 D. Preserve all the copyright notices of the Document.
37013 E. Add an appropriate copyright notice for your modifications
37014 adjacent to the other copyright notices.
37016 F. Include, immediately after the copyright notices, a license
37017 notice giving the public permission to use the Modified
37018 Version under the terms of this License, in the form shown in
37019 the Addendum below.
37021 G. Preserve in that license notice the full lists of Invariant
37022 Sections and required Cover Texts given in the Document's
37025 H. Include an unaltered copy of this License.
37027 I. Preserve the section Entitled "History", Preserve its Title,
37028 and add to it an item stating at least the title, year, new
37029 authors, and publisher of the Modified Version as given on
37030 the Title Page. If there is no section Entitled "History" in
37031 the Document, create one stating the title, year, authors,
37032 and publisher of the Document as given on its Title Page,
37033 then add an item describing the Modified Version as stated in
37034 the previous sentence.
37036 J. Preserve the network location, if any, given in the Document
37037 for public access to a Transparent copy of the Document, and
37038 likewise the network locations given in the Document for
37039 previous versions it was based on. These may be placed in
37040 the "History" section. You may omit a network location for a
37041 work that was published at least four years before the
37042 Document itself, or if the original publisher of the version
37043 it refers to gives permission.
37045 K. For any section Entitled "Acknowledgements" or "Dedications",
37046 Preserve the Title of the section, and preserve in the
37047 section all the substance and tone of each of the contributor
37048 acknowledgements and/or dedications given therein.
37050 L. Preserve all the Invariant Sections of the Document,
37051 unaltered in their text and in their titles. Section numbers
37052 or the equivalent are not considered part of the section
37055 M. Delete any section Entitled "Endorsements". Such a section
37056 may not be included in the Modified Version.
37058 N. Do not retitle any existing section to be Entitled
37059 "Endorsements" or to conflict in title with any Invariant
37062 O. Preserve any Warranty Disclaimers.
37064 If the Modified Version includes new front-matter sections or
37065 appendices that qualify as Secondary Sections and contain no
37066 material copied from the Document, you may at your option
37067 designate some or all of these sections as invariant. To do this,
37068 add their titles to the list of Invariant Sections in the Modified
37069 Version's license notice. These titles must be distinct from any
37070 other section titles.
37072 You may add a section Entitled "Endorsements", provided it contains
37073 nothing but endorsements of your Modified Version by various
37074 parties--for example, statements of peer review or that the text
37075 has been approved by an organization as the authoritative
37076 definition of a standard.
37078 You may add a passage of up to five words as a Front-Cover Text,
37079 and a passage of up to 25 words as a Back-Cover Text, to the end
37080 of the list of Cover Texts in the Modified Version. Only one
37081 passage of Front-Cover Text and one of Back-Cover Text may be
37082 added by (or through arrangements made by) any one entity. If the
37083 Document already includes a cover text for the same cover,
37084 previously added by you or by arrangement made by the same entity
37085 you are acting on behalf of, you may not add another; but you may
37086 replace the old one, on explicit permission from the previous
37087 publisher that added the old one.
37089 The author(s) and publisher(s) of the Document do not by this
37090 License give permission to use their names for publicity for or to
37091 assert or imply endorsement of any Modified Version.
37093 5. COMBINING DOCUMENTS
37095 You may combine the Document with other documents released under
37096 this License, under the terms defined in section 4 above for
37097 modified versions, provided that you include in the combination
37098 all of the Invariant Sections of all of the original documents,
37099 unmodified, and list them all as Invariant Sections of your
37100 combined work in its license notice, and that you preserve all
37101 their Warranty Disclaimers.
37103 The combined work need only contain one copy of this License, and
37104 multiple identical Invariant Sections may be replaced with a single
37105 copy. If there are multiple Invariant Sections with the same name
37106 but different contents, make the title of each such section unique
37107 by adding at the end of it, in parentheses, the name of the
37108 original author or publisher of that section if known, or else a
37109 unique number. Make the same adjustment to the section titles in
37110 the list of Invariant Sections in the license notice of the
37113 In the combination, you must combine any sections Entitled
37114 "History" in the various original documents, forming one section
37115 Entitled "History"; likewise combine any sections Entitled
37116 "Acknowledgements", and any sections Entitled "Dedications". You
37117 must delete all sections Entitled "Endorsements."
37119 6. COLLECTIONS OF DOCUMENTS
37121 You may make a collection consisting of the Document and other
37122 documents released under this License, and replace the individual
37123 copies of this License in the various documents with a single copy
37124 that is included in the collection, provided that you follow the
37125 rules of this License for verbatim copying of each of the
37126 documents in all other respects.
37128 You may extract a single document from such a collection, and
37129 distribute it individually under this License, provided you insert
37130 a copy of this License into the extracted document, and follow
37131 this License in all other respects regarding verbatim copying of
37134 7. AGGREGATION WITH INDEPENDENT WORKS
37136 A compilation of the Document or its derivatives with other
37137 separate and independent documents or works, in or on a volume of
37138 a storage or distribution medium, is called an "aggregate" if the
37139 copyright resulting from the compilation is not used to limit the
37140 legal rights of the compilation's users beyond what the individual
37141 works permit. When the Document is included in an aggregate, this
37142 License does not apply to the other works in the aggregate which
37143 are not themselves derivative works of the Document.
37145 If the Cover Text requirement of section 3 is applicable to these
37146 copies of the Document, then if the Document is less than one half
37147 of the entire aggregate, the Document's Cover Texts may be placed
37148 on covers that bracket the Document within the aggregate, or the
37149 electronic equivalent of covers if the Document is in electronic
37150 form. Otherwise they must appear on printed covers that bracket
37151 the whole aggregate.
37155 Translation is considered a kind of modification, so you may
37156 distribute translations of the Document under the terms of section
37157 4. Replacing Invariant Sections with translations requires special
37158 permission from their copyright holders, but you may include
37159 translations of some or all Invariant Sections in addition to the
37160 original versions of these Invariant Sections. You may include a
37161 translation of this License, and all the license notices in the
37162 Document, and any Warranty Disclaimers, provided that you also
37163 include the original English version of this License and the
37164 original versions of those notices and disclaimers. In case of a
37165 disagreement between the translation and the original version of
37166 this License or a notice or disclaimer, the original version will
37169 If a section in the Document is Entitled "Acknowledgements",
37170 "Dedications", or "History", the requirement (section 4) to
37171 Preserve its Title (section 1) will typically require changing the
37176 You may not copy, modify, sublicense, or distribute the Document
37177 except as expressly provided for under this License. Any other
37178 attempt to copy, modify, sublicense or distribute the Document is
37179 void, and will automatically terminate your rights under this
37180 License. However, parties who have received copies, or rights,
37181 from you under this License will not have their licenses
37182 terminated so long as such parties remain in full compliance.
37184 10. FUTURE REVISIONS OF THIS LICENSE
37186 The Free Software Foundation may publish new, revised versions of
37187 the GNU Free Documentation License from time to time. Such new
37188 versions will be similar in spirit to the present version, but may
37189 differ in detail to address new problems or concerns. See
37190 `http://www.gnu.org/copyleft/'.
37192 Each version of the License is given a distinguishing version
37193 number. If the Document specifies that a particular numbered
37194 version of this License "or any later version" applies to it, you
37195 have the option of following the terms and conditions either of
37196 that specified version or of any later version that has been
37197 published (not as a draft) by the Free Software Foundation. If
37198 the Document does not specify a version number of this License,
37199 you may choose any version ever published (not as a draft) by the
37200 Free Software Foundation.
37202 ADDENDUM: How to use this License for your documents
37203 ====================================================
37205 To use this License in a document you have written, include a copy of
37206 the License in the document and put the following copyright and license
37207 notices just after the title page:
37209 Copyright (C) YEAR YOUR NAME.
37210 Permission is granted to copy, distribute and/or modify this document
37211 under the terms of the GNU Free Documentation License, Version 1.2
37212 or any later version published by the Free Software Foundation;
37213 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
37214 Texts. A copy of the license is included in the section entitled ``GNU
37215 Free Documentation License''.
37217 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts,
37218 replace the "with...Texts." line with this:
37220 with the Invariant Sections being LIST THEIR TITLES, with
37221 the Front-Cover Texts being LIST, and with the Back-Cover Texts
37224 If you have Invariant Sections without Cover Texts, or some other
37225 combination of the three, merge those two alternatives to suit the
37228 If your document contains nontrivial examples of program code, we
37229 recommend releasing these examples in parallel under your choice of
37230 free software license, such as the GNU General Public License, to
37231 permit their use in free software.
37234 File: gcc.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top
37236 Contributors to GCC
37237 *******************
37239 The GCC project would like to thank its many contributors. Without
37240 them the project would not have been nearly as successful as it has
37241 been. Any omissions in this list are accidental. Feel free to contact
37242 <law@redhat.com> or <gerald@pfeifer.com> if you have been left out or
37243 some of your contributions are not listed. Please keep this list in
37244 alphabetical order.
37246 * Analog Devices helped implement the support for complex data types
37249 * John David Anglin for threading-related fixes and improvements to
37250 libstdc++-v3, and the HP-UX port.
37252 * James van Artsdalen wrote the code that makes efficient use of the
37253 Intel 80387 register stack.
37255 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta
37258 * Alasdair Baird for various bug fixes.
37260 * Giovanni Bajo for analyzing lots of complicated C++ problem
37263 * Peter Barada for his work to improve code generation for new
37266 * Gerald Baumgartner added the signature extension to the C++ front
37269 * Godmar Back for his Java improvements and encouragement.
37271 * Scott Bambrough for help porting the Java compiler.
37273 * Wolfgang Bangerth for processing tons of bug reports.
37275 * Jon Beniston for his Microsoft Windows port of Java.
37277 * Daniel Berlin for better DWARF2 support, faster/better
37278 optimizations, improved alias analysis, plus migrating GCC to
37281 * Geoff Berry for his Java object serialization work and various
37284 * Uros Bizjak for the implementation of x87 math built-in functions
37285 and for various middle end and i386 back end improvements and bug
37288 * Eric Blake for helping to make GCJ and libgcj conform to the
37291 * Janne Blomqvist for contributions to GNU Fortran.
37293 * Segher Boessenkool for various fixes.
37295 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and
37298 * Neil Booth for work on cpplib, lang hooks, debug hooks and other
37299 miscellaneous clean-ups.
37301 * Steven Bosscher for integrating the GNU Fortran front end into GCC
37302 and for contributing to the tree-ssa branch.
37304 * Eric Botcazou for fixing middle- and backend bugs left and right.
37306 * Per Bothner for his direction via the steering committee and
37307 various improvements to the infrastructure for supporting new
37308 languages. Chill front end implementation. Initial
37309 implementations of cpplib, fix-header, config.guess, libio, and
37310 past C++ library (libg++) maintainer. Dreaming up, designing and
37311 implementing much of GCJ.
37313 * Devon Bowen helped port GCC to the Tahoe.
37315 * Don Bowman for mips-vxworks contributions.
37317 * Dave Brolley for work on cpplib and Chill.
37319 * Paul Brook for work on the ARM architecture and maintaining GNU
37322 * Robert Brown implemented the support for Encore 32000 systems.
37324 * Christian Bruel for improvements to local store elimination.
37326 * Herman A.J. ten Brugge for various fixes.
37328 * Joerg Brunsmann for Java compiler hacking and help with the GCJ
37331 * Joe Buck for his direction via the steering committee.
37333 * Craig Burley for leadership of the G77 Fortran effort.
37335 * Stephan Buys for contributing Doxygen notes for libstdc++.
37337 * Paolo Carlini for libstdc++ work: lots of efficiency improvements
37338 to the C++ strings, streambufs and formatted I/O, hard detective
37339 work on the frustrating localization issues, and keeping up with
37340 the problem reports.
37342 * John Carr for his alias work, SPARC hacking, infrastructure
37343 improvements, previous contributions to the steering committee,
37344 loop optimizations, etc.
37346 * Stephane Carrez for 68HC11 and 68HC12 ports.
37348 * Steve Chamberlain for support for the Renesas SH and H8 processors
37349 and the PicoJava processor, and for GCJ config fixes.
37351 * Glenn Chambers for help with the GCJ FAQ.
37353 * John-Marc Chandonia for various libgcj patches.
37355 * Scott Christley for his Objective-C contributions.
37357 * Eric Christopher for his Java porting help and clean-ups.
37359 * Branko Cibej for more warning contributions.
37361 * The GNU Classpath project for all of their merged runtime code.
37363 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and
37364 other random hacking.
37366 * Michael Cook for libstdc++ cleanup patches to reduce warnings.
37368 * R. Kelley Cook for making GCC buildable from a read-only directory
37369 as well as other miscellaneous build process and documentation
37372 * Ralf Corsepius for SH testing and minor bug fixing.
37374 * Stan Cox for care and feeding of the x86 port and lots of behind
37375 the scenes hacking.
37377 * Alex Crain provided changes for the 3b1.
37379 * Ian Dall for major improvements to the NS32k port.
37381 * Paul Dale for his work to add uClinux platform support to the m68k
37384 * Dario Dariol contributed the four varieties of sample programs
37385 that print a copy of their source.
37387 * Russell Davidson for fstream and stringstream fixes in libstdc++.
37389 * Bud Davis for work on the G77 and GNU Fortran compilers.
37391 * Mo DeJong for GCJ and libgcj bug fixes.
37393 * DJ Delorie for the DJGPP port, build and libiberty maintenance,
37394 various bug fixes, and the M32C port.
37396 * Arnaud Desitter for helping to debug GNU Fortran.
37398 * Gabriel Dos Reis for contributions to G++, contributions and
37399 maintenance of GCC diagnostics infrastructure, libstdc++-v3,
37400 including `valarray<>', `complex<>', maintaining the numerics
37401 library (including that pesky `<limits>' :-) and keeping
37402 up-to-date anything to do with numbers.
37404 * Ulrich Drepper for his work on glibc, testing of GCC using glibc,
37405 ISO C99 support, CFG dumping support, etc., plus support of the
37406 C++ runtime libraries including for all kinds of C interface
37407 issues, contributing and maintaining `complex<>', sanity checking
37408 and disbursement, configuration architecture, libio maintenance,
37409 and early math work.
37411 * Zdenek Dvorak for a new loop unroller and various fixes.
37413 * Richard Earnshaw for his ongoing work with the ARM.
37415 * David Edelsohn for his direction via the steering committee,
37416 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa
37417 loop changes, doing the entire AIX port of libstdc++ with his bare
37418 hands, and for ensuring GCC properly keeps working on AIX.
37420 * Kevin Ediger for the floating point formatting of num_put::do_put
37423 * Phil Edwards for libstdc++ work including configuration hackery,
37424 documentation maintainer, chief breaker of the web pages, the
37425 occasional iostream bug fix, and work on shared library symbol
37428 * Paul Eggert for random hacking all over GCC.
37430 * Mark Elbrecht for various DJGPP improvements, and for libstdc++
37431 configuration support for locales and fstream-related fixes.
37433 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and
37436 * Christian Ehrhardt for dealing with bug reports.
37438 * Ben Elliston for his work to move the Objective-C runtime into its
37439 own subdirectory and for his work on autoconf.
37441 * Revital Eres for work on the PowerPC 750CL port.
37443 * Marc Espie for OpenBSD support.
37445 * Doug Evans for much of the global optimization framework, arc,
37446 m32r, and SPARC work.
37448 * Christopher Faylor for his work on the Cygwin port and for caring
37449 and feeding the gcc.gnu.org box and saving its users tons of spam.
37451 * Fred Fish for BeOS support and Ada fixes.
37453 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ.
37455 * Peter Gerwinski for various bug fixes and the Pascal front end.
37457 * Kaveh R. Ghazi for his direction via the steering committee,
37458 amazing work to make `-W -Wall -W* -Werror' useful, and
37459 continuously testing GCC on a plethora of platforms. Kaveh
37460 extends his gratitude to the CAIP Center at Rutgers University for
37461 providing him with computing resources to work on Free Software
37462 since the late 1980s.
37464 * John Gilmore for a donation to the FSF earmarked improving GNU
37467 * Judy Goldberg for c++ contributions.
37469 * Torbjorn Granlund for various fixes and the c-torture testsuite,
37470 multiply- and divide-by-constant optimization, improved long long
37471 support, improved leaf function register allocation, and his
37472 direction via the steering committee.
37474 * Anthony Green for his `-Os' contributions and Java front end work.
37476 * Stu Grossman for gdb hacking, allowing GCJ developers to debug
37479 * Michael K. Gschwind contributed the port to the PDP-11.
37481 * Ron Guilmette implemented the `protoize' and `unprotoize' tools,
37482 the support for Dwarf symbolic debugging information, and much of
37483 the support for System V Release 4. He has also worked heavily on
37484 the Intel 386 and 860 support.
37486 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload
37489 * Bruno Haible for improvements in the runtime overhead for EH, new
37490 warnings and assorted bug fixes.
37492 * Andrew Haley for his amazing Java compiler and library efforts.
37494 * Chris Hanson assisted in making GCC work on HP-UX for the 9000
37497 * Michael Hayes for various thankless work he's done trying to get
37498 the c30/c40 ports functional. Lots of loop and unroll
37499 improvements and fixes.
37501 * Dara Hazeghi for wading through myriads of target-specific bug
37504 * Kate Hedstrom for staking the G77 folks with an initial testsuite.
37506 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64
37507 work, loop opts, and generally fixing lots of old problems we've
37508 ignored for years, flow rewrite and lots of further stuff,
37509 including reviewing tons of patches.
37511 * Aldy Hernandez for working on the PowerPC port, SIMD support, and
37514 * Nobuyuki Hikichi of Software Research Associates, Tokyo,
37515 contributed the support for the Sony NEWS machine.
37517 * Kazu Hirata for caring and feeding the Renesas H8/300 port and
37520 * Katherine Holcomb for work on GNU Fortran.
37522 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots
37523 of testing and bug fixing, particularly of GCC configury code.
37525 * Steve Holmgren for MachTen patches.
37527 * Jan Hubicka for his x86 port improvements.
37529 * Falk Hueffner for working on C and optimization bug reports.
37531 * Bernardo Innocenti for his m68k work, including merging of
37532 ColdFire improvements and uClinux support.
37534 * Christian Iseli for various bug fixes.
37536 * Kamil Iskra for general m68k hacking.
37538 * Lee Iverson for random fixes and MIPS testing.
37540 * Andreas Jaeger for testing and benchmarking of GCC and various bug
37543 * Jakub Jelinek for his SPARC work and sibling call optimizations as
37544 well as lots of bug fixes and test cases, and for improving the
37547 * Janis Johnson for ia64 testing and fixes, her quality improvement
37548 sidetracks, and web page maintenance.
37550 * Kean Johnston for SCO OpenServer support and various fixes.
37552 * Tim Josling for the sample language treelang based originally on
37553 Richard Kenner's "toy" language.
37555 * Nicolai Josuttis for additional libstdc++ documentation.
37557 * Klaus Kaempf for his ongoing work to make alpha-vms a viable
37560 * Steven G. Kargl for work on GNU Fortran.
37562 * David Kashtan of SRI adapted GCC to VMS.
37564 * Ryszard Kabatek for many, many libstdc++ bug fixes and
37565 optimizations of strings, especially member functions, and for
37568 * Geoffrey Keating for his ongoing work to make the PPC work for
37569 GNU/Linux and his automatic regression tester.
37571 * Brendan Kehoe for his ongoing work with G++ and for a lot of early
37572 work in just about every part of libstdc++.
37574 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the
37577 * Richard Kenner of the New York University Ultracomputer Research
37578 Laboratory wrote the machine descriptions for the AMD 29000, the
37579 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the
37580 support for instruction attributes. He also made changes to
37581 better support RISC processors including changes to common
37582 subexpression elimination, strength reduction, function calling
37583 sequence handling, and condition code support, in addition to
37584 generalizing the code for frame pointer elimination and delay slot
37585 scheduling. Richard Kenner was also the head maintainer of GCC
37588 * Mumit Khan for various contributions to the Cygwin and Mingw32
37589 ports and maintaining binary releases for Microsoft Windows hosts,
37590 and for massive libstdc++ porting work to Cygwin/Mingw32.
37592 * Robin Kirkham for cpu32 support.
37594 * Mark Klein for PA improvements.
37596 * Thomas Koenig for various bug fixes.
37598 * Bruce Korb for the new and improved fixincludes code.
37600 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3
37603 * Charles LaBrec contributed the support for the Integrated Solutions
37606 * Asher Langton and Mike Kumbera for contributing Cray pointer
37607 support to GNU Fortran, and for other GNU Fortran improvements.
37609 * Jeff Law for his direction via the steering committee,
37610 coordinating the entire egcs project and GCC 2.95, rolling out
37611 snapshots and releases, handling merges from GCC2, reviewing tons
37612 of patches that might have fallen through the cracks else, and
37613 random but extensive hacking.
37615 * Marc Lehmann for his direction via the steering committee and
37616 helping with analysis and improvements of x86 performance.
37618 * Victor Leikehman for work on GNU Fortran.
37620 * Ted Lemon wrote parts of the RTL reader and printer.
37622 * Kriang Lerdsuwanakij for C++ improvements including template as
37623 template parameter support, and many C++ fixes.
37625 * Warren Levy for tremendous work on libgcj (Java Runtime Library)
37626 and random work on the Java front end.
37628 * Alain Lichnewsky ported GCC to the MIPS CPU.
37630 * Oskar Liljeblad for hacking on AWT and his many Java bug reports
37633 * Robert Lipe for OpenServer support, new testsuites, testing, etc.
37635 * Chen Liqin for various S+core related fixes/improvement, and for
37636 maintaining the S+core port.
37638 * Weiwen Liu for testing and various bug fixes.
37640 * Manuel Lo'pez-Iba'n~ez for improving `-Wconversion' and many other
37641 diagnostics fixes and improvements.
37643 * Dave Love for his ongoing work with the Fortran front end and
37646 * Martin von Lo"wis for internal consistency checking infrastructure,
37647 various C++ improvements including namespace support, and tons of
37648 assistance with libstdc++/compiler merges.
37650 * H.J. Lu for his previous contributions to the steering committee,
37651 many x86 bug reports, prototype patches, and keeping the GNU/Linux
37654 * Greg McGary for random fixes and (someday) bounded pointers.
37656 * Andrew MacLeod for his ongoing work in building a real EH system,
37657 various code generation improvements, work on the global
37660 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC
37661 hacking improvements to compile-time performance, overall
37662 knowledge and direction in the area of instruction scheduling, and
37663 design and implementation of the automaton based instruction
37666 * Bob Manson for his behind the scenes work on dejagnu.
37668 * Philip Martin for lots of libstdc++ string and vector iterator
37669 fixes and improvements, and string clean up and testsuites.
37671 * All of the Mauve project contributors, for Java test code.
37673 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements.
37675 * Adam Megacz for his work on the Microsoft Windows port of GCJ.
37677 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS,
37678 powerpc, haifa, ECOFF debug support, and other assorted hacking.
37680 * Jason Merrill for his direction via the steering committee and
37681 leading the G++ effort.
37683 * Martin Michlmayr for testing GCC on several architectures using the
37684 entire Debian archive.
37686 * David Miller for his direction via the steering committee, lots of
37687 SPARC work, improvements in jump.c and interfacing with the Linux
37690 * Gary Miller ported GCC to Charles River Data Systems machines.
37692 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning
37693 the entire libstdc++ testsuite namespace-compatible.
37695 * Mark Mitchell for his direction via the steering committee,
37696 mountains of C++ work, load/store hoisting out of loops, alias
37697 analysis improvements, ISO C `restrict' support, and serving as
37698 release manager for GCC 3.x.
37700 * Alan Modra for various GNU/Linux bits and testing.
37702 * Toon Moene for his direction via the steering committee, Fortran
37703 maintenance, and his ongoing work to make us make Fortran run fast.
37705 * Jason Molenda for major help in the care and feeding of all the
37706 services on the gcc.gnu.org (formerly egcs.cygnus.com)
37707 machine--mail, web services, ftp services, etc etc. Doing all
37708 this work on scrap paper and the backs of envelopes would have
37711 * Catherine Moore for fixing various ugly problems we have sent her
37712 way, including the haifa bug which was killing the Alpha & PowerPC
37715 * Mike Moreton for his various Java patches.
37717 * David Mosberger-Tang for various Alpha improvements, and for the
37718 initial IA-64 port.
37720 * Stephen Moshier contributed the floating point emulator that
37721 assists in cross-compilation and permits support for floating
37722 point numbers wider than 64 bits and for ISO C99 support.
37724 * Bill Moyer for his behind the scenes work on various issues.
37726 * Philippe De Muyter for his work on the m68k port.
37728 * Joseph S. Myers for his work on the PDP-11 port, format checking
37729 and ISO C99 support, and continuous emphasis on (and contributions
37732 * Nathan Myers for his work on libstdc++-v3: architecture and
37733 authorship through the first three snapshots, including
37734 implementation of locale infrastructure, string, shadow C headers,
37735 and the initial project documentation (DESIGN, CHECKLIST, and so
37736 forth). Later, more work on MT-safe string and shadow headers.
37738 * Felix Natter for documentation on porting libstdc++.
37740 * Nathanael Nerode for cleaning up the configuration/build process.
37742 * NeXT, Inc. donated the front end that supports the Objective-C
37745 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to
37746 the search engine setup, various documentation fixes and other
37749 * Geoff Noer for his work on getting cygwin native builds working.
37751 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance
37752 tracking web pages and assorted fixes.
37754 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64,
37755 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and
37756 related infrastructure improvements.
37758 * Alexandre Oliva for various build infrastructure improvements,
37759 scripts and amazing testing work, including keeping libtool issues
37762 * Stefan Olsson for work on mt_alloc.
37764 * Melissa O'Neill for various NeXT fixes.
37766 * Rainer Orth for random MIPS work, including improvements to GCC's
37767 o32 ABI support, improvements to dejagnu's MIPS support, Java
37768 configuration clean-ups and porting work, etc.
37770 * Hartmut Penner for work on the s390 port.
37772 * Paul Petersen wrote the machine description for the Alliant FX/8.
37774 * Alexandre Petit-Bianco for implementing much of the Java compiler
37775 and continued Java maintainership.
37777 * Matthias Pfaller for major improvements to the NS32k port.
37779 * Gerald Pfeifer for his direction via the steering committee,
37780 pointing out lots of problems we need to solve, maintenance of the
37781 web pages, and taking care of documentation maintenance in general.
37783 * Andrew Pinski for processing bug reports by the dozen.
37785 * Ovidiu Predescu for his work on the Objective-C front end and
37788 * Jerry Quinn for major performance improvements in C++ formatted
37791 * Ken Raeburn for various improvements to checker, MIPS ports and
37792 various cleanups in the compiler.
37794 * Rolf W. Rasmussen for hacking on AWT.
37796 * David Reese of Sun Microsystems contributed to the Solaris on
37799 * Volker Reichelt for keeping up with the problem reports.
37801 * Joern Rennecke for maintaining the sh port, loop, regmove & reload
37804 * Loren J. Rittle for improvements to libstdc++-v3 including the
37805 FreeBSD port, threading fixes, thread-related configury changes,
37806 critical threading documentation, and solutions to really tricky
37807 I/O problems, as well as keeping GCC properly working on FreeBSD
37808 and continuous testing.
37810 * Craig Rodrigues for processing tons of bug reports.
37812 * Ola Ro"nnerup for work on mt_alloc.
37814 * Gavin Romig-Koch for lots of behind the scenes MIPS work.
37816 * David Ronis inspired and encouraged Craig to rewrite the G77
37817 documentation in texinfo format by contributing a first pass at a
37818 translation of the old `g77-0.5.16/f/DOC' file.
37820 * Ken Rose for fixes to GCC's delay slot filling code.
37822 * Paul Rubin wrote most of the preprocessor.
37824 * Pe'tur Runo'lfsson for major performance improvements in C++
37825 formatted I/O and large file support in C++ filebuf.
37827 * Chip Salzenberg for libstdc++ patches and improvements to locales,
37828 traits, Makefiles, libio, libtool hackery, and "long long" support.
37830 * Juha Sarlin for improvements to the H8 code generator.
37832 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series
37835 * Roger Sayle for improvements to constant folding and GCC's RTL
37836 optimizers as well as for fixing numerous bugs.
37838 * Bradley Schatz for his work on the GCJ FAQ.
37840 * Peter Schauer wrote the code to allow debugging to work on the
37843 * William Schelter did most of the work on the Intel 80386 support.
37845 * Tobias Schlu"ter for work on GNU Fortran.
37847 * Bernd Schmidt for various code generation improvements and major
37848 work in the reload pass as well a serving as release manager for
37851 * Peter Schmid for constant testing of libstdc++--especially
37852 application testing, going above and beyond what was requested for
37853 the release criteria--and libstdc++ header file tweaks.
37855 * Jason Schroeder for jcf-dump patches.
37857 * Andreas Schwab for his work on the m68k port.
37859 * Lars Segerlund for work on GNU Fortran.
37861 * Joel Sherrill for his direction via the steering committee, RTEMS
37862 contributions and RTEMS testing.
37864 * Nathan Sidwell for many C++ fixes/improvements.
37866 * Jeffrey Siegal for helping RMS with the original design of GCC,
37867 some code which handles the parse tree and RTL data structures,
37868 constant folding and help with the original VAX & m68k ports.
37870 * Kenny Simpson for prompting libstdc++ fixes due to defect reports
37871 from the LWG (thereby keeping GCC in line with updates from the
37874 * Franz Sirl for his ongoing work with making the PPC port stable
37877 * Andrey Slepuhin for assorted AIX hacking.
37879 * Trevor Smigiel for contributing the SPU port.
37881 * Christopher Smith did the port for Convex machines.
37883 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports.
37885 * Randy Smith finished the Sun FPA support.
37887 * Scott Snyder for queue, iterator, istream, and string fixes and
37888 libstdc++ testsuite entries. Also for providing the patch to G77
37889 to add rudimentary support for `INTEGER*1', `INTEGER*2', and
37892 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique.
37894 * Richard Stallman, for writing the original GCC and launching the
37897 * Jan Stein of the Chalmers Computer Society provided support for
37898 Genix, as well as part of the 32000 machine description.
37900 * Nigel Stephens for various mips16 related fixes/improvements.
37902 * Jonathan Stone wrote the machine description for the Pyramid
37905 * Graham Stott for various infrastructure improvements.
37907 * John Stracke for his Java HTTP protocol fixes.
37909 * Mike Stump for his Elxsi port, G++ contributions over the years
37910 and more recently his vxworks contributions
37912 * Jeff Sturm for Java porting help, bug fixes, and encouragement.
37914 * Shigeya Suzuki for this fixes for the bsdi platforms.
37916 * Ian Lance Taylor for his mips16 work, general configury hacking,
37919 * Holger Teutsch provided the support for the Clipper CPU.
37921 * Gary Thomas for his ongoing work to make the PPC work for
37924 * Philipp Thomas for random bug fixes throughout the compiler
37926 * Jason Thorpe for thread support in libstdc++ on NetBSD.
37928 * Kresten Krab Thorup wrote the run time support for the Objective-C
37929 language and the fantastic Java bytecode interpreter.
37931 * Michael Tiemann for random bug fixes, the first instruction
37932 scheduler, initial C++ support, function integration, NS32k, SPARC
37933 and M88k machine description work, delay slot scheduling.
37935 * Andreas Tobler for his work porting libgcj to Darwin.
37937 * Teemu Torma for thread safe exception handling support.
37939 * Leonard Tower wrote parts of the parser, RTL generator, and RTL
37940 definitions, and of the VAX machine description.
37942 * Tom Tromey for internationalization support and for his many Java
37943 contributions and libgcj maintainership.
37945 * Lassi Tuura for improvements to config.guess to determine HP
37948 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes.
37950 * Andy Vaught for the design and initial implementation of the GNU
37953 * Brent Verner for work with the libstdc++ cshadow files and their
37954 associated configure steps.
37956 * Todd Vierling for contributions for NetBSD ports.
37958 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML
37961 * Dean Wakerley for converting the install documentation from HTML
37962 to texinfo in time for GCC 3.0.
37964 * Krister Walfridsson for random bug fixes.
37966 * Feng Wang for contributions to GNU Fortran.
37968 * Stephen M. Webb for time and effort on making libstdc++ shadow
37969 files work with the tricky Solaris 8+ headers, and for pushing the
37970 build-time header tree.
37972 * John Wehle for various improvements for the x86 code generator,
37973 related infrastructure improvements to help x86 code generation,
37974 value range propagation and other work, WE32k port.
37976 * Ulrich Weigand for work on the s390 port.
37978 * Zack Weinberg for major work on cpplib and various other bug fixes.
37980 * Matt Welsh for help with Linux Threads support in GCJ.
37982 * Urban Widmark for help fixing java.io.
37984 * Mark Wielaard for new Java library code and his work integrating
37987 * Dale Wiles helped port GCC to the Tahoe.
37989 * Bob Wilson from Tensilica, Inc. for the Xtensa port.
37991 * Jim Wilson for his direction via the steering committee, tackling
37992 hard problems in various places that nobody else wanted to work
37993 on, strength reduction and other loop optimizations.
37995 * Paul Woegerer and Tal Agmon for the CRX port.
37997 * Carlo Wood for various fixes.
37999 * Tom Wood for work on the m88k port.
38001 * Canqun Yang for work on GNU Fortran.
38003 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine
38004 description for the Tron architecture (specifically, the Gmicro).
38006 * Kevin Zachmann helped port GCC to the Tahoe.
38008 * Ayal Zaks for Swing Modulo Scheduling (SMS).
38010 * Xiaoqiang Zhang for work on GNU Fortran.
38012 * Gilles Zunino for help porting Java to Irix.
38015 The following people are recognized for their contributions to GNAT,
38016 the Ada front end of GCC:
38019 * Romain Berrendonner
38069 * Hristian Kirtchev
38112 The following people are recognized for their contributions of new
38113 features, bug reports, testing and integration of classpath/libgcj for
38115 * Lillian Angel for `JTree' implementation and lots Free Swing
38116 additions and bug fixes.
38118 * Wolfgang Baer for `GapContent' bug fixes.
38120 * Anthony Balkissoon for `JList', Free Swing 1.5 updates and mouse
38121 event fixes, lots of Free Swing work including `JTable' editing.
38123 * Stuart Ballard for RMI constant fixes.
38125 * Goffredo Baroncelli for `HTTPURLConnection' fixes.
38127 * Gary Benson for `MessageFormat' fixes.
38129 * Daniel Bonniot for `Serialization' fixes.
38131 * Chris Burdess for lots of gnu.xml and http protocol fixes, `StAX'
38132 and `DOM xml:id' support.
38134 * Ka-Hing Cheung for `TreePath' and `TreeSelection' fixes.
38136 * Archie Cobbs for build fixes, VM interface updates,
38137 `URLClassLoader' updates.
38139 * Kelley Cook for build fixes.
38141 * Martin Cordova for Suggestions for better `SocketTimeoutException'.
38143 * David Daney for `BitSet' bug fixes, `HttpURLConnection' rewrite
38146 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo
38147 2D support. Lots of imageio framework additions, lots of AWT and
38148 Free Swing bug fixes.
38150 * Jeroen Frijters for `ClassLoader' and nio cleanups, serialization
38151 fixes, better `Proxy' support, bug fixes and IKVM integration.
38153 * Santiago Gala for `AccessControlContext' fixes.
38155 * Nicolas Geoffray for `VMClassLoader' and `AccessController'
38158 * David Gilbert for `basic' and `metal' icon and plaf support and
38159 lots of documenting, Lots of Free Swing and metal theme additions.
38160 `MetalIconFactory' implementation.
38162 * Anthony Green for `MIDI' framework, `ALSA' and `DSSI' providers.
38164 * Andrew Haley for `Serialization' and `URLClassLoader' fixes, gcj
38167 * Kim Ho for `JFileChooser' implementation.
38169 * Andrew John Hughes for `Locale' and net fixes, URI RFC2986
38170 updates, `Serialization' fixes, `Properties' XML support and
38171 generic branch work, VMIntegration guide update.
38173 * Bastiaan Huisman for `TimeZone' bug fixing.
38175 * Andreas Jaeger for mprec updates.
38177 * Paul Jenner for better `-Werror' support.
38179 * Ito Kazumitsu for `NetworkInterface' implementation and updates.
38181 * Roman Kennke for `BoxLayout', `GrayFilter' and `SplitPane', plus
38182 bug fixes all over. Lots of Free Swing work including styled text.
38184 * Simon Kitching for `String' cleanups and optimization suggestions.
38186 * Michael Koch for configuration fixes, `Locale' updates, bug and
38189 * Guilhem Lavaux for configuration, thread and channel fixes and
38190 Kaffe integration. JCL native `Pointer' updates. Logger bug fixes.
38192 * David Lichteblau for JCL support library global/local reference
38195 * Aaron Luchko for JDWP updates and documentation fixes.
38197 * Ziga Mahkovec for `Graphics2D' upgraded to Cairo 0.5 and new regex
38200 * Sven de Marothy for BMP imageio support, CSS and `TextLayout'
38201 fixes. `GtkImage' rewrite, 2D, awt, free swing and date/time fixes
38202 and implementing the Qt4 peers.
38204 * Casey Marshall for crypto algorithm fixes, `FileChannel' lock,
38205 `SystemLogger' and `FileHandler' rotate implementations, NIO
38206 `FileChannel.map' support, security and policy updates.
38208 * Bryce McKinlay for RMI work.
38210 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus
38211 testing and documenting.
38213 * Kalle Olavi Niemitalo for build fixes.
38215 * Rainer Orth for build fixes.
38217 * Andrew Overholt for `File' locking fixes.
38219 * Ingo Proetel for `Image', `Logger' and `URLClassLoader' updates.
38221 * Olga Rodimina for `MenuSelectionManager' implementation.
38223 * Jan Roehrich for `BasicTreeUI' and `JTree' fixes.
38225 * Julian Scheid for documentation updates and gjdoc support.
38227 * Christian Schlichtherle for zip fixes and cleanups.
38229 * Robert Schuster for documentation updates and beans fixes,
38230 `TreeNode' enumerations and `ActionCommand' and various fixes, XML
38231 and URL, AWT and Free Swing bug fixes.
38233 * Keith Seitz for lots of JDWP work.
38235 * Christian Thalinger for 64-bit cleanups, Configuration and VM
38236 interface fixes and `CACAO' integration, `fdlibm' updates.
38238 * Gael Thomas for `VMClassLoader' boot packages support suggestions.
38240 * Andreas Tobler for Darwin and Solaris testing and fixing, `Qt4'
38241 support for Darwin/OS X, `Graphics2D' support, `gtk+' updates.
38243 * Dalibor Topic for better `DEBUG' support, build cleanups and Kaffe
38244 integration. `Qt4' build infrastructure, `SHA1PRNG' and
38245 `GdkPixbugDecoder' updates.
38247 * Tom Tromey for Eclipse integration, generics work, lots of bug
38248 fixes and gcj integration including coordinating The Big Merge.
38250 * Mark Wielaard for bug fixes, packaging and release management,
38251 `Clipboard' implementation, system call interrupts and network
38252 timeouts and `GdkPixpufDecoder' fixes.
38255 In addition to the above, all of which also contributed time and
38256 energy in testing GCC, we would like to thank the following for their
38257 contributions to testing:
38259 * Michael Abd-El-Malek
38269 * David Billinghurst
38273 * Stephane Bortzmeyer
38283 * Bradford Castalia
38303 * Charles-Antoine Gauthier
38325 * Kevin B. Hendricks
38329 * Christian Joensson
38337 * Anand Krishnaswamy
38339 * A. O. V. Le Blanc
38403 * Pedro A. M. Vazquez
38413 And finally we'd like to thank everyone who uses the compiler, submits
38414 bug reports and generally reminds us why we're doing this work in the
38418 File: gcc.info, Node: Option Index, Next: Keyword Index, Prev: Contributors, Up: Top
38423 GCC's command line options are indexed here without any initial `-' or
38424 `--'. Where an option has both positive and negative forms (such as
38425 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are
38426 indexed under the most appropriate form; it may sometimes be useful to
38427 look up both forms.
38432 * ###: Overall Options. (line 199)
38433 * A: Preprocessor Options.
38435 * all_load: Darwin Options. (line 112)
38436 * allowable_client: Darwin Options. (line 199)
38437 * ansi <1>: Non-bugs. (line 107)
38438 * ansi <2>: Other Builtins. (line 22)
38439 * ansi <3>: Preprocessor Options.
38441 * ansi <4>: C Dialect Options. (line 11)
38442 * ansi: Standards. (line 16)
38443 * arch_errors_fatal: Darwin Options. (line 116)
38444 * aux-info: C Dialect Options. (line 140)
38445 * b: Target Options. (line 13)
38446 * B: Directory Options. (line 41)
38447 * bcopy-builtin: PDP-11 Options. (line 32)
38448 * Bdynamic: VxWorks Options. (line 22)
38449 * bind_at_load: Darwin Options. (line 120)
38450 * Bstatic: VxWorks Options. (line 22)
38451 * bundle: Darwin Options. (line 125)
38452 * bundle_loader: Darwin Options. (line 129)
38453 * c: Link Options. (line 20)
38454 * C: Preprocessor Options.
38456 * c: Overall Options. (line 154)
38457 * client_name: Darwin Options. (line 199)
38458 * combine: Overall Options. (line 210)
38459 * compatibility_version: Darwin Options. (line 199)
38460 * coverage: Debugging Options. (line 259)
38461 * current_version: Darwin Options. (line 199)
38462 * D: Preprocessor Options.
38464 * d: Debugging Options. (line 324)
38465 * da: Debugging Options. (line 492)
38466 * dA: Debugging Options. (line 338)
38467 * dB: Debugging Options. (line 343)
38468 * dC: Debugging Options. (line 353)
38469 * dc: Debugging Options. (line 347)
38470 * dD <1>: Preprocessor Options.
38472 * dD: Debugging Options. (line 367)
38473 * dd: Debugging Options. (line 361)
38474 * dE: Debugging Options. (line 372)
38475 * dead_strip: Darwin Options. (line 199)
38476 * dependency-file: Darwin Options. (line 199)
38477 * df: Debugging Options. (line 377)
38478 * dG: Debugging Options. (line 389)
38479 * dg: Debugging Options. (line 384)
38480 * dH: Debugging Options. (line 495)
38481 * dh: Debugging Options. (line 396)
38482 * dI: Preprocessor Options.
38484 * di: Debugging Options. (line 400)
38485 * dj: Debugging Options. (line 404)
38486 * dk: Debugging Options. (line 408)
38487 * dL: Debugging Options. (line 418)
38488 * dl: Debugging Options. (line 414)
38489 * dM: Preprocessor Options.
38491 * dm: Debugging Options. (line 498)
38492 * dM: Debugging Options. (line 429)
38493 * dm: Debugging Options. (line 425)
38494 * dN <1>: Preprocessor Options.
38496 * dN: Debugging Options. (line 438)
38497 * dn: Debugging Options. (line 434)
38498 * do: Debugging Options. (line 442)
38499 * dP: Debugging Options. (line 507)
38500 * dp: Debugging Options. (line 502)
38501 * dR: Debugging Options. (line 450)
38502 * dr: Debugging Options. (line 446)
38503 * dS: Debugging Options. (line 459)
38504 * ds: Debugging Options. (line 454)
38505 * dT: Debugging Options. (line 468)
38506 * dt: Debugging Options. (line 463)
38507 * dumpmachine: Debugging Options. (line 879)
38508 * dumpspecs: Debugging Options. (line 887)
38509 * dumpversion: Debugging Options. (line 883)
38510 * dv: Debugging Options. (line 511)
38511 * dV: Debugging Options. (line 473)
38512 * dw: Debugging Options. (line 480)
38513 * dx: Debugging Options. (line 516)
38514 * dy: Debugging Options. (line 520)
38515 * dylib_file: Darwin Options. (line 199)
38516 * dylinker_install_name: Darwin Options. (line 199)
38517 * dynamic: Darwin Options. (line 199)
38518 * dynamiclib: Darwin Options. (line 133)
38519 * dZ: Debugging Options. (line 488)
38520 * dz: Debugging Options. (line 484)
38521 * E <1>: Link Options. (line 20)
38522 * E: Overall Options. (line 175)
38523 * EB <1>: MIPS Options. (line 7)
38524 * EB: ARC Options. (line 12)
38525 * EL <1>: MIPS Options. (line 10)
38526 * EL: ARC Options. (line 9)
38527 * exported_symbols_list: Darwin Options. (line 199)
38528 * F: Darwin Options. (line 32)
38529 * fabi-version: C++ Dialect Options.
38531 * falign-functions: Optimize Options. (line 999)
38532 * falign-jumps: Optimize Options. (line 1049)
38533 * falign-labels: Optimize Options. (line 1017)
38534 * falign-loops: Optimize Options. (line 1035)
38535 * fargument-alias: Code Gen Options. (line 378)
38536 * fargument-noalias: Code Gen Options. (line 378)
38537 * fargument-noalias-anything: Code Gen Options. (line 378)
38538 * fargument-noalias-global: Code Gen Options. (line 378)
38539 * fassociative-math: Optimize Options. (line 1222)
38540 * fasynchronous-unwind-tables: Code Gen Options. (line 64)
38541 * fauto-inc-dec: Optimize Options. (line 447)
38542 * fbounds-check: Code Gen Options. (line 15)
38543 * fbranch-probabilities: Optimize Options. (line 1343)
38544 * fbranch-target-load-optimize: Optimize Options. (line 1451)
38545 * fbranch-target-load-optimize2: Optimize Options. (line 1457)
38546 * fbtr-bb-exclusive: Optimize Options. (line 1461)
38547 * fcall-saved: Code Gen Options. (line 251)
38548 * fcall-used: Code Gen Options. (line 237)
38549 * fcaller-saves: Optimize Options. (line 608)
38550 * fcheck-data-deps: Optimize Options. (line 739)
38551 * fcheck-new: C++ Dialect Options.
38553 * fcommon: Variable Attributes.
38555 * fcond-mismatch: C Dialect Options. (line 258)
38556 * fconserve-space: C++ Dialect Options.
38558 * fconstant-string-class: Objective-C and Objective-C++ Dialect Options.
38560 * fcprop-registers: Optimize Options. (line 1127)
38561 * fcrossjumping: Optimize Options. (line 440)
38562 * fcse-follow-jumps: Optimize Options. (line 368)
38563 * fcse-skip-blocks: Optimize Options. (line 377)
38564 * fcx-limited-range: Optimize Options. (line 1329)
38565 * fdata-sections: Optimize Options. (line 1432)
38566 * fdbg-cnt: Debugging Options. (line 312)
38567 * fdbg-cnt-list: Debugging Options. (line 309)
38568 * fdce: Optimize Options. (line 453)
38569 * fdebug-prefix-map: Debugging Options. (line 210)
38570 * fdelayed-branch: Optimize Options. (line 507)
38571 * fdelete-null-pointer-checks: Optimize Options. (line 476)
38572 * fdiagnostics-show-location: Language Independent Options.
38574 * fdiagnostics-show-option: Language Independent Options.
38576 * fdirectives-only: Preprocessor Options.
38578 * fdollars-in-identifiers <1>: Interoperation. (line 146)
38579 * fdollars-in-identifiers: Preprocessor Options.
38581 * fdse: Optimize Options. (line 457)
38582 * fdump-class-hierarchy: Debugging Options. (line 545)
38583 * fdump-ipa: Debugging Options. (line 552)
38584 * fdump-noaddr: Debugging Options. (line 523)
38585 * fdump-rtl-all: Debugging Options. (line 492)
38586 * fdump-rtl-bbro: Debugging Options. (line 343)
38587 * fdump-rtl-btl: Debugging Options. (line 361)
38588 * fdump-rtl-bypass: Debugging Options. (line 389)
38589 * fdump-rtl-ce1: Debugging Options. (line 353)
38590 * fdump-rtl-ce2: Debugging Options. (line 353)
38591 * fdump-rtl-ce3: Debugging Options. (line 372)
38592 * fdump-rtl-cfg: Debugging Options. (line 377)
38593 * fdump-rtl-combine: Debugging Options. (line 347)
38594 * fdump-rtl-cse: Debugging Options. (line 454)
38595 * fdump-rtl-cse2: Debugging Options. (line 463)
38596 * fdump-rtl-dbr: Debugging Options. (line 361)
38597 * fdump-rtl-eh: Debugging Options. (line 396)
38598 * fdump-rtl-expand: Debugging Options. (line 446)
38599 * fdump-rtl-flow2: Debugging Options. (line 480)
38600 * fdump-rtl-gcse: Debugging Options. (line 389)
38601 * fdump-rtl-greg: Debugging Options. (line 384)
38602 * fdump-rtl-jump: Debugging Options. (line 404)
38603 * fdump-rtl-life: Debugging Options. (line 377)
38604 * fdump-rtl-loop2: Debugging Options. (line 418)
38605 * fdump-rtl-lreg: Debugging Options. (line 414)
38606 * fdump-rtl-mach: Debugging Options. (line 429)
38607 * fdump-rtl-peephole2: Debugging Options. (line 484)
38608 * fdump-rtl-postreload: Debugging Options. (line 442)
38609 * fdump-rtl-regmove: Debugging Options. (line 438)
38610 * fdump-rtl-rnreg: Debugging Options. (line 434)
38611 * fdump-rtl-sched1: Debugging Options. (line 459)
38612 * fdump-rtl-sched2: Debugging Options. (line 450)
38613 * fdump-rtl-sibling: Debugging Options. (line 400)
38614 * fdump-rtl-sms: Debugging Options. (line 425)
38615 * fdump-rtl-stack: Debugging Options. (line 408)
38616 * fdump-rtl-tracer: Debugging Options. (line 468)
38617 * fdump-rtl-vartrack: Debugging Options. (line 473)
38618 * fdump-rtl-vpt: Debugging Options. (line 473)
38619 * fdump-rtl-web: Debugging Options. (line 488)
38620 * fdump-translation-unit: Debugging Options. (line 537)
38621 * fdump-tree: Debugging Options. (line 570)
38622 * fdump-tree-alias: Debugging Options. (line 655)
38623 * fdump-tree-all: Debugging Options. (line 740)
38624 * fdump-tree-ccp: Debugging Options. (line 659)
38625 * fdump-tree-cfg: Debugging Options. (line 630)
38626 * fdump-tree-ch: Debugging Options. (line 642)
38627 * fdump-tree-copyprop: Debugging Options. (line 675)
38628 * fdump-tree-copyrename: Debugging Options. (line 721)
38629 * fdump-tree-dce: Debugging Options. (line 683)
38630 * fdump-tree-dom: Debugging Options. (line 701)
38631 * fdump-tree-dse: Debugging Options. (line 706)
38632 * fdump-tree-forwprop: Debugging Options. (line 716)
38633 * fdump-tree-fre: Debugging Options. (line 671)
38634 * fdump-tree-gimple: Debugging Options. (line 625)
38635 * fdump-tree-mudflap: Debugging Options. (line 687)
38636 * fdump-tree-nrv: Debugging Options. (line 726)
38637 * fdump-tree-phiopt: Debugging Options. (line 711)
38638 * fdump-tree-pre: Debugging Options. (line 667)
38639 * fdump-tree-salias: Debugging Options. (line 650)
38640 * fdump-tree-sink: Debugging Options. (line 697)
38641 * fdump-tree-sra: Debugging Options. (line 692)
38642 * fdump-tree-ssa: Debugging Options. (line 646)
38643 * fdump-tree-store_copyprop: Debugging Options. (line 679)
38644 * fdump-tree-storeccp: Debugging Options. (line 663)
38645 * fdump-tree-vcg: Debugging Options. (line 634)
38646 * fdump-tree-vect: Debugging Options. (line 731)
38647 * fdump-tree-vrp: Debugging Options. (line 736)
38648 * fdump-unnumbered: Debugging Options. (line 530)
38649 * fearly-inlining: Optimize Options. (line 213)
38650 * feliminate-dwarf2-dups: Debugging Options. (line 128)
38651 * feliminate-unused-debug-symbols: Debugging Options. (line 52)
38652 * feliminate-unused-debug-types: Debugging Options. (line 891)
38653 * fexceptions: Code Gen Options. (line 34)
38654 * fexec-charset: Preprocessor Options.
38656 * fexpensive-optimizations: Optimize Options. (line 489)
38657 * fextended-identifiers: Preprocessor Options.
38659 * ffast-math: Optimize Options. (line 1173)
38660 * ffinite-math-only: Optimize Options. (line 1246)
38661 * ffix-and-continue: Darwin Options. (line 106)
38662 * ffixed: Code Gen Options. (line 225)
38663 * ffloat-store <1>: Disappointments. (line 77)
38664 * ffloat-store: Optimize Options. (line 1159)
38665 * ffor-scope: C++ Dialect Options.
38667 * fforward-propagate: Optimize Options. (line 150)
38668 * ffreestanding <1>: Function Attributes.
38670 * ffreestanding <2>: Warning Options. (line 199)
38671 * ffreestanding <3>: C Dialect Options. (line 211)
38672 * ffreestanding: Standards. (line 84)
38673 * ffriend-injection: C++ Dialect Options.
38675 * ffunction-sections: Optimize Options. (line 1432)
38676 * fgcse: Optimize Options. (line 391)
38677 * fgcse-after-reload: Optimize Options. (line 427)
38678 * fgcse-las: Optimize Options. (line 420)
38679 * fgcse-lm: Optimize Options. (line 402)
38680 * fgcse-sm: Optimize Options. (line 411)
38681 * fgnu-runtime: Objective-C and Objective-C++ Dialect Options.
38683 * fgnu89-inline: C Dialect Options. (line 120)
38684 * fhosted: C Dialect Options. (line 204)
38685 * fif-conversion: Optimize Options. (line 461)
38686 * fif-conversion2: Optimize Options. (line 470)
38687 * filelist: Darwin Options. (line 199)
38688 * findirect-data: Darwin Options. (line 106)
38689 * finhibit-size-directive: Code Gen Options. (line 147)
38690 * finline-functions: Optimize Options. (line 194)
38691 * finline-functions-called-once: Optimize Options. (line 205)
38692 * finline-limit: Optimize Options. (line 223)
38693 * finline-small-functions: Optimize Options. (line 186)
38694 * finput-charset: Preprocessor Options.
38696 * finstrument-functions <1>: Function Attributes.
38698 * finstrument-functions: Code Gen Options. (line 281)
38699 * finstrument-functions-exclude-file-list: Code Gen Options. (line 318)
38700 * finstrument-functions-exclude-function-list: Code Gen Options.
38702 * fipa-cp: Optimize Options. (line 671)
38703 * fipa-matrix-reorg: Optimize Options. (line 680)
38704 * fipa-pta: Optimize Options. (line 668)
38705 * fipa-pure-const: Optimize Options. (line 645)
38706 * fipa-reference: Optimize Options. (line 649)
38707 * fipa-struct-reorg: Optimize Options. (line 653)
38708 * fivopts: Optimize Options. (line 758)
38709 * fkeep-inline-functions <1>: Inline. (line 51)
38710 * fkeep-inline-functions: Optimize Options. (line 249)
38711 * fkeep-static-consts: Optimize Options. (line 256)
38712 * flat_namespace: Darwin Options. (line 199)
38713 * flax-vector-conversions: C Dialect Options. (line 263)
38714 * fleading-underscore: Code Gen Options. (line 395)
38715 * fmem-report: Debugging Options. (line 234)
38716 * fmerge-all-constants: Optimize Options. (line 275)
38717 * fmerge-constants: Optimize Options. (line 265)
38718 * fmerge-debug-strings: Debugging Options. (line 203)
38719 * fmessage-length: Language Independent Options.
38721 * fmodulo-sched: Optimize Options. (line 285)
38722 * fmodulo-sched-allow-regmoves: Optimize Options. (line 290)
38723 * fmove-loop-invariants: Optimize Options. (line 1422)
38724 * fms-extensions <1>: Unnamed Fields. (line 37)
38725 * fms-extensions <2>: C++ Dialect Options.
38727 * fms-extensions: C Dialect Options. (line 229)
38728 * fmudflap: Optimize Options. (line 330)
38729 * fmudflapir: Optimize Options. (line 330)
38730 * fmudflapth: Optimize Options. (line 330)
38731 * fnext-runtime: Objective-C and Objective-C++ Dialect Options.
38733 * fno-access-control: C++ Dialect Options.
38735 * fno-asm: C Dialect Options. (line 156)
38736 * fno-branch-count-reg: Optimize Options. (line 297)
38737 * fno-builtin <1>: Other Builtins. (line 14)
38738 * fno-builtin <2>: Function Attributes.
38740 * fno-builtin <3>: Warning Options. (line 199)
38741 * fno-builtin: C Dialect Options. (line 170)
38742 * fno-common <1>: Variable Attributes.
38744 * fno-common: Code Gen Options. (line 135)
38745 * fno-default-inline <1>: Inline. (line 71)
38746 * fno-default-inline <2>: Optimize Options. (line 135)
38747 * fno-default-inline: C++ Dialect Options.
38749 * fno-defer-pop: Optimize Options. (line 142)
38750 * fno-elide-constructors: C++ Dialect Options.
38752 * fno-enforce-eh-specs: C++ Dialect Options.
38754 * fno-for-scope: C++ Dialect Options.
38756 * fno-function-cse: Optimize Options. (line 307)
38757 * fno-gnu-keywords: C++ Dialect Options.
38759 * fno-guess-branch-probability: Optimize Options. (line 881)
38760 * fno-ident: Code Gen Options. (line 144)
38761 * fno-implement-inlines <1>: C++ Interface. (line 75)
38762 * fno-implement-inlines: C++ Dialect Options.
38764 * fno-implicit-inline-templates: C++ Dialect Options.
38766 * fno-implicit-templates <1>: Template Instantiation.
38768 * fno-implicit-templates: C++ Dialect Options.
38770 * fno-inline: Optimize Options. (line 180)
38771 * fno-jump-tables: Code Gen Options. (line 217)
38772 * fno-math-errno: Optimize Options. (line 1187)
38773 * fno-merge-debug-strings: Debugging Options. (line 203)
38774 * fno-nil-receivers: Objective-C and Objective-C++ Dialect Options.
38776 * fno-nonansi-builtins: C++ Dialect Options.
38778 * fno-operator-names: C++ Dialect Options.
38780 * fno-optional-diags: C++ Dialect Options.
38782 * fno-peephole: Optimize Options. (line 872)
38783 * fno-peephole2: Optimize Options. (line 872)
38784 * fno-rtti: C++ Dialect Options.
38786 * fno-sched-interblock: Optimize Options. (line 533)
38787 * fno-sched-spec: Optimize Options. (line 538)
38788 * fno-show-column: Preprocessor Options.
38790 * fno-signed-bitfields: C Dialect Options. (line 296)
38791 * fno-signed-zeros: Optimize Options. (line 1258)
38792 * fno-stack-limit: Code Gen Options. (line 361)
38793 * fno-threadsafe-statics: C++ Dialect Options.
38795 * fno-toplevel-reorder: Optimize Options. (line 1093)
38796 * fno-trapping-math: Optimize Options. (line 1268)
38797 * fno-unsigned-bitfields: C Dialect Options. (line 296)
38798 * fno-use-cxa-get-exception-ptr: C++ Dialect Options.
38800 * fno-weak: C++ Dialect Options.
38802 * fno-working-directory: Preprocessor Options.
38804 * fno-zero-initialized-in-bss: Optimize Options. (line 318)
38805 * fnon-call-exceptions: Code Gen Options. (line 48)
38806 * fobjc-call-cxx-cdtors: Objective-C and Objective-C++ Dialect Options.
38808 * fobjc-direct-dispatch: Objective-C and Objective-C++ Dialect Options.
38810 * fobjc-exceptions: Objective-C and Objective-C++ Dialect Options.
38812 * fobjc-gc: Objective-C and Objective-C++ Dialect Options.
38814 * fomit-frame-pointer: Optimize Options. (line 159)
38815 * fopenmp: C Dialect Options. (line 221)
38816 * foptimize-register-move: Optimize Options. (line 496)
38817 * foptimize-sibling-calls: Optimize Options. (line 175)
38818 * force_cpusubtype_ALL: Darwin Options. (line 138)
38819 * force_flat_namespace: Darwin Options. (line 199)
38820 * fpack-struct: Code Gen Options. (line 268)
38821 * fpcc-struct-return <1>: Incompatibilities. (line 170)
38822 * fpcc-struct-return: Code Gen Options. (line 70)
38823 * fpch-deps: Preprocessor Options.
38825 * fpch-preprocess: Preprocessor Options.
38827 * fpeel-loops: Optimize Options. (line 1414)
38828 * fpermissive: C++ Dialect Options.
38830 * fPIC: Code Gen Options. (line 194)
38831 * fpic: Code Gen Options. (line 173)
38832 * fPIE: Code Gen Options. (line 207)
38833 * fpie: Code Gen Options. (line 207)
38834 * fpost-ipa-mem-report: Debugging Options. (line 240)
38835 * fpre-ipa-mem-report: Debugging Options. (line 238)
38836 * fpredictive-commoning: Optimize Options. (line 854)
38837 * fprefetch-loop-arrays: Optimize Options. (line 861)
38838 * fpreprocessed: Preprocessor Options.
38840 * fprofile-arcs <1>: Other Builtins. (line 240)
38841 * fprofile-arcs: Debugging Options. (line 244)
38842 * fprofile-generate: Optimize Options. (line 1134)
38843 * fprofile-use: Optimize Options. (line 1143)
38844 * fprofile-values: Optimize Options. (line 1362)
38845 * frandom-string: Debugging Options. (line 769)
38846 * freciprocal-math: Optimize Options. (line 1237)
38847 * frecord-gcc-switches: Code Gen Options. (line 163)
38848 * freg-struct-return: Code Gen Options. (line 88)
38849 * fregmove: Optimize Options. (line 496)
38850 * frename-registers: Optimize Options. (line 1381)
38851 * freorder-blocks: Optimize Options. (line 898)
38852 * freorder-blocks-and-partition: Optimize Options. (line 904)
38853 * freorder-functions: Optimize Options. (line 915)
38854 * freplace-objc-classes: Objective-C and Objective-C++ Dialect Options.
38856 * frepo <1>: Template Instantiation.
38858 * frepo: C++ Dialect Options.
38860 * frerun-cse-after-loop: Optimize Options. (line 385)
38861 * freschedule-modulo-scheduled-loops: Optimize Options. (line 602)
38862 * frounding-math: Optimize Options. (line 1283)
38863 * frtl-abstract-sequences: Optimize Options. (line 1303)
38864 * fsched-spec-load: Optimize Options. (line 543)
38865 * fsched-spec-load-dangerous: Optimize Options. (line 548)
38866 * fsched-stalled-insns: Optimize Options. (line 554)
38867 * fsched-stalled-insns-dep: Optimize Options. (line 564)
38868 * fsched-verbose: Debugging Options. (line 779)
38869 * fsched2-use-superblocks: Optimize Options. (line 574)
38870 * fsched2-use-traces: Optimize Options. (line 585)
38871 * fschedule-insns: Optimize Options. (line 514)
38872 * fschedule-insns2: Optimize Options. (line 524)
38873 * fsection-anchors: Optimize Options. (line 1477)
38874 * fsee: Optimize Options. (line 597)
38875 * fshort-double: Code Gen Options. (line 117)
38876 * fshort-enums <1>: Non-bugs. (line 42)
38877 * fshort-enums <2>: Type Attributes. (line 113)
38878 * fshort-enums <3>: Structures unions enumerations and bit-fields implementation.
38880 * fshort-enums: Code Gen Options. (line 106)
38881 * fshort-wchar: Code Gen Options. (line 125)
38882 * fsignaling-nans: Optimize Options. (line 1310)
38883 * fsigned-bitfields <1>: Non-bugs. (line 57)
38884 * fsigned-bitfields: C Dialect Options. (line 296)
38885 * fsigned-char <1>: Characters implementation.
38887 * fsigned-char: C Dialect Options. (line 286)
38888 * fsingle-precision-constant: Optimize Options. (line 1325)
38889 * fsplit-ivs-in-unroller: Optimize Options. (line 835)
38890 * fsplit-wide-types: Optimize Options. (line 360)
38891 * fstack-check: Code Gen Options. (line 346)
38892 * fstack-limit-register: Code Gen Options. (line 361)
38893 * fstack-limit-symbol: Code Gen Options. (line 361)
38894 * fstack-protector: Optimize Options. (line 1465)
38895 * fstack-protector-all: Optimize Options. (line 1474)
38896 * fstats: C++ Dialect Options.
38898 * fstrict-aliasing: Optimize Options. (line 928)
38899 * fstrict-overflow: Optimize Options. (line 964)
38900 * fsyntax-only: Warning Options. (line 14)
38901 * ftabstop: Preprocessor Options.
38903 * ftemplate-depth: C++ Dialect Options.
38905 * ftest-coverage: Debugging Options. (line 300)
38906 * fthread-jumps: Optimize Options. (line 351)
38907 * ftime-report: Debugging Options. (line 230)
38908 * ftls-model: Code Gen Options. (line 406)
38909 * ftracer: Optimize Options. (line 818)
38910 * ftrapv: Code Gen Options. (line 22)
38911 * ftree-ccp: Optimize Options. (line 694)
38912 * ftree-ch: Optimize Options. (line 723)
38913 * ftree-copy-prop: Optimize Options. (line 636)
38914 * ftree-copyrename: Optimize Options. (line 778)
38915 * ftree-dce: Optimize Options. (line 705)
38916 * ftree-dominator-opts: Optimize Options. (line 709)
38917 * ftree-dse: Optimize Options. (line 716)
38918 * ftree-fre: Optimize Options. (line 629)
38919 * ftree-loop-im: Optimize Options. (line 743)
38920 * ftree-loop-ivcanon: Optimize Options. (line 752)
38921 * ftree-loop-linear: Optimize Options. (line 734)
38922 * ftree-loop-optimize: Optimize Options. (line 730)
38923 * ftree-parallelize-loops: Optimize Options. (line 763)
38924 * ftree-pre: Optimize Options. (line 625)
38925 * ftree-reassoc: Optimize Options. (line 621)
38926 * ftree-salias: Optimize Options. (line 641)
38927 * ftree-sink: Optimize Options. (line 690)
38928 * ftree-sra: Optimize Options. (line 772)
38929 * ftree-store-ccp: Optimize Options. (line 699)
38930 * ftree-ter: Optimize Options. (line 785)
38931 * ftree-vect-loop-version: Optimize Options. (line 797)
38932 * ftree-vectorize: Optimize Options. (line 793)
38933 * ftree-vectorizer-verbose: Debugging Options. (line 744)
38934 * ftree-vrp: Optimize Options. (line 809)
38935 * funit-at-a-time: Optimize Options. (line 1062)
38936 * funroll-all-loops: Optimize Options. (line 829)
38937 * funroll-loops: Optimize Options. (line 823)
38938 * funsafe-loop-optimizations: Optimize Options. (line 432)
38939 * funsafe-math-optimizations: Optimize Options. (line 1205)
38940 * funsigned-bitfields <1>: Non-bugs. (line 57)
38941 * funsigned-bitfields <2>: Structures unions enumerations and bit-fields implementation.
38943 * funsigned-bitfields: C Dialect Options. (line 296)
38944 * funsigned-char <1>: Characters implementation.
38946 * funsigned-char: C Dialect Options. (line 268)
38947 * funswitch-loops: Optimize Options. (line 1426)
38948 * funwind-tables: Code Gen Options. (line 57)
38949 * fuse-cxa-atexit: C++ Dialect Options.
38951 * fvar-tracking: Debugging Options. (line 822)
38952 * fvariable-expansion-in-unroller: Optimize Options. (line 849)
38953 * fvect-cost-model: Optimize Options. (line 806)
38954 * fverbose-asm: Code Gen Options. (line 154)
38955 * fvisibility: Code Gen Options. (line 414)
38956 * fvisibility-inlines-hidden: C++ Dialect Options.
38958 * fvisibility-ms-compat: C++ Dialect Options.
38960 * fvpt: Optimize Options. (line 1372)
38961 * fweb: Optimize Options. (line 1101)
38962 * fwhole-program: Optimize Options. (line 1112)
38963 * fwide-exec-charset: Preprocessor Options.
38965 * fworking-directory: Preprocessor Options.
38967 * fwrapv: Code Gen Options. (line 26)
38968 * fzero-link: Objective-C and Objective-C++ Dialect Options.
38970 * G <1>: System V Options. (line 10)
38971 * G <2>: RS/6000 and PowerPC Options.
38973 * G <3>: MIPS Options. (line 295)
38974 * G: M32R/D Options. (line 57)
38975 * g: Debugging Options. (line 10)
38976 * gcoff: Debugging Options. (line 70)
38977 * gdwarf-2: Debugging Options. (line 88)
38978 * gen-decls: Objective-C and Objective-C++ Dialect Options.
38980 * gfull: Darwin Options. (line 71)
38981 * ggdb: Debugging Options. (line 38)
38982 * gnu-ld: HPPA Options. (line 113)
38983 * gstabs: Debugging Options. (line 44)
38984 * gstabs+: Debugging Options. (line 64)
38985 * gused: Darwin Options. (line 66)
38986 * gvms: Debugging Options. (line 95)
38987 * gxcoff: Debugging Options. (line 75)
38988 * gxcoff+: Debugging Options. (line 80)
38989 * H: Preprocessor Options.
38991 * headerpad_max_install_names: Darwin Options. (line 199)
38992 * help <1>: Preprocessor Options.
38994 * help: Overall Options. (line 226)
38995 * hp-ld: HPPA Options. (line 125)
38996 * I <1>: Directory Options. (line 10)
38997 * I: Preprocessor Options.
38999 * I- <1>: Directory Options. (line 107)
39000 * I-: Preprocessor Options.
39002 * idirafter: Preprocessor Options.
39004 * iframework: Darwin Options. (line 59)
39005 * imacros: Preprocessor Options.
39007 * image_base: Darwin Options. (line 199)
39008 * imultilib: Preprocessor Options.
39010 * include: Preprocessor Options.
39012 * init: Darwin Options. (line 199)
39013 * install_name: Darwin Options. (line 199)
39014 * iprefix: Preprocessor Options.
39016 * iquote <1>: Directory Options. (line 31)
39017 * iquote: Preprocessor Options.
39019 * isysroot: Preprocessor Options.
39021 * isystem: Preprocessor Options.
39023 * iwithprefix: Preprocessor Options.
39025 * iwithprefixbefore: Preprocessor Options.
39027 * keep_private_externs: Darwin Options. (line 199)
39028 * L: Directory Options. (line 37)
39029 * l: Link Options. (line 26)
39030 * lobjc: Link Options. (line 53)
39031 * M: Preprocessor Options.
39033 * m1: SH Options. (line 9)
39034 * m10: PDP-11 Options. (line 29)
39035 * m128bit-long-double: i386 and x86-64 Options.
39037 * m16-bit: CRIS Options. (line 69)
39038 * m2: SH Options. (line 12)
39039 * m210: MCore Options. (line 43)
39040 * m3: SH Options. (line 18)
39041 * m31: S/390 and zSeries Options.
39043 * m32 <1>: SPARC Options. (line 191)
39044 * m32 <2>: RS/6000 and PowerPC Options.
39046 * m32: i386 and x86-64 Options.
39048 * m32-bit: CRIS Options. (line 69)
39049 * m32r: M32R/D Options. (line 15)
39050 * m32r2: M32R/D Options. (line 9)
39051 * m32rx: M32R/D Options. (line 12)
39052 * m340: MCore Options. (line 43)
39053 * m3dnow: i386 and x86-64 Options.
39055 * m3e: SH Options. (line 21)
39056 * m4: SH Options. (line 35)
39057 * m4-nofpu: SH Options. (line 24)
39058 * m4-single: SH Options. (line 31)
39059 * m4-single-only: SH Options. (line 27)
39060 * m40: PDP-11 Options. (line 23)
39061 * m45: PDP-11 Options. (line 26)
39062 * m4a: SH Options. (line 50)
39063 * m4a-nofpu: SH Options. (line 38)
39064 * m4a-single: SH Options. (line 46)
39065 * m4a-single-only: SH Options. (line 42)
39066 * m4al: SH Options. (line 53)
39067 * m4byte-functions: MCore Options. (line 27)
39068 * m5200: M680x0 Options. (line 143)
39069 * m5206e: M680x0 Options. (line 152)
39070 * m528x: M680x0 Options. (line 156)
39071 * m5307: M680x0 Options. (line 160)
39072 * m5407: M680x0 Options. (line 164)
39073 * m64 <1>: SPARC Options. (line 191)
39074 * m64 <2>: S/390 and zSeries Options.
39076 * m64 <3>: RS/6000 and PowerPC Options.
39078 * m64: i386 and x86-64 Options.
39080 * m68000: M680x0 Options. (line 91)
39081 * m68010: M680x0 Options. (line 99)
39082 * m68020: M680x0 Options. (line 105)
39083 * m68020-40: M680x0 Options. (line 174)
39084 * m68020-60: M680x0 Options. (line 183)
39085 * m68030: M680x0 Options. (line 110)
39086 * m68040: M680x0 Options. (line 115)
39087 * m68060: M680x0 Options. (line 124)
39088 * m6811: M68hc1x Options. (line 13)
39089 * m6812: M68hc1x Options. (line 18)
39090 * m68881: M680x0 Options. (line 193)
39091 * m68hc11: M68hc1x Options. (line 13)
39092 * m68hc12: M68hc1x Options. (line 18)
39093 * m68hcs12: M68hc1x Options. (line 23)
39094 * m68S12: M68hc1x Options. (line 23)
39095 * m8-bit: CRIS Options. (line 69)
39096 * m96bit-long-double: i386 and x86-64 Options.
39098 * mabi <1>: RS/6000 and PowerPC Options.
39100 * mabi: ARM Options. (line 10)
39101 * mabi-mmixware: MMIX Options. (line 20)
39102 * mabi=32: MIPS Options. (line 120)
39103 * mabi=64: MIPS Options. (line 120)
39104 * mabi=eabi: MIPS Options. (line 120)
39105 * mabi=gnu: MMIX Options. (line 20)
39106 * mabi=ibmlongdouble: RS/6000 and PowerPC Options.
39108 * mabi=ieeelongdouble: RS/6000 and PowerPC Options.
39110 * mabi=n32: MIPS Options. (line 120)
39111 * mabi=no-spe: RS/6000 and PowerPC Options.
39113 * mabi=o64: MIPS Options. (line 120)
39114 * mabi=spe: RS/6000 and PowerPC Options.
39116 * mabicalls: MIPS Options. (line 144)
39117 * mabort-on-noreturn: ARM Options. (line 147)
39118 * mabshi: PDP-11 Options. (line 55)
39119 * mac0: PDP-11 Options. (line 16)
39120 * macc-4: FRV Options. (line 113)
39121 * macc-8: FRV Options. (line 116)
39122 * maccumulate-outgoing-args: i386 and x86-64 Options.
39124 * madjust-unroll: SH Options. (line 186)
39125 * mads: RS/6000 and PowerPC Options.
39127 * maix-struct-return: RS/6000 and PowerPC Options.
39129 * maix32: RS/6000 and PowerPC Options.
39131 * maix64: RS/6000 and PowerPC Options.
39133 * malign-300: H8/300 Options. (line 31)
39134 * malign-double: i386 and x86-64 Options.
39136 * malign-int: M680x0 Options. (line 263)
39137 * malign-labels: FRV Options. (line 104)
39138 * malign-loops: M32R/D Options. (line 73)
39139 * malign-natural: RS/6000 and PowerPC Options.
39141 * malign-power: RS/6000 and PowerPC Options.
39143 * malloc-cc: FRV Options. (line 25)
39144 * malpha-as: DEC Alpha Options. (line 159)
39145 * maltivec: RS/6000 and PowerPC Options.
39147 * mam33: MN10300 Options. (line 17)
39148 * maout: CRIS Options. (line 92)
39149 * mapcs: ARM Options. (line 22)
39150 * mapcs-frame: ARM Options. (line 14)
39151 * mapp-regs <1>: V850 Options. (line 57)
39152 * mapp-regs: SPARC Options. (line 10)
39153 * march <1>: S/390 and zSeries Options.
39155 * march <2>: MT Options. (line 9)
39156 * march <3>: MIPS Options. (line 14)
39157 * march <4>: M680x0 Options. (line 12)
39158 * march <5>: i386 and x86-64 Options.
39160 * march <6>: HPPA Options. (line 9)
39161 * march <7>: CRIS Options. (line 10)
39162 * march: ARM Options. (line 110)
39163 * masm=DIALECT: i386 and x86-64 Options.
39165 * mauto-incdec: M68hc1x Options. (line 26)
39166 * mauto-pic: IA-64 Options. (line 50)
39167 * mb: SH Options. (line 58)
39168 * mbacc: MT Options. (line 16)
39169 * mbackchain: S/390 and zSeries Options.
39171 * mbase-addresses: MMIX Options. (line 54)
39172 * mbcopy: PDP-11 Options. (line 36)
39173 * mbig: RS/6000 and PowerPC Options.
39175 * mbig-endian <1>: RS/6000 and PowerPC Options.
39177 * mbig-endian <2>: MCore Options. (line 39)
39178 * mbig-endian <3>: IA-64 Options. (line 9)
39179 * mbig-endian: ARM Options. (line 72)
39180 * mbig-switch <1>: V850 Options. (line 52)
39181 * mbig-switch: HPPA Options. (line 23)
39182 * mbigtable: SH Options. (line 74)
39183 * mbit-align: RS/6000 and PowerPC Options.
39185 * mbitfield: M680x0 Options. (line 231)
39186 * mbranch-cheap: PDP-11 Options. (line 65)
39187 * mbranch-cost: MIPS Options. (line 519)
39188 * mbranch-cost=NUMBER: M32R/D Options. (line 82)
39189 * mbranch-expensive: PDP-11 Options. (line 61)
39190 * mbranch-hints: SPU Options. (line 27)
39191 * mbranch-likely: MIPS Options. (line 526)
39192 * mbranch-predict: MMIX Options. (line 49)
39193 * mbss-plt: RS/6000 and PowerPC Options.
39195 * mbuild-constants: DEC Alpha Options. (line 142)
39196 * mbwx: DEC Alpha Options. (line 171)
39197 * mc68000: M680x0 Options. (line 91)
39198 * mc68020: M680x0 Options. (line 105)
39199 * mcall-gnu: RS/6000 and PowerPC Options.
39201 * mcall-linux: RS/6000 and PowerPC Options.
39203 * mcall-netbsd: RS/6000 and PowerPC Options.
39205 * mcall-prologues: AVR Options. (line 43)
39206 * mcall-solaris: RS/6000 and PowerPC Options.
39208 * mcall-sysv: RS/6000 and PowerPC Options.
39210 * mcall-sysv-eabi: RS/6000 and PowerPC Options.
39212 * mcall-sysv-noeabi: RS/6000 and PowerPC Options.
39214 * mcallee-super-interworking: ARM Options. (line 239)
39215 * mcaller-super-interworking: ARM Options. (line 245)
39216 * mcallgraph-data: MCore Options. (line 31)
39217 * mcc-init: CRIS Options. (line 46)
39218 * mcfv4e: M680x0 Options. (line 168)
39219 * mcheck-zero-division: MIPS Options. (line 406)
39220 * mcirrus-fix-invalid-insns: ARM Options. (line 190)
39221 * mcix: DEC Alpha Options. (line 171)
39222 * mcld: i386 and x86-64 Options.
39224 * mcmodel=embmedany: SPARC Options. (line 213)
39225 * mcmodel=kernel: i386 and x86-64 Options.
39227 * mcmodel=large: i386 and x86-64 Options.
39229 * mcmodel=medany: SPARC Options. (line 207)
39230 * mcmodel=medium: i386 and x86-64 Options.
39232 * mcmodel=medlow: SPARC Options. (line 196)
39233 * mcmodel=medmid: SPARC Options. (line 201)
39234 * mcmodel=small: i386 and x86-64 Options.
39236 * mcmpb: RS/6000 and PowerPC Options.
39238 * mcode-readable: MIPS Options. (line 366)
39239 * mcond-exec: FRV Options. (line 152)
39240 * mcond-move: FRV Options. (line 128)
39241 * mconst-align: CRIS Options. (line 60)
39242 * mconst16: Xtensa Options. (line 10)
39243 * mconstant-gp: IA-64 Options. (line 46)
39244 * mcpu <1>: SPARC Options. (line 96)
39245 * mcpu <2>: RS/6000 and PowerPC Options.
39247 * mcpu <3>: M680x0 Options. (line 28)
39248 * mcpu <4>: i386 and x86-64 Options.
39250 * mcpu <5>: FRV Options. (line 212)
39251 * mcpu <6>: DEC Alpha Options. (line 223)
39252 * mcpu <7>: CRIS Options. (line 10)
39253 * mcpu <8>: ARM Options. (line 84)
39254 * mcpu: ARC Options. (line 23)
39255 * mcpu32: M680x0 Options. (line 134)
39256 * mcpu= <1>: M32C Options. (line 7)
39257 * mcpu=: Blackfin Options. (line 7)
39258 * mcsync-anomaly: Blackfin Options. (line 55)
39259 * mcx16: i386 and x86-64 Options.
39261 * MD: Preprocessor Options.
39263 * mdalign: SH Options. (line 64)
39264 * mdata: ARC Options. (line 30)
39265 * mdata-align: CRIS Options. (line 60)
39266 * mdebug <1>: S/390 and zSeries Options.
39268 * mdebug: M32R/D Options. (line 69)
39269 * mdec-asm: PDP-11 Options. (line 78)
39270 * mdisable-callt: V850 Options. (line 80)
39271 * mdisable-fpregs: HPPA Options. (line 33)
39272 * mdisable-indexing: HPPA Options. (line 40)
39273 * mdiv <1>: MCore Options. (line 15)
39274 * mdiv: M680x0 Options. (line 205)
39275 * mdiv=STRATEGY: SH Options. (line 138)
39276 * mdivide-breaks: MIPS Options. (line 412)
39277 * mdivide-traps: MIPS Options. (line 412)
39278 * mdivsi3_libfunc=NAME: SH Options. (line 179)
39279 * mdlmzb: RS/6000 and PowerPC Options.
39281 * mdmx: MIPS Options. (line 259)
39282 * mdouble: FRV Options. (line 38)
39283 * mdouble-float: MIPS Options. (line 217)
39284 * mdsp: MIPS Options. (line 236)
39285 * mdspr2: MIPS Options. (line 242)
39286 * mdwarf2-asm: IA-64 Options. (line 79)
39287 * mdword: FRV Options. (line 32)
39288 * mdynamic-no-pic: RS/6000 and PowerPC Options.
39290 * meabi: RS/6000 and PowerPC Options.
39292 * mearly-stop-bits: IA-64 Options. (line 85)
39293 * meb: Score Options. (line 9)
39294 * mel: Score Options. (line 12)
39295 * melf <1>: MMIX Options. (line 44)
39296 * melf: CRIS Options. (line 95)
39297 * melinux: CRIS Options. (line 99)
39298 * melinux-stacksize: CRIS Options. (line 25)
39299 * memb: RS/6000 and PowerPC Options.
39301 * membedded-data: MIPS Options. (line 353)
39302 * memregs=: M32C Options. (line 21)
39303 * mep: V850 Options. (line 16)
39304 * mepsilon: MMIX Options. (line 15)
39305 * merror-reloc: SPU Options. (line 10)
39306 * mesa: S/390 and zSeries Options.
39308 * metrax100: CRIS Options. (line 31)
39309 * metrax4: CRIS Options. (line 31)
39310 * mexplicit-relocs <1>: MIPS Options. (line 397)
39311 * mexplicit-relocs: DEC Alpha Options. (line 184)
39312 * mextern-sdata: MIPS Options. (line 315)
39313 * MF: Preprocessor Options.
39315 * mfast-fp: Blackfin Options. (line 128)
39316 * mfast-indirect-calls: HPPA Options. (line 52)
39317 * mfaster-structs: SPARC Options. (line 71)
39318 * mfdpic: FRV Options. (line 56)
39319 * mfix: DEC Alpha Options. (line 171)
39320 * mfix-and-continue: Darwin Options. (line 106)
39321 * mfix-r4000: MIPS Options. (line 462)
39322 * mfix-r4400: MIPS Options. (line 476)
39323 * mfix-sb1: MIPS Options. (line 504)
39324 * mfix-vr4120: MIPS Options. (line 483)
39325 * mfix-vr4130: MIPS Options. (line 497)
39326 * mfixed-cc: FRV Options. (line 28)
39327 * mfixed-range <1>: SPU Options. (line 47)
39328 * mfixed-range <2>: IA-64 Options. (line 90)
39329 * mfixed-range: HPPA Options. (line 59)
39330 * mflip-mips16: MIPS Options. (line 100)
39331 * mfloat-abi: ARM Options. (line 59)
39332 * mfloat-gprs: RS/6000 and PowerPC Options.
39334 * mfloat-ieee: DEC Alpha Options. (line 179)
39335 * mfloat-vax: DEC Alpha Options. (line 179)
39336 * mfloat32: PDP-11 Options. (line 52)
39337 * mfloat64: PDP-11 Options. (line 48)
39338 * mflush-func: MIPS Options. (line 510)
39339 * mflush-func=NAME: M32R/D Options. (line 94)
39340 * mflush-trap=NUMBER: M32R/D Options. (line 87)
39341 * mfmovd: SH Options. (line 78)
39342 * mfp: ARM Options. (line 122)
39343 * mfp-exceptions: MIPS Options. (line 537)
39344 * mfp-reg: DEC Alpha Options. (line 25)
39345 * mfp-rounding-mode: DEC Alpha Options. (line 85)
39346 * mfp-trap-mode: DEC Alpha Options. (line 63)
39347 * mfp32: MIPS Options. (line 200)
39348 * mfp64: MIPS Options. (line 203)
39349 * mfpe: ARM Options. (line 122)
39350 * mfpr-32: FRV Options. (line 13)
39351 * mfpr-64: FRV Options. (line 16)
39352 * mfprnd: RS/6000 and PowerPC Options.
39354 * mfpu <1>: SPARC Options. (line 20)
39355 * mfpu <2>: PDP-11 Options. (line 9)
39356 * mfpu: ARM Options. (line 122)
39357 * mfull-toc: RS/6000 and PowerPC Options.
39359 * mfused-madd <1>: Xtensa Options. (line 19)
39360 * mfused-madd <2>: S/390 and zSeries Options.
39362 * mfused-madd <3>: RS/6000 and PowerPC Options.
39364 * mfused-madd <4>: MIPS Options. (line 447)
39365 * mfused-madd: i386 and x86-64 Options.
39367 * mg: VAX Options. (line 17)
39368 * MG: Preprocessor Options.
39370 * mgas <1>: HPPA Options. (line 75)
39371 * mgas: DEC Alpha Options. (line 159)
39372 * mgettrcost=NUMBER: SH Options. (line 201)
39373 * mglibc: GNU/Linux Options. (line 9)
39374 * mgnu: VAX Options. (line 13)
39375 * mgnu-as: IA-64 Options. (line 18)
39376 * mgnu-ld: IA-64 Options. (line 23)
39377 * mgotplt: CRIS Options. (line 86)
39378 * mgp32: MIPS Options. (line 194)
39379 * mgp64: MIPS Options. (line 197)
39380 * mgpopt: MIPS Options. (line 338)
39381 * mgpr-32: FRV Options. (line 7)
39382 * mgpr-64: FRV Options. (line 10)
39383 * mgprel-ro: FRV Options. (line 79)
39384 * mh: H8/300 Options. (line 14)
39385 * mhard-dfp: RS/6000 and PowerPC Options.
39387 * mhard-float <1>: SPARC Options. (line 20)
39388 * mhard-float <2>: S/390 and zSeries Options.
39390 * mhard-float <3>: RS/6000 and PowerPC Options.
39392 * mhard-float <4>: MIPS Options. (line 206)
39393 * mhard-float <5>: M680x0 Options. (line 193)
39394 * mhard-float <6>: FRV Options. (line 19)
39395 * mhard-float: ARM Options. (line 41)
39396 * mhard-quad-float: SPARC Options. (line 41)
39397 * mhardlit: MCore Options. (line 10)
39398 * mhitachi: SH Options. (line 81)
39399 * mid-shared-library: Blackfin Options. (line 76)
39400 * mieee <1>: SH Options. (line 96)
39401 * mieee: DEC Alpha Options. (line 39)
39402 * mieee-conformant: DEC Alpha Options. (line 134)
39403 * mieee-fp: i386 and x86-64 Options.
39405 * mieee-with-inexact: DEC Alpha Options. (line 52)
39406 * milp32: IA-64 Options. (line 114)
39407 * mimpure-text: SPARC Options. (line 81)
39408 * mindexed-addressing: SH Options. (line 191)
39409 * minit-stack: AVR Options. (line 35)
39410 * minline-all-stringops: i386 and x86-64 Options.
39412 * minline-float-divide-max-throughput: IA-64 Options. (line 58)
39413 * minline-float-divide-min-latency: IA-64 Options. (line 54)
39414 * minline-ic_invalidate: SH Options. (line 103)
39415 * minline-int-divide-max-throughput: IA-64 Options. (line 66)
39416 * minline-int-divide-min-latency: IA-64 Options. (line 62)
39417 * minline-plt <1>: FRV Options. (line 64)
39418 * minline-plt: Blackfin Options. (line 133)
39419 * minline-sqrt-max-throughput: IA-64 Options. (line 74)
39420 * minline-sqrt-min-latency: IA-64 Options. (line 70)
39421 * minline-stringops-dynamically: i386 and x86-64 Options.
39423 * minmax: M68hc1x Options. (line 31)
39424 * minsert-sched-nops: RS/6000 and PowerPC Options.
39426 * mint16: PDP-11 Options. (line 40)
39427 * mint32 <1>: PDP-11 Options. (line 44)
39428 * mint32: H8/300 Options. (line 28)
39429 * mint8: AVR Options. (line 53)
39430 * minterlink-mips16: MIPS Options. (line 107)
39431 * minvalid-symbols: SH Options. (line 224)
39432 * mips1: MIPS Options. (line 70)
39433 * mips16: MIPS Options. (line 92)
39434 * mips2: MIPS Options. (line 73)
39435 * mips3: MIPS Options. (line 76)
39436 * mips32: MIPS Options. (line 82)
39437 * mips32r2: MIPS Options. (line 85)
39438 * mips3d: MIPS Options. (line 265)
39439 * mips4: MIPS Options. (line 79)
39440 * mips64: MIPS Options. (line 88)
39441 * misel: RS/6000 and PowerPC Options.
39443 * misize: SH Options. (line 115)
39444 * missue-rate=NUMBER: M32R/D Options. (line 79)
39445 * mjump-in-delay: HPPA Options. (line 28)
39446 * mkernel: Darwin Options. (line 84)
39447 * mknuthdiv: MMIX Options. (line 33)
39448 * ml: SH Options. (line 61)
39449 * mlarge-data: DEC Alpha Options. (line 195)
39450 * mlarge-data-threshold=NUMBER: i386 and x86-64 Options.
39452 * mlarge-mem: SPU Options. (line 35)
39453 * mlarge-text: DEC Alpha Options. (line 213)
39454 * mleaf-id-shared-library: Blackfin Options. (line 87)
39455 * mlibfuncs: MMIX Options. (line 10)
39456 * mlibrary-pic: FRV Options. (line 110)
39457 * mlinked-fp: FRV Options. (line 94)
39458 * mlinker-opt: HPPA Options. (line 85)
39459 * mlinux: CRIS Options. (line 104)
39460 * mlittle: RS/6000 and PowerPC Options.
39462 * mlittle-endian <1>: SPARC Options. (line 185)
39463 * mlittle-endian <2>: RS/6000 and PowerPC Options.
39465 * mlittle-endian <3>: MCore Options. (line 39)
39466 * mlittle-endian <4>: IA-64 Options. (line 13)
39467 * mlittle-endian: ARM Options. (line 68)
39468 * mllsc: MIPS Options. (line 222)
39469 * mlocal-sdata: MIPS Options. (line 303)
39470 * mlong-calls <1>: V850 Options. (line 10)
39471 * mlong-calls <2>: MIPS Options. (line 433)
39472 * mlong-calls <3>: M68hc1x Options. (line 35)
39473 * mlong-calls <4>: FRV Options. (line 99)
39474 * mlong-calls <5>: Blackfin Options. (line 116)
39475 * mlong-calls: ARM Options. (line 152)
39476 * mlong-double-128: S/390 and zSeries Options.
39478 * mlong-double-64: S/390 and zSeries Options.
39480 * mlong-load-store: HPPA Options. (line 66)
39481 * mlong32: MIPS Options. (line 278)
39482 * mlong64: MIPS Options. (line 273)
39483 * mlongcall: RS/6000 and PowerPC Options.
39485 * mlongcalls: Xtensa Options. (line 60)
39486 * mlow-64k: Blackfin Options. (line 65)
39487 * mlp64: IA-64 Options. (line 114)
39488 * MM: Preprocessor Options.
39490 * mmac <1>: Score Options. (line 21)
39491 * mmac: CRX Options. (line 9)
39492 * mmad: MIPS Options. (line 442)
39493 * mmangle-cpu: ARC Options. (line 15)
39494 * mmax: DEC Alpha Options. (line 171)
39495 * mmax-stack-frame: CRIS Options. (line 22)
39496 * mmcu: AVR Options. (line 9)
39497 * MMD: Preprocessor Options.
39499 * mmedia: FRV Options. (line 44)
39500 * mmemcpy: MIPS Options. (line 427)
39501 * mmemory-latency: DEC Alpha Options. (line 266)
39502 * mmfcrf: RS/6000 and PowerPC Options.
39504 * mmfpgpr: RS/6000 and PowerPC Options.
39506 * mminimal-toc: RS/6000 and PowerPC Options.
39508 * mmmx: i386 and x86-64 Options.
39510 * mmodel=large: M32R/D Options. (line 33)
39511 * mmodel=medium: M32R/D Options. (line 27)
39512 * mmodel=small: M32R/D Options. (line 18)
39513 * mmt: MIPS Options. (line 270)
39514 * mmul-bug-workaround: CRIS Options. (line 36)
39515 * mmuladd: FRV Options. (line 50)
39516 * mmulhw: RS/6000 and PowerPC Options.
39518 * mmult-bug: MN10300 Options. (line 9)
39519 * mmulti-cond-exec: FRV Options. (line 176)
39520 * mmultiple: RS/6000 and PowerPC Options.
39522 * mmvcle: S/390 and zSeries Options.
39524 * mmvme: RS/6000 and PowerPC Options.
39526 * mn: H8/300 Options. (line 20)
39527 * mnested-cond-exec: FRV Options. (line 189)
39528 * mnew-mnemonics: RS/6000 and PowerPC Options.
39530 * mnhwloop: Score Options. (line 15)
39531 * mno-3dnow: i386 and x86-64 Options.
39533 * mno-4byte-functions: MCore Options. (line 27)
39534 * mno-abicalls: MIPS Options. (line 144)
39535 * mno-abshi: PDP-11 Options. (line 58)
39536 * mno-ac0: PDP-11 Options. (line 20)
39537 * mno-align-double: i386 and x86-64 Options.
39539 * mno-align-int: M680x0 Options. (line 263)
39540 * mno-align-loops: M32R/D Options. (line 76)
39541 * mno-align-stringops: i386 and x86-64 Options.
39543 * mno-altivec: RS/6000 and PowerPC Options.
39545 * mno-am33: MN10300 Options. (line 20)
39546 * mno-app-regs <1>: V850 Options. (line 61)
39547 * mno-app-regs: SPARC Options. (line 10)
39548 * mno-bacc: MT Options. (line 19)
39549 * mno-backchain: S/390 and zSeries Options.
39551 * mno-base-addresses: MMIX Options. (line 54)
39552 * mno-bit-align: RS/6000 and PowerPC Options.
39554 * mno-bitfield: M680x0 Options. (line 227)
39555 * mno-branch-likely: MIPS Options. (line 526)
39556 * mno-branch-predict: MMIX Options. (line 49)
39557 * mno-bwx: DEC Alpha Options. (line 171)
39558 * mno-callgraph-data: MCore Options. (line 31)
39559 * mno-check-zero-division: MIPS Options. (line 406)
39560 * mno-cirrus-fix-invalid-insns: ARM Options. (line 190)
39561 * mno-cix: DEC Alpha Options. (line 171)
39562 * mno-cmpb: RS/6000 and PowerPC Options.
39564 * mno-cond-exec: FRV Options. (line 158)
39565 * mno-cond-move: FRV Options. (line 134)
39566 * mno-const-align: CRIS Options. (line 60)
39567 * mno-const16: Xtensa Options. (line 10)
39568 * mno-crt0 <1>: MT Options. (line 25)
39569 * mno-crt0: MN10300 Options. (line 31)
39570 * mno-csync-anomaly: Blackfin Options. (line 61)
39571 * mno-data-align: CRIS Options. (line 60)
39572 * mno-debug: S/390 and zSeries Options.
39574 * mno-div <1>: MCore Options. (line 15)
39575 * mno-div: M680x0 Options. (line 205)
39576 * mno-dlmzb: RS/6000 and PowerPC Options.
39578 * mno-double: FRV Options. (line 41)
39579 * mno-dsp: MIPS Options. (line 236)
39580 * mno-dspr2: MIPS Options. (line 242)
39581 * mno-dwarf2-asm: IA-64 Options. (line 79)
39582 * mno-dword: FRV Options. (line 35)
39583 * mno-eabi: RS/6000 and PowerPC Options.
39585 * mno-early-stop-bits: IA-64 Options. (line 85)
39586 * mno-eflags: FRV Options. (line 125)
39587 * mno-embedded-data: MIPS Options. (line 353)
39588 * mno-ep: V850 Options. (line 16)
39589 * mno-epsilon: MMIX Options. (line 15)
39590 * mno-explicit-relocs <1>: MIPS Options. (line 397)
39591 * mno-explicit-relocs: DEC Alpha Options. (line 184)
39592 * mno-extern-sdata: MIPS Options. (line 315)
39593 * mno-fancy-math-387: i386 and x86-64 Options.
39595 * mno-faster-structs: SPARC Options. (line 71)
39596 * mno-fix: DEC Alpha Options. (line 171)
39597 * mno-fix-r4000: MIPS Options. (line 462)
39598 * mno-fix-r4400: MIPS Options. (line 476)
39599 * mno-float32: PDP-11 Options. (line 48)
39600 * mno-float64: PDP-11 Options. (line 52)
39601 * mno-flush-func: M32R/D Options. (line 99)
39602 * mno-flush-trap: M32R/D Options. (line 91)
39603 * mno-fp-in-toc: RS/6000 and PowerPC Options.
39605 * mno-fp-regs: DEC Alpha Options. (line 25)
39606 * mno-fp-ret-in-387: i386 and x86-64 Options.
39608 * mno-fprnd: RS/6000 and PowerPC Options.
39610 * mno-fpu: SPARC Options. (line 25)
39611 * mno-fused-madd <1>: Xtensa Options. (line 19)
39612 * mno-fused-madd <2>: S/390 and zSeries Options.
39614 * mno-fused-madd <3>: RS/6000 and PowerPC Options.
39616 * mno-fused-madd: MIPS Options. (line 447)
39617 * mno-gnu-as: IA-64 Options. (line 18)
39618 * mno-gnu-ld: IA-64 Options. (line 23)
39619 * mno-gotplt: CRIS Options. (line 86)
39620 * mno-gpopt: MIPS Options. (line 338)
39621 * mno-hard-dfp: RS/6000 and PowerPC Options.
39623 * mno-hardlit: MCore Options. (line 10)
39624 * mno-id-shared-library: Blackfin Options. (line 83)
39625 * mno-ieee-fp: i386 and x86-64 Options.
39627 * mno-int16: PDP-11 Options. (line 44)
39628 * mno-int32: PDP-11 Options. (line 40)
39629 * mno-interlink-mips16: MIPS Options. (line 107)
39630 * mno-interrupts: AVR Options. (line 39)
39631 * mno-isel: RS/6000 and PowerPC Options.
39633 * mno-knuthdiv: MMIX Options. (line 33)
39634 * mno-leaf-id-shared-library: Blackfin Options. (line 93)
39635 * mno-libfuncs: MMIX Options. (line 10)
39636 * mno-llsc: MIPS Options. (line 222)
39637 * mno-local-sdata: MIPS Options. (line 303)
39638 * mno-long-calls <1>: V850 Options. (line 10)
39639 * mno-long-calls <2>: MIPS Options. (line 433)
39640 * mno-long-calls <3>: M68hc1x Options. (line 35)
39641 * mno-long-calls <4>: HPPA Options. (line 138)
39642 * mno-long-calls <5>: Blackfin Options. (line 116)
39643 * mno-long-calls: ARM Options. (line 152)
39644 * mno-longcall: RS/6000 and PowerPC Options.
39646 * mno-longcalls: Xtensa Options. (line 60)
39647 * mno-low-64k: Blackfin Options. (line 69)
39648 * mno-mad: MIPS Options. (line 442)
39649 * mno-max: DEC Alpha Options. (line 171)
39650 * mno-mdmx: MIPS Options. (line 259)
39651 * mno-media: FRV Options. (line 47)
39652 * mno-memcpy: MIPS Options. (line 427)
39653 * mno-mfcrf: RS/6000 and PowerPC Options.
39655 * mno-mfpgpr: RS/6000 and PowerPC Options.
39657 * mno-mips16: MIPS Options. (line 92)
39658 * mno-mips3d: MIPS Options. (line 265)
39659 * mno-mmx: i386 and x86-64 Options.
39661 * mno-mt: MIPS Options. (line 270)
39662 * mno-mul-bug-workaround: CRIS Options. (line 36)
39663 * mno-muladd: FRV Options. (line 53)
39664 * mno-mulhw: RS/6000 and PowerPC Options.
39666 * mno-mult-bug: MN10300 Options. (line 13)
39667 * mno-multi-cond-exec: FRV Options. (line 183)
39668 * mno-multiple: RS/6000 and PowerPC Options.
39670 * mno-mvcle: S/390 and zSeries Options.
39672 * mno-nested-cond-exec: FRV Options. (line 195)
39673 * mno-optimize-membar: FRV Options. (line 205)
39674 * mno-pack: FRV Options. (line 122)
39675 * mno-packed-stack: S/390 and zSeries Options.
39677 * mno-paired: RS/6000 and PowerPC Options.
39679 * mno-paired-single: MIPS Options. (line 253)
39680 * mno-pic: IA-64 Options. (line 26)
39681 * mno-popcntb: RS/6000 and PowerPC Options.
39683 * mno-power: RS/6000 and PowerPC Options.
39685 * mno-power2: RS/6000 and PowerPC Options.
39687 * mno-powerpc: RS/6000 and PowerPC Options.
39689 * mno-powerpc-gfxopt: RS/6000 and PowerPC Options.
39691 * mno-powerpc-gpopt: RS/6000 and PowerPC Options.
39693 * mno-powerpc64: RS/6000 and PowerPC Options.
39695 * mno-prolog-function: V850 Options. (line 23)
39696 * mno-prologue-epilogue: CRIS Options. (line 76)
39697 * mno-prototype: RS/6000 and PowerPC Options.
39699 * mno-push-args: i386 and x86-64 Options.
39701 * mno-register-names: IA-64 Options. (line 37)
39702 * mno-regnames: RS/6000 and PowerPC Options.
39704 * mno-relax-immediate: MCore Options. (line 19)
39705 * mno-relocatable: RS/6000 and PowerPC Options.
39707 * mno-relocatable-lib: RS/6000 and PowerPC Options.
39709 * mno-rtd: M680x0 Options. (line 258)
39710 * mno-scc: FRV Options. (line 146)
39711 * mno-sched-ar-data-spec: IA-64 Options. (line 128)
39712 * mno-sched-ar-in-data-spec: IA-64 Options. (line 149)
39713 * mno-sched-br-data-spec: IA-64 Options. (line 121)
39714 * mno-sched-br-in-data-spec: IA-64 Options. (line 142)
39715 * mno-sched-control-ldc: IA-64 Options. (line 168)
39716 * mno-sched-control-spec: IA-64 Options. (line 135)
39717 * mno-sched-count-spec-in-critical-path: IA-64 Options. (line 194)
39718 * mno-sched-in-control-spec: IA-64 Options. (line 156)
39719 * mno-sched-ldc: IA-64 Options. (line 162)
39720 * mno-sched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
39721 * mno-sched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
39722 * mno-sched-prolog: ARM Options. (line 32)
39723 * mno-sched-spec-verbose: IA-64 Options. (line 176)
39724 * mno-sdata <1>: RS/6000 and PowerPC Options.
39726 * mno-sdata: IA-64 Options. (line 42)
39727 * mno-sep-data: Blackfin Options. (line 111)
39728 * mno-short: M680x0 Options. (line 222)
39729 * mno-side-effects: CRIS Options. (line 51)
39730 * mno-single-exit: MMIX Options. (line 66)
39731 * mno-slow-bytes: MCore Options. (line 35)
39732 * mno-small-exec: S/390 and zSeries Options.
39734 * mno-smartmips: MIPS Options. (line 249)
39735 * mno-soft-float: DEC Alpha Options. (line 10)
39736 * mno-space-regs: HPPA Options. (line 45)
39737 * mno-spe: RS/6000 and PowerPC Options.
39739 * mno-specld-anomaly: Blackfin Options. (line 51)
39740 * mno-split: PDP-11 Options. (line 71)
39741 * mno-split-addresses: MIPS Options. (line 391)
39742 * mno-sse: i386 and x86-64 Options.
39744 * mno-stack-align: CRIS Options. (line 60)
39745 * mno-stack-bias: SPARC Options. (line 222)
39746 * mno-strict-align <1>: RS/6000 and PowerPC Options.
39748 * mno-strict-align: M680x0 Options. (line 283)
39749 * mno-string: RS/6000 and PowerPC Options.
39751 * mno-sum-in-toc: RS/6000 and PowerPC Options.
39753 * mno-swdiv: RS/6000 and PowerPC Options.
39755 * mno-sym32: MIPS Options. (line 288)
39756 * mno-tablejump: AVR Options. (line 47)
39757 * mno-target-align: Xtensa Options. (line 47)
39758 * mno-text-section-literals: Xtensa Options. (line 35)
39759 * mno-toc: RS/6000 and PowerPC Options.
39761 * mno-toplevel-symbols: MMIX Options. (line 40)
39762 * mno-tpf-trace: S/390 and zSeries Options.
39764 * mno-unaligned-doubles: SPARC Options. (line 59)
39765 * mno-uninit-const-in-rodata: MIPS Options. (line 361)
39766 * mno-update: RS/6000 and PowerPC Options.
39768 * mno-v8plus: SPARC Options. (line 170)
39769 * mno-vis: SPARC Options. (line 177)
39770 * mno-vliw-branch: FRV Options. (line 170)
39771 * mno-volatile-asm-stop: IA-64 Options. (line 32)
39772 * mno-vrsave: RS/6000 and PowerPC Options.
39774 * mno-wide-bitfields: MCore Options. (line 23)
39775 * mno-xgot: MIPS Options. (line 171)
39776 * mno-xl-compat: RS/6000 and PowerPC Options.
39778 * mno-zero-extend: MMIX Options. (line 27)
39779 * mnobitfield: M680x0 Options. (line 227)
39780 * mnomacsave: SH Options. (line 92)
39781 * mnominmax: M68hc1x Options. (line 31)
39782 * mnop-fun-dllimport: ARM Options. (line 177)
39783 * mold-mnemonics: RS/6000 and PowerPC Options.
39785 * momit-leaf-frame-pointer <1>: i386 and x86-64 Options.
39787 * momit-leaf-frame-pointer: Blackfin Options. (line 39)
39788 * mone-byte-bool: Darwin Options. (line 92)
39789 * moptimize-membar: FRV Options. (line 201)
39790 * MP: Preprocessor Options.
39792 * mpa-risc-1-0: HPPA Options. (line 19)
39793 * mpa-risc-1-1: HPPA Options. (line 19)
39794 * mpa-risc-2-0: HPPA Options. (line 19)
39795 * mpack: FRV Options. (line 119)
39796 * mpacked-stack: S/390 and zSeries Options.
39798 * mpadstruct: SH Options. (line 118)
39799 * mpaired: RS/6000 and PowerPC Options.
39801 * mpaired-single: MIPS Options. (line 253)
39802 * mpc32: i386 and x86-64 Options.
39804 * mpc64: i386 and x86-64 Options.
39806 * mpc80: i386 and x86-64 Options.
39808 * mpcrel: M680x0 Options. (line 275)
39809 * mpdebug: CRIS Options. (line 40)
39810 * mpe: RS/6000 and PowerPC Options.
39812 * mpic-register: ARM Options. (line 186)
39813 * mpoke-function-name: ARM Options. (line 200)
39814 * mpopcntb: RS/6000 and PowerPC Options.
39816 * mportable-runtime: HPPA Options. (line 71)
39817 * mpower: RS/6000 and PowerPC Options.
39819 * mpower2: RS/6000 and PowerPC Options.
39821 * mpowerpc: RS/6000 and PowerPC Options.
39823 * mpowerpc-gfxopt: RS/6000 and PowerPC Options.
39825 * mpowerpc-gpopt: RS/6000 and PowerPC Options.
39827 * mpowerpc64: RS/6000 and PowerPC Options.
39829 * mprefergot: SH Options. (line 125)
39830 * mpreferred-stack-boundary: i386 and x86-64 Options.
39832 * mprioritize-restricted-insns: RS/6000 and PowerPC Options.
39834 * mprolog-function: V850 Options. (line 23)
39835 * mprologue-epilogue: CRIS Options. (line 76)
39836 * mprototype: RS/6000 and PowerPC Options.
39838 * mpt-fixed: SH Options. (line 205)
39839 * mpush-args <1>: i386 and x86-64 Options.
39841 * mpush-args: CRX Options. (line 13)
39842 * MQ: Preprocessor Options.
39844 * mrecip: i386 and x86-64 Options.
39846 * mregister-names: IA-64 Options. (line 37)
39847 * mregnames: RS/6000 and PowerPC Options.
39849 * mregparm: i386 and x86-64 Options.
39851 * mrelax <1>: SH Options. (line 70)
39852 * mrelax <2>: MN10300 Options. (line 34)
39853 * mrelax: H8/300 Options. (line 9)
39854 * mrelax-immediate: MCore Options. (line 19)
39855 * mrelocatable: RS/6000 and PowerPC Options.
39857 * mrelocatable-lib: RS/6000 and PowerPC Options.
39859 * mreturn-pointer-on-d0: MN10300 Options. (line 24)
39860 * mrodata: ARC Options. (line 30)
39861 * mrtd <1>: Function Attributes.
39863 * mrtd <2>: M680x0 Options. (line 236)
39864 * mrtd: i386 and x86-64 Options.
39866 * mrtp: VxWorks Options. (line 11)
39867 * ms: H8/300 Options. (line 17)
39868 * ms2600: H8/300 Options. (line 24)
39869 * msafe-dma: SPU Options. (line 17)
39870 * msahf: i386 and x86-64 Options.
39872 * mscc: FRV Options. (line 140)
39873 * msched-ar-data-spec: IA-64 Options. (line 128)
39874 * msched-ar-in-data-spec: IA-64 Options. (line 149)
39875 * msched-br-data-spec: IA-64 Options. (line 121)
39876 * msched-br-in-data-spec: IA-64 Options. (line 142)
39877 * msched-control-ldc: IA-64 Options. (line 168)
39878 * msched-control-spec: IA-64 Options. (line 135)
39879 * msched-costly-dep: RS/6000 and PowerPC Options.
39881 * msched-count-spec-in-critical-path: IA-64 Options. (line 194)
39882 * msched-in-control-spec: IA-64 Options. (line 156)
39883 * msched-ldc: IA-64 Options. (line 162)
39884 * msched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
39885 * msched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
39886 * msched-spec-verbose: IA-64 Options. (line 176)
39887 * mschedule: HPPA Options. (line 78)
39888 * mscore5: Score Options. (line 25)
39889 * mscore5u: Score Options. (line 28)
39890 * mscore7: Score Options. (line 31)
39891 * mscore7d: Score Options. (line 34)
39892 * msda: V850 Options. (line 40)
39893 * msdata <1>: RS/6000 and PowerPC Options.
39895 * msdata: IA-64 Options. (line 42)
39896 * msdata-data: RS/6000 and PowerPC Options.
39898 * msdata=default: RS/6000 and PowerPC Options.
39900 * msdata=eabi: RS/6000 and PowerPC Options.
39902 * msdata=none <1>: RS/6000 and PowerPC Options.
39904 * msdata=none: M32R/D Options. (line 40)
39905 * msdata=sdata: M32R/D Options. (line 49)
39906 * msdata=sysv: RS/6000 and PowerPC Options.
39908 * msdata=use: M32R/D Options. (line 53)
39909 * msecure-plt: RS/6000 and PowerPC Options.
39911 * msep-data: Blackfin Options. (line 105)
39912 * mshared-library-id: Blackfin Options. (line 98)
39913 * mshort <1>: M68hc1x Options. (line 40)
39914 * mshort: M680x0 Options. (line 216)
39915 * msim <1>: Xstormy16 Options. (line 9)
39916 * msim <2>: RS/6000 and PowerPC Options.
39918 * msim <3>: MT Options. (line 22)
39919 * msim <4>: M32C Options. (line 13)
39920 * msim: Blackfin Options. (line 32)
39921 * msingle-exit: MMIX Options. (line 66)
39922 * msingle-float: MIPS Options. (line 213)
39923 * msingle-pic-base: ARM Options. (line 180)
39924 * msio: HPPA Options. (line 107)
39925 * msize: AVR Options. (line 32)
39926 * mslow-bytes: MCore Options. (line 35)
39927 * msmall-data: DEC Alpha Options. (line 195)
39928 * msmall-exec: S/390 and zSeries Options.
39930 * msmall-mem: SPU Options. (line 35)
39931 * msmall-text: DEC Alpha Options. (line 213)
39932 * msmartmips: MIPS Options. (line 249)
39933 * msoft-float <1>: SPARC Options. (line 25)
39934 * msoft-float <2>: S/390 and zSeries Options.
39936 * msoft-float <3>: RS/6000 and PowerPC Options.
39938 * msoft-float <4>: PDP-11 Options. (line 13)
39939 * msoft-float <5>: MIPS Options. (line 209)
39940 * msoft-float <6>: M680x0 Options. (line 199)
39941 * msoft-float <7>: i386 and x86-64 Options.
39943 * msoft-float <8>: HPPA Options. (line 91)
39944 * msoft-float <9>: FRV Options. (line 22)
39945 * msoft-float <10>: DEC Alpha Options. (line 10)
39946 * msoft-float: ARM Options. (line 45)
39947 * msoft-quad-float: SPARC Options. (line 45)
39948 * msoft-reg-count: M68hc1x Options. (line 43)
39949 * mspace <1>: V850 Options. (line 30)
39950 * mspace: SH Options. (line 122)
39951 * mspe: RS/6000 and PowerPC Options.
39953 * mspecld-anomaly: Blackfin Options. (line 46)
39954 * msplit: PDP-11 Options. (line 68)
39955 * msplit-addresses: MIPS Options. (line 391)
39956 * msse: i386 and x86-64 Options.
39958 * msseregparm: i386 and x86-64 Options.
39960 * mstack-align: CRIS Options. (line 60)
39961 * mstack-bias: SPARC Options. (line 222)
39962 * mstack-check-l1: Blackfin Options. (line 72)
39963 * mstack-guard: S/390 and zSeries Options.
39965 * mstack-size: S/390 and zSeries Options.
39967 * mstackrealign: i386 and x86-64 Options.
39969 * mstdmain: SPU Options. (line 40)
39970 * mstrict-align <1>: RS/6000 and PowerPC Options.
39972 * mstrict-align: M680x0 Options. (line 283)
39973 * mstring: RS/6000 and PowerPC Options.
39975 * mstringop-strategy=ALG: i386 and x86-64 Options.
39977 * mstructure-size-boundary: ARM Options. (line 132)
39978 * msvr4-struct-return: RS/6000 and PowerPC Options.
39980 * mswdiv: RS/6000 and PowerPC Options.
39982 * msym32: MIPS Options. (line 288)
39983 * mt: IA-64 Options. (line 106)
39984 * MT: Preprocessor Options.
39986 * mtarget-align: Xtensa Options. (line 47)
39987 * mtda: V850 Options. (line 34)
39988 * mtext: ARC Options. (line 30)
39989 * mtext-section-literals: Xtensa Options. (line 35)
39990 * mthreads: i386 and x86-64 Options.
39992 * mthumb: ARM Options. (line 221)
39993 * mthumb-interwork: ARM Options. (line 25)
39994 * mtiny-stack: AVR Options. (line 50)
39995 * mtls-direct-seg-refs: i386 and x86-64 Options.
39997 * mtls-size: IA-64 Options. (line 97)
39998 * mtoc: RS/6000 and PowerPC Options.
40000 * mtomcat-stats: FRV Options. (line 209)
40001 * mtoplevel-symbols: MMIX Options. (line 40)
40002 * mtp: ARM Options. (line 251)
40003 * mtpcs-frame: ARM Options. (line 227)
40004 * mtpcs-leaf-frame: ARM Options. (line 233)
40005 * mtpf-trace: S/390 and zSeries Options.
40007 * mtrap-precision: DEC Alpha Options. (line 109)
40008 * mtune <1>: SPARC Options. (line 158)
40009 * mtune <2>: S/390 and zSeries Options.
40011 * mtune <3>: RS/6000 and PowerPC Options.
40013 * mtune <4>: MIPS Options. (line 55)
40014 * mtune <5>: M680x0 Options. (line 66)
40015 * mtune <6>: IA-64 Options. (line 101)
40016 * mtune <7>: i386 and x86-64 Options.
40018 * mtune <8>: DEC Alpha Options. (line 262)
40019 * mtune <9>: CRIS Options. (line 16)
40020 * mtune: ARM Options. (line 100)
40021 * muclibc: GNU/Linux Options. (line 13)
40022 * muls: Score Options. (line 18)
40023 * multcost=NUMBER: SH Options. (line 135)
40024 * multi_module: Darwin Options. (line 199)
40025 * multilib-library-pic: FRV Options. (line 89)
40026 * multiply_defined: Darwin Options. (line 199)
40027 * multiply_defined_unused: Darwin Options. (line 199)
40028 * munaligned-doubles: SPARC Options. (line 59)
40029 * muninit-const-in-rodata: MIPS Options. (line 361)
40030 * munix: VAX Options. (line 9)
40031 * munix-asm: PDP-11 Options. (line 74)
40032 * munsafe-dma: SPU Options. (line 17)
40033 * mupdate: RS/6000 and PowerPC Options.
40035 * musermode: SH Options. (line 130)
40036 * mv850: V850 Options. (line 49)
40037 * mv850e: V850 Options. (line 69)
40038 * mv850e1: V850 Options. (line 64)
40039 * mv8plus: SPARC Options. (line 170)
40040 * mveclibabi: i386 and x86-64 Options.
40042 * mvis: SPARC Options. (line 177)
40043 * mvliw-branch: FRV Options. (line 164)
40044 * mvms-return-codes: DEC Alpha/VMS Options.
40046 * mvolatile-asm-stop: IA-64 Options. (line 32)
40047 * mvr4130-align: MIPS Options. (line 547)
40048 * mvrsave: RS/6000 and PowerPC Options.
40050 * mvxworks: RS/6000 and PowerPC Options.
40052 * mwarn-dynamicstack: S/390 and zSeries Options.
40054 * mwarn-framesize: S/390 and zSeries Options.
40056 * mwarn-reloc: SPU Options. (line 10)
40057 * mwide-bitfields: MCore Options. (line 23)
40058 * mwindiss: RS/6000 and PowerPC Options.
40060 * mwords-little-endian: ARM Options. (line 76)
40061 * mxgot: MIPS Options. (line 171)
40062 * mxl-compat: RS/6000 and PowerPC Options.
40064 * myellowknife: RS/6000 and PowerPC Options.
40066 * mzarch: S/390 and zSeries Options.
40068 * mzda: V850 Options. (line 45)
40069 * mzero-extend: MMIX Options. (line 27)
40070 * no-integrated-cpp: C Dialect Options. (line 240)
40071 * no-red-zone: i386 and x86-64 Options.
40073 * no_dead_strip_inits_and_terms: Darwin Options. (line 199)
40074 * noall_load: Darwin Options. (line 199)
40075 * nocpp: MIPS Options. (line 457)
40076 * nodefaultlibs: Link Options. (line 62)
40077 * nofixprebinding: Darwin Options. (line 199)
40078 * nolibdld: HPPA Options. (line 190)
40079 * nomultidefs: Darwin Options. (line 199)
40080 * non-static: VxWorks Options. (line 16)
40081 * noprebind: Darwin Options. (line 199)
40082 * noseglinkedit: Darwin Options. (line 199)
40083 * nostartfiles: Link Options. (line 57)
40084 * nostdinc: Preprocessor Options.
40086 * nostdinc++ <1>: Preprocessor Options.
40088 * nostdinc++: C++ Dialect Options.
40090 * nostdlib: Link Options. (line 71)
40091 * o: Preprocessor Options.
40093 * O: Optimize Options. (line 32)
40094 * o: Overall Options. (line 182)
40095 * O0: Optimize Options. (line 107)
40096 * O1: Optimize Options. (line 32)
40097 * O2: Optimize Options. (line 69)
40098 * O3: Optimize Options. (line 101)
40099 * Os: Optimize Options. (line 111)
40100 * P: Preprocessor Options.
40102 * p: Debugging Options. (line 214)
40103 * pagezero_size: Darwin Options. (line 199)
40104 * param: Optimize Options. (line 1501)
40105 * pass-exit-codes: Overall Options. (line 140)
40106 * pedantic <1>: Warnings and Errors.
40108 * pedantic <2>: Alternate Keywords. (line 29)
40109 * pedantic <3>: C Extensions. (line 6)
40110 * pedantic <4>: Preprocessor Options.
40112 * pedantic <5>: Warning Options. (line 53)
40113 * pedantic: Standards. (line 16)
40114 * pedantic-errors <1>: Warnings and Errors.
40116 * pedantic-errors <2>: Non-bugs. (line 216)
40117 * pedantic-errors <3>: Preprocessor Options.
40119 * pedantic-errors <4>: Warning Options. (line 95)
40120 * pedantic-errors: Standards. (line 16)
40121 * pg: Debugging Options. (line 220)
40122 * pie: Link Options. (line 92)
40123 * pipe: Overall Options. (line 204)
40124 * prebind: Darwin Options. (line 199)
40125 * prebind_all_twolevel_modules: Darwin Options. (line 199)
40126 * preprocessor: Preprocessor Options.
40128 * print-file-name: Debugging Options. (line 832)
40129 * print-libgcc-file-name: Debugging Options. (line 853)
40130 * print-multi-directory: Debugging Options. (line 838)
40131 * print-multi-lib: Debugging Options. (line 843)
40132 * print-objc-runtime-info: Objective-C and Objective-C++ Dialect Options.
40134 * print-prog-name: Debugging Options. (line 850)
40135 * print-search-dirs: Debugging Options. (line 861)
40136 * print-sysroot-headers-suffix: Debugging Options. (line 874)
40137 * private_bundle: Darwin Options. (line 199)
40138 * pthread <1>: SPARC Options. (line 242)
40139 * pthread <2>: RS/6000 and PowerPC Options.
40141 * pthread: IA-64 Options. (line 106)
40142 * pthreads: SPARC Options. (line 236)
40143 * Q: Debugging Options. (line 226)
40144 * Qn: System V Options. (line 18)
40145 * Qy: System V Options. (line 14)
40146 * rdynamic: Link Options. (line 98)
40147 * read_only_relocs: Darwin Options. (line 199)
40148 * remap: Preprocessor Options.
40150 * s: Link Options. (line 105)
40151 * S <1>: Link Options. (line 20)
40152 * S: Overall Options. (line 165)
40153 * save-temps: Debugging Options. (line 794)
40154 * sectalign: Darwin Options. (line 199)
40155 * sectcreate: Darwin Options. (line 199)
40156 * sectobjectsymbols: Darwin Options. (line 199)
40157 * sectorder: Darwin Options. (line 199)
40158 * seg1addr: Darwin Options. (line 199)
40159 * seg_addr_table: Darwin Options. (line 199)
40160 * seg_addr_table_filename: Darwin Options. (line 199)
40161 * segaddr: Darwin Options. (line 199)
40162 * seglinkedit: Darwin Options. (line 199)
40163 * segprot: Darwin Options. (line 199)
40164 * segs_read_only_addr: Darwin Options. (line 199)
40165 * segs_read_write_addr: Darwin Options. (line 199)
40166 * shared: Link Options. (line 114)
40167 * shared-libgcc: Link Options. (line 122)
40168 * sim: CRIS Options. (line 108)
40169 * sim2: CRIS Options. (line 114)
40170 * single_module: Darwin Options. (line 199)
40171 * specs: Directory Options. (line 84)
40172 * static <1>: HPPA Options. (line 194)
40173 * static <2>: Darwin Options. (line 199)
40174 * static: Link Options. (line 109)
40175 * static-libgcc: Link Options. (line 122)
40176 * std <1>: Non-bugs. (line 107)
40177 * std <2>: Other Builtins. (line 22)
40178 * std <3>: C Dialect Options. (line 47)
40179 * std: Standards. (line 16)
40180 * std=: Preprocessor Options.
40182 * sub_library: Darwin Options. (line 199)
40183 * sub_umbrella: Darwin Options. (line 199)
40184 * symbolic: Link Options. (line 157)
40185 * sysroot: Directory Options. (line 92)
40186 * target-help <1>: Preprocessor Options.
40188 * target-help: Overall Options. (line 235)
40189 * threads <1>: SPARC Options. (line 230)
40190 * threads: HPPA Options. (line 207)
40191 * time: Debugging Options. (line 808)
40192 * tls: FRV Options. (line 75)
40193 * TLS: FRV Options. (line 72)
40194 * traditional <1>: Incompatibilities. (line 6)
40195 * traditional: C Dialect Options. (line 252)
40196 * traditional-cpp <1>: Preprocessor Options.
40198 * traditional-cpp: C Dialect Options. (line 252)
40199 * trigraphs <1>: Preprocessor Options.
40201 * trigraphs: C Dialect Options. (line 236)
40202 * twolevel_namespace: Darwin Options. (line 199)
40203 * u: Link Options. (line 179)
40204 * U: Preprocessor Options.
40206 * umbrella: Darwin Options. (line 199)
40207 * undef: Preprocessor Options.
40209 * undefined: Darwin Options. (line 199)
40210 * unexported_symbols_list: Darwin Options. (line 199)
40211 * V: Target Options. (line 24)
40212 * v <1>: Preprocessor Options.
40214 * v: Overall Options. (line 193)
40215 * version <1>: Preprocessor Options.
40217 * version: Overall Options. (line 343)
40218 * W: Incompatibilities. (line 64)
40219 * w: Preprocessor Options.
40221 * W: Warning Options. (line 145)
40222 * w: Warning Options. (line 18)
40223 * Wa: Assembler Options. (line 9)
40224 * Wabi: C++ Dialect Options.
40226 * Waddress: Warning Options. (line 935)
40227 * Waggregate-return: Warning Options. (line 953)
40228 * Wall <1>: Standard Libraries. (line 6)
40229 * Wall <2>: Preprocessor Options.
40231 * Wall: Warning Options. (line 99)
40232 * Warray-bounds: Warning Options. (line 692)
40233 * Wassign-intercept: Objective-C and Objective-C++ Dialect Options.
40235 * Wattributes: Warning Options. (line 958)
40236 * Wbad-function-cast: Warning Options. (line 855)
40237 * Wcast-align: Warning Options. (line 875)
40238 * Wcast-qual: Warning Options. (line 870)
40239 * Wchar-subscripts: Warning Options. (line 189)
40240 * Wclobbered: Warning Options. (line 893)
40241 * Wcomment <1>: Preprocessor Options.
40243 * Wcomment: Warning Options. (line 194)
40244 * Wcomments: Preprocessor Options.
40246 * Wconversion: Warning Options. (line 897)
40247 * Wcoverage-mismatch: Language Independent Options.
40249 * Wctor-dtor-privacy: C++ Dialect Options.
40251 * Wdeclaration-after-statement: Warning Options. (line 813)
40252 * Wdeprecated: C++ Dialect Options.
40254 * Wdeprecated-declarations: Warning Options. (line 1096)
40255 * Wdisabled-optimization: Warning Options. (line 1227)
40256 * Wdiv-by-zero: Warning Options. (line 697)
40257 * weak_reference_mismatches: Darwin Options. (line 199)
40258 * Weffc++: C++ Dialect Options.
40260 * Wempty-body: Warning Options. (line 916)
40261 * Wendif-labels <1>: Preprocessor Options.
40263 * Wendif-labels: Warning Options. (line 823)
40264 * Werror <1>: Preprocessor Options.
40266 * Werror: Warning Options. (line 21)
40267 * Werror=: Warning Options. (line 24)
40268 * Wextra: Warning Options. (line 145)
40269 * Wfatal-errors: Warning Options. (line 38)
40270 * Wfloat-equal: Warning Options. (line 713)
40271 * Wformat <1>: Function Attributes.
40273 * Wformat: Warning Options. (line 199)
40274 * Wformat-extra-args: Warning Options. (line 238)
40275 * Wformat-nonliteral <1>: Function Attributes.
40277 * Wformat-nonliteral: Warning Options. (line 256)
40278 * Wformat-security: Warning Options. (line 261)
40279 * Wformat-y2k: Warning Options. (line 234)
40280 * Wformat-zero-length: Warning Options. (line 252)
40281 * Wformat=2: Warning Options. (line 272)
40282 * whatsloaded: Darwin Options. (line 199)
40283 * whyload: Darwin Options. (line 199)
40284 * Wignored-qualifiers: Warning Options. (line 312)
40285 * Wimplicit: Warning Options. (line 308)
40286 * Wimplicit-function-declaration: Warning Options. (line 302)
40287 * Wimplicit-int: Warning Options. (line 298)
40288 * Wimport <1>: Preprocessor Options.
40290 * Wimport: Warning Options. (line 186)
40291 * Winit-self: Warning Options. (line 284)
40292 * Winline <1>: Inline. (line 63)
40293 * Winline: Warning Options. (line 1167)
40294 * Wint-to-pointer-cast: Warning Options. (line 1194)
40295 * Winvalid-offsetof: Warning Options. (line 1180)
40296 * Winvalid-pch: Warning Options. (line 1202)
40297 * Wl: Link Options. (line 175)
40298 * Wlarger-than-LEN: Warning Options. (line 832)
40299 * Wlogical-op: Warning Options. (line 948)
40300 * Wlong-long: Warning Options. (line 1206)
40301 * Wmain: Warning Options. (line 323)
40302 * Wmissing-braces: Warning Options. (line 329)
40303 * Wmissing-declarations: Warning Options. (line 994)
40304 * Wmissing-field-initializers: Warning Options. (line 1002)
40305 * Wmissing-format-attribute: Warning Options. (line 1028)
40306 * Wmissing-include-dirs: Warning Options. (line 339)
40307 * Wmissing-noreturn: Warning Options. (line 1020)
40308 * Wmissing-parameter-type: Warning Options. (line 980)
40309 * Wmissing-prototypes: Warning Options. (line 988)
40310 * Wmultichar: Warning Options. (line 1047)
40311 * Wnested-externs: Warning Options. (line 1142)
40312 * Wno-abi: C++ Dialect Options.
40314 * Wno-address: Warning Options. (line 935)
40315 * Wno-aggregate-return: Warning Options. (line 953)
40316 * Wno-all: Warning Options. (line 99)
40317 * Wno-array-bounds: Warning Options. (line 692)
40318 * Wno-assign-intercept: Objective-C and Objective-C++ Dialect Options.
40320 * Wno-attributes: Warning Options. (line 958)
40321 * Wno-bad-function-cast: Warning Options. (line 855)
40322 * Wno-cast-align: Warning Options. (line 875)
40323 * Wno-cast-qual: Warning Options. (line 870)
40324 * Wno-char-subscripts: Warning Options. (line 189)
40325 * Wno-clobbered: Warning Options. (line 893)
40326 * Wno-comment: Warning Options. (line 194)
40327 * Wno-conversion: Warning Options. (line 897)
40328 * Wno-ctor-dtor-privacy: C++ Dialect Options.
40330 * Wno-declaration-after-statement: Warning Options. (line 813)
40331 * Wno-deprecated: C++ Dialect Options.
40333 * Wno-deprecated-declarations: Warning Options. (line 1096)
40334 * Wno-disabled-optimization: Warning Options. (line 1227)
40335 * Wno-div-by-zero: Warning Options. (line 697)
40336 * Wno-effc++: C++ Dialect Options.
40338 * Wno-empty-body: Warning Options. (line 916)
40339 * Wno-endif-labels: Warning Options. (line 823)
40340 * Wno-error: Warning Options. (line 21)
40341 * Wno-error=: Warning Options. (line 24)
40342 * Wno-extra: Warning Options. (line 145)
40343 * Wno-fatal-errors: Warning Options. (line 38)
40344 * Wno-float-equal: Warning Options. (line 713)
40345 * Wno-format: Warning Options. (line 199)
40346 * Wno-format-extra-args: Warning Options. (line 238)
40347 * Wno-format-nonliteral: Warning Options. (line 256)
40348 * Wno-format-security: Warning Options. (line 261)
40349 * Wno-format-y2k: Warning Options. (line 234)
40350 * Wno-format-zero-length: Warning Options. (line 252)
40351 * Wno-format=2: Warning Options. (line 272)
40352 * Wno-ignored-qualifiers: Warning Options. (line 312)
40353 * Wno-implicit: Warning Options. (line 308)
40354 * Wno-implicit-function-declaration: Warning Options. (line 302)
40355 * Wno-implicit-int: Warning Options. (line 298)
40356 * Wno-import: Warning Options. (line 186)
40357 * Wno-init-self: Warning Options. (line 284)
40358 * Wno-inline: Warning Options. (line 1167)
40359 * Wno-int-to-pointer-cast: Warning Options. (line 1194)
40360 * Wno-invalid-offsetof: Warning Options. (line 1180)
40361 * Wno-invalid-pch: Warning Options. (line 1202)
40362 * Wno-logical-op: Warning Options. (line 948)
40363 * Wno-long-long: Warning Options. (line 1206)
40364 * Wno-main: Warning Options. (line 323)
40365 * Wno-missing-braces: Warning Options. (line 329)
40366 * Wno-missing-declarations: Warning Options. (line 994)
40367 * Wno-missing-field-initializers: Warning Options. (line 1002)
40368 * Wno-missing-format-attribute: Warning Options. (line 1028)
40369 * Wno-missing-include-dirs: Warning Options. (line 339)
40370 * Wno-missing-noreturn: Warning Options. (line 1020)
40371 * Wno-missing-parameter-type: Warning Options. (line 980)
40372 * Wno-missing-prototypes: Warning Options. (line 988)
40373 * Wno-multichar: Warning Options. (line 1047)
40374 * Wno-nested-externs: Warning Options. (line 1142)
40375 * Wno-non-template-friend: C++ Dialect Options.
40377 * Wno-non-virtual-dtor: C++ Dialect Options.
40379 * Wno-nonnull: Warning Options. (line 277)
40380 * Wno-old-style-cast: C++ Dialect Options.
40382 * Wno-old-style-declaration: Warning Options. (line 970)
40383 * Wno-old-style-definition: Warning Options. (line 976)
40384 * Wno-overflow: Warning Options. (line 1102)
40385 * Wno-overlength-strings: Warning Options. (line 1247)
40386 * Wno-overloaded-virtual: C++ Dialect Options.
40388 * Wno-override-init: Warning Options. (line 1105)
40389 * Wno-packed: Warning Options. (line 1113)
40390 * Wno-padded: Warning Options. (line 1130)
40391 * Wno-parentheses: Warning Options. (line 342)
40392 * Wno-pmf-conversions <1>: Bound member functions.
40394 * Wno-pmf-conversions: C++ Dialect Options.
40396 * Wno-pointer-arith: Warning Options. (line 841)
40397 * Wno-pointer-sign: Warning Options. (line 1236)
40398 * Wno-pointer-to-int-cast: Warning Options. (line 1198)
40399 * Wno-pragmas: Warning Options. (line 595)
40400 * Wno-protocol: Objective-C and Objective-C++ Dialect Options.
40402 * Wno-redundant-decls: Warning Options. (line 1137)
40403 * Wno-reorder: C++ Dialect Options.
40405 * Wno-return-type: Warning Options. (line 432)
40406 * Wno-selector: Objective-C and Objective-C++ Dialect Options.
40408 * Wno-sequence-point: Warning Options. (line 386)
40409 * Wno-shadow: Warning Options. (line 827)
40410 * Wno-sign-compare: Warning Options. (line 922)
40411 * Wno-sign-conversion: Warning Options. (line 929)
40412 * Wno-sign-promo: C++ Dialect Options.
40414 * Wno-stack-protector: Warning Options. (line 1242)
40415 * Wno-strict-aliasing: Warning Options. (line 600)
40416 * Wno-strict-aliasing=n: Warning Options. (line 608)
40417 * Wno-strict-null-sentinel: C++ Dialect Options.
40419 * Wno-strict-overflow: Warning Options. (line 641)
40420 * Wno-strict-prototypes: Warning Options. (line 964)
40421 * Wno-strict-selector-match: Objective-C and Objective-C++ Dialect Options.
40423 * Wno-switch: Warning Options. (line 447)
40424 * Wno-switch-default: Warning Options. (line 455)
40425 * Wno-switch-enum: Warning Options. (line 458)
40426 * Wno-system-headers: Warning Options. (line 702)
40427 * Wno-traditional: Warning Options. (line 728)
40428 * Wno-traditional-conversion: Warning Options. (line 805)
40429 * Wno-trigraphs: Warning Options. (line 464)
40430 * Wno-type-limits: Warning Options. (line 848)
40431 * Wno-undeclared-selector: Objective-C and Objective-C++ Dialect Options.
40433 * Wno-undef: Warning Options. (line 820)
40434 * Wno-uninitialized: Warning Options. (line 513)
40435 * Wno-unknown-pragmas: Warning Options. (line 588)
40436 * Wno-unreachable-code: Warning Options. (line 1145)
40437 * Wno-unsafe-loop-optimizations: Warning Options. (line 835)
40438 * Wno-unused: Warning Options. (line 506)
40439 * Wno-unused-function: Warning Options. (line 469)
40440 * Wno-unused-label: Warning Options. (line 474)
40441 * Wno-unused-parameter: Warning Options. (line 481)
40442 * Wno-unused-value: Warning Options. (line 496)
40443 * Wno-unused-variable: Warning Options. (line 488)
40444 * Wno-variadic-macros: Warning Options. (line 1212)
40445 * Wno-vla: Warning Options. (line 1218)
40446 * Wno-volatile-register-var: Warning Options. (line 1222)
40447 * Wno-write-strings: Warning Options. (line 881)
40448 * Wnon-template-friend: C++ Dialect Options.
40450 * Wnon-virtual-dtor: C++ Dialect Options.
40452 * Wnonnull: Warning Options. (line 277)
40453 * Wnormalized=: Warning Options. (line 1053)
40454 * Wold-style-cast: C++ Dialect Options.
40456 * Wold-style-declaration: Warning Options. (line 970)
40457 * Wold-style-definition: Warning Options. (line 976)
40458 * Woverflow: Warning Options. (line 1102)
40459 * Woverlength-strings: Warning Options. (line 1247)
40460 * Woverloaded-virtual: C++ Dialect Options.
40462 * Woverride-init: Warning Options. (line 1105)
40463 * Wp: Preprocessor Options.
40465 * Wpacked: Warning Options. (line 1113)
40466 * Wpadded: Warning Options. (line 1130)
40467 * Wparentheses: Warning Options. (line 342)
40468 * Wpmf-conversions: C++ Dialect Options.
40470 * Wpointer-arith <1>: Pointer Arith. (line 13)
40471 * Wpointer-arith: Warning Options. (line 841)
40472 * Wpointer-sign: Warning Options. (line 1236)
40473 * Wpointer-to-int-cast: Warning Options. (line 1198)
40474 * Wpragmas: Warning Options. (line 595)
40475 * Wprotocol: Objective-C and Objective-C++ Dialect Options.
40477 * Wredundant-decls: Warning Options. (line 1137)
40478 * Wreorder: C++ Dialect Options.
40480 * Wreturn-type: Warning Options. (line 432)
40481 * Wselector: Objective-C and Objective-C++ Dialect Options.
40483 * Wsequence-point: Warning Options. (line 386)
40484 * Wshadow: Warning Options. (line 827)
40485 * Wsign-compare: Warning Options. (line 922)
40486 * Wsign-conversion: Warning Options. (line 929)
40487 * Wsign-promo: C++ Dialect Options.
40489 * Wstack-protector: Warning Options. (line 1242)
40490 * Wstrict-aliasing: Warning Options. (line 600)
40491 * Wstrict-aliasing=n: Warning Options. (line 608)
40492 * Wstrict-null-sentinel: C++ Dialect Options.
40494 * Wstrict-overflow: Warning Options. (line 641)
40495 * Wstrict-prototypes: Warning Options. (line 964)
40496 * Wstrict-selector-match: Objective-C and Objective-C++ Dialect Options.
40498 * Wswitch: Warning Options. (line 447)
40499 * Wswitch-default: Warning Options. (line 455)
40500 * Wswitch-enum: Warning Options. (line 458)
40501 * Wsystem-headers <1>: Preprocessor Options.
40503 * Wsystem-headers: Warning Options. (line 702)
40504 * Wtraditional <1>: Preprocessor Options.
40506 * Wtraditional: Warning Options. (line 728)
40507 * Wtraditional-conversion <1>: Protoize Caveats. (line 31)
40508 * Wtraditional-conversion: Warning Options. (line 805)
40509 * Wtrigraphs <1>: Preprocessor Options.
40511 * Wtrigraphs: Warning Options. (line 464)
40512 * Wtype-limits: Warning Options. (line 848)
40513 * Wundeclared-selector: Objective-C and Objective-C++ Dialect Options.
40515 * Wundef <1>: Preprocessor Options.
40517 * Wundef: Warning Options. (line 820)
40518 * Wuninitialized: Warning Options. (line 513)
40519 * Wunknown-pragmas: Warning Options. (line 588)
40520 * Wunreachable-code: Warning Options. (line 1145)
40521 * Wunsafe-loop-optimizations: Warning Options. (line 835)
40522 * Wunused: Warning Options. (line 506)
40523 * Wunused-function: Warning Options. (line 469)
40524 * Wunused-label: Warning Options. (line 474)
40525 * Wunused-macros: Preprocessor Options.
40527 * Wunused-parameter: Warning Options. (line 481)
40528 * Wunused-value: Warning Options. (line 496)
40529 * Wunused-variable: Warning Options. (line 488)
40530 * Wvariadic-macros: Warning Options. (line 1212)
40531 * Wvla: Warning Options. (line 1218)
40532 * Wvolatile-register-var: Warning Options. (line 1222)
40533 * Wwrite-strings: Warning Options. (line 881)
40534 * x <1>: Preprocessor Options.
40536 * x: Overall Options. (line 116)
40537 * Xassembler: Assembler Options. (line 13)
40538 * Xbind-lazy: VxWorks Options. (line 26)
40539 * Xbind-now: VxWorks Options. (line 30)
40540 * Xlinker: Link Options. (line 163)
40541 * Ym: System V Options. (line 26)
40542 * YP: System V Options. (line 22)
40545 File: gcc.info, Node: Keyword Index, Prev: Option Index, Up: Top
40553 * ! in constraint: Multi-Alternative. (line 33)
40554 * # in constraint: Modifiers. (line 57)
40555 * #pragma: Pragmas. (line 6)
40556 * #pragma implementation: C++ Interface. (line 39)
40557 * #pragma implementation, implied: C++ Interface. (line 46)
40558 * #pragma interface: C++ Interface. (line 20)
40559 * #pragma, reason for not using: Function Attributes.
40561 * $: Dollar Signs. (line 6)
40562 * % in constraint: Modifiers. (line 45)
40563 * %include: Spec Files. (line 27)
40564 * %include_noerr: Spec Files. (line 31)
40565 * %rename: Spec Files. (line 35)
40566 * & in constraint: Modifiers. (line 25)
40567 * ': Incompatibilities. (line 116)
40568 * (: Constructing Calls. (line 53)
40569 * * in constraint: Modifiers. (line 62)
40570 * + in constraint: Modifiers. (line 12)
40571 * -lgcc, use with -nodefaultlibs: Link Options. (line 79)
40572 * -lgcc, use with -nostdlib: Link Options. (line 79)
40573 * -nodefaultlibs and unresolved references: Link Options. (line 79)
40574 * -nostdlib and unresolved references: Link Options. (line 79)
40575 * .sdata/.sdata2 references (PowerPC): RS/6000 and PowerPC Options.
40577 * //: C++ Comments. (line 6)
40578 * 0 in constraint: Simple Constraints. (line 115)
40579 * < in constraint: Simple Constraints. (line 46)
40580 * = in constraint: Modifiers. (line 8)
40581 * > in constraint: Simple Constraints. (line 50)
40582 * ? in constraint: Multi-Alternative. (line 27)
40583 * ?: extensions: Conditionals. (line 6)
40584 * ?: side effect: Conditionals. (line 20)
40585 * _ in variables in macros: Typeof. (line 42)
40586 * __builtin___clear_cache: Other Builtins. (line 272)
40587 * __builtin___fprintf_chk: Object Size Checking.
40589 * __builtin___memcpy_chk: Object Size Checking.
40591 * __builtin___memmove_chk: Object Size Checking.
40593 * __builtin___mempcpy_chk: Object Size Checking.
40595 * __builtin___memset_chk: Object Size Checking.
40597 * __builtin___printf_chk: Object Size Checking.
40599 * __builtin___snprintf_chk: Object Size Checking.
40601 * __builtin___sprintf_chk: Object Size Checking.
40603 * __builtin___stpcpy_chk: Object Size Checking.
40605 * __builtin___strcat_chk: Object Size Checking.
40607 * __builtin___strcpy_chk: Object Size Checking.
40609 * __builtin___strncat_chk: Object Size Checking.
40611 * __builtin___strncpy_chk: Object Size Checking.
40613 * __builtin___vfprintf_chk: Object Size Checking.
40615 * __builtin___vprintf_chk: Object Size Checking.
40617 * __builtin___vsnprintf_chk: Object Size Checking.
40619 * __builtin___vsprintf_chk: Object Size Checking.
40621 * __builtin_apply: Constructing Calls. (line 31)
40622 * __builtin_apply_args: Constructing Calls. (line 20)
40623 * __builtin_bswap32: Other Builtins. (line 473)
40624 * __builtin_bswap64: Other Builtins. (line 478)
40625 * __builtin_choose_expr: Other Builtins. (line 154)
40626 * __builtin_clz: Other Builtins. (line 406)
40627 * __builtin_clzl: Other Builtins. (line 424)
40628 * __builtin_clzll: Other Builtins. (line 444)
40629 * __builtin_constant_p: Other Builtins. (line 194)
40630 * __builtin_ctz: Other Builtins. (line 410)
40631 * __builtin_ctzl: Other Builtins. (line 428)
40632 * __builtin_ctzll: Other Builtins. (line 448)
40633 * __builtin_expect: Other Builtins. (line 240)
40634 * __builtin_ffs: Other Builtins. (line 402)
40635 * __builtin_ffsl: Other Builtins. (line 420)
40636 * __builtin_ffsll: Other Builtins. (line 440)
40637 * __builtin_frame_address: Return Address. (line 34)
40638 * __builtin_huge_val: Other Builtins. (line 323)
40639 * __builtin_huge_valf: Other Builtins. (line 328)
40640 * __builtin_huge_vall: Other Builtins. (line 331)
40641 * __builtin_inf: Other Builtins. (line 335)
40642 * __builtin_infd128: Other Builtins. (line 345)
40643 * __builtin_infd32: Other Builtins. (line 339)
40644 * __builtin_infd64: Other Builtins. (line 342)
40645 * __builtin_inff: Other Builtins. (line 349)
40646 * __builtin_infl: Other Builtins. (line 354)
40647 * __builtin_isfinite: Other Builtins. (line 6)
40648 * __builtin_isgreater: Other Builtins. (line 6)
40649 * __builtin_isgreaterequal: Other Builtins. (line 6)
40650 * __builtin_isless: Other Builtins. (line 6)
40651 * __builtin_islessequal: Other Builtins. (line 6)
40652 * __builtin_islessgreater: Other Builtins. (line 6)
40653 * __builtin_isnormal: Other Builtins. (line 6)
40654 * __builtin_isunordered: Other Builtins. (line 6)
40655 * __builtin_nan: Other Builtins. (line 358)
40656 * __builtin_nand128: Other Builtins. (line 380)
40657 * __builtin_nand32: Other Builtins. (line 374)
40658 * __builtin_nand64: Other Builtins. (line 377)
40659 * __builtin_nanf: Other Builtins. (line 384)
40660 * __builtin_nanl: Other Builtins. (line 387)
40661 * __builtin_nans: Other Builtins. (line 391)
40662 * __builtin_nansf: Other Builtins. (line 395)
40663 * __builtin_nansl: Other Builtins. (line 398)
40664 * __builtin_object_size: Object Size Checking.
40666 * __builtin_offsetof: Offsetof. (line 6)
40667 * __builtin_parity: Other Builtins. (line 417)
40668 * __builtin_parityl: Other Builtins. (line 436)
40669 * __builtin_parityll: Other Builtins. (line 456)
40670 * __builtin_popcount: Other Builtins. (line 414)
40671 * __builtin_popcountl: Other Builtins. (line 432)
40672 * __builtin_popcountll: Other Builtins. (line 452)
40673 * __builtin_powi: Other Builtins. (line 6)
40674 * __builtin_powif: Other Builtins. (line 6)
40675 * __builtin_powil: Other Builtins. (line 6)
40676 * __builtin_prefetch: Other Builtins. (line 284)
40677 * __builtin_return: Constructing Calls. (line 48)
40678 * __builtin_return_address: Return Address. (line 11)
40679 * __builtin_trap: Other Builtins. (line 264)
40680 * __builtin_types_compatible_p: Other Builtins. (line 108)
40681 * __complex__ keyword: Complex. (line 6)
40682 * __declspec(dllexport): Function Attributes.
40684 * __declspec(dllimport): Function Attributes.
40686 * __extension__: Alternate Keywords. (line 29)
40687 * __float128 data type: Floating Types. (line 6)
40688 * __float80 data type: Floating Types. (line 6)
40689 * __func__ identifier: Function Names. (line 6)
40690 * __FUNCTION__ identifier: Function Names. (line 6)
40691 * __imag__ keyword: Complex. (line 27)
40692 * __PRETTY_FUNCTION__ identifier: Function Names. (line 6)
40693 * __real__ keyword: Complex. (line 27)
40694 * __STDC_HOSTED__: Standards. (line 13)
40695 * __sync_add_and_fetch: Atomic Builtins. (line 57)
40696 * __sync_and_and_fetch: Atomic Builtins. (line 57)
40697 * __sync_bool_compare_and_swap: Atomic Builtins. (line 65)
40698 * __sync_fetch_and_add: Atomic Builtins. (line 45)
40699 * __sync_fetch_and_and: Atomic Builtins. (line 45)
40700 * __sync_fetch_and_nand: Atomic Builtins. (line 45)
40701 * __sync_fetch_and_or: Atomic Builtins. (line 45)
40702 * __sync_fetch_and_sub: Atomic Builtins. (line 45)
40703 * __sync_fetch_and_xor: Atomic Builtins. (line 45)
40704 * __sync_lock_release: Atomic Builtins. (line 95)
40705 * __sync_lock_test_and_set: Atomic Builtins. (line 77)
40706 * __sync_nand_and_fetch: Atomic Builtins. (line 57)
40707 * __sync_or_and_fetch: Atomic Builtins. (line 57)
40708 * __sync_sub_and_fetch: Atomic Builtins. (line 57)
40709 * __sync_synchronize: Atomic Builtins. (line 74)
40710 * __sync_val_compare_and_swap: Atomic Builtins. (line 65)
40711 * __sync_xor_and_fetch: Atomic Builtins. (line 57)
40712 * __thread: Thread-Local. (line 6)
40713 * _Accum data type: Fixed-Point. (line 6)
40714 * _Complex keyword: Complex. (line 6)
40715 * _Decimal128 data type: Decimal Float. (line 6)
40716 * _Decimal32 data type: Decimal Float. (line 6)
40717 * _Decimal64 data type: Decimal Float. (line 6)
40718 * _exit: Other Builtins. (line 6)
40719 * _Exit: Other Builtins. (line 6)
40720 * _Fract data type: Fixed-Point. (line 6)
40721 * _Sat data type: Fixed-Point. (line 6)
40722 * ABI: Compatibility. (line 6)
40723 * abort: Other Builtins. (line 6)
40724 * abs: Other Builtins. (line 6)
40725 * accessing volatiles: Volatiles. (line 6)
40726 * acos: Other Builtins. (line 6)
40727 * acosf: Other Builtins. (line 6)
40728 * acosh: Other Builtins. (line 6)
40729 * acoshf: Other Builtins. (line 6)
40730 * acoshl: Other Builtins. (line 6)
40731 * acosl: Other Builtins. (line 6)
40732 * Ada: G++ and GCC. (line 6)
40733 * additional floating types: Floating Types. (line 6)
40734 * address constraints: Simple Constraints. (line 142)
40735 * address of a label: Labels as Values. (line 6)
40736 * address_operand: Simple Constraints. (line 146)
40737 * alias attribute: Function Attributes.
40739 * aliasing of parameters: Code Gen Options. (line 374)
40740 * aligned attribute <1>: Type Attributes. (line 31)
40741 * aligned attribute <2>: Variable Attributes.
40743 * aligned attribute: Function Attributes.
40745 * alignment: Alignment. (line 6)
40746 * alloc_size attribute: Function Attributes.
40748 * alloca: Other Builtins. (line 6)
40749 * alloca vs variable-length arrays: Variable Length. (line 27)
40750 * Allow nesting in an interrupt handler on the Blackfin processor.: Function Attributes.
40752 * alternate keywords: Alternate Keywords. (line 6)
40753 * always_inline function attribute: Function Attributes.
40755 * AMD x86-64 Options: i386 and x86-64 Options.
40757 * AMD1: Standards. (line 13)
40758 * ANSI C: Standards. (line 13)
40759 * ANSI C standard: Standards. (line 13)
40760 * ANSI C89: Standards. (line 13)
40761 * ANSI support: C Dialect Options. (line 10)
40762 * ANSI X3.159-1989: Standards. (line 13)
40763 * apostrophes: Incompatibilities. (line 116)
40764 * application binary interface: Compatibility. (line 6)
40765 * ARC Options: ARC Options. (line 6)
40766 * ARM [Annotated C++ Reference Manual]: Backwards Compatibility.
40768 * ARM options: ARM Options. (line 6)
40769 * arrays of length zero: Zero Length. (line 6)
40770 * arrays of variable length: Variable Length. (line 6)
40771 * arrays, non-lvalue: Subscripting. (line 6)
40772 * artificial function attribute: Function Attributes.
40774 * asin: Other Builtins. (line 6)
40775 * asinf: Other Builtins. (line 6)
40776 * asinh: Other Builtins. (line 6)
40777 * asinhf: Other Builtins. (line 6)
40778 * asinhl: Other Builtins. (line 6)
40779 * asinl: Other Builtins. (line 6)
40780 * asm constraints: Constraints. (line 6)
40781 * asm expressions: Extended Asm. (line 6)
40782 * assembler instructions: Extended Asm. (line 6)
40783 * assembler names for identifiers: Asm Labels. (line 6)
40784 * assembly code, invalid: Bug Criteria. (line 12)
40785 * atan: Other Builtins. (line 6)
40786 * atan2: Other Builtins. (line 6)
40787 * atan2f: Other Builtins. (line 6)
40788 * atan2l: Other Builtins. (line 6)
40789 * atanf: Other Builtins. (line 6)
40790 * atanh: Other Builtins. (line 6)
40791 * atanhf: Other Builtins. (line 6)
40792 * atanhl: Other Builtins. (line 6)
40793 * atanl: Other Builtins. (line 6)
40794 * attribute of types: Type Attributes. (line 6)
40795 * attribute of variables: Variable Attributes.
40797 * attribute syntax: Attribute Syntax. (line 6)
40798 * autoincrement/decrement addressing: Simple Constraints. (line 28)
40799 * automatic inline for C++ member fns: Inline. (line 71)
40800 * AVR Options: AVR Options. (line 6)
40801 * Backwards Compatibility: Backwards Compatibility.
40803 * base class members: Name lookup. (line 6)
40804 * bcmp: Other Builtins. (line 6)
40805 * below100 attribute: Variable Attributes.
40807 * binary compatibility: Compatibility. (line 6)
40808 * Binary constants using the 0b prefix: Binary constants. (line 6)
40809 * Blackfin Options: Blackfin Options. (line 6)
40810 * bound pointer to member function: Bound member functions.
40812 * bounds checking: Optimize Options. (line 330)
40813 * bug criteria: Bug Criteria. (line 6)
40814 * bugs: Bugs. (line 6)
40815 * bugs, known: Trouble. (line 6)
40816 * built-in functions <1>: Other Builtins. (line 6)
40817 * built-in functions: C Dialect Options. (line 170)
40818 * bzero: Other Builtins. (line 6)
40819 * C compilation options: Invoking GCC. (line 17)
40820 * C intermediate output, nonexistent: G++ and GCC. (line 35)
40821 * C language extensions: C Extensions. (line 6)
40822 * C language, traditional: C Dialect Options. (line 250)
40823 * C standard: Standards. (line 13)
40824 * C standards: Standards. (line 13)
40825 * c++: Invoking G++. (line 14)
40826 * C++: G++ and GCC. (line 30)
40827 * C++ comments: C++ Comments. (line 6)
40828 * C++ compilation options: Invoking GCC. (line 23)
40829 * C++ interface and implementation headers: C++ Interface. (line 6)
40830 * C++ language extensions: C++ Extensions. (line 6)
40831 * C++ member fns, automatically inline: Inline. (line 71)
40832 * C++ misunderstandings: C++ Misunderstandings.
40834 * C++ options, command line: C++ Dialect Options.
40836 * C++ pragmas, effect on inlining: C++ Interface. (line 66)
40837 * C++ source file suffixes: Invoking G++. (line 6)
40838 * C++ static data, declaring and defining: Static Definitions.
40840 * C89: Standards. (line 13)
40841 * C90: Standards. (line 13)
40842 * C94: Standards. (line 13)
40843 * C95: Standards. (line 13)
40844 * C99: Standards. (line 13)
40845 * C9X: Standards. (line 13)
40846 * C_INCLUDE_PATH: Environment Variables.
40848 * cabs: Other Builtins. (line 6)
40849 * cabsf: Other Builtins. (line 6)
40850 * cabsl: Other Builtins. (line 6)
40851 * cacos: Other Builtins. (line 6)
40852 * cacosf: Other Builtins. (line 6)
40853 * cacosh: Other Builtins. (line 6)
40854 * cacoshf: Other Builtins. (line 6)
40855 * cacoshl: Other Builtins. (line 6)
40856 * cacosl: Other Builtins. (line 6)
40857 * calling functions through the function vector on H8/300, M16C, and M32C processors: Function Attributes.
40859 * calloc: Other Builtins. (line 6)
40860 * carg: Other Builtins. (line 6)
40861 * cargf: Other Builtins. (line 6)
40862 * cargl: Other Builtins. (line 6)
40863 * case labels in initializers: Designated Inits. (line 6)
40864 * case ranges: Case Ranges. (line 6)
40865 * casin: Other Builtins. (line 6)
40866 * casinf: Other Builtins. (line 6)
40867 * casinh: Other Builtins. (line 6)
40868 * casinhf: Other Builtins. (line 6)
40869 * casinhl: Other Builtins. (line 6)
40870 * casinl: Other Builtins. (line 6)
40871 * cast to a union: Cast to Union. (line 6)
40872 * catan: Other Builtins. (line 6)
40873 * catanf: Other Builtins. (line 6)
40874 * catanh: Other Builtins. (line 6)
40875 * catanhf: Other Builtins. (line 6)
40876 * catanhl: Other Builtins. (line 6)
40877 * catanl: Other Builtins. (line 6)
40878 * cbrt: Other Builtins. (line 6)
40879 * cbrtf: Other Builtins. (line 6)
40880 * cbrtl: Other Builtins. (line 6)
40881 * ccos: Other Builtins. (line 6)
40882 * ccosf: Other Builtins. (line 6)
40883 * ccosh: Other Builtins. (line 6)
40884 * ccoshf: Other Builtins. (line 6)
40885 * ccoshl: Other Builtins. (line 6)
40886 * ccosl: Other Builtins. (line 6)
40887 * ceil: Other Builtins. (line 6)
40888 * ceilf: Other Builtins. (line 6)
40889 * ceill: Other Builtins. (line 6)
40890 * cexp: Other Builtins. (line 6)
40891 * cexpf: Other Builtins. (line 6)
40892 * cexpl: Other Builtins. (line 6)
40893 * character set, execution: Preprocessor Options.
40895 * character set, input: Preprocessor Options.
40897 * character set, input normalization: Warning Options. (line 1053)
40898 * character set, wide execution: Preprocessor Options.
40900 * cimag: Other Builtins. (line 6)
40901 * cimagf: Other Builtins. (line 6)
40902 * cimagl: Other Builtins. (line 6)
40903 * cleanup attribute: Variable Attributes.
40905 * clog: Other Builtins. (line 6)
40906 * clogf: Other Builtins. (line 6)
40907 * clogl: Other Builtins. (line 6)
40908 * COBOL: G++ and GCC. (line 23)
40909 * code generation conventions: Code Gen Options. (line 6)
40910 * code, mixed with declarations: Mixed Declarations. (line 6)
40911 * cold function attribute: Function Attributes.
40913 * command options: Invoking GCC. (line 6)
40914 * comments, C++ style: C++ Comments. (line 6)
40915 * common attribute: Variable Attributes.
40917 * comparison of signed and unsigned values, warning: Warning Options.
40919 * compiler bugs, reporting: Bug Reporting. (line 6)
40920 * compiler compared to C++ preprocessor: G++ and GCC. (line 35)
40921 * compiler options, C++: C++ Dialect Options.
40923 * compiler options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
40925 * compiler version, specifying: Target Options. (line 6)
40926 * COMPILER_PATH: Environment Variables.
40928 * complex conjugation: Complex. (line 34)
40929 * complex numbers: Complex. (line 6)
40930 * compound literals: Compound Literals. (line 6)
40931 * computed gotos: Labels as Values. (line 6)
40932 * conditional expressions, extensions: Conditionals. (line 6)
40933 * conflicting types: Disappointments. (line 21)
40934 * conj: Other Builtins. (line 6)
40935 * conjf: Other Builtins. (line 6)
40936 * conjl: Other Builtins. (line 6)
40937 * const applied to function: Function Attributes.
40939 * const function attribute: Function Attributes.
40941 * constants in constraints: Simple Constraints. (line 58)
40942 * constraint modifier characters: Modifiers. (line 6)
40943 * constraint, matching: Simple Constraints. (line 127)
40944 * constraints, asm: Constraints. (line 6)
40945 * constraints, machine specific: Machine Constraints.
40947 * constructing calls: Constructing Calls. (line 6)
40948 * constructor expressions: Compound Literals. (line 6)
40949 * constructor function attribute: Function Attributes.
40951 * contributors: Contributors. (line 6)
40952 * copysign: Other Builtins. (line 6)
40953 * copysignf: Other Builtins. (line 6)
40954 * copysignl: Other Builtins. (line 6)
40955 * core dump: Bug Criteria. (line 9)
40956 * cos: Other Builtins. (line 6)
40957 * cosf: Other Builtins. (line 6)
40958 * cosh: Other Builtins. (line 6)
40959 * coshf: Other Builtins. (line 6)
40960 * coshl: Other Builtins. (line 6)
40961 * cosl: Other Builtins. (line 6)
40962 * CPATH: Environment Variables.
40964 * CPLUS_INCLUDE_PATH: Environment Variables.
40966 * cpow: Other Builtins. (line 6)
40967 * cpowf: Other Builtins. (line 6)
40968 * cpowl: Other Builtins. (line 6)
40969 * cproj: Other Builtins. (line 6)
40970 * cprojf: Other Builtins. (line 6)
40971 * cprojl: Other Builtins. (line 6)
40972 * creal: Other Builtins. (line 6)
40973 * crealf: Other Builtins. (line 6)
40974 * creall: Other Builtins. (line 6)
40975 * CRIS Options: CRIS Options. (line 6)
40976 * cross compiling: Target Options. (line 6)
40977 * CRX Options: CRX Options. (line 6)
40978 * csin: Other Builtins. (line 6)
40979 * csinf: Other Builtins. (line 6)
40980 * csinh: Other Builtins. (line 6)
40981 * csinhf: Other Builtins. (line 6)
40982 * csinhl: Other Builtins. (line 6)
40983 * csinl: Other Builtins. (line 6)
40984 * csqrt: Other Builtins. (line 6)
40985 * csqrtf: Other Builtins. (line 6)
40986 * csqrtl: Other Builtins. (line 6)
40987 * ctan: Other Builtins. (line 6)
40988 * ctanf: Other Builtins. (line 6)
40989 * ctanh: Other Builtins. (line 6)
40990 * ctanhf: Other Builtins. (line 6)
40991 * ctanhl: Other Builtins. (line 6)
40992 * ctanl: Other Builtins. (line 6)
40993 * Darwin options: Darwin Options. (line 6)
40994 * dcgettext: Other Builtins. (line 6)
40995 * DD integer suffix: Decimal Float. (line 6)
40996 * dd integer suffix: Decimal Float. (line 6)
40997 * deallocating variable length arrays: Variable Length. (line 23)
40998 * debugging information options: Debugging Options. (line 6)
40999 * decimal floating types: Decimal Float. (line 6)
41000 * declaration scope: Incompatibilities. (line 80)
41001 * declarations inside expressions: Statement Exprs. (line 6)
41002 * declarations, mixed with code: Mixed Declarations. (line 6)
41003 * declaring attributes of functions: Function Attributes.
41005 * declaring static data in C++: Static Definitions. (line 6)
41006 * defining static data in C++: Static Definitions. (line 6)
41007 * dependencies for make as output: Environment Variables.
41009 * dependencies, make: Preprocessor Options.
41011 * DEPENDENCIES_OUTPUT: Environment Variables.
41013 * dependent name lookup: Name lookup. (line 6)
41014 * deprecated attribute: Variable Attributes.
41016 * deprecated attribute.: Function Attributes.
41018 * designated initializers: Designated Inits. (line 6)
41019 * designator lists: Designated Inits. (line 94)
41020 * designators: Designated Inits. (line 61)
41021 * destructor function attribute: Function Attributes.
41023 * DF integer suffix: Decimal Float. (line 6)
41024 * df integer suffix: Decimal Float. (line 6)
41025 * dgettext: Other Builtins. (line 6)
41026 * diagnostic messages: Language Independent Options.
41028 * dialect options: C Dialect Options. (line 6)
41029 * digits in constraint: Simple Constraints. (line 115)
41030 * directory options: Directory Options. (line 6)
41031 * DL integer suffix: Decimal Float. (line 6)
41032 * dl integer suffix: Decimal Float. (line 6)
41033 * dollar signs in identifier names: Dollar Signs. (line 6)
41034 * double-word arithmetic: Long Long. (line 6)
41035 * downward funargs: Nested Functions. (line 6)
41036 * drem: Other Builtins. (line 6)
41037 * dremf: Other Builtins. (line 6)
41038 * dreml: Other Builtins. (line 6)
41039 * E in constraint: Simple Constraints. (line 77)
41040 * earlyclobber operand: Modifiers. (line 25)
41041 * eight bit data on the H8/300, H8/300H, and H8S: Function Attributes.
41043 * empty structures: Empty Structures. (line 6)
41044 * environment variables: Environment Variables.
41046 * erf: Other Builtins. (line 6)
41047 * erfc: Other Builtins. (line 6)
41048 * erfcf: Other Builtins. (line 6)
41049 * erfcl: Other Builtins. (line 6)
41050 * erff: Other Builtins. (line 6)
41051 * erfl: Other Builtins. (line 6)
41052 * error function attribute: Function Attributes.
41054 * error messages: Warnings and Errors.
41056 * escaped newlines: Escaped Newlines. (line 6)
41057 * exception handler functions on the Blackfin processor: Function Attributes.
41059 * exclamation point: Multi-Alternative. (line 33)
41060 * exit: Other Builtins. (line 6)
41061 * exp: Other Builtins. (line 6)
41062 * exp10: Other Builtins. (line 6)
41063 * exp10f: Other Builtins. (line 6)
41064 * exp10l: Other Builtins. (line 6)
41065 * exp2: Other Builtins. (line 6)
41066 * exp2f: Other Builtins. (line 6)
41067 * exp2l: Other Builtins. (line 6)
41068 * expf: Other Builtins. (line 6)
41069 * expl: Other Builtins. (line 6)
41070 * explicit register variables: Explicit Reg Vars. (line 6)
41071 * expm1: Other Builtins. (line 6)
41072 * expm1f: Other Builtins. (line 6)
41073 * expm1l: Other Builtins. (line 6)
41074 * expressions containing statements: Statement Exprs. (line 6)
41075 * expressions, constructor: Compound Literals. (line 6)
41076 * extended asm: Extended Asm. (line 6)
41077 * extensible constraints: Simple Constraints. (line 151)
41078 * extensions, ?:: Conditionals. (line 6)
41079 * extensions, C language: C Extensions. (line 6)
41080 * extensions, C++ language: C++ Extensions. (line 6)
41081 * external declaration scope: Incompatibilities. (line 80)
41082 * externally_visible attribute.: Function Attributes.
41084 * F in constraint: Simple Constraints. (line 82)
41085 * fabs: Other Builtins. (line 6)
41086 * fabsf: Other Builtins. (line 6)
41087 * fabsl: Other Builtins. (line 6)
41088 * fatal signal: Bug Criteria. (line 9)
41089 * fdim: Other Builtins. (line 6)
41090 * fdimf: Other Builtins. (line 6)
41091 * fdiml: Other Builtins. (line 6)
41092 * FDL, GNU Free Documentation License: GNU Free Documentation License.
41094 * ffs: Other Builtins. (line 6)
41095 * file name suffix: Overall Options. (line 14)
41096 * file names: Link Options. (line 10)
41097 * fixed-point types: Fixed-Point. (line 6)
41098 * flatten function attribute: Function Attributes.
41100 * flexible array members: Zero Length. (line 6)
41101 * float as function value type: Incompatibilities. (line 141)
41102 * floating point precision <1>: Disappointments. (line 68)
41103 * floating point precision: Optimize Options. (line 1163)
41104 * floor: Other Builtins. (line 6)
41105 * floorf: Other Builtins. (line 6)
41106 * floorl: Other Builtins. (line 6)
41107 * fma: Other Builtins. (line 6)
41108 * fmaf: Other Builtins. (line 6)
41109 * fmal: Other Builtins. (line 6)
41110 * fmax: Other Builtins. (line 6)
41111 * fmaxf: Other Builtins. (line 6)
41112 * fmaxl: Other Builtins. (line 6)
41113 * fmin: Other Builtins. (line 6)
41114 * fminf: Other Builtins. (line 6)
41115 * fminl: Other Builtins. (line 6)
41116 * fmod: Other Builtins. (line 6)
41117 * fmodf: Other Builtins. (line 6)
41118 * fmodl: Other Builtins. (line 6)
41119 * force_align_arg_pointer attribute: Function Attributes.
41121 * format function attribute: Function Attributes.
41123 * format_arg function attribute: Function Attributes.
41125 * Fortran: G++ and GCC. (line 6)
41126 * forwarding calls: Constructing Calls. (line 6)
41127 * fprintf: Other Builtins. (line 6)
41128 * fprintf_unlocked: Other Builtins. (line 6)
41129 * fputs: Other Builtins. (line 6)
41130 * fputs_unlocked: Other Builtins. (line 6)
41131 * freestanding environment: Standards. (line 13)
41132 * freestanding implementation: Standards. (line 13)
41133 * frexp: Other Builtins. (line 6)
41134 * frexpf: Other Builtins. (line 6)
41135 * frexpl: Other Builtins. (line 6)
41136 * FRV Options: FRV Options. (line 6)
41137 * fscanf: Other Builtins. (line 6)
41138 * fscanf, and constant strings: Incompatibilities. (line 17)
41139 * function addressability on the M32R/D: Function Attributes.
41141 * function attributes: Function Attributes.
41143 * function pointers, arithmetic: Pointer Arith. (line 6)
41144 * function prototype declarations: Function Prototypes.
41146 * function without a prologue/epilogue code: Function Attributes.
41148 * function, size of pointer to: Pointer Arith. (line 6)
41149 * functions called via pointer on the RS/6000 and PowerPC: Function Attributes.
41151 * functions in arbitrary sections: Function Attributes.
41153 * functions that are passed arguments in registers on the 386: Function Attributes.
41155 * functions that behave like malloc: Function Attributes.
41157 * functions that do not pop the argument stack on the 386: Function Attributes.
41159 * functions that do pop the argument stack on the 386: Function Attributes.
41161 * functions that have no side effects: Function Attributes.
41163 * functions that never return: Function Attributes.
41165 * functions that pop the argument stack on the 386: Function Attributes.
41167 * functions that return more than once: Function Attributes.
41169 * functions which do not handle memory bank switching on 68HC11/68HC12: Function Attributes.
41171 * functions which handle memory bank switching: Function Attributes.
41173 * functions with non-null pointer arguments: Function Attributes.
41175 * functions with printf, scanf, strftime or strfmon style arguments: Function Attributes.
41177 * g in constraint: Simple Constraints. (line 108)
41178 * G in constraint: Simple Constraints. (line 86)
41179 * g++: Invoking G++. (line 14)
41180 * G++: G++ and GCC. (line 30)
41181 * gamma: Other Builtins. (line 6)
41182 * gamma_r: Other Builtins. (line 6)
41183 * gammaf: Other Builtins. (line 6)
41184 * gammaf_r: Other Builtins. (line 6)
41185 * gammal: Other Builtins. (line 6)
41186 * gammal_r: Other Builtins. (line 6)
41187 * GCC: G++ and GCC. (line 6)
41188 * GCC command options: Invoking GCC. (line 6)
41189 * GCC_EXEC_PREFIX: Environment Variables.
41191 * gcc_struct: Type Attributes. (line 303)
41192 * gcc_struct attribute: Variable Attributes.
41194 * gcov: Debugging Options. (line 258)
41195 * gettext: Other Builtins. (line 6)
41196 * global offset table: Code Gen Options. (line 173)
41197 * global register after longjmp: Global Reg Vars. (line 66)
41198 * global register variables: Global Reg Vars. (line 6)
41199 * GNAT: G++ and GCC. (line 30)
41200 * GNU C Compiler: G++ and GCC. (line 6)
41201 * GNU Compiler Collection: G++ and GCC. (line 6)
41202 * gnu_inline function attribute: Function Attributes.
41204 * goto with computed label: Labels as Values. (line 6)
41205 * gprof: Debugging Options. (line 219)
41206 * grouping options: Invoking GCC. (line 26)
41207 * H in constraint: Simple Constraints. (line 86)
41208 * hardware models and configurations, specifying: Submodel Options.
41210 * hex floats: Hex Floats. (line 6)
41211 * HK fixed-suffix: Fixed-Point. (line 6)
41212 * hk fixed-suffix: Fixed-Point. (line 6)
41213 * hosted environment <1>: C Dialect Options. (line 204)
41214 * hosted environment: Standards. (line 13)
41215 * hosted implementation: Standards. (line 13)
41216 * hot function attribute: Function Attributes.
41218 * HPPA Options: HPPA Options. (line 6)
41219 * HR fixed-suffix: Fixed-Point. (line 6)
41220 * hr fixed-suffix: Fixed-Point. (line 6)
41221 * hypot: Other Builtins. (line 6)
41222 * hypotf: Other Builtins. (line 6)
41223 * hypotl: Other Builtins. (line 6)
41224 * I in constraint: Simple Constraints. (line 69)
41225 * i in constraint: Simple Constraints. (line 58)
41226 * i386 Options: i386 and x86-64 Options.
41228 * IA-64 Options: IA-64 Options. (line 6)
41229 * IBM RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
41231 * identifier names, dollar signs in: Dollar Signs. (line 6)
41232 * identifiers, names in assembler code: Asm Labels. (line 6)
41233 * ilogb: Other Builtins. (line 6)
41234 * ilogbf: Other Builtins. (line 6)
41235 * ilogbl: Other Builtins. (line 6)
41236 * imaxabs: Other Builtins. (line 6)
41237 * implementation-defined behavior, C language: C Implementation.
41239 * implied #pragma implementation: C++ Interface. (line 46)
41240 * incompatibilities of GCC: Incompatibilities. (line 6)
41241 * increment operators: Bug Criteria. (line 17)
41242 * index: Other Builtins. (line 6)
41243 * indirect calls on ARM: Function Attributes.
41245 * indirect calls on MIPS: Function Attributes.
41247 * init_priority attribute: C++ Attributes. (line 9)
41248 * initializations in expressions: Compound Literals. (line 6)
41249 * initializers with labeled elements: Designated Inits. (line 6)
41250 * initializers, non-constant: Initializers. (line 6)
41251 * inline automatic for C++ member fns: Inline. (line 71)
41252 * inline functions: Inline. (line 6)
41253 * inline functions, omission of: Inline. (line 51)
41254 * inlining and C++ pragmas: C++ Interface. (line 66)
41255 * installation trouble: Trouble. (line 6)
41256 * integrating function code: Inline. (line 6)
41257 * Intel 386 Options: i386 and x86-64 Options.
41259 * interface and implementation headers, C++: C++ Interface. (line 6)
41260 * intermediate C version, nonexistent: G++ and GCC. (line 35)
41261 * interrupt handler functions: Function Attributes.
41263 * interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors: Function Attributes.
41265 * interrupt thread functions on fido: Function Attributes.
41267 * introduction: Top. (line 6)
41268 * invalid assembly code: Bug Criteria. (line 12)
41269 * invalid input: Bug Criteria. (line 42)
41270 * invoking g++: Invoking G++. (line 22)
41271 * isalnum: Other Builtins. (line 6)
41272 * isalpha: Other Builtins. (line 6)
41273 * isascii: Other Builtins. (line 6)
41274 * isblank: Other Builtins. (line 6)
41275 * iscntrl: Other Builtins. (line 6)
41276 * isdigit: Other Builtins. (line 6)
41277 * isgraph: Other Builtins. (line 6)
41278 * islower: Other Builtins. (line 6)
41279 * ISO 9899: Standards. (line 13)
41280 * ISO C: Standards. (line 13)
41281 * ISO C standard: Standards. (line 13)
41282 * ISO C90: Standards. (line 13)
41283 * ISO C94: Standards. (line 13)
41284 * ISO C95: Standards. (line 13)
41285 * ISO C99: Standards. (line 13)
41286 * ISO C9X: Standards. (line 13)
41287 * ISO support: C Dialect Options. (line 10)
41288 * ISO/IEC 9899: Standards. (line 13)
41289 * isprint: Other Builtins. (line 6)
41290 * ispunct: Other Builtins. (line 6)
41291 * isspace: Other Builtins. (line 6)
41292 * isupper: Other Builtins. (line 6)
41293 * iswalnum: Other Builtins. (line 6)
41294 * iswalpha: Other Builtins. (line 6)
41295 * iswblank: Other Builtins. (line 6)
41296 * iswcntrl: Other Builtins. (line 6)
41297 * iswdigit: Other Builtins. (line 6)
41298 * iswgraph: Other Builtins. (line 6)
41299 * iswlower: Other Builtins. (line 6)
41300 * iswprint: Other Builtins. (line 6)
41301 * iswpunct: Other Builtins. (line 6)
41302 * iswspace: Other Builtins. (line 6)
41303 * iswupper: Other Builtins. (line 6)
41304 * iswxdigit: Other Builtins. (line 6)
41305 * isxdigit: Other Builtins. (line 6)
41306 * j0: Other Builtins. (line 6)
41307 * j0f: Other Builtins. (line 6)
41308 * j0l: Other Builtins. (line 6)
41309 * j1: Other Builtins. (line 6)
41310 * j1f: Other Builtins. (line 6)
41311 * j1l: Other Builtins. (line 6)
41312 * Java: G++ and GCC. (line 6)
41313 * java_interface attribute: C++ Attributes. (line 29)
41314 * jn: Other Builtins. (line 6)
41315 * jnf: Other Builtins. (line 6)
41316 * jnl: Other Builtins. (line 6)
41317 * K fixed-suffix: Fixed-Point. (line 6)
41318 * k fixed-suffix: Fixed-Point. (line 6)
41319 * keywords, alternate: Alternate Keywords. (line 6)
41320 * known causes of trouble: Trouble. (line 6)
41321 * l1_data variable attribute: Variable Attributes.
41323 * l1_data_A variable attribute: Variable Attributes.
41325 * l1_data_B variable attribute: Variable Attributes.
41327 * l1_text function attribute: Function Attributes.
41329 * labeled elements in initializers: Designated Inits. (line 6)
41330 * labels as values: Labels as Values. (line 6)
41331 * labs: Other Builtins. (line 6)
41332 * LANG: Environment Variables.
41334 * language dialect options: C Dialect Options. (line 6)
41335 * LC_ALL: Environment Variables.
41337 * LC_CTYPE: Environment Variables.
41339 * LC_MESSAGES: Environment Variables.
41341 * ldexp: Other Builtins. (line 6)
41342 * ldexpf: Other Builtins. (line 6)
41343 * ldexpl: Other Builtins. (line 6)
41344 * length-zero arrays: Zero Length. (line 6)
41345 * lgamma: Other Builtins. (line 6)
41346 * lgamma_r: Other Builtins. (line 6)
41347 * lgammaf: Other Builtins. (line 6)
41348 * lgammaf_r: Other Builtins. (line 6)
41349 * lgammal: Other Builtins. (line 6)
41350 * lgammal_r: Other Builtins. (line 6)
41351 * Libraries: Link Options. (line 24)
41352 * LIBRARY_PATH: Environment Variables.
41354 * link options: Link Options. (line 6)
41355 * LK fixed-suffix: Fixed-Point. (line 6)
41356 * lk fixed-suffix: Fixed-Point. (line 6)
41357 * LL integer suffix: Long Long. (line 6)
41358 * llabs: Other Builtins. (line 6)
41359 * LLK fixed-suffix: Fixed-Point. (line 6)
41360 * llk fixed-suffix: Fixed-Point. (line 6)
41361 * LLR fixed-suffix: Fixed-Point. (line 6)
41362 * llr fixed-suffix: Fixed-Point. (line 6)
41363 * llrint: Other Builtins. (line 6)
41364 * llrintf: Other Builtins. (line 6)
41365 * llrintl: Other Builtins. (line 6)
41366 * llround: Other Builtins. (line 6)
41367 * llroundf: Other Builtins. (line 6)
41368 * llroundl: Other Builtins. (line 6)
41369 * load address instruction: Simple Constraints. (line 142)
41370 * local labels: Local Labels. (line 6)
41371 * local variables in macros: Typeof. (line 42)
41372 * local variables, specifying registers: Local Reg Vars. (line 6)
41373 * locale: Environment Variables.
41375 * locale definition: Environment Variables.
41377 * log: Other Builtins. (line 6)
41378 * log10: Other Builtins. (line 6)
41379 * log10f: Other Builtins. (line 6)
41380 * log10l: Other Builtins. (line 6)
41381 * log1p: Other Builtins. (line 6)
41382 * log1pf: Other Builtins. (line 6)
41383 * log1pl: Other Builtins. (line 6)
41384 * log2: Other Builtins. (line 6)
41385 * log2f: Other Builtins. (line 6)
41386 * log2l: Other Builtins. (line 6)
41387 * logb: Other Builtins. (line 6)
41388 * logbf: Other Builtins. (line 6)
41389 * logbl: Other Builtins. (line 6)
41390 * logf: Other Builtins. (line 6)
41391 * logl: Other Builtins. (line 6)
41392 * long long data types: Long Long. (line 6)
41393 * longjmp: Global Reg Vars. (line 66)
41394 * longjmp incompatibilities: Incompatibilities. (line 39)
41395 * longjmp warnings: Warning Options. (line 570)
41396 * LR fixed-suffix: Fixed-Point. (line 6)
41397 * lr fixed-suffix: Fixed-Point. (line 6)
41398 * lrint: Other Builtins. (line 6)
41399 * lrintf: Other Builtins. (line 6)
41400 * lrintl: Other Builtins. (line 6)
41401 * lround: Other Builtins. (line 6)
41402 * lroundf: Other Builtins. (line 6)
41403 * lroundl: Other Builtins. (line 6)
41404 * m in constraint: Simple Constraints. (line 17)
41405 * M32C options: M32C Options. (line 6)
41406 * M32R/D options: M32R/D Options. (line 6)
41407 * M680x0 options: M680x0 Options. (line 6)
41408 * M68hc1x options: M68hc1x Options. (line 6)
41409 * machine dependent options: Submodel Options. (line 6)
41410 * machine specific constraints: Machine Constraints.
41412 * macro with variable arguments: Variadic Macros. (line 6)
41413 * macros containing asm: Extended Asm. (line 239)
41414 * macros, inline alternative: Inline. (line 6)
41415 * macros, local labels: Local Labels. (line 6)
41416 * macros, local variables in: Typeof. (line 42)
41417 * macros, statements in expressions: Statement Exprs. (line 6)
41418 * macros, types of arguments: Typeof. (line 6)
41419 * make: Preprocessor Options.
41421 * malloc: Other Builtins. (line 6)
41422 * malloc attribute: Function Attributes.
41424 * matching constraint: Simple Constraints. (line 127)
41425 * MCore options: MCore Options. (line 6)
41426 * member fns, automatically inline: Inline. (line 71)
41427 * memchr: Other Builtins. (line 6)
41428 * memcmp: Other Builtins. (line 6)
41429 * memcpy: Other Builtins. (line 6)
41430 * memory references in constraints: Simple Constraints. (line 17)
41431 * mempcpy: Other Builtins. (line 6)
41432 * memset: Other Builtins. (line 6)
41433 * Mercury: G++ and GCC. (line 23)
41434 * message formatting: Language Independent Options.
41436 * messages, warning: Warning Options. (line 6)
41437 * messages, warning and error: Warnings and Errors.
41439 * middle-operands, omitted: Conditionals. (line 6)
41440 * MIPS options: MIPS Options. (line 6)
41441 * mips16 attribute: Function Attributes.
41443 * misunderstandings in C++: C++ Misunderstandings.
41445 * mixed declarations and code: Mixed Declarations. (line 6)
41446 * mktemp, and constant strings: Incompatibilities. (line 13)
41447 * MMIX Options: MMIX Options. (line 6)
41448 * MN10300 options: MN10300 Options. (line 6)
41449 * mode attribute: Variable Attributes.
41451 * modf: Other Builtins. (line 6)
41452 * modff: Other Builtins. (line 6)
41453 * modfl: Other Builtins. (line 6)
41454 * modifiers in constraints: Modifiers. (line 6)
41455 * ms_struct: Type Attributes. (line 303)
41456 * ms_struct attribute: Variable Attributes.
41458 * MT options: MT Options. (line 6)
41459 * mudflap: Optimize Options. (line 330)
41460 * multiple alternative constraints: Multi-Alternative. (line 6)
41461 * multiprecision arithmetic: Long Long. (line 6)
41462 * n in constraint: Simple Constraints. (line 63)
41463 * names used in assembler code: Asm Labels. (line 6)
41464 * naming convention, implementation headers: C++ Interface. (line 46)
41465 * nearbyint: Other Builtins. (line 6)
41466 * nearbyintf: Other Builtins. (line 6)
41467 * nearbyintl: Other Builtins. (line 6)
41468 * nested functions: Nested Functions. (line 6)
41469 * newlines (escaped): Escaped Newlines. (line 6)
41470 * nextafter: Other Builtins. (line 6)
41471 * nextafterf: Other Builtins. (line 6)
41472 * nextafterl: Other Builtins. (line 6)
41473 * nexttoward: Other Builtins. (line 6)
41474 * nexttowardf: Other Builtins. (line 6)
41475 * nexttowardl: Other Builtins. (line 6)
41476 * NFC: Warning Options. (line 1053)
41477 * NFKC: Warning Options. (line 1053)
41478 * NMI handler functions on the Blackfin processor: Function Attributes.
41480 * no_instrument_function function attribute: Function Attributes.
41482 * nocommon attribute: Variable Attributes.
41484 * noinline function attribute: Function Attributes.
41486 * nomips16 attribute: Function Attributes.
41488 * non-constant initializers: Initializers. (line 6)
41489 * non-static inline function: Inline. (line 85)
41490 * nonnull function attribute: Function Attributes.
41492 * noreturn function attribute: Function Attributes.
41494 * nothrow function attribute: Function Attributes.
41496 * o in constraint: Simple Constraints. (line 21)
41497 * OBJC_INCLUDE_PATH: Environment Variables.
41499 * Objective-C <1>: Standards. (line 153)
41500 * Objective-C: G++ and GCC. (line 6)
41501 * Objective-C and Objective-C++ options, command line: Objective-C and Objective-C++ Dialect Options.
41503 * Objective-C++ <1>: Standards. (line 153)
41504 * Objective-C++: G++ and GCC. (line 6)
41505 * offsettable address: Simple Constraints. (line 21)
41506 * old-style function definitions: Function Prototypes.
41508 * omitted middle-operands: Conditionals. (line 6)
41509 * open coding: Inline. (line 6)
41510 * openmp parallel: C Dialect Options. (line 221)
41511 * operand constraints, asm: Constraints. (line 6)
41512 * optimize options: Optimize Options. (line 6)
41513 * options to control diagnostics formatting: Language Independent Options.
41515 * options to control warnings: Warning Options. (line 6)
41516 * options, C++: C++ Dialect Options.
41518 * options, code generation: Code Gen Options. (line 6)
41519 * options, debugging: Debugging Options. (line 6)
41520 * options, dialect: C Dialect Options. (line 6)
41521 * options, directory search: Directory Options. (line 6)
41522 * options, GCC command: Invoking GCC. (line 6)
41523 * options, grouping: Invoking GCC. (line 26)
41524 * options, linking: Link Options. (line 6)
41525 * options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
41527 * options, optimization: Optimize Options. (line 6)
41528 * options, order: Invoking GCC. (line 30)
41529 * options, preprocessor: Preprocessor Options.
41531 * order of evaluation, side effects: Non-bugs. (line 196)
41532 * order of options: Invoking GCC. (line 30)
41533 * other register constraints: Simple Constraints. (line 151)
41534 * output file option: Overall Options. (line 181)
41535 * overloaded virtual fn, warning: C++ Dialect Options.
41537 * p in constraint: Simple Constraints. (line 142)
41538 * packed attribute: Variable Attributes.
41540 * parameter forward declaration: Variable Length. (line 60)
41541 * parameters, aliased: Code Gen Options. (line 374)
41542 * Pascal: G++ and GCC. (line 23)
41543 * PDP-11 Options: PDP-11 Options. (line 6)
41544 * PIC: Code Gen Options. (line 173)
41545 * pmf: Bound member functions.
41547 * pointer arguments: Function Attributes.
41549 * pointer to member function: Bound member functions.
41551 * portions of temporary objects, pointers to: Temporaries. (line 6)
41552 * pow: Other Builtins. (line 6)
41553 * pow10: Other Builtins. (line 6)
41554 * pow10f: Other Builtins. (line 6)
41555 * pow10l: Other Builtins. (line 6)
41556 * PowerPC options: PowerPC Options. (line 6)
41557 * powf: Other Builtins. (line 6)
41558 * powl: Other Builtins. (line 6)
41559 * pragma, align: Solaris Pragmas. (line 11)
41560 * pragma, diagnostic: Diagnostic Pragmas. (line 14)
41561 * pragma, extern_prefix: Symbol-Renaming Pragmas.
41563 * pragma, fini: Solaris Pragmas. (line 19)
41564 * pragma, init: Solaris Pragmas. (line 24)
41565 * pragma, long_calls: ARM Pragmas. (line 11)
41566 * pragma, long_calls_off: ARM Pragmas. (line 17)
41567 * pragma, longcall: RS/6000 and PowerPC Pragmas.
41569 * pragma, mark: Darwin Pragmas. (line 11)
41570 * pragma, memregs: M32C Pragmas. (line 7)
41571 * pragma, no_long_calls: ARM Pragmas. (line 14)
41572 * pragma, options align: Darwin Pragmas. (line 14)
41573 * pragma, reason for not using: Function Attributes.
41575 * pragma, redefine_extname: Symbol-Renaming Pragmas.
41577 * pragma, segment: Darwin Pragmas. (line 21)
41578 * pragma, unused: Darwin Pragmas. (line 24)
41579 * pragma, visibility: Visibility Pragmas. (line 8)
41580 * pragma, weak: Weak Pragmas. (line 10)
41581 * pragmas: Pragmas. (line 6)
41582 * pragmas in C++, effect on inlining: C++ Interface. (line 66)
41583 * pragmas, interface and implementation: C++ Interface. (line 6)
41584 * pragmas, warning of unknown: Warning Options. (line 588)
41585 * precompiled headers: Precompiled Headers.
41587 * preprocessing numbers: Incompatibilities. (line 173)
41588 * preprocessing tokens: Incompatibilities. (line 173)
41589 * preprocessor options: Preprocessor Options.
41591 * printf: Other Builtins. (line 6)
41592 * printf_unlocked: Other Builtins. (line 6)
41593 * prof: Debugging Options. (line 213)
41594 * progmem variable attribute: Variable Attributes.
41596 * promotion of formal parameters: Function Prototypes.
41598 * pure function attribute: Function Attributes.
41600 * push address instruction: Simple Constraints. (line 142)
41601 * putchar: Other Builtins. (line 6)
41602 * puts: Other Builtins. (line 6)
41603 * Q floating point suffix: Floating Types. (line 6)
41604 * q floating point suffix: Floating Types. (line 6)
41605 * qsort, and global register variables: Global Reg Vars. (line 42)
41606 * question mark: Multi-Alternative. (line 27)
41607 * R fixed-suffix: Fixed-Point. (line 6)
41608 * r fixed-suffix: Fixed-Point. (line 6)
41609 * r in constraint: Simple Constraints. (line 54)
41610 * ranges in case statements: Case Ranges. (line 6)
41611 * read-only strings: Incompatibilities. (line 9)
41612 * register variable after longjmp: Global Reg Vars. (line 66)
41613 * registers: Extended Asm. (line 6)
41614 * registers for local variables: Local Reg Vars. (line 6)
41615 * registers in constraints: Simple Constraints. (line 54)
41616 * registers, global allocation: Explicit Reg Vars. (line 6)
41617 * registers, global variables in: Global Reg Vars. (line 6)
41618 * regparm attribute: Function Attributes.
41620 * relocation truncated to fit (MIPS): MIPS Options. (line 179)
41621 * remainder: Other Builtins. (line 6)
41622 * remainderf: Other Builtins. (line 6)
41623 * remainderl: Other Builtins. (line 6)
41624 * remquo: Other Builtins. (line 6)
41625 * remquof: Other Builtins. (line 6)
41626 * remquol: Other Builtins. (line 6)
41627 * reordering, warning: C++ Dialect Options.
41629 * reporting bugs: Bugs. (line 6)
41630 * rest argument (in macro): Variadic Macros. (line 6)
41631 * restricted pointers: Restricted Pointers.
41633 * restricted references: Restricted Pointers.
41635 * restricted this pointer: Restricted Pointers.
41637 * returns_twice attribute: Function Attributes.
41639 * rindex: Other Builtins. (line 6)
41640 * rint: Other Builtins. (line 6)
41641 * rintf: Other Builtins. (line 6)
41642 * rintl: Other Builtins. (line 6)
41643 * round: Other Builtins. (line 6)
41644 * roundf: Other Builtins. (line 6)
41645 * roundl: Other Builtins. (line 6)
41646 * RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
41648 * RTTI: Vague Linkage. (line 43)
41649 * run-time options: Code Gen Options. (line 6)
41650 * s in constraint: Simple Constraints. (line 90)
41651 * S/390 and zSeries Options: S/390 and zSeries Options.
41653 * save all registers on the Blackfin, H8/300, H8/300H, and H8S: Function Attributes.
41655 * scalb: Other Builtins. (line 6)
41656 * scalbf: Other Builtins. (line 6)
41657 * scalbl: Other Builtins. (line 6)
41658 * scalbln: Other Builtins. (line 6)
41659 * scalblnf: Other Builtins. (line 6)
41660 * scalbn: Other Builtins. (line 6)
41661 * scalbnf: Other Builtins. (line 6)
41662 * scanf, and constant strings: Incompatibilities. (line 17)
41663 * scanfnl: Other Builtins. (line 6)
41664 * scope of a variable length array: Variable Length. (line 23)
41665 * scope of declaration: Disappointments. (line 21)
41666 * scope of external declarations: Incompatibilities. (line 80)
41667 * Score Options: Score Options. (line 6)
41668 * search path: Directory Options. (line 6)
41669 * section function attribute: Function Attributes.
41671 * section variable attribute: Variable Attributes.
41673 * sentinel function attribute: Function Attributes.
41675 * setjmp: Global Reg Vars. (line 66)
41676 * setjmp incompatibilities: Incompatibilities. (line 39)
41677 * shared strings: Incompatibilities. (line 9)
41678 * shared variable attribute: Variable Attributes.
41680 * side effect in ?:: Conditionals. (line 20)
41681 * side effects, macro argument: Statement Exprs. (line 35)
41682 * side effects, order of evaluation: Non-bugs. (line 196)
41683 * signal handler functions on the AVR processors: Function Attributes.
41685 * signbit: Other Builtins. (line 6)
41686 * signbitd128: Other Builtins. (line 6)
41687 * signbitd32: Other Builtins. (line 6)
41688 * signbitd64: Other Builtins. (line 6)
41689 * signbitf: Other Builtins. (line 6)
41690 * signbitl: Other Builtins. (line 6)
41691 * signed and unsigned values, comparison warning: Warning Options.
41693 * significand: Other Builtins. (line 6)
41694 * significandf: Other Builtins. (line 6)
41695 * significandl: Other Builtins. (line 6)
41696 * simple constraints: Simple Constraints. (line 6)
41697 * sin: Other Builtins. (line 6)
41698 * sincos: Other Builtins. (line 6)
41699 * sincosf: Other Builtins. (line 6)
41700 * sincosl: Other Builtins. (line 6)
41701 * sinf: Other Builtins. (line 6)
41702 * sinh: Other Builtins. (line 6)
41703 * sinhf: Other Builtins. (line 6)
41704 * sinhl: Other Builtins. (line 6)
41705 * sinl: Other Builtins. (line 6)
41706 * sizeof: Typeof. (line 6)
41707 * smaller data references: M32R/D Options. (line 57)
41708 * smaller data references (PowerPC): RS/6000 and PowerPC Options.
41710 * snprintf: Other Builtins. (line 6)
41711 * SPARC options: SPARC Options. (line 6)
41712 * Spec Files: Spec Files. (line 6)
41713 * specified registers: Explicit Reg Vars. (line 6)
41714 * specifying compiler version and target machine: Target Options.
41716 * specifying hardware config: Submodel Options. (line 6)
41717 * specifying machine version: Target Options. (line 6)
41718 * specifying registers for local variables: Local Reg Vars. (line 6)
41719 * speed of compilation: Precompiled Headers.
41721 * sprintf: Other Builtins. (line 6)
41722 * SPU options: SPU Options. (line 6)
41723 * sqrt: Other Builtins. (line 6)
41724 * sqrtf: Other Builtins. (line 6)
41725 * sqrtl: Other Builtins. (line 6)
41726 * sscanf: Other Builtins. (line 6)
41727 * sscanf, and constant strings: Incompatibilities. (line 17)
41728 * sseregparm attribute: Function Attributes.
41730 * statements inside expressions: Statement Exprs. (line 6)
41731 * static data in C++, declaring and defining: Static Definitions.
41733 * stpcpy: Other Builtins. (line 6)
41734 * stpncpy: Other Builtins. (line 6)
41735 * strcasecmp: Other Builtins. (line 6)
41736 * strcat: Other Builtins. (line 6)
41737 * strchr: Other Builtins. (line 6)
41738 * strcmp: Other Builtins. (line 6)
41739 * strcpy: Other Builtins. (line 6)
41740 * strcspn: Other Builtins. (line 6)
41741 * strdup: Other Builtins. (line 6)
41742 * strfmon: Other Builtins. (line 6)
41743 * strftime: Other Builtins. (line 6)
41744 * string constants: Incompatibilities. (line 9)
41745 * strlen: Other Builtins. (line 6)
41746 * strncasecmp: Other Builtins. (line 6)
41747 * strncat: Other Builtins. (line 6)
41748 * strncmp: Other Builtins. (line 6)
41749 * strncpy: Other Builtins. (line 6)
41750 * strndup: Other Builtins. (line 6)
41751 * strpbrk: Other Builtins. (line 6)
41752 * strrchr: Other Builtins. (line 6)
41753 * strspn: Other Builtins. (line 6)
41754 * strstr: Other Builtins. (line 6)
41755 * struct: Unnamed Fields. (line 6)
41756 * structures: Incompatibilities. (line 146)
41757 * structures, constructor expression: Compound Literals. (line 6)
41758 * submodel options: Submodel Options. (line 6)
41759 * subscripting: Subscripting. (line 6)
41760 * subscripting and function values: Subscripting. (line 6)
41761 * suffixes for C++ source: Invoking G++. (line 6)
41762 * SUNPRO_DEPENDENCIES: Environment Variables.
41764 * suppressing warnings: Warning Options. (line 6)
41765 * surprises in C++: C++ Misunderstandings.
41767 * syntax checking: Warning Options. (line 13)
41768 * system headers, warnings from: Warning Options. (line 702)
41769 * tan: Other Builtins. (line 6)
41770 * tanf: Other Builtins. (line 6)
41771 * tanh: Other Builtins. (line 6)
41772 * tanhf: Other Builtins. (line 6)
41773 * tanhl: Other Builtins. (line 6)
41774 * tanl: Other Builtins. (line 6)
41775 * target machine, specifying: Target Options. (line 6)
41776 * target options: Target Options. (line 6)
41777 * TC1: Standards. (line 13)
41778 * TC2: Standards. (line 13)
41779 * TC3: Standards. (line 13)
41780 * Technical Corrigenda: Standards. (line 13)
41781 * Technical Corrigendum 1: Standards. (line 13)
41782 * Technical Corrigendum 2: Standards. (line 13)
41783 * Technical Corrigendum 3: Standards. (line 13)
41784 * template instantiation: Template Instantiation.
41786 * temporaries, lifetime of: Temporaries. (line 6)
41787 * tgamma: Other Builtins. (line 6)
41788 * tgammaf: Other Builtins. (line 6)
41789 * tgammal: Other Builtins. (line 6)
41790 * Thread-Local Storage: Thread-Local. (line 6)
41791 * thunks: Nested Functions. (line 6)
41792 * tiny data section on the H8/300H and H8S: Function Attributes.
41794 * TLS: Thread-Local. (line 6)
41795 * tls_model attribute: Variable Attributes.
41797 * TMPDIR: Environment Variables.
41799 * toascii: Other Builtins. (line 6)
41800 * tolower: Other Builtins. (line 6)
41801 * toupper: Other Builtins. (line 6)
41802 * towlower: Other Builtins. (line 6)
41803 * towupper: Other Builtins. (line 6)
41804 * traditional C language: C Dialect Options. (line 250)
41805 * treelang <1>: Standards. (line 169)
41806 * treelang: G++ and GCC. (line 6)
41807 * trunc: Other Builtins. (line 6)
41808 * truncf: Other Builtins. (line 6)
41809 * truncl: Other Builtins. (line 6)
41810 * two-stage name lookup: Name lookup. (line 6)
41811 * type alignment: Alignment. (line 6)
41812 * type attributes: Type Attributes. (line 6)
41813 * type_info: Vague Linkage. (line 43)
41814 * typedef names as function parameters: Incompatibilities. (line 97)
41815 * typeof: Typeof. (line 6)
41816 * UHK fixed-suffix: Fixed-Point. (line 6)
41817 * uhk fixed-suffix: Fixed-Point. (line 6)
41818 * UHR fixed-suffix: Fixed-Point. (line 6)
41819 * uhr fixed-suffix: Fixed-Point. (line 6)
41820 * UK fixed-suffix: Fixed-Point. (line 6)
41821 * uk fixed-suffix: Fixed-Point. (line 6)
41822 * ULK fixed-suffix: Fixed-Point. (line 6)
41823 * ulk fixed-suffix: Fixed-Point. (line 6)
41824 * ULL integer suffix: Long Long. (line 6)
41825 * ULLK fixed-suffix: Fixed-Point. (line 6)
41826 * ullk fixed-suffix: Fixed-Point. (line 6)
41827 * ULLR fixed-suffix: Fixed-Point. (line 6)
41828 * ullr fixed-suffix: Fixed-Point. (line 6)
41829 * ULR fixed-suffix: Fixed-Point. (line 6)
41830 * ulr fixed-suffix: Fixed-Point. (line 6)
41831 * undefined behavior: Bug Criteria. (line 17)
41832 * undefined function value: Bug Criteria. (line 17)
41833 * underscores in variables in macros: Typeof. (line 42)
41834 * union: Unnamed Fields. (line 6)
41835 * union, casting to a: Cast to Union. (line 6)
41836 * unions: Incompatibilities. (line 146)
41837 * unknown pragmas, warning: Warning Options. (line 588)
41838 * unresolved references and -nodefaultlibs: Link Options. (line 79)
41839 * unresolved references and -nostdlib: Link Options. (line 79)
41840 * unused attribute.: Function Attributes.
41842 * UR fixed-suffix: Fixed-Point. (line 6)
41843 * ur fixed-suffix: Fixed-Point. (line 6)
41844 * used attribute.: Function Attributes.
41846 * User stack pointer in interrupts on the Blackfin: Function Attributes.
41848 * V in constraint: Simple Constraints. (line 41)
41849 * V850 Options: V850 Options. (line 6)
41850 * vague linkage: Vague Linkage. (line 6)
41851 * value after longjmp: Global Reg Vars. (line 66)
41852 * variable addressability on the IA-64: Function Attributes.
41854 * variable addressability on the M32R/D: Variable Attributes.
41856 * variable alignment: Alignment. (line 6)
41857 * variable attributes: Variable Attributes.
41859 * variable number of arguments: Variadic Macros. (line 6)
41860 * variable-length array scope: Variable Length. (line 23)
41861 * variable-length arrays: Variable Length. (line 6)
41862 * variables in specified registers: Explicit Reg Vars. (line 6)
41863 * variables, local, in macros: Typeof. (line 42)
41864 * variadic macros: Variadic Macros. (line 6)
41865 * VAX options: VAX Options. (line 6)
41866 * version_id attribute on IA64 HP-UX: Function Attributes.
41868 * vfprintf: Other Builtins. (line 6)
41869 * vfscanf: Other Builtins. (line 6)
41870 * visibility attribute: Function Attributes.
41872 * VLAs: Variable Length. (line 6)
41873 * void pointers, arithmetic: Pointer Arith. (line 6)
41874 * void, size of pointer to: Pointer Arith. (line 6)
41875 * volatile access: Volatiles. (line 6)
41876 * volatile applied to function: Function Attributes.
41878 * volatile read: Volatiles. (line 6)
41879 * volatile write: Volatiles. (line 6)
41880 * vprintf: Other Builtins. (line 6)
41881 * vscanf: Other Builtins. (line 6)
41882 * vsnprintf: Other Builtins. (line 6)
41883 * vsprintf: Other Builtins. (line 6)
41884 * vsscanf: Other Builtins. (line 6)
41885 * vtable: Vague Linkage. (line 28)
41886 * VxWorks Options: VxWorks Options. (line 6)
41887 * W floating point suffix: Floating Types. (line 6)
41888 * w floating point suffix: Floating Types. (line 6)
41889 * warn_unused_result attribute: Function Attributes.
41891 * warning for comparison of signed and unsigned values: Warning Options.
41893 * warning for overloaded virtual fn: C++ Dialect Options.
41895 * warning for reordering of member initializers: C++ Dialect Options.
41897 * warning for unknown pragmas: Warning Options. (line 588)
41898 * warning function attribute: Function Attributes.
41900 * warning messages: Warning Options. (line 6)
41901 * warnings from system headers: Warning Options. (line 702)
41902 * warnings vs errors: Warnings and Errors.
41904 * weak attribute: Function Attributes.
41906 * weakref attribute: Function Attributes.
41908 * whitespace: Incompatibilities. (line 112)
41909 * X in constraint: Simple Constraints. (line 112)
41910 * X3.159-1989: Standards. (line 13)
41911 * x86-64 options: x86-64 Options. (line 6)
41912 * x86-64 Options: i386 and x86-64 Options.
41914 * Xstormy16 Options: Xstormy16 Options. (line 6)
41915 * Xtensa Options: Xtensa Options. (line 6)
41916 * y0: Other Builtins. (line 6)
41917 * y0f: Other Builtins. (line 6)
41918 * y0l: Other Builtins. (line 6)
41919 * y1: Other Builtins. (line 6)
41920 * y1f: Other Builtins. (line 6)
41921 * y1l: Other Builtins. (line 6)
41922 * yn: Other Builtins. (line 6)
41923 * ynf: Other Builtins. (line 6)
41924 * ynl: Other Builtins. (line 6)
41925 * zero-length arrays: Zero Length. (line 6)
41926 * zero-size structures: Empty Structures. (line 6)
41927 * zSeries options: zSeries Options. (line 6)
41933 Node: G++ and GCC
\7f3821
41934 Node: Standards
\7f5886
41935 Node: Invoking GCC
\7f15179
41936 Node: Option Summary
\7f18996
41937 Node: Overall Options
\7f49780
41938 Node: Invoking G++
\7f63300
41939 Node: C Dialect Options
\7f64823
41940 Node: C++ Dialect Options
\7f78712
41941 Node: Objective-C and Objective-C++ Dialect Options
\7f99422
41942 Node: Language Independent Options
\7f111203
41943 Node: Warning Options
\7f113973
41944 Node: Debugging Options
\7f170532
41945 Node: Optimize Options
\7f207355
41946 Node: Preprocessor Options
\7f299864
41947 Ref: Wtrigraphs
\7f303949
41948 Ref: dashMF
\7f308753
41949 Ref: fdollars-in-identifiers
\7f319272
41950 Node: Assembler Options
\7f327508
41951 Node: Link Options
\7f328213
41952 Ref: Link Options-Footnote-1
\7f336781
41953 Node: Directory Options
\7f337115
41954 Node: Spec Files
\7f343177
41955 Node: Target Options
\7f363516
41956 Node: Submodel Options
\7f364940
41957 Node: ARC Options
\7f366580
41958 Node: ARM Options
\7f367770
41959 Node: AVR Options
\7f379677
41960 Node: Blackfin Options
\7f381810
41961 Node: CRIS Options
\7f387852
41962 Node: CRX Options
\7f392071
41963 Node: Darwin Options
\7f392496
41964 Node: DEC Alpha Options
\7f399989
41965 Node: DEC Alpha/VMS Options
\7f411466
41966 Node: FRV Options
\7f411851
41967 Node: GNU/Linux Options
\7f418577
41968 Node: H8/300 Options
\7f419035
41969 Node: HPPA Options
\7f420102
41970 Node: i386 and x86-64 Options
\7f429695
41971 Node: IA-64 Options
\7f456610
41972 Node: M32C Options
\7f463927
41973 Node: M32R/D Options
\7f465218
41974 Node: M680x0 Options
\7f468805
41975 Node: M68hc1x Options
\7f481378
41976 Node: MCore Options
\7f482946
41977 Node: MIPS Options
\7f483967
41978 Node: MMIX Options
\7f506201
41979 Node: MN10300 Options
\7f508683
41980 Node: MT Options
\7f510101
41981 Node: PDP-11 Options
\7f511015
41982 Node: PowerPC Options
\7f512849
41983 Node: RS/6000 and PowerPC Options
\7f513083
41984 Node: S/390 and zSeries Options
\7f542504
41985 Node: Score Options
\7f549987
41986 Node: SH Options
\7f550815
41987 Node: SPARC Options
\7f560652
41988 Node: SPU Options
\7f571625
41989 Node: System V Options
\7f573944
41990 Node: V850 Options
\7f574767
41991 Node: VAX Options
\7f577907
41992 Node: VxWorks Options
\7f578455
41993 Node: x86-64 Options
\7f579610
41994 Node: Xstormy16 Options
\7f579828
41995 Node: Xtensa Options
\7f580117
41996 Node: zSeries Options
\7f583957
41997 Node: Code Gen Options
\7f584153
41998 Node: Environment Variables
\7f606922
41999 Node: Precompiled Headers
\7f614818
42000 Node: Running Protoize
\7f621061
42001 Node: C Implementation
\7f627398
42002 Node: Translation implementation
\7f629061
42003 Node: Environment implementation
\7f629635
42004 Node: Identifiers implementation
\7f630185
42005 Node: Characters implementation
\7f631239
42006 Node: Integers implementation
\7f634045
42007 Node: Floating point implementation
\7f635870
42008 Node: Arrays and pointers implementation
\7f638799
42009 Ref: Arrays and pointers implementation-Footnote-1
\7f640234
42010 Node: Hints implementation
\7f640358
42011 Node: Structures unions enumerations and bit-fields implementation
\7f641824
42012 Node: Qualifiers implementation
\7f643787
42013 Node: Declarators implementation
\7f645559
42014 Node: Statements implementation
\7f645901
42015 Node: Preprocessing directives implementation
\7f646228
42016 Node: Library functions implementation
\7f648333
42017 Node: Architecture implementation
\7f648973
42018 Node: Locale-specific behavior implementation
\7f649676
42019 Node: C Extensions
\7f649981
42020 Node: Statement Exprs
\7f654537
42021 Node: Local Labels
\7f659050
42022 Node: Labels as Values
\7f662029
42023 Ref: Labels as Values-Footnote-1
\7f664402
42024 Node: Nested Functions
\7f664585
42025 Node: Constructing Calls
\7f668479
42026 Node: Typeof
\7f673202
42027 Node: Conditionals
\7f676368
42028 Node: Long Long
\7f677259
42029 Node: Complex
\7f678760
42030 Node: Floating Types
\7f681330
42031 Node: Decimal Float
\7f682449
42032 Node: Hex Floats
\7f684021
42033 Node: Fixed-Point
\7f685062
42034 Node: Zero Length
\7f688192
42035 Node: Empty Structures
\7f691470
42036 Node: Variable Length
\7f691886
42037 Node: Variadic Macros
\7f694653
42038 Node: Escaped Newlines
\7f697035
42039 Node: Subscripting
\7f697874
42040 Node: Pointer Arith
\7f698597
42041 Node: Initializers
\7f699165
42042 Node: Compound Literals
\7f699661
42043 Node: Designated Inits
\7f701836
42044 Node: Case Ranges
\7f705491
42045 Node: Cast to Union
\7f706174
42046 Node: Mixed Declarations
\7f707270
42047 Node: Function Attributes
\7f707776
42048 Node: Attribute Syntax
\7f761301
42049 Node: Function Prototypes
\7f771571
42050 Node: C++ Comments
\7f773352
42051 Node: Dollar Signs
\7f773871
42052 Node: Character Escapes
\7f774336
42053 Node: Alignment
\7f774630
42054 Node: Variable Attributes
\7f776004
42055 Ref: i386 Variable Attributes
\7f789904
42056 Node: Type Attributes
\7f795927
42057 Ref: i386 Type Attributes
\7f809239
42058 Ref: PowerPC Type Attributes
\7f810083
42059 Ref: SPU Type Attributes
\7f810936
42060 Node: Inline
\7f811227
42061 Node: Extended Asm
\7f816174
42062 Ref: Example of asm with clobbered asm reg
\7f822260
42063 Node: Constraints
\7f836356
42064 Node: Simple Constraints
\7f837206
42065 Node: Multi-Alternative
\7f843733
42066 Node: Modifiers
\7f845450
42067 Node: Machine Constraints
\7f848344
42068 Node: Asm Labels
\7f879774
42069 Node: Explicit Reg Vars
\7f881450
42070 Node: Global Reg Vars
\7f883058
42071 Node: Local Reg Vars
\7f887608
42072 Node: Alternate Keywords
\7f890049
42073 Node: Incomplete Enums
\7f891477
42074 Node: Function Names
\7f892234
42075 Node: Return Address
\7f894424
42076 Node: Vector Extensions
\7f897221
42077 Node: Offsetof
\7f900723
42078 Node: Atomic Builtins
\7f901509
42079 Node: Object Size Checking
\7f906594
42080 Node: Other Builtins
\7f911952
42081 Node: Target Builtins
\7f935623
42082 Node: Alpha Built-in Functions
\7f936414
42083 Node: ARM iWMMXt Built-in Functions
\7f939413
42084 Node: ARM NEON Intrinsics
\7f946132
42085 Node: Blackfin Built-in Functions
\7f1153970
42086 Node: FR-V Built-in Functions
\7f1154584
42087 Node: Argument Types
\7f1155443
42088 Node: Directly-mapped Integer Functions
\7f1157199
42089 Node: Directly-mapped Media Functions
\7f1158281
42090 Node: Raw read/write Functions
\7f1165313
42091 Node: Other Built-in Functions
\7f1166225
42092 Node: X86 Built-in Functions
\7f1167414
42093 Node: MIPS DSP Built-in Functions
\7f1204012
42094 Node: MIPS Paired-Single Support
\7f1216166
42095 Node: Paired-Single Arithmetic
\7f1217776
42096 Node: Paired-Single Built-in Functions
\7f1218716
42097 Node: MIPS-3D Built-in Functions
\7f1221380
42098 Node: PowerPC AltiVec Built-in Functions
\7f1226749
42099 Node: SPARC VIS Built-in Functions
\7f1328053
42100 Node: SPU Built-in Functions
\7f1329745
42101 Node: Target Format Checks
\7f1331527
42102 Node: Solaris Format Checks
\7f1331934
42103 Node: Pragmas
\7f1332331
42104 Node: ARM Pragmas
\7f1332961
42105 Node: M32C Pragmas
\7f1333564
42106 Node: RS/6000 and PowerPC Pragmas
\7f1334140
42107 Node: Darwin Pragmas
\7f1334882
42108 Node: Solaris Pragmas
\7f1335949
42109 Node: Symbol-Renaming Pragmas
\7f1337110
42110 Node: Structure-Packing Pragmas
\7f1339732
42111 Node: Weak Pragmas
\7f1341363
42112 Node: Diagnostic Pragmas
\7f1342165
42113 Node: Visibility Pragmas
\7f1344158
42114 Node: Unnamed Fields
\7f1344879
42115 Node: Thread-Local
\7f1346389
42116 Node: C99 Thread-Local Edits
\7f1348498
42117 Node: C++98 Thread-Local Edits
\7f1350510
42118 Node: Binary constants
\7f1353955
42119 Node: C++ Extensions
\7f1354626
42120 Node: Volatiles
\7f1356259
42121 Node: Restricted Pointers
\7f1358935
42122 Node: Vague Linkage
\7f1360529
42123 Node: C++ Interface
\7f1364185
42124 Ref: C++ Interface-Footnote-1
\7f1368482
42125 Node: Template Instantiation
\7f1368619
42126 Node: Bound member functions
\7f1375631
42127 Node: C++ Attributes
\7f1377174
42128 Node: Namespace Association
\7f1378832
42129 Node: Type Traits
\7f1380246
42130 Node: Java Exceptions
\7f1385805
42131 Node: Deprecated Features
\7f1387202
42132 Node: Backwards Compatibility
\7f1390166
42133 Node: Objective-C
\7f1391521
42134 Node: Executing code before main
\7f1392102
42135 Node: What you can and what you cannot do in +load
\7f1394708
42136 Node: Type encoding
\7f1396875
42137 Node: Garbage Collection
\7f1400262
42138 Node: Constant string objects
\7f1402886
42139 Node: compatibility_alias
\7f1405394
42140 Node: Compatibility
\7f1406272
42141 Node: Gcov
\7f1412839
42142 Node: Gcov Intro
\7f1413363
42143 Node: Invoking Gcov
\7f1416079
42144 Node: Gcov and Optimization
\7f1427940
42145 Node: Gcov Data Files
\7f1430593
42146 Node: Cross-profiling
\7f1431731
42147 Node: Trouble
\7f1433557
42148 Node: Actual Bugs
\7f1435097
42149 Node: Cross-Compiler Problems
\7f1435837
42150 Node: Interoperation
\7f1436251
42151 Node: Incompatibilities
\7f1443388
42152 Node: Fixed Headers
\7f1451538
42153 Node: Standard Libraries
\7f1453201
42154 Node: Disappointments
\7f1454573
42155 Node: C++ Misunderstandings
\7f1458931
42156 Node: Static Definitions
\7f1459750
42157 Node: Name lookup
\7f1460803
42158 Ref: Name lookup-Footnote-1
\7f1465581
42159 Node: Temporaries
\7f1465768
42160 Node: Copy Assignment
\7f1467744
42161 Node: Protoize Caveats
\7f1469551
42162 Node: Non-bugs
\7f1473524
42163 Node: Warnings and Errors
\7f1484028
42164 Node: Bugs
\7f1485792
42165 Node: Bug Criteria
\7f1486356
42166 Node: Bug Reporting
\7f1488566
42167 Node: Service
\7f1488787
42168 Node: Contributing
\7f1489606
42169 Node: Funding
\7f1490346
42170 Node: GNU Project
\7f1492835
42171 Node: Copying
\7f1493481
42172 Node: GNU Free Documentation License
\7f1531009
42173 Node: Contributors
\7f1553415
42174 Node: Option Index
\7f1589610
42175 Node: Keyword Index
\7f1742164