1 This is doc/gcc.info, produced by makeinfo version 4.8 from
2 ../../../../android-toolchain/gcc-4.4.0/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
6 Software Foundation, Inc.
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 "Funding Free Software", the Front-Cover Texts
12 being (a) (see below), and with the Back-Cover Texts being (b) (see
13 below). A copy of the license is included in the section entitled "GNU
14 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
35 Software Foundation, Inc.
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 "Funding Free Software", the Front-Cover Texts
41 being (a) (see below), and with the Back-Cover Texts being (b) (see
42 below). A copy of the license is included in the section entitled "GNU
43 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 4.4.0. 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-4.4/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 *Note GNAT Reference Manual: (gnat_rm)Top, for information on standard
306 conformance and compatibility of the Ada compiler.
308 *Note Standards: (gfortran)Standards, for details of standards
309 supported by GNU Fortran.
311 *Note Compatibility with the Java Platform: (gcj)Compatibility, for
312 details of compatibility between `gcj' and the Java Platform.
315 File: gcc.info, Node: Invoking GCC, Next: C Implementation, Prev: Standards, Up: Top
317 3 GCC Command Options
318 *********************
320 When you invoke GCC, it normally does preprocessing, compilation,
321 assembly and linking. The "overall options" allow you to stop this
322 process at an intermediate stage. For example, the `-c' option says
323 not to run the linker. Then the output consists of object files output
326 Other options are passed on to one stage of processing. Some options
327 control the preprocessor and others the compiler itself. Yet other
328 options control the assembler and linker; most of these are not
329 documented here, since you rarely need to use any of them.
331 Most of the command line options that you can use with GCC are useful
332 for C programs; when an option is only useful with another language
333 (usually C++), the explanation says so explicitly. If the description
334 for a particular option does not mention a source language, you can use
335 that option with all supported languages.
337 *Note Compiling C++ Programs: Invoking G++, for a summary of special
338 options for compiling C++ programs.
340 The `gcc' program accepts options and file names as operands. Many
341 options have multi-letter names; therefore multiple single-letter
342 options may _not_ be grouped: `-dv' is very different from `-d -v'.
344 You can mix options and other arguments. For the most part, the order
345 you use doesn't matter. Order does matter when you use several options
346 of the same kind; for example, if you specify `-L' more than once, the
347 directories are searched in the order specified. Also, the placement
348 of the `-l' option is significant.
350 Many options have long names starting with `-f' or with `-W'--for
351 example, `-fmove-loop-invariants', `-Wformat' and so on. Most of these
352 have both positive and negative forms; the negative form of `-ffoo'
353 would be `-fno-foo'. This manual documents only one of these two
354 forms, whichever one is not the default.
356 *Note Option Index::, for an index to GCC's options.
360 * Option Summary:: Brief list of all options, without explanations.
361 * Overall Options:: Controlling the kind of output:
362 an executable, object files, assembler files,
363 or preprocessed source.
364 * Invoking G++:: Compiling C++ programs.
365 * C Dialect Options:: Controlling the variant of C language compiled.
366 * C++ Dialect Options:: Variations on C++.
367 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
369 * Language Independent Options:: Controlling how diagnostics should be
371 * Warning Options:: How picky should the compiler be?
372 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
373 * Optimize Options:: How much optimization?
374 * Preprocessor Options:: Controlling header files and macro definitions.
375 Also, getting dependency information for Make.
376 * Assembler Options:: Passing options to the assembler.
377 * Link Options:: Specifying libraries and so on.
378 * Directory Options:: Where to find header files and libraries.
379 Where to find the compiler executable files.
380 * Spec Files:: How to pass switches to sub-processes.
381 * Target Options:: Running a cross-compiler, or an old version of GCC.
382 * Submodel Options:: Specifying minor hardware or convention variations,
383 such as 68010 vs 68020.
384 * Code Gen Options:: Specifying conventions for function calls, data layout
386 * Environment Variables:: Env vars that affect GCC.
387 * Precompiled Headers:: Compiling a header once, and using it many times.
388 * Running Protoize:: Automatically adding or removing function prototypes.
391 File: gcc.info, Node: Option Summary, Next: Overall Options, Up: Invoking GCC
396 Here is a summary of all the options, grouped by type. Explanations are
397 in the following sections.
400 *Note Options Controlling the Kind of Output: Overall Options.
401 -c -S -E -o FILE -combine -no-canonical-prefixes
402 -pipe -pass-exit-codes
403 -x LANGUAGE -v -### --help[=CLASS[,...]] --target-help
404 --version -wrapper@FILE -fplugin=FILE -fplugin-arg-NAME=ARG
407 *Note Options Controlling C Dialect: C Dialect Options.
408 -ansi -std=STANDARD -fgnu89-inline
410 -fno-asm -fno-builtin -fno-builtin-FUNCTION
411 -fhosted -ffreestanding -fopenmp -fms-extensions
412 -trigraphs -no-integrated-cpp -traditional -traditional-cpp
413 -fallow-single-precision -fcond-mismatch -flax-vector-conversions
414 -fsigned-bitfields -fsigned-char
415 -funsigned-bitfields -funsigned-char
417 _C++ Language Options_
418 *Note Options Controlling C++ Dialect: C++ Dialect Options.
419 -fabi-version=N -fno-access-control -fcheck-new
420 -fconserve-space -ffriend-injection
421 -fno-elide-constructors
422 -fno-enforce-eh-specs
423 -ffor-scope -fno-for-scope -fno-gnu-keywords
424 -fno-implicit-templates
425 -fno-implicit-inline-templates
426 -fno-implement-inlines -fms-extensions
427 -fno-nonansi-builtins -fno-operator-names
428 -fno-optional-diags -fpermissive
429 -frepo -fno-rtti -fstats -ftemplate-depth-N
430 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++
431 -fno-default-inline -fvisibility-inlines-hidden
432 -fvisibility-ms-compat
433 -Wabi -Wctor-dtor-privacy
434 -Wnon-virtual-dtor -Wreorder
435 -Weffc++ -Wstrict-null-sentinel
436 -Wno-non-template-friend -Wold-style-cast
437 -Woverloaded-virtual -Wno-pmf-conversions
440 _Objective-C and Objective-C++ Language Options_
441 *Note Options Controlling Objective-C and Objective-C++ Dialects:
442 Objective-C and Objective-C++ Dialect Options.
443 -fconstant-string-class=CLASS-NAME
444 -fgnu-runtime -fnext-runtime
446 -fobjc-call-cxx-cdtors
447 -fobjc-direct-dispatch
450 -freplace-objc-classes
454 -Wno-protocol -Wselector
455 -Wstrict-selector-match
456 -Wundeclared-selector
458 _Language Independent Options_
459 *Note Options to Control Diagnostic Messages Formatting: Language
462 -fdiagnostics-show-location=[once|every-line]
463 -fdiagnostics-show-option
466 *Note Options to Request or Suppress Warnings: Warning Options.
467 -fsyntax-only -pedantic -pedantic-errors
468 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds
469 -Wno-attributes -Wno-builtin-macro-redefined
470 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual
471 -Wchar-subscripts -Wclobbered -Wcomment
472 -Wconversion -Wcoverage-mismatch -Wno-deprecated
473 -Wno-deprecated-declarations -Wdisabled-optimization
474 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels
476 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2
477 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral
478 -Wformat-security -Wformat-y2k
479 -Wframe-larger-than=LEN -Wignored-qualifiers
480 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int
482 -Wno-int-to-pointer-cast -Wno-invalid-offsetof
483 -Winvalid-pch -Wlarger-than=LEN -Wunsafe-loop-optimizations
484 -Wlogical-op -Wlong-long
485 -Wmain -Wmissing-braces -Wmissing-field-initializers
486 -Wmissing-format-attribute -Wmissing-include-dirs
487 -Wmissing-noreturn -Wno-mudflap
488 -Wno-multichar -Wnonnull -Wno-overflow
489 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded
490 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format
491 -Wpointer-arith -Wno-pointer-to-int-cast
493 -Wreturn-type -Wsequence-point -Wshadow
494 -Wsign-compare -Wsign-conversion -Wstack-protector
495 -Wstrict-aliasing -Wstrict-aliasing=n
496 -Wstrict-overflow -Wstrict-overflow=N
497 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand
498 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized
499 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code
500 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
501 -Wunused-value -Wunused-variable
502 -Wvariadic-macros -Wvla
503 -Wvolatile-register-var -Wwrite-strings
505 _C and Objective-C-only Warning Options_
506 -Wbad-function-cast -Wmissing-declarations
507 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs
508 -Wold-style-declaration -Wold-style-definition
509 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion
510 -Wdeclaration-after-statement -Wpointer-sign
513 *Note Options for Debugging Your Program or GCC: Debugging Options.
514 -dLETTERS -dumpspecs -dumpmachine -dumpversion
515 -fdbg-cnt-list -fdbg-cnt=COUNTER-VALUE-LIST
516 -fdump-noaddr -fdump-unnumbered
517 -fdump-translation-unit[-N]
518 -fdump-class-hierarchy[-N]
519 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline
522 -fdump-tree-original[-N]
523 -fdump-tree-optimized[-N]
524 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias
526 -fdump-tree-ssa[-N] -fdump-tree-pre[-N]
527 -fdump-tree-ccp[-N] -fdump-tree-dce[-N]
528 -fdump-tree-gimple[-raw] -fdump-tree-mudflap[-N]
531 -fdump-tree-phiopt[-N]
532 -fdump-tree-forwprop[-N]
533 -fdump-tree-copyrename[-N]
534 -fdump-tree-nrv -fdump-tree-vect
539 -ftree-vectorizer-verbose=N
540 -fdump-tree-storeccp[-N]
541 -feliminate-dwarf2-dups -feliminate-unused-debug-types
542 -feliminate-unused-debug-symbols -femit-class-debug-always
543 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs
544 -frandom-seed=STRING -fsched-verbose=N
545 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose
546 -ftest-coverage -ftime-report -fvar-tracking
547 -g -gLEVEL -gcoff -gdwarf-2
548 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+
549 -fno-merge-debug-strings -fno-dwarf2-cfi-asm
550 -fdebug-prefix-map=OLD=NEW
551 -femit-struct-debug-baseonly -femit-struct-debug-reduced
552 -femit-struct-debug-detailed[=SPEC-LIST]
553 -p -pg -print-file-name=LIBRARY -print-libgcc-file-name
554 -print-multi-directory -print-multi-lib
555 -print-prog-name=PROGRAM -print-search-dirs -Q
556 -print-sysroot -print-sysroot-headers-suffix
559 _Optimization Options_
560 *Note Options that Control Optimization: Optimize Options.
561 -falign-functions[=N] -falign-jumps[=N]
562 -falign-labels[=N] -falign-loops[=N] -fassociative-math
563 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize
564 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves
565 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping
566 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range
567 -fdata-sections -fdce -fdce
568 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse -fdyn-ipa
569 -fearly-inlining -fexpensive-optimizations -ffast-math
570 -ffinite-math-only -ffloat-store -fforward-propagate
571 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm
572 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining
573 -finline-functions -finline-functions-called-once -finline-limit=N
574 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta
575 -fipa-pure-const -fipa-reference -fipa-struct-reorg
576 -fipa-type-escape -fira-algorithm=ALGORITHM
577 -fira-region=REGION -fira-coalesce -fno-ira-share-save-slots
578 -fno-ira-share-spill-slots -fira-verbose=N
579 -fivopts -fkeep-inline-functions -fkeep-static-consts
580 -floop-block -floop-interchange -floop-strip-mine
581 -fmerge-all-constants -fmerge-constants -fmodulo-sched
582 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap
583 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline
584 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability
585 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2
586 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros
587 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss
588 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls
589 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays
590 -fprofile-correction -fprofile-dir=PATH -fprofile-generate
591 -fprofile-generate=PATH
592 -fprofile-use -fprofile-use=PATH -fprofile-values
593 -freciprocal-math -fregmove -frename-registers -freorder-blocks
594 -freorder-blocks-and-partition -freorder-functions
595 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops
596 -frounding-math -fsched2-use-superblocks
597 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous
598 -fsched-stalled-insns-dep[=N] -fsched-stalled-insns[=N]
599 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee
600 -fselective-scheduling -fselective-scheduling2
601 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops
602 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller
603 -fsplit-wide-types -fstack-protector -fstack-protector-all
604 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer
605 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop
606 -ftree-copyrename -ftree-dce
607 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im
608 -ftree-loop-distribution
609 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize
610 -ftree-parallelize-loops=N -ftree-pre -ftree-reassoc
611 -ftree-sink -ftree-sra -ftree-switch-conversion
612 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp
613 -funit-at-a-time -funroll-all-loops -funroll-loops
614 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops
615 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb
618 -O -O0 -O1 -O2 -O3 -Os
620 _Preprocessor Options_
621 *Note Options Controlling the Preprocessor: Preprocessor Options.
627 -include FILE -imacros FILE
628 -iprefix FILE -iwithprefix DIR
629 -iwithprefixbefore DIR -isystem DIR
630 -imultilib DIR -isysroot DIR
631 -M -MM -MF -MG -MP -MQ -MT -nostdinc
632 -P -fworking-directory -remap
633 -trigraphs -undef -UMACRO -Wp,OPTION
634 -Xpreprocessor OPTION
637 *Note Passing Options to the Assembler: Assembler Options.
638 -Wa,OPTION -Xassembler OPTION
641 *Note Options for Linking: Link Options.
642 OBJECT-FILE-NAME -lLIBRARY
643 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic
644 -s -static -static-libgcc -shared -shared-libgcc -symbolic
645 -T SCRIPT -Wl,OPTION -Xlinker OPTION
649 *Note Options for Directory Search: Directory Options.
650 -BPREFIX -IDIR -iquoteDIR -LDIR
651 -specs=FILE -I- --sysroot=DIR
654 *Note Target Options::.
655 -V VERSION -b MACHINE
657 _Machine Dependent Options_
658 *Note Hardware Models and Configurations: Submodel Options.
662 -mmangle-cpu -mcpu=CPU -mtext=TEXT-SECTION
663 -mdata=DATA-SECTION -mrodata=READONLY-DATA-SECTION
666 -mapcs-frame -mno-apcs-frame
668 -mapcs-stack-check -mno-apcs-stack-check
669 -mapcs-float -mno-apcs-float
670 -mapcs-reentrant -mno-apcs-reentrant
671 -msched-prolog -mno-sched-prolog
672 -mlittle-endian -mbig-endian -mwords-little-endian
673 -mfloat-abi=NAME -msoft-float -mhard-float -mfpe
674 -mthumb-interwork -mno-thumb-interwork
675 -mcpu=NAME -march=NAME -mfpu=NAME
676 -mstructure-size-boundary=N
678 -mlong-calls -mno-long-calls
679 -msingle-pic-base -mno-single-pic-base
682 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns
685 -mtpcs-frame -mtpcs-leaf-frame
686 -mcaller-super-interworking -mcallee-super-interworking
693 -mmcu=MCU -msize -minit-stack=N -mno-interrupts
694 -mcall-prologues -mno-tablejump -mtiny-stack -mint8
697 -mcpu=CPU[-SIREVISION]
698 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer
699 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly
700 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library
701 -mno-id-shared-library -mshared-library-id=N
702 -mleaf-id-shared-library -mno-leaf-id-shared-library
703 -msep-data -mno-sep-data -mlong-calls -mno-long-calls
704 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram
708 -mcpu=CPU -march=CPU -mtune=CPU
709 -mmax-stack-frame=N -melinux-stacksize=N
710 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects
711 -mstack-align -mdata-align -mconst-align
712 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt
713 -melf -maout -melinux -mlinux -sim -sim2
714 -mmul-bug-workaround -mno-mul-bug-workaround
720 -all_load -allowable_client -arch -arch_errors_fatal
721 -arch_only -bind_at_load -bundle -bundle_loader
722 -client_name -compatibility_version -current_version
724 -dependency-file -dylib_file -dylinker_install_name
725 -dynamic -dynamiclib -exported_symbols_list
726 -filelist -flat_namespace -force_cpusubtype_ALL
727 -force_flat_namespace -headerpad_max_install_names
729 -image_base -init -install_name -keep_private_externs
730 -multi_module -multiply_defined -multiply_defined_unused
731 -noall_load -no_dead_strip_inits_and_terms
732 -nofixprebinding -nomultidefs -noprebind -noseglinkedit
733 -pagezero_size -prebind -prebind_all_twolevel_modules
734 -private_bundle -read_only_relocs -sectalign
735 -sectobjectsymbols -whyload -seg1addr
736 -sectcreate -sectobjectsymbols -sectorder
737 -segaddr -segs_read_only_addr -segs_read_write_addr
738 -seg_addr_table -seg_addr_table_filename -seglinkedit
739 -segprot -segs_read_only_addr -segs_read_write_addr
740 -single_module -static -sub_library -sub_umbrella
741 -twolevel_namespace -umbrella -undefined
742 -unexported_symbols_list -weak_reference_mismatches
743 -whatsloaded -F -gused -gfull -mmacosx-version-min=VERSION
744 -mkernel -mone-byte-bool
747 -mno-fp-regs -msoft-float -malpha-as -mgas
748 -mieee -mieee-with-inexact -mieee-conformant
749 -mfp-trap-mode=MODE -mfp-rounding-mode=MODE
750 -mtrap-precision=MODE -mbuild-constants
751 -mcpu=CPU-TYPE -mtune=CPU-TYPE
752 -mbwx -mmax -mfix -mcix
753 -mfloat-vax -mfloat-ieee
754 -mexplicit-relocs -msmall-data -mlarge-data
755 -msmall-text -mlarge-text
756 -mmemory-latency=TIME
758 _DEC Alpha/VMS Options_
762 -msmall-model -mno-lsim
765 -mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64
766 -mhard-float -msoft-float
767 -malloc-cc -mfixed-cc -mdword -mno-dword
769 -mmedia -mno-media -mmuladd -mno-muladd
770 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic
771 -mlinked-fp -mlong-calls -malign-labels
772 -mlibrary-pic -macc-4 -macc-8
773 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move
774 -moptimize-membar -mno-optimize-membar
775 -mscc -mno-scc -mcond-exec -mno-cond-exec
776 -mvliw-branch -mno-vliw-branch
777 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec
778 -mno-nested-cond-exec -mtomcat-stats
786 -mrelax -mh -ms -mn -mint32 -malign-300
789 -march=ARCHITECTURE-TYPE
790 -mbig-switch -mdisable-fpregs -mdisable-indexing
791 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld
792 -mfixed-range=REGISTER-RANGE
793 -mjump-in-delay -mlinker-opt -mlong-calls
794 -mlong-load-store -mno-big-switch -mno-disable-fpregs
795 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
796 -mno-jump-in-delay -mno-long-load-store
797 -mno-portable-runtime -mno-soft-float
798 -mno-space-regs -msoft-float -mpa-risc-1-0
799 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
800 -mschedule=CPU-TYPE -mspace-regs -msio -mwsio
801 -munix=UNIX-STD -nolibdld -static -threads
803 _i386 and x86-64 Options_
804 -mtune=CPU-TYPE -march=CPU-TYPE
806 -masm=DIALECT -mno-fancy-math-387
807 -mno-fp-ret-in-387 -msoft-float
808 -mno-wide-multiply -mrtd -malign-double
809 -mpreferred-stack-boundary=NUM
810 -mincoming-stack-boundary=NUM
811 -mcld -mcx16 -msahf -mrecip
812 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx
814 -msse4a -m3dnow -mpopcnt -mabm -msse5
815 -mthreads -mno-align-stringops -minline-all-stringops
816 -minline-stringops-dynamically -minline-compares
817 -mstringop-strategy=ALG -mpush-args -maccumulate-outgoing-args
818 -m128bit-long-double -m96bit-long-double -mregparm=NUM -msseregparm
819 -mveclibabi=TYPE -mpc32 -mpc64 -mpc80 -mstackrealign
820 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs
822 -m32 -m64 -mlarge-data-threshold=NUM
823 -mfused-madd -mno-fused-madd -msse2avx
826 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic
827 -mvolatile-asm-stop -mregister-names -mno-sdata
828 -mconstant-gp -mauto-pic -minline-float-divide-min-latency
829 -minline-float-divide-max-throughput
830 -minline-int-divide-min-latency
831 -minline-int-divide-max-throughput
832 -minline-sqrt-min-latency -minline-sqrt-max-throughput
833 -mno-dwarf2-asm -mearly-stop-bits
834 -mfixed-range=REGISTER-RANGE -mtls-size=TLS-SIZE
835 -mtune=CPU-TYPE -mt -pthread -milp32 -mlp64
836 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec
837 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec
838 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose
839 -mno-sched-prefer-non-data-spec-insns
840 -mno-sched-prefer-non-control-spec-insns
841 -mno-sched-count-spec-in-critical-path
846 -malign-loops -mno-align-loops
849 -mmodel=CODE-SIZE-MODEL-TYPE
851 -mno-flush-func -mflush-func=NAME
852 -mno-flush-trap -mflush-trap=NUMBER
856 -mcpu=CPU -msim -memregs=NUMBER
859 -march=ARCH -mcpu=CPU -mtune=TUNE
860 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
861 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407
862 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020
863 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort
864 -mno-short -mhard-float -m68881 -msoft-float -mpcrel
865 -malign-int -mstrict-align -msep-data -mno-sep-data
866 -mshared-library-id=n -mid-shared-library -mno-id-shared-library
870 -m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12
871 -mauto-incdec -minmax -mlong-calls -mshort
872 -msoft-reg-count=COUNT
875 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates
876 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
877 -m4byte-functions -mno-4byte-functions -mcallgraph-data
878 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
879 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
882 -EL -EB -march=ARCH -mtune=ARCH
883 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2
885 -mips16 -mno-mips16 -mflip-mips16
886 -minterlink-mips16 -mno-interlink-mips16
887 -mabi=ABI -mabicalls -mno-abicalls
888 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot
889 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float
890 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2
892 -msmartmips -mno-smartmips
893 -mpaired-single -mno-paired-single -mdmx -mno-mdmx
894 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc
895 -mlong64 -mlong32 -msym32 -mno-sym32
896 -GNUM -mlocal-sdata -mno-local-sdata
897 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt
898 -membedded-data -mno-embedded-data
899 -muninit-const-in-rodata -mno-uninit-const-in-rodata
900 -mcode-readable=SETTING
901 -msplit-addresses -mno-split-addresses
902 -mexplicit-relocs -mno-explicit-relocs
903 -mcheck-zero-division -mno-check-zero-division
904 -mdivide-traps -mdivide-breaks
905 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls
906 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp
907 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400
908 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120
909 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1
910 -mflush-func=FUNC -mno-flush-func
911 -mbranch-cost=NUM -mbranch-likely -mno-branch-likely
912 -mfp-exceptions -mno-fp-exceptions
913 -mvr4130-align -mno-vr4130-align
916 -mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu
917 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols
918 -melf -mbranch-predict -mno-branch-predict -mbase-addresses
919 -mno-base-addresses -msingle-exit -mno-single-exit
922 -mmult-bug -mno-mult-bug
925 -mreturn-pointer-on-d0
929 -mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10
930 -mbcopy -mbcopy-builtin -mint32 -mno-int16
931 -mint16 -mno-int32 -mfloat32 -mno-float64
932 -mfloat64 -mno-float32 -mabshi -mno-abshi
933 -mbranch-expensive -mbranch-cheap
934 -msplit -mno-split -munix-asm -mdec-asm
937 -mae=AE_TYPE -mvliw-lookahead=N
938 -msymbol-as-address -mno-inefficient-warnings
940 _PowerPC Options_ See RS/6000 and PowerPC Options.
942 _RS/6000 and PowerPC Options_
945 -mpower -mno-power -mpower2 -mno-power2
946 -mpowerpc -mpowerpc64 -mno-powerpc
947 -maltivec -mno-altivec
948 -mpowerpc-gpopt -mno-powerpc-gpopt
949 -mpowerpc-gfxopt -mno-powerpc-gfxopt
950 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd
951 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp
952 -mnew-mnemonics -mold-mnemonics
953 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc
954 -m64 -m32 -mxl-compat -mno-xl-compat -mpe
955 -malign-power -malign-natural
956 -msoft-float -mhard-float -mmultiple -mno-multiple
957 -msingle-float -mdouble-float -msimple-fpu
958 -mstring -mno-string -mupdate -mno-update
959 -mavoid-indexed-addresses -mno-avoid-indexed-addresses
960 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
961 -mstrict-align -mno-strict-align -mrelocatable
962 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
963 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
964 -mdynamic-no-pic -maltivec -mswdiv
965 -mprioritize-restricted-insns=PRIORITY
966 -msched-costly-dep=DEPENDENCE_TYPE
967 -minsert-sched-nops=SCHEME
968 -mcall-sysv -mcall-netbsd
969 -maix-struct-return -msvr4-struct-return
970 -mabi=ABI-TYPE -msecure-plt -mbss-plt
976 -mgen-cell-microcode -mwarn-cell-microcode
980 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double
981 -mprototype -mno-prototype
982 -msim -mmvme -mads -myellowknife -memb -msdata
983 -msdata=OPT -mvxworks -G NUM -pthread
985 _S/390 and zSeries Options_
986 -mtune=CPU-TYPE -march=CPU-TYPE
987 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp
988 -mlong-double-64 -mlong-double-128
989 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack
990 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle
991 -m64 -m31 -mdebug -mno-debug -mesa -mzarch
992 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd
993 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard
1000 -mscore5 -mscore5u -mscore7 -mscore7d
1003 -m1 -m2 -m2e -m3 -m3e
1004 -m4-nofpu -m4-single-only -m4-single -m4
1005 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al
1006 -m5-64media -m5-64media-nofpu
1007 -m5-32media -m5-32media-nofpu
1008 -m5-compact -m5-compact-nofpu
1009 -mb -ml -mdalign -mrelax
1010 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave
1011 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace
1012 -mprefergot -musermode -multcost=NUMBER -mdiv=STRATEGY
1013 -mdivsi3_libfunc=NAME -mfixed-range=REGISTER-RANGE
1014 -madjust-unroll -mindexed-addressing -mgettrcost=NUMBER -mpt-fixed
1021 -m32 -m64 -mapp-regs -mno-app-regs
1022 -mfaster-structs -mno-faster-structs
1023 -mfpu -mno-fpu -mhard-float -msoft-float
1024 -mhard-quad-float -msoft-quad-float
1025 -mimpure-text -mno-impure-text -mlittle-endian
1026 -mstack-bias -mno-stack-bias
1027 -munaligned-doubles -mno-unaligned-doubles
1028 -mv8plus -mno-v8plus -mvis -mno-vis
1029 -threads -pthreads -pthread
1032 -mwarn-reloc -merror-reloc
1033 -msafe-dma -munsafe-dma
1035 -msmall-mem -mlarge-mem -mstdmain
1036 -mfixed-range=REGISTER-RANGE
1039 -Qy -Qn -YP,PATHS -Ym,DIR
1042 -mlong-calls -mno-long-calls -mep -mno-ep
1043 -mprolog-function -mno-prolog-function -mspace
1044 -mtda=N -msda=N -mzda=N
1045 -mapp-regs -mno-app-regs
1046 -mdisable-callt -mno-disable-callt
1055 -mrtp -non-static -Bstatic -Bdynamic
1056 -Xbind-lazy -Xbind-now
1058 _x86-64 Options_ See i386 and x86-64 Options.
1060 _i386 and x86-64 Windows Options_
1061 -mconsole -mcygwin -mno-cygwin -mdll
1062 -mnop-fun-dllimport -mthread -mwin32 -mwindows
1068 -mconst16 -mno-const16
1069 -mfused-madd -mno-fused-madd
1070 -mserialize-volatile -mno-serialize-volatile
1071 -mtext-section-literals -mno-text-section-literals
1072 -mtarget-align -mno-target-align
1073 -mlongcalls -mno-longcalls
1075 _zSeries Options_ See S/390 and zSeries Options.
1077 _Code Generation Options_
1078 *Note Options for Code Generation Conventions: Code Gen Options.
1079 -fcall-saved-REG -fcall-used-REG
1080 -ffixed-REG -fexceptions
1081 -fnon-call-exceptions -funwind-tables
1082 -fasynchronous-unwind-tables
1083 -finhibit-size-directive -finstrument-functions
1084 -finstrument-functions-exclude-function-list=SYM,SYM,...
1085 -finstrument-functions-exclude-file-list=FILE,FILE,...
1086 -fno-common -fno-ident
1087 -fpcc-struct-return -fpic -fPIC -fpie -fPIE
1089 -frecord-gcc-switches
1090 -freg-struct-return -fshort-enums
1091 -fshort-double -fshort-wchar
1092 -fverbose-asm -fpack-struct[=N] -fstack-check
1093 -fstack-limit-register=REG -fstack-limit-symbol=SYM
1094 -fno-stack-limit -fargument-alias -fargument-noalias
1095 -fargument-noalias-global -fargument-noalias-anything
1096 -fleading-underscore -ftls-model=MODEL
1097 -ftrapv -fwrapv -fbounds-check
1103 * Overall Options:: Controlling the kind of output:
1104 an executable, object files, assembler files,
1105 or preprocessed source.
1106 * C Dialect Options:: Controlling the variant of C language compiled.
1107 * C++ Dialect Options:: Variations on C++.
1108 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1110 * Language Independent Options:: Controlling how diagnostics should be
1112 * Warning Options:: How picky should the compiler be?
1113 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1114 * Optimize Options:: How much optimization?
1115 * Preprocessor Options:: Controlling header files and macro definitions.
1116 Also, getting dependency information for Make.
1117 * Assembler Options:: Passing options to the assembler.
1118 * Link Options:: Specifying libraries and so on.
1119 * Directory Options:: Where to find header files and libraries.
1120 Where to find the compiler executable files.
1121 * Spec Files:: How to pass switches to sub-processes.
1122 * Target Options:: Running a cross-compiler, or an old version of GCC.
1125 File: gcc.info, Node: Overall Options, Next: Invoking G++, Prev: Option Summary, Up: Invoking GCC
1127 3.2 Options Controlling the Kind of Output
1128 ==========================================
1130 Compilation can involve up to four stages: preprocessing, compilation
1131 proper, assembly and linking, always in that order. GCC is capable of
1132 preprocessing and compiling several files either into several assembler
1133 input files, or into one assembler input file; then each assembler
1134 input file produces an object file, and linking combines all the object
1135 files (those newly compiled, and those specified as input) into an
1138 For any given input file, the file name suffix determines what kind of
1139 compilation is done:
1142 C source code which must be preprocessed.
1145 C source code which should not be preprocessed.
1148 C++ source code which should not be preprocessed.
1151 Objective-C source code. Note that you must link with the
1152 `libobjc' library to make an Objective-C program work.
1155 Objective-C source code which should not be preprocessed.
1159 Objective-C++ source code. Note that you must link with the
1160 `libobjc' library to make an Objective-C++ program work. Note
1161 that `.M' refers to a literal capital M.
1164 Objective-C++ source code which should not be preprocessed.
1167 C, C++, Objective-C or Objective-C++ header file to be turned into
1168 a precompiled header.
1177 C++ source code which must be preprocessed. Note that in `.cxx',
1178 the last two letters must both be literally `x'. Likewise, `.C'
1179 refers to a literal capital C.
1183 Objective-C++ source code which must be preprocessed.
1186 Objective-C++ source code which should not be preprocessed.
1196 C++ header file to be turned into a precompiled header.
1201 Fixed form Fortran source code which should not be preprocessed.
1208 Fixed form Fortran source code which must be preprocessed (with
1209 the traditional preprocessor).
1215 Free form Fortran source code which should not be preprocessed.
1221 Free form Fortran source code which must be preprocessed (with the
1222 traditional preprocessor).
1225 Ada source code file which contains a library unit declaration (a
1226 declaration of a package, subprogram, or generic, or a generic
1227 instantiation), or a library unit renaming declaration (a package,
1228 generic, or subprogram renaming declaration). Such files are also
1232 Ada source code file containing a library unit body (a subprogram
1233 or package body). Such files are also called "bodies".
1240 Assembler code which must be preprocessed.
1243 An object file to be fed straight into linking. Any file name
1244 with no recognized suffix is treated this way.
1246 You can specify the input language explicitly with the `-x' option:
1249 Specify explicitly the LANGUAGE for the following input files
1250 (rather than letting the compiler choose a default based on the
1251 file name suffix). This option applies to all following input
1252 files until the next `-x' option. Possible values for LANGUAGE
1254 c c-header c-cpp-output
1255 c++ c++-header c++-cpp-output
1256 objective-c objective-c-header objective-c-cpp-output
1257 objective-c++ objective-c++-header objective-c++-cpp-output
1258 assembler assembler-with-cpp
1260 f77 f77-cpp-input f95 f95-cpp-input
1264 Turn off any specification of a language, so that subsequent files
1265 are handled according to their file name suffixes (as they are if
1266 `-x' has not been used at all).
1269 Normally the `gcc' program will exit with the code of 1 if any
1270 phase of the compiler returns a non-success return code. If you
1271 specify `-pass-exit-codes', the `gcc' program will instead return
1272 with numerically highest error produced by any phase that returned
1273 an error indication. The C, C++, and Fortran frontends return 4,
1274 if an internal compiler error is encountered.
1276 If you only want some of the stages of compilation, you can use `-x'
1277 (or filename suffixes) to tell `gcc' where to start, and one of the
1278 options `-c', `-S', or `-E' to say where `gcc' is to stop. Note that
1279 some combinations (for example, `-x cpp-output -E') instruct `gcc' to
1283 Compile or assemble the source files, but do not link. The linking
1284 stage simply is not done. The ultimate output is in the form of an
1285 object file for each source file.
1287 By default, the object file name for a source file is made by
1288 replacing the suffix `.c', `.i', `.s', etc., with `.o'.
1290 Unrecognized input files, not requiring compilation or assembly,
1294 Stop after the stage of compilation proper; do not assemble. The
1295 output is in the form of an assembler code file for each
1296 non-assembler input file specified.
1298 By default, the assembler file name for a source file is made by
1299 replacing the suffix `.c', `.i', etc., with `.s'.
1301 Input files that don't require compilation are ignored.
1304 Stop after the preprocessing stage; do not run the compiler
1305 proper. The output is in the form of preprocessed source code,
1306 which is sent to the standard output.
1308 Input files which don't require preprocessing are ignored.
1311 Place output in file FILE. This applies regardless to whatever
1312 sort of output is being produced, whether it be an executable file,
1313 an object file, an assembler file or preprocessed C code.
1315 If `-o' is not specified, the default is to put an executable file
1316 in `a.out', the object file for `SOURCE.SUFFIX' in `SOURCE.o', its
1317 assembler file in `SOURCE.s', a precompiled header file in
1318 `SOURCE.SUFFIX.gch', and all preprocessed C source on standard
1322 Print (on standard error output) the commands executed to run the
1323 stages of compilation. Also print the version number of the
1324 compiler driver program and of the preprocessor and the compiler
1328 Like `-v' except the commands are not executed and all command
1329 arguments are quoted. This is useful for shell scripts to capture
1330 the driver-generated command lines.
1333 Use pipes rather than temporary files for communication between the
1334 various stages of compilation. This fails to work on some systems
1335 where the assembler is unable to read from a pipe; but the GNU
1336 assembler has no trouble.
1339 If you are compiling multiple source files, this option tells the
1340 driver to pass all the source files to the compiler at once (for
1341 those languages for which the compiler can handle this). This
1342 will allow intermodule analysis (IMA) to be performed by the
1343 compiler. Currently the only language for which this is supported
1344 is C. If you pass source files for multiple languages to the
1345 driver, using this option, the driver will invoke the compiler(s)
1346 that support IMA once each, passing each compiler all the source
1347 files appropriate for it. For those languages that do not support
1348 IMA this option will be ignored, and the compiler will be invoked
1349 once for each source file in that language. If you use this
1350 option in conjunction with `-save-temps', the compiler will
1351 generate multiple pre-processed files (one for each source file),
1352 but only one (combined) `.o' or `.s' file.
1355 Print (on the standard output) a description of the command line
1356 options understood by `gcc'. If the `-v' option is also specified
1357 then `--help' will also be passed on to the various processes
1358 invoked by `gcc', so that they can display the command line options
1359 they accept. If the `-Wextra' option has also been specified
1360 (prior to the `--help' option), then command line options which
1361 have no documentation associated with them will also be displayed.
1364 Print (on the standard output) a description of target-specific
1365 command line options for each tool. For some targets extra
1366 target-specific information may also be printed.
1368 `--help={CLASS|[^]QUALIFIER}[,...]'
1369 Print (on the standard output) a description of the command line
1370 options understood by the compiler that fit into all specified
1371 classes and qualifiers. These are the supported classes:
1374 This will display all of the optimization options supported
1378 This will display all of the options controlling warning
1379 messages produced by the compiler.
1382 This will display target-specific options. Unlike the
1383 `--target-help' option however, target-specific options of the
1384 linker and assembler will not be displayed. This is because
1385 those tools do not currently support the extended `--help='
1389 This will display the values recognized by the `--param'
1393 This will display the options supported for LANGUAGE, where
1394 LANGUAGE is the name of one of the languages supported in this
1398 This will display the options that are common to all
1401 These are the supported qualifiers:
1404 Display only those options which are undocumented.
1407 Display options which take an argument that appears after an
1408 equal sign in the same continuous piece of text, such as:
1412 Display options which take an argument that appears as a
1413 separate word following the original option, such as: `-o
1416 Thus for example to display all the undocumented target-specific
1417 switches supported by the compiler the following can be used:
1419 --help=target,undocumented
1421 The sense of a qualifier can be inverted by prefixing it with the
1422 `^' character, so for example to display all binary warning
1423 options (i.e., ones that are either on or off and that do not take
1424 an argument), which have a description the following can be used:
1426 --help=warnings,^joined,^undocumented
1428 The argument to `--help=' should not consist solely of inverted
1431 Combining several classes is possible, although this usually
1432 restricts the output by so much that there is nothing to display.
1433 One case where it does work however is when one of the classes is
1434 TARGET. So for example to display all the target-specific
1435 optimization options the following can be used:
1437 --help=target,optimizers
1439 The `--help=' option can be repeated on the command line. Each
1440 successive use will display its requested class of options,
1441 skipping those that have already been displayed.
1443 If the `-Q' option appears on the command line before the
1444 `--help=' option, then the descriptive text displayed by `--help='
1445 is changed. Instead of describing the displayed options, an
1446 indication is given as to whether the option is enabled, disabled
1447 or set to a specific value (assuming that the compiler knows this
1448 at the point where the `--help=' option is used).
1450 Here is a truncated example from the ARM port of `gcc':
1452 % gcc -Q -mabi=2 --help=target -c
1453 The following options are target specific:
1455 -mabort-on-noreturn [disabled]
1458 The output is sensitive to the effects of previous command line
1459 options, so for example it is possible to find out which
1460 optimizations are enabled at `-O2' by using:
1462 -Q -O2 --help=optimizers
1464 Alternatively you can discover which binary optimizations are
1465 enabled by `-O3' by using:
1467 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1468 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1469 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1471 `-no-canonical-prefixes'
1472 Do not expand any symbolic links, resolve references to `/../' or
1473 `/./', or make the path absolute when generating a relative prefix.
1476 Display the version number and copyrights of the invoked GCC.
1479 Invoke all subcommands under a wrapper program. It takes a single
1480 comma separated list as an argument, which will be used to invoke
1483 gcc -c t.c -wrapper gdb,--args
1485 This will invoke all subprograms of gcc under "gdb -args", thus
1486 cc1 invocation will be "gdb -args cc1 ...".
1489 Load the plugin code in file NAME.so, assumed to be a shared
1490 object to be dlopen'd by the compiler. The base name of the
1491 shared object file is used to identify the plugin for the purposes
1492 of argument parsing (See `-fplugin-arg-NAME-KEY=VALUE' below).
1493 Each plugin should define the callback functions specified in the
1496 `-fplugin-arg-NAME-KEY=VALUE'
1497 Define an argument called KEY with a value of VALUE for the plugin
1501 Read command-line options from FILE. The options read are
1502 inserted in place of the original @FILE option. If FILE does not
1503 exist, or cannot be read, then the option will be treated
1504 literally, and not removed.
1506 Options in FILE are separated by whitespace. A whitespace
1507 character may be included in an option by surrounding the entire
1508 option in either single or double quotes. Any character
1509 (including a backslash) may be included by prefixing the character
1510 to be included with a backslash. The FILE may itself contain
1511 additional @FILE options; any such options will be processed
1515 File: gcc.info, Node: Invoking G++, Next: C Dialect Options, Prev: Overall Options, Up: Invoking GCC
1517 3.3 Compiling C++ Programs
1518 ==========================
1520 C++ source files conventionally use one of the suffixes `.C', `.cc',
1521 `.cpp', `.CPP', `.c++', `.cp', or `.cxx'; C++ header files often use
1522 `.hh', `.hpp', `.H', or (for shared template code) `.tcc'; and
1523 preprocessed C++ files use the suffix `.ii'. GCC recognizes files with
1524 these names and compiles them as C++ programs even if you call the
1525 compiler the same way as for compiling C programs (usually with the
1528 However, the use of `gcc' does not add the C++ library. `g++' is a
1529 program that calls GCC and treats `.c', `.h' and `.i' files as C++
1530 source files instead of C source files unless `-x' is used, and
1531 automatically specifies linking against the C++ library. This program
1532 is also useful when precompiling a C header file with a `.h' extension
1533 for use in C++ compilations. On many systems, `g++' is also installed
1534 with the name `c++'.
1536 When you compile C++ programs, you may specify many of the same
1537 command-line options that you use for compiling programs in any
1538 language; or command-line options meaningful for C and related
1539 languages; or options that are meaningful only for C++ programs. *Note
1540 Options Controlling C Dialect: C Dialect Options, for explanations of
1541 options for languages related to C. *Note Options Controlling C++
1542 Dialect: C++ Dialect Options, for explanations of options that are
1543 meaningful only for C++ programs.
1546 File: gcc.info, Node: C Dialect Options, Next: C++ Dialect Options, Prev: Invoking G++, Up: Invoking GCC
1548 3.4 Options Controlling C Dialect
1549 =================================
1551 The following options control the dialect of C (or languages derived
1552 from C, such as C++, Objective-C and Objective-C++) that the compiler
1556 In C mode, this is equivalent to `-std=c89'. In C++ mode, it is
1557 equivalent to `-std=c++98'.
1559 This turns off certain features of GCC that are incompatible with
1560 ISO C90 (when compiling C code), or of standard C++ (when
1561 compiling C++ code), such as the `asm' and `typeof' keywords, and
1562 predefined macros such as `unix' and `vax' that identify the type
1563 of system you are using. It also enables the undesirable and
1564 rarely used ISO trigraph feature. For the C compiler, it disables
1565 recognition of C++ style `//' comments as well as the `inline'
1568 The alternate keywords `__asm__', `__extension__', `__inline__'
1569 and `__typeof__' continue to work despite `-ansi'. You would not
1570 want to use them in an ISO C program, of course, but it is useful
1571 to put them in header files that might be included in compilations
1572 done with `-ansi'. Alternate predefined macros such as `__unix__'
1573 and `__vax__' are also available, with or without `-ansi'.
1575 The `-ansi' option does not cause non-ISO programs to be rejected
1576 gratuitously. For that, `-pedantic' is required in addition to
1577 `-ansi'. *Note Warning Options::.
1579 The macro `__STRICT_ANSI__' is predefined when the `-ansi' option
1580 is used. Some header files may notice this macro and refrain from
1581 declaring certain functions or defining certain macros that the
1582 ISO standard doesn't call for; this is to avoid interfering with
1583 any programs that might use these names for other things.
1585 Functions that would normally be built in but do not have semantics
1586 defined by ISO C (such as `alloca' and `ffs') are not built-in
1587 functions when `-ansi' is used. *Note Other built-in functions
1588 provided by GCC: Other Builtins, for details of the functions
1592 Determine the language standard. *Note Language Standards
1593 Supported by GCC: Standards, for details of these standard
1594 versions. This option is currently only supported when compiling
1597 The compiler can accept several base standards, such as `c89' or
1598 `c++98', and GNU dialects of those standards, such as `gnu89' or
1599 `gnu++98'. By specifying a base standard, the compiler will
1600 accept all programs following that standard and those using GNU
1601 extensions that do not contradict it. For example, `-std=c89'
1602 turns off certain features of GCC that are incompatible with ISO
1603 C90, such as the `asm' and `typeof' keywords, but not other GNU
1604 extensions that do not have a meaning in ISO C90, such as omitting
1605 the middle term of a `?:' expression. On the other hand, by
1606 specifying a GNU dialect of a standard, all features the compiler
1607 support are enabled, even when those features change the meaning
1608 of the base standard and some strict-conforming programs may be
1609 rejected. The particular standard is used by `-pedantic' to
1610 identify which features are GNU extensions given that version of
1611 the standard. For example `-std=gnu89 -pedantic' would warn about
1612 C++ style `//' comments, while `-std=gnu99 -pedantic' would not.
1614 A value for this option must be provided; possible values are
1618 Support all ISO C90 programs (certain GNU extensions that
1619 conflict with ISO C90 are disabled). Same as `-ansi' for C
1623 ISO C90 as modified in amendment 1.
1629 ISO C99. Note that this standard is not yet fully supported;
1630 see `http://gcc.gnu.org/gcc-4.4/c99status.html' for more
1631 information. The names `c9x' and `iso9899:199x' are
1635 GNU dialect of ISO C90 (including some C99 features). This is
1636 the default for C code.
1640 GNU dialect of ISO C99. When ISO C99 is fully implemented in
1641 GCC, this will become the default. The name `gnu9x' is
1645 The 1998 ISO C++ standard plus amendments. Same as `-ansi' for
1649 GNU dialect of `-std=c++98'. This is the default for C++
1653 The working draft of the upcoming ISO C++0x standard. This
1654 option enables experimental features that are likely to be
1655 included in C++0x. The working draft is constantly changing,
1656 and any feature that is enabled by this flag may be removed
1657 from future versions of GCC if it is not part of the C++0x
1661 GNU dialect of `-std=c++0x'. This option enables experimental
1662 features that may be removed in future versions of GCC.
1665 The option `-fgnu89-inline' tells GCC to use the traditional GNU
1666 semantics for `inline' functions when in C99 mode. *Note An
1667 Inline Function is As Fast As a Macro: Inline. This option is
1668 accepted and ignored by GCC versions 4.1.3 up to but not including
1669 4.3. In GCC versions 4.3 and later it changes the behavior of GCC
1670 in C99 mode. Using this option is roughly equivalent to adding the
1671 `gnu_inline' function attribute to all inline functions (*note
1672 Function Attributes::).
1674 The option `-fno-gnu89-inline' explicitly tells GCC to use the C99
1675 semantics for `inline' when in C99 or gnu99 mode (i.e., it
1676 specifies the default behavior). This option was first supported
1677 in GCC 4.3. This option is not supported in C89 or gnu89 mode.
1679 The preprocessor macros `__GNUC_GNU_INLINE__' and
1680 `__GNUC_STDC_INLINE__' may be used to check which semantics are in
1681 effect for `inline' functions. *Note Common Predefined Macros:
1682 (cpp)Common Predefined Macros.
1684 `-aux-info FILENAME'
1685 Output to the given filename prototyped declarations for all
1686 functions declared and/or defined in a translation unit, including
1687 those in header files. This option is silently ignored in any
1688 language other than C.
1690 Besides declarations, the file indicates, in comments, the origin
1691 of each declaration (source file and line), whether the
1692 declaration was implicit, prototyped or unprototyped (`I', `N' for
1693 new or `O' for old, respectively, in the first character after the
1694 line number and the colon), and whether it came from a declaration
1695 or a definition (`C' or `F', respectively, in the following
1696 character). In the case of function definitions, a K&R-style list
1697 of arguments followed by their declarations is also provided,
1698 inside comments, after the declaration.
1701 Do not recognize `asm', `inline' or `typeof' as a keyword, so that
1702 code can use these words as identifiers. You can use the keywords
1703 `__asm__', `__inline__' and `__typeof__' instead. `-ansi' implies
1706 In C++, this switch only affects the `typeof' keyword, since `asm'
1707 and `inline' are standard keywords. You may want to use the
1708 `-fno-gnu-keywords' flag instead, which has the same effect. In
1709 C99 mode (`-std=c99' or `-std=gnu99'), this switch only affects
1710 the `asm' and `typeof' keywords, since `inline' is a standard
1714 `-fno-builtin-FUNCTION'
1715 Don't recognize built-in functions that do not begin with
1716 `__builtin_' as prefix. *Note Other built-in functions provided
1717 by GCC: Other Builtins, for details of the functions affected,
1718 including those which are not built-in functions when `-ansi' or
1719 `-std' options for strict ISO C conformance are used because they
1720 do not have an ISO standard meaning.
1722 GCC normally generates special code to handle certain built-in
1723 functions more efficiently; for instance, calls to `alloca' may
1724 become single instructions that adjust the stack directly, and
1725 calls to `memcpy' may become inline copy loops. The resulting
1726 code is often both smaller and faster, but since the function
1727 calls no longer appear as such, you cannot set a breakpoint on
1728 those calls, nor can you change the behavior of the functions by
1729 linking with a different library. In addition, when a function is
1730 recognized as a built-in function, GCC may use information about
1731 that function to warn about problems with calls to that function,
1732 or to generate more efficient code, even if the resulting code
1733 still contains calls to that function. For example, warnings are
1734 given with `-Wformat' for bad calls to `printf', when `printf' is
1735 built in, and `strlen' is known not to modify global memory.
1737 With the `-fno-builtin-FUNCTION' option only the built-in function
1738 FUNCTION is disabled. FUNCTION must not begin with `__builtin_'.
1739 If a function is named that is not built-in in this version of
1740 GCC, this option is ignored. There is no corresponding
1741 `-fbuiltin-FUNCTION' option; if you wish to enable built-in
1742 functions selectively when using `-fno-builtin' or
1743 `-ffreestanding', you may define macros such as:
1745 #define abs(n) __builtin_abs ((n))
1746 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1749 Assert that compilation takes place in a hosted environment. This
1750 implies `-fbuiltin'. A hosted environment is one in which the
1751 entire standard library is available, and in which `main' has a
1752 return type of `int'. Examples are nearly everything except a
1753 kernel. This is equivalent to `-fno-freestanding'.
1756 Assert that compilation takes place in a freestanding environment.
1757 This implies `-fno-builtin'. A freestanding environment is one
1758 in which the standard library may not exist, and program startup
1759 may not necessarily be at `main'. The most obvious example is an
1760 OS kernel. This is equivalent to `-fno-hosted'.
1762 *Note Language Standards Supported by GCC: Standards, for details
1763 of freestanding and hosted environments.
1766 Enable handling of OpenMP directives `#pragma omp' in C/C++ and
1767 `!$omp' in Fortran. When `-fopenmp' is specified, the compiler
1768 generates parallel code according to the OpenMP Application
1769 Program Interface v2.5 `http://www.openmp.org/'. This option
1770 implies `-pthread', and thus is only supported on targets that
1771 have support for `-pthread'.
1774 Accept some non-standard constructs used in Microsoft header files.
1776 Some cases of unnamed fields in structures and unions are only
1777 accepted with this option. *Note Unnamed struct/union fields
1778 within structs/unions: Unnamed Fields, for details.
1781 Support ISO C trigraphs. The `-ansi' option (and `-std' options
1782 for strict ISO C conformance) implies `-trigraphs'.
1784 `-no-integrated-cpp'
1785 Performs a compilation in two passes: preprocessing and compiling.
1786 This option allows a user supplied "cc1", "cc1plus", or "cc1obj"
1787 via the `-B' option. The user supplied compilation step can then
1788 add in an additional preprocessing step after normal preprocessing
1789 but before compiling. The default is to use the integrated cpp
1792 The semantics of this option will change if "cc1", "cc1plus", and
1793 "cc1obj" are merged.
1797 Formerly, these options caused GCC to attempt to emulate a
1798 pre-standard C compiler. They are now only supported with the
1799 `-E' switch. The preprocessor continues to support a pre-standard
1800 mode. See the GNU CPP manual for details.
1803 Allow conditional expressions with mismatched types in the second
1804 and third arguments. The value of such an expression is void.
1805 This option is not supported for C++.
1807 `-flax-vector-conversions'
1808 Allow implicit conversions between vectors with differing numbers
1809 of elements and/or incompatible element types. This option should
1810 not be used for new code.
1813 Let the type `char' be unsigned, like `unsigned char'.
1815 Each kind of machine has a default for what `char' should be. It
1816 is either like `unsigned char' by default or like `signed char' by
1819 Ideally, a portable program should always use `signed char' or
1820 `unsigned char' when it depends on the signedness of an object.
1821 But many programs have been written to use plain `char' and expect
1822 it to be signed, or expect it to be unsigned, depending on the
1823 machines they were written for. This option, and its inverse, let
1824 you make such a program work with the opposite default.
1826 The type `char' is always a distinct type from each of `signed
1827 char' or `unsigned char', even though its behavior is always just
1828 like one of those two.
1831 Let the type `char' be signed, like `signed char'.
1833 Note that this is equivalent to `-fno-unsigned-char', which is the
1834 negative form of `-funsigned-char'. Likewise, the option
1835 `-fno-signed-char' is equivalent to `-funsigned-char'.
1837 `-fsigned-bitfields'
1838 `-funsigned-bitfields'
1839 `-fno-signed-bitfields'
1840 `-fno-unsigned-bitfields'
1841 These options control whether a bit-field is signed or unsigned,
1842 when the declaration does not use either `signed' or `unsigned'.
1843 By default, such a bit-field is signed, because this is
1844 consistent: the basic integer types such as `int' are signed types.
1847 File: gcc.info, Node: C++ Dialect Options, Next: Objective-C and Objective-C++ Dialect Options, Prev: C Dialect Options, Up: Invoking GCC
1849 3.5 Options Controlling C++ Dialect
1850 ===================================
1852 This section describes the command-line options that are only meaningful
1853 for C++ programs; but you can also use most of the GNU compiler options
1854 regardless of what language your program is in. For example, you might
1855 compile a file `firstClass.C' like this:
1857 g++ -g -frepo -O -c firstClass.C
1859 In this example, only `-frepo' is an option meant only for C++
1860 programs; you can use the other options with any language supported by
1863 Here is a list of options that are _only_ for compiling C++ programs:
1866 Use version N of the C++ ABI. Version 2 is the version of the C++
1867 ABI that first appeared in G++ 3.4. Version 1 is the version of
1868 the C++ ABI that first appeared in G++ 3.2. Version 0 will always
1869 be the version that conforms most closely to the C++ ABI
1870 specification. Therefore, the ABI obtained using version 0 will
1871 change as ABI bugs are fixed.
1873 The default is version 2.
1875 `-fno-access-control'
1876 Turn off all access checking. This switch is mainly useful for
1877 working around bugs in the access control code.
1880 Check that the pointer returned by `operator new' is non-null
1881 before attempting to modify the storage allocated. This check is
1882 normally unnecessary because the C++ standard specifies that
1883 `operator new' will only return `0' if it is declared `throw()',
1884 in which case the compiler will always check the return value even
1885 without this option. In all other cases, when `operator new' has
1886 a non-empty exception specification, memory exhaustion is
1887 signalled by throwing `std::bad_alloc'. See also `new (nothrow)'.
1890 Put uninitialized or runtime-initialized global variables into the
1891 common segment, as C does. This saves space in the executable at
1892 the cost of not diagnosing duplicate definitions. If you compile
1893 with this flag and your program mysteriously crashes after
1894 `main()' has completed, you may have an object that is being
1895 destroyed twice because two definitions were merged.
1897 This option is no longer useful on most targets, now that support
1898 has been added for putting variables into BSS without making them
1901 `-ffriend-injection'
1902 Inject friend functions into the enclosing namespace, so that they
1903 are visible outside the scope of the class in which they are
1904 declared. Friend functions were documented to work this way in
1905 the old Annotated C++ Reference Manual, and versions of G++ before
1906 4.1 always worked that way. However, in ISO C++ a friend function
1907 which is not declared in an enclosing scope can only be found
1908 using argument dependent lookup. This option causes friends to be
1909 injected as they were in earlier releases.
1911 This option is for compatibility, and may be removed in a future
1914 `-fno-elide-constructors'
1915 The C++ standard allows an implementation to omit creating a
1916 temporary which is only used to initialize another object of the
1917 same type. Specifying this option disables that optimization, and
1918 forces G++ to call the copy constructor in all cases.
1920 `-fno-enforce-eh-specs'
1921 Don't generate code to check for violation of exception
1922 specifications at runtime. This option violates the C++ standard,
1923 but may be useful for reducing code size in production builds,
1924 much like defining `NDEBUG'. This does not give user code
1925 permission to throw exceptions in violation of the exception
1926 specifications; the compiler will still optimize based on the
1927 specifications, so throwing an unexpected exception will result in
1932 If `-ffor-scope' is specified, the scope of variables declared in
1933 a for-init-statement is limited to the `for' loop itself, as
1934 specified by the C++ standard. If `-fno-for-scope' is specified,
1935 the scope of variables declared in a for-init-statement extends to
1936 the end of the enclosing scope, as was the case in old versions of
1937 G++, and other (traditional) implementations of C++.
1939 The default if neither flag is given to follow the standard, but
1940 to allow and give a warning for old-style code that would
1941 otherwise be invalid, or have different behavior.
1944 Do not recognize `typeof' as a keyword, so that code can use this
1945 word as an identifier. You can use the keyword `__typeof__'
1946 instead. `-ansi' implies `-fno-gnu-keywords'.
1948 `-fno-implicit-templates'
1949 Never emit code for non-inline templates which are instantiated
1950 implicitly (i.e. by use); only emit code for explicit
1951 instantiations. *Note Template Instantiation::, for more
1954 `-fno-implicit-inline-templates'
1955 Don't emit code for implicit instantiations of inline templates,
1956 either. The default is to handle inlines differently so that
1957 compiles with and without optimization will need the same set of
1958 explicit instantiations.
1960 `-fno-implement-inlines'
1961 To save space, do not emit out-of-line copies of inline functions
1962 controlled by `#pragma implementation'. This will cause linker
1963 errors if these functions are not inlined everywhere they are
1967 Disable pedantic warnings about constructs used in MFC, such as
1968 implicit int and getting a pointer to member function via
1969 non-standard syntax.
1971 `-fno-nonansi-builtins'
1972 Disable built-in declarations of functions that are not mandated by
1973 ANSI/ISO C. These include `ffs', `alloca', `_exit', `index',
1974 `bzero', `conjf', and other related functions.
1976 `-fno-operator-names'
1977 Do not treat the operator name keywords `and', `bitand', `bitor',
1978 `compl', `not', `or' and `xor' as synonyms as keywords.
1980 `-fno-optional-diags'
1981 Disable diagnostics that the standard says a compiler does not
1982 need to issue. Currently, the only such diagnostic issued by G++
1983 is the one for a name having multiple meanings within a class.
1986 Downgrade some diagnostics about nonconformant code from errors to
1987 warnings. Thus, using `-fpermissive' will allow some
1988 nonconforming code to compile.
1991 Enable automatic template instantiation at link time. This option
1992 also implies `-fno-implicit-templates'. *Note Template
1993 Instantiation::, for more information.
1996 Disable generation of information about every class with virtual
1997 functions for use by the C++ runtime type identification features
1998 (`dynamic_cast' and `typeid'). If you don't use those parts of
1999 the language, you can save some space by using this flag. Note
2000 that exception handling uses the same information, but it will
2001 generate it as needed. The `dynamic_cast' operator can still be
2002 used for casts that do not require runtime type information, i.e.
2003 casts to `void *' or to unambiguous base classes.
2006 Emit statistics about front-end processing at the end of the
2007 compilation. This information is generally only useful to the G++
2010 `-ftemplate-depth-N'
2011 Set the maximum instantiation depth for template classes to N. A
2012 limit on the template instantiation depth is needed to detect
2013 endless recursions during template class instantiation. ANSI/ISO
2014 C++ conforming programs must not rely on a maximum depth greater
2017 `-fno-threadsafe-statics'
2018 Do not emit the extra code to use the routines specified in the C++
2019 ABI for thread-safe initialization of local statics. You can use
2020 this option to reduce code size slightly in code that doesn't need
2024 Register destructors for objects with static storage duration with
2025 the `__cxa_atexit' function rather than the `atexit' function.
2026 This option is required for fully standards-compliant handling of
2027 static destructors, but will only work if your C library supports
2030 `-fno-use-cxa-get-exception-ptr'
2031 Don't use the `__cxa_get_exception_ptr' runtime routine. This
2032 will cause `std::uncaught_exception' to be incorrect, but is
2033 necessary if the runtime routine is not available.
2035 `-fvisibility-inlines-hidden'
2036 This switch declares that the user does not attempt to compare
2037 pointers to inline methods where the addresses of the two functions
2038 were taken in different shared objects.
2040 The effect of this is that GCC may, effectively, mark inline
2041 methods with `__attribute__ ((visibility ("hidden")))' so that
2042 they do not appear in the export table of a DSO and do not require
2043 a PLT indirection when used within the DSO. Enabling this option
2044 can have a dramatic effect on load and link times of a DSO as it
2045 massively reduces the size of the dynamic export table when the
2046 library makes heavy use of templates.
2048 The behavior of this switch is not quite the same as marking the
2049 methods as hidden directly, because it does not affect static
2050 variables local to the function or cause the compiler to deduce
2051 that the function is defined in only one shared object.
2053 You may mark a method as having a visibility explicitly to negate
2054 the effect of the switch for that method. For example, if you do
2055 want to compare pointers to a particular inline method, you might
2056 mark it as having default visibility. Marking the enclosing class
2057 with explicit visibility will have no effect.
2059 Explicitly instantiated inline methods are unaffected by this
2060 option as their linkage might otherwise cross a shared library
2061 boundary. *Note Template Instantiation::.
2063 `-fvisibility-ms-compat'
2064 This flag attempts to use visibility settings to make GCC's C++
2065 linkage model compatible with that of Microsoft Visual Studio.
2067 The flag makes these changes to GCC's linkage model:
2069 1. It sets the default visibility to `hidden', like
2070 `-fvisibility=hidden'.
2072 2. Types, but not their members, are not hidden by default.
2074 3. The One Definition Rule is relaxed for types without explicit
2075 visibility specifications which are defined in more than one
2076 different shared object: those declarations are permitted if
2077 they would have been permitted when this option was not used.
2079 In new code it is better to use `-fvisibility=hidden' and export
2080 those classes which are intended to be externally visible.
2081 Unfortunately it is possible for code to rely, perhaps
2082 accidentally, on the Visual Studio behavior.
2084 Among the consequences of these changes are that static data
2085 members of the same type with the same name but defined in
2086 different shared objects will be different, so changing one will
2087 not change the other; and that pointers to function members
2088 defined in different shared objects may not compare equal. When
2089 this flag is given, it is a violation of the ODR to define types
2090 with the same name differently.
2093 Do not use weak symbol support, even if it is provided by the
2094 linker. By default, G++ will use weak symbols if they are
2095 available. This option exists only for testing, and should not be
2096 used by end-users; it will result in inferior code and has no
2097 benefits. This option may be removed in a future release of G++.
2100 Do not search for header files in the standard directories
2101 specific to C++, but do still search the other standard
2102 directories. (This option is used when building the C++ library.)
2104 In addition, these optimization, warning, and code generation options
2105 have meanings only for C++ programs:
2107 `-fno-default-inline'
2108 Do not assume `inline' for functions defined inside a class scope.
2109 *Note Options That Control Optimization: Optimize Options. Note
2110 that these functions will have linkage like inline functions; they
2111 just won't be inlined by default.
2113 `-Wabi (C, Objective-C, C++ and Objective-C++ only)'
2114 Warn when G++ generates code that is probably not compatible with
2115 the vendor-neutral C++ ABI. Although an effort has been made to
2116 warn about all such cases, there are probably some cases that are
2117 not warned about, even though G++ is generating incompatible code.
2118 There may also be cases where warnings are emitted even though
2119 the code that is generated will be compatible.
2121 You should rewrite your code to avoid these warnings if you are
2122 concerned about the fact that code generated by G++ may not be
2123 binary compatible with code generated by other compilers.
2125 The known incompatibilities at this point include:
2127 * Incorrect handling of tail-padding for bit-fields. G++ may
2128 attempt to pack data into the same byte as a base class. For
2131 struct A { virtual void f(); int f1 : 1; };
2132 struct B : public A { int f2 : 1; };
2134 In this case, G++ will place `B::f2' into the same byte
2135 as`A::f1'; other compilers will not. You can avoid this
2136 problem by explicitly padding `A' so that its size is a
2137 multiple of the byte size on your platform; that will cause
2138 G++ and other compilers to layout `B' identically.
2140 * Incorrect handling of tail-padding for virtual bases. G++
2141 does not use tail padding when laying out virtual bases. For
2144 struct A { virtual void f(); char c1; };
2145 struct B { B(); char c2; };
2146 struct C : public A, public virtual B {};
2148 In this case, G++ will not place `B' into the tail-padding for
2149 `A'; other compilers will. You can avoid this problem by
2150 explicitly padding `A' so that its size is a multiple of its
2151 alignment (ignoring virtual base classes); that will cause
2152 G++ and other compilers to layout `C' identically.
2154 * Incorrect handling of bit-fields with declared widths greater
2155 than that of their underlying types, when the bit-fields
2156 appear in a union. For example:
2158 union U { int i : 4096; };
2160 Assuming that an `int' does not have 4096 bits, G++ will make
2161 the union too small by the number of bits in an `int'.
2163 * Empty classes can be placed at incorrect offsets. For
2173 struct C : public B, public A {};
2175 G++ will place the `A' base class of `C' at a nonzero offset;
2176 it should be placed at offset zero. G++ mistakenly believes
2177 that the `A' data member of `B' is already at offset zero.
2179 * Names of template functions whose types involve `typename' or
2180 template template parameters can be mangled incorrectly.
2182 template <typename Q>
2183 void f(typename Q::X) {}
2185 template <template <typename> class Q>
2186 void f(typename Q<int>::X) {}
2188 Instantiations of these templates may be mangled incorrectly.
2191 It also warns psABI related changes. The known psABI changes at
2194 * For SYSV/x86-64, when passing union with long double, it is
2195 changed to pass in memory as specified in psABI. For example:
2202 `union U' will always be passed in memory.
2205 `-Wctor-dtor-privacy (C++ and Objective-C++ only)'
2206 Warn when a class seems unusable because all the constructors or
2207 destructors in that class are private, and it has neither friends
2208 nor public static member functions.
2210 `-Wnon-virtual-dtor (C++ and Objective-C++ only)'
2211 Warn when a class has virtual functions and accessible non-virtual
2212 destructor, in which case it would be possible but unsafe to delete
2213 an instance of a derived class through a pointer to the base class.
2214 This warning is also enabled if -Weffc++ is specified.
2216 `-Wreorder (C++ and Objective-C++ only)'
2217 Warn when the order of member initializers given in the code does
2218 not match the order in which they must be executed. For instance:
2223 A(): j (0), i (1) { }
2226 The compiler will rearrange the member initializers for `i' and
2227 `j' to match the declaration order of the members, emitting a
2228 warning to that effect. This warning is enabled by `-Wall'.
2230 The following `-W...' options are not affected by `-Wall'.
2232 `-Weffc++ (C++ and Objective-C++ only)'
2233 Warn about violations of the following style guidelines from Scott
2234 Meyers' `Effective C++' book:
2236 * Item 11: Define a copy constructor and an assignment
2237 operator for classes with dynamically allocated memory.
2239 * Item 12: Prefer initialization to assignment in constructors.
2241 * Item 14: Make destructors virtual in base classes.
2243 * Item 15: Have `operator=' return a reference to `*this'.
2245 * Item 23: Don't try to return a reference when you must
2249 Also warn about violations of the following style guidelines from
2250 Scott Meyers' `More Effective C++' book:
2252 * Item 6: Distinguish between prefix and postfix forms of
2253 increment and decrement operators.
2255 * Item 7: Never overload `&&', `||', or `,'.
2258 When selecting this option, be aware that the standard library
2259 headers do not obey all of these guidelines; use `grep -v' to
2260 filter out those warnings.
2262 `-Wstrict-null-sentinel (C++ and Objective-C++ only)'
2263 Warn also about the use of an uncasted `NULL' as sentinel. When
2264 compiling only with GCC this is a valid sentinel, as `NULL' is
2265 defined to `__null'. Although it is a null pointer constant not a
2266 null pointer, it is guaranteed to be of the same size as a
2267 pointer. But this use is not portable across different compilers.
2269 `-Wno-non-template-friend (C++ and Objective-C++ only)'
2270 Disable warnings when non-templatized friend functions are declared
2271 within a template. Since the advent of explicit template
2272 specification support in G++, if the name of the friend is an
2273 unqualified-id (i.e., `friend foo(int)'), the C++ language
2274 specification demands that the friend declare or define an
2275 ordinary, nontemplate function. (Section 14.5.3). Before G++
2276 implemented explicit specification, unqualified-ids could be
2277 interpreted as a particular specialization of a templatized
2278 function. Because this non-conforming behavior is no longer the
2279 default behavior for G++, `-Wnon-template-friend' allows the
2280 compiler to check existing code for potential trouble spots and is
2281 on by default. This new compiler behavior can be turned off with
2282 `-Wno-non-template-friend' which keeps the conformant compiler code
2283 but disables the helpful warning.
2285 `-Wold-style-cast (C++ and Objective-C++ only)'
2286 Warn if an old-style (C-style) cast to a non-void type is used
2287 within a C++ program. The new-style casts (`dynamic_cast',
2288 `static_cast', `reinterpret_cast', and `const_cast') are less
2289 vulnerable to unintended effects and much easier to search for.
2291 `-Woverloaded-virtual (C++ and Objective-C++ only)'
2292 Warn when a function declaration hides virtual functions from a
2293 base class. For example, in:
2299 struct B: public A {
2303 the `A' class version of `f' is hidden in `B', and code like:
2308 will fail to compile.
2310 `-Wno-pmf-conversions (C++ and Objective-C++ only)'
2311 Disable the diagnostic for converting a bound pointer to member
2312 function to a plain pointer.
2314 `-Wsign-promo (C++ and Objective-C++ only)'
2315 Warn when overload resolution chooses a promotion from unsigned or
2316 enumerated type to a signed type, over a conversion to an unsigned
2317 type of the same size. Previous versions of G++ would try to
2318 preserve unsignedness, but the standard mandates the current
2323 A& operator = (int);
2332 In this example, G++ will synthesize a default `A& operator =
2333 (const A&);', while cfront will use the user-defined `operator ='.
2336 File: gcc.info, Node: Objective-C and Objective-C++ Dialect Options, Next: Language Independent Options, Prev: C++ Dialect Options, Up: Invoking GCC
2338 3.6 Options Controlling Objective-C and Objective-C++ Dialects
2339 ==============================================================
2341 (NOTE: This manual does not describe the Objective-C and Objective-C++
2342 languages themselves. See *Note Language Standards Supported by GCC:
2343 Standards, for references.)
2345 This section describes the command-line options that are only
2346 meaningful for Objective-C and Objective-C++ programs, but you can also
2347 use most of the language-independent GNU compiler options. For
2348 example, you might compile a file `some_class.m' like this:
2350 gcc -g -fgnu-runtime -O -c some_class.m
2352 In this example, `-fgnu-runtime' is an option meant only for
2353 Objective-C and Objective-C++ programs; you can use the other options
2354 with any language supported by GCC.
2356 Note that since Objective-C is an extension of the C language,
2357 Objective-C compilations may also use options specific to the C
2358 front-end (e.g., `-Wtraditional'). Similarly, Objective-C++
2359 compilations may use C++-specific options (e.g., `-Wabi').
2361 Here is a list of options that are _only_ for compiling Objective-C
2362 and Objective-C++ programs:
2364 `-fconstant-string-class=CLASS-NAME'
2365 Use CLASS-NAME as the name of the class to instantiate for each
2366 literal string specified with the syntax `@"..."'. The default
2367 class name is `NXConstantString' if the GNU runtime is being used,
2368 and `NSConstantString' if the NeXT runtime is being used (see
2369 below). The `-fconstant-cfstrings' option, if also present, will
2370 override the `-fconstant-string-class' setting and cause `@"..."'
2371 literals to be laid out as constant CoreFoundation strings.
2374 Generate object code compatible with the standard GNU Objective-C
2375 runtime. This is the default for most types of systems.
2378 Generate output compatible with the NeXT runtime. This is the
2379 default for NeXT-based systems, including Darwin and Mac OS X.
2380 The macro `__NEXT_RUNTIME__' is predefined if (and only if) this
2383 `-fno-nil-receivers'
2384 Assume that all Objective-C message dispatches (e.g., `[receiver
2385 message:arg]') in this translation unit ensure that the receiver
2386 is not `nil'. This allows for more efficient entry points in the
2387 runtime to be used. Currently, this option is only available in
2388 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2390 `-fobjc-call-cxx-cdtors'
2391 For each Objective-C class, check if any of its instance variables
2392 is a C++ object with a non-trivial default constructor. If so,
2393 synthesize a special `- (id) .cxx_construct' instance method that
2394 will run non-trivial default constructors on any such instance
2395 variables, in order, and then return `self'. Similarly, check if
2396 any instance variable is a C++ object with a non-trivial
2397 destructor, and if so, synthesize a special `- (void)
2398 .cxx_destruct' method that will run all such default destructors,
2401 The `- (id) .cxx_construct' and/or `- (void) .cxx_destruct' methods
2402 thusly generated will only operate on instance variables declared
2403 in the current Objective-C class, and not those inherited from
2404 superclasses. It is the responsibility of the Objective-C runtime
2405 to invoke all such methods in an object's inheritance hierarchy.
2406 The `- (id) .cxx_construct' methods will be invoked by the runtime
2407 immediately after a new object instance is allocated; the `-
2408 (void) .cxx_destruct' methods will be invoked immediately before
2409 the runtime deallocates an object instance.
2411 As of this writing, only the NeXT runtime on Mac OS X 10.4 and
2412 later has support for invoking the `- (id) .cxx_construct' and `-
2413 (void) .cxx_destruct' methods.
2415 `-fobjc-direct-dispatch'
2416 Allow fast jumps to the message dispatcher. On Darwin this is
2417 accomplished via the comm page.
2420 Enable syntactic support for structured exception handling in
2421 Objective-C, similar to what is offered by C++ and Java. This
2422 option is unavailable in conjunction with the NeXT runtime on Mac
2423 OS X 10.2 and earlier.
2430 @catch (AnObjCClass *exc) {
2437 @catch (AnotherClass *exc) {
2440 @catch (id allOthers) {
2449 The `@throw' statement may appear anywhere in an Objective-C or
2450 Objective-C++ program; when used inside of a `@catch' block, the
2451 `@throw' may appear without an argument (as shown above), in which
2452 case the object caught by the `@catch' will be rethrown.
2454 Note that only (pointers to) Objective-C objects may be thrown and
2455 caught using this scheme. When an object is thrown, it will be
2456 caught by the nearest `@catch' clause capable of handling objects
2457 of that type, analogously to how `catch' blocks work in C++ and
2458 Java. A `@catch(id ...)' clause (as shown above) may also be
2459 provided to catch any and all Objective-C exceptions not caught by
2460 previous `@catch' clauses (if any).
2462 The `@finally' clause, if present, will be executed upon exit from
2463 the immediately preceding `@try ... @catch' section. This will
2464 happen regardless of whether any exceptions are thrown, caught or
2465 rethrown inside the `@try ... @catch' section, analogously to the
2466 behavior of the `finally' clause in Java.
2468 There are several caveats to using the new exception mechanism:
2470 * Although currently designed to be binary compatible with
2471 `NS_HANDLER'-style idioms provided by the `NSException'
2472 class, the new exceptions can only be used on Mac OS X 10.3
2473 (Panther) and later systems, due to additional functionality
2474 needed in the (NeXT) Objective-C runtime.
2476 * As mentioned above, the new exceptions do not support handling
2477 types other than Objective-C objects. Furthermore, when
2478 used from Objective-C++, the Objective-C exception model does
2479 not interoperate with C++ exceptions at this time. This
2480 means you cannot `@throw' an exception from Objective-C and
2481 `catch' it in C++, or vice versa (i.e., `throw ... @catch').
2483 The `-fobjc-exceptions' switch also enables the use of
2484 synchronization blocks for thread-safe execution:
2486 @synchronized (ObjCClass *guard) {
2490 Upon entering the `@synchronized' block, a thread of execution
2491 shall first check whether a lock has been placed on the
2492 corresponding `guard' object by another thread. If it has, the
2493 current thread shall wait until the other thread relinquishes its
2494 lock. Once `guard' becomes available, the current thread will
2495 place its own lock on it, execute the code contained in the
2496 `@synchronized' block, and finally relinquish the lock (thereby
2497 making `guard' available to other threads).
2499 Unlike Java, Objective-C does not allow for entire methods to be
2500 marked `@synchronized'. Note that throwing exceptions out of
2501 `@synchronized' blocks is allowed, and will cause the guarding
2502 object to be unlocked properly.
2505 Enable garbage collection (GC) in Objective-C and Objective-C++
2508 `-freplace-objc-classes'
2509 Emit a special marker instructing `ld(1)' not to statically link in
2510 the resulting object file, and allow `dyld(1)' to load it in at
2511 run time instead. This is used in conjunction with the
2512 Fix-and-Continue debugging mode, where the object file in question
2513 may be recompiled and dynamically reloaded in the course of
2514 program execution, without the need to restart the program itself.
2515 Currently, Fix-and-Continue functionality is only available in
2516 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2519 When compiling for the NeXT runtime, the compiler ordinarily
2520 replaces calls to `objc_getClass("...")' (when the name of the
2521 class is known at compile time) with static class references that
2522 get initialized at load time, which improves run-time performance.
2523 Specifying the `-fzero-link' flag suppresses this behavior and
2524 causes calls to `objc_getClass("...")' to be retained. This is
2525 useful in Zero-Link debugging mode, since it allows for individual
2526 class implementations to be modified during program execution.
2529 Dump interface declarations for all classes seen in the source
2530 file to a file named `SOURCENAME.decl'.
2532 `-Wassign-intercept (Objective-C and Objective-C++ only)'
2533 Warn whenever an Objective-C assignment is being intercepted by the
2536 `-Wno-protocol (Objective-C and Objective-C++ only)'
2537 If a class is declared to implement a protocol, a warning is
2538 issued for every method in the protocol that is not implemented by
2539 the class. The default behavior is to issue a warning for every
2540 method not explicitly implemented in the class, even if a method
2541 implementation is inherited from the superclass. If you use the
2542 `-Wno-protocol' option, then methods inherited from the superclass
2543 are considered to be implemented, and no warning is issued for
2546 `-Wselector (Objective-C and Objective-C++ only)'
2547 Warn if multiple methods of different types for the same selector
2548 are found during compilation. The check is performed on the list
2549 of methods in the final stage of compilation. Additionally, a
2550 check is performed for each selector appearing in a
2551 `@selector(...)' expression, and a corresponding method for that
2552 selector has been found during compilation. Because these checks
2553 scan the method table only at the end of compilation, these
2554 warnings are not produced if the final stage of compilation is not
2555 reached, for example because an error is found during compilation,
2556 or because the `-fsyntax-only' option is being used.
2558 `-Wstrict-selector-match (Objective-C and Objective-C++ only)'
2559 Warn if multiple methods with differing argument and/or return
2560 types are found for a given selector when attempting to send a
2561 message using this selector to a receiver of type `id' or `Class'.
2562 When this flag is off (which is the default behavior), the
2563 compiler will omit such warnings if any differences found are
2564 confined to types which share the same size and alignment.
2566 `-Wundeclared-selector (Objective-C and Objective-C++ only)'
2567 Warn if a `@selector(...)' expression referring to an undeclared
2568 selector is found. A selector is considered undeclared if no
2569 method with that name has been declared before the
2570 `@selector(...)' expression, either explicitly in an `@interface'
2571 or `@protocol' declaration, or implicitly in an `@implementation'
2572 section. This option always performs its checks as soon as a
2573 `@selector(...)' expression is found, while `-Wselector' only
2574 performs its checks in the final stage of compilation. This also
2575 enforces the coding style convention that methods and selectors
2576 must be declared before being used.
2578 `-print-objc-runtime-info'
2579 Generate C header describing the largest structure that is passed
2584 File: gcc.info, Node: Language Independent Options, Next: Warning Options, Prev: Objective-C and Objective-C++ Dialect Options, Up: Invoking GCC
2586 3.7 Options to Control Diagnostic Messages Formatting
2587 =====================================================
2589 Traditionally, diagnostic messages have been formatted irrespective of
2590 the output device's aspect (e.g. its width, ...). The options described
2591 below can be used to control the diagnostic messages formatting
2592 algorithm, e.g. how many characters per line, how often source location
2593 information should be reported. Right now, only the C++ front end can
2594 honor these options. However it is expected, in the near future, that
2595 the remaining front ends would be able to digest them correctly.
2597 `-fmessage-length=N'
2598 Try to format error messages so that they fit on lines of about N
2599 characters. The default is 72 characters for `g++' and 0 for the
2600 rest of the front ends supported by GCC. If N is zero, then no
2601 line-wrapping will be done; each error message will appear on a
2604 `-fdiagnostics-show-location=once'
2605 Only meaningful in line-wrapping mode. Instructs the diagnostic
2606 messages reporter to emit _once_ source location information; that
2607 is, in case the message is too long to fit on a single physical
2608 line and has to be wrapped, the source location won't be emitted
2609 (as prefix) again, over and over, in subsequent continuation
2610 lines. This is the default behavior.
2612 `-fdiagnostics-show-location=every-line'
2613 Only meaningful in line-wrapping mode. Instructs the diagnostic
2614 messages reporter to emit the same source location information (as
2615 prefix) for physical lines that result from the process of breaking
2616 a message which is too long to fit on a single line.
2618 `-fdiagnostics-show-option'
2619 This option instructs the diagnostic machinery to add text to each
2620 diagnostic emitted, which indicates which command line option
2621 directly controls that diagnostic, when such an option is known to
2622 the diagnostic machinery.
2624 `-Wcoverage-mismatch'
2625 Warn if feedback profiles do not match when using the
2626 `-fprofile-use' option. If a source file was changed between
2627 `-fprofile-gen' and `-fprofile-use', the files with the profile
2628 feedback can fail to match the source file and GCC can not use the
2629 profile feedback information. By default, GCC emits an error
2630 message in this case. The option `-Wcoverage-mismatch' emits a
2631 warning instead of an error. GCC does not use appropriate
2632 feedback profiles, so using this option can result in poorly
2633 optimized code. This option is useful only in the case of very
2634 minor changes such as bug fixes to an existing code-base.
2638 File: gcc.info, Node: Warning Options, Next: Debugging Options, Prev: Language Independent Options, Up: Invoking GCC
2640 3.8 Options to Request or Suppress Warnings
2641 ===========================================
2643 Warnings are diagnostic messages that report constructions which are
2644 not inherently erroneous but which are risky or suggest there may have
2647 The following language-independent options do not enable specific
2648 warnings but control the kinds of diagnostics produced by GCC.
2651 Check the code for syntax errors, but don't do anything beyond
2655 Inhibit all warning messages.
2658 Make all warnings into errors.
2661 Make the specified warning into an error. The specifier for a
2662 warning is appended, for example `-Werror=switch' turns the
2663 warnings controlled by `-Wswitch' into errors. This switch takes a
2664 negative form, to be used to negate `-Werror' for specific
2665 warnings, for example `-Wno-error=switch' makes `-Wswitch'
2666 warnings not be errors, even when `-Werror' is in effect. You can
2667 use the `-fdiagnostics-show-option' option to have each
2668 controllable warning amended with the option which controls it, to
2669 determine what to use with this option.
2671 Note that specifying `-Werror='FOO automatically implies `-W'FOO.
2672 However, `-Wno-error='FOO does not imply anything.
2675 This option causes the compiler to abort compilation on the first
2676 error occurred rather than trying to keep going and printing
2677 further error messages.
2680 You can request many specific warnings with options beginning `-W',
2681 for example `-Wimplicit' to request warnings on implicit declarations.
2682 Each of these specific warning options also has a negative form
2683 beginning `-Wno-' to turn off warnings; for example, `-Wno-implicit'.
2684 This manual lists only one of the two forms, whichever is not the
2685 default. For further, language-specific options also refer to *Note
2686 C++ Dialect Options:: and *Note Objective-C and Objective-C++ Dialect
2690 Issue all the warnings demanded by strict ISO C and ISO C++;
2691 reject all programs that use forbidden extensions, and some other
2692 programs that do not follow ISO C and ISO C++. For ISO C, follows
2693 the version of the ISO C standard specified by any `-std' option
2696 Valid ISO C and ISO C++ programs should compile properly with or
2697 without this option (though a rare few will require `-ansi' or a
2698 `-std' option specifying the required version of ISO C). However,
2699 without this option, certain GNU extensions and traditional C and
2700 C++ features are supported as well. With this option, they are
2703 `-pedantic' does not cause warning messages for use of the
2704 alternate keywords whose names begin and end with `__'. Pedantic
2705 warnings are also disabled in the expression that follows
2706 `__extension__'. However, only system header files should use
2707 these escape routes; application programs should avoid them.
2708 *Note Alternate Keywords::.
2710 Some users try to use `-pedantic' to check programs for strict ISO
2711 C conformance. They soon find that it does not do quite what they
2712 want: it finds some non-ISO practices, but not all--only those for
2713 which ISO C _requires_ a diagnostic, and some others for which
2714 diagnostics have been added.
2716 A feature to report any failure to conform to ISO C might be
2717 useful in some instances, but would require considerable
2718 additional work and would be quite different from `-pedantic'. We
2719 don't have plans to support such a feature in the near future.
2721 Where the standard specified with `-std' represents a GNU extended
2722 dialect of C, such as `gnu89' or `gnu99', there is a corresponding
2723 "base standard", the version of ISO C on which the GNU extended
2724 dialect is based. Warnings from `-pedantic' are given where they
2725 are required by the base standard. (It would not make sense for
2726 such warnings to be given only for features not in the specified
2727 GNU C dialect, since by definition the GNU dialects of C include
2728 all features the compiler supports with the given option, and
2729 there would be nothing to warn about.)
2732 Like `-pedantic', except that errors are produced rather than
2736 This enables all the warnings about constructions that some users
2737 consider questionable, and that are easy to avoid (or modify to
2738 prevent the warning), even in conjunction with macros. This also
2739 enables some language-specific warnings described in *Note C++
2740 Dialect Options:: and *Note Objective-C and Objective-C++ Dialect
2743 `-Wall' turns on the following warning flags:
2746 -Warray-bounds (only with `-O2')
2750 -Wimplicit-function-declaration
2753 -Wmain (only for C/ObjC and unless `-ffreestanding')
2761 -Wsign-compare (only in C++)
2772 -Wvolatile-register-var
2774 Note that some warning flags are not implied by `-Wall'. Some of
2775 them warn about constructions that users generally do not consider
2776 questionable, but which occasionally you might wish to check for;
2777 others warn about constructions that are necessary or hard to
2778 avoid in some cases, and there is no simple way to modify the code
2779 to suppress the warning. Some of them are enabled by `-Wextra' but
2780 many of them must be enabled individually.
2783 This enables some extra warning flags that are not enabled by
2784 `-Wall'. (This option used to be called `-W'. The older name is
2785 still supported, but the newer name is more descriptive.)
2789 -Wignored-qualifiers
2790 -Wmissing-field-initializers
2791 -Wmissing-parameter-type (C only)
2792 -Wold-style-declaration (C only)
2797 -Wunused-parameter (only with `-Wunused' or `-Wall')
2799 The option `-Wextra' also prints warning messages for the
2802 * A pointer is compared against integer zero with `<', `<=',
2805 * (C++ only) An enumerator and a non-enumerator both appear in a
2806 conditional expression.
2808 * (C++ only) Ambiguous virtual bases.
2810 * (C++ only) Subscripting an array which has been declared
2813 * (C++ only) Taking the address of a variable which has been
2814 declared `register'.
2816 * (C++ only) A base class is not initialized in a derived
2817 class' copy constructor.
2821 Warn if an array subscript has type `char'. This is a common cause
2822 of error, as programmers often forget that this type is signed on
2823 some machines. This warning is enabled by `-Wall'.
2826 Warn whenever a comment-start sequence `/*' appears in a `/*'
2827 comment, or whenever a Backslash-Newline appears in a `//' comment.
2828 This warning is enabled by `-Wall'.
2831 Check calls to `printf' and `scanf', etc., to make sure that the
2832 arguments supplied have types appropriate to the format string
2833 specified, and that the conversions specified in the format string
2834 make sense. This includes standard functions, and others
2835 specified by format attributes (*note Function Attributes::), in
2836 the `printf', `scanf', `strftime' and `strfmon' (an X/Open
2837 extension, not in the C standard) families (or other
2838 target-specific families). Which functions are checked without
2839 format attributes having been specified depends on the standard
2840 version selected, and such checks of functions without the
2841 attribute specified are disabled by `-ffreestanding' or
2844 The formats are checked against the format features supported by
2845 GNU libc version 2.2. These include all ISO C90 and C99 features,
2846 as well as features from the Single Unix Specification and some
2847 BSD and GNU extensions. Other library implementations may not
2848 support all these features; GCC does not support warning about
2849 features that go beyond a particular library's limitations.
2850 However, if `-pedantic' is used with `-Wformat', warnings will be
2851 given about format features not in the selected standard version
2852 (but not for `strfmon' formats, since those are not in any version
2853 of the C standard). *Note Options Controlling C Dialect: C
2856 Since `-Wformat' also checks for null format arguments for several
2857 functions, `-Wformat' also implies `-Wnonnull'.
2859 `-Wformat' is included in `-Wall'. For more control over some
2860 aspects of format checking, the options `-Wformat-y2k',
2861 `-Wno-format-extra-args', `-Wno-format-zero-length',
2862 `-Wformat-nonliteral', `-Wformat-security', and `-Wformat=2' are
2863 available, but are not included in `-Wall'.
2866 If `-Wformat' is specified, also warn about `strftime' formats
2867 which may yield only a two-digit year.
2869 `-Wno-format-contains-nul'
2870 If `-Wformat' is specified, do not warn about format strings that
2873 `-Wno-format-extra-args'
2874 If `-Wformat' is specified, do not warn about excess arguments to a
2875 `printf' or `scanf' format function. The C standard specifies
2876 that such arguments are ignored.
2878 Where the unused arguments lie between used arguments that are
2879 specified with `$' operand number specifications, normally
2880 warnings are still given, since the implementation could not know
2881 what type to pass to `va_arg' to skip the unused arguments.
2882 However, in the case of `scanf' formats, this option will suppress
2883 the warning if the unused arguments are all pointers, since the
2884 Single Unix Specification says that such unused arguments are
2887 `-Wno-format-zero-length (C and Objective-C only)'
2888 If `-Wformat' is specified, do not warn about zero-length formats.
2889 The C standard specifies that zero-length formats are allowed.
2891 `-Wformat-nonliteral'
2892 If `-Wformat' is specified, also warn if the format string is not a
2893 string literal and so cannot be checked, unless the format function
2894 takes its format arguments as a `va_list'.
2897 If `-Wformat' is specified, also warn about uses of format
2898 functions that represent possible security problems. At present,
2899 this warns about calls to `printf' and `scanf' functions where the
2900 format string is not a string literal and there are no format
2901 arguments, as in `printf (foo);'. This may be a security hole if
2902 the format string came from untrusted input and contains `%n'.
2903 (This is currently a subset of what `-Wformat-nonliteral' warns
2904 about, but in future warnings may be added to `-Wformat-security'
2905 that are not included in `-Wformat-nonliteral'.)
2908 Enable `-Wformat' plus format checks not included in `-Wformat'.
2909 Currently equivalent to `-Wformat -Wformat-nonliteral
2910 -Wformat-security -Wformat-y2k'.
2912 `-Wnonnull (C and Objective-C only)'
2913 Warn about passing a null pointer for arguments marked as
2914 requiring a non-null value by the `nonnull' function attribute.
2916 `-Wnonnull' is included in `-Wall' and `-Wformat'. It can be
2917 disabled with the `-Wno-nonnull' option.
2919 `-Winit-self (C, C++, Objective-C and Objective-C++ only)'
2920 Warn about uninitialized variables which are initialized with
2921 themselves. Note this option can only be used with the
2922 `-Wuninitialized' option.
2924 For example, GCC will warn about `i' being uninitialized in the
2925 following snippet only when `-Winit-self' has been specified:
2932 `-Wimplicit-int (C and Objective-C only)'
2933 Warn when a declaration does not specify a type. This warning is
2936 `-Wimplicit-function-declaration (C and Objective-C only)'
2937 Give a warning whenever a function is used before being declared.
2938 In C99 mode (`-std=c99' or `-std=gnu99'), this warning is enabled
2939 by default and it is made into an error by `-pedantic-errors'.
2940 This warning is also enabled by `-Wall'.
2943 Same as `-Wimplicit-int' and `-Wimplicit-function-declaration'.
2944 This warning is enabled by `-Wall'.
2946 `-Wignored-qualifiers (C and C++ only)'
2947 Warn if the return type of a function has a type qualifier such as
2948 `const'. For ISO C such a type qualifier has no effect, since the
2949 value returned by a function is not an lvalue. For C++, the
2950 warning is only emitted for scalar types or `void'. ISO C
2951 prohibits qualified `void' return types on function definitions,
2952 so such return types always receive a warning even without this
2955 This warning is also enabled by `-Wextra'.
2958 Warn if the type of `main' is suspicious. `main' should be a
2959 function with external linkage, returning int, taking either zero
2960 arguments, two, or three arguments of appropriate types. This
2961 warning is enabled by default in C++ and is enabled by either
2962 `-Wall' or `-pedantic'.
2965 Warn if an aggregate or union initializer is not fully bracketed.
2966 In the following example, the initializer for `a' is not fully
2967 bracketed, but that for `b' is fully bracketed.
2969 int a[2][2] = { 0, 1, 2, 3 };
2970 int b[2][2] = { { 0, 1 }, { 2, 3 } };
2972 This warning is enabled by `-Wall'.
2974 `-Wmissing-include-dirs (C, C++, Objective-C and Objective-C++ only)'
2975 Warn if a user-supplied include directory does not exist.
2978 Warn if parentheses are omitted in certain contexts, such as when
2979 there is an assignment in a context where a truth value is
2980 expected, or when operators are nested whose precedence people
2981 often get confused about.
2983 Also warn if a comparison like `x<=y<=z' appears; this is
2984 equivalent to `(x<=y ? 1 : 0) <= z', which is a different
2985 interpretation from that of ordinary mathematical notation.
2987 Also warn about constructions where there may be confusion to which
2988 `if' statement an `else' branch belongs. Here is an example of
2999 In C/C++, every `else' branch belongs to the innermost possible
3000 `if' statement, which in this example is `if (b)'. This is often
3001 not what the programmer expected, as illustrated in the above
3002 example by indentation the programmer chose. When there is the
3003 potential for this confusion, GCC will issue a warning when this
3004 flag is specified. To eliminate the warning, add explicit braces
3005 around the innermost `if' statement so there is no way the `else'
3006 could belong to the enclosing `if'. The resulting code would look
3019 This warning is enabled by `-Wall'.
3022 Warn about code that may have undefined semantics because of
3023 violations of sequence point rules in the C and C++ standards.
3025 The C and C++ standards defines the order in which expressions in
3026 a C/C++ program are evaluated in terms of "sequence points", which
3027 represent a partial ordering between the execution of parts of the
3028 program: those executed before the sequence point, and those
3029 executed after it. These occur after the evaluation of a full
3030 expression (one which is not part of a larger expression), after
3031 the evaluation of the first operand of a `&&', `||', `? :' or `,'
3032 (comma) operator, before a function is called (but after the
3033 evaluation of its arguments and the expression denoting the called
3034 function), and in certain other places. Other than as expressed
3035 by the sequence point rules, the order of evaluation of
3036 subexpressions of an expression is not specified. All these rules
3037 describe only a partial order rather than a total order, since,
3038 for example, if two functions are called within one expression
3039 with no sequence point between them, the order in which the
3040 functions are called is not specified. However, the standards
3041 committee have ruled that function calls do not overlap.
3043 It is not specified when between sequence points modifications to
3044 the values of objects take effect. Programs whose behavior
3045 depends on this have undefined behavior; the C and C++ standards
3046 specify that "Between the previous and next sequence point an
3047 object shall have its stored value modified at most once by the
3048 evaluation of an expression. Furthermore, the prior value shall
3049 be read only to determine the value to be stored.". If a program
3050 breaks these rules, the results on any particular implementation
3051 are entirely unpredictable.
3053 Examples of code with undefined behavior are `a = a++;', `a[n] =
3054 b[n++]' and `a[i++] = i;'. Some more complicated cases are not
3055 diagnosed by this option, and it may give an occasional false
3056 positive result, but in general it has been found fairly effective
3057 at detecting this sort of problem in programs.
3059 The standard is worded confusingly, therefore there is some debate
3060 over the precise meaning of the sequence point rules in subtle
3061 cases. Links to discussions of the problem, including proposed
3062 formal definitions, may be found on the GCC readings page, at
3063 `http://gcc.gnu.org/readings.html'.
3065 This warning is enabled by `-Wall' for C and C++.
3068 Warn whenever a function is defined with a return-type that
3069 defaults to `int'. Also warn about any `return' statement with no
3070 return-value in a function whose return-type is not `void'
3071 (falling off the end of the function body is considered returning
3072 without a value), and about a `return' statement with a expression
3073 in a function whose return-type is `void'.
3075 For C++, a function without return type always produces a
3076 diagnostic message, even when `-Wno-return-type' is specified.
3077 The only exceptions are `main' and functions defined in system
3080 This warning is enabled by `-Wall'.
3083 Warn whenever a `switch' statement has an index of enumerated type
3084 and lacks a `case' for one or more of the named codes of that
3085 enumeration. (The presence of a `default' label prevents this
3086 warning.) `case' labels outside the enumeration range also
3087 provoke warnings when this option is used. This warning is
3091 Warn whenever a `switch' statement does not have a `default' case.
3094 Warn whenever a `switch' statement has an index of enumerated type
3095 and lacks a `case' for one or more of the named codes of that
3096 enumeration. `case' labels outside the enumeration range also
3097 provoke warnings when this option is used.
3099 `-Wsync-nand (C and C++ only)'
3100 Warn when `__sync_fetch_and_nand' and `__sync_nand_and_fetch'
3101 built-in functions are used. These functions changed semantics in
3105 Warn if any trigraphs are encountered that might change the
3106 meaning of the program (trigraphs within comments are not warned
3107 about). This warning is enabled by `-Wall'.
3110 Warn whenever a static function is declared but not defined or a
3111 non-inline static function is unused. This warning is enabled by
3115 Warn whenever a label is declared but not used. This warning is
3118 To suppress this warning use the `unused' attribute (*note
3119 Variable Attributes::).
3121 `-Wunused-parameter'
3122 Warn whenever a function parameter is unused aside from its
3125 To suppress this warning use the `unused' attribute (*note
3126 Variable Attributes::).
3129 Warn whenever a local variable or non-constant static variable is
3130 unused aside from its declaration. This warning is enabled by
3133 To suppress this warning use the `unused' attribute (*note
3134 Variable Attributes::).
3137 Warn whenever a statement computes a result that is explicitly not
3138 used. To suppress this warning cast the unused expression to
3139 `void'. This includes an expression-statement or the left-hand
3140 side of a comma expression that contains no side effects. For
3141 example, an expression such as `x[i,j]' will cause a warning, while
3142 `x[(void)i,j]' will not.
3144 This warning is enabled by `-Wall'.
3147 All the above `-Wunused' options combined.
3149 In order to get a warning about an unused function parameter, you
3150 must either specify `-Wextra -Wunused' (note that `-Wall' implies
3151 `-Wunused'), or separately specify `-Wunused-parameter'.
3154 Warn if an automatic variable is used without first being
3155 initialized or if a variable may be clobbered by a `setjmp' call.
3156 In C++, warn if a non-static reference or non-static `const' member
3157 appears in a class without constructors.
3159 If you want to warn about code which uses the uninitialized value
3160 of the variable in its own initializer, use the `-Winit-self'
3163 These warnings occur for individual uninitialized or clobbered
3164 elements of structure, union or array variables as well as for
3165 variables which are uninitialized or clobbered as a whole. They do
3166 not occur for variables or elements declared `volatile'. Because
3167 these warnings depend on optimization, the exact variables or
3168 elements for which there are warnings will depend on the precise
3169 optimization options and version of GCC used.
3171 Note that there may be no warning about a variable that is used
3172 only to compute a value that itself is never used, because such
3173 computations may be deleted by data flow analysis before the
3174 warnings are printed.
3176 These warnings are made optional because GCC is not smart enough
3177 to see all the reasons why the code might be correct despite
3178 appearing to have an error. Here is one example of how this can
3194 If the value of `y' is always 1, 2 or 3, then `x' is always
3195 initialized, but GCC doesn't know this. Here is another common
3200 if (change_y) save_y = y, y = new_y;
3202 if (change_y) y = save_y;
3205 This has no bug because `save_y' is used only if it is set.
3207 This option also warns when a non-volatile automatic variable
3208 might be changed by a call to `longjmp'. These warnings as well
3209 are possible only in optimizing compilation.
3211 The compiler sees only the calls to `setjmp'. It cannot know
3212 where `longjmp' will be called; in fact, a signal handler could
3213 call it at any point in the code. As a result, you may get a
3214 warning even when there is in fact no problem because `longjmp'
3215 cannot in fact be called at the place which would cause a problem.
3217 Some spurious warnings can be avoided if you declare all the
3218 functions you use that never return as `noreturn'. *Note Function
3221 This warning is enabled by `-Wall' or `-Wextra'.
3224 Warn when a #pragma directive is encountered which is not
3225 understood by GCC. If this command line option is used, warnings
3226 will even be issued for unknown pragmas in system header files.
3227 This is not the case if the warnings were only enabled by the
3228 `-Wall' command line option.
3231 Do not warn about misuses of pragmas, such as incorrect parameters,
3232 invalid syntax, or conflicts between pragmas. See also
3233 `-Wunknown-pragmas'.
3236 This option is only active when `-fstrict-aliasing' is active. It
3237 warns about code which might break the strict aliasing rules that
3238 the compiler is using for optimization. The warning does not
3239 catch all cases, but does attempt to catch the more common
3240 pitfalls. It is included in `-Wall'. It is equivalent to
3241 `-Wstrict-aliasing=3'
3243 `-Wstrict-aliasing=n'
3244 This option is only active when `-fstrict-aliasing' is active. It
3245 warns about code which might break the strict aliasing rules that
3246 the compiler is using for optimization. Higher levels correspond
3247 to higher accuracy (fewer false positives). Higher levels also
3248 correspond to more effort, similar to the way -O works.
3249 `-Wstrict-aliasing' is equivalent to `-Wstrict-aliasing=n', with
3252 Level 1: Most aggressive, quick, least accurate. Possibly useful
3253 when higher levels do not warn but -fstrict-aliasing still breaks
3254 the code, as it has very few false negatives. However, it has
3255 many false positives. Warns for all pointer conversions between
3256 possibly incompatible types, even if never dereferenced. Runs in
3259 Level 2: Aggressive, quick, not too precise. May still have many
3260 false positives (not as many as level 1 though), and few false
3261 negatives (but possibly more than level 1). Unlike level 1, it
3262 only warns when an address is taken. Warns about incomplete
3263 types. Runs in the frontend only.
3265 Level 3 (default for `-Wstrict-aliasing'): Should have very few
3266 false positives and few false negatives. Slightly slower than
3267 levels 1 or 2 when optimization is enabled. Takes care of the
3268 common punn+dereference pattern in the frontend:
3269 `*(int*)&some_float'. If optimization is enabled, it also runs in
3270 the backend, where it deals with multiple statement cases using
3271 flow-sensitive points-to information. Only warns when the
3272 converted pointer is dereferenced. Does not warn about incomplete
3276 `-Wstrict-overflow=N'
3277 This option is only active when `-fstrict-overflow' is active. It
3278 warns about cases where the compiler optimizes based on the
3279 assumption that signed overflow does not occur. Note that it does
3280 not warn about all cases where the code might overflow: it only
3281 warns about cases where the compiler implements some optimization.
3282 Thus this warning depends on the optimization level.
3284 An optimization which assumes that signed overflow does not occur
3285 is perfectly safe if the values of the variables involved are such
3286 that overflow never does, in fact, occur. Therefore this warning
3287 can easily give a false positive: a warning about code which is not
3288 actually a problem. To help focus on important issues, several
3289 warning levels are defined. No warnings are issued for the use of
3290 undefined signed overflow when estimating how many iterations a
3291 loop will require, in particular when determining whether a loop
3292 will be executed at all.
3294 `-Wstrict-overflow=1'
3295 Warn about cases which are both questionable and easy to
3296 avoid. For example: `x + 1 > x'; with `-fstrict-overflow',
3297 the compiler will simplify this to `1'. This level of
3298 `-Wstrict-overflow' is enabled by `-Wall'; higher levels are
3299 not, and must be explicitly requested.
3301 `-Wstrict-overflow=2'
3302 Also warn about other cases where a comparison is simplified
3303 to a constant. For example: `abs (x) >= 0'. This can only be
3304 simplified when `-fstrict-overflow' is in effect, because
3305 `abs (INT_MIN)' overflows to `INT_MIN', which is less than
3306 zero. `-Wstrict-overflow' (with no level) is the same as
3307 `-Wstrict-overflow=2'.
3309 `-Wstrict-overflow=3'
3310 Also warn about other cases where a comparison is simplified.
3311 For example: `x + 1 > 1' will be simplified to `x > 0'.
3313 `-Wstrict-overflow=4'
3314 Also warn about other simplifications not covered by the
3315 above cases. For example: `(x * 10) / 5' will be simplified
3318 `-Wstrict-overflow=5'
3319 Also warn about cases where the compiler reduces the
3320 magnitude of a constant involved in a comparison. For
3321 example: `x + 2 > y' will be simplified to `x + 1 >= y'.
3322 This is reported only at the highest warning level because
3323 this simplification applies to many comparisons, so this
3324 warning level will give a very large number of false
3328 This option is only active when `-ftree-vrp' is active (default
3329 for -O2 and above). It warns about subscripts to arrays that are
3330 always out of bounds. This warning is enabled by `-Wall'.
3333 Do not warn about compile-time integer division by zero. Floating
3334 point division by zero is not warned about, as it can be a
3335 legitimate way of obtaining infinities and NaNs.
3338 Print warning messages for constructs found in system header files.
3339 Warnings from system headers are normally suppressed, on the
3340 assumption that they usually do not indicate real problems and
3341 would only make the compiler output harder to read. Using this
3342 command line option tells GCC to emit warnings from system headers
3343 as if they occurred in user code. However, note that using
3344 `-Wall' in conjunction with this option will _not_ warn about
3345 unknown pragmas in system headers--for that, `-Wunknown-pragmas'
3349 Warn if floating point values are used in equality comparisons.
3351 The idea behind this is that sometimes it is convenient (for the
3352 programmer) to consider floating-point values as approximations to
3353 infinitely precise real numbers. If you are doing this, then you
3354 need to compute (by analyzing the code, or in some other way) the
3355 maximum or likely maximum error that the computation introduces,
3356 and allow for it when performing comparisons (and when producing
3357 output, but that's a different problem). In particular, instead
3358 of testing for equality, you would check to see whether the two
3359 values have ranges that overlap; and this is done with the
3360 relational operators, so equality comparisons are probably
3363 `-Wtraditional (C and Objective-C only)'
3364 Warn about certain constructs that behave differently in
3365 traditional and ISO C. Also warn about ISO C constructs that have
3366 no traditional C equivalent, and/or problematic constructs which
3369 * Macro parameters that appear within string literals in the
3370 macro body. In traditional C macro replacement takes place
3371 within string literals, but does not in ISO C.
3373 * In traditional C, some preprocessor directives did not exist.
3374 Traditional preprocessors would only consider a line to be a
3375 directive if the `#' appeared in column 1 on the line.
3376 Therefore `-Wtraditional' warns about directives that
3377 traditional C understands but would ignore because the `#'
3378 does not appear as the first character on the line. It also
3379 suggests you hide directives like `#pragma' not understood by
3380 traditional C by indenting them. Some traditional
3381 implementations would not recognize `#elif', so it suggests
3382 avoiding it altogether.
3384 * A function-like macro that appears without arguments.
3386 * The unary plus operator.
3388 * The `U' integer constant suffix, or the `F' or `L' floating
3389 point constant suffixes. (Traditional C does support the `L'
3390 suffix on integer constants.) Note, these suffixes appear in
3391 macros defined in the system headers of most modern systems,
3392 e.g. the `_MIN'/`_MAX' macros in `<limits.h>'. Use of these
3393 macros in user code might normally lead to spurious warnings,
3394 however GCC's integrated preprocessor has enough context to
3395 avoid warning in these cases.
3397 * A function declared external in one block and then used after
3398 the end of the block.
3400 * A `switch' statement has an operand of type `long'.
3402 * A non-`static' function declaration follows a `static' one.
3403 This construct is not accepted by some traditional C
3406 * The ISO type of an integer constant has a different width or
3407 signedness from its traditional type. This warning is only
3408 issued if the base of the constant is ten. I.e. hexadecimal
3409 or octal values, which typically represent bit patterns, are
3412 * Usage of ISO string concatenation is detected.
3414 * Initialization of automatic aggregates.
3416 * Identifier conflicts with labels. Traditional C lacks a
3417 separate namespace for labels.
3419 * Initialization of unions. If the initializer is zero, the
3420 warning is omitted. This is done under the assumption that
3421 the zero initializer in user code appears conditioned on e.g.
3422 `__STDC__' to avoid missing initializer warnings and relies
3423 on default initialization to zero in the traditional C case.
3425 * Conversions by prototypes between fixed/floating point values
3426 and vice versa. The absence of these prototypes when
3427 compiling with traditional C would cause serious problems.
3428 This is a subset of the possible conversion warnings, for the
3429 full set use `-Wtraditional-conversion'.
3431 * Use of ISO C style function definitions. This warning
3432 intentionally is _not_ issued for prototype declarations or
3433 variadic functions because these ISO C features will appear
3434 in your code when using libiberty's traditional C
3435 compatibility macros, `PARAMS' and `VPARAMS'. This warning
3436 is also bypassed for nested functions because that feature is
3437 already a GCC extension and thus not relevant to traditional
3440 `-Wtraditional-conversion (C and Objective-C only)'
3441 Warn if a prototype causes a type conversion that is different
3442 from what would happen to the same argument in the absence of a
3443 prototype. This includes conversions of fixed point to floating
3444 and vice versa, and conversions changing the width or signedness
3445 of a fixed point argument except when the same as the default
3448 `-Wdeclaration-after-statement (C and Objective-C only)'
3449 Warn when a declaration is found after a statement in a block.
3450 This construct, known from C++, was introduced with ISO C99 and is
3451 by default allowed in GCC. It is not supported by ISO C90 and was
3452 not supported by GCC versions before GCC 3.0. *Note Mixed
3456 Warn if an undefined identifier is evaluated in an `#if' directive.
3459 Do not warn whenever an `#else' or an `#endif' are followed by
3463 Warn whenever a local variable shadows another local variable,
3464 parameter or global variable or whenever a built-in function is
3468 Warn whenever an object of larger than LEN bytes is defined.
3470 `-Wframe-larger-than=LEN'
3471 Warn if the size of a function frame is larger than LEN bytes.
3472 The computation done to determine the stack frame size is
3473 approximate and not conservative. The actual requirements may be
3474 somewhat greater than LEN even if you do not get a warning. In
3475 addition, any space allocated via `alloca', variable-length
3476 arrays, or related constructs is not included by the compiler when
3477 determining whether or not to issue a warning.
3479 `-Wunsafe-loop-optimizations'
3480 Warn if the loop cannot be optimized because the compiler could not
3481 assume anything on the bounds of the loop indices. With
3482 `-funsafe-loop-optimizations' warn if the compiler made such
3485 `-Wno-pedantic-ms-format (MinGW targets only)'
3486 Disables the warnings about non-ISO `printf' / `scanf' format
3487 width specifiers `I32', `I64', and `I' used on Windows targets
3488 depending on the MS runtime, when you are using the options
3489 `-Wformat' and `-pedantic' without gnu-extensions.
3492 Warn about anything that depends on the "size of" a function type
3493 or of `void'. GNU C assigns these types a size of 1, for
3494 convenience in calculations with `void *' pointers and pointers to
3495 functions. In C++, warn also when an arithmetic operation involves
3496 `NULL'. This warning is also enabled by `-pedantic'.
3499 Warn if a comparison is always true or always false due to the
3500 limited range of the data type, but do not warn for constant
3501 expressions. For example, warn if an unsigned variable is
3502 compared against zero with `<' or `>='. This warning is also
3503 enabled by `-Wextra'.
3505 `-Wbad-function-cast (C and Objective-C only)'
3506 Warn whenever a function call is cast to a non-matching type. For
3507 example, warn if `int malloc()' is cast to `anything *'.
3509 `-Wc++-compat (C and Objective-C only)'
3510 Warn about ISO C constructs that are outside of the common subset
3511 of ISO C and ISO C++, e.g. request for implicit conversion from
3512 `void *' to a pointer to non-`void' type.
3514 `-Wc++0x-compat (C++ and Objective-C++ only)'
3515 Warn about C++ constructs whose meaning differs between ISO C++
3516 1998 and ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will
3517 become keywords in ISO C++ 200x. This warning is enabled by
3521 Warn whenever a pointer is cast so as to remove a type qualifier
3522 from the target type. For example, warn if a `const char *' is
3523 cast to an ordinary `char *'.
3526 Warn whenever a pointer is cast such that the required alignment
3527 of the target is increased. For example, warn if a `char *' is
3528 cast to an `int *' on machines where integers can only be accessed
3529 at two- or four-byte boundaries.
3532 When compiling C, give string constants the type `const
3533 char[LENGTH]' so that copying the address of one into a
3534 non-`const' `char *' pointer will get a warning. These warnings
3535 will help you find at compile time code that can try to write into
3536 a string constant, but only if you have been very careful about
3537 using `const' in declarations and prototypes. Otherwise, it will
3538 just be a nuisance. This is why we did not make `-Wall' request
3541 When compiling C++, warn about the deprecated conversion from
3542 string literals to `char *'. This warning is enabled by default
3546 Warn for variables that might be changed by `longjmp' or `vfork'.
3547 This warning is also enabled by `-Wextra'.
3550 Warn for implicit conversions that may alter a value. This includes
3551 conversions between real and integer, like `abs (x)' when `x' is
3552 `double'; conversions between signed and unsigned, like `unsigned
3553 ui = -1'; and conversions to smaller types, like `sqrtf (M_PI)'.
3554 Do not warn for explicit casts like `abs ((int) x)' and `ui =
3555 (unsigned) -1', or if the value is not changed by the conversion
3556 like in `abs (2.0)'. Warnings about conversions between signed
3557 and unsigned integers can be disabled by using
3558 `-Wno-sign-conversion'.
3560 For C++, also warn for conversions between `NULL' and non-pointer
3561 types; confusing overload resolution for user-defined conversions;
3562 and conversions that will never use a type conversion operator:
3563 conversions to `void', the same type, a base class or a reference
3564 to them. Warnings about conversions between signed and unsigned
3565 integers are disabled by default in C++ unless `-Wsign-conversion'
3566 is explicitly enabled.
3569 Warn if an empty body occurs in an `if', `else' or `do while'
3570 statement. This warning is also enabled by `-Wextra'.
3572 `-Wenum-compare (C++ and Objective-C++ only)'
3573 Warn about a comparison between values of different enum types.
3574 This warning is enabled by default.
3577 Warn when a comparison between signed and unsigned values could
3578 produce an incorrect result when the signed value is converted to
3579 unsigned. This warning is also enabled by `-Wextra'; to get the
3580 other warnings of `-Wextra' without this warning, use `-Wextra
3584 Warn for implicit conversions that may change the sign of an
3585 integer value, like assigning a signed integer expression to an
3586 unsigned integer variable. An explicit cast silences the warning.
3587 In C, this option is enabled also by `-Wconversion'.
3590 Warn about suspicious uses of memory addresses. These include using
3591 the address of a function in a conditional expression, such as
3592 `void func(void); if (func)', and comparisons against the memory
3593 address of a string literal, such as `if (x == "abc")'. Such uses
3594 typically indicate a programmer error: the address of a function
3595 always evaluates to true, so their use in a conditional usually
3596 indicate that the programmer forgot the parentheses in a function
3597 call; and comparisons against string literals result in unspecified
3598 behavior and are not portable in C, so they usually indicate that
3599 the programmer intended to use `strcmp'. This warning is enabled
3603 Warn about suspicious uses of logical operators in expressions.
3604 This includes using logical operators in contexts where a bit-wise
3605 operator is likely to be expected.
3607 `-Waggregate-return'
3608 Warn if any functions that return structures or unions are defined
3609 or called. (In languages where you can return an array, this also
3613 Do not warn if an unexpected `__attribute__' is used, such as
3614 unrecognized attributes, function attributes applied to variables,
3615 etc. This will not stop errors for incorrect use of supported
3618 `-Wno-builtin-macro-redefined'
3619 Do not warn if certain built-in macros are redefined. This
3620 suppresses warnings for redefinition of `__TIMESTAMP__',
3621 `__TIME__', `__DATE__', `__FILE__', and `__BASE_FILE__'.
3623 `-Wstrict-prototypes (C and Objective-C only)'
3624 Warn if a function is declared or defined without specifying the
3625 argument types. (An old-style function definition is permitted
3626 without a warning if preceded by a declaration which specifies the
3629 `-Wold-style-declaration (C and Objective-C only)'
3630 Warn for obsolescent usages, according to the C Standard, in a
3631 declaration. For example, warn if storage-class specifiers like
3632 `static' are not the first things in a declaration. This warning
3633 is also enabled by `-Wextra'.
3635 `-Wold-style-definition (C and Objective-C only)'
3636 Warn if an old-style function definition is used. A warning is
3637 given even if there is a previous prototype.
3639 `-Wmissing-parameter-type (C and Objective-C only)'
3640 A function parameter is declared without a type specifier in
3641 K&R-style functions:
3645 This warning is also enabled by `-Wextra'.
3647 `-Wmissing-prototypes (C and Objective-C only)'
3648 Warn if a global function is defined without a previous prototype
3649 declaration. This warning is issued even if the definition itself
3650 provides a prototype. The aim is to detect global functions that
3651 fail to be declared in header files.
3653 `-Wmissing-declarations'
3654 Warn if a global function is defined without a previous
3655 declaration. Do so even if the definition itself provides a
3656 prototype. Use this option to detect global functions that are
3657 not declared in header files. In C++, no warnings are issued for
3658 function templates, or for inline functions, or for functions in
3659 anonymous namespaces.
3661 `-Wmissing-field-initializers'
3662 Warn if a structure's initializer has some fields missing. For
3663 example, the following code would cause such a warning, because
3664 `x.h' is implicitly zero:
3666 struct s { int f, g, h; };
3667 struct s x = { 3, 4 };
3669 This option does not warn about designated initializers, so the
3670 following modification would not trigger a warning:
3672 struct s { int f, g, h; };
3673 struct s x = { .f = 3, .g = 4 };
3675 This warning is included in `-Wextra'. To get other `-Wextra'
3676 warnings without this one, use `-Wextra
3677 -Wno-missing-field-initializers'.
3679 `-Wmissing-noreturn'
3680 Warn about functions which might be candidates for attribute
3681 `noreturn'. Note these are only possible candidates, not absolute
3682 ones. Care should be taken to manually verify functions actually
3683 do not ever return before adding the `noreturn' attribute,
3684 otherwise subtle code generation bugs could be introduced. You
3685 will not get a warning for `main' in hosted C environments.
3687 `-Wmissing-format-attribute'
3688 Warn about function pointers which might be candidates for `format'
3689 attributes. Note these are only possible candidates, not absolute
3690 ones. GCC will guess that function pointers with `format'
3691 attributes that are used in assignment, initialization, parameter
3692 passing or return statements should have a corresponding `format'
3693 attribute in the resulting type. I.e. the left-hand side of the
3694 assignment or initialization, the type of the parameter variable,
3695 or the return type of the containing function respectively should
3696 also have a `format' attribute to avoid the warning.
3698 GCC will also warn about function definitions which might be
3699 candidates for `format' attributes. Again, these are only
3700 possible candidates. GCC will guess that `format' attributes
3701 might be appropriate for any function that calls a function like
3702 `vprintf' or `vscanf', but this might not always be the case, and
3703 some functions for which `format' attributes are appropriate may
3707 Do not warn if a multicharacter constant (`'FOOF'') is used.
3708 Usually they indicate a typo in the user's code, as they have
3709 implementation-defined values, and should not be used in portable
3712 `-Wnormalized=<none|id|nfc|nfkc>'
3713 In ISO C and ISO C++, two identifiers are different if they are
3714 different sequences of characters. However, sometimes when
3715 characters outside the basic ASCII character set are used, you can
3716 have two different character sequences that look the same. To
3717 avoid confusion, the ISO 10646 standard sets out some
3718 "normalization rules" which when applied ensure that two sequences
3719 that look the same are turned into the same sequence. GCC can
3720 warn you if you are using identifiers which have not been
3721 normalized; this option controls that warning.
3723 There are four levels of warning that GCC supports. The default is
3724 `-Wnormalized=nfc', which warns about any identifier which is not
3725 in the ISO 10646 "C" normalized form, "NFC". NFC is the
3726 recommended form for most uses.
3728 Unfortunately, there are some characters which ISO C and ISO C++
3729 allow in identifiers that when turned into NFC aren't allowable as
3730 identifiers. That is, there's no way to use these symbols in
3731 portable ISO C or C++ and have all your identifiers in NFC.
3732 `-Wnormalized=id' suppresses the warning for these characters. It
3733 is hoped that future versions of the standards involved will
3734 correct this, which is why this option is not the default.
3736 You can switch the warning off for all characters by writing
3737 `-Wnormalized=none'. You would only want to do this if you were
3738 using some other normalization scheme (like "D"), because
3739 otherwise you can easily create bugs that are literally impossible
3742 Some characters in ISO 10646 have distinct meanings but look
3743 identical in some fonts or display methodologies, especially once
3744 formatting has been applied. For instance `\u207F', "SUPERSCRIPT
3745 LATIN SMALL LETTER N", will display just like a regular `n' which
3746 has been placed in a superscript. ISO 10646 defines the "NFKC"
3747 normalization scheme to convert all these into a standard form as
3748 well, and GCC will warn if your code is not in NFKC if you use
3749 `-Wnormalized=nfkc'. This warning is comparable to warning about
3750 every identifier that contains the letter O because it might be
3751 confused with the digit 0, and so is not the default, but may be
3752 useful as a local coding convention if the programming environment
3753 is unable to be fixed to display these characters distinctly.
3756 Do not warn about usage of deprecated features. *Note Deprecated
3759 `-Wno-deprecated-declarations'
3760 Do not warn about uses of functions (*note Function Attributes::),
3761 variables (*note Variable Attributes::), and types (*note Type
3762 Attributes::) marked as deprecated by using the `deprecated'
3766 Do not warn about compile-time overflow in constant expressions.
3768 `-Woverride-init (C and Objective-C only)'
3769 Warn if an initialized field without side effects is overridden
3770 when using designated initializers (*note Designated Initializers:
3773 This warning is included in `-Wextra'. To get other `-Wextra'
3774 warnings without this one, use `-Wextra -Wno-override-init'.
3777 Warn if a structure is given the packed attribute, but the packed
3778 attribute has no effect on the layout or size of the structure.
3779 Such structures may be mis-aligned for little benefit. For
3780 instance, in this code, the variable `f.x' in `struct bar' will be
3781 misaligned even though `struct bar' does not itself have the
3787 } __attribute__((packed));
3793 `-Wpacked-bitfield-compat'
3794 The 4.1, 4.2 and 4.3 series of GCC ignore the `packed' attribute
3795 on bit-fields of type `char'. This has been fixed in GCC 4.4 but
3796 the change can lead to differences in the structure layout. GCC
3797 informs you when the offset of such a field has changed in GCC 4.4.
3798 For example there is no longer a 4-bit padding between field `a'
3799 and `b' in this structure:
3805 } __attribute__ ((packed));
3807 This warning is enabled by default. Use
3808 `-Wno-packed-bitfield-compat' to disable this warning.
3811 Warn if padding is included in a structure, either to align an
3812 element of the structure or to align the whole structure.
3813 Sometimes when this happens it is possible to rearrange the fields
3814 of the structure to reduce the padding and so make the structure
3818 Warn if anything is declared more than once in the same scope,
3819 even in cases where multiple declaration is valid and changes
3822 `-Wnested-externs (C and Objective-C only)'
3823 Warn if an `extern' declaration is encountered within a function.
3825 `-Wunreachable-code'
3826 Warn if the compiler detects that code will never be executed.
3828 This option is intended to warn when the compiler detects that at
3829 least a whole line of source code will never be executed, because
3830 some condition is never satisfied or because it is after a
3831 procedure that never returns.
3833 It is possible for this option to produce a warning even though
3834 there are circumstances under which part of the affected line can
3835 be executed, so care should be taken when removing
3836 apparently-unreachable code.
3838 For instance, when a function is inlined, a warning may mean that
3839 the line is unreachable in only one inlined copy of the function.
3841 This option is not made part of `-Wall' because in a debugging
3842 version of a program there is often substantial code which checks
3843 correct functioning of the program and is, hopefully, unreachable
3844 because the program does work. Another common use of unreachable
3845 code is to provide behavior which is selectable at compile-time.
3848 Warn if a function can not be inlined and it was declared as
3849 inline. Even with this option, the compiler will not warn about
3850 failures to inline functions declared in system headers.
3852 The compiler uses a variety of heuristics to determine whether or
3853 not to inline a function. For example, the compiler takes into
3854 account the size of the function being inlined and the amount of
3855 inlining that has already been done in the current function.
3856 Therefore, seemingly insignificant changes in the source program
3857 can cause the warnings produced by `-Winline' to appear or
3860 `-Wno-invalid-offsetof (C++ and Objective-C++ only)'
3861 Suppress warnings from applying the `offsetof' macro to a non-POD
3862 type. According to the 1998 ISO C++ standard, applying `offsetof'
3863 to a non-POD type is undefined. In existing C++ implementations,
3864 however, `offsetof' typically gives meaningful results even when
3865 applied to certain kinds of non-POD types. (Such as a simple
3866 `struct' that fails to be a POD type only by virtue of having a
3867 constructor.) This flag is for users who are aware that they are
3868 writing nonportable code and who have deliberately chosen to
3869 ignore the warning about it.
3871 The restrictions on `offsetof' may be relaxed in a future version
3872 of the C++ standard.
3874 `-Wno-int-to-pointer-cast (C and Objective-C only)'
3875 Suppress warnings from casts to pointer type of an integer of a
3878 `-Wno-pointer-to-int-cast (C and Objective-C only)'
3879 Suppress warnings from casts from a pointer to an integer type of a
3883 Warn if a precompiled header (*note Precompiled Headers::) is
3884 found in the search path but can't be used.
3887 Warn if `long long' type is used. This is default. To inhibit
3888 the warning messages, use `-Wno-long-long'. Flags `-Wlong-long'
3889 and `-Wno-long-long' are taken into account only when `-pedantic'
3893 Warn if variadic macros are used in pedantic ISO C90 mode, or the
3894 GNU alternate syntax when in pedantic ISO C99 mode. This is
3895 default. To inhibit the warning messages, use
3896 `-Wno-variadic-macros'.
3899 Warn if variable length array is used in the code. `-Wno-vla'
3900 will prevent the `-pedantic' warning of the variable length array.
3902 `-Wvolatile-register-var'
3903 Warn if a register variable is declared volatile. The volatile
3904 modifier does not inhibit all optimizations that may eliminate
3905 reads and/or writes to register variables. This warning is
3908 `-Wdisabled-optimization'
3909 Warn if a requested optimization pass is disabled. This warning
3910 does not generally indicate that there is anything wrong with your
3911 code; it merely indicates that GCC's optimizers were unable to
3912 handle the code effectively. Often, the problem is that your code
3913 is too big or too complex; GCC will refuse to optimize programs
3914 when the optimization itself is likely to take inordinate amounts
3917 `-Wpointer-sign (C and Objective-C only)'
3918 Warn for pointer argument passing or assignment with different
3919 signedness. This option is only supported for C and Objective-C.
3920 It is implied by `-Wall' and by `-pedantic', which can be disabled
3921 with `-Wno-pointer-sign'.
3924 This option is only active when `-fstack-protector' is active. It
3925 warns about functions that will not be protected against stack
3929 Suppress warnings about constructs that cannot be instrumented by
3932 `-Woverlength-strings'
3933 Warn about string constants which are longer than the "minimum
3934 maximum" length specified in the C standard. Modern compilers
3935 generally allow string constants which are much longer than the
3936 standard's minimum limit, but very portable programs should avoid
3937 using longer strings.
3939 The limit applies _after_ string constant concatenation, and does
3940 not count the trailing NUL. In C89, the limit was 509 characters;
3941 in C99, it was raised to 4095. C++98 does not specify a normative
3942 minimum maximum, so we do not diagnose overlength strings in C++.
3944 This option is implied by `-pedantic', and can be disabled with
3945 `-Wno-overlength-strings'.
3948 File: gcc.info, Node: Debugging Options, Next: Optimize Options, Prev: Warning Options, Up: Invoking GCC
3950 3.9 Options for Debugging Your Program or GCC
3951 =============================================
3953 GCC has various special options that are used for debugging either your
3957 Produce debugging information in the operating system's native
3958 format (stabs, COFF, XCOFF, or DWARF 2). GDB can work with this
3959 debugging information.
3961 On most systems that use stabs format, `-g' enables use of extra
3962 debugging information that only GDB can use; this extra information
3963 makes debugging work better in GDB but will probably make other
3964 debuggers crash or refuse to read the program. If you want to
3965 control for certain whether to generate the extra information, use
3966 `-gstabs+', `-gstabs', `-gxcoff+', `-gxcoff', or `-gvms' (see
3969 GCC allows you to use `-g' with `-O'. The shortcuts taken by
3970 optimized code may occasionally produce surprising results: some
3971 variables you declared may not exist at all; flow of control may
3972 briefly move where you did not expect it; some statements may not
3973 be executed because they compute constant results or their values
3974 were already at hand; some statements may execute in different
3975 places because they were moved out of loops.
3977 Nevertheless it proves possible to debug optimized output. This
3978 makes it reasonable to use the optimizer for programs that might
3981 The following options are useful when GCC is generated with the
3982 capability for more than one debugging format.
3985 Produce debugging information for use by GDB. This means to use
3986 the most expressive format available (DWARF 2, stabs, or the
3987 native format if neither of those are supported), including GDB
3988 extensions if at all possible.
3991 Produce debugging information in stabs format (if that is
3992 supported), without GDB extensions. This is the format used by
3993 DBX on most BSD systems. On MIPS, Alpha and System V Release 4
3994 systems this option produces stabs debugging output which is not
3995 understood by DBX or SDB. On System V Release 4 systems this
3996 option requires the GNU assembler.
3998 `-feliminate-unused-debug-symbols'
3999 Produce debugging information in stabs format (if that is
4000 supported), for only symbols that are actually used.
4002 `-femit-class-debug-always'
4003 Instead of emitting debugging information for a C++ class in only
4004 one object file, emit it in all object files using the class.
4005 This option should be used only with debuggers that are unable to
4006 handle the way GCC normally emits debugging information for
4007 classes because using this option will increase the size of
4008 debugging information by as much as a factor of two.
4011 Produce debugging information in stabs format (if that is
4012 supported), using GNU extensions understood only by the GNU
4013 debugger (GDB). The use of these extensions is likely to make
4014 other debuggers crash or refuse to read the program.
4017 Produce debugging information in COFF format (if that is
4018 supported). This is the format used by SDB on most System V
4019 systems prior to System V Release 4.
4022 Produce debugging information in XCOFF format (if that is
4023 supported). This is the format used by the DBX debugger on IBM
4027 Produce debugging information in XCOFF format (if that is
4028 supported), using GNU extensions understood only by the GNU
4029 debugger (GDB). The use of these extensions is likely to make
4030 other debuggers crash or refuse to read the program, and may cause
4031 assemblers other than the GNU assembler (GAS) to fail with an
4035 Produce debugging information in DWARF version 2 format (if that is
4036 supported). This is the format used by DBX on IRIX 6. With this
4037 option, GCC uses features of DWARF version 3 when they are useful;
4038 version 3 is upward compatible with version 2, but may still cause
4039 problems for older debuggers.
4042 Produce debugging information in DWARF version 4 format (if that is
4043 supported). With this option, GCC uses features of DWARF version 4
4044 when they are useful, including the placement of most type
4045 information in separate comdat sections. The DWARF version 4
4046 format is still a draft specification, and this option is
4047 currently experimental.
4050 Produce debugging information in VMS debug format (if that is
4051 supported). This is the format used by DEBUG on VMS systems.
4059 Request debugging information and also use LEVEL to specify how
4060 much information. The default level is 2.
4062 Level 0 produces no debug information at all. Thus, `-g0' negates
4065 Level 1 produces minimal information, enough for making backtraces
4066 in parts of the program that you don't plan to debug. This
4067 includes descriptions of functions and external variables, but no
4068 information about local variables and no line numbers.
4070 Level 3 includes extra information, such as all the macro
4071 definitions present in the program. Some debuggers support macro
4072 expansion when you use `-g3'.
4074 `-gdwarf-2' does not accept a concatenated debug level, because
4075 GCC used to support an option `-gdwarf' that meant to generate
4076 debug information in version 1 of the DWARF format (which is very
4077 different from version 2), and it would have been too confusing.
4078 That debug format is long obsolete, but the option cannot be
4079 changed now. Instead use an additional `-gLEVEL' option to change
4080 the debug level for DWARF2.
4082 `-feliminate-dwarf2-dups'
4083 Compress DWARF2 debugging information by eliminating duplicated
4084 information about each symbol. This option only makes sense when
4085 generating DWARF2 debugging information with `-gdwarf-2'.
4087 `-femit-struct-debug-baseonly'
4088 Emit debug information for struct-like types only when the base
4089 name of the compilation source file matches the base name of file
4090 in which the struct was defined.
4092 This option substantially reduces the size of debugging
4093 information, but at significant potential loss in type information
4094 to the debugger. See `-femit-struct-debug-reduced' for a less
4095 aggressive option. See `-femit-struct-debug-detailed' for more
4098 This option works only with DWARF 2.
4100 `-femit-struct-debug-reduced'
4101 Emit debug information for struct-like types only when the base
4102 name of the compilation source file matches the base name of file
4103 in which the type was defined, unless the struct is a template or
4104 defined in a system header.
4106 This option significantly reduces the size of debugging
4107 information, with some potential loss in type information to the
4108 debugger. See `-femit-struct-debug-baseonly' for a more
4109 aggressive option. See `-femit-struct-debug-detailed' for more
4112 This option works only with DWARF 2.
4114 `-femit-struct-debug-detailed[=SPEC-LIST]'
4115 Specify the struct-like types for which the compiler will generate
4116 debug information. The intent is to reduce duplicate struct debug
4117 information between different object files within the same program.
4119 This option is a detailed version of `-femit-struct-debug-reduced'
4120 and `-femit-struct-debug-baseonly', which will serve for most
4123 A specification has the syntax
4124 [`dir:'|`ind:'][`ord:'|`gen:'](`any'|`sys'|`base'|`none')
4126 The optional first word limits the specification to structs that
4127 are used directly (`dir:') or used indirectly (`ind:'). A struct
4128 type is used directly when it is the type of a variable, member.
4129 Indirect uses arise through pointers to structs. That is, when
4130 use of an incomplete struct would be legal, the use is indirect.
4131 An example is `struct one direct; struct two * indirect;'.
4133 The optional second word limits the specification to ordinary
4134 structs (`ord:') or generic structs (`gen:'). Generic structs are
4135 a bit complicated to explain. For C++, these are non-explicit
4136 specializations of template classes, or non-template classes
4137 within the above. Other programming languages have generics, but
4138 `-femit-struct-debug-detailed' does not yet implement them.
4140 The third word specifies the source files for those structs for
4141 which the compiler will emit debug information. The values `none'
4142 and `any' have the normal meaning. The value `base' means that
4143 the base of name of the file in which the type declaration appears
4144 must match the base of the name of the main compilation file. In
4145 practice, this means that types declared in `foo.c' and `foo.h'
4146 will have debug information, but types declared in other header
4147 will not. The value `sys' means those types satisfying `base' or
4148 declared in system or compiler headers.
4150 You may need to experiment to determine the best settings for your
4153 The default is `-femit-struct-debug-detailed=all'.
4155 This option works only with DWARF 2.
4157 `-fno-merge-debug-strings'
4158 Direct the linker to not merge together strings in the debugging
4159 information which are identical in different object files.
4160 Merging is not supported by all assemblers or linkers. Merging
4161 decreases the size of the debug information in the output file at
4162 the cost of increasing link processing time. Merging is enabled
4165 `-fdebug-prefix-map=OLD=NEW'
4166 When compiling files in directory `OLD', record debugging
4167 information describing them as in `NEW' instead.
4169 `-fno-dwarf2-cfi-asm'
4170 Emit DWARF 2 unwind info as compiler generated `.eh_frame' section
4171 instead of using GAS `.cfi_*' directives.
4174 Generate extra code to write profile information suitable for the
4175 analysis program `prof'. You must use this option when compiling
4176 the source files you want data about, and you must also use it when
4180 Generate extra code to write profile information suitable for the
4181 analysis program `gprof'. You must use this option when compiling
4182 the source files you want data about, and you must also use it when
4186 Makes the compiler print out each function name as it is compiled,
4187 and print some statistics about each pass when it finishes.
4190 Makes the compiler print some statistics about the time consumed
4191 by each pass when it finishes.
4194 Makes the compiler print some statistics about permanent memory
4195 allocation when it finishes.
4197 `-fpre-ipa-mem-report'
4199 `-fpost-ipa-mem-report'
4200 Makes the compiler print some statistics about permanent memory
4201 allocation before or after interprocedural optimization.
4204 Add code so that program flow "arcs" are instrumented. During
4205 execution the program records how many times each branch and call
4206 is executed and how many times it is taken or returns. When the
4207 compiled program exits it saves this data to a file called
4208 `AUXNAME.gcda' for each source file. The data may be used for
4209 profile-directed optimizations (`-fbranch-probabilities'), or for
4210 test coverage analysis (`-ftest-coverage'). Each object file's
4211 AUXNAME is generated from the name of the output file, if
4212 explicitly specified and it is not the final executable, otherwise
4213 it is the basename of the source file. In both cases any suffix
4214 is removed (e.g. `foo.gcda' for input file `dir/foo.c', or
4215 `dir/foo.gcda' for output file specified as `-o dir/foo.o').
4216 *Note Cross-profiling::.
4219 This option is used to compile and link code instrumented for
4220 coverage analysis. The option is a synonym for `-fprofile-arcs'
4221 `-ftest-coverage' (when compiling) and `-lgcov' (when linking).
4222 See the documentation for those options for more details.
4224 * Compile the source files with `-fprofile-arcs' plus
4225 optimization and code generation options. For test coverage
4226 analysis, use the additional `-ftest-coverage' option. You
4227 do not need to profile every source file in a program.
4229 * Link your object files with `-lgcov' or `-fprofile-arcs' (the
4230 latter implies the former).
4232 * Run the program on a representative workload to generate the
4233 arc profile information. This may be repeated any number of
4234 times. You can run concurrent instances of your program, and
4235 provided that the file system supports locking, the data
4236 files will be correctly updated. Also `fork' calls are
4237 detected and correctly handled (double counting will not
4240 * For profile-directed optimizations, compile the source files
4241 again with the same optimization and code generation options
4242 plus `-fbranch-probabilities' (*note Options that Control
4243 Optimization: Optimize Options.).
4245 * For test coverage analysis, use `gcov' to produce human
4246 readable information from the `.gcno' and `.gcda' files.
4247 Refer to the `gcov' documentation for further information.
4250 With `-fprofile-arcs', for each function of your program GCC
4251 creates a program flow graph, then finds a spanning tree for the
4252 graph. Only arcs that are not on the spanning tree have to be
4253 instrumented: the compiler adds code to count the number of times
4254 that these arcs are executed. When an arc is the only exit or
4255 only entrance to a block, the instrumentation code can be added to
4256 the block; otherwise, a new basic block must be created to hold
4257 the instrumentation code.
4260 Produce a notes file that the `gcov' code-coverage utility (*note
4261 `gcov'--a Test Coverage Program: Gcov.) can use to show program
4262 coverage. Each source file's note file is called `AUXNAME.gcno'.
4263 Refer to the `-fprofile-arcs' option above for a description of
4264 AUXNAME and instructions on how to generate test coverage data.
4265 Coverage data will match the source files more closely, if you do
4269 Print the name and the counter upperbound for all debug counters.
4271 `-fdbg-cnt=COUNTER-VALUE-LIST'
4272 Set the internal debug counter upperbound. COUNTER-VALUE-LIST is a
4273 comma-separated list of NAME:VALUE pairs which sets the upperbound
4274 of each debug counter NAME to VALUE. All debug counters have the
4275 initial upperbound of UINT_MAX, thus dbg_cnt() returns true always
4276 unless the upperbound is set by this option. e.g. With
4277 -fdbg-cnt=dce:10,tail_call:0 dbg_cnt(dce) will return true only
4278 for first 10 invocations and dbg_cnt(tail_call) will return false
4283 Says to make debugging dumps during compilation at times specified
4284 by LETTERS. This is used for debugging the RTL-based passes of the
4285 compiler. The file names for most of the dumps are made by
4286 appending a pass number and a word to the DUMPNAME, and the files
4287 are created in the directory of the output file. DUMPNAME is
4288 generated from the name of the output file, if explicitly specified
4289 and it is not an executable, otherwise it is the basename of the
4290 source file. These switches may have different effects when `-E'
4291 is used for preprocessing.
4293 Debug dumps can be enabled with a `-fdump-rtl' switch or some `-d'
4294 option LETTERS. Here are the possible letters for use in PASS and
4295 LETTERS, and their meanings:
4297 `-fdump-rtl-alignments'
4298 Dump after branch alignments have been computed.
4300 `-fdump-rtl-asmcons'
4301 Dump after fixing rtl statements that have unsatisfied in/out
4304 `-fdump-rtl-auto_inc_dec'
4305 Dump after auto-inc-dec discovery. This pass is only run on
4306 architectures that have auto inc or auto dec instructions.
4308 `-fdump-rtl-barriers'
4309 Dump after cleaning up the barrier instructions.
4312 Dump after partitioning hot and cold basic blocks.
4315 Dump after block reordering.
4319 `-fdump-rtl-btl1' and `-fdump-rtl-btl2' enable dumping after
4320 the two branch target load optimization passes.
4323 Dump after jump bypassing and control flow optimizations.
4325 `-fdump-rtl-combine'
4326 Dump after the RTL instruction combination pass.
4328 `-fdump-rtl-compgotos'
4329 Dump after duplicating the computed gotos.
4334 `-fdump-rtl-ce1', `-fdump-rtl-ce2', and `-fdump-rtl-ce3'
4335 enable dumping after the three if conversion passes.
4337 `-fdump-rtl-cprop_hardreg'
4338 Dump after hard register copy propagation.
4341 Dump after combining stack adjustments.
4345 `-fdump-rtl-cse1' and `-fdump-rtl-cse2' enable dumping after
4346 the two common sub-expression elimination passes.
4349 Dump after the standalone dead code elimination passes.
4352 Dump after delayed branch scheduling.
4356 `-fdump-rtl-dce1' and `-fdump-rtl-dce2' enable dumping after
4357 the two dead store elimination passes.
4360 Dump after finalization of EH handling code.
4362 `-fdump-rtl-eh_ranges'
4363 Dump after conversion of EH handling range regions.
4366 Dump after RTL generation.
4368 `-fdump-rtl-fwprop1'
4369 `-fdump-rtl-fwprop2'
4370 `-fdump-rtl-fwprop1' and `-fdump-rtl-fwprop2' enable dumping
4371 after the two forward propagation passes.
4375 `-fdump-rtl-gcse1' and `-fdump-rtl-gcse2' enable dumping
4376 after global common subexpression elimination.
4378 `-fdump-rtl-init-regs'
4379 Dump after the initialization of the registers.
4381 `-fdump-rtl-initvals'
4382 Dump after the computation of the initial value sets.
4384 `-fdump-rtl-into_cfglayout'
4385 Dump after converting to cfglayout mode.
4388 Dump after iterated register allocation.
4391 Dump after the second jump optimization.
4394 `-fdump-rtl-loop2' enables dumping after the rtl loop
4395 optimization passes.
4398 Dump after performing the machine dependent reorganization
4399 pass, if that pass exists.
4401 `-fdump-rtl-mode_sw'
4402 Dump after removing redundant mode switches.
4405 Dump after register renumbering.
4407 `-fdump-rtl-outof_cfglayout'
4408 Dump after converting from cfglayout mode.
4410 `-fdump-rtl-peephole2'
4411 Dump after the peephole pass.
4413 `-fdump-rtl-postreload'
4414 Dump after post-reload optimizations.
4416 `-fdump-rtl-pro_and_epilogue'
4417 Dump after generating the function pro and epilogues.
4419 `-fdump-rtl-regmove'
4420 Dump after the register move pass.
4424 `-fdump-rtl-sched1' and `-fdump-rtl-sched2' enable dumping
4425 after the basic block scheduling passes.
4428 Dump after sign extension elimination.
4430 `-fdump-rtl-seqabstr'
4431 Dump after common sequence discovery.
4433 `-fdump-rtl-shorten'
4434 Dump after shortening branches.
4436 `-fdump-rtl-sibling'
4437 Dump after sibling call optimizations.
4444 `-fdump-rtl-split1', `-fdump-rtl-split2',
4445 `-fdump-rtl-split3', `-fdump-rtl-split4' and
4446 `-fdump-rtl-split5' enable dumping after five rounds of
4447 instruction splitting.
4450 Dump after modulo scheduling. This pass is only run on some
4454 Dump after conversion from GCC's "flat register file"
4455 registers to the x87's stack-like registers. This pass is
4456 only run on x86 variants.
4458 `-fdump-rtl-subreg1'
4459 `-fdump-rtl-subreg2'
4460 `-fdump-rtl-subreg1' and `-fdump-rtl-subreg2' enable dumping
4461 after the two subreg expansion passes.
4463 `-fdump-rtl-unshare'
4464 Dump after all rtl has been unshared.
4466 `-fdump-rtl-vartrack'
4467 Dump after variable tracking.
4470 Dump after converting virtual registers to hard registers.
4473 Dump after live range splitting.
4475 `-fdump-rtl-regclass'
4476 `-fdump-rtl-subregs_of_mode_init'
4477 `-fdump-rtl-subregs_of_mode_finish'
4479 `-fdump-rtl-dfinish'
4480 These dumps are defined but always produce empty files.
4483 Produce all the dumps listed above.
4486 Annotate the assembler output with miscellaneous debugging
4490 Dump all macro definitions, at the end of preprocessing, in
4491 addition to normal output.
4494 Produce a core dump whenever an error occurs.
4497 Print statistics on memory usage, at the end of the run, to
4501 Annotate the assembler output with a comment indicating which
4502 pattern and alternative was used. The length of each
4503 instruction is also printed.
4506 Dump the RTL in the assembler output as a comment before each
4507 instruction. Also turns on `-dp' annotation.
4510 For each of the other indicated dump files
4511 (`-fdump-rtl-PASS'), dump a representation of the control
4512 flow graph suitable for viewing with VCG to `FILE.PASS.vcg'.
4515 Just generate RTL for a function instead of compiling it.
4516 Usually used with `-fdump-rtl-expand'.
4519 Dump debugging information during parsing, to standard error.
4522 When doing debugging dumps, suppress address output. This makes
4523 it more feasible to use diff on debugging dumps for compiler
4524 invocations with different compiler binaries and/or different text
4525 / bss / data / heap / stack / dso start locations.
4528 When doing debugging dumps, suppress instruction numbers and
4529 address output. This makes it more feasible to use diff on
4530 debugging dumps for compiler invocations with different options,
4531 in particular with and without `-g'.
4533 `-fdump-translation-unit (C++ only)'
4534 `-fdump-translation-unit-OPTIONS (C++ only)'
4535 Dump a representation of the tree structure for the entire
4536 translation unit to a file. The file name is made by appending
4537 `.tu' to the source file name, and the file is created in the same
4538 directory as the output file. If the `-OPTIONS' form is used,
4539 OPTIONS controls the details of the dump as described for the
4540 `-fdump-tree' options.
4542 `-fdump-class-hierarchy (C++ only)'
4543 `-fdump-class-hierarchy-OPTIONS (C++ only)'
4544 Dump a representation of each class's hierarchy and virtual
4545 function table layout to a file. The file name is made by
4546 appending `.class' to the source file name, and the file is
4547 created in the same directory as the output file. If the
4548 `-OPTIONS' form is used, OPTIONS controls the details of the dump
4549 as described for the `-fdump-tree' options.
4552 Control the dumping at various stages of inter-procedural analysis
4553 language tree to a file. The file name is generated by appending a
4554 switch specific suffix to the source file name, and the file is
4555 created in the same directory as the output file. The following
4559 Enables all inter-procedural analysis dumps.
4562 Dumps information about call-graph optimization, unused
4563 function removal, and inlining decisions.
4566 Dump after function inlining.
4569 `-fdump-statistics-OPTION'
4570 Enable and control dumping of pass statistics in a separate file.
4571 The file name is generated by appending a suffix ending in
4572 `.statistics' to the source file name, and the file is created in
4573 the same directory as the output file. If the `-OPTION' form is
4574 used, `-stats' will cause counters to be summed over the whole
4575 compilation unit while `-details' will dump every event as the
4576 passes generate them. The default with no option is to sum
4577 counters for each function compiled.
4579 `-fdump-tree-SWITCH'
4580 `-fdump-tree-SWITCH-OPTIONS'
4581 Control the dumping at various stages of processing the
4582 intermediate language tree to a file. The file name is generated
4583 by appending a switch specific suffix to the source file name, and
4584 the file is created in the same directory as the output file. If
4585 the `-OPTIONS' form is used, OPTIONS is a list of `-' separated
4586 options that control the details of the dump. Not all options are
4587 applicable to all dumps, those which are not meaningful will be
4588 ignored. The following options are available
4591 Print the address of each node. Usually this is not
4592 meaningful as it changes according to the environment and
4593 source file. Its primary use is for tying up a dump file
4594 with a debug environment.
4597 Inhibit dumping of members of a scope or body of a function
4598 merely because that scope has been reached. Only dump such
4599 items when they are directly reachable by some other path.
4600 When dumping pretty-printed trees, this option inhibits
4601 dumping the bodies of control structures.
4604 Print a raw representation of the tree. By default, trees are
4605 pretty-printed into a C-like representation.
4608 Enable more detailed dumps (not honored by every dump option).
4611 Enable dumping various statistics about the pass (not honored
4612 by every dump option).
4615 Enable showing basic block boundaries (disabled in raw dumps).
4618 Enable showing virtual operands for every statement.
4621 Enable showing line numbers for statements.
4624 Enable showing the unique ID (`DECL_UID') for each variable.
4627 Enable showing the tree dump for each statement.
4630 Turn on all options, except `raw', `slim', `verbose' and
4633 The following tree dumps are possible:
4635 Dump before any tree based optimization, to `FILE.original'.
4638 Dump after all tree based optimization, to `FILE.optimized'.
4641 Dump each function before and after the gimplification pass
4642 to a file. The file name is made by appending `.gimple' to
4643 the source file name.
4646 Dump the control flow graph of each function to a file. The
4647 file name is made by appending `.cfg' to the source file name.
4650 Dump the control flow graph of each function to a file in VCG
4651 format. The file name is made by appending `.vcg' to the
4652 source file name. Note that if the file contains more than
4653 one function, the generated file cannot be used directly by
4654 VCG. You will need to cut and paste each function's graph
4655 into its own separate file first.
4658 Dump each function after copying loop headers. The file name
4659 is made by appending `.ch' to the source file name.
4662 Dump SSA related information to a file. The file name is
4663 made by appending `.ssa' to the source file name.
4666 Dump aliasing information for each function. The file name
4667 is made by appending `.alias' to the source file name.
4670 Dump each function after CCP. The file name is made by
4671 appending `.ccp' to the source file name.
4674 Dump each function after STORE-CCP. The file name is made by
4675 appending `.storeccp' to the source file name.
4678 Dump trees after partial redundancy elimination. The file
4679 name is made by appending `.pre' to the source file name.
4682 Dump trees after full redundancy elimination. The file name
4683 is made by appending `.fre' to the source file name.
4686 Dump trees after copy propagation. The file name is made by
4687 appending `.copyprop' to the source file name.
4690 Dump trees after store copy-propagation. The file name is
4691 made by appending `.store_copyprop' to the source file name.
4694 Dump each function after dead code elimination. The file
4695 name is made by appending `.dce' to the source file name.
4698 Dump each function after adding mudflap instrumentation. The
4699 file name is made by appending `.mudflap' to the source file
4703 Dump each function after performing scalar replacement of
4704 aggregates. The file name is made by appending `.sra' to the
4708 Dump each function after performing code sinking. The file
4709 name is made by appending `.sink' to the source file name.
4712 Dump each function after applying dominator tree
4713 optimizations. The file name is made by appending `.dom' to
4714 the source file name.
4717 Dump each function after applying dead store elimination.
4718 The file name is made by appending `.dse' to the source file
4722 Dump each function after optimizing PHI nodes into
4723 straightline code. The file name is made by appending
4724 `.phiopt' to the source file name.
4727 Dump each function after forward propagating single use
4728 variables. The file name is made by appending `.forwprop' to
4729 the source file name.
4732 Dump each function after applying the copy rename
4733 optimization. The file name is made by appending
4734 `.copyrename' to the source file name.
4737 Dump each function after applying the named return value
4738 optimization on generic trees. The file name is made by
4739 appending `.nrv' to the source file name.
4742 Dump each function after applying vectorization of loops.
4743 The file name is made by appending `.vect' to the source file
4747 Dump each function after Value Range Propagation (VRP). The
4748 file name is made by appending `.vrp' to the source file name.
4751 Enable all the available tree dumps with the flags provided
4754 `-ftree-vectorizer-verbose=N'
4755 This option controls the amount of debugging output the vectorizer
4756 prints. This information is written to standard error, unless
4757 `-fdump-tree-all' or `-fdump-tree-vect' is specified, in which
4758 case it is output to the usual dump listing file, `.vect'. For
4759 N=0 no diagnostic information is reported. If N=1 the vectorizer
4760 reports each loop that got vectorized, and the total number of
4761 loops that got vectorized. If N=2 the vectorizer also reports
4762 non-vectorized loops that passed the first analysis phase
4763 (vect_analyze_loop_form) - i.e. countable, inner-most, single-bb,
4764 single-entry/exit loops. This is the same verbosity level that
4765 `-fdump-tree-vect-stats' uses. Higher verbosity levels mean
4766 either more information dumped for each reported loop, or same
4767 amount of information reported for more loops: If N=3, alignment
4768 related information is added to the reports. If N=4,
4769 data-references related information (e.g. memory dependences,
4770 memory access-patterns) is added to the reports. If N=5, the
4771 vectorizer reports also non-vectorized inner-most loops that did
4772 not pass the first analysis phase (i.e., may not be countable, or
4773 may have complicated control-flow). If N=6, the vectorizer
4774 reports also non-vectorized nested loops. For N=7, all the
4775 information the vectorizer generates during its analysis and
4776 transformation is reported. This is the same verbosity level that
4777 `-fdump-tree-vect-details' uses.
4779 `-frandom-seed=STRING'
4780 This option provides a seed that GCC uses when it would otherwise
4781 use random numbers. It is used to generate certain symbol names
4782 that have to be different in every compiled file. It is also used
4783 to place unique stamps in coverage data files and the object files
4784 that produce them. You can use the `-frandom-seed' option to
4785 produce reproducibly identical object files.
4787 The STRING should be different for every file you compile.
4790 On targets that use instruction scheduling, this option controls
4791 the amount of debugging output the scheduler prints. This
4792 information is written to standard error, unless
4793 `-fdump-rtl-sched1' or `-fdump-rtl-sched2' is specified, in which
4794 case it is output to the usual dump listing file, `.sched' or
4795 `.sched2' respectively. However for N greater than nine, the
4796 output is always printed to standard error.
4798 For N greater than zero, `-fsched-verbose' outputs the same
4799 information as `-fdump-rtl-sched1' and `-fdump-rtl-sched2'. For N
4800 greater than one, it also output basic block probabilities,
4801 detailed ready list information and unit/insn info. For N greater
4802 than two, it includes RTL at abort point, control-flow and regions
4803 info. And for N over four, `-fsched-verbose' also includes
4807 Store the usual "temporary" intermediate files permanently; place
4808 them in the current directory and name them based on the source
4809 file. Thus, compiling `foo.c' with `-c -save-temps' would produce
4810 files `foo.i' and `foo.s', as well as `foo.o'. This creates a
4811 preprocessed `foo.i' output file even though the compiler now
4812 normally uses an integrated preprocessor.
4814 When used in combination with the `-x' command line option,
4815 `-save-temps' is sensible enough to avoid over writing an input
4816 source file with the same extension as an intermediate file. The
4817 corresponding intermediate file may be obtained by renaming the
4818 source file before using `-save-temps'.
4821 Report the CPU time taken by each subprocess in the compilation
4822 sequence. For C source files, this is the compiler proper and
4823 assembler (plus the linker if linking is done). The output looks
4829 The first number on each line is the "user time", that is time
4830 spent executing the program itself. The second number is "system
4831 time", time spent executing operating system routines on behalf of
4832 the program. Both numbers are in seconds.
4835 Run variable tracking pass. It computes where variables are
4836 stored at each position in code. Better debugging information is
4837 then generated (if the debugging information format supports this
4840 It is enabled by default when compiling with optimization (`-Os',
4841 `-O', `-O2', ...), debugging information (`-g') and the debug info
4844 `-print-file-name=LIBRARY'
4845 Print the full absolute name of the library file LIBRARY that
4846 would be used when linking--and don't do anything else. With this
4847 option, GCC does not compile or link anything; it just prints the
4850 `-print-multi-directory'
4851 Print the directory name corresponding to the multilib selected by
4852 any other switches present in the command line. This directory is
4853 supposed to exist in `GCC_EXEC_PREFIX'.
4856 Print the mapping from multilib directory names to compiler
4857 switches that enable them. The directory name is separated from
4858 the switches by `;', and each switch starts with an `@' instead of
4859 the `-', without spaces between multiple switches. This is
4860 supposed to ease shell-processing.
4862 `-print-prog-name=PROGRAM'
4863 Like `-print-file-name', but searches for a program such as `cpp'.
4865 `-print-libgcc-file-name'
4866 Same as `-print-file-name=libgcc.a'.
4868 This is useful when you use `-nostdlib' or `-nodefaultlibs' but
4869 you do want to link with `libgcc.a'. You can do
4871 gcc -nostdlib FILES... `gcc -print-libgcc-file-name`
4873 `-print-search-dirs'
4874 Print the name of the configured installation directory and a list
4875 of program and library directories `gcc' will search--and don't do
4878 This is useful when `gcc' prints the error message `installation
4879 problem, cannot exec cpp0: No such file or directory'. To resolve
4880 this you either need to put `cpp0' and the other compiler
4881 components where `gcc' expects to find them, or you can set the
4882 environment variable `GCC_EXEC_PREFIX' to the directory where you
4883 installed them. Don't forget the trailing `/'. *Note Environment
4887 Print the target sysroot directory that will be used during
4888 compilation. This is the target sysroot specified either at
4889 configure time or using the `--sysroot' option, possibly with an
4890 extra suffix that depends on compilation options. If no target
4891 sysroot is specified, the option prints nothing.
4893 `-print-sysroot-headers-suffix'
4894 Print the suffix added to the target sysroot when searching for
4895 headers, or give an error if the compiler is not configured with
4896 such a suffix--and don't do anything else.
4899 Print the compiler's target machine (for example,
4900 `i686-pc-linux-gnu')--and don't do anything else.
4903 Print the compiler version (for example, `3.0')--and don't do
4907 Print the compiler's built-in specs--and don't do anything else.
4908 (This is used when GCC itself is being built.) *Note Spec Files::.
4910 `-feliminate-unused-debug-types'
4911 Normally, when producing DWARF2 output, GCC will emit debugging
4912 information for all types declared in a compilation unit,
4913 regardless of whether or not they are actually used in that
4914 compilation unit. Sometimes this is useful, such as if, in the
4915 debugger, you want to cast a value to a type that is not actually
4916 used in your program (but is declared). More often, however, this
4917 results in a significant amount of wasted space. With this
4918 option, GCC will avoid producing debug symbol output for types
4919 that are nowhere used in the source file being compiled.
4922 File: gcc.info, Node: Optimize Options, Next: Preprocessor Options, Prev: Debugging Options, Up: Invoking GCC
4924 3.10 Options That Control Optimization
4925 ======================================
4927 These options control various sorts of optimizations.
4929 Without any optimization option, the compiler's goal is to reduce the
4930 cost of compilation and to make debugging produce the expected results.
4931 Statements are independent: if you stop the program with a breakpoint
4932 between statements, you can then assign a new value to any variable or
4933 change the program counter to any other statement in the function and
4934 get exactly the results you would expect from the source code.
4936 Turning on optimization flags makes the compiler attempt to improve
4937 the performance and/or code size at the expense of compilation time and
4938 possibly the ability to debug the program.
4940 The compiler performs optimization based on the knowledge it has of the
4941 program. Compiling multiple files at once to a single output file mode
4942 allows the compiler to use information gained from all of the files
4943 when compiling each of them.
4945 Not all optimizations are controlled directly by a flag. Only
4946 optimizations that have a flag are listed.
4950 Optimize. Optimizing compilation takes somewhat more time, and a
4951 lot more memory for a large function.
4953 With `-O', the compiler tries to reduce code size and execution
4954 time, without performing any optimizations that take a great deal
4955 of compilation time.
4957 `-O' turns on the following optimization flags:
4964 -fguess-branch-probability
4967 -finline-small-functions
4972 -ftree-builtin-call-dce
4977 -ftree-dominator-opts
4984 `-O' also turns on `-fomit-frame-pointer' on machines where doing
4985 so does not interfere with debugging.
4988 Optimize even more. GCC performs nearly all supported
4989 optimizations that do not involve a space-speed tradeoff. As
4990 compared to `-O', this option increases both compilation time and
4991 the performance of the generated code.
4993 `-O2' turns on all optimization flags specified by `-O'. It also
4994 turns on the following optimization flags:
4996 -falign-functions -falign-jumps
4997 -falign-loops -falign-labels
5000 -fcse-follow-jumps -fcse-skip-blocks
5001 -fdelete-null-pointer-checks
5002 -fexpensive-optimizations
5005 -foptimize-sibling-calls
5008 -freorder-blocks -freorder-functions
5009 -frerun-cse-after-loop
5010 -fsched-interblock -fsched-spec
5011 -fschedule-insns -fschedule-insns2
5012 -fstrict-aliasing -fstrict-overflow
5013 -ftree-switch-conversion
5017 Please note the warning under `-fgcse' about invoking `-O2' on
5018 programs that use computed gotos.
5021 Optimize yet more. `-O3' turns on all optimizations specified by
5022 `-O2' and also turns on the `-finline-functions',
5023 `-funswitch-loops', `-fpredictive-commoning',
5024 `-fgcse-after-reload' and `-ftree-vectorize' options.
5027 Reduce compilation time and make debugging produce the expected
5028 results. This is the default.
5031 Optimize for size. `-Os' enables all `-O2' optimizations that do
5032 not typically increase code size. It also performs further
5033 optimizations designed to reduce code size.
5035 `-Os' disables the following optimization flags:
5036 -falign-functions -falign-jumps -falign-loops
5037 -falign-labels -freorder-blocks -freorder-blocks-and-partition
5038 -fprefetch-loop-arrays -ftree-vect-loop-version
5040 If you use multiple `-O' options, with or without level numbers,
5041 the last such option is the one that is effective.
5043 Options of the form `-fFLAG' specify machine-independent flags. Most
5044 flags have both positive and negative forms; the negative form of
5045 `-ffoo' would be `-fno-foo'. In the table below, only one of the forms
5046 is listed--the one you typically will use. You can figure out the
5047 other form by either removing `no-' or adding it.
5049 The following options control specific optimizations. They are either
5050 activated by `-O' options or are related to ones that are. You can use
5051 the following flags in the rare cases when "fine-tuning" of
5052 optimizations to be performed is desired.
5054 `-fno-default-inline'
5055 Do not make member functions inline by default merely because they
5056 are defined inside the class scope (C++ only). Otherwise, when
5057 you specify `-O', member functions defined inside class scope are
5058 compiled inline by default; i.e., you don't need to add `inline'
5059 in front of the member function name.
5062 Always pop the arguments to each function call as soon as that
5063 function returns. For machines which must pop arguments after a
5064 function call, the compiler normally lets arguments accumulate on
5065 the stack for several function calls and pops them all at once.
5067 Disabled at levels `-O', `-O2', `-O3', `-Os'.
5069 `-fforward-propagate'
5070 Perform a forward propagation pass on RTL. The pass tries to
5071 combine two instructions and checks if the result can be
5072 simplified. If loop unrolling is active, two passes are performed
5073 and the second is scheduled after loop unrolling.
5075 This option is enabled by default at optimization levels `-O2',
5078 `-fomit-frame-pointer'
5079 Don't keep the frame pointer in a register for functions that
5080 don't need one. This avoids the instructions to save, set up and
5081 restore frame pointers; it also makes an extra register available
5082 in many functions. *It also makes debugging impossible on some
5085 On some machines, such as the VAX, this flag has no effect, because
5086 the standard calling sequence automatically handles the frame
5087 pointer and nothing is saved by pretending it doesn't exist. The
5088 machine-description macro `FRAME_POINTER_REQUIRED' controls
5089 whether a target machine supports this flag. *Note Register
5090 Usage: (gccint)Registers.
5092 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5094 `-foptimize-sibling-calls'
5095 Optimize sibling and tail recursive calls.
5097 Enabled at levels `-O2', `-O3', `-Os'.
5100 Don't pay attention to the `inline' keyword. Normally this option
5101 is used to keep the compiler from expanding any functions inline.
5102 Note that if you are not optimizing, no functions can be expanded
5105 `-finline-small-functions'
5106 Integrate functions into their callers when their body is smaller
5107 than expected function call code (so overall size of program gets
5108 smaller). The compiler heuristically decides which functions are
5109 simple enough to be worth integrating in this way.
5111 Enabled at level `-O2'.
5113 `-findirect-inlining'
5114 Inline also indirect calls that are discovered to be known at
5115 compile time thanks to previous inlining. This option has any
5116 effect only when inlining itself is turned on by the
5117 `-finline-functions' or `-finline-small-functions' options.
5119 Enabled at level `-O2'.
5121 `-finline-functions'
5122 Integrate all simple functions into their callers. The compiler
5123 heuristically decides which functions are simple enough to be worth
5124 integrating in this way.
5126 If all calls to a given function are integrated, and the function
5127 is declared `static', then the function is normally not output as
5128 assembler code in its own right.
5130 Enabled at level `-O3'.
5132 `-finline-functions-called-once'
5133 Consider all `static' functions called once for inlining into their
5134 caller even if they are not marked `inline'. If a call to a given
5135 function is integrated, then the function is not output as
5136 assembler code in its own right.
5138 Enabled at levels `-O1', `-O2', `-O3' and `-Os'.
5141 Inline functions marked by `always_inline' and functions whose
5142 body seems smaller than the function call overhead early before
5143 doing `-fprofile-generate' instrumentation and real inlining pass.
5144 Doing so makes profiling significantly cheaper and usually
5145 inlining faster on programs having large chains of nested wrapper
5151 By default, GCC limits the size of functions that can be inlined.
5152 This flag allows coarse control of this limit. N is the size of
5153 functions that can be inlined in number of pseudo instructions.
5155 Inlining is actually controlled by a number of parameters, which
5156 may be specified individually by using `--param NAME=VALUE'. The
5157 `-finline-limit=N' option sets some of these parameters as follows:
5159 `max-inline-insns-single'
5162 `max-inline-insns-auto'
5165 See below for a documentation of the individual parameters
5166 controlling inlining and for the defaults of these parameters.
5168 _Note:_ there may be no value to `-finline-limit' that results in
5171 _Note:_ pseudo instruction represents, in this particular context,
5172 an abstract measurement of function's size. In no way does it
5173 represent a count of assembly instructions and as such its exact
5174 meaning might change from one release to an another.
5176 `-fkeep-inline-functions'
5177 In C, emit `static' functions that are declared `inline' into the
5178 object file, even if the function has been inlined into all of its
5179 callers. This switch does not affect functions using the `extern
5180 inline' extension in GNU C89. In C++, emit any and all inline
5181 functions into the object file.
5183 `-fkeep-static-consts'
5184 Emit variables declared `static const' when optimization isn't
5185 turned on, even if the variables aren't referenced.
5187 GCC enables this option by default. If you want to force the
5188 compiler to check if the variable was referenced, regardless of
5189 whether or not optimization is turned on, use the
5190 `-fno-keep-static-consts' option.
5193 Attempt to merge identical constants (string constants and
5194 floating point constants) across compilation units.
5196 This option is the default for optimized compilation if the
5197 assembler and linker support it. Use `-fno-merge-constants' to
5198 inhibit this behavior.
5200 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5202 `-fmerge-all-constants'
5203 Attempt to merge identical constants and identical variables.
5205 This option implies `-fmerge-constants'. In addition to
5206 `-fmerge-constants' this considers e.g. even constant initialized
5207 arrays or initialized constant variables with integral or floating
5208 point types. Languages like C or C++ require each variable,
5209 including multiple instances of the same variable in recursive
5210 calls, to have distinct locations, so using this option will
5211 result in non-conforming behavior.
5214 Perform swing modulo scheduling immediately before the first
5215 scheduling pass. This pass looks at innermost loops and reorders
5216 their instructions by overlapping different iterations.
5218 `-fmodulo-sched-allow-regmoves'
5219 Perform more aggressive SMS based modulo scheduling with register
5220 moves allowed. By setting this flag certain anti-dependences
5221 edges will be deleted which will trigger the generation of
5222 reg-moves based on the life-range analysis. This option is
5223 effective only with `-fmodulo-sched' enabled.
5225 `-fno-branch-count-reg'
5226 Do not use "decrement and branch" instructions on a count register,
5227 but instead generate a sequence of instructions that decrement a
5228 register, compare it against zero, then branch based upon the
5229 result. This option is only meaningful on architectures that
5230 support such instructions, which include x86, PowerPC, IA-64 and
5233 The default is `-fbranch-count-reg'.
5236 Do not put function addresses in registers; make each instruction
5237 that calls a constant function contain the function's address
5240 This option results in less efficient code, but some strange hacks
5241 that alter the assembler output may be confused by the
5242 optimizations performed when this option is not used.
5244 The default is `-ffunction-cse'
5246 `-fno-zero-initialized-in-bss'
5247 If the target supports a BSS section, GCC by default puts
5248 variables that are initialized to zero into BSS. This can save
5249 space in the resulting code.
5251 This option turns off this behavior because some programs
5252 explicitly rely on variables going to the data section. E.g., so
5253 that the resulting executable can find the beginning of that
5254 section and/or make assumptions based on that.
5256 The default is `-fzero-initialized-in-bss'.
5258 `-fmudflap -fmudflapth -fmudflapir'
5259 For front-ends that support it (C and C++), instrument all risky
5260 pointer/array dereferencing operations, some standard library
5261 string/heap functions, and some other associated constructs with
5262 range/validity tests. Modules so instrumented should be immune to
5263 buffer overflows, invalid heap use, and some other classes of C/C++
5264 programming errors. The instrumentation relies on a separate
5265 runtime library (`libmudflap'), which will be linked into a
5266 program if `-fmudflap' is given at link time. Run-time behavior
5267 of the instrumented program is controlled by the `MUDFLAP_OPTIONS'
5268 environment variable. See `env MUDFLAP_OPTIONS=-help a.out' for
5271 Use `-fmudflapth' instead of `-fmudflap' to compile and to link if
5272 your program is multi-threaded. Use `-fmudflapir', in addition to
5273 `-fmudflap' or `-fmudflapth', if instrumentation should ignore
5274 pointer reads. This produces less instrumentation (and therefore
5275 faster execution) and still provides some protection against
5276 outright memory corrupting writes, but allows erroneously read
5277 data to propagate within a program.
5280 Perform optimizations where we check to see if a jump branches to a
5281 location where another comparison subsumed by the first is found.
5282 If so, the first branch is redirected to either the destination of
5283 the second branch or a point immediately following it, depending
5284 on whether the condition is known to be true or false.
5286 Enabled at levels `-O2', `-O3', `-Os'.
5288 `-fsplit-wide-types'
5289 When using a type that occupies multiple registers, such as `long
5290 long' on a 32-bit system, split the registers apart and allocate
5291 them independently. This normally generates better code for those
5292 types, but may make debugging more difficult.
5294 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5296 `-fcse-follow-jumps'
5297 In common subexpression elimination (CSE), scan through jump
5298 instructions when the target of the jump is not reached by any
5299 other path. For example, when CSE encounters an `if' statement
5300 with an `else' clause, CSE will follow the jump when the condition
5303 Enabled at levels `-O2', `-O3', `-Os'.
5306 This is similar to `-fcse-follow-jumps', but causes CSE to follow
5307 jumps which conditionally skip over blocks. When CSE encounters a
5308 simple `if' statement with no else clause, `-fcse-skip-blocks'
5309 causes CSE to follow the jump around the body of the `if'.
5311 Enabled at levels `-O2', `-O3', `-Os'.
5313 `-frerun-cse-after-loop'
5314 Re-run common subexpression elimination after loop optimizations
5317 Enabled at levels `-O2', `-O3', `-Os'.
5320 Perform a global common subexpression elimination pass. This pass
5321 also performs global constant and copy propagation.
5323 _Note:_ When compiling a program using computed gotos, a GCC
5324 extension, you may get better runtime performance if you disable
5325 the global common subexpression elimination pass by adding
5326 `-fno-gcse' to the command line.
5328 Enabled at levels `-O2', `-O3', `-Os'.
5331 When `-fgcse-lm' is enabled, global common subexpression
5332 elimination will attempt to move loads which are only killed by
5333 stores into themselves. This allows a loop containing a
5334 load/store sequence to be changed to a load outside the loop, and
5335 a copy/store within the loop.
5337 Enabled by default when gcse is enabled.
5340 When `-fgcse-sm' is enabled, a store motion pass is run after
5341 global common subexpression elimination. This pass will attempt
5342 to move stores out of loops. When used in conjunction with
5343 `-fgcse-lm', loops containing a load/store sequence can be changed
5344 to a load before the loop and a store after the loop.
5346 Not enabled at any optimization level.
5349 When `-fgcse-las' is enabled, the global common subexpression
5350 elimination pass eliminates redundant loads that come after stores
5351 to the same memory location (both partial and full redundancies).
5353 Not enabled at any optimization level.
5355 `-fgcse-after-reload'
5356 When `-fgcse-after-reload' is enabled, a redundant load elimination
5357 pass is performed after reload. The purpose of this pass is to
5358 cleanup redundant spilling.
5360 `-funsafe-loop-optimizations'
5361 If given, the loop optimizer will assume that loop indices do not
5362 overflow, and that the loops with nontrivial exit condition are not
5363 infinite. This enables a wider range of loop optimizations even if
5364 the loop optimizer itself cannot prove that these assumptions are
5365 valid. Using `-Wunsafe-loop-optimizations', the compiler will
5366 warn you if it finds this kind of loop.
5369 Perform cross-jumping transformation. This transformation unifies
5370 equivalent code and save code size. The resulting code may or may
5371 not perform better than without cross-jumping.
5373 Enabled at levels `-O2', `-O3', `-Os'.
5376 Combine increments or decrements of addresses with memory accesses.
5377 This pass is always skipped on architectures that do not have
5378 instructions to support this. Enabled by default at `-O' and
5379 higher on architectures that support this.
5382 Perform dead code elimination (DCE) on RTL. Enabled by default at
5386 Perform dead store elimination (DSE) on RTL. Enabled by default
5390 Attempt to transform conditional jumps into branch-less
5391 equivalents. This include use of conditional moves, min, max, set
5392 flags and abs instructions, and some tricks doable by standard
5393 arithmetics. The use of conditional execution on chips where it
5394 is available is controlled by `if-conversion2'.
5396 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5399 Use conditional execution (where available) to transform
5400 conditional jumps into branch-less equivalents.
5402 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5404 `-fdelete-null-pointer-checks'
5405 Use global dataflow analysis to identify and eliminate useless
5406 checks for null pointers. The compiler assumes that dereferencing
5407 a null pointer would have halted the program. If a pointer is
5408 checked after it has already been dereferenced, it cannot be null.
5410 In some environments, this assumption is not true, and programs can
5411 safely dereference null pointers. Use
5412 `-fno-delete-null-pointer-checks' to disable this optimization for
5413 programs which depend on that behavior.
5415 Enabled at levels `-O2', `-O3', `-Os'.
5417 `-fexpensive-optimizations'
5418 Perform a number of minor optimizations that are relatively
5421 Enabled at levels `-O2', `-O3', `-Os'.
5423 `-foptimize-register-move'
5425 Attempt to reassign register numbers in move instructions and as
5426 operands of other simple instructions in order to maximize the
5427 amount of register tying. This is especially helpful on machines
5428 with two-operand instructions.
5430 Note `-fregmove' and `-foptimize-register-move' are the same
5433 Enabled at levels `-O2', `-O3', `-Os'.
5435 `-fira-algorithm=ALGORITHM'
5436 Use specified coloring algorithm for the integrated register
5437 allocator. The ALGORITHM argument should be `priority' or `CB'.
5438 The first algorithm specifies Chow's priority coloring, the second
5439 one specifies Chaitin-Briggs coloring. The second algorithm can
5440 be unimplemented for some architectures. If it is implemented, it
5441 is the default because Chaitin-Briggs coloring as a rule generates
5444 `-fira-region=REGION'
5445 Use specified regions for the integrated register allocator. The
5446 REGION argument should be one of `all', `mixed', or `one'. The
5447 first value means using all loops as register allocation regions,
5448 the second value which is the default means using all loops except
5449 for loops with small register pressure as the regions, and third
5450 one means using all function as a single region. The first value
5451 can give best result for machines with small size and irregular
5452 register set, the third one results in faster and generates decent
5453 code and the smallest size code, and the default value usually
5454 give the best results in most cases and for most architectures.
5457 Do optimistic register coalescing. This option might be
5458 profitable for architectures with big regular register files.
5460 `-fno-ira-share-save-slots'
5461 Switch off sharing stack slots used for saving call used hard
5462 registers living through a call. Each hard register will get a
5463 separate stack slot and as a result function stack frame will be
5466 `-fno-ira-share-spill-slots'
5467 Switch off sharing stack slots allocated for pseudo-registers.
5468 Each pseudo-register which did not get a hard register will get a
5469 separate stack slot and as a result function stack frame will be
5473 Set up how verbose dump file for the integrated register allocator
5474 will be. Default value is 5. If the value is greater or equal to
5475 10, the dump file will be stderr as if the value were N minus 10.
5478 If supported for the target machine, attempt to reorder
5479 instructions to exploit instruction slots available after delayed
5480 branch instructions.
5482 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5485 If supported for the target machine, attempt to reorder
5486 instructions to eliminate execution stalls due to required data
5487 being unavailable. This helps machines that have slow floating
5488 point or memory load instructions by allowing other instructions
5489 to be issued until the result of the load or floating point
5490 instruction is required.
5492 Enabled at levels `-O2', `-O3', `-Os'.
5495 Similar to `-fschedule-insns', but requests an additional pass of
5496 instruction scheduling after register allocation has been done.
5497 This is especially useful on machines with a relatively small
5498 number of registers and where memory load instructions take more
5501 Enabled at levels `-O2', `-O3', `-Os'.
5503 `-fno-sched-interblock'
5504 Don't schedule instructions across basic blocks. This is normally
5505 enabled by default when scheduling before register allocation, i.e.
5506 with `-fschedule-insns' or at `-O2' or higher.
5509 Don't allow speculative motion of non-load instructions. This is
5510 normally enabled by default when scheduling before register
5511 allocation, i.e. with `-fschedule-insns' or at `-O2' or higher.
5514 Allow speculative motion of some load instructions. This only
5515 makes sense when scheduling before register allocation, i.e. with
5516 `-fschedule-insns' or at `-O2' or higher.
5518 `-fsched-spec-load-dangerous'
5519 Allow speculative motion of more load instructions. This only
5520 makes sense when scheduling before register allocation, i.e. with
5521 `-fschedule-insns' or at `-O2' or higher.
5523 `-fsched-stalled-insns'
5524 `-fsched-stalled-insns=N'
5525 Define how many insns (if any) can be moved prematurely from the
5526 queue of stalled insns into the ready list, during the second
5527 scheduling pass. `-fno-sched-stalled-insns' means that no insns
5528 will be moved prematurely, `-fsched-stalled-insns=0' means there
5529 is no limit on how many queued insns can be moved prematurely.
5530 `-fsched-stalled-insns' without a value is equivalent to
5531 `-fsched-stalled-insns=1'.
5533 `-fsched-stalled-insns-dep'
5534 `-fsched-stalled-insns-dep=N'
5535 Define how many insn groups (cycles) will be examined for a
5536 dependency on a stalled insn that is candidate for premature
5537 removal from the queue of stalled insns. This has an effect only
5538 during the second scheduling pass, and only if
5539 `-fsched-stalled-insns' is used. `-fno-sched-stalled-insns-dep'
5540 is equivalent to `-fsched-stalled-insns-dep=0'.
5541 `-fsched-stalled-insns-dep' without a value is equivalent to
5542 `-fsched-stalled-insns-dep=1'.
5544 `-fsched2-use-superblocks'
5545 When scheduling after register allocation, do use superblock
5546 scheduling algorithm. Superblock scheduling allows motion across
5547 basic block boundaries resulting on faster schedules. This option
5548 is experimental, as not all machine descriptions used by GCC model
5549 the CPU closely enough to avoid unreliable results from the
5552 This only makes sense when scheduling after register allocation,
5553 i.e. with `-fschedule-insns2' or at `-O2' or higher.
5555 `-fsched2-use-traces'
5556 Use `-fsched2-use-superblocks' algorithm when scheduling after
5557 register allocation and additionally perform code duplication in
5558 order to increase the size of superblocks using tracer pass. See
5559 `-ftracer' for details on trace formation.
5561 This mode should produce faster but significantly longer programs.
5562 Also without `-fbranch-probabilities' the traces constructed may
5563 not match the reality and hurt the performance. This only makes
5564 sense when scheduling after register allocation, i.e. with
5565 `-fschedule-insns2' or at `-O2' or higher.
5568 Eliminate redundant sign extension instructions and move the
5569 non-redundant ones to optimal placement using lazy code motion
5572 `-freschedule-modulo-scheduled-loops'
5573 The modulo scheduling comes before the traditional scheduling, if
5574 a loop was modulo scheduled we may want to prevent the later
5575 scheduling passes from changing its schedule, we use this option
5578 `-fselective-scheduling'
5579 Schedule instructions using selective scheduling algorithm.
5580 Selective scheduling runs instead of the first scheduler pass.
5582 `-fselective-scheduling2'
5583 Schedule instructions using selective scheduling algorithm.
5584 Selective scheduling runs instead of the second scheduler pass.
5586 `-fsel-sched-pipelining'
5587 Enable software pipelining of innermost loops during selective
5588 scheduling. This option has no effect until one of
5589 `-fselective-scheduling' or `-fselective-scheduling2' is turned on.
5591 `-fsel-sched-pipelining-outer-loops'
5592 When pipelining loops during selective scheduling, also pipeline
5593 outer loops. This option has no effect until
5594 `-fsel-sched-pipelining' is turned on.
5597 Enable values to be allocated in registers that will be clobbered
5598 by function calls, by emitting extra instructions to save and
5599 restore the registers around such calls. Such allocation is done
5600 only when it seems to result in better code than would otherwise
5603 This option is always enabled by default on certain machines,
5604 usually those which have no call-preserved registers to use
5607 Enabled at levels `-O2', `-O3', `-Os'.
5610 Attempt to minimize stack usage. The compiler will attempt to use
5611 less stack space, even if that makes the program slower. This
5612 option implies setting the `large-stack-frame' parameter to 100
5613 and the `large-stack-frame-growth' parameter to 400.
5616 Perform reassociation on trees. This flag is enabled by default
5620 Perform partial redundancy elimination (PRE) on trees. This flag
5621 is enabled by default at `-O2' and `-O3'.
5624 Perform full redundancy elimination (FRE) on trees. The difference
5625 between FRE and PRE is that FRE only considers expressions that
5626 are computed on all paths leading to the redundant computation.
5627 This analysis is faster than PRE, though it exposes fewer
5628 redundancies. This flag is enabled by default at `-O' and higher.
5631 Perform copy propagation on trees. This pass eliminates
5632 unnecessary copy operations. This flag is enabled by default at
5636 Discover which functions are pure or constant. Enabled by default
5640 Discover which static variables do not escape cannot escape the
5641 compilation unit. Enabled by default at `-O' and higher.
5643 `-fipa-struct-reorg'
5644 Perform structure reorganization optimization, that change C-like
5645 structures layout in order to better utilize spatial locality.
5646 This transformation is affective for programs containing arrays of
5647 structures. Available in two compilation modes: profile-based
5648 (enabled with `-fprofile-generate') or static (which uses built-in
5649 heuristics). Require `-fipa-type-escape' to provide the safety of
5650 this transformation. It works only in whole program mode, so it
5651 requires `-fwhole-program' and `-combine' to be enabled.
5652 Structures considered `cold' by this transformation are not
5653 affected (see `--param struct-reorg-cold-struct-ratio=VALUE').
5655 With this flag, the program debug info reflects a new structure
5659 Perform interprocedural pointer analysis. This option is
5660 experimental and does not affect generated code.
5663 Perform interprocedural constant propagation. This optimization
5664 analyzes the program to determine when values passed to functions
5665 are constants and then optimizes accordingly. This optimization
5666 can substantially increase performance if the application has
5667 constants passed to functions. This flag is enabled by default at
5668 `-O2', `-Os' and `-O3'.
5671 Perform function cloning to make interprocedural constant
5672 propagation stronger. When enabled, interprocedural constant
5673 propagation will perform function cloning when externally visible
5674 function can be called with constant arguments. Because this
5675 optimization can create multiple copies of functions, it may
5676 significantly increase code size (see `--param
5677 ipcp-unit-growth=VALUE'). This flag is enabled by default at
5680 `-fipa-matrix-reorg'
5681 Perform matrix flattening and transposing. Matrix flattening
5682 tries to replace a m-dimensional matrix with its equivalent
5683 n-dimensional matrix, where n < m. This reduces the level of
5684 indirection needed for accessing the elements of the matrix. The
5685 second optimization is matrix transposing that attempts to change
5686 the order of the matrix's dimensions in order to improve cache
5687 locality. Both optimizations need the `-fwhole-program' flag.
5688 Transposing is enabled only if profiling information is available.
5691 Perform forward store motion on trees. This flag is enabled by
5692 default at `-O' and higher.
5695 Perform sparse conditional constant propagation (CCP) on trees.
5696 This pass only operates on local scalar variables and is enabled
5697 by default at `-O' and higher.
5699 `-ftree-switch-conversion'
5700 Perform conversion of simple initializations in a switch to
5701 initializations from a scalar array. This flag is enabled by
5702 default at `-O2' and higher.
5705 Perform dead code elimination (DCE) on trees. This flag is
5706 enabled by default at `-O' and higher.
5708 `-ftree-builtin-call-dce'
5709 Perform conditional dead code elimination (DCE) for calls to
5710 builtin functions that may set `errno' but are otherwise
5711 side-effect free. This flag is enabled by default at `-O2' and
5712 higher if `-Os' is not also specified.
5714 `-ftree-dominator-opts'
5715 Perform a variety of simple scalar cleanups (constant/copy
5716 propagation, redundancy elimination, range propagation and
5717 expression simplification) based on a dominator tree traversal.
5718 This also performs jump threading (to reduce jumps to jumps). This
5719 flag is enabled by default at `-O' and higher.
5722 Perform dead store elimination (DSE) on trees. A dead store is a
5723 store into a memory location which will later be overwritten by
5724 another store without any intervening loads. In this case the
5725 earlier store can be deleted. This flag is enabled by default at
5729 Perform loop header copying on trees. This is beneficial since it
5730 increases effectiveness of code motion optimizations. It also
5731 saves one jump. This flag is enabled by default at `-O' and
5732 higher. It is not enabled for `-Os', since it usually increases
5735 `-ftree-loop-optimize'
5736 Perform loop optimizations on trees. This flag is enabled by
5737 default at `-O' and higher.
5739 `-ftree-loop-linear'
5740 Perform linear loop transformations on tree. This flag can
5741 improve cache performance and allow further loop optimizations to
5744 `-floop-interchange'
5745 Perform loop interchange transformations on loops. Interchanging
5746 two nested loops switches the inner and outer loops. For example,
5750 A(J, I) = A(J, I) * C
5753 loop interchange will transform the loop as if the user had
5757 A(J, I) = A(J, I) * C
5760 which can be beneficial when `N' is larger than the caches,
5761 because in Fortran, the elements of an array are stored in memory
5762 contiguously by column, and the original loop iterates over rows,
5763 potentially creating at each access a cache miss. This
5764 optimization applies to all the languages supported by GCC and is
5765 not limited to Fortran. To use this code transformation, GCC has
5766 to be configured with `--with-ppl' and `--with-cloog' to enable the
5767 Graphite loop transformation infrastructure.
5770 Perform loop strip mining transformations on loops. Strip mining
5771 splits a loop into two nested loops. The outer loop has strides
5772 equal to the strip size and the inner loop has strides of the
5773 original loop within a strip. For example, given a loop like:
5777 loop strip mining will transform the loop as if the user had
5780 DO I = II, min (II + 3, N)
5784 This optimization applies to all the languages supported by GCC
5785 and is not limited to Fortran. To use this code transformation,
5786 GCC has to be configured with `--with-ppl' and `--with-cloog' to
5787 enable the Graphite loop transformation infrastructure.
5790 Perform loop blocking transformations on loops. Blocking strip
5791 mines each loop in the loop nest such that the memory accesses of
5792 the element loops fit inside caches. For example, given a loop
5796 A(J, I) = B(I) + C(J)
5799 loop blocking will transform the loop as if the user had written:
5802 DO I = II, min (II + 63, N)
5803 DO J = JJ, min (JJ + 63, M)
5804 A(J, I) = B(I) + C(J)
5809 which can be beneficial when `M' is larger than the caches,
5810 because the innermost loop will iterate over a smaller amount of
5811 data that can be kept in the caches. This optimization applies to
5812 all the languages supported by GCC and is not limited to Fortran.
5813 To use this code transformation, GCC has to be configured with
5814 `--with-ppl' and `--with-cloog' to enable the Graphite loop
5815 transformation infrastructure.
5818 Compare the results of several data dependence analyzers. This
5819 option is used for debugging the data dependence analyzers.
5821 `-ftree-loop-distribution'
5822 Perform loop distribution. This flag can improve cache
5823 performance on big loop bodies and allow further loop
5824 optimizations, like parallelization or vectorization, to take
5825 place. For example, the loop
5839 Perform loop invariant motion on trees. This pass moves only
5840 invariants that would be hard to handle at RTL level (function
5841 calls, operations that expand to nontrivial sequences of insns).
5842 With `-funswitch-loops' it also moves operands of conditions that
5843 are invariant out of the loop, so that we can use just trivial
5844 invariantness analysis in loop unswitching. The pass also includes
5847 `-ftree-loop-ivcanon'
5848 Create a canonical counter for number of iterations in the loop
5849 for that determining number of iterations requires complicated
5850 analysis. Later optimizations then may determine the number
5851 easily. Useful especially in connection with unrolling.
5854 Perform induction variable optimizations (strength reduction,
5855 induction variable merging and induction variable elimination) on
5858 `-ftree-parallelize-loops=n'
5859 Parallelize loops, i.e., split their iteration space to run in n
5860 threads. This is only possible for loops whose iterations are
5861 independent and can be arbitrarily reordered. The optimization is
5862 only profitable on multiprocessor machines, for loops that are
5863 CPU-intensive, rather than constrained e.g. by memory bandwidth.
5864 This option implies `-pthread', and thus is only supported on
5865 targets that have support for `-pthread'.
5868 Perform scalar replacement of aggregates. This pass replaces
5869 structure references with scalars to prevent committing structures
5870 to memory too early. This flag is enabled by default at `-O' and
5874 Perform copy renaming on trees. This pass attempts to rename
5875 compiler temporaries to other variables at copy locations, usually
5876 resulting in variable names which more closely resemble the
5877 original variables. This flag is enabled by default at `-O' and
5881 Perform temporary expression replacement during the SSA->normal
5882 phase. Single use/single def temporaries are replaced at their
5883 use location with their defining expression. This results in
5884 non-GIMPLE code, but gives the expanders much more complex trees
5885 to work on resulting in better RTL generation. This is enabled by
5886 default at `-O' and higher.
5889 Perform loop vectorization on trees. This flag is enabled by
5892 `-ftree-vect-loop-version'
5893 Perform loop versioning when doing loop vectorization on trees.
5894 When a loop appears to be vectorizable except that data alignment
5895 or data dependence cannot be determined at compile time then
5896 vectorized and non-vectorized versions of the loop are generated
5897 along with runtime checks for alignment or dependence to control
5898 which version is executed. This option is enabled by default
5899 except at level `-Os' where it is disabled.
5902 Enable cost model for vectorization.
5905 Perform Value Range Propagation on trees. This is similar to the
5906 constant propagation pass, but instead of values, ranges of values
5907 are propagated. This allows the optimizers to remove unnecessary
5908 range checks like array bound checks and null pointer checks.
5909 This is enabled by default at `-O2' and higher. Null pointer check
5910 elimination is only done if `-fdelete-null-pointer-checks' is
5914 Perform tail duplication to enlarge superblock size. This
5915 transformation simplifies the control flow of the function
5916 allowing other optimizations to do better job.
5919 Unroll loops whose number of iterations can be determined at
5920 compile time or upon entry to the loop. `-funroll-loops' implies
5921 `-frerun-cse-after-loop'. This option makes code larger, and may
5922 or may not make it run faster.
5924 `-funroll-all-loops'
5925 Unroll all loops, even if their number of iterations is uncertain
5926 when the loop is entered. This usually makes programs run more
5927 slowly. `-funroll-all-loops' implies the same options as
5930 `-fsplit-ivs-in-unroller'
5931 Enables expressing of values of induction variables in later
5932 iterations of the unrolled loop using the value in the first
5933 iteration. This breaks long dependency chains, thus improving
5934 efficiency of the scheduling passes.
5936 Combination of `-fweb' and CSE is often sufficient to obtain the
5937 same effect. However in cases the loop body is more complicated
5938 than a single basic block, this is not reliable. It also does not
5939 work at all on some of the architectures due to restrictions in
5942 This optimization is enabled by default.
5944 `-fvariable-expansion-in-unroller'
5945 With this option, the compiler will create multiple copies of some
5946 local variables when unrolling a loop which can result in superior
5949 `-fpredictive-commoning'
5950 Perform predictive commoning optimization, i.e., reusing
5951 computations (especially memory loads and stores) performed in
5952 previous iterations of loops.
5954 This option is enabled at level `-O3'.
5956 `-fprefetch-loop-arrays'
5957 If supported by the target machine, generate instructions to
5958 prefetch memory to improve the performance of loops that access
5961 This option may generate better or worse code; results are highly
5962 dependent on the structure of loops within the source code.
5964 Disabled at level `-Os'.
5968 Disable any machine-specific peephole optimizations. The
5969 difference between `-fno-peephole' and `-fno-peephole2' is in how
5970 they are implemented in the compiler; some targets use one, some
5971 use the other, a few use both.
5973 `-fpeephole' is enabled by default. `-fpeephole2' enabled at
5974 levels `-O2', `-O3', `-Os'.
5976 `-fno-guess-branch-probability'
5977 Do not guess branch probabilities using heuristics.
5979 GCC will use heuristics to guess branch probabilities if they are
5980 not provided by profiling feedback (`-fprofile-arcs'). These
5981 heuristics are based on the control flow graph. If some branch
5982 probabilities are specified by `__builtin_expect', then the
5983 heuristics will be used to guess branch probabilities for the rest
5984 of the control flow graph, taking the `__builtin_expect' info into
5985 account. The interactions between the heuristics and
5986 `__builtin_expect' can be complex, and in some cases, it may be
5987 useful to disable the heuristics so that the effects of
5988 `__builtin_expect' are easier to understand.
5990 The default is `-fguess-branch-probability' at levels `-O', `-O2',
5994 Reorder basic blocks in the compiled function in order to reduce
5995 number of taken branches and improve code locality.
5997 Enabled at levels `-O2', `-O3'.
5999 `-freorder-blocks-and-partition'
6000 In addition to reordering basic blocks in the compiled function,
6001 in order to reduce number of taken branches, partitions hot and
6002 cold basic blocks into separate sections of the assembly and .o
6003 files, to improve paging and cache locality performance.
6005 This optimization is automatically turned off in the presence of
6006 exception handling, for linkonce sections, for functions with a
6007 user-defined section attribute and on any architecture that does
6008 not support named sections.
6010 `-freorder-functions'
6011 Reorder functions in the object file in order to improve code
6012 locality. This is implemented by using special subsections
6013 `.text.hot' for most frequently executed functions and
6014 `.text.unlikely' for unlikely executed functions. Reordering is
6015 done by the linker so object file format must support named
6016 sections and linker must place them in a reasonable way.
6018 Also profile feedback must be available in to make this option
6019 effective. See `-fprofile-arcs' for details.
6021 Enabled at levels `-O2', `-O3', `-Os'.
6024 Allows the compiler to assume the strictest aliasing rules
6025 applicable to the language being compiled. For C (and C++), this
6026 activates optimizations based on the type of expressions. In
6027 particular, an object of one type is assumed never to reside at
6028 the same address as an object of a different type, unless the
6029 types are almost the same. For example, an `unsigned int' can
6030 alias an `int', but not a `void*' or a `double'. A character type
6031 may alias any other type.
6033 Pay special attention to code like this:
6044 The practice of reading from a different union member than the one
6045 most recently written to (called "type-punning") is common. Even
6046 with `-fstrict-aliasing', type-punning is allowed, provided the
6047 memory is accessed through the union type. So, the code above
6048 will work as expected. *Note Structures unions enumerations and
6049 bit-fields implementation::. However, this code might not:
6058 Similarly, access by taking the address, casting the resulting
6059 pointer and dereferencing the result has undefined behavior, even
6060 if the cast uses a union type, e.g.:
6063 return ((union a_union *) &d)->i;
6066 The `-fstrict-aliasing' option is enabled at levels `-O2', `-O3',
6070 Allow the compiler to assume strict signed overflow rules,
6071 depending on the language being compiled. For C (and C++) this
6072 means that overflow when doing arithmetic with signed numbers is
6073 undefined, which means that the compiler may assume that it will
6074 not happen. This permits various optimizations. For example, the
6075 compiler will assume that an expression like `i + 10 > i' will
6076 always be true for signed `i'. This assumption is only valid if
6077 signed overflow is undefined, as the expression is false if `i +
6078 10' overflows when using twos complement arithmetic. When this
6079 option is in effect any attempt to determine whether an operation
6080 on signed numbers will overflow must be written carefully to not
6081 actually involve overflow.
6083 This option also allows the compiler to assume strict pointer
6084 semantics: given a pointer to an object, if adding an offset to
6085 that pointer does not produce a pointer to the same object, the
6086 addition is undefined. This permits the compiler to conclude that
6087 `p + u > p' is always true for a pointer `p' and unsigned integer
6088 `u'. This assumption is only valid because pointer wraparound is
6089 undefined, as the expression is false if `p + u' overflows using
6090 twos complement arithmetic.
6092 See also the `-fwrapv' option. Using `-fwrapv' means that integer
6093 signed overflow is fully defined: it wraps. When `-fwrapv' is
6094 used, there is no difference between `-fstrict-overflow' and
6095 `-fno-strict-overflow' for integers. With `-fwrapv' certain types
6096 of overflow are permitted. For example, if the compiler gets an
6097 overflow when doing arithmetic on constants, the overflowed value
6098 can still be used with `-fwrapv', but not otherwise.
6100 The `-fstrict-overflow' option is enabled at levels `-O2', `-O3',
6104 `-falign-functions=N'
6105 Align the start of functions to the next power-of-two greater than
6106 N, skipping up to N bytes. For instance, `-falign-functions=32'
6107 aligns functions to the next 32-byte boundary, but
6108 `-falign-functions=24' would align to the next 32-byte boundary
6109 only if this can be done by skipping 23 bytes or less.
6111 `-fno-align-functions' and `-falign-functions=1' are equivalent
6112 and mean that functions will not be aligned.
6114 Some assemblers only support this flag when N is a power of two;
6115 in that case, it is rounded up.
6117 If N is not specified or is zero, use a machine-dependent default.
6119 Enabled at levels `-O2', `-O3'.
6123 Align all branch targets to a power-of-two boundary, skipping up to
6124 N bytes like `-falign-functions'. This option can easily make
6125 code slower, because it must insert dummy operations for when the
6126 branch target is reached in the usual flow of the code.
6128 `-fno-align-labels' and `-falign-labels=1' are equivalent and mean
6129 that labels will not be aligned.
6131 If `-falign-loops' or `-falign-jumps' are applicable and are
6132 greater than this value, then their values are used instead.
6134 If N is not specified or is zero, use a machine-dependent default
6135 which is very likely to be `1', meaning no alignment.
6137 Enabled at levels `-O2', `-O3'.
6141 Align loops to a power-of-two boundary, skipping up to N bytes
6142 like `-falign-functions'. The hope is that the loop will be
6143 executed many times, which will make up for any execution of the
6146 `-fno-align-loops' and `-falign-loops=1' are equivalent and mean
6147 that loops will not be aligned.
6149 If N is not specified or is zero, use a machine-dependent default.
6151 Enabled at levels `-O2', `-O3'.
6155 Align branch targets to a power-of-two boundary, for branch targets
6156 where the targets can only be reached by jumping, skipping up to N
6157 bytes like `-falign-functions'. In this case, no dummy operations
6160 `-fno-align-jumps' and `-falign-jumps=1' are equivalent and mean
6161 that loops will not be aligned.
6163 If N is not specified or is zero, use a machine-dependent default.
6165 Enabled at levels `-O2', `-O3'.
6168 This option is left for compatibility reasons. `-funit-at-a-time'
6169 has no effect, while `-fno-unit-at-a-time' implies
6170 `-fno-toplevel-reorder' and `-fno-section-anchors'.
6174 `-fno-toplevel-reorder'
6175 Do not reorder top-level functions, variables, and `asm'
6176 statements. Output them in the same order that they appear in the
6177 input file. When this option is used, unreferenced static
6178 variables will not be removed. This option is intended to support
6179 existing code which relies on a particular ordering. For new
6180 code, it is better to use attributes.
6182 Enabled at level `-O0'. When disabled explicitly, it also imply
6183 `-fno-section-anchors' that is otherwise enabled at `-O0' on some
6187 Constructs webs as commonly used for register allocation purposes
6188 and assign each web individual pseudo register. This allows the
6189 register allocation pass to operate on pseudos directly, but also
6190 strengthens several other optimization passes, such as CSE, loop
6191 optimizer and trivial dead code remover. It can, however, make
6192 debugging impossible, since variables will no longer stay in a
6195 Enabled by default with `-funroll-loops'.
6198 Assume that the current compilation unit represents whole program
6199 being compiled. All public functions and variables with the
6200 exception of `main' and those merged by attribute
6201 `externally_visible' become static functions and in a affect gets
6202 more aggressively optimized by interprocedural optimizers. While
6203 this option is equivalent to proper use of `static' keyword for
6204 programs consisting of single file, in combination with option
6205 `--combine' this flag can be used to compile most of smaller scale
6206 C programs since the functions and variables become local for the
6207 whole combined compilation unit, not for the single source file
6210 This option is not supported for Fortran programs.
6213 After register allocation and post-register allocation instruction
6214 splitting, we perform a copy-propagation pass to try to reduce
6215 scheduling dependencies and occasionally eliminate the copy.
6217 Enabled at levels `-O', `-O2', `-O3', `-Os'.
6219 `-fprofile-correction'
6220 Profiles collected using an instrumented binary for multi-threaded
6221 programs may be inconsistent due to missed counter updates. When
6222 this option is specified, GCC will use heuristics to correct or
6223 smooth out such inconsistencies. By default, GCC will emit an
6224 error message when an inconsistent profile is detected.
6226 `-fprofile-dir=PATH'
6227 Set the directory to search the profile data files in to PATH.
6228 This option affects only the profile data generated by
6229 `-fprofile-generate', `-ftest-coverage', `-fprofile-arcs' and used
6230 by `-fprofile-use' and `-fbranch-probabilities' and its related
6231 options. By default, GCC will use the current directory as PATH
6232 thus the profile data file will appear in the same directory as
6235 `-fprofile-generate'
6236 `-fprofile-generate=PATH'
6237 Enable options usually used for instrumenting application to
6238 produce profile useful for later recompilation with profile
6239 feedback based optimization. You must use `-fprofile-generate'
6240 both when compiling and when linking your program.
6242 The following options are enabled: `-fprofile-arcs',
6243 `-fprofile-values', `-fvpt'.
6245 If PATH is specified, GCC will look at the PATH to find the
6246 profile feedback data files. See `-fprofile-dir'.
6249 `-fprofile-use=PATH'
6250 Enable profile feedback directed optimizations, and optimizations
6251 generally profitable only with profile feedback available.
6253 The following options are enabled: `-fbranch-probabilities',
6254 `-fvpt', `-funroll-loops', `-fpeel-loops'
6256 By default, GCC emits an error message if the feedback profiles do
6257 not match the source code. This error can be turned into a
6258 warning by using `-Wcoverage-mismatch'. Note this may result in
6259 poorly optimized code.
6261 If PATH is specified, GCC will look at the PATH to find the
6262 profile feedback data files. See `-fprofile-dir'.
6265 Perform dynamic inter-procedural analysis. This is used in
6266 conjunction with the `-fprofile-generate' and `-fprofile-use'
6267 options. During the `-fprofile-generate' phase, this flag turns
6268 on some additional instrumentation code that enables dynamic
6269 call-graph analysis. During the `-fprofile-use' phase, this flag
6270 enables cross-module optimizations such as inlining.
6272 The following options control compiler behavior regarding floating
6273 point arithmetic. These options trade off between speed and
6274 correctness. All must be specifically enabled.
6277 Do not store floating point variables in registers, and inhibit
6278 other options that might change whether a floating point value is
6279 taken from a register or memory.
6281 This option prevents undesirable excess precision on machines such
6282 as the 68000 where the floating registers (of the 68881) keep more
6283 precision than a `double' is supposed to have. Similarly for the
6284 x86 architecture. For most programs, the excess precision does
6285 only good, but a few programs rely on the precise definition of
6286 IEEE floating point. Use `-ffloat-store' for such programs, after
6287 modifying them to store all pertinent intermediate computations
6291 Sets `-fno-math-errno', `-funsafe-math-optimizations',
6292 `-ffinite-math-only', `-fno-rounding-math', `-fno-signaling-nans'
6293 and `-fcx-limited-range'.
6295 This option causes the preprocessor macro `__FAST_MATH__' to be
6298 This option is not turned on by any `-O' option since it can
6299 result in incorrect output for programs which depend on an exact
6300 implementation of IEEE or ISO rules/specifications for math
6301 functions. It may, however, yield faster code for programs that do
6302 not require the guarantees of these specifications.
6305 Do not set ERRNO after calling math functions that are executed
6306 with a single instruction, e.g., sqrt. A program that relies on
6307 IEEE exceptions for math error handling may want to use this flag
6308 for speed while maintaining IEEE arithmetic compatibility.
6310 This option is not turned on by any `-O' option since it can
6311 result in incorrect output for programs which depend on an exact
6312 implementation of IEEE or ISO rules/specifications for math
6313 functions. It may, however, yield faster code for programs that do
6314 not require the guarantees of these specifications.
6316 The default is `-fmath-errno'.
6318 On Darwin systems, the math library never sets `errno'. There is
6319 therefore no reason for the compiler to consider the possibility
6320 that it might, and `-fno-math-errno' is the default.
6322 `-funsafe-math-optimizations'
6323 Allow optimizations for floating-point arithmetic that (a) assume
6324 that arguments and results are valid and (b) may violate IEEE or
6325 ANSI standards. When used at link-time, it may include libraries
6326 or startup files that change the default FPU control word or other
6327 similar optimizations.
6329 This option is not turned on by any `-O' option since it can
6330 result in incorrect output for programs which depend on an exact
6331 implementation of IEEE or ISO rules/specifications for math
6332 functions. It may, however, yield faster code for programs that do
6333 not require the guarantees of these specifications. Enables
6334 `-fno-signed-zeros', `-fno-trapping-math', `-fassociative-math'
6335 and `-freciprocal-math'.
6337 The default is `-fno-unsafe-math-optimizations'.
6339 `-fassociative-math'
6340 Allow re-association of operands in series of floating-point
6341 operations. This violates the ISO C and C++ language standard by
6342 possibly changing computation result. NOTE: re-ordering may
6343 change the sign of zero as well as ignore NaNs and inhibit or
6344 create underflow or overflow (and thus cannot be used on a code
6345 which relies on rounding behavior like `(x + 2**52) - 2**52)'.
6346 May also reorder floating-point comparisons and thus may not be
6347 used when ordered comparisons are required. This option requires
6348 that both `-fno-signed-zeros' and `-fno-trapping-math' be in
6349 effect. Moreover, it doesn't make much sense with
6352 The default is `-fno-associative-math'.
6355 Allow the reciprocal of a value to be used instead of dividing by
6356 the value if this enables optimizations. For example `x / y' can
6357 be replaced with `x * (1/y)' which is useful if `(1/y)' is subject
6358 to common subexpression elimination. Note that this loses
6359 precision and increases the number of flops operating on the value.
6361 The default is `-fno-reciprocal-math'.
6363 `-ffinite-math-only'
6364 Allow optimizations for floating-point arithmetic that assume that
6365 arguments and results are not NaNs or +-Infs.
6367 This option is not turned on by any `-O' option since it can
6368 result in incorrect output for programs which depend on an exact
6369 implementation of IEEE or ISO rules/specifications for math
6370 functions. It may, however, yield faster code for programs that do
6371 not require the guarantees of these specifications.
6373 The default is `-fno-finite-math-only'.
6376 Allow optimizations for floating point arithmetic that ignore the
6377 signedness of zero. IEEE arithmetic specifies the behavior of
6378 distinct +0.0 and -0.0 values, which then prohibits simplification
6379 of expressions such as x+0.0 or 0.0*x (even with
6380 `-ffinite-math-only'). This option implies that the sign of a
6381 zero result isn't significant.
6383 The default is `-fsigned-zeros'.
6385 `-fno-trapping-math'
6386 Compile code assuming that floating-point operations cannot
6387 generate user-visible traps. These traps include division by
6388 zero, overflow, underflow, inexact result and invalid operation.
6389 This option requires that `-fno-signaling-nans' be in effect.
6390 Setting this option may allow faster code if one relies on
6391 "non-stop" IEEE arithmetic, for example.
6393 This option should never be turned on by any `-O' option since it
6394 can result in incorrect output for programs which depend on an
6395 exact implementation of IEEE or ISO rules/specifications for math
6398 The default is `-ftrapping-math'.
6401 Disable transformations and optimizations that assume default
6402 floating point rounding behavior. This is round-to-zero for all
6403 floating point to integer conversions, and round-to-nearest for
6404 all other arithmetic truncations. This option should be specified
6405 for programs that change the FP rounding mode dynamically, or that
6406 may be executed with a non-default rounding mode. This option
6407 disables constant folding of floating point expressions at
6408 compile-time (which may be affected by rounding mode) and
6409 arithmetic transformations that are unsafe in the presence of
6410 sign-dependent rounding modes.
6412 The default is `-fno-rounding-math'.
6414 This option is experimental and does not currently guarantee to
6415 disable all GCC optimizations that are affected by rounding mode.
6416 Future versions of GCC may provide finer control of this setting
6417 using C99's `FENV_ACCESS' pragma. This command line option will
6418 be used to specify the default state for `FENV_ACCESS'.
6421 Compile code assuming that IEEE signaling NaNs may generate
6422 user-visible traps during floating-point operations. Setting this
6423 option disables optimizations that may change the number of
6424 exceptions visible with signaling NaNs. This option implies
6427 This option causes the preprocessor macro `__SUPPORT_SNAN__' to be
6430 The default is `-fno-signaling-nans'.
6432 This option is experimental and does not currently guarantee to
6433 disable all GCC optimizations that affect signaling NaN behavior.
6435 `-fsingle-precision-constant'
6436 Treat floating point constant as single precision constant instead
6437 of implicitly converting it to double precision constant.
6439 `-fcx-limited-range'
6440 When enabled, this option states that a range reduction step is not
6441 needed when performing complex division. Also, there is no
6442 checking whether the result of a complex multiplication or
6443 division is `NaN + I*NaN', with an attempt to rescue the situation
6444 in that case. The default is `-fno-cx-limited-range', but is
6445 enabled by `-ffast-math'.
6447 This option controls the default setting of the ISO C99
6448 `CX_LIMITED_RANGE' pragma. Nevertheless, the option applies to
6451 `-fcx-fortran-rules'
6452 Complex multiplication and division follow Fortran rules. Range
6453 reduction is done as part of complex division, but there is no
6454 checking whether the result of a complex multiplication or
6455 division is `NaN + I*NaN', with an attempt to rescue the situation
6458 The default is `-fno-cx-fortran-rules'.
6461 The following options control optimizations that may improve
6462 performance, but are not enabled by any `-O' options. This section
6463 includes experimental options that may produce broken code.
6465 `-fbranch-probabilities'
6466 After running a program compiled with `-fprofile-arcs' (*note
6467 Options for Debugging Your Program or `gcc': Debugging Options.),
6468 you can compile it a second time using `-fbranch-probabilities',
6469 to improve optimizations based on the number of times each branch
6470 was taken. When the program compiled with `-fprofile-arcs' exits
6471 it saves arc execution counts to a file called `SOURCENAME.gcda'
6472 for each source file. The information in this data file is very
6473 dependent on the structure of the generated code, so you must use
6474 the same source code and the same optimization options for both
6477 With `-fbranch-probabilities', GCC puts a `REG_BR_PROB' note on
6478 each `JUMP_INSN' and `CALL_INSN'. These can be used to improve
6479 optimization. Currently, they are only used in one place: in
6480 `reorg.c', instead of guessing which path a branch is mostly to
6481 take, the `REG_BR_PROB' values are used to exactly determine which
6482 path is taken more often.
6485 If combined with `-fprofile-arcs', it adds code so that some data
6486 about values of expressions in the program is gathered.
6488 With `-fbranch-probabilities', it reads back the data gathered
6489 from profiling values of expressions and adds `REG_VALUE_PROFILE'
6490 notes to instructions for their later usage in optimizations.
6492 Enabled with `-fprofile-generate' and `-fprofile-use'.
6495 If combined with `-fprofile-arcs', it instructs the compiler to add
6496 a code to gather information about values of expressions.
6498 With `-fbranch-probabilities', it reads back the data gathered and
6499 actually performs the optimizations based on them. Currently the
6500 optimizations include specialization of division operation using
6501 the knowledge about the value of the denominator.
6503 `-frename-registers'
6504 Attempt to avoid false dependencies in scheduled code by making use
6505 of registers left over after register allocation. This
6506 optimization will most benefit processors with lots of registers.
6507 Depending on the debug information format adopted by the target,
6508 however, it can make debugging impossible, since variables will no
6509 longer stay in a "home register".
6511 Enabled by default with `-funroll-loops'.
6514 Perform tail duplication to enlarge superblock size. This
6515 transformation simplifies the control flow of the function
6516 allowing other optimizations to do better job.
6518 Enabled with `-fprofile-use'.
6521 Unroll loops whose number of iterations can be determined at
6522 compile time or upon entry to the loop. `-funroll-loops' implies
6523 `-frerun-cse-after-loop', `-fweb' and `-frename-registers'. It
6524 also turns on complete loop peeling (i.e. complete removal of
6525 loops with small constant number of iterations). This option
6526 makes code larger, and may or may not make it run faster.
6528 Enabled with `-fprofile-use'.
6530 `-funroll-all-loops'
6531 Unroll all loops, even if their number of iterations is uncertain
6532 when the loop is entered. This usually makes programs run more
6533 slowly. `-funroll-all-loops' implies the same options as
6537 Peels the loops for that there is enough information that they do
6538 not roll much (from profile feedback). It also turns on complete
6539 loop peeling (i.e. complete removal of loops with small constant
6540 number of iterations).
6542 Enabled with `-fprofile-use'.
6544 `-fmove-loop-invariants'
6545 Enables the loop invariant motion pass in the RTL loop optimizer.
6546 Enabled at level `-O1'
6549 Move branches with loop invariant conditions out of the loop, with
6550 duplicates of the loop on both branches (modified according to
6551 result of the condition).
6553 `-ffunction-sections'
6555 Place each function or data item into its own section in the output
6556 file if the target supports arbitrary sections. The name of the
6557 function or the name of the data item determines the section's name
6560 Use these options on systems where the linker can perform
6561 optimizations to improve locality of reference in the instruction
6562 space. Most systems using the ELF object format and SPARC
6563 processors running Solaris 2 have linkers with such optimizations.
6564 AIX may have these optimizations in the future.
6566 Only use these options when there are significant benefits from
6567 doing so. When you specify these options, the assembler and
6568 linker will create larger object and executable files and will
6569 also be slower. You will not be able to use `gprof' on all
6570 systems if you specify this option and you may have problems with
6571 debugging if you specify both this option and `-g'.
6573 `-fbranch-target-load-optimize'
6574 Perform branch target register load optimization before prologue /
6575 epilogue threading. The use of target registers can typically be
6576 exposed only during reload, thus hoisting loads out of loops and
6577 doing inter-block scheduling needs a separate optimization pass.
6579 `-fbranch-target-load-optimize2'
6580 Perform branch target register load optimization after prologue /
6583 `-fbtr-bb-exclusive'
6584 When performing branch target register load optimization, don't
6585 reuse branch target registers in within any basic block.
6588 Emit extra code to check for buffer overflows, such as stack
6589 smashing attacks. This is done by adding a guard variable to
6590 functions with vulnerable objects. This includes functions that
6591 call alloca, and functions with buffers larger than 8 bytes. The
6592 guards are initialized when a function is entered and then checked
6593 when the function exits. If a guard check fails, an error message
6594 is printed and the program exits.
6596 `-fstack-protector-all'
6597 Like `-fstack-protector' except that all functions are protected.
6600 Try to reduce the number of symbolic address calculations by using
6601 shared "anchor" symbols to address nearby objects. This
6602 transformation can help to reduce the number of GOT entries and
6603 GOT accesses on some targets.
6605 For example, the implementation of the following function `foo':
6608 int foo (void) { return a + b + c; }
6610 would usually calculate the addresses of all three variables, but
6611 if you compile it with `-fsection-anchors', it will access the
6612 variables from a common anchor point instead. The effect is
6613 similar to the following pseudocode (which isn't valid C):
6617 register int *xr = &x;
6618 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6621 Not all targets support this option.
6623 `--param NAME=VALUE'
6624 In some places, GCC uses various constants to control the amount of
6625 optimization that is done. For example, GCC will not inline
6626 functions that contain more that a certain number of instructions.
6627 You can control some of these constants on the command-line using
6628 the `--param' option.
6630 The names of specific parameters, and the meaning of the values,
6631 are tied to the internals of the compiler, and are subject to
6632 change without notice in future releases.
6634 In each case, the VALUE is an integer. The allowable choices for
6635 NAME are given in the following table:
6637 `sra-max-structure-size'
6638 The maximum structure size, in bytes, at which the scalar
6639 replacement of aggregates (SRA) optimization will perform
6640 block copies. The default value, 0, implies that GCC will
6641 select the most appropriate size itself.
6643 `sra-field-structure-ratio'
6644 The threshold ratio (as a percentage) between instantiated
6645 fields and the complete structure size. We say that if the
6646 ratio of the number of bytes in instantiated fields to the
6647 number of bytes in the complete structure exceeds this
6648 parameter, then block copies are not used. The default is 75.
6650 `struct-reorg-cold-struct-ratio'
6651 The threshold ratio (as a percentage) between a structure
6652 frequency and the frequency of the hottest structure in the
6653 program. This parameter is used by struct-reorg optimization
6654 enabled by `-fipa-struct-reorg'. We say that if the ratio of
6655 a structure frequency, calculated by profiling, to the
6656 hottest structure frequency in the program is less than this
6657 parameter, then structure reorganization is not applied to
6658 this structure. The default is 10.
6660 `predictable-branch-cost-outcome'
6661 When branch is predicted to be taken with probability lower
6662 than this threshold (in percent), then it is considered well
6663 predictable. The default is 10.
6665 `max-crossjump-edges'
6666 The maximum number of incoming edges to consider for
6667 crossjumping. The algorithm used by `-fcrossjumping' is
6668 O(N^2) in the number of edges incoming to each block.
6669 Increasing values mean more aggressive optimization, making
6670 the compile time increase with probably small improvement in
6673 `min-crossjump-insns'
6674 The minimum number of instructions which must be matched at
6675 the end of two blocks before crossjumping will be performed
6676 on them. This value is ignored in the case where all
6677 instructions in the block being crossjumped from are matched.
6678 The default value is 5.
6680 `max-grow-copy-bb-insns'
6681 The maximum code size expansion factor when copying basic
6682 blocks instead of jumping. The expansion is relative to a
6683 jump instruction. The default value is 8.
6685 `max-goto-duplication-insns'
6686 The maximum number of instructions to duplicate to a block
6687 that jumps to a computed goto. To avoid O(N^2) behavior in a
6688 number of passes, GCC factors computed gotos early in the
6689 compilation process, and unfactors them as late as possible.
6690 Only computed jumps at the end of a basic blocks with no more
6691 than max-goto-duplication-insns are unfactored. The default
6694 `max-delay-slot-insn-search'
6695 The maximum number of instructions to consider when looking
6696 for an instruction to fill a delay slot. If more than this
6697 arbitrary number of instructions is searched, the time
6698 savings from filling the delay slot will be minimal so stop
6699 searching. Increasing values mean more aggressive
6700 optimization, making the compile time increase with probably
6701 small improvement in executable run time.
6703 `max-delay-slot-live-search'
6704 When trying to fill delay slots, the maximum number of
6705 instructions to consider when searching for a block with
6706 valid live register information. Increasing this arbitrarily
6707 chosen value means more aggressive optimization, increasing
6708 the compile time. This parameter should be removed when the
6709 delay slot code is rewritten to maintain the control-flow
6713 The approximate maximum amount of memory that will be
6714 allocated in order to perform the global common subexpression
6715 elimination optimization. If more memory than specified is
6716 required, the optimization will not be done.
6719 The maximum number of passes of GCSE to run. The default is
6722 `max-pending-list-length'
6723 The maximum number of pending dependencies scheduling will
6724 allow before flushing the current state and starting over.
6725 Large functions with few branches or calls can create
6726 excessively large lists which needlessly consume memory and
6729 `max-inline-insns-single'
6730 Several parameters control the tree inliner used in gcc.
6731 This number sets the maximum number of instructions (counted
6732 in GCC's internal representation) in a single function that
6733 the tree inliner will consider for inlining. This only
6734 affects functions declared inline and methods implemented in
6735 a class declaration (C++). The default value is 450.
6737 `max-inline-insns-auto'
6738 When you use `-finline-functions' (included in `-O3'), a lot
6739 of functions that would otherwise not be considered for
6740 inlining by the compiler will be investigated. To those
6741 functions, a different (potentially more restrictive) limit
6742 compared to functions declared inline can be applied. The
6743 default value is 450.
6745 `inline-limit-increase-with-profile'
6746 When profile information is available, such as when compiling
6747 with `-fprofile-use', the maximum function size limits
6748 `--param max-inline-insns-single' and `--param
6749 max-inline-insns-auto' are increased by this percentage
6750 amount. Profile information increases the selectivity and
6751 quality of the inlining decisions, so having a larger set of
6752 candidate functions available for inlining can improve
6753 performance. The default value is 100.
6755 `large-function-insns'
6756 The limit specifying really large functions. For functions
6757 larger than this limit after inlining, inlining is
6758 constrained by `--param large-function-growth'. This
6759 parameter is useful primarily to avoid extreme compilation
6760 time caused by non-linear algorithms used by the backend.
6761 The default value is 2700.
6763 `large-function-growth'
6764 Specifies maximal growth of large function caused by inlining
6765 in percents. The default value is 100 which limits large
6766 function growth to 2.0 times the original size.
6769 The limit specifying large translation unit. Growth caused
6770 by inlining of units larger than this limit is limited by
6771 `--param inline-unit-growth'. For small units this might be
6772 too tight (consider unit consisting of function A that is
6773 inline and B that just calls A three time. If B is small
6774 relative to A, the growth of unit is 300\% and yet such
6775 inlining is very sane. For very large units consisting of
6776 small inlineable functions however the overall unit growth
6777 limit is needed to avoid exponential explosion of code size.
6778 Thus for smaller units, the size is increased to `--param
6779 large-unit-insns' before applying `--param
6780 inline-unit-growth'. The default is 10000
6782 `inline-unit-growth'
6783 Specifies maximal overall growth of the compilation unit
6784 caused by inlining. The default value is 30 which limits
6785 unit growth to 1.3 times the original size.
6788 Specifies maximal overall growth of the compilation unit
6789 caused by interprocedural constant propagation. The default
6790 value is 10 which limits unit growth to 1.1 times the
6794 The limit specifying large stack frames. While inlining the
6795 algorithm is trying to not grow past this limit too much.
6796 Default value is 256 bytes.
6798 `large-stack-frame-growth'
6799 Specifies maximal growth of large stack frames caused by
6800 inlining in percents. The default value is 1000 which limits
6801 large stack frame growth to 11 times the original size.
6803 `max-inline-insns-recursive'
6804 `max-inline-insns-recursive-auto'
6805 Specifies maximum number of instructions out-of-line copy of
6806 self recursive inline function can grow into by performing
6809 For functions declared inline `--param
6810 max-inline-insns-recursive' is taken into account. For
6811 function not declared inline, recursive inlining happens only
6812 when `-finline-functions' (included in `-O3') is enabled and
6813 `--param max-inline-insns-recursive-auto' is used. The
6814 default value is 450.
6816 `max-inline-recursive-depth'
6817 `max-inline-recursive-depth-auto'
6818 Specifies maximum recursion depth used by the recursive
6821 For functions declared inline `--param
6822 max-inline-recursive-depth' is taken into account. For
6823 function not declared inline, recursive inlining happens only
6824 when `-finline-functions' (included in `-O3') is enabled and
6825 `--param max-inline-recursive-depth-auto' is used. The
6828 `min-inline-recursive-probability'
6829 Recursive inlining is profitable only for function having
6830 deep recursion in average and can hurt for function having
6831 little recursion depth by increasing the prologue size or
6832 complexity of function body to other optimizers.
6834 When profile feedback is available (see `-fprofile-generate')
6835 the actual recursion depth can be guessed from probability
6836 that function will recurse via given call expression. This
6837 parameter limits inlining only to call expression whose
6838 probability exceeds given threshold (in percents). The
6839 default value is 10.
6842 Specify cost of call instruction relative to simple
6843 arithmetics operations (having cost of 1). Increasing this
6844 cost disqualifies inlining of non-leaf functions and at the
6845 same time increases size of leaf function that is believed to
6846 reduce function size by being inlined. In effect it
6847 increases amount of inlining for code having large
6848 abstraction penalty (many functions that just pass the
6849 arguments to other functions) and decrease inlining for code
6850 with low abstraction penalty. The default value is 12.
6852 `min-vect-loop-bound'
6853 The minimum number of iterations under which a loop will not
6854 get vectorized when `-ftree-vectorize' is used. The number
6855 of iterations after vectorization needs to be greater than
6856 the value specified by this option to allow vectorization.
6857 The default value is 0.
6859 `max-unrolled-insns'
6860 The maximum number of instructions that a loop should have if
6861 that loop is unrolled, and if the loop is unrolled, it
6862 determines how many times the loop code is unrolled.
6864 `max-average-unrolled-insns'
6865 The maximum number of instructions biased by probabilities of
6866 their execution that a loop should have if that loop is
6867 unrolled, and if the loop is unrolled, it determines how many
6868 times the loop code is unrolled.
6871 The maximum number of unrollings of a single loop.
6874 The maximum number of instructions that a loop should have if
6875 that loop is peeled, and if the loop is peeled, it determines
6876 how many times the loop code is peeled.
6879 The maximum number of peelings of a single loop.
6881 `max-completely-peeled-insns'
6883 `max-completely-peeled-insns-feedback'
6884 The maximum number of insns of a completely peeled loop.
6886 The `max-completely-peeled-insns-feedback' is used only when
6887 profile feedback is available and the loop is hot. Because of
6888 the real profiles, this value may set to be larger for hot
6891 `max-once-peeled-insns'
6893 `max-once-peeled-insns-feedback'
6894 The maximum number of insns of a peeled loop that rolls only
6895 once. The `max-once-peeled-insns-feedback' is used only
6896 when profile feedback is available and the loop is hot.
6897 Because of the real profiles, this value may set to be larger
6900 `max-completely-peel-times'
6902 `max-completely-peel-times-feedback'
6903 The maximum number of iterations of a loop to be suitable for
6906 The `max-completely-peel-times-feedback' is used only when
6907 profile feedback is available and the loop is hot. Because of
6908 the real profiles, this value may set to be larger for hot
6911 `max-unswitch-insns'
6912 The maximum number of insns of an unswitched loop.
6914 `max-unswitch-level'
6915 The maximum number of branches unswitched in a single loop.
6918 The minimum cost of an expensive expression in the loop
6921 `iv-consider-all-candidates-bound'
6922 Bound on number of candidates for induction variables below
6923 that all candidates are considered for each use in induction
6924 variable optimizations. Only the most relevant candidates
6925 are considered if there are more candidates, to avoid
6926 quadratic time complexity.
6928 `iv-max-considered-uses'
6929 The induction variable optimizations give up on loops that
6930 contain more induction variable uses.
6932 `iv-always-prune-cand-set-bound'
6933 If number of candidates in the set is smaller than this value,
6934 we always try to remove unnecessary ivs from the set during
6935 its optimization when a new iv is added to the set.
6937 `scev-max-expr-size'
6938 Bound on size of expressions used in the scalar evolutions
6939 analyzer. Large expressions slow the analyzer.
6942 The maximum number of variables in an Omega constraint system.
6943 The default value is 128.
6946 The maximum number of inequalities in an Omega constraint
6947 system. The default value is 256.
6950 The maximum number of equalities in an Omega constraint
6951 system. The default value is 128.
6953 `omega-max-wild-cards'
6954 The maximum number of wildcard variables that the Omega
6955 solver will be able to insert. The default value is 18.
6957 `omega-hash-table-size'
6958 The size of the hash table in the Omega solver. The default
6962 The maximal number of keys used by the Omega solver. The
6963 default value is 500.
6965 `omega-eliminate-redundant-constraints'
6966 When set to 1, use expensive methods to eliminate all
6967 redundant constraints. The default value is 0.
6969 `vect-max-version-for-alignment-checks'
6970 The maximum number of runtime checks that can be performed
6971 when doing loop versioning for alignment in the vectorizer.
6972 See option ftree-vect-loop-version for more information.
6974 `vect-max-version-for-alias-checks'
6975 The maximum number of runtime checks that can be performed
6976 when doing loop versioning for alias in the vectorizer. See
6977 option ftree-vect-loop-version for more information.
6979 `max-iterations-to-track'
6980 The maximum number of iterations of a loop the brute force
6981 algorithm for analysis of # of iterations of the loop tries
6984 `hot-bb-count-fraction'
6985 Select fraction of the maximal count of repetitions of basic
6986 block in program given basic block needs to have to be
6989 `hot-bb-frequency-fraction'
6990 Select fraction of the maximal frequency of executions of
6991 basic block in function given basic block needs to have to be
6994 `max-predicted-iterations'
6995 The maximum number of loop iterations we predict statically.
6996 This is useful in cases where function contain single loop
6997 with known bound and other loop with unknown. We predict the
6998 known number of iterations correctly, while the unknown
6999 number of iterations average to roughly 10. This means that
7000 the loop without bounds would appear artificially cold
7001 relative to the other one.
7004 Select fraction of the maximal frequency of executions of
7005 basic block in function given basic block will get aligned.
7007 `align-loop-iterations'
7008 A loop expected to iterate at lest the selected number of
7009 iterations will get aligned.
7011 `tracer-dynamic-coverage'
7012 `tracer-dynamic-coverage-feedback'
7013 This value is used to limit superblock formation once the
7014 given percentage of executed instructions is covered. This
7015 limits unnecessary code size expansion.
7017 The `tracer-dynamic-coverage-feedback' is used only when
7018 profile feedback is available. The real profiles (as opposed
7019 to statically estimated ones) are much less balanced allowing
7020 the threshold to be larger value.
7022 `tracer-max-code-growth'
7023 Stop tail duplication once code growth has reached given
7024 percentage. This is rather hokey argument, as most of the
7025 duplicates will be eliminated later in cross jumping, so it
7026 may be set to much higher values than is the desired code
7029 `tracer-min-branch-ratio'
7030 Stop reverse growth when the reverse probability of best edge
7031 is less than this threshold (in percent).
7033 `tracer-min-branch-ratio'
7034 `tracer-min-branch-ratio-feedback'
7035 Stop forward growth if the best edge do have probability
7036 lower than this threshold.
7038 Similarly to `tracer-dynamic-coverage' two values are
7039 present, one for compilation for profile feedback and one for
7040 compilation without. The value for compilation with profile
7041 feedback needs to be more conservative (higher) in order to
7042 make tracer effective.
7044 `max-cse-path-length'
7045 Maximum number of basic blocks on path that cse considers.
7049 The maximum instructions CSE process before flushing. The
7053 Maximum number of virtual operands per function allowed to
7054 represent aliases before triggering the alias partitioning
7055 heuristic. Alias partitioning reduces compile times and
7056 memory consumption needed for aliasing at the expense of
7057 precision loss in alias information. The default value for
7058 this parameter is 100 for -O1, 500 for -O2 and 1000 for -O3.
7060 Notice that if a function contains more memory statements
7061 than the value of this parameter, it is not really possible
7062 to achieve this reduction. In this case, the compiler will
7063 use the number of memory statements as the value for
7067 Average number of virtual operands per statement allowed to
7068 represent aliases before triggering the alias partitioning
7069 heuristic. This works in conjunction with
7070 `max-aliased-vops'. If a function contains more than
7071 `max-aliased-vops' virtual operators, then memory symbols
7072 will be grouped into memory partitions until either the total
7073 number of virtual operators is below `max-aliased-vops' or
7074 the average number of virtual operators per memory statement
7075 is below `avg-aliased-vops'. The default value for this
7076 parameter is 1 for -O1 and -O2, and 3 for -O3.
7079 GCC uses a garbage collector to manage its own memory
7080 allocation. This parameter specifies the minimum percentage
7081 by which the garbage collector's heap should be allowed to
7082 expand between collections. Tuning this may improve
7083 compilation speed; it has no effect on code generation.
7085 The default is 30% + 70% * (RAM/1GB) with an upper bound of
7086 100% when RAM >= 1GB. If `getrlimit' is available, the
7087 notion of "RAM" is the smallest of actual RAM and
7088 `RLIMIT_DATA' or `RLIMIT_AS'. If GCC is not able to
7089 calculate RAM on a particular platform, the lower bound of
7090 30% is used. Setting this parameter and `ggc-min-heapsize'
7091 to zero causes a full collection to occur at every
7092 opportunity. This is extremely slow, but can be useful for
7096 Minimum size of the garbage collector's heap before it begins
7097 bothering to collect garbage. The first collection occurs
7098 after the heap expands by `ggc-min-expand'% beyond
7099 `ggc-min-heapsize'. Again, tuning this may improve
7100 compilation speed, and has no effect on code generation.
7102 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit
7103 which tries to ensure that RLIMIT_DATA or RLIMIT_AS are not
7104 exceeded, but with a lower bound of 4096 (four megabytes) and
7105 an upper bound of 131072 (128 megabytes). If GCC is not able
7106 to calculate RAM on a particular platform, the lower bound is
7107 used. Setting this parameter very large effectively disables
7108 garbage collection. Setting this parameter and
7109 `ggc-min-expand' to zero causes a full collection to occur at
7112 `max-reload-search-insns'
7113 The maximum number of instruction reload should look backward
7114 for equivalent register. Increasing values mean more
7115 aggressive optimization, making the compile time increase
7116 with probably slightly better performance. The default value
7119 `max-cselib-memory-locations'
7120 The maximum number of memory locations cselib should take
7121 into account. Increasing values mean more aggressive
7122 optimization, making the compile time increase with probably
7123 slightly better performance. The default value is 500.
7125 `reorder-blocks-duplicate'
7126 `reorder-blocks-duplicate-feedback'
7127 Used by basic block reordering pass to decide whether to use
7128 unconditional branch or duplicate the code on its
7129 destination. Code is duplicated when its estimated size is
7130 smaller than this value multiplied by the estimated size of
7131 unconditional jump in the hot spots of the program.
7133 The `reorder-block-duplicate-feedback' is used only when
7134 profile feedback is available and may be set to higher values
7135 than `reorder-block-duplicate' since information about the
7136 hot spots is more accurate.
7138 `max-sched-ready-insns'
7139 The maximum number of instructions ready to be issued the
7140 scheduler should consider at any given time during the first
7141 scheduling pass. Increasing values mean more thorough
7142 searches, making the compilation time increase with probably
7143 little benefit. The default value is 100.
7145 `max-sched-region-blocks'
7146 The maximum number of blocks in a region to be considered for
7147 interblock scheduling. The default value is 10.
7149 `max-pipeline-region-blocks'
7150 The maximum number of blocks in a region to be considered for
7151 pipelining in the selective scheduler. The default value is
7154 `max-sched-region-insns'
7155 The maximum number of insns in a region to be considered for
7156 interblock scheduling. The default value is 100.
7158 `max-pipeline-region-insns'
7159 The maximum number of insns in a region to be considered for
7160 pipelining in the selective scheduler. The default value is
7164 The minimum probability (in percents) of reaching a source
7165 block for interblock speculative scheduling. The default
7168 `max-sched-extend-regions-iters'
7169 The maximum number of iterations through CFG to extend
7170 regions. 0 - disable region extension, N - do at most N
7171 iterations. The default value is 0.
7173 `max-sched-insn-conflict-delay'
7174 The maximum conflict delay for an insn to be considered for
7175 speculative motion. The default value is 3.
7177 `sched-spec-prob-cutoff'
7178 The minimal probability of speculation success (in percents),
7179 so that speculative insn will be scheduled. The default
7182 `sched-mem-true-dep-cost'
7183 Minimal distance (in CPU cycles) between store and load
7184 targeting same memory locations. The default value is 1.
7186 `selsched-max-lookahead'
7187 The maximum size of the lookahead window of selective
7188 scheduling. It is a depth of search for available
7189 instructions. The default value is 50.
7191 `selsched-max-sched-times'
7192 The maximum number of times that an instruction will be
7193 scheduled during selective scheduling. This is the limit on
7194 the number of iterations through which the instruction may be
7195 pipelined. The default value is 2.
7197 `selsched-max-insns-to-rename'
7198 The maximum number of best instructions in the ready list
7199 that are considered for renaming in the selective scheduler.
7200 The default value is 2.
7202 `max-last-value-rtl'
7203 The maximum size measured as number of RTLs that can be
7204 recorded in an expression in combiner for a pseudo register
7205 as last known value of that register. The default is 10000.
7207 `integer-share-limit'
7208 Small integer constants can use a shared data structure,
7209 reducing the compiler's memory usage and increasing its
7210 speed. This sets the maximum value of a shared integer
7211 constant. The default value is 256.
7213 `min-virtual-mappings'
7214 Specifies the minimum number of virtual mappings in the
7215 incremental SSA updater that should be registered to trigger
7216 the virtual mappings heuristic defined by
7217 virtual-mappings-ratio. The default value is 100.
7219 `virtual-mappings-ratio'
7220 If the number of virtual mappings is virtual-mappings-ratio
7221 bigger than the number of virtual symbols to be updated, then
7222 the incremental SSA updater switches to a full update for
7223 those symbols. The default ratio is 3.
7226 The minimum size of buffers (i.e. arrays) that will receive
7227 stack smashing protection when `-fstack-protection' is used.
7229 `max-jump-thread-duplication-stmts'
7230 Maximum number of statements allowed in a block that needs to
7231 be duplicated when threading jumps.
7233 `max-fields-for-field-sensitive'
7234 Maximum number of fields in a structure we will treat in a
7235 field sensitive manner during pointer analysis. The default
7236 is zero for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7239 Estimate on average number of instructions that are executed
7240 before prefetch finishes. The distance we prefetch ahead is
7241 proportional to this constant. Increasing this number may
7242 also lead to less streams being prefetched (see
7243 `simultaneous-prefetches').
7245 `simultaneous-prefetches'
7246 Maximum number of prefetches that can run at the same time.
7248 `l1-cache-line-size'
7249 The size of cache line in L1 cache, in bytes.
7252 The size of L1 cache, in kilobytes.
7255 The size of L2 cache, in kilobytes.
7257 `use-canonical-types'
7258 Whether the compiler should use the "canonical" type system.
7259 By default, this should always be 1, which uses a more
7260 efficient internal mechanism for comparing types in C++ and
7261 Objective-C++. However, if bugs in the canonical type system
7262 are causing compilation failures, set this value to 0 to
7263 disable canonical types.
7265 `switch-conversion-max-branch-ratio'
7266 Switch initialization conversion will refuse to create arrays
7267 that are bigger than `switch-conversion-max-branch-ratio'
7268 times the number of branches in the switch.
7270 `max-partial-antic-length'
7271 Maximum length of the partial antic set computed during the
7272 tree partial redundancy elimination optimization
7273 (`-ftree-pre') when optimizing at `-O3' and above. For some
7274 sorts of source code the enhanced partial redundancy
7275 elimination optimization can run away, consuming all of the
7276 memory available on the host machine. This parameter sets a
7277 limit on the length of the sets that are computed, which
7278 prevents the runaway behavior. Setting a value of 0 for this
7279 parameter will allow an unlimited set length.
7281 `sccvn-max-scc-size'
7282 Maximum size of a strongly connected component (SCC) during
7283 SCCVN processing. If this limit is hit, SCCVN processing for
7284 the whole function will not be done and optimizations
7285 depending on it will be disabled. The default maximum SCC
7289 IRA uses a regional register allocation by default. If a
7290 function contains loops more than number given by the
7291 parameter, only at most given number of the most frequently
7292 executed loops will form regions for the regional register
7293 allocation. The default value of the parameter is 100.
7295 `ira-max-conflict-table-size'
7296 Although IRA uses a sophisticated algorithm of compression
7297 conflict table, the table can be still big for huge
7298 functions. If the conflict table for a function could be
7299 more than size in MB given by the parameter, the conflict
7300 table is not built and faster, simpler, and lower quality
7301 register allocation algorithm will be used. The algorithm do
7302 not use pseudo-register conflicts. The default value of the
7305 `loop-invariant-max-bbs-in-loop'
7306 Loop invariant motion can be very expensive, both in compile
7307 time and in amount of needed compile time memory, with very
7308 large loops. Loops with more basic blocks than this
7309 parameter won't have loop invariant motion optimization
7310 performed on them. The default value of the parameter is
7311 1000 for -O1 and 10000 for -O2 and above.
7314 This is a switch to turn on live range shrinking optimization.
7317 This is used as a control knob to enable different
7318 transformations in the live range shrinking phase. Values of
7319 1, 2, and 4 are used to enable upward motion, downward
7320 motion, and tree reshaping transformations respectively. The
7321 values can be bitwise ORed.
7323 `reg-pressure-min-bb-factor'
7324 A performance tuning knob to control register pressure. When
7325 the size (in the number of gimple statements) of a basic
7326 block in a loop is larger than the threshold specified by
7327 this parameter multiplied by the number of available
7328 registers, live range shrinking optimization is enabled.
7330 `reg-pressure-min-tree'
7331 The minimal size (number of leaves) of a tree to be reshaped
7332 in the Live Range Shrinking optimization.
7336 File: gcc.info, Node: Preprocessor Options, Next: Assembler Options, Prev: Optimize Options, Up: Invoking GCC
7338 3.11 Options Controlling the Preprocessor
7339 =========================================
7341 These options control the C preprocessor, which is run on each C source
7342 file before actual compilation.
7344 If you use the `-E' option, nothing is done except preprocessing.
7345 Some of these options make sense only together with `-E' because they
7346 cause the preprocessor output to be unsuitable for actual compilation.
7348 You can use `-Wp,OPTION' to bypass the compiler driver and pass
7349 OPTION directly through to the preprocessor. If OPTION contains
7350 commas, it is split into multiple options at the commas. However,
7351 many options are modified, translated or interpreted by the
7352 compiler driver before being passed to the preprocessor, and `-Wp'
7353 forcibly bypasses this phase. The preprocessor's direct interface
7354 is undocumented and subject to change, so whenever possible you
7355 should avoid using `-Wp' and let the driver handle the options
7358 `-Xpreprocessor OPTION'
7359 Pass OPTION as an option to the preprocessor. You can use this to
7360 supply system-specific preprocessor options which GCC does not
7361 know how to recognize.
7363 If you want to pass an option that takes an argument, you must use
7364 `-Xpreprocessor' twice, once for the option and once for the
7368 Predefine NAME as a macro, with definition `1'.
7370 `-D NAME=DEFINITION'
7371 The contents of DEFINITION are tokenized and processed as if they
7372 appeared during translation phase three in a `#define' directive.
7373 In particular, the definition will be truncated by embedded
7376 If you are invoking the preprocessor from a shell or shell-like
7377 program you may need to use the shell's quoting syntax to protect
7378 characters such as spaces that have a meaning in the shell syntax.
7380 If you wish to define a function-like macro on the command line,
7381 write its argument list with surrounding parentheses before the
7382 equals sign (if any). Parentheses are meaningful to most shells,
7383 so you will need to quote the option. With `sh' and `csh',
7384 `-D'NAME(ARGS...)=DEFINITION'' works.
7386 `-D' and `-U' options are processed in the order they are given on
7387 the command line. All `-imacros FILE' and `-include FILE' options
7388 are processed after all `-D' and `-U' options.
7391 Cancel any previous definition of NAME, either built in or
7392 provided with a `-D' option.
7395 Do not predefine any system-specific or GCC-specific macros. The
7396 standard predefined macros remain defined.
7399 Add the directory DIR to the list of directories to be searched
7400 for header files. Directories named by `-I' are searched before
7401 the standard system include directories. If the directory DIR is
7402 a standard system include directory, the option is ignored to
7403 ensure that the default search order for system directories and
7404 the special treatment of system headers are not defeated . If DIR
7405 begins with `=', then the `=' will be replaced by the sysroot
7406 prefix; see `--sysroot' and `-isysroot'.
7409 Write output to FILE. This is the same as specifying FILE as the
7410 second non-option argument to `cpp'. `gcc' has a different
7411 interpretation of a second non-option argument, so you must use
7412 `-o' to specify the output file.
7415 Turns on all optional warnings which are desirable for normal code.
7416 At present this is `-Wcomment', `-Wtrigraphs', `-Wmultichar' and a
7417 warning about integer promotion causing a change of sign in `#if'
7418 expressions. Note that many of the preprocessor's warnings are on
7419 by default and have no options to control them.
7423 Warn whenever a comment-start sequence `/*' appears in a `/*'
7424 comment, or whenever a backslash-newline appears in a `//' comment.
7425 (Both forms have the same effect.)
7428 Most trigraphs in comments cannot affect the meaning of the
7429 program. However, a trigraph that would form an escaped newline
7430 (`??/' at the end of a line) can, by changing where the comment
7431 begins or ends. Therefore, only trigraphs that would form escaped
7432 newlines produce warnings inside a comment.
7434 This option is implied by `-Wall'. If `-Wall' is not given, this
7435 option is still enabled unless trigraphs are enabled. To get
7436 trigraph conversion without warnings, but get the other `-Wall'
7437 warnings, use `-trigraphs -Wall -Wno-trigraphs'.
7440 Warn about certain constructs that behave differently in
7441 traditional and ISO C. Also warn about ISO C constructs that have
7442 no traditional C equivalent, and problematic constructs which
7446 Warn whenever an identifier which is not a macro is encountered in
7447 an `#if' directive, outside of `defined'. Such identifiers are
7451 Warn about macros defined in the main file that are unused. A
7452 macro is "used" if it is expanded or tested for existence at least
7453 once. The preprocessor will also warn if the macro has not been
7454 used at the time it is redefined or undefined.
7456 Built-in macros, macros defined on the command line, and macros
7457 defined in include files are not warned about.
7459 _Note:_ If a macro is actually used, but only used in skipped
7460 conditional blocks, then CPP will report it as unused. To avoid
7461 the warning in such a case, you might improve the scope of the
7462 macro's definition by, for example, moving it into the first
7463 skipped block. Alternatively, you could provide a dummy use with
7466 #if defined the_macro_causing_the_warning
7470 Warn whenever an `#else' or an `#endif' are followed by text.
7471 This usually happens in code of the form
7479 The second and third `FOO' should be in comments, but often are not
7480 in older programs. This warning is on by default.
7483 Make all warnings into hard errors. Source code which triggers
7484 warnings will be rejected.
7487 Issue warnings for code in system headers. These are normally
7488 unhelpful in finding bugs in your own code, therefore suppressed.
7489 If you are responsible for the system library, you may want to see
7493 Suppress all warnings, including those which GNU CPP issues by
7497 Issue all the mandatory diagnostics listed in the C standard.
7498 Some of them are left out by default, since they trigger
7499 frequently on harmless code.
7502 Issue all the mandatory diagnostics, and make all mandatory
7503 diagnostics into errors. This includes mandatory diagnostics that
7504 GCC issues without `-pedantic' but treats as warnings.
7507 Instead of outputting the result of preprocessing, output a rule
7508 suitable for `make' describing the dependencies of the main source
7509 file. The preprocessor outputs one `make' rule containing the
7510 object file name for that source file, a colon, and the names of
7511 all the included files, including those coming from `-include' or
7512 `-imacros' command line options.
7514 Unless specified explicitly (with `-MT' or `-MQ'), the object file
7515 name consists of the name of the source file with any suffix
7516 replaced with object file suffix and with any leading directory
7517 parts removed. If there are many included files then the rule is
7518 split into several lines using `\'-newline. The rule has no
7521 This option does not suppress the preprocessor's debug output,
7522 such as `-dM'. To avoid mixing such debug output with the
7523 dependency rules you should explicitly specify the dependency
7524 output file with `-MF', or use an environment variable like
7525 `DEPENDENCIES_OUTPUT' (*note Environment Variables::). Debug
7526 output will still be sent to the regular output stream as normal.
7528 Passing `-M' to the driver implies `-E', and suppresses warnings
7529 with an implicit `-w'.
7532 Like `-M' but do not mention header files that are found in system
7533 header directories, nor header files that are included, directly
7534 or indirectly, from such a header.
7536 This implies that the choice of angle brackets or double quotes in
7537 an `#include' directive does not in itself determine whether that
7538 header will appear in `-MM' dependency output. This is a slight
7539 change in semantics from GCC versions 3.0 and earlier.
7542 When used with `-M' or `-MM', specifies a file to write the
7543 dependencies to. If no `-MF' switch is given the preprocessor
7544 sends the rules to the same place it would have sent preprocessed
7547 When used with the driver options `-MD' or `-MMD', `-MF' overrides
7548 the default dependency output file.
7551 In conjunction with an option such as `-M' requesting dependency
7552 generation, `-MG' assumes missing header files are generated files
7553 and adds them to the dependency list without raising an error.
7554 The dependency filename is taken directly from the `#include'
7555 directive without prepending any path. `-MG' also suppresses
7556 preprocessed output, as a missing header file renders this useless.
7558 This feature is used in automatic updating of makefiles.
7561 This option instructs CPP to add a phony target for each dependency
7562 other than the main file, causing each to depend on nothing. These
7563 dummy rules work around errors `make' gives if you remove header
7564 files without updating the `Makefile' to match.
7566 This is typical output:
7568 test.o: test.c test.h
7573 Change the target of the rule emitted by dependency generation. By
7574 default CPP takes the name of the main input file, deletes any
7575 directory components and any file suffix such as `.c', and appends
7576 the platform's usual object suffix. The result is the target.
7578 An `-MT' option will set the target to be exactly the string you
7579 specify. If you want multiple targets, you can specify them as a
7580 single argument to `-MT', or use multiple `-MT' options.
7582 For example, `-MT '$(objpfx)foo.o'' might give
7584 $(objpfx)foo.o: foo.c
7587 Same as `-MT', but it quotes any characters which are special to
7588 Make. `-MQ '$(objpfx)foo.o'' gives
7590 $$(objpfx)foo.o: foo.c
7592 The default target is automatically quoted, as if it were given
7596 `-MD' is equivalent to `-M -MF FILE', except that `-E' is not
7597 implied. The driver determines FILE based on whether an `-o'
7598 option is given. If it is, the driver uses its argument but with
7599 a suffix of `.d', otherwise it takes the name of the input file,
7600 removes any directory components and suffix, and applies a `.d'
7603 If `-MD' is used in conjunction with `-E', any `-o' switch is
7604 understood to specify the dependency output file (*note -MF:
7605 dashMF.), but if used without `-E', each `-o' is understood to
7606 specify a target object file.
7608 Since `-E' is not implied, `-MD' can be used to generate a
7609 dependency output file as a side-effect of the compilation process.
7612 Like `-MD' except mention only user header files, not system
7616 When using precompiled headers (*note Precompiled Headers::), this
7617 flag will cause the dependency-output flags to also list the files
7618 from the precompiled header's dependencies. If not specified only
7619 the precompiled header would be listed and not the files that were
7620 used to create it because those files are not consulted when a
7621 precompiled header is used.
7624 This option allows use of a precompiled header (*note Precompiled
7625 Headers::) together with `-E'. It inserts a special `#pragma',
7626 `#pragma GCC pch_preprocess "<filename>"' in the output to mark
7627 the place where the precompiled header was found, and its
7628 filename. When `-fpreprocessed' is in use, GCC recognizes this
7629 `#pragma' and loads the PCH.
7631 This option is off by default, because the resulting preprocessed
7632 output is only really suitable as input to GCC. It is switched on
7635 You should not write this `#pragma' in your own code, but it is
7636 safe to edit the filename if the PCH file is available in a
7637 different location. The filename may be absolute or it may be
7638 relative to GCC's current directory.
7643 `-x assembler-with-cpp'
7644 Specify the source language: C, C++, Objective-C, or assembly.
7645 This has nothing to do with standards conformance or extensions;
7646 it merely selects which base syntax to expect. If you give none
7647 of these options, cpp will deduce the language from the extension
7648 of the source file: `.c', `.cc', `.m', or `.S'. Some other common
7649 extensions for C++ and assembly are also recognized. If cpp does
7650 not recognize the extension, it will treat the file as C; this is
7651 the most generic mode.
7653 _Note:_ Previous versions of cpp accepted a `-lang' option which
7654 selected both the language and the standards conformance level.
7655 This option has been removed, because it conflicts with the `-l'
7660 Specify the standard to which the code should conform. Currently
7661 CPP knows about C and C++ standards; others may be added in the
7664 STANDARD may be one of:
7667 The ISO C standard from 1990. `c89' is the customary
7668 shorthand for this version of the standard.
7670 The `-ansi' option is equivalent to `-std=c89'.
7673 The 1990 C standard, as amended in 1994.
7679 The revised ISO C standard, published in December 1999.
7680 Before publication, this was known as C9X.
7683 The 1990 C standard plus GNU extensions. This is the default.
7687 The 1999 C standard plus GNU extensions.
7690 The 1998 ISO C++ standard plus amendments.
7693 The same as `-std=c++98' plus GNU extensions. This is the
7694 default for C++ code.
7697 Split the include path. Any directories specified with `-I'
7698 options before `-I-' are searched only for headers requested with
7699 `#include "FILE"'; they are not searched for `#include <FILE>'.
7700 If additional directories are specified with `-I' options after
7701 the `-I-', those directories are searched for all `#include'
7704 In addition, `-I-' inhibits the use of the directory of the current
7705 file directory as the first search directory for `#include "FILE"'.
7706 This option has been deprecated.
7709 Do not search the standard system directories for header files.
7710 Only the directories you have specified with `-I' options (and the
7711 directory of the current file, if appropriate) are searched.
7714 Do not search for header files in the C++-specific standard
7715 directories, but do still search the other standard directories.
7716 (This option is used when building the C++ library.)
7719 Process FILE as if `#include "file"' appeared as the first line of
7720 the primary source file. However, the first directory searched
7721 for FILE is the preprocessor's working directory _instead of_ the
7722 directory containing the main source file. If not found there, it
7723 is searched for in the remainder of the `#include "..."' search
7726 If multiple `-include' options are given, the files are included
7727 in the order they appear on the command line.
7730 Exactly like `-include', except that any output produced by
7731 scanning FILE is thrown away. Macros it defines remain defined.
7732 This allows you to acquire all the macros from a header without
7733 also processing its declarations.
7735 All files specified by `-imacros' are processed before all files
7736 specified by `-include'.
7739 Search DIR for header files, but do it _after_ all directories
7740 specified with `-I' and the standard system directories have been
7741 exhausted. DIR is treated as a system include directory. If DIR
7742 begins with `=', then the `=' will be replaced by the sysroot
7743 prefix; see `--sysroot' and `-isysroot'.
7746 Specify PREFIX as the prefix for subsequent `-iwithprefix'
7747 options. If the prefix represents a directory, you should include
7751 `-iwithprefixbefore DIR'
7752 Append DIR to the prefix specified previously with `-iprefix', and
7753 add the resulting directory to the include search path.
7754 `-iwithprefixbefore' puts it in the same place `-I' would;
7755 `-iwithprefix' puts it where `-idirafter' would.
7758 This option is like the `--sysroot' option, but applies only to
7759 header files. See the `--sysroot' option for more information.
7762 Use DIR as a subdirectory of the directory containing
7763 target-specific C++ headers.
7766 Search DIR for header files, after all directories specified by
7767 `-I' but before the standard system directories. Mark it as a
7768 system directory, so that it gets the same special treatment as is
7769 applied to the standard system directories. If DIR begins with
7770 `=', then the `=' will be replaced by the sysroot prefix; see
7771 `--sysroot' and `-isysroot'.
7774 Search DIR only for header files requested with `#include "FILE"';
7775 they are not searched for `#include <FILE>', before all
7776 directories specified by `-I' and before the standard system
7777 directories. If DIR begins with `=', then the `=' will be replaced
7778 by the sysroot prefix; see `--sysroot' and `-isysroot'.
7781 When preprocessing, handle directives, but do not expand macros.
7783 The option's behavior depends on the `-E' and `-fpreprocessed'
7786 With `-E', preprocessing is limited to the handling of directives
7787 such as `#define', `#ifdef', and `#error'. Other preprocessor
7788 operations, such as macro expansion and trigraph conversion are
7789 not performed. In addition, the `-dD' option is implicitly
7792 With `-fpreprocessed', predefinition of command line and most
7793 builtin macros is disabled. Macros such as `__LINE__', which are
7794 contextually dependent, are handled normally. This enables
7795 compilation of files previously preprocessed with `-E
7798 With both `-E' and `-fpreprocessed', the rules for
7799 `-fpreprocessed' take precedence. This enables full preprocessing
7800 of files previously preprocessed with `-E -fdirectives-only'.
7802 `-fdollars-in-identifiers'
7803 Accept `$' in identifiers.
7805 `-fextended-identifiers'
7806 Accept universal character names in identifiers. This option is
7807 experimental; in a future version of GCC, it will be enabled by
7808 default for C99 and C++.
7811 Indicate to the preprocessor that the input file has already been
7812 preprocessed. This suppresses things like macro expansion,
7813 trigraph conversion, escaped newline splicing, and processing of
7814 most directives. The preprocessor still recognizes and removes
7815 comments, so that you can pass a file preprocessed with `-C' to
7816 the compiler without problems. In this mode the integrated
7817 preprocessor is little more than a tokenizer for the front ends.
7819 `-fpreprocessed' is implicit if the input file has one of the
7820 extensions `.i', `.ii' or `.mi'. These are the extensions that
7821 GCC uses for preprocessed files created by `-save-temps'.
7824 Set the distance between tab stops. This helps the preprocessor
7825 report correct column numbers in warnings or errors, even if tabs
7826 appear on the line. If the value is less than 1 or greater than
7827 100, the option is ignored. The default is 8.
7829 `-fexec-charset=CHARSET'
7830 Set the execution character set, used for string and character
7831 constants. The default is UTF-8. CHARSET can be any encoding
7832 supported by the system's `iconv' library routine.
7834 `-fwide-exec-charset=CHARSET'
7835 Set the wide execution character set, used for wide string and
7836 character constants. The default is UTF-32 or UTF-16, whichever
7837 corresponds to the width of `wchar_t'. As with `-fexec-charset',
7838 CHARSET can be any encoding supported by the system's `iconv'
7839 library routine; however, you will have problems with encodings
7840 that do not fit exactly in `wchar_t'.
7842 `-finput-charset=CHARSET'
7843 Set the input character set, used for translation from the
7844 character set of the input file to the source character set used
7845 by GCC. If the locale does not specify, or GCC cannot get this
7846 information from the locale, the default is UTF-8. This can be
7847 overridden by either the locale or this command line option.
7848 Currently the command line option takes precedence if there's a
7849 conflict. CHARSET can be any encoding supported by the system's
7850 `iconv' library routine.
7852 `-fworking-directory'
7853 Enable generation of linemarkers in the preprocessor output that
7854 will let the compiler know the current working directory at the
7855 time of preprocessing. When this option is enabled, the
7856 preprocessor will emit, after the initial linemarker, a second
7857 linemarker with the current working directory followed by two
7858 slashes. GCC will use this directory, when it's present in the
7859 preprocessed input, as the directory emitted as the current
7860 working directory in some debugging information formats. This
7861 option is implicitly enabled if debugging information is enabled,
7862 but this can be inhibited with the negated form
7863 `-fno-working-directory'. If the `-P' flag is present in the
7864 command line, this option has no effect, since no `#line'
7865 directives are emitted whatsoever.
7868 Do not print column numbers in diagnostics. This may be necessary
7869 if diagnostics are being scanned by a program that does not
7870 understand the column numbers, such as `dejagnu'.
7872 `-A PREDICATE=ANSWER'
7873 Make an assertion with the predicate PREDICATE and answer ANSWER.
7874 This form is preferred to the older form `-A PREDICATE(ANSWER)',
7875 which is still supported, because it does not use shell special
7878 `-A -PREDICATE=ANSWER'
7879 Cancel an assertion with the predicate PREDICATE and answer ANSWER.
7882 CHARS is a sequence of one or more of the following characters,
7883 and must not be preceded by a space. Other characters are
7884 interpreted by the compiler proper, or reserved for future
7885 versions of GCC, and so are silently ignored. If you specify
7886 characters whose behavior conflicts, the result is undefined.
7889 Instead of the normal output, generate a list of `#define'
7890 directives for all the macros defined during the execution of
7891 the preprocessor, including predefined macros. This gives
7892 you a way of finding out what is predefined in your version
7893 of the preprocessor. Assuming you have no file `foo.h', the
7896 touch foo.h; cpp -dM foo.h
7898 will show all the predefined macros.
7900 If you use `-dM' without the `-E' option, `-dM' is
7901 interpreted as a synonym for `-fdump-rtl-mach'. *Note
7902 Debugging Options: (gcc)Debugging Options.
7905 Like `M' except in two respects: it does _not_ include the
7906 predefined macros, and it outputs _both_ the `#define'
7907 directives and the result of preprocessing. Both kinds of
7908 output go to the standard output file.
7911 Like `D', but emit only the macro names, not their expansions.
7914 Output `#include' directives in addition to the result of
7918 Like `D' except that only macros that are expanded, or whose
7919 definedness is tested in preprocessor directives, are output;
7920 the output is delayed until the use or test of the macro; and
7921 `#undef' directives are also output for macros tested but
7922 undefined at the time.
7925 Inhibit generation of linemarkers in the output from the
7926 preprocessor. This might be useful when running the preprocessor
7927 on something that is not C code, and will be sent to a program
7928 which might be confused by the linemarkers.
7931 Do not discard comments. All comments are passed through to the
7932 output file, except for comments in processed directives, which
7933 are deleted along with the directive.
7935 You should be prepared for side effects when using `-C'; it causes
7936 the preprocessor to treat comments as tokens in their own right.
7937 For example, comments appearing at the start of what would be a
7938 directive line have the effect of turning that line into an
7939 ordinary source line, since the first token on the line is no
7943 Do not discard comments, including during macro expansion. This is
7944 like `-C', except that comments contained within macros are also
7945 passed through to the output file where the macro is expanded.
7947 In addition to the side-effects of the `-C' option, the `-CC'
7948 option causes all C++-style comments inside a macro to be
7949 converted to C-style comments. This is to prevent later use of
7950 that macro from inadvertently commenting out the remainder of the
7953 The `-CC' option is generally used to support lint comments.
7956 Try to imitate the behavior of old-fashioned C preprocessors, as
7957 opposed to ISO C preprocessors.
7960 Process trigraph sequences. These are three-character sequences,
7961 all starting with `??', that are defined by ISO C to stand for
7962 single characters. For example, `??/' stands for `\', so `'??/n''
7963 is a character constant for a newline. By default, GCC ignores
7964 trigraphs, but in standard-conforming modes it converts them. See
7965 the `-std' and `-ansi' options.
7967 The nine trigraphs and their replacements are
7969 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??-
7970 Replacement: [ ] { } # \ ^ | ~
7973 Enable special code to work around file systems which only permit
7974 very short file names, such as MS-DOS.
7978 Print text describing all the command line options instead of
7979 preprocessing anything.
7982 Verbose mode. Print out GNU CPP's version number at the beginning
7983 of execution, and report the final form of the include path.
7986 Print the name of each header file used, in addition to other
7987 normal activities. Each name is indented to show how deep in the
7988 `#include' stack it is. Precompiled header files are also
7989 printed, even if they are found to be invalid; an invalid
7990 precompiled header file is printed with `...x' and a valid one
7995 Print out GNU CPP's version number. With one dash, proceed to
7996 preprocess as normal. With two dashes, exit immediately.
7999 File: gcc.info, Node: Assembler Options, Next: Link Options, Prev: Preprocessor Options, Up: Invoking GCC
8001 3.12 Passing Options to the Assembler
8002 =====================================
8004 You can pass options to the assembler.
8007 Pass OPTION as an option to the assembler. If OPTION contains
8008 commas, it is split into multiple options at the commas.
8010 `-Xassembler OPTION'
8011 Pass OPTION as an option to the assembler. You can use this to
8012 supply system-specific assembler options which GCC does not know
8015 If you want to pass an option that takes an argument, you must use
8016 `-Xassembler' twice, once for the option and once for the argument.
8020 File: gcc.info, Node: Link Options, Next: Directory Options, Prev: Assembler Options, Up: Invoking GCC
8022 3.13 Options for Linking
8023 ========================
8025 These options come into play when the compiler links object files into
8026 an executable output file. They are meaningless if the compiler is not
8030 A file name that does not end in a special recognized suffix is
8031 considered to name an object file or library. (Object files are
8032 distinguished from libraries by the linker according to the file
8033 contents.) If linking is done, these object files are used as
8034 input to the linker.
8039 If any of these options is used, then the linker is not run, and
8040 object file names should not be used as arguments. *Note Overall
8045 Search the library named LIBRARY when linking. (The second
8046 alternative with the library as a separate argument is only for
8047 POSIX compliance and is not recommended.)
8049 It makes a difference where in the command you write this option;
8050 the linker searches and processes libraries and object files in
8051 the order they are specified. Thus, `foo.o -lz bar.o' searches
8052 library `z' after file `foo.o' but before `bar.o'. If `bar.o'
8053 refers to functions in `z', those functions may not be loaded.
8055 The linker searches a standard list of directories for the library,
8056 which is actually a file named `libLIBRARY.a'. The linker then
8057 uses this file as if it had been specified precisely by name.
8059 The directories searched include several standard system
8060 directories plus any that you specify with `-L'.
8062 Normally the files found this way are library files--archive files
8063 whose members are object files. The linker handles an archive
8064 file by scanning through it for members which define symbols that
8065 have so far been referenced but not defined. But if the file that
8066 is found is an ordinary object file, it is linked in the usual
8067 fashion. The only difference between using an `-l' option and
8068 specifying a file name is that `-l' surrounds LIBRARY with `lib'
8069 and `.a' and searches several directories.
8072 You need this special case of the `-l' option in order to link an
8073 Objective-C or Objective-C++ program.
8076 Do not use the standard system startup files when linking. The
8077 standard system libraries are used normally, unless `-nostdlib' or
8078 `-nodefaultlibs' is used.
8081 Do not use the standard system libraries when linking. Only the
8082 libraries you specify will be passed to the linker. The standard
8083 startup files are used normally, unless `-nostartfiles' is used.
8084 The compiler may generate calls to `memcmp', `memset', `memcpy'
8085 and `memmove'. These entries are usually resolved by entries in
8086 libc. These entry points should be supplied through some other
8087 mechanism when this option is specified.
8090 Do not use the standard system startup files or libraries when
8091 linking. No startup files and only the libraries you specify will
8092 be passed to the linker. The compiler may generate calls to
8093 `memcmp', `memset', `memcpy' and `memmove'. These entries are
8094 usually resolved by entries in libc. These entry points should be
8095 supplied through some other mechanism when this option is
8098 One of the standard libraries bypassed by `-nostdlib' and
8099 `-nodefaultlibs' is `libgcc.a', a library of internal subroutines
8100 that GCC uses to overcome shortcomings of particular machines, or
8101 special needs for some languages. (*Note Interfacing to GCC
8102 Output: (gccint)Interface, for more discussion of `libgcc.a'.) In
8103 most cases, you need `libgcc.a' even when you want to avoid other
8104 standard libraries. In other words, when you specify `-nostdlib'
8105 or `-nodefaultlibs' you should usually specify `-lgcc' as well.
8106 This ensures that you have no unresolved references to internal GCC
8107 library subroutines. (For example, `__main', used to ensure C++
8108 constructors will be called; *note `collect2': (gccint)Collect2.)
8111 Produce a position independent executable on targets which support
8112 it. For predictable results, you must also specify the same set
8113 of options that were used to generate code (`-fpie', `-fPIE', or
8114 model suboptions) when you specify this option.
8117 Pass the flag `-export-dynamic' to the ELF linker, on targets that
8118 support it. This instructs the linker to add all symbols, not only
8119 used ones, to the dynamic symbol table. This option is needed for
8120 some uses of `dlopen' or to allow obtaining backtraces from within
8124 Remove all symbol table and relocation information from the
8128 On systems that support dynamic linking, this prevents linking
8129 with the shared libraries. On other systems, this option has no
8133 Produce a shared object which can then be linked with other
8134 objects to form an executable. Not all systems support this
8135 option. For predictable results, you must also specify the same
8136 set of options that were used to generate code (`-fpic', `-fPIC',
8137 or model suboptions) when you specify this option.(1)
8141 On systems that provide `libgcc' as a shared library, these options
8142 force the use of either the shared or static version respectively.
8143 If no shared version of `libgcc' was built when the compiler was
8144 configured, these options have no effect.
8146 There are several situations in which an application should use the
8147 shared `libgcc' instead of the static version. The most common of
8148 these is when the application wishes to throw and catch exceptions
8149 across different shared libraries. In that case, each of the
8150 libraries as well as the application itself should use the shared
8153 Therefore, the G++ and GCJ drivers automatically add
8154 `-shared-libgcc' whenever you build a shared library or a main
8155 executable, because C++ and Java programs typically use
8156 exceptions, so this is the right thing to do.
8158 If, instead, you use the GCC driver to create shared libraries,
8159 you may find that they will not always be linked with the shared
8160 `libgcc'. If GCC finds, at its configuration time, that you have
8161 a non-GNU linker or a GNU linker that does not support option
8162 `--eh-frame-hdr', it will link the shared version of `libgcc' into
8163 shared libraries by default. Otherwise, it will take advantage of
8164 the linker and optimize away the linking with the shared version
8165 of `libgcc', linking with the static version of libgcc by default.
8166 This allows exceptions to propagate through such shared
8167 libraries, without incurring relocation costs at library load time.
8169 However, if a library or main executable is supposed to throw or
8170 catch exceptions, you must link it using the G++ or GCJ driver, as
8171 appropriate for the languages used in the program, or using the
8172 option `-shared-libgcc', such that it is linked with the shared
8176 Bind references to global symbols when building a shared object.
8177 Warn about any unresolved references (unless overridden by the
8178 link editor option `-Xlinker -z -Xlinker defs'). Only a few
8179 systems support this option.
8182 Use SCRIPT as the linker script. This option is supported by most
8183 systems using the GNU linker. On some targets, such as bare-board
8184 targets without an operating system, the `-T' option may be
8185 required when linking to avoid references to undefined symbols.
8188 Pass OPTION as an option to the linker. You can use this to
8189 supply system-specific linker options which GCC does not know how
8192 If you want to pass an option that takes a separate argument, you
8193 must use `-Xlinker' twice, once for the option and once for the
8194 argument. For example, to pass `-assert definitions', you must
8195 write `-Xlinker -assert -Xlinker definitions'. It does not work
8196 to write `-Xlinker "-assert definitions"', because this passes the
8197 entire string as a single argument, which is not what the linker
8200 When using the GNU linker, it is usually more convenient to pass
8201 arguments to linker options using the `OPTION=VALUE' syntax than
8202 as separate arguments. For example, you can specify `-Xlinker
8203 -Map=output.map' rather than `-Xlinker -Map -Xlinker output.map'.
8204 Other linkers may not support this syntax for command-line options.
8207 Pass OPTION as an option to the linker. If OPTION contains
8208 commas, it is split into multiple options at the commas. You can
8209 use this syntax to pass an argument to the option. For example,
8210 `-Wl,-Map,output.map' passes `-Map output.map' to the linker.
8211 When using the GNU linker, you can also get the same effect with
8212 `-Wl,-Map=output.map'.
8215 Pretend the symbol SYMBOL is undefined, to force linking of
8216 library modules to define it. You can use `-u' multiple times with
8217 different symbols to force loading of additional library modules.
8219 ---------- Footnotes ----------
8221 (1) On some systems, `gcc -shared' needs to build supplementary stub
8222 code for constructors to work. On multi-libbed systems, `gcc -shared'
8223 must select the correct support libraries to link against. Failing to
8224 supply the correct flags may lead to subtle defects. Supplying them in
8225 cases where they are not necessary is innocuous.
8228 File: gcc.info, Node: Directory Options, Next: Spec Files, Prev: Link Options, Up: Invoking GCC
8230 3.14 Options for Directory Search
8231 =================================
8233 These options specify directories to search for header files, for
8234 libraries and for parts of the compiler:
8237 Add the directory DIR to the head of the list of directories to be
8238 searched for header files. This can be used to override a system
8239 header file, substituting your own version, since these
8240 directories are searched before the system header file
8241 directories. However, you should not use this option to add
8242 directories that contain vendor-supplied system header files (use
8243 `-isystem' for that). If you use more than one `-I' option, the
8244 directories are scanned in left-to-right order; the standard
8245 system directories come after.
8247 If a standard system include directory, or a directory specified
8248 with `-isystem', is also specified with `-I', the `-I' option will
8249 be ignored. The directory will still be searched but as a system
8250 directory at its normal position in the system include chain.
8251 This is to ensure that GCC's procedure to fix buggy system headers
8252 and the ordering for the include_next directive are not
8253 inadvertently changed. If you really need to change the search
8254 order for system directories, use the `-nostdinc' and/or
8258 Add the directory DIR to the head of the list of directories to be
8259 searched for header files only for the case of `#include "FILE"';
8260 they are not searched for `#include <FILE>', otherwise just like
8264 Add directory DIR to the list of directories to be searched for
8268 This option specifies where to find the executables, libraries,
8269 include files, and data files of the compiler itself.
8271 The compiler driver program runs one or more of the subprograms
8272 `cpp', `cc1', `as' and `ld'. It tries PREFIX as a prefix for each
8273 program it tries to run, both with and without `MACHINE/VERSION/'
8274 (*note Target Options::).
8276 For each subprogram to be run, the compiler driver first tries the
8277 `-B' prefix, if any. If that name is not found, or if `-B' was
8278 not specified, the driver tries two standard prefixes, which are
8279 `/usr/lib/gcc/' and `/usr/local/lib/gcc/'. If neither of those
8280 results in a file name that is found, the unmodified program name
8281 is searched for using the directories specified in your `PATH'
8282 environment variable.
8284 The compiler will check to see if the path provided by the `-B'
8285 refers to a directory, and if necessary it will add a directory
8286 separator character at the end of the path.
8288 `-B' prefixes that effectively specify directory names also apply
8289 to libraries in the linker, because the compiler translates these
8290 options into `-L' options for the linker. They also apply to
8291 includes files in the preprocessor, because the compiler
8292 translates these options into `-isystem' options for the
8293 preprocessor. In this case, the compiler appends `include' to the
8296 The run-time support file `libgcc.a' can also be searched for using
8297 the `-B' prefix, if needed. If it is not found there, the two
8298 standard prefixes above are tried, and that is all. The file is
8299 left out of the link if it is not found by those means.
8301 Another way to specify a prefix much like the `-B' prefix is to use
8302 the environment variable `GCC_EXEC_PREFIX'. *Note Environment
8305 As a special kludge, if the path provided by `-B' is
8306 `[dir/]stageN/', where N is a number in the range 0 to 9, then it
8307 will be replaced by `[dir/]include'. This is to help with
8308 boot-strapping the compiler.
8311 Process FILE after the compiler reads in the standard `specs'
8312 file, in order to override the defaults that the `gcc' driver
8313 program uses when determining what switches to pass to `cc1',
8314 `cc1plus', `as', `ld', etc. More than one `-specs=FILE' can be
8315 specified on the command line, and they are processed in order,
8319 Use DIR as the logical root directory for headers and libraries.
8320 For example, if the compiler would normally search for headers in
8321 `/usr/include' and libraries in `/usr/lib', it will instead search
8322 `DIR/usr/include' and `DIR/usr/lib'.
8324 If you use both this option and the `-isysroot' option, then the
8325 `--sysroot' option will apply to libraries, but the `-isysroot'
8326 option will apply to header files.
8328 The GNU linker (beginning with version 2.16) has the necessary
8329 support for this option. If your linker does not support this
8330 option, the header file aspect of `--sysroot' will still work, but
8331 the library aspect will not.
8334 This option has been deprecated. Please use `-iquote' instead for
8335 `-I' directories before the `-I-' and remove the `-I-'. Any
8336 directories you specify with `-I' options before the `-I-' option
8337 are searched only for the case of `#include "FILE"'; they are not
8338 searched for `#include <FILE>'.
8340 If additional directories are specified with `-I' options after
8341 the `-I-', these directories are searched for all `#include'
8342 directives. (Ordinarily _all_ `-I' directories are used this way.)
8344 In addition, the `-I-' option inhibits the use of the current
8345 directory (where the current input file came from) as the first
8346 search directory for `#include "FILE"'. There is no way to
8347 override this effect of `-I-'. With `-I.' you can specify
8348 searching the directory which was current when the compiler was
8349 invoked. That is not exactly the same as what the preprocessor
8350 does by default, but it is often satisfactory.
8352 `-I-' does not inhibit the use of the standard system directories
8353 for header files. Thus, `-I-' and `-nostdinc' are independent.
8356 File: gcc.info, Node: Spec Files, Next: Target Options, Prev: Directory Options, Up: Invoking GCC
8358 3.15 Specifying subprocesses and the switches to pass to them
8359 =============================================================
8361 `gcc' is a driver program. It performs its job by invoking a sequence
8362 of other programs to do the work of compiling, assembling and linking.
8363 GCC interprets its command-line parameters and uses these to deduce
8364 which programs it should invoke, and which command-line options it
8365 ought to place on their command lines. This behavior is controlled by
8366 "spec strings". In most cases there is one spec string for each
8367 program that GCC can invoke, but a few programs have multiple spec
8368 strings to control their behavior. The spec strings built into GCC can
8369 be overridden by using the `-specs=' command-line switch to specify a
8372 "Spec files" are plaintext files that are used to construct spec
8373 strings. They consist of a sequence of directives separated by blank
8374 lines. The type of directive is determined by the first non-whitespace
8375 character on the line and it can be one of the following:
8378 Issues a COMMAND to the spec file processor. The commands that can
8382 Search for FILE and insert its text at the current point in
8385 `%include_noerr <FILE>'
8386 Just like `%include', but do not generate an error message if
8387 the include file cannot be found.
8389 `%rename OLD_NAME NEW_NAME'
8390 Rename the spec string OLD_NAME to NEW_NAME.
8394 This tells the compiler to create, override or delete the named
8395 spec string. All lines after this directive up to the next
8396 directive or blank line are considered to be the text for the spec
8397 string. If this results in an empty string then the spec will be
8398 deleted. (Or, if the spec did not exist, then nothing will
8399 happened.) Otherwise, if the spec does not currently exist a new
8400 spec will be created. If the spec does exist then its contents
8401 will be overridden by the text of this directive, unless the first
8402 character of that text is the `+' character, in which case the
8403 text will be appended to the spec.
8406 Creates a new `[SUFFIX] spec' pair. All lines after this directive
8407 and up to the next directive or blank line are considered to make
8408 up the spec string for the indicated suffix. When the compiler
8409 encounters an input file with the named suffix, it will processes
8410 the spec string in order to work out how to compile that file.
8416 This says that any input file whose name ends in `.ZZ' should be
8417 passed to the program `z-compile', which should be invoked with the
8418 command-line switch `-input' and with the result of performing the
8419 `%i' substitution. (See below.)
8421 As an alternative to providing a spec string, the text that
8422 follows a suffix directive can be one of the following:
8425 This says that the suffix is an alias for a known LANGUAGE.
8426 This is similar to using the `-x' command-line switch to GCC
8427 to specify a language explicitly. For example:
8432 Says that .ZZ files are, in fact, C++ source files.
8435 This causes an error messages saying:
8437 NAME compiler not installed on this system.
8439 GCC already has an extensive list of suffixes built into it. This
8440 directive will add an entry to the end of the list of suffixes, but
8441 since the list is searched from the end backwards, it is
8442 effectively possible to override earlier entries using this
8446 GCC has the following spec strings built into it. Spec files can
8447 override these strings or create their own. Note that individual
8448 targets can also add their own spec strings to this list.
8450 asm Options to pass to the assembler
8451 asm_final Options to pass to the assembler post-processor
8452 cpp Options to pass to the C preprocessor
8453 cc1 Options to pass to the C compiler
8454 cc1plus Options to pass to the C++ compiler
8455 endfile Object files to include at the end of the link
8456 link Options to pass to the linker
8457 lib Libraries to include on the command line to the linker
8458 libgcc Decides which GCC support library to pass to the linker
8459 linker Sets the name of the linker
8460 predefines Defines to be passed to the C preprocessor
8461 signed_char Defines to pass to CPP to say whether `char' is signed
8463 startfile Object files to include at the start of the link
8465 Here is a small example of a spec file:
8470 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8472 This example renames the spec called `lib' to `old_lib' and then
8473 overrides the previous definition of `lib' with a new one. The new
8474 definition adds in some extra command-line options before including the
8475 text of the old definition.
8477 "Spec strings" are a list of command-line options to be passed to their
8478 corresponding program. In addition, the spec strings can contain
8479 `%'-prefixed sequences to substitute variable text or to conditionally
8480 insert text into the command line. Using these constructs it is
8481 possible to generate quite complex command lines.
8483 Here is a table of all defined `%'-sequences for spec strings. Note
8484 that spaces are not generated automatically around the results of
8485 expanding these sequences. Therefore you can concatenate them together
8486 or combine them with constant text in a single argument.
8489 Substitute one `%' into the program name or argument.
8492 Substitute the name of the input file being processed.
8495 Substitute the basename of the input file being processed. This
8496 is the substring up to (and not including) the last period and not
8497 including the directory.
8500 This is the same as `%b', but include the file suffix (text after
8504 Marks the argument containing or following the `%d' as a temporary
8505 file name, so that that file will be deleted if GCC exits
8506 successfully. Unlike `%g', this contributes no text to the
8510 Substitute a file name that has suffix SUFFIX and is chosen once
8511 per compilation, and mark the argument in the same way as `%d'.
8512 To reduce exposure to denial-of-service attacks, the file name is
8513 now chosen in a way that is hard to predict even when previously
8514 chosen file names are known. For example, `%g.s ... %g.o ... %g.s'
8515 might turn into `ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s'. SUFFIX
8516 matches the regexp `[.A-Za-z]*' or the special string `%O', which
8517 is treated exactly as if `%O' had been preprocessed. Previously,
8518 `%g' was simply substituted with a file name chosen once per
8519 compilation, without regard to any appended suffix (which was
8520 therefore treated just like ordinary text), making such attacks
8521 more likely to succeed.
8524 Like `%g', but generates a new temporary file name even if
8525 `%uSUFFIX' was already seen.
8528 Substitutes the last file name generated with `%uSUFFIX',
8529 generating a new one if there is no such last file name. In the
8530 absence of any `%uSUFFIX', this is just like `%gSUFFIX', except
8531 they don't share the same suffix _space_, so `%g.s ... %U.s ...
8532 %g.s ... %U.s' would involve the generation of two distinct file
8533 names, one for each `%g.s' and another for each `%U.s'.
8534 Previously, `%U' was simply substituted with a file name chosen
8535 for the previous `%u', without regard to any appended suffix.
8538 Substitutes the name of the `HOST_BIT_BUCKET', if any, and if it is
8539 writable, and if save-temps is off; otherwise, substitute the name
8540 of a temporary file, just like `%u'. This temporary file is not
8541 meant for communication between processes, but rather as a junk
8546 Like `%g', except if `-pipe' is in effect. In that case `%|'
8547 substitutes a single dash and `%m' substitutes nothing at all.
8548 These are the two most common ways to instruct a program that it
8549 should read from standard input or write to standard output. If
8550 you need something more elaborate you can use an `%{pipe:`X'}'
8551 construct: see for example `f/lang-specs.h'.
8554 Substitutes .SUFFIX for the suffixes of a matched switch's args
8555 when it is subsequently output with `%*'. SUFFIX is terminated by
8556 the next space or %.
8559 Marks the argument containing or following the `%w' as the
8560 designated output file of this compilation. This puts the argument
8561 into the sequence of arguments that `%o' will substitute later.
8564 Substitutes the names of all the output files, with spaces
8565 automatically placed around them. You should write spaces around
8566 the `%o' as well or the results are undefined. `%o' is for use in
8567 the specs for running the linker. Input files whose names have no
8568 recognized suffix are not compiled at all, but they are included
8569 among the output files, so they will be linked.
8572 Substitutes the suffix for object files. Note that this is
8573 handled specially when it immediately follows `%g, %u, or %U',
8574 because of the need for those to form complete file names. The
8575 handling is such that `%O' is treated exactly as if it had already
8576 been substituted, except that `%g, %u, and %U' do not currently
8577 support additional SUFFIX characters following `%O' as they would
8578 following, for example, `.o'.
8581 Substitutes the standard macro predefinitions for the current
8582 target machine. Use this when running `cpp'.
8585 Like `%p', but puts `__' before and after the name of each
8586 predefined macro, except for macros that start with `__' or with
8587 `_L', where L is an uppercase letter. This is for ISO C.
8590 Substitute any of `-iprefix' (made from `GCC_EXEC_PREFIX'),
8591 `-isysroot' (made from `TARGET_SYSTEM_ROOT'), `-isystem' (made
8592 from `COMPILER_PATH' and `-B' options) and `-imultilib' as
8596 Current argument is the name of a library or startup file of some
8597 sort. Search for that file in a standard list of directories and
8598 substitute the full name found.
8601 Print STR as an error message. STR is terminated by a newline.
8602 Use this when inconsistent options are detected.
8605 Substitute the contents of spec string NAME at this point.
8608 Like `%(...)' but put `__' around `-D' arguments.
8611 Accumulate an option for `%X'.
8614 Output the accumulated linker options specified by `-Wl' or a `%x'
8618 Output the accumulated assembler options specified by `-Wa'.
8621 Output the accumulated preprocessor options specified by `-Wp'.
8624 Process the `asm' spec. This is used to compute the switches to
8625 be passed to the assembler.
8628 Process the `asm_final' spec. This is a spec string for passing
8629 switches to an assembler post-processor, if such a program is
8633 Process the `link' spec. This is the spec for computing the
8634 command line passed to the linker. Typically it will make use of
8635 the `%L %G %S %D and %E' sequences.
8638 Dump out a `-L' option for each directory that GCC believes might
8639 contain startup files. If the target supports multilibs then the
8640 current multilib directory will be prepended to each of these
8644 Process the `lib' spec. This is a spec string for deciding which
8645 libraries should be included on the command line to the linker.
8648 Process the `libgcc' spec. This is a spec string for deciding
8649 which GCC support library should be included on the command line
8653 Process the `startfile' spec. This is a spec for deciding which
8654 object files should be the first ones passed to the linker.
8655 Typically this might be a file named `crt0.o'.
8658 Process the `endfile' spec. This is a spec string that specifies
8659 the last object files that will be passed to the linker.
8662 Process the `cpp' spec. This is used to construct the arguments
8663 to be passed to the C preprocessor.
8666 Process the `cc1' spec. This is used to construct the options to
8667 be passed to the actual C compiler (`cc1').
8670 Process the `cc1plus' spec. This is used to construct the options
8671 to be passed to the actual C++ compiler (`cc1plus').
8674 Substitute the variable part of a matched option. See below.
8675 Note that each comma in the substituted string is replaced by a
8679 Remove all occurrences of `-S' from the command line. Note--this
8680 command is position dependent. `%' commands in the spec string
8681 before this one will see `-S', `%' commands in the spec string
8682 after this one will not.
8685 Call the named function FUNCTION, passing it ARGS. ARGS is first
8686 processed as a nested spec string, then split into an argument
8687 vector in the usual fashion. The function returns a string which
8688 is processed as if it had appeared literally as part of the
8691 The following built-in spec functions are provided:
8694 The `getenv' spec function takes two arguments: an environment
8695 variable name and a string. If the environment variable is
8696 not defined, a fatal error is issued. Otherwise, the return
8697 value is the value of the environment variable concatenated
8698 with the string. For example, if `TOPDIR' is defined as
8699 `/path/to/top', then:
8701 %:getenv(TOPDIR /include)
8703 expands to `/path/to/top/include'.
8706 The `if-exists' spec function takes one argument, an absolute
8707 pathname to a file. If the file exists, `if-exists' returns
8708 the pathname. Here is a small example of its usage:
8711 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8714 The `if-exists-else' spec function is similar to the
8715 `if-exists' spec function, except that it takes two
8716 arguments. The first argument is an absolute pathname to a
8717 file. If the file exists, `if-exists-else' returns the
8718 pathname. If it does not exist, it returns the second
8719 argument. This way, `if-exists-else' can be used to select
8720 one file or another, based on the existence of the first.
8721 Here is a small example of its usage:
8724 crt0%O%s %:if-exists(crti%O%s) \
8725 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8728 The `replace-outfile' spec function takes two arguments. It
8729 looks for the first argument in the outfiles array and
8730 replaces it with the second argument. Here is a small
8731 example of its usage:
8733 %{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)}
8735 ``print-asm-header''
8736 The `print-asm-header' function takes no arguments and simply
8737 prints a banner like:
8742 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8744 It is used to separate compiler options from assembler options
8745 in the `--target-help' output.
8748 Substitutes the `-S' switch, if that switch was given to GCC. If
8749 that switch was not specified, this substitutes nothing. Note that
8750 the leading dash is omitted when specifying this option, and it is
8751 automatically inserted if the substitution is performed. Thus the
8752 spec string `%{foo}' would match the command-line option `-foo'
8753 and would output the command line option `-foo'.
8756 Like %{`S'} but mark last argument supplied within as a file to be
8760 Substitutes all the switches specified to GCC whose names start
8761 with `-S', but which also take an argument. This is used for
8762 switches like `-o', `-D', `-I', etc. GCC considers `-o foo' as
8763 being one switch whose names starts with `o'. %{o*} would
8764 substitute this text, including the space. Thus two arguments
8768 Like %{`S'*}, but preserve order of `S' and `T' options (the order
8769 of `S' and `T' in the spec is not significant). There can be any
8770 number of ampersand-separated variables; for each the wild card is
8771 optional. Useful for CPP as `%{D*&U*&A*}'.
8774 Substitutes `X', if the `-S' switch was given to GCC.
8777 Substitutes `X', if the `-S' switch was _not_ given to GCC.
8780 Substitutes `X' if one or more switches whose names start with
8781 `-S' are specified to GCC. Normally `X' is substituted only once,
8782 no matter how many such switches appeared. However, if `%*'
8783 appears somewhere in `X', then `X' will be substituted once for
8784 each matching switch, with the `%*' replaced by the part of that
8785 switch that matched the `*'.
8788 Substitutes `X', if processing a file with suffix `S'.
8791 Substitutes `X', if _not_ processing a file with suffix `S'.
8794 Substitutes `X', if processing a file for language `S'.
8797 Substitutes `X', if not processing a file for language `S'.
8800 Substitutes `X' if either `-S' or `-P' was given to GCC. This may
8801 be combined with `!', `.', `,', and `*' sequences as well,
8802 although they have a stronger binding than the `|'. If `%*'
8803 appears in `X', all of the alternatives must be starred, and only
8804 the first matching alternative is substituted.
8806 For example, a spec string like this:
8808 %{.c:-foo} %{!.c:-bar} %{.c|d:-baz} %{!.c|d:-boggle}
8810 will output the following command-line options from the following
8811 input command-line options:
8815 -d fred.c -foo -baz -boggle
8816 -d jim.d -bar -baz -boggle
8819 If `S' was given to GCC, substitutes `X'; else if `T' was given to
8820 GCC, substitutes `Y'; else substitutes `D'. There can be as many
8821 clauses as you need. This may be combined with `.', `,', `!',
8822 `|', and `*' as needed.
8825 The conditional text `X' in a %{`S':`X'} or similar construct may
8826 contain other nested `%' constructs or spaces, or even newlines. They
8827 are processed as usual, as described above. Trailing white space in
8828 `X' is ignored. White space may also appear anywhere on the left side
8829 of the colon in these constructs, except between `.' or `*' and the
8832 The `-O', `-f', `-m', and `-W' switches are handled specifically in
8833 these constructs. If another value of `-O' or the negated form of a
8834 `-f', `-m', or `-W' switch is found later in the command line, the
8835 earlier switch value is ignored, except with {`S'*} where `S' is just
8836 one letter, which passes all matching options.
8838 The character `|' at the beginning of the predicate text is used to
8839 indicate that a command should be piped to the following command, but
8840 only if `-pipe' is specified.
8842 It is built into GCC which switches take arguments and which do not.
8843 (You might think it would be useful to generalize this to allow each
8844 compiler's spec to say which switches take arguments. But this cannot
8845 be done in a consistent fashion. GCC cannot even decide which input
8846 files have been specified without knowing which switches take arguments,
8847 and it must know which input files to compile in order to tell which
8850 GCC also knows implicitly that arguments starting in `-l' are to be
8851 treated as compiler output files, and passed to the linker in their
8852 proper position among the other output files.
8855 File: gcc.info, Node: Target Options, Next: Submodel Options, Prev: Spec Files, Up: Invoking GCC
8857 3.16 Specifying Target Machine and Compiler Version
8858 ===================================================
8860 The usual way to run GCC is to run the executable called `gcc', or
8861 `<machine>-gcc' when cross-compiling, or `<machine>-gcc-<version>' to
8862 run a version other than the one that was installed last. Sometimes
8863 this is inconvenient, so GCC provides options that will switch to
8864 another cross-compiler or version.
8867 The argument MACHINE specifies the target machine for compilation.
8869 The value to use for MACHINE is the same as was specified as the
8870 machine type when configuring GCC as a cross-compiler. For
8871 example, if a cross-compiler was configured with `configure
8872 arm-elf', meaning to compile for an arm processor with elf
8873 binaries, then you would specify `-b arm-elf' to run that cross
8874 compiler. Because there are other options beginning with `-b', the
8875 configuration must contain a hyphen, or `-b' alone should be one
8876 argument followed by the configuration in the next argument.
8879 The argument VERSION specifies which version of GCC to run. This
8880 is useful when multiple versions are installed. For example,
8881 VERSION might be `4.0', meaning to run GCC version 4.0.
8883 The `-V' and `-b' options work by running the
8884 `<machine>-gcc-<version>' executable, so there's no real reason to use
8885 them if you can just run that directly.
8888 File: gcc.info, Node: Submodel Options, Next: Code Gen Options, Prev: Target Options, Up: Invoking GCC
8890 3.17 Hardware Models and Configurations
8891 =======================================
8893 Earlier we discussed the standard option `-b' which chooses among
8894 different installed compilers for completely different target machines,
8895 such as VAX vs. 68000 vs. 80386.
8897 In addition, each of these target machine types can have its own
8898 special options, starting with `-m', to choose among various hardware
8899 models or configurations--for example, 68010 vs 68020, floating
8900 coprocessor or none. A single installed version of the compiler can
8901 compile for any model or configuration, according to the options
8904 Some configurations of the compiler also support additional special
8905 options, usually for compatibility with other compilers on the same
8913 * Blackfin Options::
8917 * DEC Alpha Options::
8918 * DEC Alpha/VMS Options::
8921 * GNU/Linux Options::
8924 * i386 and x86-64 Options::
8925 * i386 and x86-64 Windows Options::
8936 * picoChip Options::
8938 * RS/6000 and PowerPC Options::
8939 * S/390 and zSeries Options::
8944 * System V Options::
8949 * Xstormy16 Options::
8954 File: gcc.info, Node: ARC Options, Next: ARM Options, Up: Submodel Options
8959 These options are defined for ARC implementations:
8962 Compile code for little endian mode. This is the default.
8965 Compile code for big endian mode.
8968 Prepend the name of the cpu to all public symbol names. In
8969 multiple-processor systems, there are many ARC variants with
8970 different instruction and register set characteristics. This flag
8971 prevents code compiled for one cpu to be linked with code compiled
8972 for another. No facility exists for handling variants that are
8973 "almost identical". This is an all or nothing option.
8976 Compile code for ARC variant CPU. Which variants are supported
8977 depend on the configuration. All variants support `-mcpu=base',
8978 this is the default.
8980 `-mtext=TEXT-SECTION'
8981 `-mdata=DATA-SECTION'
8982 `-mrodata=READONLY-DATA-SECTION'
8983 Put functions, data, and readonly data in TEXT-SECTION,
8984 DATA-SECTION, and READONLY-DATA-SECTION respectively by default.
8985 This can be overridden with the `section' attribute. *Note
8986 Variable Attributes::.
8988 `-mfix-cortex-m3-ldrd'
8989 Some Cortex-M3 cores can cause data corruption when `ldrd'
8990 instructions with overlapping destination and base registers are
8991 used. This option avoids generating these instructions. This
8992 option is enabled by default when `-mcpu=cortex-m3' is specified.
8996 File: gcc.info, Node: ARM Options, Next: AVR Options, Prev: ARC Options, Up: Submodel Options
9001 These `-m' options are defined for Advanced RISC Machines (ARM)
9005 Generate code for the specified ABI. Permissible values are:
9006 `apcs-gnu', `atpcs', `aapcs', `aapcs-linux' and `iwmmxt'.
9009 Generate a stack frame that is compliant with the ARM Procedure
9010 Call Standard for all functions, even if this is not strictly
9011 necessary for correct execution of the code. Specifying
9012 `-fomit-frame-pointer' with this option will cause the stack
9013 frames not to be generated for leaf functions. The default is
9017 This is a synonym for `-mapcs-frame'.
9020 Generate code which supports calling between the ARM and Thumb
9021 instruction sets. Without this option the two instruction sets
9022 cannot be reliably used inside one program. The default is
9023 `-mno-thumb-interwork', since slightly larger code is generated
9024 when `-mthumb-interwork' is specified.
9027 Prevent the reordering of instructions in the function prolog, or
9028 the merging of those instruction with the instructions in the
9029 function's body. This means that all functions will start with a
9030 recognizable set of instructions (or in fact one of a choice from
9031 a small set of different function prologues), and this information
9032 can be used to locate the start if functions inside an executable
9033 piece of code. The default is `-msched-prolog'.
9036 Specifies which floating-point ABI to use. Permissible values
9037 are: `soft', `softfp' and `hard'.
9039 Specifying `soft' causes GCC to generate output containing library
9040 calls for floating-point operations. `softfp' allows the
9041 generation of code using hardware floating-point instructions, but
9042 still uses the soft-float calling conventions. `hard' allows
9043 generation of floating-point instructions and uses FPU-specific
9044 calling conventions.
9046 Using `-mfloat-abi=hard' with VFP coprocessors is not supported.
9047 Use `-mfloat-abi=softfp' with the appropriate `-mfpu' option to
9048 allow the compiler to generate code that makes use of the hardware
9049 floating-point capabilities for these CPUs.
9051 The default depends on the specific target configuration. Note
9052 that the hard-float and soft-float ABIs are not link-compatible;
9053 you must compile your entire program with the same ABI, and link
9054 with a compatible set of libraries.
9057 Equivalent to `-mfloat-abi=hard'.
9060 Equivalent to `-mfloat-abi=soft'.
9063 Generate code for a processor running in little-endian mode. This
9064 is the default for all standard configurations.
9067 Generate code for a processor running in big-endian mode; the
9068 default is to compile code for a little-endian processor.
9070 `-mwords-little-endian'
9071 This option only applies when generating code for big-endian
9072 processors. Generate code for a little-endian word order but a
9073 big-endian byte order. That is, a byte order of the form
9074 `32107654'. Note: this option should only be used if you require
9075 compatibility with code for big-endian ARM processors generated by
9076 versions of the compiler prior to 2.8.
9079 This specifies the name of the target ARM processor. GCC uses
9080 this name to determine what kind of instructions it can emit when
9081 generating assembly code. Permissible names are: `arm2', `arm250',
9082 `arm3', `arm6', `arm60', `arm600', `arm610', `arm620', `arm7',
9083 `arm7m', `arm7d', `arm7dm', `arm7di', `arm7dmi', `arm70', `arm700',
9084 `arm700i', `arm710', `arm710c', `arm7100', `arm720', `arm7500',
9085 `arm7500fe', `arm7tdmi', `arm7tdmi-s', `arm710t', `arm720t',
9086 `arm740t', `strongarm', `strongarm110', `strongarm1100',
9087 `strongarm1110', `arm8', `arm810', `arm9', `arm9e', `arm920',
9088 `arm920t', `arm922t', `arm946e-s', `arm966e-s', `arm968e-s',
9089 `arm926ej-s', `arm940t', `arm9tdmi', `arm10tdmi', `arm1020t',
9090 `arm1026ej-s', `arm10e', `arm1020e', `arm1022e', `arm1136j-s',
9091 `arm1136jf-s', `mpcore', `mpcorenovfp', `arm1156t2-s',
9092 `arm1176jz-s', `arm1176jzf-s', `cortex-a8', `cortex-a9',
9093 `cortex-r4', `cortex-r4f', `cortex-m3', `cortex-m1', `xscale',
9094 `iwmmxt', `iwmmxt2', `ep9312'.
9097 This option is very similar to the `-mcpu=' option, except that
9098 instead of specifying the actual target processor type, and hence
9099 restricting which instructions can be used, it specifies that GCC
9100 should tune the performance of the code as if the target were of
9101 the type specified in this option, but still choosing the
9102 instructions that it will generate based on the cpu specified by a
9103 `-mcpu=' option. For some ARM implementations better performance
9104 can be obtained by using this option.
9107 This specifies the name of the target ARM architecture. GCC uses
9108 this name to determine what kind of instructions it can emit when
9109 generating assembly code. This option can be used in conjunction
9110 with or instead of the `-mcpu=' option. Permissible names are:
9111 `armv2', `armv2a', `armv3', `armv3m', `armv4', `armv4t', `armv5',
9112 `armv5t', `armv5e', `armv5te', `armv6', `armv6j', `armv6t2',
9113 `armv6z', `armv6zk', `armv6-m', `armv7', `armv7-a', `armv7-r',
9114 `armv7-m', `iwmmxt', `iwmmxt2', `ep9312'.
9119 This specifies what floating point hardware (or hardware
9120 emulation) is available on the target. Permissible names are:
9121 `fpa', `fpe2', `fpe3', `maverick', `vfp', `vfpv3', `vfpv3-d16' and
9122 `neon'. `-mfp' and `-mfpe' are synonyms for `-mfpu'=`fpe'NUMBER,
9123 for compatibility with older versions of GCC.
9125 If `-msoft-float' is specified this specifies the format of
9126 floating point values.
9128 `-mstructure-size-boundary=N'
9129 The size of all structures and unions will be rounded up to a
9130 multiple of the number of bits set by this option. Permissible
9131 values are 8, 32 and 64. The default value varies for different
9132 toolchains. For the COFF targeted toolchain the default value is
9133 8. A value of 64 is only allowed if the underlying ABI supports
9136 Specifying the larger number can produce faster, more efficient
9137 code, but can also increase the size of the program. Different
9138 values are potentially incompatible. Code compiled with one value
9139 cannot necessarily expect to work with code or libraries compiled
9140 with another value, if they exchange information using structures
9143 `-mabort-on-noreturn'
9144 Generate a call to the function `abort' at the end of a `noreturn'
9145 function. It will be executed if the function tries to return.
9149 Tells the compiler to perform function calls by first loading the
9150 address of the function into a register and then performing a
9151 subroutine call on this register. This switch is needed if the
9152 target function will lie outside of the 64 megabyte addressing
9153 range of the offset based version of subroutine call instruction.
9155 Even if this switch is enabled, not all function calls will be
9156 turned into long calls. The heuristic is that static functions,
9157 functions which have the `short-call' attribute, functions that
9158 are inside the scope of a `#pragma no_long_calls' directive and
9159 functions whose definitions have already been compiled within the
9160 current compilation unit, will not be turned into long calls. The
9161 exception to this rule is that weak function definitions,
9162 functions with the `long-call' attribute or the `section'
9163 attribute, and functions that are within the scope of a `#pragma
9164 long_calls' directive, will always be turned into long calls.
9166 This feature is not enabled by default. Specifying
9167 `-mno-long-calls' will restore the default behavior, as will
9168 placing the function calls within the scope of a `#pragma
9169 long_calls_off' directive. Note these switches have no effect on
9170 how the compiler generates code to handle function calls via
9174 Treat the register used for PIC addressing as read-only, rather
9175 than loading it in the prologue for each function. The run-time
9176 system is responsible for initializing this register with an
9177 appropriate value before execution begins.
9179 `-mpic-register=REG'
9180 Specify the register to be used for PIC addressing. The default
9181 is R10 unless stack-checking is enabled, when R9 is used.
9183 `-mcirrus-fix-invalid-insns'
9184 Insert NOPs into the instruction stream to in order to work around
9185 problems with invalid Maverick instruction combinations. This
9186 option is only valid if the `-mcpu=ep9312' option has been used to
9187 enable generation of instructions for the Cirrus Maverick floating
9188 point co-processor. This option is not enabled by default, since
9189 the problem is only present in older Maverick implementations.
9190 The default can be re-enabled by use of the
9191 `-mno-cirrus-fix-invalid-insns' switch.
9193 `-mpoke-function-name'
9194 Write the name of each function into the text section, directly
9195 preceding the function prologue. The generated code is similar to
9199 .ascii "arm_poke_function_name", 0
9202 .word 0xff000000 + (t1 - t0)
9203 arm_poke_function_name
9205 stmfd sp!, {fp, ip, lr, pc}
9208 When performing a stack backtrace, code can inspect the value of
9209 `pc' stored at `fp + 0'. If the trace function then looks at
9210 location `pc - 12' and the top 8 bits are set, then we know that
9211 there is a function name embedded immediately preceding this
9212 location and has length `((pc[-3]) & 0xff000000)'.
9215 Generate code for the Thumb instruction set. The default is to
9216 use the 32-bit ARM instruction set. This option automatically
9217 enables either 16-bit Thumb-1 or mixed 16/32-bit Thumb-2
9218 instructions based on the `-mcpu=NAME' and `-march=NAME' options.
9221 Generate a stack frame that is compliant with the Thumb Procedure
9222 Call Standard for all non-leaf functions. (A leaf function is one
9223 that does not call any other functions.) The default is
9227 Generate a stack frame that is compliant with the Thumb Procedure
9228 Call Standard for all leaf functions. (A leaf function is one
9229 that does not call any other functions.) The default is
9230 `-mno-apcs-leaf-frame'.
9232 `-mcallee-super-interworking'
9233 Gives all externally visible functions in the file being compiled
9234 an ARM instruction set header which switches to Thumb mode before
9235 executing the rest of the function. This allows these functions
9236 to be called from non-interworking code.
9238 `-mcaller-super-interworking'
9239 Allows calls via function pointers (including virtual functions) to
9240 execute correctly regardless of whether the target code has been
9241 compiled for interworking or not. There is a small overhead in
9242 the cost of executing a function pointer if this option is enabled.
9245 Specify the access model for the thread local storage pointer.
9246 The valid models are `soft', which generates calls to
9247 `__aeabi_read_tp', `cp15', which fetches the thread pointer from
9248 `cp15' directly (supported in the arm6k architecture), and `auto',
9249 which uses the best available method for the selected processor.
9250 The default setting is `auto'.
9252 `-mword-relocations'
9253 Only generate absolute relocations on word sized values (i.e.
9254 R_ARM_ABS32). This is enabled by default on targets (uClinux,
9255 SymbianOS) where the runtime loader imposes this restriction, and
9256 when `-fpic' or `-fPIC' is specified.
9259 Enable Android specific compilier options.
9261 If this option is used, a preprocessor macro `__ANDROID__' is
9262 defined and has the value 1 during compilation. The option also
9263 implies C/C++ options `-fno-exceptions' `-fpic' `-mthumb-interwork'
9264 `-fno-short-enums' and C++ option `-fno-rtti'. These implied
9265 options can be overridden. For example RTTI in C++ code can still
9266 be enabled with -frtti even when -mandroid is also used.
9268 Linking options depend on whether a static executable, a dynamic
9269 executable or a shared library is built. When `-static' is given,
9270 `-mandroid' implies linking flag `-Bstatic', start file
9271 `crtbegin_static.o' and end file `crtend_android.o'.
9273 When `-shared' is given, `-mandroid' implies the linking flag
9274 `-Bsymbolic' and no start and end files.
9276 When none of `-static' and `-shared' is given, `-mandroid' implies
9277 linking flags `-Bdynamic -dynamic-linker /system/bin/linker',
9278 start file `crtbegin_dynamic.o' and end file `crtend_android.o'.
9279 The dynamic linker used can be overriden by another
9280 `-dynamic-linker' in command line.
9282 The linking option `-ldl' is also added if `-static' is not given.
9284 If more than one of `-dynamic', `-static' and `-shared' are given,
9285 behaviour of `-mandroid' is undefined.
9289 File: gcc.info, Node: AVR Options, Next: Blackfin Options, Prev: ARM Options, Up: Submodel Options
9294 These options are defined for AVR implementations:
9297 Specify ATMEL AVR instruction set or MCU type.
9299 Instruction set avr1 is for the minimal AVR core, not supported by
9300 the C compiler, only for assembler programs (MCU types: at90s1200,
9301 attiny10, attiny11, attiny12, attiny15, attiny28).
9303 Instruction set avr2 (default) is for the classic AVR core with up
9304 to 8K program memory space (MCU types: at90s2313, at90s2323,
9305 attiny22, at90s2333, at90s2343, at90s4414, at90s4433, at90s4434,
9306 at90s8515, at90c8534, at90s8535).
9308 Instruction set avr3 is for the classic AVR core with up to 128K
9309 program memory space (MCU types: atmega103, atmega603, at43usb320,
9312 Instruction set avr4 is for the enhanced AVR core with up to 8K
9313 program memory space (MCU types: atmega8, atmega83, atmega85).
9315 Instruction set avr5 is for the enhanced AVR core with up to 128K
9316 program memory space (MCU types: atmega16, atmega161, atmega163,
9317 atmega32, atmega323, atmega64, atmega128, at43usb355, at94k).
9320 Output instruction sizes to the asm file.
9323 Specify the initial stack address, which may be a symbol or
9324 numeric value, `__stack' is the default.
9327 Generated code is not compatible with hardware interrupts. Code
9328 size will be smaller.
9331 Functions prologues/epilogues expanded as call to appropriate
9332 subroutines. Code size will be smaller.
9335 Do not generate tablejump insns which sometimes increase code size.
9336 The option is now deprecated in favor of the equivalent
9340 Change only the low 8 bits of the stack pointer.
9343 Assume int to be 8 bit integer. This affects the sizes of all
9344 types: A char will be 1 byte, an int will be 1 byte, an long will
9345 be 2 bytes and long long will be 4 bytes. Please note that this
9346 option does not comply to the C standards, but it will provide you
9347 with smaller code size.
9350 File: gcc.info, Node: Blackfin Options, Next: CRIS Options, Prev: AVR Options, Up: Submodel Options
9352 3.17.4 Blackfin Options
9353 -----------------------
9355 `-mcpu=CPU[-SIREVISION]'
9356 Specifies the name of the target Blackfin processor. Currently,
9357 CPU can be one of `bf512', `bf514', `bf516', `bf518', `bf522',
9358 `bf523', `bf524', `bf525', `bf526', `bf527', `bf531', `bf532',
9359 `bf533', `bf534', `bf536', `bf537', `bf538', `bf539', `bf542',
9360 `bf544', `bf547', `bf548', `bf549', `bf561'. The optional
9361 SIREVISION specifies the silicon revision of the target Blackfin
9362 processor. Any workarounds available for the targeted silicon
9363 revision will be enabled. If SIREVISION is `none', no workarounds
9364 are enabled. If SIREVISION is `any', all workarounds for the
9365 targeted processor will be enabled. The `__SILICON_REVISION__'
9366 macro is defined to two hexadecimal digits representing the major
9367 and minor numbers in the silicon revision. If SIREVISION is
9368 `none', the `__SILICON_REVISION__' is not defined. If SIREVISION
9369 is `any', the `__SILICON_REVISION__' is defined to be `0xffff'.
9370 If this optional SIREVISION is not used, GCC assumes the latest
9371 known silicon revision of the targeted Blackfin processor.
9373 Support for `bf561' is incomplete. For `bf561', Only the
9374 processor macro is defined. Without this option, `bf532' is used
9375 as the processor by default. The corresponding predefined
9376 processor macros for CPU is to be defined. And for `bfin-elf'
9377 toolchain, this causes the hardware BSP provided by libgloss to be
9378 linked in if `-msim' is not given.
9381 Specifies that the program will be run on the simulator. This
9382 causes the simulator BSP provided by libgloss to be linked in.
9383 This option has effect only for `bfin-elf' toolchain. Certain
9384 other options, such as `-mid-shared-library' and `-mfdpic', imply
9387 `-momit-leaf-frame-pointer'
9388 Don't keep the frame pointer in a register for leaf functions.
9389 This avoids the instructions to save, set up and restore frame
9390 pointers and makes an extra register available in leaf functions.
9391 The option `-fomit-frame-pointer' removes the frame pointer for
9392 all functions which might make debugging harder.
9395 When enabled, the compiler will ensure that the generated code
9396 does not contain speculative loads after jump instructions. If
9397 this option is used, `__WORKAROUND_SPECULATIVE_LOADS' is defined.
9399 `-mno-specld-anomaly'
9400 Don't generate extra code to prevent speculative loads from
9404 When enabled, the compiler will ensure that the generated code
9405 does not contain CSYNC or SSYNC instructions too soon after
9406 conditional branches. If this option is used,
9407 `__WORKAROUND_SPECULATIVE_SYNCS' is defined.
9409 `-mno-csync-anomaly'
9410 Don't generate extra code to prevent CSYNC or SSYNC instructions
9411 from occurring too soon after a conditional branch.
9414 When enabled, the compiler is free to take advantage of the
9415 knowledge that the entire program fits into the low 64k of memory.
9418 Assume that the program is arbitrarily large. This is the default.
9421 Do stack checking using information placed into L1 scratchpad
9422 memory by the uClinux kernel.
9424 `-mid-shared-library'
9425 Generate code that supports shared libraries via the library ID
9426 method. This allows for execute in place and shared libraries in
9427 an environment without virtual memory management. This option
9428 implies `-fPIC'. With a `bfin-elf' target, this option implies
9431 `-mno-id-shared-library'
9432 Generate code that doesn't assume ID based shared libraries are
9433 being used. This is the default.
9435 `-mleaf-id-shared-library'
9436 Generate code that supports shared libraries via the library ID
9437 method, but assumes that this library or executable won't link
9438 against any other ID shared libraries. That allows the compiler
9439 to use faster code for jumps and calls.
9441 `-mno-leaf-id-shared-library'
9442 Do not assume that the code being compiled won't link against any
9443 ID shared libraries. Slower code will be generated for jump and
9446 `-mshared-library-id=n'
9447 Specified the identification number of the ID based shared library
9448 being compiled. Specifying a value of 0 will generate more
9449 compact code, specifying other values will force the allocation of
9450 that number to the current library but is no more space or time
9451 efficient than omitting this option.
9454 Generate code that allows the data segment to be located in a
9455 different area of memory from the text segment. This allows for
9456 execute in place in an environment without virtual memory
9457 management by eliminating relocations against the text section.
9460 Generate code that assumes that the data segment follows the text
9461 segment. This is the default.
9465 Tells the compiler to perform function calls by first loading the
9466 address of the function into a register and then performing a
9467 subroutine call on this register. This switch is needed if the
9468 target function will lie outside of the 24 bit addressing range of
9469 the offset based version of subroutine call instruction.
9471 This feature is not enabled by default. Specifying
9472 `-mno-long-calls' will restore the default behavior. Note these
9473 switches have no effect on how the compiler generates code to
9474 handle function calls via function pointers.
9477 Link with the fast floating-point library. This library relaxes
9478 some of the IEEE floating-point standard's rules for checking
9479 inputs against Not-a-Number (NAN), in the interest of performance.
9482 Enable inlining of PLT entries in function calls to functions that
9483 are not known to bind locally. It has no effect without `-mfdpic'.
9486 Build standalone application for multicore Blackfin processor.
9487 Proper start files and link scripts will be used to support
9488 multicore. This option defines `__BFIN_MULTICORE'. It can only be
9489 used with `-mcpu=bf561[-SIREVISION]'. It can be used with
9490 `-mcorea' or `-mcoreb'. If it's used without `-mcorea' or
9491 `-mcoreb', single application/dual core programming model is used.
9492 In this model, the main function of Core B should be named as
9493 coreb_main. If it's used with `-mcorea' or `-mcoreb', one
9494 application per core programming model is used. If this option is
9495 not used, single core application programming model is used.
9498 Build standalone application for Core A of BF561 when using one
9499 application per core programming model. Proper start files and
9500 link scripts will be used to support Core A. This option defines
9501 `__BFIN_COREA'. It must be used with `-mmulticore'.
9504 Build standalone application for Core B of BF561 when using one
9505 application per core programming model. Proper start files and
9506 link scripts will be used to support Core B. This option defines
9507 `__BFIN_COREB'. When this option is used, coreb_main should be
9508 used instead of main. It must be used with `-mmulticore'.
9511 Build standalone application for SDRAM. Proper start files and
9512 link scripts will be used to put the application into SDRAM.
9513 Loader should initialize SDRAM before loading the application into
9514 SDRAM. This option defines `__BFIN_SDRAM'.
9517 Assume that ICPLBs are enabled at runtime. This has an effect on
9518 certain anomaly workarounds. For Linux targets, the default is to
9519 assume ICPLBs are enabled; for standalone applications the default
9523 File: gcc.info, Node: CRIS Options, Next: CRX Options, Prev: Blackfin Options, Up: Submodel Options
9528 These options are defined specifically for the CRIS ports.
9530 `-march=ARCHITECTURE-TYPE'
9531 `-mcpu=ARCHITECTURE-TYPE'
9532 Generate code for the specified architecture. The choices for
9533 ARCHITECTURE-TYPE are `v3', `v8' and `v10' for respectively
9534 ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is `v0' except for
9535 cris-axis-linux-gnu, where the default is `v10'.
9537 `-mtune=ARCHITECTURE-TYPE'
9538 Tune to ARCHITECTURE-TYPE everything applicable about the generated
9539 code, except for the ABI and the set of available instructions.
9540 The choices for ARCHITECTURE-TYPE are the same as for
9541 `-march=ARCHITECTURE-TYPE'.
9543 `-mmax-stack-frame=N'
9544 Warn when the stack frame of a function exceeds N bytes.
9548 The options `-metrax4' and `-metrax100' are synonyms for
9549 `-march=v3' and `-march=v8' respectively.
9551 `-mmul-bug-workaround'
9552 `-mno-mul-bug-workaround'
9553 Work around a bug in the `muls' and `mulu' instructions for CPU
9554 models where it applies. This option is active by default.
9557 Enable CRIS-specific verbose debug-related information in the
9558 assembly code. This option also has the effect to turn off the
9559 `#NO_APP' formatted-code indicator to the assembler at the
9560 beginning of the assembly file.
9563 Do not use condition-code results from previous instruction;
9564 always emit compare and test instructions before use of condition
9568 Do not emit instructions with side-effects in addressing modes
9569 other than post-increment.
9577 These options (no-options) arranges (eliminate arrangements) for
9578 the stack-frame, individual data and constants to be aligned for
9579 the maximum single data access size for the chosen CPU model. The
9580 default is to arrange for 32-bit alignment. ABI details such as
9581 structure layout are not affected by these options.
9586 Similar to the stack- data- and const-align options above, these
9587 options arrange for stack-frame, writable data and constants to
9588 all be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit
9591 `-mno-prologue-epilogue'
9592 `-mprologue-epilogue'
9593 With `-mno-prologue-epilogue', the normal function prologue and
9594 epilogue that sets up the stack-frame are omitted and no return
9595 instructions or return sequences are generated in the code. Use
9596 this option only together with visual inspection of the compiled
9597 code: no warnings or errors are generated when call-saved
9598 registers must be saved, or storage for local variable needs to be
9603 With `-fpic' and `-fPIC', don't generate (do generate) instruction
9604 sequences that load addresses for functions from the PLT part of
9605 the GOT rather than (traditional on other architectures) calls to
9606 the PLT. The default is `-mgotplt'.
9609 Legacy no-op option only recognized with the cris-axis-elf and
9610 cris-axis-linux-gnu targets.
9613 Legacy no-op option only recognized with the cris-axis-linux-gnu
9617 This option, recognized for the cris-axis-elf arranges to link
9618 with input-output functions from a simulator library. Code,
9619 initialized data and zero-initialized data are allocated
9623 Like `-sim', but pass linker options to locate initialized data at
9624 0x40000000 and zero-initialized data at 0x80000000.
9627 File: gcc.info, Node: CRX Options, Next: Darwin Options, Prev: CRIS Options, Up: Submodel Options
9632 These options are defined specifically for the CRX ports.
9635 Enable the use of multiply-accumulate instructions. Disabled by
9639 Push instructions will be used to pass outgoing arguments when
9640 functions are called. Enabled by default.
9643 File: gcc.info, Node: Darwin Options, Next: DEC Alpha Options, Prev: CRX Options, Up: Submodel Options
9645 3.17.7 Darwin Options
9646 ---------------------
9648 These options are defined for all architectures running the Darwin
9651 FSF GCC on Darwin does not create "fat" object files; it will create
9652 an object file for the single architecture that it was built to target.
9653 Apple's GCC on Darwin does create "fat" files if multiple `-arch'
9654 options are used; it does so by running the compiler or linker multiple
9655 times and joining the results together with `lipo'.
9657 The subtype of the file created (like `ppc7400' or `ppc970' or `i686')
9658 is determined by the flags that specify the ISA that GCC is targetting,
9659 like `-mcpu' or `-march'. The `-force_cpusubtype_ALL' option can be
9660 used to override this.
9662 The Darwin tools vary in their behavior when presented with an ISA
9663 mismatch. The assembler, `as', will only permit instructions to be
9664 used that are valid for the subtype of the file it is generating, so
9665 you cannot put 64-bit instructions in an `ppc750' object file. The
9666 linker for shared libraries, `/usr/bin/libtool', will fail and print an
9667 error if asked to create a shared library with a less restrictive
9668 subtype than its input files (for instance, trying to put a `ppc970'
9669 object file in a `ppc7400' library). The linker for executables, `ld',
9670 will quietly give the executable the most restrictive subtype of any of
9674 Add the framework directory DIR to the head of the list of
9675 directories to be searched for header files. These directories are
9676 interleaved with those specified by `-I' options and are scanned
9677 in a left-to-right order.
9679 A framework directory is a directory with frameworks in it. A
9680 framework is a directory with a `"Headers"' and/or
9681 `"PrivateHeaders"' directory contained directly in it that ends in
9682 `".framework"'. The name of a framework is the name of this
9683 directory excluding the `".framework"'. Headers associated with
9684 the framework are found in one of those two directories, with
9685 `"Headers"' being searched first. A subframework is a framework
9686 directory that is in a framework's `"Frameworks"' directory.
9687 Includes of subframework headers can only appear in a header of a
9688 framework that contains the subframework, or in a sibling
9689 subframework header. Two subframeworks are siblings if they occur
9690 in the same framework. A subframework should not have the same
9691 name as a framework, a warning will be issued if this is violated.
9692 Currently a subframework cannot have subframeworks, in the
9693 future, the mechanism may be extended to support this. The
9694 standard frameworks can be found in `"/System/Library/Frameworks"'
9695 and `"/Library/Frameworks"'. An example include looks like
9696 `#include <Framework/header.h>', where `Framework' denotes the
9697 name of the framework and header.h is found in the
9698 `"PrivateHeaders"' or `"Headers"' directory.
9701 Like `-F' except the directory is a treated as a system directory.
9702 The main difference between this `-iframework' and `-F' is that
9703 with `-iframework' the compiler does not warn about constructs
9704 contained within header files found via DIR. This option is valid
9705 only for the C family of languages.
9708 Emit debugging information for symbols that are used. For STABS
9709 debugging format, this enables `-feliminate-unused-debug-symbols'.
9710 This is by default ON.
9713 Emit debugging information for all symbols and types.
9715 `-mmacosx-version-min=VERSION'
9716 The earliest version of MacOS X that this executable will run on
9717 is VERSION. Typical values of VERSION include `10.1', `10.2', and
9720 If the compiler was built to use the system's headers by default,
9721 then the default for this option is the system version on which the
9722 compiler is running, otherwise the default is to make choices which
9723 are compatible with as many systems and code bases as possible.
9726 Enable kernel development mode. The `-mkernel' option sets
9727 `-static', `-fno-common', `-fno-cxa-atexit', `-fno-exceptions',
9728 `-fno-non-call-exceptions', `-fapple-kext', `-fno-weak' and
9729 `-fno-rtti' where applicable. This mode also sets `-mno-altivec',
9730 `-msoft-float', `-fno-builtin' and `-mlong-branch' for PowerPC
9734 Override the defaults for `bool' so that `sizeof(bool)==1'. By
9735 default `sizeof(bool)' is `4' when compiling for Darwin/PowerPC
9736 and `1' when compiling for Darwin/x86, so this option has no
9739 *Warning:* The `-mone-byte-bool' switch causes GCC to generate
9740 code that is not binary compatible with code generated without
9741 that switch. Using this switch may require recompiling all other
9742 modules in a program, including system libraries. Use this switch
9743 to conform to a non-default data model.
9745 `-mfix-and-continue'
9746 `-ffix-and-continue'
9748 Generate code suitable for fast turn around development. Needed to
9749 enable gdb to dynamically load `.o' files into already running
9750 programs. `-findirect-data' and `-ffix-and-continue' are provided
9751 for backwards compatibility.
9754 Loads all members of static archive libraries. See man ld(1) for
9757 `-arch_errors_fatal'
9758 Cause the errors having to do with files that have the wrong
9759 architecture to be fatal.
9762 Causes the output file to be marked such that the dynamic linker
9763 will bind all undefined references when the file is loaded or
9767 Produce a Mach-o bundle format file. See man ld(1) for more
9770 `-bundle_loader EXECUTABLE'
9771 This option specifies the EXECUTABLE that will be loading the build
9772 output file being linked. See man ld(1) for more information.
9775 When passed this option, GCC will produce a dynamic library
9776 instead of an executable when linking, using the Darwin `libtool'
9779 `-force_cpusubtype_ALL'
9780 This causes GCC's output file to have the ALL subtype, instead of
9781 one controlled by the `-mcpu' or `-march' option.
9783 `-allowable_client CLIENT_NAME'
9785 `-compatibility_version'
9790 `-dylinker_install_name'
9792 `-exported_symbols_list'
9795 `-force_flat_namespace'
9796 `-headerpad_max_install_names'
9800 `-keep_private_externs'
9803 `-multiply_defined_unused'
9805 `-no_dead_strip_inits_and_terms'
9812 `-prebind_all_twolevel_modules'
9816 `-sectobjectsymbols'
9820 `-sectobjectsymbols'
9823 `-segs_read_only_addr'
9824 `-segs_read_write_addr'
9826 `-seg_addr_table_filename'
9829 `-segs_read_only_addr'
9830 `-segs_read_write_addr'
9835 `-twolevel_namespace'
9838 `-unexported_symbols_list'
9839 `-weak_reference_mismatches'
9841 These options are passed to the Darwin linker. The Darwin linker
9842 man page describes them in detail.
9845 File: gcc.info, Node: DEC Alpha Options, Next: DEC Alpha/VMS Options, Prev: Darwin Options, Up: Submodel Options
9847 3.17.8 DEC Alpha Options
9848 ------------------------
9850 These `-m' options are defined for the DEC Alpha implementations:
9854 Use (do not use) the hardware floating-point instructions for
9855 floating-point operations. When `-msoft-float' is specified,
9856 functions in `libgcc.a' will be used to perform floating-point
9857 operations. Unless they are replaced by routines that emulate the
9858 floating-point operations, or compiled in such a way as to call
9859 such emulations routines, these routines will issue floating-point
9860 operations. If you are compiling for an Alpha without
9861 floating-point operations, you must ensure that the library is
9862 built so as not to call them.
9864 Note that Alpha implementations without floating-point operations
9865 are required to have floating-point registers.
9869 Generate code that uses (does not use) the floating-point register
9870 set. `-mno-fp-regs' implies `-msoft-float'. If the floating-point
9871 register set is not used, floating point operands are passed in
9872 integer registers as if they were integers and floating-point
9873 results are passed in `$0' instead of `$f0'. This is a
9874 non-standard calling sequence, so any function with a
9875 floating-point argument or return value called by code compiled
9876 with `-mno-fp-regs' must also be compiled with that option.
9878 A typical use of this option is building a kernel that does not
9879 use, and hence need not save and restore, any floating-point
9883 The Alpha architecture implements floating-point hardware
9884 optimized for maximum performance. It is mostly compliant with
9885 the IEEE floating point standard. However, for full compliance,
9886 software assistance is required. This option generates code fully
9887 IEEE compliant code _except_ that the INEXACT-FLAG is not
9888 maintained (see below). If this option is turned on, the
9889 preprocessor macro `_IEEE_FP' is defined during compilation. The
9890 resulting code is less efficient but is able to correctly support
9891 denormalized numbers and exceptional IEEE values such as
9892 not-a-number and plus/minus infinity. Other Alpha compilers call
9893 this option `-ieee_with_no_inexact'.
9895 `-mieee-with-inexact'
9896 This is like `-mieee' except the generated code also maintains the
9897 IEEE INEXACT-FLAG. Turning on this option causes the generated
9898 code to implement fully-compliant IEEE math. In addition to
9899 `_IEEE_FP', `_IEEE_FP_EXACT' is defined as a preprocessor macro.
9900 On some Alpha implementations the resulting code may execute
9901 significantly slower than the code generated by default. Since
9902 there is very little code that depends on the INEXACT-FLAG, you
9903 should normally not specify this option. Other Alpha compilers
9904 call this option `-ieee_with_inexact'.
9906 `-mfp-trap-mode=TRAP-MODE'
9907 This option controls what floating-point related traps are enabled.
9908 Other Alpha compilers call this option `-fptm TRAP-MODE'. The
9909 trap mode can be set to one of four values:
9912 This is the default (normal) setting. The only traps that
9913 are enabled are the ones that cannot be disabled in software
9914 (e.g., division by zero trap).
9917 In addition to the traps enabled by `n', underflow traps are
9921 Like `u', but the instructions are marked to be safe for
9922 software completion (see Alpha architecture manual for
9926 Like `su', but inexact traps are enabled as well.
9928 `-mfp-rounding-mode=ROUNDING-MODE'
9929 Selects the IEEE rounding mode. Other Alpha compilers call this
9930 option `-fprm ROUNDING-MODE'. The ROUNDING-MODE can be one of:
9933 Normal IEEE rounding mode. Floating point numbers are
9934 rounded towards the nearest machine number or towards the
9935 even machine number in case of a tie.
9938 Round towards minus infinity.
9941 Chopped rounding mode. Floating point numbers are rounded
9945 Dynamic rounding mode. A field in the floating point control
9946 register (FPCR, see Alpha architecture reference manual)
9947 controls the rounding mode in effect. The C library
9948 initializes this register for rounding towards plus infinity.
9949 Thus, unless your program modifies the FPCR, `d' corresponds
9950 to round towards plus infinity.
9952 `-mtrap-precision=TRAP-PRECISION'
9953 In the Alpha architecture, floating point traps are imprecise.
9954 This means without software assistance it is impossible to recover
9955 from a floating trap and program execution normally needs to be
9956 terminated. GCC can generate code that can assist operating
9957 system trap handlers in determining the exact location that caused
9958 a floating point trap. Depending on the requirements of an
9959 application, different levels of precisions can be selected:
9962 Program precision. This option is the default and means a
9963 trap handler can only identify which program caused a
9964 floating point exception.
9967 Function precision. The trap handler can determine the
9968 function that caused a floating point exception.
9971 Instruction precision. The trap handler can determine the
9972 exact instruction that caused a floating point exception.
9974 Other Alpha compilers provide the equivalent options called
9975 `-scope_safe' and `-resumption_safe'.
9978 This option marks the generated code as IEEE conformant. You must
9979 not use this option unless you also specify `-mtrap-precision=i'
9980 and either `-mfp-trap-mode=su' or `-mfp-trap-mode=sui'. Its only
9981 effect is to emit the line `.eflag 48' in the function prologue of
9982 the generated assembly file. Under DEC Unix, this has the effect
9983 that IEEE-conformant math library routines will be linked in.
9986 Normally GCC examines a 32- or 64-bit integer constant to see if
9987 it can construct it from smaller constants in two or three
9988 instructions. If it cannot, it will output the constant as a
9989 literal and generate code to load it from the data segment at
9992 Use this option to require GCC to construct _all_ integer constants
9993 using code, even if it takes more instructions (the maximum is
9996 You would typically use this option to build a shared library
9997 dynamic loader. Itself a shared library, it must relocate itself
9998 in memory before it can find the variables and constants in its
10003 Select whether to generate code to be assembled by the
10004 vendor-supplied assembler (`-malpha-as') or by the GNU assembler
10015 Indicate whether GCC should generate code to use the optional BWX,
10016 CIX, FIX and MAX instruction sets. The default is to use the
10017 instruction sets supported by the CPU type specified via `-mcpu='
10018 option or that of the CPU on which GCC was built if none was
10023 Generate code that uses (does not use) VAX F and G floating point
10024 arithmetic instead of IEEE single and double precision.
10026 `-mexplicit-relocs'
10027 `-mno-explicit-relocs'
10028 Older Alpha assemblers provided no way to generate symbol
10029 relocations except via assembler macros. Use of these macros does
10030 not allow optimal instruction scheduling. GNU binutils as of
10031 version 2.12 supports a new syntax that allows the compiler to
10032 explicitly mark which relocations should apply to which
10033 instructions. This option is mostly useful for debugging, as GCC
10034 detects the capabilities of the assembler when it is built and
10035 sets the default accordingly.
10039 When `-mexplicit-relocs' is in effect, static data is accessed via
10040 "gp-relative" relocations. When `-msmall-data' is used, objects 8
10041 bytes long or smaller are placed in a "small data area" (the
10042 `.sdata' and `.sbss' sections) and are accessed via 16-bit
10043 relocations off of the `$gp' register. This limits the size of
10044 the small data area to 64KB, but allows the variables to be
10045 directly accessed via a single instruction.
10047 The default is `-mlarge-data'. With this option the data area is
10048 limited to just below 2GB. Programs that require more than 2GB of
10049 data must use `malloc' or `mmap' to allocate the data in the heap
10050 instead of in the program's data segment.
10052 When generating code for shared libraries, `-fpic' implies
10053 `-msmall-data' and `-fPIC' implies `-mlarge-data'.
10057 When `-msmall-text' is used, the compiler assumes that the code of
10058 the entire program (or shared library) fits in 4MB, and is thus
10059 reachable with a branch instruction. When `-msmall-data' is used,
10060 the compiler can assume that all local symbols share the same
10061 `$gp' value, and thus reduce the number of instructions required
10062 for a function call from 4 to 1.
10064 The default is `-mlarge-text'.
10067 Set the instruction set and instruction scheduling parameters for
10068 machine type CPU_TYPE. You can specify either the `EV' style name
10069 or the corresponding chip number. GCC supports scheduling
10070 parameters for the EV4, EV5 and EV6 family of processors and will
10071 choose the default values for the instruction set from the
10072 processor you specify. If you do not specify a processor type,
10073 GCC will default to the processor on which the compiler was built.
10075 Supported values for CPU_TYPE are
10080 Schedules as an EV4 and has no instruction set extensions.
10084 Schedules as an EV5 and has no instruction set extensions.
10088 Schedules as an EV5 and supports the BWX extension.
10093 Schedules as an EV5 and supports the BWX and MAX extensions.
10097 Schedules as an EV6 and supports the BWX, FIX, and MAX
10102 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX
10105 Native Linux/GNU toolchains also support the value `native', which
10106 selects the best architecture option for the host processor.
10107 `-mcpu=native' has no effect if GCC does not recognize the
10111 Set only the instruction scheduling parameters for machine type
10112 CPU_TYPE. The instruction set is not changed.
10114 Native Linux/GNU toolchains also support the value `native', which
10115 selects the best architecture option for the host processor.
10116 `-mtune=native' has no effect if GCC does not recognize the
10119 `-mmemory-latency=TIME'
10120 Sets the latency the scheduler should assume for typical memory
10121 references as seen by the application. This number is highly
10122 dependent on the memory access patterns used by the application
10123 and the size of the external cache on the machine.
10125 Valid options for TIME are
10128 A decimal number representing clock cycles.
10134 The compiler contains estimates of the number of clock cycles
10135 for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10136 (also called Dcache, Scache, and Bcache), as well as to main
10137 memory. Note that L3 is only valid for EV5.
10141 File: gcc.info, Node: DEC Alpha/VMS Options, Next: FR30 Options, Prev: DEC Alpha Options, Up: Submodel Options
10143 3.17.9 DEC Alpha/VMS Options
10144 ----------------------------
10146 These `-m' options are defined for the DEC Alpha/VMS implementations:
10148 `-mvms-return-codes'
10149 Return VMS condition codes from main. The default is to return
10150 POSIX style condition (e.g. error) codes.
10153 File: gcc.info, Node: FR30 Options, Next: FRV Options, Prev: DEC Alpha/VMS Options, Up: Submodel Options
10155 3.17.10 FR30 Options
10156 --------------------
10158 These options are defined specifically for the FR30 port.
10161 Use the small address space model. This can produce smaller code,
10162 but it does assume that all symbolic values and addresses will fit
10163 into a 20-bit range.
10166 Assume that run-time support has been provided and so there is no
10167 need to include the simulator library (`libsim.a') on the linker
10172 File: gcc.info, Node: FRV Options, Next: GNU/Linux Options, Prev: FR30 Options, Up: Submodel Options
10174 3.17.11 FRV Options
10175 -------------------
10178 Only use the first 32 general purpose registers.
10181 Use all 64 general purpose registers.
10184 Use only the first 32 floating point registers.
10187 Use all 64 floating point registers
10190 Use hardware instructions for floating point operations.
10193 Use library routines for floating point operations.
10196 Dynamically allocate condition code registers.
10199 Do not try to dynamically allocate condition code registers, only
10200 use `icc0' and `fcc0'.
10203 Change ABI to use double word insns.
10206 Do not use double word instructions.
10209 Use floating point double instructions.
10212 Do not use floating point double instructions.
10215 Use media instructions.
10218 Do not use media instructions.
10221 Use multiply and add/subtract instructions.
10224 Do not use multiply and add/subtract instructions.
10227 Select the FDPIC ABI, that uses function descriptors to represent
10228 pointers to functions. Without any PIC/PIE-related options, it
10229 implies `-fPIE'. With `-fpic' or `-fpie', it assumes GOT entries
10230 and small data are within a 12-bit range from the GOT base
10231 address; with `-fPIC' or `-fPIE', GOT offsets are computed with 32
10232 bits. With a `bfin-elf' target, this option implies `-msim'.
10235 Enable inlining of PLT entries in function calls to functions that
10236 are not known to bind locally. It has no effect without `-mfdpic'.
10237 It's enabled by default if optimizing for speed and compiling for
10238 shared libraries (i.e., `-fPIC' or `-fpic'), or when an
10239 optimization option such as `-O3' or above is present in the
10243 Assume a large TLS segment when generating thread-local code.
10246 Do not assume a large TLS segment when generating thread-local
10250 Enable the use of `GPREL' relocations in the FDPIC ABI for data
10251 that is known to be in read-only sections. It's enabled by
10252 default, except for `-fpic' or `-fpie': even though it may help
10253 make the global offset table smaller, it trades 1 instruction for
10254 4. With `-fPIC' or `-fPIE', it trades 3 instructions for 4, one
10255 of which may be shared by multiple symbols, and it avoids the need
10256 for a GOT entry for the referenced symbol, so it's more likely to
10257 be a win. If it is not, `-mno-gprel-ro' can be used to disable it.
10259 `-multilib-library-pic'
10260 Link with the (library, not FD) pic libraries. It's implied by
10261 `-mlibrary-pic', as well as by `-fPIC' and `-fpic' without
10262 `-mfdpic'. You should never have to use it explicitly.
10265 Follow the EABI requirement of always creating a frame pointer
10266 whenever a stack frame is allocated. This option is enabled by
10267 default and can be disabled with `-mno-linked-fp'.
10270 Use indirect addressing to call functions outside the current
10271 compilation unit. This allows the functions to be placed anywhere
10272 within the 32-bit address space.
10275 Try to align labels to an 8-byte boundary by inserting nops into
10276 the previous packet. This option only has an effect when VLIW
10277 packing is enabled. It doesn't create new packets; it merely adds
10278 nops to existing ones.
10281 Generate position-independent EABI code.
10284 Use only the first four media accumulator registers.
10287 Use all eight media accumulator registers.
10290 Pack VLIW instructions.
10293 Do not pack VLIW instructions.
10296 Do not mark ABI switches in e_flags.
10299 Enable the use of conditional-move instructions (default).
10301 This switch is mainly for debugging the compiler and will likely
10302 be removed in a future version.
10305 Disable the use of conditional-move instructions.
10307 This switch is mainly for debugging the compiler and will likely
10308 be removed in a future version.
10311 Enable the use of conditional set instructions (default).
10313 This switch is mainly for debugging the compiler and will likely
10314 be removed in a future version.
10317 Disable the use of conditional set instructions.
10319 This switch is mainly for debugging the compiler and will likely
10320 be removed in a future version.
10323 Enable the use of conditional execution (default).
10325 This switch is mainly for debugging the compiler and will likely
10326 be removed in a future version.
10329 Disable the use of conditional execution.
10331 This switch is mainly for debugging the compiler and will likely
10332 be removed in a future version.
10335 Run a pass to pack branches into VLIW instructions (default).
10337 This switch is mainly for debugging the compiler and will likely
10338 be removed in a future version.
10341 Do not run a pass to pack branches into VLIW instructions.
10343 This switch is mainly for debugging the compiler and will likely
10344 be removed in a future version.
10346 `-mmulti-cond-exec'
10347 Enable optimization of `&&' and `||' in conditional execution
10350 This switch is mainly for debugging the compiler and will likely
10351 be removed in a future version.
10353 `-mno-multi-cond-exec'
10354 Disable optimization of `&&' and `||' in conditional execution.
10356 This switch is mainly for debugging the compiler and will likely
10357 be removed in a future version.
10359 `-mnested-cond-exec'
10360 Enable nested conditional execution optimizations (default).
10362 This switch is mainly for debugging the compiler and will likely
10363 be removed in a future version.
10365 `-mno-nested-cond-exec'
10366 Disable nested conditional execution optimizations.
10368 This switch is mainly for debugging the compiler and will likely
10369 be removed in a future version.
10371 `-moptimize-membar'
10372 This switch removes redundant `membar' instructions from the
10373 compiler generated code. It is enabled by default.
10375 `-mno-optimize-membar'
10376 This switch disables the automatic removal of redundant `membar'
10377 instructions from the generated code.
10380 Cause gas to print out tomcat statistics.
10383 Select the processor type for which to generate code. Possible
10384 values are `frv', `fr550', `tomcat', `fr500', `fr450', `fr405',
10385 `fr400', `fr300' and `simple'.
10389 File: gcc.info, Node: GNU/Linux Options, Next: H8/300 Options, Prev: FRV Options, Up: Submodel Options
10391 3.17.12 GNU/Linux Options
10392 -------------------------
10394 These `-m' options are defined for GNU/Linux targets:
10397 Use the GNU C library instead of uClibc. This is the default
10398 except on `*-*-linux-*uclibc*' targets.
10401 Use uClibc instead of the GNU C library. This is the default on
10402 `*-*-linux-*uclibc*' targets.
10405 File: gcc.info, Node: H8/300 Options, Next: HPPA Options, Prev: GNU/Linux Options, Up: Submodel Options
10407 3.17.13 H8/300 Options
10408 ----------------------
10410 These `-m' options are defined for the H8/300 implementations:
10413 Shorten some address references at link time, when possible; uses
10414 the linker option `-relax'. *Note `ld' and the H8/300:
10415 (ld)H8/300, for a fuller description.
10418 Generate code for the H8/300H.
10421 Generate code for the H8S.
10424 Generate code for the H8S and H8/300H in the normal mode. This
10425 switch must be used either with `-mh' or `-ms'.
10428 Generate code for the H8S/2600. This switch must be used with
10432 Make `int' data 32 bits by default.
10435 On the H8/300H and H8S, use the same alignment rules as for the
10436 H8/300. The default for the H8/300H and H8S is to align longs and
10437 floats on 4 byte boundaries. `-malign-300' causes them to be
10438 aligned on 2 byte boundaries. This option has no effect on the
10442 File: gcc.info, Node: HPPA Options, Next: i386 and x86-64 Options, Prev: H8/300 Options, Up: Submodel Options
10444 3.17.14 HPPA Options
10445 --------------------
10447 These `-m' options are defined for the HPPA family of computers:
10449 `-march=ARCHITECTURE-TYPE'
10450 Generate code for the specified architecture. The choices for
10451 ARCHITECTURE-TYPE are `1.0' for PA 1.0, `1.1' for PA 1.1, and
10452 `2.0' for PA 2.0 processors. Refer to `/usr/lib/sched.models' on
10453 an HP-UX system to determine the proper architecture option for
10454 your machine. Code compiled for lower numbered architectures will
10455 run on higher numbered architectures, but not the other way around.
10460 Synonyms for `-march=1.0', `-march=1.1', and `-march=2.0'
10464 Generate code suitable for big switch tables. Use this option
10465 only if the assembler/linker complain about out of range branches
10466 within a switch table.
10469 Fill delay slots of function calls with unconditional jump
10470 instructions by modifying the return pointer for the function call
10471 to be the target of the conditional jump.
10474 Prevent floating point registers from being used in any manner.
10475 This is necessary for compiling kernels which perform lazy context
10476 switching of floating point registers. If you use this option and
10477 attempt to perform floating point operations, the compiler will
10480 `-mdisable-indexing'
10481 Prevent the compiler from using indexing address modes. This
10482 avoids some rather obscure problems when compiling MIG generated
10486 Generate code that assumes the target has no space registers.
10487 This allows GCC to generate faster indirect calls and use unscaled
10488 index address modes.
10490 Such code is suitable for level 0 PA systems and kernels.
10492 `-mfast-indirect-calls'
10493 Generate code that assumes calls never cross space boundaries.
10494 This allows GCC to emit code which performs faster indirect calls.
10496 This option will not work in the presence of shared libraries or
10499 `-mfixed-range=REGISTER-RANGE'
10500 Generate code treating the given register range as fixed registers.
10501 A fixed register is one that the register allocator can not use.
10502 This is useful when compiling kernel code. A register range is
10503 specified as two registers separated by a dash. Multiple register
10504 ranges can be specified separated by a comma.
10506 `-mlong-load-store'
10507 Generate 3-instruction load and store sequences as sometimes
10508 required by the HP-UX 10 linker. This is equivalent to the `+k'
10509 option to the HP compilers.
10511 `-mportable-runtime'
10512 Use the portable calling conventions proposed by HP for ELF
10516 Enable the use of assembler directives only GAS understands.
10518 `-mschedule=CPU-TYPE'
10519 Schedule code according to the constraints for the machine type
10520 CPU-TYPE. The choices for CPU-TYPE are `700' `7100', `7100LC',
10521 `7200', `7300' and `8000'. Refer to `/usr/lib/sched.models' on an
10522 HP-UX system to determine the proper scheduling option for your
10523 machine. The default scheduling is `8000'.
10526 Enable the optimization pass in the HP-UX linker. Note this makes
10527 symbolic debugging impossible. It also triggers a bug in the
10528 HP-UX 8 and HP-UX 9 linkers in which they give bogus error
10529 messages when linking some programs.
10532 Generate output containing library calls for floating point.
10533 *Warning:* the requisite libraries are not available for all HPPA
10534 targets. Normally the facilities of the machine's usual C
10535 compiler are used, but this cannot be done directly in
10536 cross-compilation. You must make your own arrangements to provide
10537 suitable library functions for cross-compilation.
10539 `-msoft-float' changes the calling convention in the output file;
10540 therefore, it is only useful if you compile _all_ of a program with
10541 this option. In particular, you need to compile `libgcc.a', the
10542 library that comes with GCC, with `-msoft-float' in order for this
10546 Generate the predefine, `_SIO', for server IO. The default is
10547 `-mwsio'. This generates the predefines, `__hp9000s700',
10548 `__hp9000s700__' and `_WSIO', for workstation IO. These options
10549 are available under HP-UX and HI-UX.
10552 Use GNU ld specific options. This passes `-shared' to ld when
10553 building a shared library. It is the default when GCC is
10554 configured, explicitly or implicitly, with the GNU linker. This
10555 option does not have any affect on which ld is called, it only
10556 changes what parameters are passed to that ld. The ld that is
10557 called is determined by the `--with-ld' configure option, GCC's
10558 program search path, and finally by the user's `PATH'. The linker
10559 used by GCC can be printed using `which `gcc
10560 -print-prog-name=ld`'. This option is only available on the 64
10561 bit HP-UX GCC, i.e. configured with `hppa*64*-*-hpux*'.
10564 Use HP ld specific options. This passes `-b' to ld when building
10565 a shared library and passes `+Accept TypeMismatch' to ld on all
10566 links. It is the default when GCC is configured, explicitly or
10567 implicitly, with the HP linker. This option does not have any
10568 affect on which ld is called, it only changes what parameters are
10569 passed to that ld. The ld that is called is determined by the
10570 `--with-ld' configure option, GCC's program search path, and
10571 finally by the user's `PATH'. The linker used by GCC can be
10572 printed using `which `gcc -print-prog-name=ld`'. This option is
10573 only available on the 64 bit HP-UX GCC, i.e. configured with
10574 `hppa*64*-*-hpux*'.
10577 Generate code that uses long call sequences. This ensures that a
10578 call is always able to reach linker generated stubs. The default
10579 is to generate long calls only when the distance from the call
10580 site to the beginning of the function or translation unit, as the
10581 case may be, exceeds a predefined limit set by the branch type
10582 being used. The limits for normal calls are 7,600,000 and 240,000
10583 bytes, respectively for the PA 2.0 and PA 1.X architectures.
10584 Sibcalls are always limited at 240,000 bytes.
10586 Distances are measured from the beginning of functions when using
10587 the `-ffunction-sections' option, or when using the `-mgas' and
10588 `-mno-portable-runtime' options together under HP-UX with the SOM
10591 It is normally not desirable to use this option as it will degrade
10592 performance. However, it may be useful in large applications,
10593 particularly when partial linking is used to build the application.
10595 The types of long calls used depends on the capabilities of the
10596 assembler and linker, and the type of code being generated. The
10597 impact on systems that support long absolute calls, and long pic
10598 symbol-difference or pc-relative calls should be relatively small.
10599 However, an indirect call is used on 32-bit ELF systems in pic code
10600 and it is quite long.
10603 Generate compiler predefines and select a startfile for the
10604 specified UNIX standard. The choices for UNIX-STD are `93', `95'
10605 and `98'. `93' is supported on all HP-UX versions. `95' is
10606 available on HP-UX 10.10 and later. `98' is available on HP-UX
10607 11.11 and later. The default values are `93' for HP-UX 10.00,
10608 `95' for HP-UX 10.10 though to 11.00, and `98' for HP-UX 11.11 and
10611 `-munix=93' provides the same predefines as GCC 3.3 and 3.4.
10612 `-munix=95' provides additional predefines for `XOPEN_UNIX' and
10613 `_XOPEN_SOURCE_EXTENDED', and the startfile `unix95.o'.
10614 `-munix=98' provides additional predefines for `_XOPEN_UNIX',
10615 `_XOPEN_SOURCE_EXTENDED', `_INCLUDE__STDC_A1_SOURCE' and
10616 `_INCLUDE_XOPEN_SOURCE_500', and the startfile `unix98.o'.
10618 It is _important_ to note that this option changes the interfaces
10619 for various library routines. It also affects the operational
10620 behavior of the C library. Thus, _extreme_ care is needed in
10623 Library code that is intended to operate with more than one UNIX
10624 standard must test, set and restore the variable
10625 __XPG4_EXTENDED_MASK as appropriate. Most GNU software doesn't
10626 provide this capability.
10629 Suppress the generation of link options to search libdld.sl when
10630 the `-static' option is specified on HP-UX 10 and later.
10633 The HP-UX implementation of setlocale in libc has a dependency on
10634 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10635 when the `-static' option is specified, special link options are
10636 needed to resolve this dependency.
10638 On HP-UX 10 and later, the GCC driver adds the necessary options to
10639 link with libdld.sl when the `-static' option is specified. This
10640 causes the resulting binary to be dynamic. On the 64-bit port,
10641 the linkers generate dynamic binaries by default in any case. The
10642 `-nolibdld' option can be used to prevent the GCC driver from
10643 adding these link options.
10646 Add support for multithreading with the "dce thread" library under
10647 HP-UX. This option sets flags for both the preprocessor and
10651 File: gcc.info, Node: i386 and x86-64 Options, Next: i386 and x86-64 Windows Options, Prev: HPPA Options, Up: Submodel Options
10653 3.17.15 Intel 386 and AMD x86-64 Options
10654 ----------------------------------------
10656 These `-m' options are defined for the i386 and x86-64 family of
10660 Tune to CPU-TYPE everything applicable about the generated code,
10661 except for the ABI and the set of available instructions. The
10662 choices for CPU-TYPE are:
10664 Produce code optimized for the most common IA32/AMD64/EM64T
10665 processors. If you know the CPU on which your code will run,
10666 then you should use the corresponding `-mtune' option instead
10667 of `-mtune=generic'. But, if you do not know exactly what
10668 CPU users of your application will have, then you should use
10671 As new processors are deployed in the marketplace, the
10672 behavior of this option will change. Therefore, if you
10673 upgrade to a newer version of GCC, the code generated option
10674 will change to reflect the processors that were most common
10675 when that version of GCC was released.
10677 There is no `-march=generic' option because `-march'
10678 indicates the instruction set the compiler can use, and there
10679 is no generic instruction set applicable to all processors.
10680 In contrast, `-mtune' indicates the processor (or, in this
10681 case, collection of processors) for which the code is
10685 This selects the CPU to tune for at compilation time by
10686 determining the processor type of the compiling machine.
10687 Using `-mtune=native' will produce code optimized for the
10688 local machine under the constraints of the selected
10689 instruction set. Using `-march=native' will enable all
10690 instruction subsets supported by the local machine (hence the
10691 result might not run on different machines).
10694 Original Intel's i386 CPU.
10697 Intel's i486 CPU. (No scheduling is implemented for this
10701 Intel Pentium CPU with no MMX support.
10704 Intel PentiumMMX CPU based on Pentium core with MMX
10705 instruction set support.
10708 Intel PentiumPro CPU.
10711 Same as `generic', but when used as `march' option, PentiumPro
10712 instruction set will be used, so the code will run on all
10716 Intel Pentium2 CPU based on PentiumPro core with MMX
10717 instruction set support.
10719 _pentium3, pentium3m_
10720 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE
10721 instruction set support.
10724 Low power version of Intel Pentium3 CPU with MMX, SSE and
10725 SSE2 instruction set support. Used by Centrino notebooks.
10727 _pentium4, pentium4m_
10728 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set
10732 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2
10733 and SSE3 instruction set support.
10736 Improved version of Intel Pentium4 CPU with 64-bit
10737 extensions, MMX, SSE, SSE2 and SSE3 instruction set support.
10740 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3
10741 and SSSE3 instruction set support.
10744 AMD K6 CPU with MMX instruction set support.
10747 Improved versions of AMD K6 CPU with MMX and 3dNOW!
10748 instruction set support.
10750 _athlon, athlon-tbird_
10751 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE
10752 prefetch instructions support.
10754 _athlon-4, athlon-xp, athlon-mp_
10755 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and
10756 full SSE instruction set support.
10758 _k8, opteron, athlon64, athlon-fx_
10759 AMD K8 core based CPUs with x86-64 instruction set support.
10760 (This supersets MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and
10761 64-bit instruction set extensions.)
10763 _k8-sse3, opteron-sse3, athlon64-sse3_
10764 Improved versions of k8, opteron and athlon64 with SSE3
10765 instruction set support.
10767 _amdfam10, barcelona_
10768 AMD Family 10h core based CPUs with x86-64 instruction set
10769 support. (This supersets MMX, SSE, SSE2, SSE3, SSE4A,
10770 3dNOW!, enhanced 3dNOW!, ABM and 64-bit instruction set
10774 IDT Winchip C6 CPU, dealt in same way as i486 with additional
10775 MMX instruction set support.
10778 IDT Winchip2 CPU, dealt in same way as i486 with additional
10779 MMX and 3dNOW! instruction set support.
10782 Via C3 CPU with MMX and 3dNOW! instruction set support. (No
10783 scheduling is implemented for this chip.)
10786 Via C3-2 CPU with MMX and SSE instruction set support. (No
10787 scheduling is implemented for this chip.)
10790 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10792 While picking a specific CPU-TYPE will schedule things
10793 appropriately for that particular chip, the compiler will not
10794 generate any code that does not run on the i386 without the
10795 `-march=CPU-TYPE' option being used.
10798 Generate instructions for the machine type CPU-TYPE. The choices
10799 for CPU-TYPE are the same as for `-mtune'. Moreover, specifying
10800 `-march=CPU-TYPE' implies `-mtune=CPU-TYPE'.
10803 A deprecated synonym for `-mtune'.
10806 Generate floating point arithmetics for selected unit UNIT. The
10807 choices for UNIT are:
10810 Use the standard 387 floating point coprocessor present
10811 majority of chips and emulated otherwise. Code compiled with
10812 this option will run almost everywhere. The temporary
10813 results are computed in 80bit precision instead of precision
10814 specified by the type resulting in slightly different results
10815 compared to most of other chips. See `-ffloat-store' for
10816 more detailed description.
10818 This is the default choice for i386 compiler.
10821 Use scalar floating point instructions present in the SSE
10822 instruction set. This instruction set is supported by
10823 Pentium3 and newer chips, in the AMD line by Athlon-4,
10824 Athlon-xp and Athlon-mp chips. The earlier version of SSE
10825 instruction set supports only single precision arithmetics,
10826 thus the double and extended precision arithmetics is still
10827 done using 387. Later version, present only in Pentium4 and
10828 the future AMD x86-64 chips supports double precision
10831 For the i386 compiler, you need to use `-march=CPU-TYPE',
10832 `-msse' or `-msse2' switches to enable SSE extensions and
10833 make this option effective. For the x86-64 compiler, these
10834 extensions are enabled by default.
10836 The resulting code should be considerably faster in the
10837 majority of cases and avoid the numerical instability
10838 problems of 387 code, but may break some existing code that
10839 expects temporaries to be 80bit.
10841 This is the default choice for the x86-64 compiler.
10846 Attempt to utilize both instruction sets at once. This
10847 effectively double the amount of available registers and on
10848 chips with separate execution units for 387 and SSE the
10849 execution resources too. Use this option with care, as it is
10850 still experimental, because the GCC register allocator does
10851 not model separate functional units well resulting in
10852 instable performance.
10855 Output asm instructions using selected DIALECT. Supported choices
10856 are `intel' or `att' (the default one). Darwin does not support
10861 Control whether or not the compiler uses IEEE floating point
10862 comparisons. These handle correctly the case where the result of a
10863 comparison is unordered.
10866 Generate output containing library calls for floating point.
10867 *Warning:* the requisite libraries are not part of GCC. Normally
10868 the facilities of the machine's usual C compiler are used, but
10869 this can't be done directly in cross-compilation. You must make
10870 your own arrangements to provide suitable library functions for
10873 On machines where a function returns floating point results in the
10874 80387 register stack, some floating point opcodes may be emitted
10875 even if `-msoft-float' is used.
10877 `-mno-fp-ret-in-387'
10878 Do not use the FPU registers for return values of functions.
10880 The usual calling convention has functions return values of types
10881 `float' and `double' in an FPU register, even if there is no FPU.
10882 The idea is that the operating system should emulate an FPU.
10884 The option `-mno-fp-ret-in-387' causes such values to be returned
10885 in ordinary CPU registers instead.
10887 `-mno-fancy-math-387'
10888 Some 387 emulators do not support the `sin', `cos' and `sqrt'
10889 instructions for the 387. Specify this option to avoid generating
10890 those instructions. This option is the default on FreeBSD,
10891 OpenBSD and NetBSD. This option is overridden when `-march'
10892 indicates that the target cpu will always have an FPU and so the
10893 instruction will not need emulation. As of revision 2.6.1, these
10894 instructions are not generated unless you also use the
10895 `-funsafe-math-optimizations' switch.
10898 `-mno-align-double'
10899 Control whether GCC aligns `double', `long double', and `long
10900 long' variables on a two word boundary or a one word boundary.
10901 Aligning `double' variables on a two word boundary will produce
10902 code that runs somewhat faster on a `Pentium' at the expense of
10905 On x86-64, `-malign-double' is enabled by default.
10907 *Warning:* if you use the `-malign-double' switch, structures
10908 containing the above types will be aligned differently than the
10909 published application binary interface specifications for the 386
10910 and will not be binary compatible with structures in code compiled
10911 without that switch.
10913 `-m96bit-long-double'
10914 `-m128bit-long-double'
10915 These switches control the size of `long double' type. The i386
10916 application binary interface specifies the size to be 96 bits, so
10917 `-m96bit-long-double' is the default in 32 bit mode.
10919 Modern architectures (Pentium and newer) would prefer `long double'
10920 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10921 conforming to the ABI, this would not be possible. So specifying a
10922 `-m128bit-long-double' will align `long double' to a 16 byte
10923 boundary by padding the `long double' with an additional 32 bit
10926 In the x86-64 compiler, `-m128bit-long-double' is the default
10927 choice as its ABI specifies that `long double' is to be aligned on
10930 Notice that neither of these options enable any extra precision
10931 over the x87 standard of 80 bits for a `long double'.
10933 *Warning:* if you override the default value for your target ABI,
10934 the structures and arrays containing `long double' variables will
10935 change their size as well as function calling convention for
10936 function taking `long double' will be modified. Hence they will
10937 not be binary compatible with arrays or structures in code
10938 compiled without that switch.
10940 `-mlarge-data-threshold=NUMBER'
10941 When `-mcmodel=medium' is specified, the data greater than
10942 THRESHOLD are placed in large data section. This value must be the
10943 same across all object linked into the binary and defaults to
10947 Use a different function-calling convention, in which functions
10948 that take a fixed number of arguments return with the `ret' NUM
10949 instruction, which pops their arguments while returning. This
10950 saves one instruction in the caller since there is no need to pop
10951 the arguments there.
10953 You can specify that an individual function is called with this
10954 calling sequence with the function attribute `stdcall'. You can
10955 also override the `-mrtd' option by using the function attribute
10956 `cdecl'. *Note Function Attributes::.
10958 *Warning:* this calling convention is incompatible with the one
10959 normally used on Unix, so you cannot use it if you need to call
10960 libraries compiled with the Unix compiler.
10962 Also, you must provide function prototypes for all functions that
10963 take variable numbers of arguments (including `printf'); otherwise
10964 incorrect code will be generated for calls to those functions.
10966 In addition, seriously incorrect code will result if you call a
10967 function with too many arguments. (Normally, extra arguments are
10968 harmlessly ignored.)
10971 Control how many registers are used to pass integer arguments. By
10972 default, no registers are used to pass arguments, and at most 3
10973 registers can be used. You can control this behavior for a
10974 specific function by using the function attribute `regparm'.
10975 *Note Function Attributes::.
10977 *Warning:* if you use this switch, and NUM is nonzero, then you
10978 must build all modules with the same value, including any
10979 libraries. This includes the system libraries and startup modules.
10982 Use SSE register passing conventions for float and double arguments
10983 and return values. You can control this behavior for a specific
10984 function by using the function attribute `sseregparm'. *Note
10985 Function Attributes::.
10987 *Warning:* if you use this switch then you must build all modules
10988 with the same value, including any libraries. This includes the
10989 system libraries and startup modules.
10994 Set 80387 floating-point precision to 32, 64 or 80 bits. When
10995 `-mpc32' is specified, the significands of results of
10996 floating-point operations are rounded to 24 bits (single
10997 precision); `-mpc64' rounds the significands of results of
10998 floating-point operations to 53 bits (double precision) and
10999 `-mpc80' rounds the significands of results of floating-point
11000 operations to 64 bits (extended double precision), which is the
11001 default. When this option is used, floating-point operations in
11002 higher precisions are not available to the programmer without
11003 setting the FPU control word explicitly.
11005 Setting the rounding of floating-point operations to less than the
11006 default 80 bits can speed some programs by 2% or more. Note that
11007 some mathematical libraries assume that extended precision (80
11008 bit) floating-point operations are enabled by default; routines in
11009 such libraries could suffer significant loss of accuracy,
11010 typically through so-called "catastrophic cancellation", when this
11011 option is used to set the precision to less than extended
11015 Realign the stack at entry. On the Intel x86, the `-mstackrealign'
11016 option will generate an alternate prologue and epilogue that
11017 realigns the runtime stack if necessary. This supports mixing
11018 legacy codes that keep a 4-byte aligned stack with modern codes
11019 that keep a 16-byte stack for SSE compatibility. See also the
11020 attribute `force_align_arg_pointer', applicable to individual
11023 `-mpreferred-stack-boundary=NUM'
11024 Attempt to keep the stack boundary aligned to a 2 raised to NUM
11025 byte boundary. If `-mpreferred-stack-boundary' is not specified,
11026 the default is 4 (16 bytes or 128 bits).
11028 `-mincoming-stack-boundary=NUM'
11029 Assume the incoming stack is aligned to a 2 raised to NUM byte
11030 boundary. If `-mincoming-stack-boundary' is not specified, the
11031 one specified by `-mpreferred-stack-boundary' will be used.
11033 On Pentium and PentiumPro, `double' and `long double' values
11034 should be aligned to an 8 byte boundary (see `-malign-double') or
11035 suffer significant run time performance penalties. On Pentium
11036 III, the Streaming SIMD Extension (SSE) data type `__m128' may not
11037 work properly if it is not 16 byte aligned.
11039 To ensure proper alignment of this values on the stack, the stack
11040 boundary must be as aligned as that required by any value stored
11041 on the stack. Further, every function must be generated such that
11042 it keeps the stack aligned. Thus calling a function compiled with
11043 a higher preferred stack boundary from a function compiled with a
11044 lower preferred stack boundary will most likely misalign the
11045 stack. It is recommended that libraries that use callbacks always
11046 use the default setting.
11048 This extra alignment does consume extra stack space, and generally
11049 increases code size. Code that is sensitive to stack space usage,
11050 such as embedded systems and operating system kernels, may want to
11051 reduce the preferred alignment to `-mpreferred-stack-boundary=2'.
11085 These switches enable or disable the use of instructions in the
11086 MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A,
11087 SSE5, ABM or 3DNow! extended instruction sets. These extensions
11088 are also available as built-in functions: see *Note X86 Built-in
11089 Functions::, for details of the functions enabled and disabled by
11092 To have SSE/SSE2 instructions generated automatically from
11093 floating-point code (as opposed to 387 instructions), see
11096 GCC depresses SSEx instructions when `-mavx' is used. Instead, it
11097 generates new AVX instructions or AVX equivalence for all SSEx
11098 instructions when needed.
11100 These options will enable GCC to use these extended instructions in
11101 generated code, even without `-mfpmath=sse'. Applications which
11102 perform runtime CPU detection must compile separate files for each
11103 supported architecture, using the appropriate flags. In
11104 particular, the file containing the CPU detection code should be
11105 compiled without these options.
11108 This option instructs GCC to emit a `cld' instruction in the
11109 prologue of functions that use string instructions. String
11110 instructions depend on the DF flag to select between autoincrement
11111 or autodecrement mode. While the ABI specifies the DF flag to be
11112 cleared on function entry, some operating systems violate this
11113 specification by not clearing the DF flag in their exception
11114 dispatchers. The exception handler can be invoked with the DF flag
11115 set which leads to wrong direction mode, when string instructions
11116 are used. This option can be enabled by default on 32-bit x86
11117 targets by configuring GCC with the `--enable-cld' configure
11118 option. Generation of `cld' instructions can be suppressed with
11119 the `-mno-cld' compiler option in this case.
11122 This option will enable GCC to use CMPXCHG16B instruction in
11123 generated code. CMPXCHG16B allows for atomic operations on
11124 128-bit double quadword (or oword) data types. This is useful for
11125 high resolution counters that could be updated by multiple
11126 processors (or cores). This instruction is generated as part of
11127 atomic built-in functions: see *Note Atomic Builtins:: for details.
11130 This option will enable GCC to use SAHF instruction in generated
11131 64-bit code. Early Intel CPUs with Intel 64 lacked LAHF and SAHF
11132 instructions supported by AMD64 until introduction of Pentium 4 G1
11133 step in December 2005. LAHF and SAHF are load and store
11134 instructions, respectively, for certain status flags. In 64-bit
11135 mode, SAHF instruction is used to optimize `fmod', `drem' or
11136 `remainder' built-in functions: see *Note Other Builtins:: for
11140 This option will enable GCC to use RCPSS and RSQRTSS instructions
11141 (and their vectorized variants RCPPS and RSQRTPS) with an
11142 additional Newton-Raphson step to increase precision instead of
11143 DIVSS and SQRTSS (and their vectorized variants) for single
11144 precision floating point arguments. These instructions are
11145 generated only when `-funsafe-math-optimizations' is enabled
11146 together with `-finite-math-only' and `-fno-trapping-math'. Note
11147 that while the throughput of the sequence is higher than the
11148 throughput of the non-reciprocal instruction, the precision of the
11149 sequence can be decreased by up to 2 ulp (i.e. the inverse of 1.0
11150 equals 0.99999994).
11153 Specifies the ABI type to use for vectorizing intrinsics using an
11154 external library. Supported types are `svml' for the Intel short
11155 vector math library and `acml' for the AMD math core library style
11156 of interfacing. GCC will currently emit calls to `vmldExp2',
11157 `vmldLn2', `vmldLog102', `vmldLog102', `vmldPow2', `vmldTanh2',
11158 `vmldTan2', `vmldAtan2', `vmldAtanh2', `vmldCbrt2', `vmldSinh2',
11159 `vmldSin2', `vmldAsinh2', `vmldAsin2', `vmldCosh2', `vmldCos2',
11160 `vmldAcosh2', `vmldAcos2', `vmlsExp4', `vmlsLn4', `vmlsLog104',
11161 `vmlsLog104', `vmlsPow4', `vmlsTanh4', `vmlsTan4', `vmlsAtan4',
11162 `vmlsAtanh4', `vmlsCbrt4', `vmlsSinh4', `vmlsSin4', `vmlsAsinh4',
11163 `vmlsAsin4', `vmlsCosh4', `vmlsCos4', `vmlsAcosh4' and `vmlsAcos4'
11164 for corresponding function type when `-mveclibabi=svml' is used
11165 and `__vrd2_sin', `__vrd2_cos', `__vrd2_exp', `__vrd2_log',
11166 `__vrd2_log2', `__vrd2_log10', `__vrs4_sinf', `__vrs4_cosf',
11167 `__vrs4_expf', `__vrs4_logf', `__vrs4_log2f', `__vrs4_log10f' and
11168 `__vrs4_powf' for corresponding function type when
11169 `-mveclibabi=acml' is used. Both `-ftree-vectorize' and
11170 `-funsafe-math-optimizations' have to be enabled. A SVML or ACML
11171 ABI compatible library will have to be specified at link time.
11175 Use PUSH operations to store outgoing parameters. This method is
11176 shorter and usually equally fast as method using SUB/MOV
11177 operations and is enabled by default. In some cases disabling it
11178 may improve performance because of improved scheduling and reduced
11181 `-maccumulate-outgoing-args'
11182 If enabled, the maximum amount of space required for outgoing
11183 arguments will be computed in the function prologue. This is
11184 faster on most modern CPUs because of reduced dependencies,
11185 improved scheduling and reduced stack usage when preferred stack
11186 boundary is not equal to 2. The drawback is a notable increase in
11187 code size. This switch implies `-mno-push-args'.
11190 Support thread-safe exception handling on `Mingw32'. Code that
11191 relies on thread-safe exception handling must compile and link all
11192 code with the `-mthreads' option. When compiling, `-mthreads'
11193 defines `-D_MT'; when linking, it links in a special thread helper
11194 library `-lmingwthrd' which cleans up per thread exception
11197 `-mno-align-stringops'
11198 Do not align destination of inlined string operations. This
11199 switch reduces code size and improves performance in case the
11200 destination is already aligned, but GCC doesn't know about it.
11202 `-minline-all-stringops'
11203 By default GCC inlines string operations only when destination is
11204 known to be aligned at least to 4 byte boundary. This enables
11205 more inlining, increase code size, but may improve performance of
11206 code that depends on fast memcpy, strlen and memset for short
11209 `-minline-stringops-dynamically'
11210 For string operation of unknown size, inline runtime checks so for
11211 small blocks inline code is used, while for large blocks library
11214 `-minline-compares'
11215 This option enables GCC to inline calls to memcmp and strcmp. The
11216 inlined version does a byte-by-byte comparion using a repeat string
11219 `-mstringop-strategy=ALG'
11220 Overwrite internal decision heuristic about particular algorithm
11221 to inline string operation with. The allowed values are
11222 `rep_byte', `rep_4byte', `rep_8byte' for expanding using i386
11223 `rep' prefix of specified size, `byte_loop', `loop',
11224 `unrolled_loop' for expanding inline loop, `libcall' for always
11225 expanding library call.
11227 `-momit-leaf-frame-pointer'
11228 Don't keep the frame pointer in a register for leaf functions.
11229 This avoids the instructions to save, set up and restore frame
11230 pointers and makes an extra register available in leaf functions.
11231 The option `-fomit-frame-pointer' removes the frame pointer for
11232 all functions which might make debugging harder.
11234 `-mtls-direct-seg-refs'
11235 `-mno-tls-direct-seg-refs'
11236 Controls whether TLS variables may be accessed with offsets from
11237 the TLS segment register (`%gs' for 32-bit, `%fs' for 64-bit), or
11238 whether the thread base pointer must be added. Whether or not this
11239 is legal depends on the operating system, and whether it maps the
11240 segment to cover the entire TLS area.
11242 For systems that use GNU libc, the default is on.
11246 Enable automatic generation of fused floating point multiply-add
11247 instructions if the ISA supports such instructions. The
11248 -mfused-madd option is on by default. The fused multiply-add
11249 instructions have a different rounding behavior compared to
11250 executing a multiply followed by an add.
11254 Specify that the assembler should encode SSE instructions with VEX
11255 prefix. The option `-mavx' turns this on by default.
11257 These `-m' switches are supported in addition to the above on AMD
11258 x86-64 processors in 64-bit environments.
11262 Generate code for a 32-bit or 64-bit environment. The 32-bit
11263 environment sets int, long and pointer to 32 bits and generates
11264 code that runs on any i386 system. The 64-bit environment sets
11265 int to 32 bits and long and pointer to 64 bits and generates code
11266 for AMD's x86-64 architecture. For darwin only the -m64 option
11267 turns off the `-fno-pic' and `-mdynamic-no-pic' options.
11270 Do not use a so called red zone for x86-64 code. The red zone is
11271 mandated by the x86-64 ABI, it is a 128-byte area beyond the
11272 location of the stack pointer that will not be modified by signal
11273 or interrupt handlers and therefore can be used for temporary data
11274 without adjusting the stack pointer. The flag `-mno-red-zone'
11275 disables this red zone.
11278 Generate code for the small code model: the program and its
11279 symbols must be linked in the lower 2 GB of the address space.
11280 Pointers are 64 bits. Programs can be statically or dynamically
11281 linked. This is the default code model.
11284 Generate code for the kernel code model. The kernel runs in the
11285 negative 2 GB of the address space. This model has to be used for
11289 Generate code for the medium model: The program is linked in the
11290 lower 2 GB of the address space. Small symbols are also placed
11291 there. Symbols with sizes larger than `-mlarge-data-threshold'
11292 are put into large data or bss sections and can be located above
11293 2GB. Programs can be statically or dynamically linked.
11296 Generate code for the large model: This model makes no assumptions
11297 about addresses and sizes of sections.
11300 File: gcc.info, Node: IA-64 Options, Next: M32C Options, Prev: i386 and x86-64 Windows Options, Up: Submodel Options
11302 3.17.16 IA-64 Options
11303 ---------------------
11305 These are the `-m' options defined for the Intel IA-64 architecture.
11308 Generate code for a big endian target. This is the default for
11312 Generate code for a little endian target. This is the default for
11313 AIX5 and GNU/Linux.
11317 Generate (or don't) code for the GNU assembler. This is the
11322 Generate (or don't) code for the GNU linker. This is the default.
11325 Generate code that does not use a global pointer register. The
11326 result is not position independent code, and violates the IA-64
11329 `-mvolatile-asm-stop'
11330 `-mno-volatile-asm-stop'
11331 Generate (or don't) a stop bit immediately before and after
11332 volatile asm statements.
11335 `-mno-register-names'
11336 Generate (or don't) `in', `loc', and `out' register names for the
11337 stacked registers. This may make assembler output more readable.
11341 Disable (or enable) optimizations that use the small data section.
11342 This may be useful for working around optimizer bugs.
11345 Generate code that uses a single constant global pointer value.
11346 This is useful when compiling kernel code.
11349 Generate code that is self-relocatable. This implies
11350 `-mconstant-gp'. This is useful when compiling firmware code.
11352 `-minline-float-divide-min-latency'
11353 Generate code for inline divides of floating point values using
11354 the minimum latency algorithm.
11356 `-minline-float-divide-max-throughput'
11357 Generate code for inline divides of floating point values using
11358 the maximum throughput algorithm.
11360 `-minline-int-divide-min-latency'
11361 Generate code for inline divides of integer values using the
11362 minimum latency algorithm.
11364 `-minline-int-divide-max-throughput'
11365 Generate code for inline divides of integer values using the
11366 maximum throughput algorithm.
11368 `-minline-sqrt-min-latency'
11369 Generate code for inline square roots using the minimum latency
11372 `-minline-sqrt-max-throughput'
11373 Generate code for inline square roots using the maximum throughput
11378 Don't (or do) generate assembler code for the DWARF2 line number
11379 debugging info. This may be useful when not using the GNU
11382 `-mearly-stop-bits'
11383 `-mno-early-stop-bits'
11384 Allow stop bits to be placed earlier than immediately preceding the
11385 instruction that triggered the stop bit. This can improve
11386 instruction scheduling, but does not always do so.
11388 `-mfixed-range=REGISTER-RANGE'
11389 Generate code treating the given register range as fixed registers.
11390 A fixed register is one that the register allocator can not use.
11391 This is useful when compiling kernel code. A register range is
11392 specified as two registers separated by a dash. Multiple register
11393 ranges can be specified separated by a comma.
11395 `-mtls-size=TLS-SIZE'
11396 Specify bit size of immediate TLS offsets. Valid values are 14,
11400 Tune the instruction scheduling for a particular CPU, Valid values
11401 are itanium, itanium1, merced, itanium2, and mckinley.
11405 Add support for multithreading using the POSIX threads library.
11406 This option sets flags for both the preprocessor and linker. It
11407 does not affect the thread safety of object code produced by the
11408 compiler or that of libraries supplied with it. These are HP-UX
11413 Generate code for a 32-bit or 64-bit environment. The 32-bit
11414 environment sets int, long and pointer to 32 bits. The 64-bit
11415 environment sets int to 32 bits and long and pointer to 64 bits.
11416 These are HP-UX specific flags.
11418 `-mno-sched-br-data-spec'
11419 `-msched-br-data-spec'
11420 (Dis/En)able data speculative scheduling before reload. This will
11421 result in generation of the ld.a instructions and the
11422 corresponding check instructions (ld.c / chk.a). The default is
11425 `-msched-ar-data-spec'
11426 `-mno-sched-ar-data-spec'
11427 (En/Dis)able data speculative scheduling after reload. This will
11428 result in generation of the ld.a instructions and the
11429 corresponding check instructions (ld.c / chk.a). The default is
11432 `-mno-sched-control-spec'
11433 `-msched-control-spec'
11434 (Dis/En)able control speculative scheduling. This feature is
11435 available only during region scheduling (i.e. before reload).
11436 This will result in generation of the ld.s instructions and the
11437 corresponding check instructions chk.s . The default is 'disable'.
11439 `-msched-br-in-data-spec'
11440 `-mno-sched-br-in-data-spec'
11441 (En/Dis)able speculative scheduling of the instructions that are
11442 dependent on the data speculative loads before reload. This is
11443 effective only with `-msched-br-data-spec' enabled. The default
11446 `-msched-ar-in-data-spec'
11447 `-mno-sched-ar-in-data-spec'
11448 (En/Dis)able speculative scheduling of the instructions that are
11449 dependent on the data speculative loads after reload. This is
11450 effective only with `-msched-ar-data-spec' enabled. The default
11453 `-msched-in-control-spec'
11454 `-mno-sched-in-control-spec'
11455 (En/Dis)able speculative scheduling of the instructions that are
11456 dependent on the control speculative loads. This is effective
11457 only with `-msched-control-spec' enabled. The default is 'enable'.
11461 (En/Dis)able use of simple data speculation checks ld.c . If
11462 disabled, only chk.a instructions will be emitted to check data
11463 speculative loads. The default is 'enable'.
11465 `-mno-sched-control-ldc'
11466 `-msched-control-ldc'
11467 (Dis/En)able use of ld.c instructions to check control speculative
11468 loads. If enabled, in case of control speculative load with no
11469 speculatively scheduled dependent instructions this load will be
11470 emitted as ld.sa and ld.c will be used to check it. The default
11473 `-mno-sched-spec-verbose'
11474 `-msched-spec-verbose'
11475 (Dis/En)able printing of the information about speculative motions.
11477 `-mno-sched-prefer-non-data-spec-insns'
11478 `-msched-prefer-non-data-spec-insns'
11479 If enabled, data speculative instructions will be chosen for
11480 schedule only if there are no other choices at the moment. This
11481 will make the use of the data speculation much more conservative.
11482 The default is 'disable'.
11484 `-mno-sched-prefer-non-control-spec-insns'
11485 `-msched-prefer-non-control-spec-insns'
11486 If enabled, control speculative instructions will be chosen for
11487 schedule only if there are no other choices at the moment. This
11488 will make the use of the control speculation much more
11489 conservative. The default is 'disable'.
11491 `-mno-sched-count-spec-in-critical-path'
11492 `-msched-count-spec-in-critical-path'
11493 If enabled, speculative dependencies will be considered during
11494 computation of the instructions priorities. This will make the
11495 use of the speculation a bit more conservative. The default is
11500 File: gcc.info, Node: M32C Options, Next: M32R/D Options, Prev: IA-64 Options, Up: Submodel Options
11502 3.17.17 M32C Options
11503 --------------------
11506 Select the CPU for which code is generated. NAME may be one of
11507 `r8c' for the R8C/Tiny series, `m16c' for the M16C (up to /60)
11508 series, `m32cm' for the M16C/80 series, or `m32c' for the M32C/80
11512 Specifies that the program will be run on the simulator. This
11513 causes an alternate runtime library to be linked in which
11514 supports, for example, file I/O. You must not use this option
11515 when generating programs that will run on real hardware; you must
11516 provide your own runtime library for whatever I/O functions are
11520 Specifies the number of memory-based pseudo-registers GCC will use
11521 during code generation. These pseudo-registers will be used like
11522 real registers, so there is a tradeoff between GCC's ability to
11523 fit the code into available registers, and the performance penalty
11524 of using memory instead of registers. Note that all modules in a
11525 program must be compiled with the same value for this option.
11526 Because of that, you must not use this option with the default
11527 runtime libraries gcc builds.
11531 File: gcc.info, Node: M32R/D Options, Next: M680x0 Options, Prev: M32C Options, Up: Submodel Options
11533 3.17.18 M32R/D Options
11534 ----------------------
11536 These `-m' options are defined for Renesas M32R/D architectures:
11539 Generate code for the M32R/2.
11542 Generate code for the M32R/X.
11545 Generate code for the M32R. This is the default.
11548 Assume all objects live in the lower 16MB of memory (so that their
11549 addresses can be loaded with the `ld24' instruction), and assume
11550 all subroutines are reachable with the `bl' instruction. This is
11553 The addressability of a particular object can be set with the
11557 Assume objects may be anywhere in the 32-bit address space (the
11558 compiler will generate `seth/add3' instructions to load their
11559 addresses), and assume all subroutines are reachable with the `bl'
11563 Assume objects may be anywhere in the 32-bit address space (the
11564 compiler will generate `seth/add3' instructions to load their
11565 addresses), and assume subroutines may not be reachable with the
11566 `bl' instruction (the compiler will generate the much slower
11567 `seth/add3/jl' instruction sequence).
11570 Disable use of the small data area. Variables will be put into
11571 one of `.data', `bss', or `.rodata' (unless the `section'
11572 attribute has been specified). This is the default.
11574 The small data area consists of sections `.sdata' and `.sbss'.
11575 Objects may be explicitly put in the small data area with the
11576 `section' attribute using one of these sections.
11579 Put small global and static data in the small data area, but do not
11580 generate special code to reference them.
11583 Put small global and static data in the small data area, and
11584 generate special instructions to reference them.
11587 Put global and static objects less than or equal to NUM bytes into
11588 the small data or bss sections instead of the normal data or bss
11589 sections. The default value of NUM is 8. The `-msdata' option
11590 must be set to one of `sdata' or `use' for this option to have any
11593 All modules should be compiled with the same `-G NUM' value.
11594 Compiling with different values of NUM may or may not work; if it
11595 doesn't the linker will give an error message--incorrect code will
11599 Makes the M32R specific code in the compiler display some
11600 statistics that might help in debugging programs.
11603 Align all loops to a 32-byte boundary.
11606 Do not enforce a 32-byte alignment for loops. This is the default.
11608 `-missue-rate=NUMBER'
11609 Issue NUMBER instructions per cycle. NUMBER can only be 1 or 2.
11611 `-mbranch-cost=NUMBER'
11612 NUMBER can only be 1 or 2. If it is 1 then branches will be
11613 preferred over conditional code, if it is 2, then the opposite will
11616 `-mflush-trap=NUMBER'
11617 Specifies the trap number to use to flush the cache. The default
11618 is 12. Valid numbers are between 0 and 15 inclusive.
11621 Specifies that the cache cannot be flushed by using a trap.
11623 `-mflush-func=NAME'
11624 Specifies the name of the operating system function to call to
11625 flush the cache. The default is __flush_cache_, but a function
11626 call will only be used if a trap is not available.
11629 Indicates that there is no OS function for flushing the cache.
11633 File: gcc.info, Node: M680x0 Options, Next: M68hc1x Options, Prev: M32R/D Options, Up: Submodel Options
11635 3.17.19 M680x0 Options
11636 ----------------------
11638 These are the `-m' options defined for M680x0 and ColdFire processors.
11639 The default settings depend on which architecture was selected when the
11640 compiler was configured; the defaults for the most common choices are
11644 Generate code for a specific M680x0 or ColdFire instruction set
11645 architecture. Permissible values of ARCH for M680x0 architectures
11646 are: `68000', `68010', `68020', `68030', `68040', `68060' and
11647 `cpu32'. ColdFire architectures are selected according to
11648 Freescale's ISA classification and the permissible values are:
11649 `isaa', `isaaplus', `isab' and `isac'.
11651 gcc defines a macro `__mcfARCH__' whenever it is generating code
11652 for a ColdFire target. The ARCH in this macro is one of the
11653 `-march' arguments given above.
11655 When used together, `-march' and `-mtune' select code that runs on
11656 a family of similar processors but that is optimized for a
11657 particular microarchitecture.
11660 Generate code for a specific M680x0 or ColdFire processor. The
11661 M680x0 CPUs are: `68000', `68010', `68020', `68030', `68040',
11662 `68060', `68302', `68332' and `cpu32'. The ColdFire CPUs are
11663 given by the table below, which also classifies the CPUs into
11666 *Family* *`-mcpu' arguments*
11668 `5206' `5202' `5204' `5206'
11670 `5208' `5207' `5208'
11671 `5211a' `5210a' `5211a'
11672 `5213' `5211' `5212' `5213'
11673 `5216' `5214' `5216'
11674 `52235' `52230' `52231' `52232' `52233' `52234' `52235'
11675 `5225' `5224' `5225'
11676 `5235' `5232' `5233' `5234' `5235' `523x'
11679 `5271' `5270' `5271'
11681 `5275' `5274' `5275'
11682 `5282' `5280' `5281' `5282' `528x'
11684 `5329' `5327' `5328' `5329' `532x'
11685 `5373' `5372' `5373' `537x'
11687 `5475' `5470' `5471' `5472' `5473' `5474' `5475' `547x'
11688 `5480' `5481' `5482' `5483' `5484' `5485'
11690 `-mcpu=CPU' overrides `-march=ARCH' if ARCH is compatible with
11691 CPU. Other combinations of `-mcpu' and `-march' are rejected.
11693 gcc defines the macro `__mcf_cpu_CPU' when ColdFire target CPU is
11694 selected. It also defines `__mcf_family_FAMILY', where the value
11695 of FAMILY is given by the table above.
11698 Tune the code for a particular microarchitecture, within the
11699 constraints set by `-march' and `-mcpu'. The M680x0
11700 microarchitectures are: `68000', `68010', `68020', `68030',
11701 `68040', `68060' and `cpu32'. The ColdFire microarchitectures
11702 are: `cfv1', `cfv2', `cfv3', `cfv4' and `cfv4e'.
11704 You can also use `-mtune=68020-40' for code that needs to run
11705 relatively well on 68020, 68030 and 68040 targets.
11706 `-mtune=68020-60' is similar but includes 68060 targets as well.
11707 These two options select the same tuning decisions as `-m68020-40'
11708 and `-m68020-60' respectively.
11710 gcc defines the macros `__mcARCH' and `__mcARCH__' when tuning for
11711 680x0 architecture ARCH. It also defines `mcARCH' unless either
11712 `-ansi' or a non-GNU `-std' option is used. If gcc is tuning for
11713 a range of architectures, as selected by `-mtune=68020-40' or
11714 `-mtune=68020-60', it defines the macros for every architecture in
11717 gcc also defines the macro `__mUARCH__' when tuning for ColdFire
11718 microarchitecture UARCH, where UARCH is one of the arguments given
11723 Generate output for a 68000. This is the default when the
11724 compiler is configured for 68000-based systems. It is equivalent
11727 Use this option for microcontrollers with a 68000 or EC000 core,
11728 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11731 Generate output for a 68010. This is the default when the
11732 compiler is configured for 68010-based systems. It is equivalent
11737 Generate output for a 68020. This is the default when the
11738 compiler is configured for 68020-based systems. It is equivalent
11742 Generate output for a 68030. This is the default when the
11743 compiler is configured for 68030-based systems. It is equivalent
11747 Generate output for a 68040. This is the default when the
11748 compiler is configured for 68040-based systems. It is equivalent
11751 This option inhibits the use of 68881/68882 instructions that have
11752 to be emulated by software on the 68040. Use this option if your
11753 68040 does not have code to emulate those instructions.
11756 Generate output for a 68060. This is the default when the
11757 compiler is configured for 68060-based systems. It is equivalent
11760 This option inhibits the use of 68020 and 68881/68882 instructions
11761 that have to be emulated by software on the 68060. Use this
11762 option if your 68060 does not have code to emulate those
11766 Generate output for a CPU32. This is the default when the
11767 compiler is configured for CPU32-based systems. It is equivalent
11770 Use this option for microcontrollers with a CPU32 or CPU32+ core,
11771 including the 68330, 68331, 68332, 68333, 68334, 68336, 68340,
11772 68341, 68349 and 68360.
11775 Generate output for a 520X ColdFire CPU. This is the default when
11776 the compiler is configured for 520X-based systems. It is
11777 equivalent to `-mcpu=5206', and is now deprecated in favor of that
11780 Use this option for microcontroller with a 5200 core, including
11781 the MCF5202, MCF5203, MCF5204 and MCF5206.
11784 Generate output for a 5206e ColdFire CPU. The option is now
11785 deprecated in favor of the equivalent `-mcpu=5206e'.
11788 Generate output for a member of the ColdFire 528X family. The
11789 option is now deprecated in favor of the equivalent `-mcpu=528x'.
11792 Generate output for a ColdFire 5307 CPU. The option is now
11793 deprecated in favor of the equivalent `-mcpu=5307'.
11796 Generate output for a ColdFire 5407 CPU. The option is now
11797 deprecated in favor of the equivalent `-mcpu=5407'.
11800 Generate output for a ColdFire V4e family CPU (e.g. 547x/548x).
11801 This includes use of hardware floating point instructions. The
11802 option is equivalent to `-mcpu=547x', and is now deprecated in
11803 favor of that option.
11806 Generate output for a 68040, without using any of the new
11807 instructions. This results in code which can run relatively
11808 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11809 generated code does use the 68881 instructions that are emulated
11812 The option is equivalent to `-march=68020' `-mtune=68020-40'.
11815 Generate output for a 68060, without using any of the new
11816 instructions. This results in code which can run relatively
11817 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11818 generated code does use the 68881 instructions that are emulated
11821 The option is equivalent to `-march=68020' `-mtune=68020-60'.
11825 Generate floating-point instructions. This is the default for
11826 68020 and above, and for ColdFire devices that have an FPU. It
11827 defines the macro `__HAVE_68881__' on M680x0 targets and
11828 `__mcffpu__' on ColdFire targets.
11831 Do not generate floating-point instructions; use library calls
11832 instead. This is the default for 68000, 68010, and 68832 targets.
11833 It is also the default for ColdFire devices that have no FPU.
11837 Generate (do not generate) ColdFire hardware divide and remainder
11838 instructions. If `-march' is used without `-mcpu', the default is
11839 "on" for ColdFire architectures and "off" for M680x0
11840 architectures. Otherwise, the default is taken from the target CPU
11841 (either the default CPU, or the one specified by `-mcpu'). For
11842 example, the default is "off" for `-mcpu=5206' and "on" for
11845 gcc defines the macro `__mcfhwdiv__' when this option is enabled.
11848 Consider type `int' to be 16 bits wide, like `short int'.
11849 Additionally, parameters passed on the stack are also aligned to a
11850 16-bit boundary even on targets whose API mandates promotion to
11854 Do not consider type `int' to be 16 bits wide. This is the
11859 Do not use the bit-field instructions. The `-m68000', `-mcpu32'
11860 and `-m5200' options imply `-mnobitfield'.
11863 Do use the bit-field instructions. The `-m68020' option implies
11864 `-mbitfield'. This is the default if you use a configuration
11865 designed for a 68020.
11868 Use a different function-calling convention, in which functions
11869 that take a fixed number of arguments return with the `rtd'
11870 instruction, which pops their arguments while returning. This
11871 saves one instruction in the caller since there is no need to pop
11872 the arguments there.
11874 This calling convention is incompatible with the one normally used
11875 on Unix, so you cannot use it if you need to call libraries
11876 compiled with the Unix compiler.
11878 Also, you must provide function prototypes for all functions that
11879 take variable numbers of arguments (including `printf'); otherwise
11880 incorrect code will be generated for calls to those functions.
11882 In addition, seriously incorrect code will result if you call a
11883 function with too many arguments. (Normally, extra arguments are
11884 harmlessly ignored.)
11886 The `rtd' instruction is supported by the 68010, 68020, 68030,
11887 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11890 Do not use the calling conventions selected by `-mrtd'. This is
11895 Control whether GCC aligns `int', `long', `long long', `float',
11896 `double', and `long double' variables on a 32-bit boundary
11897 (`-malign-int') or a 16-bit boundary (`-mno-align-int'). Aligning
11898 variables on 32-bit boundaries produces code that runs somewhat
11899 faster on processors with 32-bit busses at the expense of more
11902 *Warning:* if you use the `-malign-int' switch, GCC will align
11903 structures containing the above types differently than most
11904 published application binary interface specifications for the m68k.
11907 Use the pc-relative addressing mode of the 68000 directly, instead
11908 of using a global offset table. At present, this option implies
11909 `-fpic', allowing at most a 16-bit offset for pc-relative
11910 addressing. `-fPIC' is not presently supported with `-mpcrel',
11911 though this could be supported for 68020 and higher processors.
11913 `-mno-strict-align'
11915 Do not (do) assume that unaligned memory references will be
11916 handled by the system.
11919 Generate code that allows the data segment to be located in a
11920 different area of memory from the text segment. This allows for
11921 execute in place in an environment without virtual memory
11922 management. This option implies `-fPIC'.
11925 Generate code that assumes that the data segment follows the text
11926 segment. This is the default.
11928 `-mid-shared-library'
11929 Generate code that supports shared libraries via the library ID
11930 method. This allows for execute in place and shared libraries in
11931 an environment without virtual memory management. This option
11934 `-mno-id-shared-library'
11935 Generate code that doesn't assume ID based shared libraries are
11936 being used. This is the default.
11938 `-mshared-library-id=n'
11939 Specified the identification number of the ID based shared library
11940 being compiled. Specifying a value of 0 will generate more
11941 compact code, specifying other values will force the allocation of
11942 that number to the current library but is no more space or time
11943 efficient than omitting this option.
11947 When generating position-independent code for ColdFire, generate
11948 code that works if the GOT has more than 8192 entries. This code
11949 is larger and slower than code generated without this option. On
11950 M680x0 processors, this option is not needed; `-fPIC' suffices.
11952 GCC normally uses a single instruction to load values from the GOT.
11953 While this is relatively efficient, it only works if the GOT is
11954 smaller than about 64k. Anything larger causes the linker to
11955 report an error such as:
11957 relocation truncated to fit: R_68K_GOT16O foobar
11959 If this happens, you should recompile your code with `-mxgot'. It
11960 should then work with very large GOTs. However, code generated
11961 with `-mxgot' is less efficient, since it takes 4 instructions to
11962 fetch the value of a global symbol.
11964 Note that some linkers, including newer versions of the GNU linker,
11965 can create multiple GOTs and sort GOT entries. If you have such a
11966 linker, you should only need to use `-mxgot' when compiling a
11967 single object file that accesses more than 8192 GOT entries. Very
11970 These options have no effect unless GCC is generating
11971 position-independent code.
11975 File: gcc.info, Node: M68hc1x Options, Next: MCore Options, Prev: M680x0 Options, Up: Submodel Options
11977 3.17.20 M68hc1x Options
11978 -----------------------
11980 These are the `-m' options defined for the 68hc11 and 68hc12
11981 microcontrollers. The default values for these options depends on
11982 which style of microcontroller was selected when the compiler was
11983 configured; the defaults for the most common choices are given below.
11987 Generate output for a 68HC11. This is the default when the
11988 compiler is configured for 68HC11-based systems.
11992 Generate output for a 68HC12. This is the default when the
11993 compiler is configured for 68HC12-based systems.
11997 Generate output for a 68HCS12.
12000 Enable the use of 68HC12 pre and post auto-increment and
12001 auto-decrement addressing modes.
12005 Enable the use of 68HC12 min and max instructions.
12009 Treat all calls as being far away (near). If calls are assumed to
12010 be far away, the compiler will use the `call' instruction to call
12011 a function and the `rtc' instruction for returning.
12014 Consider type `int' to be 16 bits wide, like `short int'.
12016 `-msoft-reg-count=COUNT'
12017 Specify the number of pseudo-soft registers which are used for the
12018 code generation. The maximum number is 32. Using more pseudo-soft
12019 register may or may not result in better code depending on the
12020 program. The default is 4 for 68HC11 and 2 for 68HC12.
12024 File: gcc.info, Node: MCore Options, Next: MIPS Options, Prev: M68hc1x Options, Up: Submodel Options
12026 3.17.21 MCore Options
12027 ---------------------
12029 These are the `-m' options defined for the Motorola M*Core processors.
12033 Inline constants into the code stream if it can be done in two
12034 instructions or less.
12038 Use the divide instruction. (Enabled by default).
12040 `-mrelax-immediate'
12041 `-mno-relax-immediate'
12042 Allow arbitrary sized immediates in bit operations.
12045 `-mno-wide-bitfields'
12046 Always treat bit-fields as int-sized.
12048 `-m4byte-functions'
12049 `-mno-4byte-functions'
12050 Force all functions to be aligned to a four byte boundary.
12053 `-mno-callgraph-data'
12054 Emit callgraph information.
12058 Prefer word access when reading byte quantities.
12062 Generate code for a little endian target.
12066 Generate code for the 210 processor.
12069 Assume that run-time support has been provided and so omit the
12070 simulator library (`libsim.a)' from the linker command line.
12072 `-mstack-increment=SIZE'
12073 Set the maximum amount for a single stack increment operation.
12074 Large values can increase the speed of programs which contain
12075 functions that need a large amount of stack space, but they can
12076 also trigger a segmentation fault if the stack is extended too
12077 much. The default value is 0x1000.
12081 File: gcc.info, Node: MIPS Options, Next: MMIX Options, Prev: MCore Options, Up: Submodel Options
12083 3.17.22 MIPS Options
12084 --------------------
12087 Generate big-endian code.
12090 Generate little-endian code. This is the default for `mips*el-*-*'
12094 Generate code that will run on ARCH, which can be the name of a
12095 generic MIPS ISA, or the name of a particular processor. The ISA
12096 names are: `mips1', `mips2', `mips3', `mips4', `mips32',
12097 `mips32r2', `mips64' and `mips64r2'. The processor names are:
12098 `4kc', `4km', `4kp', `4ksc', `4kec', `4kem', `4kep', `4ksd',
12099 `5kc', `5kf', `20kc', `24kc', `24kf2_1', `24kf1_1', `24kec',
12100 `24kef2_1', `24kef1_1', `34kc', `34kf2_1', `34kf1_1', `74kc',
12101 `74kf2_1', `74kf1_1', `74kf3_2', `loongson2e', `loongson2f', `m4k',
12102 `octeon', `orion', `r2000', `r3000', `r3900', `r4000', `r4400',
12103 `r4600', `r4650', `r6000', `r8000', `rm7000', `rm9000', `r10000',
12104 `r12000', `r14000', `r16000', `sb1', `sr71000', `vr4100',
12105 `vr4111', `vr4120', `vr4130', `vr4300', `vr5000', `vr5400',
12106 `vr5500' and `xlr'. The special value `from-abi' selects the most
12107 compatible architecture for the selected ABI (that is, `mips1' for
12108 32-bit ABIs and `mips3' for 64-bit ABIs).
12110 Native Linux/GNU toolchains also support the value `native', which
12111 selects the best architecture option for the host processor.
12112 `-march=native' has no effect if GCC does not recognize the
12115 In processor names, a final `000' can be abbreviated as `k' (for
12116 example, `-march=r2k'). Prefixes are optional, and `vr' may be
12119 Names of the form `Nf2_1' refer to processors with FPUs clocked at
12120 half the rate of the core, names of the form `Nf1_1' refer to
12121 processors with FPUs clocked at the same rate as the core, and
12122 names of the form `Nf3_2' refer to processors with FPUs clocked a
12123 ratio of 3:2 with respect to the core. For compatibility reasons,
12124 `Nf' is accepted as a synonym for `Nf2_1' while `Nx' and `Bfx' are
12125 accepted as synonyms for `Nf1_1'.
12127 GCC defines two macros based on the value of this option. The
12128 first is `_MIPS_ARCH', which gives the name of target
12129 architecture, as a string. The second has the form
12130 `_MIPS_ARCH_FOO', where FOO is the capitalized value of
12131 `_MIPS_ARCH'. For example, `-march=r2000' will set `_MIPS_ARCH'
12132 to `"r2000"' and define the macro `_MIPS_ARCH_R2000'.
12134 Note that the `_MIPS_ARCH' macro uses the processor names given
12135 above. In other words, it will have the full prefix and will not
12136 abbreviate `000' as `k'. In the case of `from-abi', the macro
12137 names the resolved architecture (either `"mips1"' or `"mips3"').
12138 It names the default architecture when no `-march' option is given.
12141 Optimize for ARCH. Among other things, this option controls the
12142 way instructions are scheduled, and the perceived cost of
12143 arithmetic operations. The list of ARCH values is the same as for
12146 When this option is not used, GCC will optimize for the processor
12147 specified by `-march'. By using `-march' and `-mtune' together,
12148 it is possible to generate code that will run on a family of
12149 processors, but optimize the code for one particular member of
12152 `-mtune' defines the macros `_MIPS_TUNE' and `_MIPS_TUNE_FOO',
12153 which work in the same way as the `-march' ones described above.
12156 Equivalent to `-march=mips1'.
12159 Equivalent to `-march=mips2'.
12162 Equivalent to `-march=mips3'.
12165 Equivalent to `-march=mips4'.
12168 Equivalent to `-march=mips32'.
12171 Equivalent to `-march=mips32r2'.
12174 Equivalent to `-march=mips64'.
12177 Equivalent to `-march=mips64r2'.
12181 Generate (do not generate) MIPS16 code. If GCC is targetting a
12182 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE.
12184 MIPS16 code generation can also be controlled on a per-function
12185 basis by means of `mips16' and `nomips16' attributes. *Note
12186 Function Attributes::, for more information.
12189 Generate MIPS16 code on alternating functions. This option is
12190 provided for regression testing of mixed MIPS16/non-MIPS16 code
12191 generation, and is not intended for ordinary use in compiling user
12194 `-minterlink-mips16'
12195 `-mno-interlink-mips16'
12196 Require (do not require) that non-MIPS16 code be link-compatible
12199 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12200 it must either use a call or an indirect jump.
12201 `-minterlink-mips16' therefore disables direct jumps unless GCC
12202 knows that the target of the jump is not MIPS16.
12209 Generate code for the given ABI.
12211 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12212 generates 64-bit code when you select a 64-bit architecture, but
12213 you can use `-mgp32' to get 32-bit code instead.
12215 For information about the O64 ABI, see
12216 `http://gcc.gnu.org/projects/mipso64-abi.html'.
12218 GCC supports a variant of the o32 ABI in which floating-point
12219 registers are 64 rather than 32 bits wide. You can select this
12220 combination with `-mabi=32' `-mfp64'. This ABI relies on the
12221 `mthc1' and `mfhc1' instructions and is therefore only supported
12222 for MIPS32R2 processors.
12224 The register assignments for arguments and return values remain the
12225 same, but each scalar value is passed in a single 64-bit register
12226 rather than a pair of 32-bit registers. For example, scalar
12227 floating-point values are returned in `$f0' only, not a
12228 `$f0'/`$f1' pair. The set of call-saved registers also remains
12229 the same, but all 64 bits are saved.
12233 Generate (do not generate) code that is suitable for SVR4-style
12234 dynamic objects. `-mabicalls' is the default for SVR4-based
12239 Generate (do not generate) code that is fully position-independent,
12240 and that can therefore be linked into shared libraries. This
12241 option only affects `-mabicalls'.
12243 All `-mabicalls' code has traditionally been position-independent,
12244 regardless of options like `-fPIC' and `-fpic'. However, as an
12245 extension, the GNU toolchain allows executables to use absolute
12246 accesses for locally-binding symbols. It can also use shorter GP
12247 initialization sequences and generate direct calls to
12248 locally-defined functions. This mode is selected by `-mno-shared'.
12250 `-mno-shared' depends on binutils 2.16 or higher and generates
12251 objects that can only be linked by the GNU linker. However, the
12252 option does not affect the ABI of the final executable; it only
12253 affects the ABI of relocatable objects. Using `-mno-shared' will
12254 generally make executables both smaller and quicker.
12256 `-mshared' is the default.
12260 Assume (do not assume) that the static and dynamic linkers support
12261 PLTs and copy relocations. This option only affects `-mno-shared
12262 -mabicalls'. For the n64 ABI, this option has no effect without
12265 You can make `-mplt' the default by configuring GCC with
12266 `--with-mips-plt'. The default is `-mno-plt' otherwise.
12270 Lift (do not lift) the usual restrictions on the size of the global
12273 GCC normally uses a single instruction to load values from the GOT.
12274 While this is relatively efficient, it will only work if the GOT
12275 is smaller than about 64k. Anything larger will cause the linker
12276 to report an error such as:
12278 relocation truncated to fit: R_MIPS_GOT16 foobar
12280 If this happens, you should recompile your code with `-mxgot'. It
12281 should then work with very large GOTs, although it will also be
12282 less efficient, since it will take three instructions to fetch the
12283 value of a global symbol.
12285 Note that some linkers can create multiple GOTs. If you have such
12286 a linker, you should only need to use `-mxgot' when a single object
12287 file accesses more than 64k's worth of GOT entries. Very few do.
12289 These options have no effect unless GCC is generating position
12293 Assume that general-purpose registers are 32 bits wide.
12296 Assume that general-purpose registers are 64 bits wide.
12299 Assume that floating-point registers are 32 bits wide.
12302 Assume that floating-point registers are 64 bits wide.
12305 Use floating-point coprocessor instructions.
12308 Do not use floating-point coprocessor instructions. Implement
12309 floating-point calculations using library calls instead.
12312 Assume that the floating-point coprocessor only supports
12313 single-precision operations.
12316 Assume that the floating-point coprocessor supports
12317 double-precision operations. This is the default.
12321 Use (do not use) `ll', `sc', and `sync' instructions to implement
12322 atomic memory built-in functions. When neither option is
12323 specified, GCC will use the instructions if the target architecture
12326 `-mllsc' is useful if the runtime environment can emulate the
12327 instructions and `-mno-llsc' can be useful when compiling for
12328 nonstandard ISAs. You can make either option the default by
12329 configuring GCC with `--with-llsc' and `--without-llsc'
12330 respectively. `--with-llsc' is the default for some
12331 configurations; see the installation documentation for details.
12335 Use (do not use) revision 1 of the MIPS DSP ASE. *Note MIPS DSP
12336 Built-in Functions::. This option defines the preprocessor macro
12337 `__mips_dsp'. It also defines `__mips_dsp_rev' to 1.
12341 Use (do not use) revision 2 of the MIPS DSP ASE. *Note MIPS DSP
12342 Built-in Functions::. This option defines the preprocessor macros
12343 `__mips_dsp' and `__mips_dspr2'. It also defines `__mips_dsp_rev'
12348 Use (do not use) the MIPS SmartMIPS ASE.
12351 `-mno-paired-single'
12352 Use (do not use) paired-single floating-point instructions. *Note
12353 MIPS Paired-Single Support::. This option requires hardware
12354 floating-point support to be enabled.
12358 Use (do not use) MIPS Digital Media Extension instructions. This
12359 option can only be used when generating 64-bit code and requires
12360 hardware floating-point support to be enabled.
12364 Use (do not use) the MIPS-3D ASE. *Note MIPS-3D Built-in
12365 Functions::. The option `-mips3d' implies `-mpaired-single'.
12369 Use (do not use) MT Multithreading instructions.
12372 Force `long' types to be 64 bits wide. See `-mlong32' for an
12373 explanation of the default and the way that the pointer size is
12377 Force `long', `int', and pointer types to be 32 bits wide.
12379 The default size of `int's, `long's and pointers depends on the
12380 ABI. All the supported ABIs use 32-bit `int's. The n64 ABI uses
12381 64-bit `long's, as does the 64-bit EABI; the others use 32-bit
12382 `long's. Pointers are the same size as `long's, or the same size
12383 as integer registers, whichever is smaller.
12387 Assume (do not assume) that all symbols have 32-bit values,
12388 regardless of the selected ABI. This option is useful in
12389 combination with `-mabi=64' and `-mno-abicalls' because it allows
12390 GCC to generate shorter and faster references to symbolic
12394 Put definitions of externally-visible data in a small data section
12395 if that data is no bigger than NUM bytes. GCC can then access the
12396 data more efficiently; see `-mgpopt' for details.
12398 The default `-G' option depends on the configuration.
12402 Extend (do not extend) the `-G' behavior to local data too, such
12403 as to static variables in C. `-mlocal-sdata' is the default for
12404 all configurations.
12406 If the linker complains that an application is using too much
12407 small data, you might want to try rebuilding the less
12408 performance-critical parts with `-mno-local-sdata'. You might
12409 also want to build large libraries with `-mno-local-sdata', so
12410 that the libraries leave more room for the main program.
12413 `-mno-extern-sdata'
12414 Assume (do not assume) that externally-defined data will be in a
12415 small data section if that data is within the `-G' limit.
12416 `-mextern-sdata' is the default for all configurations.
12418 If you compile a module MOD with `-mextern-sdata' `-G NUM'
12419 `-mgpopt', and MOD references a variable VAR that is no bigger
12420 than NUM bytes, you must make sure that VAR is placed in a small
12421 data section. If VAR is defined by another module, you must
12422 either compile that module with a high-enough `-G' setting or
12423 attach a `section' attribute to VAR's definition. If VAR is
12424 common, you must link the application with a high-enough `-G'
12427 The easiest way of satisfying these restrictions is to compile and
12428 link every module with the same `-G' option. However, you may
12429 wish to build a library that supports several different small data
12430 limits. You can do this by compiling the library with the highest
12431 supported `-G' setting and additionally using `-mno-extern-sdata'
12432 to stop the library from making assumptions about
12433 externally-defined data.
12437 Use (do not use) GP-relative accesses for symbols that are known
12438 to be in a small data section; see `-G', `-mlocal-sdata' and
12439 `-mextern-sdata'. `-mgpopt' is the default for all configurations.
12441 `-mno-gpopt' is useful for cases where the `$gp' register might
12442 not hold the value of `_gp'. For example, if the code is part of
12443 a library that might be used in a boot monitor, programs that call
12444 boot monitor routines will pass an unknown value in `$gp'. (In
12445 such situations, the boot monitor itself would usually be compiled
12448 `-mno-gpopt' implies `-mno-local-sdata' and `-mno-extern-sdata'.
12451 `-mno-embedded-data'
12452 Allocate variables to the read-only data section first if
12453 possible, then next in the small data section if possible,
12454 otherwise in data. This gives slightly slower code than the
12455 default, but reduces the amount of RAM required when executing,
12456 and thus may be preferred for some embedded systems.
12458 `-muninit-const-in-rodata'
12459 `-mno-uninit-const-in-rodata'
12460 Put uninitialized `const' variables in the read-only data section.
12461 This option is only meaningful in conjunction with
12464 `-mcode-readable=SETTING'
12465 Specify whether GCC may generate code that reads from executable
12466 sections. There are three possible settings:
12468 `-mcode-readable=yes'
12469 Instructions may freely access executable sections. This is
12470 the default setting.
12472 `-mcode-readable=pcrel'
12473 MIPS16 PC-relative load instructions can access executable
12474 sections, but other instructions must not do so. This option
12475 is useful on 4KSc and 4KSd processors when the code TLBs have
12476 the Read Inhibit bit set. It is also useful on processors
12477 that can be configured to have a dual instruction/data SRAM
12478 interface and that, like the M4K, automatically redirect
12479 PC-relative loads to the instruction RAM.
12481 `-mcode-readable=no'
12482 Instructions must not access executable sections. This
12483 option can be useful on targets that are configured to have a
12484 dual instruction/data SRAM interface but that (unlike the
12485 M4K) do not automatically redirect PC-relative loads to the
12488 `-msplit-addresses'
12489 `-mno-split-addresses'
12490 Enable (disable) use of the `%hi()' and `%lo()' assembler
12491 relocation operators. This option has been superseded by
12492 `-mexplicit-relocs' but is retained for backwards compatibility.
12494 `-mexplicit-relocs'
12495 `-mno-explicit-relocs'
12496 Use (do not use) assembler relocation operators when dealing with
12497 symbolic addresses. The alternative, selected by
12498 `-mno-explicit-relocs', is to use assembler macros instead.
12500 `-mexplicit-relocs' is the default if GCC was configured to use an
12501 assembler that supports relocation operators.
12503 `-mcheck-zero-division'
12504 `-mno-check-zero-division'
12505 Trap (do not trap) on integer division by zero.
12507 The default is `-mcheck-zero-division'.
12511 MIPS systems check for division by zero by generating either a
12512 conditional trap or a break instruction. Using traps results in
12513 smaller code, but is only supported on MIPS II and later. Also,
12514 some versions of the Linux kernel have a bug that prevents trap
12515 from generating the proper signal (`SIGFPE'). Use
12516 `-mdivide-traps' to allow conditional traps on architectures that
12517 support them and `-mdivide-breaks' to force the use of breaks.
12519 The default is usually `-mdivide-traps', but this can be
12520 overridden at configure time using `--with-divide=breaks'.
12521 Divide-by-zero checks can be completely disabled using
12522 `-mno-check-zero-division'.
12526 Force (do not force) the use of `memcpy()' for non-trivial block
12527 moves. The default is `-mno-memcpy', which allows GCC to inline
12528 most constant-sized copies.
12532 Disable (do not disable) use of the `jal' instruction. Calling
12533 functions using `jal' is more efficient but requires the caller
12534 and callee to be in the same 256 megabyte segment.
12536 This option has no effect on abicalls code. The default is
12541 Enable (disable) use of the `mad', `madu' and `mul' instructions,
12542 as provided by the R4650 ISA.
12546 Enable (disable) use of the floating point multiply-accumulate
12547 instructions, when they are available. The default is
12550 When multiply-accumulate instructions are used, the intermediate
12551 product is calculated to infinite precision and is not subject to
12552 the FCSR Flush to Zero bit. This may be undesirable in some
12556 Tell the MIPS assembler to not run its preprocessor over user
12557 assembler files (with a `.s' suffix) when assembling them.
12561 Work around certain R4000 CPU errata:
12562 - A double-word or a variable shift may give an incorrect
12563 result if executed immediately after starting an integer
12566 - A double-word or a variable shift may give an incorrect
12567 result if executed while an integer multiplication is in
12570 - An integer division may give an incorrect result if started
12571 in a delay slot of a taken branch or a jump.
12575 Work around certain R4400 CPU errata:
12576 - A double-word or a variable shift may give an incorrect
12577 result if executed immediately after starting an integer
12582 Work around certain R10000 errata:
12583 - `ll'/`sc' sequences may not behave atomically on revisions
12584 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
12586 This option can only be used if the target architecture supports
12587 branch-likely instructions. `-mfix-r10000' is the default when
12588 `-march=r10000' is used; `-mno-fix-r10000' is the default
12593 Work around certain VR4120 errata:
12594 - `dmultu' does not always produce the correct result.
12596 - `div' and `ddiv' do not always produce the correct result if
12597 one of the operands is negative.
12598 The workarounds for the division errata rely on special functions
12599 in `libgcc.a'. At present, these functions are only provided by
12600 the `mips64vr*-elf' configurations.
12602 Other VR4120 errata require a nop to be inserted between certain
12603 pairs of instructions. These errata are handled by the assembler,
12607 Work around the VR4130 `mflo'/`mfhi' errata. The workarounds are
12608 implemented by the assembler rather than by GCC, although GCC will
12609 avoid using `mflo' and `mfhi' if the VR4130 `macc', `macchi',
12610 `dmacc' and `dmacchi' instructions are available instead.
12614 Work around certain SB-1 CPU core errata. (This flag currently
12615 works around the SB-1 revision 2 "F1" and "F2" floating point
12618 `-mr10k-cache-barrier=SETTING'
12619 Specify whether GCC should insert cache barriers to avoid the
12620 side-effects of speculation on R10K processors.
12622 In common with many processors, the R10K tries to predict the
12623 outcome of a conditional branch and speculatively executes
12624 instructions from the "taken" branch. It later aborts these
12625 instructions if the predicted outcome was wrong. However, on the
12626 R10K, even aborted instructions can have side effects.
12628 This problem only affects kernel stores and, depending on the
12629 system, kernel loads. As an example, a speculatively-executed
12630 store may load the target memory into cache and mark the cache
12631 line as dirty, even if the store itself is later aborted. If a
12632 DMA operation writes to the same area of memory before the "dirty"
12633 line is flushed, the cached data will overwrite the DMA-ed data.
12634 See the R10K processor manual for a full description, including
12635 other potential problems.
12637 One workaround is to insert cache barrier instructions before
12638 every memory access that might be speculatively executed and that
12639 might have side effects even if aborted.
12640 `-mr10k-cache-barrier=SETTING' controls GCC's implementation of
12641 this workaround. It assumes that aborted accesses to any byte in
12642 the following regions will not have side effects:
12644 1. the memory occupied by the current function's stack frame;
12646 2. the memory occupied by an incoming stack argument;
12648 3. the memory occupied by an object with a link-time-constant
12651 It is the kernel's responsibility to ensure that speculative
12652 accesses to these regions are indeed safe.
12654 If the input program contains a function declaration such as:
12658 then the implementation of `foo' must allow `j foo' and `jal foo'
12659 to be executed speculatively. GCC honors this restriction for
12660 functions it compiles itself. It expects non-GCC functions (such
12661 as hand-written assembly code) to do the same.
12663 The option has three forms:
12665 `-mr10k-cache-barrier=load-store'
12666 Insert a cache barrier before a load or store that might be
12667 speculatively executed and that might have side effects even
12670 `-mr10k-cache-barrier=store'
12671 Insert a cache barrier before a store that might be
12672 speculatively executed and that might have side effects even
12675 `-mr10k-cache-barrier=none'
12676 Disable the insertion of cache barriers. This is the default
12679 `-mflush-func=FUNC'
12681 Specifies the function to call to flush the I and D caches, or to
12682 not call any such function. If called, the function must take the
12683 same arguments as the common `_flush_func()', that is, the address
12684 of the memory range for which the cache is being flushed, the size
12685 of the memory range, and the number 3 (to flush both caches). The
12686 default depends on the target GCC was configured for, but commonly
12687 is either `_flush_func' or `__cpu_flush'.
12690 Set the cost of branches to roughly NUM "simple" instructions.
12691 This cost is only a heuristic and is not guaranteed to produce
12692 consistent results across releases. A zero cost redundantly
12693 selects the default, which is based on the `-mtune' setting.
12696 `-mno-branch-likely'
12697 Enable or disable use of Branch Likely instructions, regardless of
12698 the default for the selected architecture. By default, Branch
12699 Likely instructions may be generated if they are supported by the
12700 selected architecture. An exception is for the MIPS32 and MIPS64
12701 architectures and processors which implement those architectures;
12702 for those, Branch Likely instructions will not be generated by
12703 default because the MIPS32 and MIPS64 architectures specifically
12704 deprecate their use.
12707 `-mno-fp-exceptions'
12708 Specifies whether FP exceptions are enabled. This affects how we
12709 schedule FP instructions for some processors. The default is that
12710 FP exceptions are enabled.
12712 For instance, on the SB-1, if FP exceptions are disabled, and we
12713 are emitting 64-bit code, then we can use both FP pipes.
12714 Otherwise, we can only use one FP pipe.
12717 `-mno-vr4130-align'
12718 The VR4130 pipeline is two-way superscalar, but can only issue two
12719 instructions together if the first one is 8-byte aligned. When
12720 this option is enabled, GCC will align pairs of instructions that
12721 it thinks should execute in parallel.
12723 This option only has an effect when optimizing for the VR4130. It
12724 normally makes code faster, but at the expense of making it bigger.
12725 It is enabled by default at optimization level `-O3'.
12728 File: gcc.info, Node: MMIX Options, Next: MN10300 Options, Prev: MIPS Options, Up: Submodel Options
12730 3.17.23 MMIX Options
12731 --------------------
12733 These options are defined for the MMIX:
12737 Specify that intrinsic library functions are being compiled,
12738 passing all values in registers, no matter the size.
12742 Generate floating-point comparison instructions that compare with
12743 respect to the `rE' epsilon register.
12747 Generate code that passes function parameters and return values
12748 that (in the called function) are seen as registers `$0' and up,
12749 as opposed to the GNU ABI which uses global registers `$231' and
12754 When reading data from memory in sizes shorter than 64 bits, use
12755 (do not use) zero-extending load instructions by default, rather
12756 than sign-extending ones.
12760 Make the result of a division yielding a remainder have the same
12761 sign as the divisor. With the default, `-mno-knuthdiv', the sign
12762 of the remainder follows the sign of the dividend. Both methods
12763 are arithmetically valid, the latter being almost exclusively used.
12765 `-mtoplevel-symbols'
12766 `-mno-toplevel-symbols'
12767 Prepend (do not prepend) a `:' to all global symbols, so the
12768 assembly code can be used with the `PREFIX' assembly directive.
12771 Generate an executable in the ELF format, rather than the default
12772 `mmo' format used by the `mmix' simulator.
12775 `-mno-branch-predict'
12776 Use (do not use) the probable-branch instructions, when static
12777 branch prediction indicates a probable branch.
12780 `-mno-base-addresses'
12781 Generate (do not generate) code that uses _base addresses_. Using
12782 a base address automatically generates a request (handled by the
12783 assembler and the linker) for a constant to be set up in a global
12784 register. The register is used for one or more base address
12785 requests within the range 0 to 255 from the value held in the
12786 register. The generally leads to short and fast code, but the
12787 number of different data items that can be addressed is limited.
12788 This means that a program that uses lots of static data may
12789 require `-mno-base-addresses'.
12793 Force (do not force) generated code to have a single exit point in
12797 File: gcc.info, Node: MN10300 Options, Next: PDP-11 Options, Prev: MMIX Options, Up: Submodel Options
12799 3.17.24 MN10300 Options
12800 -----------------------
12802 These `-m' options are defined for Matsushita MN10300 architectures:
12805 Generate code to avoid bugs in the multiply instructions for the
12806 MN10300 processors. This is the default.
12809 Do not generate code to avoid bugs in the multiply instructions
12810 for the MN10300 processors.
12813 Generate code which uses features specific to the AM33 processor.
12816 Do not generate code which uses features specific to the AM33
12817 processor. This is the default.
12819 `-mreturn-pointer-on-d0'
12820 When generating a function which returns a pointer, return the
12821 pointer in both `a0' and `d0'. Otherwise, the pointer is returned
12822 only in a0, and attempts to call such functions without a prototype
12823 would result in errors. Note that this option is on by default;
12824 use `-mno-return-pointer-on-d0' to disable it.
12827 Do not link in the C run-time initialization object file.
12830 Indicate to the linker that it should perform a relaxation
12831 optimization pass to shorten branches, calls and absolute memory
12832 addresses. This option only has an effect when used on the
12833 command line for the final link step.
12835 This option makes symbolic debugging impossible.
12838 File: gcc.info, Node: PDP-11 Options, Next: picoChip Options, Prev: MN10300 Options, Up: Submodel Options
12840 3.17.25 PDP-11 Options
12841 ----------------------
12843 These options are defined for the PDP-11:
12846 Use hardware FPP floating point. This is the default. (FIS
12847 floating point on the PDP-11/40 is not supported.)
12850 Do not use hardware floating point.
12853 Return floating-point results in ac0 (fr0 in Unix assembler
12857 Return floating-point results in memory. This is the default.
12860 Generate code for a PDP-11/40.
12863 Generate code for a PDP-11/45. This is the default.
12866 Generate code for a PDP-11/10.
12869 Use inline `movmemhi' patterns for copying memory. This is the
12873 Do not use inline `movmemhi' patterns for copying memory.
12877 Use 16-bit `int'. This is the default.
12885 Use 64-bit `float'. This is the default.
12889 Use 32-bit `float'.
12892 Use `abshi2' pattern. This is the default.
12895 Do not use `abshi2' pattern.
12897 `-mbranch-expensive'
12898 Pretend that branches are expensive. This is for experimenting
12899 with code generation only.
12902 Do not pretend that branches are expensive. This is the default.
12905 Generate code for a system with split I&D.
12908 Generate code for a system without split I&D. This is the default.
12911 Use Unix assembler syntax. This is the default when configured for
12915 Use DEC assembler syntax. This is the default when configured for
12916 any PDP-11 target other than `pdp11-*-bsd'.
12919 File: gcc.info, Node: picoChip Options, Next: PowerPC Options, Prev: PDP-11 Options, Up: Submodel Options
12921 3.17.26 picoChip Options
12922 ------------------------
12924 These `-m' options are defined for picoChip implementations:
12927 Set the instruction set, register set, and instruction scheduling
12928 parameters for array element type AE_TYPE. Supported values for
12929 AE_TYPE are `ANY', `MUL', and `MAC'.
12931 `-mae=ANY' selects a completely generic AE type. Code generated
12932 with this option will run on any of the other AE types. The code
12933 will not be as efficient as it would be if compiled for a specific
12934 AE type, and some types of operation (e.g., multiplication) will
12935 not work properly on all types of AE.
12937 `-mae=MUL' selects a MUL AE type. This is the most useful AE type
12938 for compiled code, and is the default.
12940 `-mae=MAC' selects a DSP-style MAC AE. Code compiled with this
12941 option may suffer from poor performance of byte (char)
12942 manipulation, since the DSP AE does not provide hardware support
12943 for byte load/stores.
12945 `-msymbol-as-address'
12946 Enable the compiler to directly use a symbol name as an address in
12947 a load/store instruction, without first loading it into a
12948 register. Typically, the use of this option will generate larger
12949 programs, which run faster than when the option isn't used.
12950 However, the results vary from program to program, so it is left
12951 as a user option, rather than being permanently enabled.
12953 `-mno-inefficient-warnings'
12954 Disables warnings about the generation of inefficient code. These
12955 warnings can be generated, for example, when compiling code which
12956 performs byte-level memory operations on the MAC AE type. The MAC
12957 AE has no hardware support for byte-level memory operations, so
12958 all byte load/stores must be synthesized from word load/store
12959 operations. This is inefficient and a warning will be generated
12960 indicating to the programmer that they should rewrite the code to
12961 avoid byte operations, or to target an AE type which has the
12962 necessary hardware support. This option enables the warning to be
12967 File: gcc.info, Node: PowerPC Options, Next: RS/6000 and PowerPC Options, Prev: picoChip Options, Up: Submodel Options
12969 3.17.27 PowerPC Options
12970 -----------------------
12972 These are listed under *Note RS/6000 and PowerPC Options::.
12975 File: gcc.info, Node: RS/6000 and PowerPC Options, Next: S/390 and zSeries Options, Prev: PowerPC Options, Up: Submodel Options
12977 3.17.28 IBM RS/6000 and PowerPC Options
12978 ---------------------------------------
12980 These `-m' options are defined for the IBM RS/6000 and PowerPC:
12988 `-mno-powerpc-gpopt'
12990 `-mno-powerpc-gfxopt'
13005 GCC supports two related instruction set architectures for the
13006 RS/6000 and PowerPC. The "POWER" instruction set are those
13007 instructions supported by the `rios' chip set used in the original
13008 RS/6000 systems and the "PowerPC" instruction set is the
13009 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx
13010 microprocessors, and the IBM 4xx, 6xx, and follow-on
13013 Neither architecture is a subset of the other. However there is a
13014 large common subset of instructions supported by both. An MQ
13015 register is included in processors supporting the POWER
13018 You use these options to specify which instructions are available
13019 on the processor you are using. The default value of these
13020 options is determined when configuring GCC. Specifying the
13021 `-mcpu=CPU_TYPE' overrides the specification of these options. We
13022 recommend you use the `-mcpu=CPU_TYPE' option rather than the
13023 options listed above.
13025 The `-mpower' option allows GCC to generate instructions that are
13026 found only in the POWER architecture and to use the MQ register.
13027 Specifying `-mpower2' implies `-power' and also allows GCC to
13028 generate instructions that are present in the POWER2 architecture
13029 but not the original POWER architecture.
13031 The `-mpowerpc' option allows GCC to generate instructions that
13032 are found only in the 32-bit subset of the PowerPC architecture.
13033 Specifying `-mpowerpc-gpopt' implies `-mpowerpc' and also allows
13034 GCC to use the optional PowerPC architecture instructions in the
13035 General Purpose group, including floating-point square root.
13036 Specifying `-mpowerpc-gfxopt' implies `-mpowerpc' and also allows
13037 GCC to use the optional PowerPC architecture instructions in the
13038 Graphics group, including floating-point select.
13040 The `-mmfcrf' option allows GCC to generate the move from
13041 condition register field instruction implemented on the POWER4
13042 processor and other processors that support the PowerPC V2.01
13043 architecture. The `-mpopcntb' option allows GCC to generate the
13044 popcount and double precision FP reciprocal estimate instruction
13045 implemented on the POWER5 processor and other processors that
13046 support the PowerPC V2.02 architecture. The `-mfprnd' option
13047 allows GCC to generate the FP round to integer instructions
13048 implemented on the POWER5+ processor and other processors that
13049 support the PowerPC V2.03 architecture. The `-mcmpb' option
13050 allows GCC to generate the compare bytes instruction implemented
13051 on the POWER6 processor and other processors that support the
13052 PowerPC V2.05 architecture. The `-mmfpgpr' option allows GCC to
13053 generate the FP move to/from general purpose register instructions
13054 implemented on the POWER6X processor and other processors that
13055 support the extended PowerPC V2.05 architecture. The `-mhard-dfp'
13056 option allows GCC to generate the decimal floating point
13057 instructions implemented on some POWER processors.
13059 The `-mpowerpc64' option allows GCC to generate the additional
13060 64-bit instructions that are found in the full PowerPC64
13061 architecture and to treat GPRs as 64-bit, doubleword quantities.
13062 GCC defaults to `-mno-powerpc64'.
13064 If you specify both `-mno-power' and `-mno-powerpc', GCC will use
13065 only the instructions in the common subset of both architectures
13066 plus some special AIX common-mode calls, and will not use the MQ
13067 register. Specifying both `-mpower' and `-mpowerpc' permits GCC
13068 to use any instruction from either architecture and to allow use
13069 of the MQ register; specify this for the Motorola MPC601.
13073 Select which mnemonics to use in the generated assembler code.
13074 With `-mnew-mnemonics', GCC uses the assembler mnemonics defined
13075 for the PowerPC architecture. With `-mold-mnemonics' it uses the
13076 assembler mnemonics defined for the POWER architecture.
13077 Instructions defined in only one architecture have only one
13078 mnemonic; GCC uses that mnemonic irrespective of which of these
13079 options is specified.
13081 GCC defaults to the mnemonics appropriate for the architecture in
13082 use. Specifying `-mcpu=CPU_TYPE' sometimes overrides the value of
13083 these option. Unless you are building a cross-compiler, you
13084 should normally not specify either `-mnew-mnemonics' or
13085 `-mold-mnemonics', but should instead accept the default.
13088 Set architecture type, register usage, choice of mnemonics, and
13089 instruction scheduling parameters for machine type CPU_TYPE.
13090 Supported values for CPU_TYPE are `401', `403', `405', `405fp',
13091 `440', `440fp', `464', `464fp', `505', `601', `602', `603',
13092 `603e', `604', `604e', `620', `630', `740', `7400', `7450', `750',
13093 `801', `821', `823', `860', `970', `8540', `e300c2', `e300c3',
13094 `e500mc', `ec603e', `G3', `G4', `G5', `power', `power2', `power3',
13095 `power4', `power5', `power5+', `power6', `power6x', `power7'
13096 `common', `powerpc', `powerpc64', `rios', `rios1', `rios2', `rsc',
13099 `-mcpu=common' selects a completely generic processor. Code
13100 generated under this option will run on any POWER or PowerPC
13101 processor. GCC will use only the instructions in the common
13102 subset of both architectures, and will not use the MQ register.
13103 GCC assumes a generic processor model for scheduling purposes.
13105 `-mcpu=power', `-mcpu=power2', `-mcpu=powerpc', and
13106 `-mcpu=powerpc64' specify generic POWER, POWER2, pure 32-bit
13107 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13108 types, with an appropriate, generic processor model assumed for
13109 scheduling purposes.
13111 The other options specify a specific processor. Code generated
13112 under those options will run best on that processor, and may not
13113 run at all on others.
13115 The `-mcpu' options automatically enable or disable the following
13118 -maltivec -mfprnd -mhard-float -mmfcrf -mmultiple
13119 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64
13120 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float
13121 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr
13123 The particular options set for any particular CPU will vary between
13124 compiler versions, depending on what setting seems to produce
13125 optimal code for that CPU; it doesn't necessarily reflect the
13126 actual hardware's capabilities. If you wish to set an individual
13127 option to a particular value, you may specify it after the `-mcpu'
13128 option, like `-mcpu=970 -mno-altivec'.
13130 On AIX, the `-maltivec' and `-mpowerpc64' options are not enabled
13131 or disabled by the `-mcpu' option at present because AIX does not
13132 have full support for these options. You may still enable or
13133 disable them individually if you're sure it'll work in your
13137 Set the instruction scheduling parameters for machine type
13138 CPU_TYPE, but do not set the architecture type, register usage, or
13139 choice of mnemonics, as `-mcpu=CPU_TYPE' would. The same values
13140 for CPU_TYPE are used for `-mtune' as for `-mcpu'. If both are
13141 specified, the code generated will use the architecture,
13142 registers, and mnemonics set by `-mcpu', but the scheduling
13143 parameters set by `-mtune'.
13147 Generate code to compute division as reciprocal estimate and
13148 iterative refinement, creating opportunities for increased
13149 throughput. This feature requires: optional PowerPC Graphics
13150 instruction set for single precision and FRE instruction for
13151 double precision, assuming divides cannot generate user-visible
13152 traps, and the domain values not include Infinities, denormals or
13157 Generate code that uses (does not use) AltiVec instructions, and
13158 also enable the use of built-in functions that allow more direct
13159 access to the AltiVec instruction set. You may also need to set
13160 `-mabi=altivec' to adjust the current ABI with AltiVec ABI
13165 Generate VRSAVE instructions when generating AltiVec code.
13167 `-mgen-cell-microcode'
13168 Generate Cell microcode instructions
13170 `-mwarn-cell-microcode'
13171 Warning when a Cell microcode instruction is going to emitted. An
13172 example of a Cell microcode instruction is a variable shift.
13175 Generate code that allows ld and ld.so to build executables and
13176 shared libraries with non-exec .plt and .got sections. This is a
13177 PowerPC 32-bit SYSV ABI option.
13180 Generate code that uses a BSS .plt section that ld.so fills in, and
13181 requires .plt and .got sections that are both writable and
13182 executable. This is a PowerPC 32-bit SYSV ABI option.
13186 This switch enables or disables the generation of ISEL
13190 This switch has been deprecated. Use `-misel' and `-mno-isel'
13195 This switch enables or disables the generation of SPE simd
13200 This switch enables or disables the generation of PAIRED simd
13204 This option has been deprecated. Use `-mspe' and `-mno-spe'
13207 `-mfloat-gprs=YES/SINGLE/DOUBLE/NO'
13209 This switch enables or disables the generation of floating point
13210 operations on the general purpose registers for architectures that
13213 The argument YES or SINGLE enables the use of single-precision
13214 floating point operations.
13216 The argument DOUBLE enables the use of single and double-precision
13217 floating point operations.
13219 The argument NO disables floating point operations on the general
13222 This option is currently only available on the MPC854x.
13226 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13227 targets (including GNU/Linux). The 32-bit environment sets int,
13228 long and pointer to 32 bits and generates code that runs on any
13229 PowerPC variant. The 64-bit environment sets int to 32 bits and
13230 long and pointer to 64 bits, and generates code for PowerPC64, as
13237 Modify generation of the TOC (Table Of Contents), which is created
13238 for every executable file. The `-mfull-toc' option is selected by
13239 default. In that case, GCC will allocate at least one TOC entry
13240 for each unique non-automatic variable reference in your program.
13241 GCC will also place floating-point constants in the TOC. However,
13242 only 16,384 entries are available in the TOC.
13244 If you receive a linker error message that saying you have
13245 overflowed the available TOC space, you can reduce the amount of
13246 TOC space used with the `-mno-fp-in-toc' and `-mno-sum-in-toc'
13247 options. `-mno-fp-in-toc' prevents GCC from putting floating-point
13248 constants in the TOC and `-mno-sum-in-toc' forces GCC to generate
13249 code to calculate the sum of an address and a constant at run-time
13250 instead of putting that sum into the TOC. You may specify one or
13251 both of these options. Each causes GCC to produce very slightly
13252 slower and larger code at the expense of conserving TOC space.
13254 If you still run out of space in the TOC even when you specify
13255 both of these options, specify `-mminimal-toc' instead. This
13256 option causes GCC to make only one TOC entry for every file. When
13257 you specify this option, GCC will produce code that is slower and
13258 larger but which uses extremely little TOC space. You may wish to
13259 use this option only on files that contain less frequently
13264 Enable 64-bit AIX ABI and calling convention: 64-bit pointers,
13265 64-bit `long' type, and the infrastructure needed to support them.
13266 Specifying `-maix64' implies `-mpowerpc64' and `-mpowerpc', while
13267 `-maix32' disables the 64-bit ABI and implies `-mno-powerpc64'.
13268 GCC defaults to `-maix32'.
13272 Produce code that conforms more closely to IBM XL compiler
13273 semantics when using AIX-compatible ABI. Pass floating-point
13274 arguments to prototyped functions beyond the register save area
13275 (RSA) on the stack in addition to argument FPRs. Do not assume
13276 that most significant double in 128-bit long double value is
13277 properly rounded when comparing values and converting to double.
13278 Use XL symbol names for long double support routines.
13280 The AIX calling convention was extended but not initially
13281 documented to handle an obscure K&R C case of calling a function
13282 that takes the address of its arguments with fewer arguments than
13283 declared. IBM XL compilers access floating point arguments which
13284 do not fit in the RSA from the stack when a subroutine is compiled
13285 without optimization. Because always storing floating-point
13286 arguments on the stack is inefficient and rarely needed, this
13287 option is not enabled by default and only is necessary when
13288 calling subroutines compiled by IBM XL compilers without
13292 Support "IBM RS/6000 SP" "Parallel Environment" (PE). Link an
13293 application written to use message passing with special startup
13294 code to enable the application to run. The system must have PE
13295 installed in the standard location (`/usr/lpp/ppe.poe/'), or the
13296 `specs' file must be overridden with the `-specs=' option to
13297 specify the appropriate directory location. The Parallel
13298 Environment does not support threads, so the `-mpe' option and the
13299 `-pthread' option are incompatible.
13303 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13304 `-malign-natural' overrides the ABI-defined alignment of larger
13305 types, such as floating-point doubles, on their natural size-based
13306 boundary. The option `-malign-power' instructs GCC to follow the
13307 ABI-specified alignment rules. GCC defaults to the standard
13308 alignment defined in the ABI.
13310 On 64-bit Darwin, natural alignment is the default, and
13311 `-malign-power' is not supported.
13315 Generate code that does not use (uses) the floating-point register
13316 set. Software floating point emulation is provided if you use the
13317 `-msoft-float' option, and pass the option to GCC when linking.
13321 Generate code for single or double-precision floating point
13322 operations. `-mdouble-float' implies `-msingle-float'.
13325 Do not generate sqrt and div instructions for hardware floating
13329 Specify type of floating point unit. Valid values are SP_LITE
13330 (equivalent to -msingle-float -msimple-fpu), DP_LITE (equivalent
13331 to -mdouble-float -msimple-fpu), SP_FULL (equivalent to
13332 -msingle-float), and DP_FULL (equivalent to -mdouble-float).
13335 Perform optimizations for floating point unit on Xilinx PPC
13340 Generate code that uses (does not use) the load multiple word
13341 instructions and the store multiple word instructions. These
13342 instructions are generated by default on POWER systems, and not
13343 generated on PowerPC systems. Do not use `-mmultiple' on little
13344 endian PowerPC systems, since those instructions do not work when
13345 the processor is in little endian mode. The exceptions are PPC740
13346 and PPC750 which permit the instructions usage in little endian
13351 Generate code that uses (does not use) the load string instructions
13352 and the store string word instructions to save multiple registers
13353 and do small block moves. These instructions are generated by
13354 default on POWER systems, and not generated on PowerPC systems.
13355 Do not use `-mstring' on little endian PowerPC systems, since those
13356 instructions do not work when the processor is in little endian
13357 mode. The exceptions are PPC740 and PPC750 which permit the
13358 instructions usage in little endian mode.
13362 Generate code that uses (does not use) the load or store
13363 instructions that update the base register to the address of the
13364 calculated memory location. These instructions are generated by
13365 default. If you use `-mno-update', there is a small window
13366 between the time that the stack pointer is updated and the address
13367 of the previous frame is stored, which means code that walks the
13368 stack frame across interrupts or signals may get corrupted data.
13370 `-mavoid-indexed-addresses'
13372 `-mno-avoid-indexed-addresses'
13373 Generate code that tries to avoid (not avoid) the use of indexed
13374 load or store instructions. These instructions can incur a
13375 performance penalty on Power6 processors in certain situations,
13376 such as when stepping through large arrays that cross a 16M
13377 boundary. This option is enabled by default when targetting
13378 Power6 and disabled otherwise.
13382 Generate code that uses (does not use) the floating point multiply
13383 and accumulate instructions. These instructions are generated by
13384 default if hardware floating is used.
13388 Generate code that uses (does not use) the half-word multiply and
13389 multiply-accumulate instructions on the IBM 405, 440 and 464
13390 processors. These instructions are generated by default when
13391 targetting those processors.
13395 Generate code that uses (does not use) the string-search `dlmzb'
13396 instruction on the IBM 405, 440 and 464 processors. This
13397 instruction is generated by default when targetting those
13402 On System V.4 and embedded PowerPC systems do not (do) force
13403 structures and unions that contain bit-fields to be aligned to the
13404 base type of the bit-field.
13406 For example, by default a structure containing nothing but 8
13407 `unsigned' bit-fields of length 1 would be aligned to a 4 byte
13408 boundary and have a size of 4 bytes. By using `-mno-bit-align',
13409 the structure would be aligned to a 1 byte boundary and be one
13412 `-mno-strict-align'
13414 On System V.4 and embedded PowerPC systems do not (do) assume that
13415 unaligned memory references will be handled by the system.
13419 On embedded PowerPC systems generate code that allows (does not
13420 allow) the program to be relocated to a different address at
13421 runtime. If you use `-mrelocatable' on any module, all objects
13422 linked together must be compiled with `-mrelocatable' or
13423 `-mrelocatable-lib'.
13425 `-mrelocatable-lib'
13426 `-mno-relocatable-lib'
13427 On embedded PowerPC systems generate code that allows (does not
13428 allow) the program to be relocated to a different address at
13429 runtime. Modules compiled with `-mrelocatable-lib' can be linked
13430 with either modules compiled without `-mrelocatable' and
13431 `-mrelocatable-lib' or with modules compiled with the
13432 `-mrelocatable' options.
13436 On System V.4 and embedded PowerPC systems do not (do) assume that
13437 register 2 contains a pointer to a global area pointing to the
13438 addresses used in the program.
13442 On System V.4 and embedded PowerPC systems compile code for the
13443 processor in little endian mode. The `-mlittle-endian' option is
13444 the same as `-mlittle'.
13448 On System V.4 and embedded PowerPC systems compile code for the
13449 processor in big endian mode. The `-mbig-endian' option is the
13453 On Darwin and Mac OS X systems, compile code so that it is not
13454 relocatable, but that its external references are relocatable. The
13455 resulting code is suitable for applications, but not shared
13458 `-mprioritize-restricted-insns=PRIORITY'
13459 This option controls the priority that is assigned to
13460 dispatch-slot restricted instructions during the second scheduling
13461 pass. The argument PRIORITY takes the value 0/1/2 to assign
13462 NO/HIGHEST/SECOND-HIGHEST priority to dispatch slot restricted
13465 `-msched-costly-dep=DEPENDENCE_TYPE'
13466 This option controls which dependences are considered costly by
13467 the target during instruction scheduling. The argument
13468 DEPENDENCE_TYPE takes one of the following values: NO: no
13469 dependence is costly, ALL: all dependences are costly,
13470 TRUE_STORE_TO_LOAD: a true dependence from store to load is costly,
13471 STORE_TO_LOAD: any dependence from store to load is costly,
13472 NUMBER: any dependence which latency >= NUMBER is costly.
13474 `-minsert-sched-nops=SCHEME'
13475 This option controls which nop insertion scheme will be used during
13476 the second scheduling pass. The argument SCHEME takes one of the
13477 following values: NO: Don't insert nops. PAD: Pad with nops any
13478 dispatch group which has vacant issue slots, according to the
13479 scheduler's grouping. REGROUP_EXACT: Insert nops to force costly
13480 dependent insns into separate groups. Insert exactly as many nops
13481 as needed to force an insn to a new group, according to the
13482 estimated processor grouping. NUMBER: Insert nops to force costly
13483 dependent insns into separate groups. Insert NUMBER nops to force
13484 an insn to a new group.
13487 On System V.4 and embedded PowerPC systems compile code using
13488 calling conventions that adheres to the March 1995 draft of the
13489 System V Application Binary Interface, PowerPC processor
13490 supplement. This is the default unless you configured GCC using
13491 `powerpc-*-eabiaix'.
13494 Specify both `-mcall-sysv' and `-meabi' options.
13496 `-mcall-sysv-noeabi'
13497 Specify both `-mcall-sysv' and `-mno-eabi' options.
13500 On System V.4 and embedded PowerPC systems compile code for the
13501 Solaris operating system.
13504 On System V.4 and embedded PowerPC systems compile code for the
13505 Linux-based GNU system.
13508 On System V.4 and embedded PowerPC systems compile code for the
13509 Hurd-based GNU system.
13512 On System V.4 and embedded PowerPC systems compile code for the
13513 NetBSD operating system.
13515 `-maix-struct-return'
13516 Return all structures in memory (as specified by the AIX ABI).
13518 `-msvr4-struct-return'
13519 Return structures smaller than 8 bytes in registers (as specified
13523 Extend the current ABI with a particular extension, or remove such
13524 extension. Valid values are ALTIVEC, NO-ALTIVEC, SPE, NO-SPE,
13525 IBMLONGDOUBLE, IEEELONGDOUBLE.
13528 Extend the current ABI with SPE ABI extensions. This does not
13529 change the default ABI, instead it adds the SPE ABI extensions to
13533 Disable Booke SPE ABI extensions for the current ABI.
13535 `-mabi=ibmlongdouble'
13536 Change the current ABI to use IBM extended precision long double.
13537 This is a PowerPC 32-bit SYSV ABI option.
13539 `-mabi=ieeelongdouble'
13540 Change the current ABI to use IEEE extended precision long double.
13541 This is a PowerPC 32-bit Linux ABI option.
13545 On System V.4 and embedded PowerPC systems assume that all calls to
13546 variable argument functions are properly prototyped. Otherwise,
13547 the compiler must insert an instruction before every non
13548 prototyped call to set or clear bit 6 of the condition code
13549 register (CR) to indicate whether floating point values were
13550 passed in the floating point registers in case the function takes
13551 a variable arguments. With `-mprototype', only calls to
13552 prototyped variable argument functions will set or clear the bit.
13555 On embedded PowerPC systems, assume that the startup module is
13556 called `sim-crt0.o' and that the standard C libraries are
13557 `libsim.a' and `libc.a'. This is the default for
13558 `powerpc-*-eabisim' configurations.
13561 On embedded PowerPC systems, assume that the startup module is
13562 called `crt0.o' and the standard C libraries are `libmvme.a' and
13566 On embedded PowerPC systems, assume that the startup module is
13567 called `crt0.o' and the standard C libraries are `libads.a' and
13571 On embedded PowerPC systems, assume that the startup module is
13572 called `crt0.o' and the standard C libraries are `libyk.a' and
13576 On System V.4 and embedded PowerPC systems, specify that you are
13577 compiling for a VxWorks system.
13580 On embedded PowerPC systems, set the PPC_EMB bit in the ELF flags
13581 header to indicate that `eabi' extended relocations are used.
13585 On System V.4 and embedded PowerPC systems do (do not) adhere to
13586 the Embedded Applications Binary Interface (eabi) which is a set of
13587 modifications to the System V.4 specifications. Selecting `-meabi'
13588 means that the stack is aligned to an 8 byte boundary, a function
13589 `__eabi' is called to from `main' to set up the eabi environment,
13590 and the `-msdata' option can use both `r2' and `r13' to point to
13591 two separate small data areas. Selecting `-mno-eabi' means that
13592 the stack is aligned to a 16 byte boundary, do not call an
13593 initialization function from `main', and the `-msdata' option will
13594 only use `r13' to point to a single small data area. The `-meabi'
13595 option is on by default if you configured GCC using one of the
13596 `powerpc*-*-eabi*' options.
13599 On System V.4 and embedded PowerPC systems, put small initialized
13600 `const' global and static data in the `.sdata2' section, which is
13601 pointed to by register `r2'. Put small initialized non-`const'
13602 global and static data in the `.sdata' section, which is pointed
13603 to by register `r13'. Put small uninitialized global and static
13604 data in the `.sbss' section, which is adjacent to the `.sdata'
13605 section. The `-msdata=eabi' option is incompatible with the
13606 `-mrelocatable' option. The `-msdata=eabi' option also sets the
13610 On System V.4 and embedded PowerPC systems, put small global and
13611 static data in the `.sdata' section, which is pointed to by
13612 register `r13'. Put small uninitialized global and static data in
13613 the `.sbss' section, which is adjacent to the `.sdata' section.
13614 The `-msdata=sysv' option is incompatible with the `-mrelocatable'
13619 On System V.4 and embedded PowerPC systems, if `-meabi' is used,
13620 compile code the same as `-msdata=eabi', otherwise compile code the
13621 same as `-msdata=sysv'.
13624 On System V.4 and embedded PowerPC systems, put small global data
13625 in the `.sdata' section. Put small uninitialized global data in
13626 the `.sbss' section. Do not use register `r13' to address small
13627 data however. This is the default behavior unless other `-msdata'
13632 On embedded PowerPC systems, put all initialized global and static
13633 data in the `.data' section, and all uninitialized data in the
13637 On embedded PowerPC systems, put global and static items less than
13638 or equal to NUM bytes into the small data or bss sections instead
13639 of the normal data or bss section. By default, NUM is 8. The `-G
13640 NUM' switch is also passed to the linker. All modules should be
13641 compiled with the same `-G NUM' value.
13645 On System V.4 and embedded PowerPC systems do (do not) emit
13646 register names in the assembly language output using symbolic
13651 By default assume that all calls are far away so that a longer more
13652 expensive calling sequence is required. This is required for calls
13653 further than 32 megabytes (33,554,432 bytes) from the current
13654 location. A short call will be generated if the compiler knows
13655 the call cannot be that far away. This setting can be overridden
13656 by the `shortcall' function attribute, or by `#pragma longcall(0)'.
13658 Some linkers are capable of detecting out-of-range calls and
13659 generating glue code on the fly. On these systems, long calls are
13660 unnecessary and generate slower code. As of this writing, the AIX
13661 linker can do this, as can the GNU linker for PowerPC/64. It is
13662 planned to add this feature to the GNU linker for 32-bit PowerPC
13665 On Darwin/PPC systems, `#pragma longcall' will generate "jbsr
13666 callee, L42", plus a "branch island" (glue code). The two target
13667 addresses represent the callee and the "branch island". The
13668 Darwin/PPC linker will prefer the first address and generate a "bl
13669 callee" if the PPC "bl" instruction will reach the callee directly;
13670 otherwise, the linker will generate "bl L42" to call the "branch
13671 island". The "branch island" is appended to the body of the
13672 calling function; it computes the full 32-bit address of the callee
13675 On Mach-O (Darwin) systems, this option directs the compiler emit
13676 to the glue for every direct call, and the Darwin linker decides
13677 whether to use or discard it.
13679 In the future, we may cause GCC to ignore all longcall
13680 specifications when the linker is known to generate glue.
13683 Adds support for multithreading with the "pthreads" library. This
13684 option sets flags for both the preprocessor and linker.
13688 File: gcc.info, Node: S/390 and zSeries Options, Next: Score Options, Prev: RS/6000 and PowerPC Options, Up: Submodel Options
13690 3.17.29 S/390 and zSeries Options
13691 ---------------------------------
13693 These are the `-m' options defined for the S/390 and zSeries
13698 Use (do not use) the hardware floating-point instructions and
13699 registers for floating-point operations. When `-msoft-float' is
13700 specified, functions in `libgcc.a' will be used to perform
13701 floating-point operations. When `-mhard-float' is specified, the
13702 compiler generates IEEE floating-point instructions. This is the
13707 Use (do not use) the hardware decimal-floating-point instructions
13708 for decimal-floating-point operations. When `-mno-hard-dfp' is
13709 specified, functions in `libgcc.a' will be used to perform
13710 decimal-floating-point operations. When `-mhard-dfp' is
13711 specified, the compiler generates decimal-floating-point hardware
13712 instructions. This is the default for `-march=z9-ec' or higher.
13715 `-mlong-double-128'
13716 These switches control the size of `long double' type. A size of
13717 64bit makes the `long double' type equivalent to the `double'
13718 type. This is the default.
13722 Store (do not store) the address of the caller's frame as
13723 backchain pointer into the callee's stack frame. A backchain may
13724 be needed to allow debugging using tools that do not understand
13725 DWARF-2 call frame information. When `-mno-packed-stack' is in
13726 effect, the backchain pointer is stored at the bottom of the stack
13727 frame; when `-mpacked-stack' is in effect, the backchain is placed
13728 into the topmost word of the 96/160 byte register save area.
13730 In general, code compiled with `-mbackchain' is call-compatible
13731 with code compiled with `-mmo-backchain'; however, use of the
13732 backchain for debugging purposes usually requires that the whole
13733 binary is built with `-mbackchain'. Note that the combination of
13734 `-mbackchain', `-mpacked-stack' and `-mhard-float' is not
13735 supported. In order to build a linux kernel use `-msoft-float'.
13737 The default is to not maintain the backchain.
13740 `-mno-packed-stack'
13741 Use (do not use) the packed stack layout. When
13742 `-mno-packed-stack' is specified, the compiler uses the all fields
13743 of the 96/160 byte register save area only for their default
13744 purpose; unused fields still take up stack space. When
13745 `-mpacked-stack' is specified, register save slots are densely
13746 packed at the top of the register save area; unused space is
13747 reused for other purposes, allowing for more efficient use of the
13748 available stack space. However, when `-mbackchain' is also in
13749 effect, the topmost word of the save area is always used to store
13750 the backchain, and the return address register is always saved two
13751 words below the backchain.
13753 As long as the stack frame backchain is not used, code generated
13754 with `-mpacked-stack' is call-compatible with code generated with
13755 `-mno-packed-stack'. Note that some non-FSF releases of GCC 2.95
13756 for S/390 or zSeries generated code that uses the stack frame
13757 backchain at run time, not just for debugging purposes. Such code
13758 is not call-compatible with code compiled with `-mpacked-stack'.
13759 Also, note that the combination of `-mbackchain', `-mpacked-stack'
13760 and `-mhard-float' is not supported. In order to build a linux
13761 kernel use `-msoft-float'.
13763 The default is to not use the packed stack layout.
13767 Generate (or do not generate) code using the `bras' instruction to
13768 do subroutine calls. This only works reliably if the total
13769 executable size does not exceed 64k. The default is to use the
13770 `basr' instruction instead, which does not have this limitation.
13774 When `-m31' is specified, generate code compliant to the GNU/Linux
13775 for S/390 ABI. When `-m64' is specified, generate code compliant
13776 to the GNU/Linux for zSeries ABI. This allows GCC in particular
13777 to generate 64-bit instructions. For the `s390' targets, the
13778 default is `-m31', while the `s390x' targets default to `-m64'.
13782 When `-mzarch' is specified, generate code using the instructions
13783 available on z/Architecture. When `-mesa' is specified, generate
13784 code using the instructions available on ESA/390. Note that
13785 `-mesa' is not possible with `-m64'. When generating code
13786 compliant to the GNU/Linux for S/390 ABI, the default is `-mesa'.
13787 When generating code compliant to the GNU/Linux for zSeries ABI,
13788 the default is `-mzarch'.
13792 Generate (or do not generate) code using the `mvcle' instruction
13793 to perform block moves. When `-mno-mvcle' is specified, use a
13794 `mvc' loop instead. This is the default unless optimizing for
13799 Print (or do not print) additional debug information when
13800 compiling. The default is to not print debug information.
13803 Generate code that will run on CPU-TYPE, which is the name of a
13804 system representing a certain processor type. Possible values for
13805 CPU-TYPE are `g5', `g6', `z900', `z990', `z9-109', `z9-ec' and
13806 `z10'. When generating code using the instructions available on
13807 z/Architecture, the default is `-march=z900'. Otherwise, the
13808 default is `-march=g5'.
13811 Tune to CPU-TYPE everything applicable about the generated code,
13812 except for the ABI and the set of available instructions. The
13813 list of CPU-TYPE values is the same as for `-march'. The default
13814 is the value used for `-march'.
13818 Generate code that adds (does not add) in TPF OS specific branches
13819 to trace routines in the operating system. This option is off by
13820 default, even when compiling for the TPF OS.
13824 Generate code that uses (does not use) the floating point multiply
13825 and accumulate instructions. These instructions are generated by
13826 default if hardware floating point is used.
13828 `-mwarn-framesize=FRAMESIZE'
13829 Emit a warning if the current function exceeds the given frame
13830 size. Because this is a compile time check it doesn't need to be
13831 a real problem when the program runs. It is intended to identify
13832 functions which most probably cause a stack overflow. It is
13833 useful to be used in an environment with limited stack size e.g.
13836 `-mwarn-dynamicstack'
13837 Emit a warning if the function calls alloca or uses dynamically
13838 sized arrays. This is generally a bad idea with a limited stack
13841 `-mstack-guard=STACK-GUARD'
13842 `-mstack-size=STACK-SIZE'
13843 If these options are provided the s390 back end emits additional
13844 instructions in the function prologue which trigger a trap if the
13845 stack size is STACK-GUARD bytes above the STACK-SIZE (remember
13846 that the stack on s390 grows downward). If the STACK-GUARD option
13847 is omitted the smallest power of 2 larger than the frame size of
13848 the compiled function is chosen. These options are intended to be
13849 used to help debugging stack overflow problems. The additionally
13850 emitted code causes only little overhead and hence can also be
13851 used in production like systems without greater performance
13852 degradation. The given values have to be exact powers of 2 and
13853 STACK-SIZE has to be greater than STACK-GUARD without exceeding
13854 64k. In order to be efficient the extra code makes the assumption
13855 that the stack starts at an address aligned to the value given by
13856 STACK-SIZE. The STACK-GUARD option can only be used in
13857 conjunction with STACK-SIZE.
13860 File: gcc.info, Node: Score Options, Next: SH Options, Prev: S/390 and zSeries Options, Up: Submodel Options
13862 3.17.30 Score Options
13863 ---------------------
13865 These options are defined for Score implementations:
13868 Compile code for big endian mode. This is the default.
13871 Compile code for little endian mode.
13874 Disable generate bcnz instruction.
13877 Enable generate unaligned load and store instruction.
13880 Enable the use of multiply-accumulate instructions. Disabled by
13884 Specify the SCORE5 as the target architecture.
13887 Specify the SCORE5U of the target architecture.
13890 Specify the SCORE7 as the target architecture. This is the default.
13893 Specify the SCORE7D as the target architecture.
13896 File: gcc.info, Node: SH Options, Next: SPARC Options, Prev: Score Options, Up: Submodel Options
13901 These `-m' options are defined for the SH implementations:
13904 Generate code for the SH1.
13907 Generate code for the SH2.
13910 Generate code for the SH2e.
13913 Generate code for the SH3.
13916 Generate code for the SH3e.
13919 Generate code for the SH4 without a floating-point unit.
13922 Generate code for the SH4 with a floating-point unit that only
13923 supports single-precision arithmetic.
13926 Generate code for the SH4 assuming the floating-point unit is in
13927 single-precision mode by default.
13930 Generate code for the SH4.
13933 Generate code for the SH4al-dsp, or for a SH4a in such a way that
13934 the floating-point unit is not used.
13937 Generate code for the SH4a, in such a way that no double-precision
13938 floating point operations are used.
13941 Generate code for the SH4a assuming the floating-point unit is in
13942 single-precision mode by default.
13945 Generate code for the SH4a.
13948 Same as `-m4a-nofpu', except that it implicitly passes `-dsp' to
13949 the assembler. GCC doesn't generate any DSP instructions at the
13953 Compile code for the processor in big endian mode.
13956 Compile code for the processor in little endian mode.
13959 Align doubles at 64-bit boundaries. Note that this changes the
13960 calling conventions, and thus some functions from the standard C
13961 library will not work unless you recompile it first with
13965 Shorten some address references at link time, when possible; uses
13966 the linker option `-relax'.
13969 Use 32-bit offsets in `switch' tables. The default is to use
13973 Enable the use of bit manipulation instructions on SH2A.
13976 Enable the use of the instruction `fmovd'.
13979 Comply with the calling conventions defined by Renesas.
13982 Comply with the calling conventions defined by Renesas.
13985 Comply with the calling conventions defined for GCC before the
13986 Renesas conventions were available. This option is the default
13987 for all targets of the SH toolchain except for `sh-symbianelf'.
13990 Mark the `MAC' register as call-clobbered, even if `-mhitachi' is
13994 Increase IEEE-compliance of floating-point code. At the moment,
13995 this is equivalent to `-fno-finite-math-only'. When generating 16
13996 bit SH opcodes, getting IEEE-conforming results for comparisons of
13997 NANs / infinities incurs extra overhead in every floating point
13998 comparison, therefore the default is set to `-ffinite-math-only'.
14000 `-minline-ic_invalidate'
14001 Inline code to invalidate instruction cache entries after setting
14002 up nested function trampolines. This option has no effect if
14003 -musermode is in effect and the selected code generation option
14004 (e.g. -m4) does not allow the use of the icbi instruction. If the
14005 selected code generation option does not allow the use of the icbi
14006 instruction, and -musermode is not in effect, the inlined code will
14007 manipulate the instruction cache address array directly with an
14008 associative write. This not only requires privileged mode, but it
14009 will also fail if the cache line had been mapped via the TLB and
14010 has become unmapped.
14013 Dump instruction size and location in the assembly code.
14016 This option is deprecated. It pads structures to multiple of 4
14017 bytes, which is incompatible with the SH ABI.
14020 Optimize for space instead of speed. Implied by `-Os'.
14023 When generating position-independent code, emit function calls
14024 using the Global Offset Table instead of the Procedure Linkage
14028 Don't generate privileged mode only code; implies
14029 -mno-inline-ic_invalidate if the inlined code would not work in
14030 user mode. This is the default when the target is `sh-*-linux*'.
14033 Set the cost to assume for a multiply insn.
14036 Set the division strategy to use for SHmedia code. STRATEGY must
14037 be one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l,
14038 inv:call, inv:call2, inv:fp . "fp" performs the operation in
14039 floating point. This has a very high latency, but needs only a
14040 few instructions, so it might be a good choice if your code has
14041 enough easily exploitable ILP to allow the compiler to schedule
14042 the floating point instructions together with other instructions.
14043 Division by zero causes a floating point exception. "inv" uses
14044 integer operations to calculate the inverse of the divisor, and
14045 then multiplies the dividend with the inverse. This strategy
14046 allows cse and hoisting of the inverse calculation. Division by
14047 zero calculates an unspecified result, but does not trap.
14048 "inv:minlat" is a variant of "inv" where if no cse / hoisting
14049 opportunities have been found, or if the entire operation has been
14050 hoisted to the same place, the last stages of the inverse
14051 calculation are intertwined with the final multiply to reduce the
14052 overall latency, at the expense of using a few more instructions,
14053 and thus offering fewer scheduling opportunities with other code.
14054 "call" calls a library function that usually implements the
14055 inv:minlat strategy. This gives high code density for
14056 m5-*media-nofpu compilations. "call2" uses a different entry
14057 point of the same library function, where it assumes that a
14058 pointer to a lookup table has already been set up, which exposes
14059 the pointer load to cse / code hoisting optimizations.
14060 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm
14061 for initial code generation, but if the code stays unoptimized,
14062 revert to the "call", "call2", or "fp" strategies, respectively.
14063 Note that the potentially-trapping side effect of division by zero
14064 is carried by a separate instruction, so it is possible that all
14065 the integer instructions are hoisted out, but the marker for the
14066 side effect stays where it is. A recombination to fp operations
14067 or a call is not possible in that case. "inv20u" and "inv20l" are
14068 variants of the "inv:minlat" strategy. In the case that the
14069 inverse calculation was nor separated from the multiply, they speed
14070 up division where the dividend fits into 20 bits (plus sign where
14071 applicable), by inserting a test to skip a number of operations in
14072 this case; this test slows down the case of larger dividends.
14073 inv20u assumes the case of a such a small dividend to be unlikely,
14074 and inv20l assumes it to be likely.
14076 `-mdivsi3_libfunc=NAME'
14077 Set the name of the library function used for 32 bit signed
14078 division to NAME. This only affect the name used in the call and
14079 inv:call division strategies, and the compiler will still expect
14080 the same sets of input/output/clobbered registers as if this
14081 option was not present.
14083 `-mfixed-range=REGISTER-RANGE'
14084 Generate code treating the given register range as fixed registers.
14085 A fixed register is one that the register allocator can not use.
14086 This is useful when compiling kernel code. A register range is
14087 specified as two registers separated by a dash. Multiple register
14088 ranges can be specified separated by a comma.
14091 Throttle unrolling to avoid thrashing target registers. This
14092 option only has an effect if the gcc code base supports the
14093 TARGET_ADJUST_UNROLL_MAX target hook.
14095 `-mindexed-addressing'
14096 Enable the use of the indexed addressing mode for
14097 SHmedia32/SHcompact. This is only safe if the hardware and/or OS
14098 implement 32 bit wrap-around semantics for the indexed addressing
14099 mode. The architecture allows the implementation of processors
14100 with 64 bit MMU, which the OS could use to get 32 bit addressing,
14101 but since no current hardware implementation supports this or any
14102 other way to make the indexed addressing mode safe to use in the
14103 32 bit ABI, the default is -mno-indexed-addressing.
14105 `-mgettrcost=NUMBER'
14106 Set the cost assumed for the gettr instruction to NUMBER. The
14107 default is 2 if `-mpt-fixed' is in effect, 100 otherwise.
14110 Assume pt* instructions won't trap. This will generally generate
14111 better scheduled code, but is unsafe on current hardware. The
14112 current architecture definition says that ptabs and ptrel trap
14113 when the target anded with 3 is 3. This has the unintentional
14114 effect of making it unsafe to schedule ptabs / ptrel before a
14115 branch, or hoist it out of a loop. For example,
14116 __do_global_ctors, a part of libgcc that runs constructors at
14117 program startup, calls functions in a list which is delimited by
14118 -1. With the -mpt-fixed option, the ptabs will be done before
14119 testing against -1. That means that all the constructors will be
14120 run a bit quicker, but when the loop comes to the end of the list,
14121 the program crashes because ptabs loads -1 into a target register.
14122 Since this option is unsafe for any hardware implementing the
14123 current architecture specification, the default is -mno-pt-fixed.
14124 Unless the user specifies a specific cost with `-mgettrcost',
14125 -mno-pt-fixed also implies `-mgettrcost=100'; this deters register
14126 allocation using target registers for storing ordinary integers.
14128 `-minvalid-symbols'
14129 Assume symbols might be invalid. Ordinary function symbols
14130 generated by the compiler will always be valid to load with
14131 movi/shori/ptabs or movi/shori/ptrel, but with assembler and/or
14132 linker tricks it is possible to generate symbols that will cause
14133 ptabs / ptrel to trap. This option is only meaningful when
14134 `-mno-pt-fixed' is in effect. It will then prevent
14135 cross-basic-block cse, hoisting and most scheduling of symbol
14136 loads. The default is `-mno-invalid-symbols'.
14139 File: gcc.info, Node: SPARC Options, Next: SPU Options, Prev: SH Options, Up: Submodel Options
14141 3.17.32 SPARC Options
14142 ---------------------
14144 These `-m' options are supported on the SPARC:
14148 Specify `-mapp-regs' to generate output using the global registers
14149 2 through 4, which the SPARC SVR4 ABI reserves for applications.
14150 This is the default.
14152 To be fully SVR4 ABI compliant at the cost of some performance
14153 loss, specify `-mno-app-regs'. You should compile libraries and
14154 system software with this option.
14158 Generate output containing floating point instructions. This is
14163 Generate output containing library calls for floating point.
14164 *Warning:* the requisite libraries are not available for all SPARC
14165 targets. Normally the facilities of the machine's usual C
14166 compiler are used, but this cannot be done directly in
14167 cross-compilation. You must make your own arrangements to provide
14168 suitable library functions for cross-compilation. The embedded
14169 targets `sparc-*-aout' and `sparclite-*-*' do provide software
14170 floating point support.
14172 `-msoft-float' changes the calling convention in the output file;
14173 therefore, it is only useful if you compile _all_ of a program with
14174 this option. In particular, you need to compile `libgcc.a', the
14175 library that comes with GCC, with `-msoft-float' in order for this
14178 `-mhard-quad-float'
14179 Generate output containing quad-word (long double) floating point
14182 `-msoft-quad-float'
14183 Generate output containing library calls for quad-word (long
14184 double) floating point instructions. The functions called are
14185 those specified in the SPARC ABI. This is the default.
14187 As of this writing, there are no SPARC implementations that have
14188 hardware support for the quad-word floating point instructions.
14189 They all invoke a trap handler for one of these instructions, and
14190 then the trap handler emulates the effect of the instruction.
14191 Because of the trap handler overhead, this is much slower than
14192 calling the ABI library routines. Thus the `-msoft-quad-float'
14193 option is the default.
14195 `-mno-unaligned-doubles'
14196 `-munaligned-doubles'
14197 Assume that doubles have 8 byte alignment. This is the default.
14199 With `-munaligned-doubles', GCC assumes that doubles have 8 byte
14200 alignment only if they are contained in another type, or if they
14201 have an absolute address. Otherwise, it assumes they have 4 byte
14202 alignment. Specifying this option avoids some rare compatibility
14203 problems with code generated by other compilers. It is not the
14204 default because it results in a performance loss, especially for
14205 floating point code.
14207 `-mno-faster-structs'
14209 With `-mfaster-structs', the compiler assumes that structures
14210 should have 8 byte alignment. This enables the use of pairs of
14211 `ldd' and `std' instructions for copies in structure assignment,
14212 in place of twice as many `ld' and `st' pairs. However, the use
14213 of this changed alignment directly violates the SPARC ABI. Thus,
14214 it's intended only for use on targets where the developer
14215 acknowledges that their resulting code will not be directly in
14216 line with the rules of the ABI.
14219 `-mimpure-text', used in addition to `-shared', tells the compiler
14220 to not pass `-z text' to the linker when linking a shared object.
14221 Using this option, you can link position-dependent code into a
14224 `-mimpure-text' suppresses the "relocations remain against
14225 allocatable but non-writable sections" linker error message.
14226 However, the necessary relocations will trigger copy-on-write, and
14227 the shared object is not actually shared across processes.
14228 Instead of using `-mimpure-text', you should compile all source
14229 code with `-fpic' or `-fPIC'.
14231 This option is only available on SunOS and Solaris.
14234 Set the instruction set, register set, and instruction scheduling
14235 parameters for machine type CPU_TYPE. Supported values for
14236 CPU_TYPE are `v7', `cypress', `v8', `supersparc', `sparclite',
14237 `f930', `f934', `hypersparc', `sparclite86x', `sparclet',
14238 `tsc701', `v9', `ultrasparc', `ultrasparc3', `niagara' and
14241 Default instruction scheduling parameters are used for values that
14242 select an architecture and not an implementation. These are `v7',
14243 `v8', `sparclite', `sparclet', `v9'.
14245 Here is a list of each supported architecture and their supported
14249 v8: supersparc, hypersparc
14250 sparclite: f930, f934, sparclite86x
14252 v9: ultrasparc, ultrasparc3, niagara, niagara2
14254 By default (unless configured otherwise), GCC generates code for
14255 the V7 variant of the SPARC architecture. With `-mcpu=cypress',
14256 the compiler additionally optimizes it for the Cypress CY7C602
14257 chip, as used in the SPARCStation/SPARCServer 3xx series. This is
14258 also appropriate for the older SPARCStation 1, 2, IPX etc.
14260 With `-mcpu=v8', GCC generates code for the V8 variant of the SPARC
14261 architecture. The only difference from V7 code is that the
14262 compiler emits the integer multiply and integer divide
14263 instructions which exist in SPARC-V8 but not in SPARC-V7. With
14264 `-mcpu=supersparc', the compiler additionally optimizes it for the
14265 SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000
14268 With `-mcpu=sparclite', GCC generates code for the SPARClite
14269 variant of the SPARC architecture. This adds the integer
14270 multiply, integer divide step and scan (`ffs') instructions which
14271 exist in SPARClite but not in SPARC-V7. With `-mcpu=f930', the
14272 compiler additionally optimizes it for the Fujitsu MB86930 chip,
14273 which is the original SPARClite, with no FPU. With `-mcpu=f934',
14274 the compiler additionally optimizes it for the Fujitsu MB86934
14275 chip, which is the more recent SPARClite with FPU.
14277 With `-mcpu=sparclet', GCC generates code for the SPARClet variant
14278 of the SPARC architecture. This adds the integer multiply,
14279 multiply/accumulate, integer divide step and scan (`ffs')
14280 instructions which exist in SPARClet but not in SPARC-V7. With
14281 `-mcpu=tsc701', the compiler additionally optimizes it for the
14282 TEMIC SPARClet chip.
14284 With `-mcpu=v9', GCC generates code for the V9 variant of the SPARC
14285 architecture. This adds 64-bit integer and floating-point move
14286 instructions, 3 additional floating-point condition code registers
14287 and conditional move instructions. With `-mcpu=ultrasparc', the
14288 compiler additionally optimizes it for the Sun UltraSPARC I/II/IIi
14289 chips. With `-mcpu=ultrasparc3', the compiler additionally
14290 optimizes it for the Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+
14291 chips. With `-mcpu=niagara', the compiler additionally optimizes
14292 it for Sun UltraSPARC T1 chips. With `-mcpu=niagara2', the
14293 compiler additionally optimizes it for Sun UltraSPARC T2 chips.
14296 Set the instruction scheduling parameters for machine type
14297 CPU_TYPE, but do not set the instruction set or register set that
14298 the option `-mcpu=CPU_TYPE' would.
14300 The same values for `-mcpu=CPU_TYPE' can be used for
14301 `-mtune=CPU_TYPE', but the only useful values are those that
14302 select a particular cpu implementation. Those are `cypress',
14303 `supersparc', `hypersparc', `f930', `f934', `sparclite86x',
14304 `tsc701', `ultrasparc', `ultrasparc3', `niagara', and `niagara2'.
14308 With `-mv8plus', GCC generates code for the SPARC-V8+ ABI. The
14309 difference from the V8 ABI is that the global and out registers are
14310 considered 64-bit wide. This is enabled by default on Solaris in
14311 32-bit mode for all SPARC-V9 processors.
14315 With `-mvis', GCC generates code that takes advantage of the
14316 UltraSPARC Visual Instruction Set extensions. The default is
14319 These `-m' options are supported in addition to the above on SPARC-V9
14320 processors in 64-bit environments:
14323 Generate code for a processor running in little-endian mode. It
14324 is only available for a few configurations and most notably not on
14329 Generate code for a 32-bit or 64-bit environment. The 32-bit
14330 environment sets int, long and pointer to 32 bits. The 64-bit
14331 environment sets int to 32 bits and long and pointer to 64 bits.
14334 Generate code for the Medium/Low code model: 64-bit addresses,
14335 programs must be linked in the low 32 bits of memory. Programs
14336 can be statically or dynamically linked.
14339 Generate code for the Medium/Middle code model: 64-bit addresses,
14340 programs must be linked in the low 44 bits of memory, the text and
14341 data segments must be less than 2GB in size and the data segment
14342 must be located within 2GB of the text segment.
14345 Generate code for the Medium/Anywhere code model: 64-bit
14346 addresses, programs may be linked anywhere in memory, the text and
14347 data segments must be less than 2GB in size and the data segment
14348 must be located within 2GB of the text segment.
14350 `-mcmodel=embmedany'
14351 Generate code for the Medium/Anywhere code model for embedded
14352 systems: 64-bit addresses, the text and data segments must be less
14353 than 2GB in size, both starting anywhere in memory (determined at
14354 link time). The global register %g4 points to the base of the
14355 data segment. Programs are statically linked and PIC is not
14360 With `-mstack-bias', GCC assumes that the stack pointer, and frame
14361 pointer if present, are offset by -2047 which must be added back
14362 when making stack frame references. This is the default in 64-bit
14363 mode. Otherwise, assume no such offset is present.
14365 These switches are supported in addition to the above on Solaris:
14368 Add support for multithreading using the Solaris threads library.
14369 This option sets flags for both the preprocessor and linker. This
14370 option does not affect the thread safety of object code produced
14371 by the compiler or that of libraries supplied with it.
14374 Add support for multithreading using the POSIX threads library.
14375 This option sets flags for both the preprocessor and linker. This
14376 option does not affect the thread safety of object code produced
14377 by the compiler or that of libraries supplied with it.
14380 This is a synonym for `-pthreads'.
14383 File: gcc.info, Node: SPU Options, Next: System V Options, Prev: SPARC Options, Up: Submodel Options
14385 3.17.33 SPU Options
14386 -------------------
14388 These `-m' options are supported on the SPU:
14392 The loader for SPU does not handle dynamic relocations. By
14393 default, GCC will give an error when it generates code that
14394 requires a dynamic relocation. `-mno-error-reloc' disables the
14395 error, `-mwarn-reloc' will generate a warning instead.
14399 Instructions which initiate or test completion of DMA must not be
14400 reordered with respect to loads and stores of the memory which is
14401 being accessed. Users typically address this problem using the
14402 volatile keyword, but that can lead to inefficient code in places
14403 where the memory is known to not change. Rather than mark the
14404 memory as volatile we treat the DMA instructions as potentially
14405 effecting all memory. With `-munsafe-dma' users must use the
14406 volatile keyword to protect memory accesses.
14409 By default, GCC will generate a branch hint instruction to avoid
14410 pipeline stalls for always taken or probably taken branches. A
14411 hint will not be generated closer than 8 instructions away from
14412 its branch. There is little reason to disable them, except for
14413 debugging purposes, or to make an object a little bit smaller.
14417 By default, GCC generates code assuming that addresses are never
14418 larger than 18 bits. With `-mlarge-mem' code is generated that
14419 assumes a full 32 bit address.
14422 By default, GCC links against startup code that assumes the
14423 SPU-style main function interface (which has an unconventional
14424 parameter list). With `-mstdmain', GCC will link your program
14425 against startup code that assumes a C99-style interface to `main',
14426 including a local copy of `argv' strings.
14428 `-mfixed-range=REGISTER-RANGE'
14429 Generate code treating the given register range as fixed registers.
14430 A fixed register is one that the register allocator can not use.
14431 This is useful when compiling kernel code. A register range is
14432 specified as two registers separated by a dash. Multiple register
14433 ranges can be specified separated by a comma.
14437 By default, GCC will insert nops to increase dual issue when it
14438 expects it to increase performance. N can be a value from 0 to
14439 10. A smaller N will insert fewer nops. 10 is the default, 0 is
14440 the same as `-mno-dual-nops'. Disabled with `-Os'.
14442 `-mhint-max-nops=N'
14443 Maximum number of nops to insert for a branch hint. A branch hint
14444 must be at least 8 instructions away from the branch it is
14445 effecting. GCC will insert up to N nops to enforce this,
14446 otherwise it will not generate the branch hint.
14448 `-mhint-max-distance=N'
14449 The encoding of the branch hint instruction limits the hint to be
14450 within 256 instructions of the branch it is effecting. By
14451 default, GCC makes sure it is within 125.
14454 Work around a hardware bug which causes the SPU to stall
14455 indefinitely. By default, GCC will insert the `hbrp' instruction
14456 to make sure this stall won't happen.
14460 File: gcc.info, Node: System V Options, Next: V850 Options, Prev: SPU Options, Up: Submodel Options
14462 3.17.34 Options for System V
14463 ----------------------------
14465 These additional options are available on System V Release 4 for
14466 compatibility with other compilers on those systems:
14469 Create a shared object. It is recommended that `-symbolic' or
14470 `-shared' be used instead.
14473 Identify the versions of each tool used by the compiler, in a
14474 `.ident' assembler directive in the output.
14477 Refrain from adding `.ident' directives to the output file (this is
14481 Search the directories DIRS, and no others, for libraries
14482 specified with `-l'.
14485 Look in the directory DIR to find the M4 preprocessor. The
14486 assembler uses this option.
14489 File: gcc.info, Node: V850 Options, Next: VAX Options, Prev: System V Options, Up: Submodel Options
14491 3.17.35 V850 Options
14492 --------------------
14494 These `-m' options are defined for V850 implementations:
14498 Treat all calls as being far away (near). If calls are assumed to
14499 be far away, the compiler will always load the functions address
14500 up into a register, and call indirect through the pointer.
14504 Do not optimize (do optimize) basic blocks that use the same index
14505 pointer 4 or more times to copy pointer into the `ep' register, and
14506 use the shorter `sld' and `sst' instructions. The `-mep' option
14507 is on by default if you optimize.
14509 `-mno-prolog-function'
14510 `-mprolog-function'
14511 Do not use (do use) external functions to save and restore
14512 registers at the prologue and epilogue of a function. The
14513 external functions are slower, but use less code space if more
14514 than one function saves the same number of registers. The
14515 `-mprolog-function' option is on by default if you optimize.
14518 Try to make the code as small as possible. At present, this just
14519 turns on the `-mep' and `-mprolog-function' options.
14522 Put static or global variables whose size is N bytes or less into
14523 the tiny data area that register `ep' points to. The tiny data
14524 area can hold up to 256 bytes in total (128 bytes for byte
14528 Put static or global variables whose size is N bytes or less into
14529 the small data area that register `gp' points to. The small data
14530 area can hold up to 64 kilobytes.
14533 Put static or global variables whose size is N bytes or less into
14534 the first 32 kilobytes of memory.
14537 Specify that the target processor is the V850.
14540 Generate code suitable for big switch tables. Use this option
14541 only if the assembler/linker complain about out of range branches
14542 within a switch table.
14545 This option will cause r2 and r5 to be used in the code generated
14546 by the compiler. This setting is the default.
14549 This option will cause r2 and r5 to be treated as fixed registers.
14552 Specify that the target processor is the V850E1. The preprocessor
14553 constants `__v850e1__' and `__v850e__' will be defined if this
14557 Specify that the target processor is the V850E. The preprocessor
14558 constant `__v850e__' will be defined if this option is used.
14560 If neither `-mv850' nor `-mv850e' nor `-mv850e1' are defined then
14561 a default target processor will be chosen and the relevant
14562 `__v850*__' preprocessor constant will be defined.
14564 The preprocessor constants `__v850' and `__v851__' are always
14565 defined, regardless of which processor variant is the target.
14568 This option will suppress generation of the CALLT instruction for
14569 the v850e and v850e1 flavors of the v850 architecture. The
14570 default is `-mno-disable-callt' which allows the CALLT instruction
14575 File: gcc.info, Node: VAX Options, Next: VxWorks Options, Prev: V850 Options, Up: Submodel Options
14577 3.17.36 VAX Options
14578 -------------------
14580 These `-m' options are defined for the VAX:
14583 Do not output certain jump instructions (`aobleq' and so on) that
14584 the Unix assembler for the VAX cannot handle across long ranges.
14587 Do output those jump instructions, on the assumption that you will
14588 assemble with the GNU assembler.
14591 Output code for g-format floating point numbers instead of
14595 File: gcc.info, Node: VxWorks Options, Next: x86-64 Options, Prev: VAX Options, Up: Submodel Options
14597 3.17.37 VxWorks Options
14598 -----------------------
14600 The options in this section are defined for all VxWorks targets.
14601 Options specific to the target hardware are listed with the other
14602 options for that target.
14605 GCC can generate code for both VxWorks kernels and real time
14606 processes (RTPs). This option switches from the former to the
14607 latter. It also defines the preprocessor macro `__RTP__'.
14610 Link an RTP executable against shared libraries rather than static
14611 libraries. The options `-static' and `-shared' can also be used
14612 for RTPs (*note Link Options::); `-static' is the default.
14616 These options are passed down to the linker. They are defined for
14617 compatibility with Diab.
14620 Enable lazy binding of function calls. This option is equivalent
14621 to `-Wl,-z,now' and is defined for compatibility with Diab.
14624 Disable lazy binding of function calls. This option is the
14625 default and is defined for compatibility with Diab.
14628 File: gcc.info, Node: x86-64 Options, Next: Xstormy16 Options, Prev: VxWorks Options, Up: Submodel Options
14630 3.17.38 x86-64 Options
14631 ----------------------
14633 These are listed under *Note i386 and x86-64 Options::.
14636 File: gcc.info, Node: i386 and x86-64 Windows Options, Next: IA-64 Options, Prev: i386 and x86-64 Options, Up: Submodel Options
14638 3.17.39 i386 and x86-64 Windows Options
14639 ---------------------------------------
14641 These additional options are available for Windows targets:
14644 This option is available for Cygwin and MinGW targets. It
14645 specifies that a console application is to be generated, by
14646 instructing the linker to set the PE header subsystem type
14647 required for console applications. This is the default behaviour
14648 for Cygwin and MinGW targets.
14651 This option is available for Cygwin targets. It specifies that
14652 the Cygwin internal interface is to be used for predefined
14653 preprocessor macros, C runtime libraries and related linker paths
14654 and options. For Cygwin targets this is the default behaviour.
14655 This option is deprecated and will be removed in a future release.
14658 This option is available for Cygwin targets. It specifies that
14659 the MinGW internal interface is to be used instead of Cygwin's, by
14660 setting MinGW-related predefined macros and linker paths and
14661 default library options. This option is deprecated and will be
14662 removed in a future release.
14665 This option is available for Cygwin and MinGW targets. It
14666 specifies that a DLL - a dynamic link library - is to be
14667 generated, enabling the selection of the required runtime startup
14668 object and entry point.
14670 `-mnop-fun-dllimport'
14671 This option is available for Cygwin and MinGW targets. It
14672 specifies that the dllimport attribute should be ignored.
14675 This option is available for MinGW targets. It specifies that
14676 MinGW-specific thread support is to be used.
14679 This option is available for Cygwin and MinGW targets. It
14680 specifies that the typical Windows pre-defined macros are to be
14681 set in the pre-processor, but does not influence the choice of
14682 runtime library/startup code.
14685 This option is available for Cygwin and MinGW targets. It
14686 specifies that a GUI application is to be generated by instructing
14687 the linker to set the PE header subsystem type appropriately.
14689 See also under *Note i386 and x86-64 Options:: for standard options.
14692 File: gcc.info, Node: Xstormy16 Options, Next: Xtensa Options, Prev: x86-64 Options, Up: Submodel Options
14694 3.17.40 Xstormy16 Options
14695 -------------------------
14697 These options are defined for Xstormy16:
14700 Choose startup files and linker script suitable for the simulator.
14703 File: gcc.info, Node: Xtensa Options, Next: zSeries Options, Prev: Xstormy16 Options, Up: Submodel Options
14705 3.17.41 Xtensa Options
14706 ----------------------
14708 These options are supported for Xtensa targets:
14712 Enable or disable use of `CONST16' instructions for loading
14713 constant values. The `CONST16' instruction is currently not a
14714 standard option from Tensilica. When enabled, `CONST16'
14715 instructions are always used in place of the standard `L32R'
14716 instructions. The use of `CONST16' is enabled by default only if
14717 the `L32R' instruction is not available.
14721 Enable or disable use of fused multiply/add and multiply/subtract
14722 instructions in the floating-point option. This has no effect if
14723 the floating-point option is not also enabled. Disabling fused
14724 multiply/add and multiply/subtract instructions forces the
14725 compiler to use separate instructions for the multiply and
14726 add/subtract operations. This may be desirable in some cases
14727 where strict IEEE 754-compliant results are required: the fused
14728 multiply add/subtract instructions do not round the intermediate
14729 result, thereby producing results with _more_ bits of precision
14730 than specified by the IEEE standard. Disabling fused multiply
14731 add/subtract instructions also ensures that the program output is
14732 not sensitive to the compiler's ability to combine multiply and
14733 add/subtract operations.
14735 `-mserialize-volatile'
14736 `-mno-serialize-volatile'
14737 When this option is enabled, GCC inserts `MEMW' instructions before
14738 `volatile' memory references to guarantee sequential consistency.
14739 The default is `-mserialize-volatile'. Use
14740 `-mno-serialize-volatile' to omit the `MEMW' instructions.
14742 `-mtext-section-literals'
14743 `-mno-text-section-literals'
14744 Control the treatment of literal pools. The default is
14745 `-mno-text-section-literals', which places literals in a separate
14746 section in the output file. This allows the literal pool to be
14747 placed in a data RAM/ROM, and it also allows the linker to combine
14748 literal pools from separate object files to remove redundant
14749 literals and improve code size. With `-mtext-section-literals',
14750 the literals are interspersed in the text section in order to keep
14751 them as close as possible to their references. This may be
14752 necessary for large assembly files.
14755 `-mno-target-align'
14756 When this option is enabled, GCC instructs the assembler to
14757 automatically align instructions to reduce branch penalties at the
14758 expense of some code density. The assembler attempts to widen
14759 density instructions to align branch targets and the instructions
14760 following call instructions. If there are not enough preceding
14761 safe density instructions to align a target, no widening will be
14762 performed. The default is `-mtarget-align'. These options do not
14763 affect the treatment of auto-aligned instructions like `LOOP',
14764 which the assembler will always align, either by widening density
14765 instructions or by inserting no-op instructions.
14769 When this option is enabled, GCC instructs the assembler to
14770 translate direct calls to indirect calls unless it can determine
14771 that the target of a direct call is in the range allowed by the
14772 call instruction. This translation typically occurs for calls to
14773 functions in other source files. Specifically, the assembler
14774 translates a direct `CALL' instruction into an `L32R' followed by
14775 a `CALLX' instruction. The default is `-mno-longcalls'. This
14776 option should be used in programs where the call target can
14777 potentially be out of range. This option is implemented in the
14778 assembler, not the compiler, so the assembly code generated by GCC
14779 will still show direct call instructions--look at the disassembled
14780 object code to see the actual instructions. Note that the
14781 assembler will use an indirect call for every cross-file call, not
14782 just those that really will be out of range.
14785 File: gcc.info, Node: zSeries Options, Prev: Xtensa Options, Up: Submodel Options
14787 3.17.42 zSeries Options
14788 -----------------------
14790 These are listed under *Note S/390 and zSeries Options::.
14793 File: gcc.info, Node: Code Gen Options, Next: Environment Variables, Prev: Submodel Options, Up: Invoking GCC
14795 3.18 Options for Code Generation Conventions
14796 ============================================
14798 These machine-independent options control the interface conventions
14799 used in code generation.
14801 Most of them have both positive and negative forms; the negative form
14802 of `-ffoo' would be `-fno-foo'. In the table below, only one of the
14803 forms is listed--the one which is not the default. You can figure out
14804 the other form by either removing `no-' or adding it.
14807 For front-ends that support it, generate additional code to check
14808 that indices used to access arrays are within the declared range.
14809 This is currently only supported by the Java and Fortran
14810 front-ends, where this option defaults to true and false
14814 This option generates traps for signed overflow on addition,
14815 subtraction, multiplication operations.
14818 This option instructs the compiler to assume that signed arithmetic
14819 overflow of addition, subtraction and multiplication wraps around
14820 using twos-complement representation. This flag enables some
14821 optimizations and disables others. This option is enabled by
14822 default for the Java front-end, as required by the Java language
14826 Enable exception handling. Generates extra code needed to
14827 propagate exceptions. For some targets, this implies GCC will
14828 generate frame unwind information for all functions, which can
14829 produce significant data size overhead, although it does not
14830 affect execution. If you do not specify this option, GCC will
14831 enable it by default for languages like C++ which normally require
14832 exception handling, and disable it for languages like C that do
14833 not normally require it. However, you may need to enable this
14834 option when compiling C code that needs to interoperate properly
14835 with exception handlers written in C++. You may also wish to
14836 disable this option if you are compiling older C++ programs that
14837 don't use exception handling.
14839 `-fnon-call-exceptions'
14840 Generate code that allows trapping instructions to throw
14841 exceptions. Note that this requires platform-specific runtime
14842 support that does not exist everywhere. Moreover, it only allows
14843 _trapping_ instructions to throw exceptions, i.e. memory
14844 references or floating point instructions. It does not allow
14845 exceptions to be thrown from arbitrary signal handlers such as
14849 Similar to `-fexceptions', except that it will just generate any
14850 needed static data, but will not affect the generated code in any
14851 other way. You will normally not enable this option; instead, a
14852 language processor that needs this handling would enable it on
14855 `-fasynchronous-unwind-tables'
14856 Generate unwind table in dwarf2 format, if supported by target
14857 machine. The table is exact at each instruction boundary, so it
14858 can be used for stack unwinding from asynchronous events (such as
14859 debugger or garbage collector).
14861 `-fpcc-struct-return'
14862 Return "short" `struct' and `union' values in memory like longer
14863 ones, rather than in registers. This convention is less
14864 efficient, but it has the advantage of allowing intercallability
14865 between GCC-compiled files and files compiled with other
14866 compilers, particularly the Portable C Compiler (pcc).
14868 The precise convention for returning structures in memory depends
14869 on the target configuration macros.
14871 Short structures and unions are those whose size and alignment
14872 match that of some integer type.
14874 *Warning:* code compiled with the `-fpcc-struct-return' switch is
14875 not binary compatible with code compiled with the
14876 `-freg-struct-return' switch. Use it to conform to a non-default
14877 application binary interface.
14879 `-freg-struct-return'
14880 Return `struct' and `union' values in registers when possible.
14881 This is more efficient for small structures than
14882 `-fpcc-struct-return'.
14884 If you specify neither `-fpcc-struct-return' nor
14885 `-freg-struct-return', GCC defaults to whichever convention is
14886 standard for the target. If there is no standard convention, GCC
14887 defaults to `-fpcc-struct-return', except on targets where GCC is
14888 the principal compiler. In those cases, we can choose the
14889 standard, and we chose the more efficient register return
14892 *Warning:* code compiled with the `-freg-struct-return' switch is
14893 not binary compatible with code compiled with the
14894 `-fpcc-struct-return' switch. Use it to conform to a non-default
14895 application binary interface.
14898 Allocate to an `enum' type only as many bytes as it needs for the
14899 declared range of possible values. Specifically, the `enum' type
14900 will be equivalent to the smallest integer type which has enough
14903 *Warning:* the `-fshort-enums' switch causes GCC to generate code
14904 that is not binary compatible with code generated without that
14905 switch. Use it to conform to a non-default application binary
14909 Use the same size for `double' as for `float'.
14911 *Warning:* the `-fshort-double' switch causes GCC to generate code
14912 that is not binary compatible with code generated without that
14913 switch. Use it to conform to a non-default application binary
14917 Override the underlying type for `wchar_t' to be `short unsigned
14918 int' instead of the default for the target. This option is useful
14919 for building programs to run under WINE.
14921 *Warning:* the `-fshort-wchar' switch causes GCC to generate code
14922 that is not binary compatible with code generated without that
14923 switch. Use it to conform to a non-default application binary
14927 In C code, controls the placement of uninitialized global
14928 variables. Unix C compilers have traditionally permitted multiple
14929 definitions of such variables in different compilation units by
14930 placing the variables in a common block. This is the behavior
14931 specified by `-fcommon', and is the default for GCC on most
14932 targets. On the other hand, this behavior is not required by ISO
14933 C, and on some targets may carry a speed or code size penalty on
14934 variable references. The `-fno-common' option specifies that the
14935 compiler should place uninitialized global variables in the data
14936 section of the object file, rather than generating them as common
14937 blocks. This has the effect that if the same variable is declared
14938 (without `extern') in two different compilations, you will get a
14939 multiple-definition error when you link them. In this case, you
14940 must compile with `-fcommon' instead. Compiling with
14941 `-fno-common' is useful on targets for which it provides better
14942 performance, or if you wish to verify that the program will work
14943 on other systems which always treat uninitialized variable
14944 declarations this way.
14947 Ignore the `#ident' directive.
14949 `-finhibit-size-directive'
14950 Don't output a `.size' assembler directive, or anything else that
14951 would cause trouble if the function is split in the middle, and the
14952 two halves are placed at locations far apart in memory. This
14953 option is used when compiling `crtstuff.c'; you should not need to
14954 use it for anything else.
14957 Put extra commentary information in the generated assembly code to
14958 make it more readable. This option is generally only of use to
14959 those who actually need to read the generated assembly code
14960 (perhaps while debugging the compiler itself).
14962 `-fno-verbose-asm', the default, causes the extra information to
14963 be omitted and is useful when comparing two assembler files.
14965 `-frecord-gcc-switches'
14966 This switch causes the command line that was used to invoke the
14967 compiler to be recorded into the object file that is being created.
14968 This switch is only implemented on some targets and the exact
14969 format of the recording is target and binary file format
14970 dependent, but it usually takes the form of a section containing
14971 ASCII text. This switch is related to the `-fverbose-asm' switch,
14972 but that switch only records information in the assembler output
14973 file as comments, so it never reaches the object file.
14976 Generate position-independent code (PIC) suitable for use in a
14977 shared library, if supported for the target machine. Such code
14978 accesses all constant addresses through a global offset table
14979 (GOT). The dynamic loader resolves the GOT entries when the
14980 program starts (the dynamic loader is not part of GCC; it is part
14981 of the operating system). If the GOT size for the linked
14982 executable exceeds a machine-specific maximum size, you get an
14983 error message from the linker indicating that `-fpic' does not
14984 work; in that case, recompile with `-fPIC' instead. (These
14985 maximums are 8k on the SPARC and 32k on the m68k and RS/6000. The
14986 386 has no such limit.)
14988 Position-independent code requires special support, and therefore
14989 works only on certain machines. For the 386, GCC supports PIC for
14990 System V but not for the Sun 386i. Code generated for the IBM
14991 RS/6000 is always position-independent.
14993 When this flag is set, the macros `__pic__' and `__PIC__' are
14997 If supported for the target machine, emit position-independent
14998 code, suitable for dynamic linking and avoiding any limit on the
14999 size of the global offset table. This option makes a difference
15000 on the m68k, PowerPC and SPARC.
15002 Position-independent code requires special support, and therefore
15003 works only on certain machines.
15005 When this flag is set, the macros `__pic__' and `__PIC__' are
15010 These options are similar to `-fpic' and `-fPIC', but generated
15011 position independent code can be only linked into executables.
15012 Usually these options are used when `-pie' GCC option will be used
15015 `-fpie' and `-fPIE' both define the macros `__pie__' and
15016 `__PIE__'. The macros have the value 1 for `-fpie' and 2 for
15020 Do not use jump tables for switch statements even where it would be
15021 more efficient than other code generation strategies. This option
15022 is of use in conjunction with `-fpic' or `-fPIC' for building code
15023 which forms part of a dynamic linker and cannot reference the
15024 address of a jump table. On some targets, jump tables do not
15025 require a GOT and this option is not needed.
15028 Treat the register named REG as a fixed register; generated code
15029 should never refer to it (except perhaps as a stack pointer, frame
15030 pointer or in some other fixed role).
15032 REG must be the name of a register. The register names accepted
15033 are machine-specific and are defined in the `REGISTER_NAMES' macro
15034 in the machine description macro file.
15036 This flag does not have a negative form, because it specifies a
15040 Treat the register named REG as an allocable register that is
15041 clobbered by function calls. It may be allocated for temporaries
15042 or variables that do not live across a call. Functions compiled
15043 this way will not save and restore the register REG.
15045 It is an error to used this flag with the frame pointer or stack
15046 pointer. Use of this flag for other registers that have fixed
15047 pervasive roles in the machine's execution model will produce
15048 disastrous results.
15050 This flag does not have a negative form, because it specifies a
15054 Treat the register named REG as an allocable register saved by
15055 functions. It may be allocated even for temporaries or variables
15056 that live across a call. Functions compiled this way will save
15057 and restore the register REG if they use it.
15059 It is an error to used this flag with the frame pointer or stack
15060 pointer. Use of this flag for other registers that have fixed
15061 pervasive roles in the machine's execution model will produce
15062 disastrous results.
15064 A different sort of disaster will result from the use of this flag
15065 for a register in which function values may be returned.
15067 This flag does not have a negative form, because it specifies a
15070 `-fpack-struct[=N]'
15071 Without a value specified, pack all structure members together
15072 without holes. When a value is specified (which must be a small
15073 power of two), pack structure members according to this value,
15074 representing the maximum alignment (that is, objects with default
15075 alignment requirements larger than this will be output potentially
15076 unaligned at the next fitting location.
15078 *Warning:* the `-fpack-struct' switch causes GCC to generate code
15079 that is not binary compatible with code generated without that
15080 switch. Additionally, it makes the code suboptimal. Use it to
15081 conform to a non-default application binary interface.
15083 `-finstrument-functions'
15084 Generate instrumentation calls for entry and exit to functions.
15085 Just after function entry and just before function exit, the
15086 following profiling functions will be called with the address of
15087 the current function and its call site. (On some platforms,
15088 `__builtin_return_address' does not work beyond the current
15089 function, so the call site information may not be available to the
15090 profiling functions otherwise.)
15092 void __cyg_profile_func_enter (void *this_fn,
15094 void __cyg_profile_func_exit (void *this_fn,
15097 The first argument is the address of the start of the current
15098 function, which may be looked up exactly in the symbol table.
15100 This instrumentation is also done for functions expanded inline in
15101 other functions. The profiling calls will indicate where,
15102 conceptually, the inline function is entered and exited. This
15103 means that addressable versions of such functions must be
15104 available. If all your uses of a function are expanded inline,
15105 this may mean an additional expansion of code size. If you use
15106 `extern inline' in your C code, an addressable version of such
15107 functions must be provided. (This is normally the case anyways,
15108 but if you get lucky and the optimizer always expands the
15109 functions inline, you might have gotten away without providing
15112 A function may be given the attribute `no_instrument_function', in
15113 which case this instrumentation will not be done. This can be
15114 used, for example, for the profiling functions listed above,
15115 high-priority interrupt routines, and any functions from which the
15116 profiling functions cannot safely be called (perhaps signal
15117 handlers, if the profiling routines generate output or allocate
15120 `-finstrument-functions-exclude-file-list=FILE,FILE,...'
15121 Set the list of functions that are excluded from instrumentation
15122 (see the description of `-finstrument-functions'). If the file
15123 that contains a function definition matches with one of FILE, then
15124 that function is not instrumented. The match is done on
15125 substrings: if the FILE parameter is a substring of the file name,
15126 it is considered to be a match.
15129 `-finstrument-functions-exclude-file-list=/bits/stl,include/sys'
15130 will exclude any inline function defined in files whose pathnames
15131 contain `/bits/stl' or `include/sys'.
15133 If, for some reason, you want to include letter `','' in one of
15134 SYM, write `'\,''. For example,
15135 `-finstrument-functions-exclude-file-list='\,\,tmp'' (note the
15136 single quote surrounding the option).
15138 `-finstrument-functions-exclude-function-list=SYM,SYM,...'
15139 This is similar to `-finstrument-functions-exclude-file-list', but
15140 this option sets the list of function names to be excluded from
15141 instrumentation. The function name to be matched is its
15142 user-visible name, such as `vector<int> blah(const vector<int>
15143 &)', not the internal mangled name (e.g.,
15144 `_Z4blahRSt6vectorIiSaIiEE'). The match is done on substrings: if
15145 the SYM parameter is a substring of the function name, it is
15146 considered to be a match.
15149 Generate code to verify that you do not go beyond the boundary of
15150 the stack. You should specify this flag if you are running in an
15151 environment with multiple threads, but only rarely need to specify
15152 it in a single-threaded environment since stack overflow is
15153 automatically detected on nearly all systems if there is only one
15156 Note that this switch does not actually cause checking to be done;
15157 the operating system or the language runtime must do that. The
15158 switch causes generation of code to ensure that they see the stack
15161 You can additionally specify a string parameter: `no' means no
15162 checking, `generic' means force the use of old-style checking,
15163 `specific' means use the best checking method and is equivalent to
15164 bare `-fstack-check'.
15166 Old-style checking is a generic mechanism that requires no specific
15167 target support in the compiler but comes with the following
15170 1. Modified allocation strategy for large objects: they will
15171 always be allocated dynamically if their size exceeds a fixed
15174 2. Fixed limit on the size of the static frame of functions:
15175 when it is topped by a particular function, stack checking is
15176 not reliable and a warning is issued by the compiler.
15178 3. Inefficiency: because of both the modified allocation
15179 strategy and the generic implementation, the performances of
15180 the code are hampered.
15182 Note that old-style stack checking is also the fallback method for
15183 `specific' if no target support has been added in the compiler.
15185 `-fstack-limit-register=REG'
15186 `-fstack-limit-symbol=SYM'
15188 Generate code to ensure that the stack does not grow beyond a
15189 certain value, either the value of a register or the address of a
15190 symbol. If the stack would grow beyond the value, a signal is
15191 raised. For most targets, the signal is raised before the stack
15192 overruns the boundary, so it is possible to catch the signal
15193 without taking special precautions.
15195 For instance, if the stack starts at absolute address `0x80000000'
15196 and grows downwards, you can use the flags
15197 `-fstack-limit-symbol=__stack_limit' and
15198 `-Wl,--defsym,__stack_limit=0x7ffe0000' to enforce a stack limit
15199 of 128KB. Note that this may only work with the GNU linker.
15202 `-fargument-noalias'
15203 `-fargument-noalias-global'
15204 `-fargument-noalias-anything'
15205 Specify the possible relationships among parameters and between
15206 parameters and global data.
15208 `-fargument-alias' specifies that arguments (parameters) may alias
15209 each other and may alias global storage.
15210 `-fargument-noalias' specifies that arguments do not alias each
15211 other, but may alias global storage.
15212 `-fargument-noalias-global' specifies that arguments do not alias
15213 each other and do not alias global storage.
15214 `-fargument-noalias-anything' specifies that arguments do not
15215 alias any other storage.
15217 Each language will automatically use whatever option is required by
15218 the language standard. You should not need to use these options
15221 `-fleading-underscore'
15222 This option and its counterpart, `-fno-leading-underscore',
15223 forcibly change the way C symbols are represented in the object
15224 file. One use is to help link with legacy assembly code.
15226 *Warning:* the `-fleading-underscore' switch causes GCC to
15227 generate code that is not binary compatible with code generated
15228 without that switch. Use it to conform to a non-default
15229 application binary interface. Not all targets provide complete
15230 support for this switch.
15232 `-ftls-model=MODEL'
15233 Alter the thread-local storage model to be used (*note
15234 Thread-Local::). The MODEL argument should be one of
15235 `global-dynamic', `local-dynamic', `initial-exec' or `local-exec'.
15237 The default without `-fpic' is `initial-exec'; with `-fpic' the
15238 default is `global-dynamic'.
15240 `-fvisibility=DEFAULT|INTERNAL|HIDDEN|PROTECTED'
15241 Set the default ELF image symbol visibility to the specified
15242 option--all symbols will be marked with this unless overridden
15243 within the code. Using this feature can very substantially
15244 improve linking and load times of shared object libraries, produce
15245 more optimized code, provide near-perfect API export and prevent
15246 symbol clashes. It is *strongly* recommended that you use this in
15247 any shared objects you distribute.
15249 Despite the nomenclature, `default' always means public ie;
15250 available to be linked against from outside the shared object.
15251 `protected' and `internal' are pretty useless in real-world usage
15252 so the only other commonly used option will be `hidden'. The
15253 default if `-fvisibility' isn't specified is `default', i.e., make
15254 every symbol public--this causes the same behavior as previous
15257 A good explanation of the benefits offered by ensuring ELF symbols
15258 have the correct visibility is given by "How To Write Shared
15259 Libraries" by Ulrich Drepper (which can be found at
15260 `http://people.redhat.com/~drepper/')--however a superior solution
15261 made possible by this option to marking things hidden when the
15262 default is public is to make the default hidden and mark things
15263 public. This is the norm with DLL's on Windows and with
15264 `-fvisibility=hidden' and `__attribute__
15265 ((visibility("default")))' instead of `__declspec(dllexport)' you
15266 get almost identical semantics with identical syntax. This is a
15267 great boon to those working with cross-platform projects.
15269 For those adding visibility support to existing code, you may find
15270 `#pragma GCC visibility' of use. This works by you enclosing the
15271 declarations you wish to set visibility for with (for example)
15272 `#pragma GCC visibility push(hidden)' and `#pragma GCC visibility
15273 pop'. Bear in mind that symbol visibility should be viewed *as
15274 part of the API interface contract* and thus all new code should
15275 always specify visibility when it is not the default ie;
15276 declarations only for use within the local DSO should *always* be
15277 marked explicitly as hidden as so to avoid PLT indirection
15278 overheads--making this abundantly clear also aids readability and
15279 self-documentation of the code. Note that due to ISO C++
15280 specification requirements, operator new and operator delete must
15281 always be of default visibility.
15283 Be aware that headers from outside your project, in particular
15284 system headers and headers from any other library you use, may not
15285 be expecting to be compiled with visibility other than the
15286 default. You may need to explicitly say `#pragma GCC visibility
15287 push(default)' before including any such headers.
15289 `extern' declarations are not affected by `-fvisibility', so a lot
15290 of code can be recompiled with `-fvisibility=hidden' with no
15291 modifications. However, this means that calls to `extern'
15292 functions with no explicit visibility will use the PLT, so it is
15293 more effective to use `__attribute ((visibility))' and/or `#pragma
15294 GCC visibility' to tell the compiler which `extern' declarations
15295 should be treated as hidden.
15297 Note that `-fvisibility' does affect C++ vague linkage entities.
15298 This means that, for instance, an exception class that will be
15299 thrown between DSOs must be explicitly marked with default
15300 visibility so that the `type_info' nodes will be unified between
15303 An overview of these techniques, their benefits and how to use them
15304 is at `http://gcc.gnu.org/wiki/Visibility'.
15308 File: gcc.info, Node: Environment Variables, Next: Precompiled Headers, Prev: Code Gen Options, Up: Invoking GCC
15310 3.19 Environment Variables Affecting GCC
15311 ========================================
15313 This section describes several environment variables that affect how GCC
15314 operates. Some of them work by specifying directories or prefixes to
15315 use when searching for various kinds of files. Some are used to
15316 specify other aspects of the compilation environment.
15318 Note that you can also specify places to search using options such as
15319 `-B', `-I' and `-L' (*note Directory Options::). These take precedence
15320 over places specified using environment variables, which in turn take
15321 precedence over those specified by the configuration of GCC. *Note
15322 Controlling the Compilation Driver `gcc': (gccint)Driver.
15328 These environment variables control the way that GCC uses
15329 localization information that allow GCC to work with different
15330 national conventions. GCC inspects the locale categories
15331 `LC_CTYPE' and `LC_MESSAGES' if it has been configured to do so.
15332 These locale categories can be set to any value supported by your
15333 installation. A typical value is `en_GB.UTF-8' for English in the
15334 United Kingdom encoded in UTF-8.
15336 The `LC_CTYPE' environment variable specifies character
15337 classification. GCC uses it to determine the character boundaries
15338 in a string; this is needed for some multibyte encodings that
15339 contain quote and escape characters that would otherwise be
15340 interpreted as a string end or escape.
15342 The `LC_MESSAGES' environment variable specifies the language to
15343 use in diagnostic messages.
15345 If the `LC_ALL' environment variable is set, it overrides the value
15346 of `LC_CTYPE' and `LC_MESSAGES'; otherwise, `LC_CTYPE' and
15347 `LC_MESSAGES' default to the value of the `LANG' environment
15348 variable. If none of these variables are set, GCC defaults to
15349 traditional C English behavior.
15352 If `TMPDIR' is set, it specifies the directory to use for temporary
15353 files. GCC uses temporary files to hold the output of one stage of
15354 compilation which is to be used as input to the next stage: for
15355 example, the output of the preprocessor, which is the input to the
15359 If `GCC_EXEC_PREFIX' is set, it specifies a prefix to use in the
15360 names of the subprograms executed by the compiler. No slash is
15361 added when this prefix is combined with the name of a subprogram,
15362 but you can specify a prefix that ends with a slash if you wish.
15364 If `GCC_EXEC_PREFIX' is not set, GCC will attempt to figure out an
15365 appropriate prefix to use based on the pathname it was invoked
15368 If GCC cannot find the subprogram using the specified prefix, it
15369 tries looking in the usual places for the subprogram.
15371 The default value of `GCC_EXEC_PREFIX' is `PREFIX/lib/gcc/' where
15372 PREFIX is the prefix to the installed compiler. In many cases
15373 PREFIX is the value of `prefix' when you ran the `configure'
15376 Other prefixes specified with `-B' take precedence over this
15379 This prefix is also used for finding files such as `crt0.o' that
15380 are used for linking.
15382 In addition, the prefix is used in an unusual way in finding the
15383 directories to search for header files. For each of the standard
15384 directories whose name normally begins with `/usr/local/lib/gcc'
15385 (more precisely, with the value of `GCC_INCLUDE_DIR'), GCC tries
15386 replacing that beginning with the specified prefix to produce an
15387 alternate directory name. Thus, with `-Bfoo/', GCC will search
15388 `foo/bar' where it would normally search `/usr/local/lib/bar'.
15389 These alternate directories are searched first; the standard
15390 directories come next. If a standard directory begins with the
15391 configured PREFIX then the value of PREFIX is replaced by
15392 `GCC_EXEC_PREFIX' when looking for header files.
15395 The value of `COMPILER_PATH' is a colon-separated list of
15396 directories, much like `PATH'. GCC tries the directories thus
15397 specified when searching for subprograms, if it can't find the
15398 subprograms using `GCC_EXEC_PREFIX'.
15401 The value of `LIBRARY_PATH' is a colon-separated list of
15402 directories, much like `PATH'. When configured as a native
15403 compiler, GCC tries the directories thus specified when searching
15404 for special linker files, if it can't find them using
15405 `GCC_EXEC_PREFIX'. Linking using GCC also uses these directories
15406 when searching for ordinary libraries for the `-l' option (but
15407 directories specified with `-L' come first).
15410 This variable is used to pass locale information to the compiler.
15411 One way in which this information is used is to determine the
15412 character set to be used when character literals, string literals
15413 and comments are parsed in C and C++. When the compiler is
15414 configured to allow multibyte characters, the following values for
15415 `LANG' are recognized:
15418 Recognize JIS characters.
15421 Recognize SJIS characters.
15424 Recognize EUCJP characters.
15426 If `LANG' is not defined, or if it has some other value, then the
15427 compiler will use mblen and mbtowc as defined by the default
15428 locale to recognize and translate multibyte characters.
15430 Some additional environments variables affect the behavior of the
15435 `CPLUS_INCLUDE_PATH'
15436 `OBJC_INCLUDE_PATH'
15437 Each variable's value is a list of directories separated by a
15438 special character, much like `PATH', in which to look for header
15439 files. The special character, `PATH_SEPARATOR', is
15440 target-dependent and determined at GCC build time. For Microsoft
15441 Windows-based targets it is a semicolon, and for almost all other
15442 targets it is a colon.
15444 `CPATH' specifies a list of directories to be searched as if
15445 specified with `-I', but after any paths given with `-I' options
15446 on the command line. This environment variable is used regardless
15447 of which language is being preprocessed.
15449 The remaining environment variables apply only when preprocessing
15450 the particular language indicated. Each specifies a list of
15451 directories to be searched as if specified with `-isystem', but
15452 after any paths given with `-isystem' options on the command line.
15454 In all these variables, an empty element instructs the compiler to
15455 search its current working directory. Empty elements can appear
15456 at the beginning or end of a path. For instance, if the value of
15457 `CPATH' is `:/special/include', that has the same effect as
15458 `-I. -I/special/include'.
15460 `DEPENDENCIES_OUTPUT'
15461 If this variable is set, its value specifies how to output
15462 dependencies for Make based on the non-system header files
15463 processed by the compiler. System header files are ignored in the
15466 The value of `DEPENDENCIES_OUTPUT' can be just a file name, in
15467 which case the Make rules are written to that file, guessing the
15468 target name from the source file name. Or the value can have the
15469 form `FILE TARGET', in which case the rules are written to file
15470 FILE using TARGET as the target name.
15472 In other words, this environment variable is equivalent to
15473 combining the options `-MM' and `-MF' (*note Preprocessor
15474 Options::), with an optional `-MT' switch too.
15476 `SUNPRO_DEPENDENCIES'
15477 This variable is the same as `DEPENDENCIES_OUTPUT' (see above),
15478 except that system header files are not ignored, so it implies
15479 `-M' rather than `-MM'. However, the dependence on the main input
15480 file is omitted. *Note Preprocessor Options::.
15483 File: gcc.info, Node: Precompiled Headers, Next: Running Protoize, Prev: Environment Variables, Up: Invoking GCC
15485 3.20 Using Precompiled Headers
15486 ==============================
15488 Often large projects have many header files that are included in every
15489 source file. The time the compiler takes to process these header files
15490 over and over again can account for nearly all of the time required to
15491 build the project. To make builds faster, GCC allows users to
15492 `precompile' a header file; then, if builds can use the precompiled
15493 header file they will be much faster.
15495 To create a precompiled header file, simply compile it as you would any
15496 other file, if necessary using the `-x' option to make the driver treat
15497 it as a C or C++ header file. You will probably want to use a tool
15498 like `make' to keep the precompiled header up-to-date when the headers
15499 it contains change.
15501 A precompiled header file will be searched for when `#include' is seen
15502 in the compilation. As it searches for the included file (*note Search
15503 Path: (cpp)Search Path.) the compiler looks for a precompiled header in
15504 each directory just before it looks for the include file in that
15505 directory. The name searched for is the name specified in the
15506 `#include' with `.gch' appended. If the precompiled header file can't
15507 be used, it is ignored.
15509 For instance, if you have `#include "all.h"', and you have `all.h.gch'
15510 in the same directory as `all.h', then the precompiled header file will
15511 be used if possible, and the original header will be used otherwise.
15513 Alternatively, you might decide to put the precompiled header file in a
15514 directory and use `-I' to ensure that directory is searched before (or
15515 instead of) the directory containing the original header. Then, if you
15516 want to check that the precompiled header file is always used, you can
15517 put a file of the same name as the original header in this directory
15518 containing an `#error' command.
15520 This also works with `-include'. So yet another way to use
15521 precompiled headers, good for projects not designed with precompiled
15522 header files in mind, is to simply take most of the header files used by
15523 a project, include them from another header file, precompile that header
15524 file, and `-include' the precompiled header. If the header files have
15525 guards against multiple inclusion, they will be skipped because they've
15526 already been included (in the precompiled header).
15528 If you need to precompile the same header file for different
15529 languages, targets, or compiler options, you can instead make a
15530 _directory_ named like `all.h.gch', and put each precompiled header in
15531 the directory, perhaps using `-o'. It doesn't matter what you call the
15532 files in the directory, every precompiled header in the directory will
15533 be considered. The first precompiled header encountered in the
15534 directory that is valid for this compilation will be used; they're
15535 searched in no particular order.
15537 There are many other possibilities, limited only by your imagination,
15538 good sense, and the constraints of your build system.
15540 A precompiled header file can be used only when these conditions apply:
15542 * Only one precompiled header can be used in a particular
15545 * A precompiled header can't be used once the first C token is seen.
15546 You can have preprocessor directives before a precompiled header;
15547 you can even include a precompiled header from inside another
15548 header, so long as there are no C tokens before the `#include'.
15550 * The precompiled header file must be produced for the same language
15551 as the current compilation. You can't use a C precompiled header
15552 for a C++ compilation.
15554 * The precompiled header file must have been produced by the same
15555 compiler binary as the current compilation is using.
15557 * Any macros defined before the precompiled header is included must
15558 either be defined in the same way as when the precompiled header
15559 was generated, or must not affect the precompiled header, which
15560 usually means that they don't appear in the precompiled header at
15563 The `-D' option is one way to define a macro before a precompiled
15564 header is included; using a `#define' can also do it. There are
15565 also some options that define macros implicitly, like `-O' and
15566 `-Wdeprecated'; the same rule applies to macros defined this way.
15568 * If debugging information is output when using the precompiled
15569 header, using `-g' or similar, the same kind of debugging
15570 information must have been output when building the precompiled
15571 header. However, a precompiled header built using `-g' can be
15572 used in a compilation when no debugging information is being
15575 * The same `-m' options must generally be used when building and
15576 using the precompiled header. *Note Submodel Options::, for any
15577 cases where this rule is relaxed.
15579 * Each of the following options must be the same when building and
15580 using the precompiled header:
15584 * Some other command-line options starting with `-f', `-p', or `-O'
15585 must be defined in the same way as when the precompiled header was
15586 generated. At present, it's not clear which options are safe to
15587 change and which are not; the safest choice is to use exactly the
15588 same options when generating and using the precompiled header.
15589 The following are known to be safe:
15591 -fmessage-length= -fpreprocessed -fsched-interblock
15592 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
15593 -fsched-verbose=<number> -fschedule-insns -fvisibility=
15597 For all of these except the last, the compiler will automatically
15598 ignore the precompiled header if the conditions aren't met. If you
15599 find an option combination that doesn't work and doesn't cause the
15600 precompiled header to be ignored, please consider filing a bug report,
15603 If you do use differing options when generating and using the
15604 precompiled header, the actual behavior will be a mixture of the
15605 behavior for the options. For instance, if you use `-g' to generate
15606 the precompiled header but not when using it, you may or may not get
15607 debugging information for routines in the precompiled header.
15610 File: gcc.info, Node: Running Protoize, Prev: Precompiled Headers, Up: Invoking GCC
15612 3.21 Running Protoize
15613 =====================
15615 The program `protoize' is an optional part of GCC. You can use it to
15616 add prototypes to a program, thus converting the program to ISO C in
15617 one respect. The companion program `unprotoize' does the reverse: it
15618 removes argument types from any prototypes that are found.
15620 When you run these programs, you must specify a set of source files as
15621 command line arguments. The conversion programs start out by compiling
15622 these files to see what functions they define. The information gathered
15623 about a file FOO is saved in a file named `FOO.X'.
15625 After scanning comes actual conversion. The specified files are all
15626 eligible to be converted; any files they include (whether sources or
15627 just headers) are eligible as well.
15629 But not all the eligible files are converted. By default, `protoize'
15630 and `unprotoize' convert only source and header files in the current
15631 directory. You can specify additional directories whose files should
15632 be converted with the `-d DIRECTORY' option. You can also specify
15633 particular files to exclude with the `-x FILE' option. A file is
15634 converted if it is eligible, its directory name matches one of the
15635 specified directory names, and its name within the directory has not
15638 Basic conversion with `protoize' consists of rewriting most function
15639 definitions and function declarations to specify the types of the
15640 arguments. The only ones not rewritten are those for varargs functions.
15642 `protoize' optionally inserts prototype declarations at the beginning
15643 of the source file, to make them available for any calls that precede
15644 the function's definition. Or it can insert prototype declarations
15645 with block scope in the blocks where undeclared functions are called.
15647 Basic conversion with `unprotoize' consists of rewriting most function
15648 declarations to remove any argument types, and rewriting function
15649 definitions to the old-style pre-ISO form.
15651 Both conversion programs print a warning for any function declaration
15652 or definition that they can't convert. You can suppress these warnings
15655 The output from `protoize' or `unprotoize' replaces the original
15656 source file. The original file is renamed to a name ending with
15657 `.save' (for DOS, the saved filename ends in `.sav' without the
15658 original `.c' suffix). If the `.save' (`.sav' for DOS) file already
15659 exists, then the source file is simply discarded.
15661 `protoize' and `unprotoize' both depend on GCC itself to scan the
15662 program and collect information about the functions it uses. So
15663 neither of these programs will work until GCC is installed.
15665 Here is a table of the options you can use with `protoize' and
15666 `unprotoize'. Each option works with both programs unless otherwise
15670 Look for the file `SYSCALLS.c.X' in DIRECTORY, instead of the
15671 usual directory (normally `/usr/local/lib'). This file contains
15672 prototype information about standard system functions. This option
15673 applies only to `protoize'.
15675 `-c COMPILATION-OPTIONS'
15676 Use COMPILATION-OPTIONS as the options when running `gcc' to
15677 produce the `.X' files. The special option `-aux-info' is always
15678 passed in addition, to tell `gcc' to write a `.X' file.
15680 Note that the compilation options must be given as a single
15681 argument to `protoize' or `unprotoize'. If you want to specify
15682 several `gcc' options, you must quote the entire set of
15683 compilation options to make them a single word in the shell.
15685 There are certain `gcc' arguments that you cannot use, because they
15686 would produce the wrong kind of output. These include `-g', `-O',
15687 `-c', `-S', and `-o' If you include these in the
15688 COMPILATION-OPTIONS, they are ignored.
15691 Rename files to end in `.C' (`.cc' for DOS-based file systems)
15692 instead of `.c'. This is convenient if you are converting a C
15693 program to C++. This option applies only to `protoize'.
15696 Add explicit global declarations. This means inserting explicit
15697 declarations at the beginning of each source file for each function
15698 that is called in the file and was not declared. These
15699 declarations precede the first function definition that contains a
15700 call to an undeclared function. This option applies only to
15704 Indent old-style parameter declarations with the string STRING.
15705 This option applies only to `protoize'.
15707 `unprotoize' converts prototyped function definitions to old-style
15708 function definitions, where the arguments are declared between the
15709 argument list and the initial `{'. By default, `unprotoize' uses
15710 five spaces as the indentation. If you want to indent with just
15711 one space instead, use `-i " "'.
15714 Keep the `.X' files. Normally, they are deleted after conversion
15718 Add explicit local declarations. `protoize' with `-l' inserts a
15719 prototype declaration for each function in each block which calls
15720 the function without any declaration. This option applies only to
15724 Make no real changes. This mode just prints information about the
15725 conversions that would have been done without `-n'.
15728 Make no `.save' files. The original files are simply deleted.
15729 Use this option with caution.
15732 Use the program PROGRAM as the compiler. Normally, the name `gcc'
15736 Work quietly. Most warnings are suppressed.
15739 Print the version number, just like `-v' for `gcc'.
15741 If you need special compiler options to compile one of your program's
15742 source files, then you should generate that file's `.X' file specially,
15743 by running `gcc' on that source file with the appropriate options and
15744 the option `-aux-info'. Then run `protoize' on the entire set of
15745 files. `protoize' will use the existing `.X' file because it is newer
15746 than the source file. For example:
15748 gcc -Dfoo=bar file1.c -aux-info file1.X
15751 You need to include the special files along with the rest in the
15752 `protoize' command, even though their `.X' files already exist, because
15753 otherwise they won't get converted.
15755 *Note Protoize Caveats::, for more information on how to use
15756 `protoize' successfully.
15759 File: gcc.info, Node: C Implementation, Next: C Extensions, Prev: Invoking GCC, Up: Top
15761 4 C Implementation-defined behavior
15762 ***********************************
15764 A conforming implementation of ISO C is required to document its choice
15765 of behavior in each of the areas that are designated "implementation
15766 defined". The following lists all such areas, along with the section
15767 numbers from the ISO/IEC 9899:1990 and ISO/IEC 9899:1999 standards.
15768 Some areas are only implementation-defined in one version of the
15771 Some choices depend on the externally determined ABI for the platform
15772 (including standard character encodings) which GCC follows; these are
15773 listed as "determined by ABI" below. *Note Binary Compatibility:
15774 Compatibility, and `http://gcc.gnu.org/readings.html'. Some choices
15775 are documented in the preprocessor manual. *Note
15776 Implementation-defined behavior: (cpp)Implementation-defined behavior.
15777 Some choices are made by the library and operating system (or other
15778 environment when compiling for a freestanding environment); refer to
15779 their documentation for details.
15783 * Translation implementation::
15784 * Environment implementation::
15785 * Identifiers implementation::
15786 * Characters implementation::
15787 * Integers implementation::
15788 * Floating point implementation::
15789 * Arrays and pointers implementation::
15790 * Hints implementation::
15791 * Structures unions enumerations and bit-fields implementation::
15792 * Qualifiers implementation::
15793 * Declarators implementation::
15794 * Statements implementation::
15795 * Preprocessing directives implementation::
15796 * Library functions implementation::
15797 * Architecture implementation::
15798 * Locale-specific behavior implementation::
15801 File: gcc.info, Node: Translation implementation, Next: Environment implementation, Up: C Implementation
15806 * `How a diagnostic is identified (C90 3.7, C99 3.10, C90 and C99
15809 Diagnostics consist of all the output sent to stderr by GCC.
15811 * `Whether each nonempty sequence of white-space characters other
15812 than new-line is retained or replaced by one space character in
15813 translation phase 3 (C90 and C99 5.1.1.2).'
15815 *Note Implementation-defined behavior: (cpp)Implementation-defined
15820 File: gcc.info, Node: Environment implementation, Next: Identifiers implementation, Prev: Translation implementation, Up: C Implementation
15825 The behavior of most of these points are dependent on the implementation
15826 of the C library, and are not defined by GCC itself.
15828 * `The mapping between physical source file multibyte characters and
15829 the source character set in translation phase 1 (C90 and C99
15832 *Note Implementation-defined behavior: (cpp)Implementation-defined
15837 File: gcc.info, Node: Identifiers implementation, Next: Characters implementation, Prev: Environment implementation, Up: C Implementation
15842 * `Which additional multibyte characters may appear in identifiers
15843 and their correspondence to universal character names (C99 6.4.2).'
15845 *Note Implementation-defined behavior: (cpp)Implementation-defined
15848 * `The number of significant initial characters in an identifier
15849 (C90 6.1.2, C90 and C99 5.2.4.1, C99 6.4.2).'
15851 For internal names, all characters are significant. For external
15852 names, the number of significant characters are defined by the
15853 linker; for almost all targets, all characters are significant.
15855 * `Whether case distinctions are significant in an identifier with
15856 external linkage (C90 6.1.2).'
15858 This is a property of the linker. C99 requires that case
15859 distinctions are always significant in identifiers with external
15860 linkage and systems without this property are not supported by GCC.
15864 File: gcc.info, Node: Characters implementation, Next: Integers implementation, Prev: Identifiers implementation, Up: C Implementation
15869 * `The number of bits in a byte (C90 3.4, C99 3.6).'
15873 * `The values of the members of the execution character set (C90 and
15878 * `The unique value of the member of the execution character set
15879 produced for each of the standard alphabetic escape sequences (C90
15884 * `The value of a `char' object into which has been stored any
15885 character other than a member of the basic execution character set
15886 (C90 6.1.2.5, C99 6.2.5).'
15890 * `Which of `signed char' or `unsigned char' has the same range,
15891 representation, and behavior as "plain" `char' (C90 6.1.2.5, C90
15892 6.2.1.1, C99 6.2.5, C99 6.3.1.1).'
15894 Determined by ABI. The options `-funsigned-char' and
15895 `-fsigned-char' change the default. *Note Options Controlling C
15896 Dialect: C Dialect Options.
15898 * `The mapping of members of the source character set (in character
15899 constants and string literals) to members of the execution
15900 character set (C90 6.1.3.4, C99 6.4.4.4, C90 and C99 5.1.1.2).'
15904 * `The value of an integer character constant containing more than
15905 one character or containing a character or escape sequence that
15906 does not map to a single-byte execution character (C90 6.1.3.4,
15909 *Note Implementation-defined behavior: (cpp)Implementation-defined
15912 * `The value of a wide character constant containing more than one
15913 multibyte character, or containing a multibyte character or escape
15914 sequence not represented in the extended execution character set
15915 (C90 6.1.3.4, C99 6.4.4.4).'
15917 *Note Implementation-defined behavior: (cpp)Implementation-defined
15920 * `The current locale used to convert a wide character constant
15921 consisting of a single multibyte character that maps to a member
15922 of the extended execution character set into a corresponding wide
15923 character code (C90 6.1.3.4, C99 6.4.4.4).'
15925 *Note Implementation-defined behavior: (cpp)Implementation-defined
15928 * `The current locale used to convert a wide string literal into
15929 corresponding wide character codes (C90 6.1.4, C99 6.4.5).'
15931 *Note Implementation-defined behavior: (cpp)Implementation-defined
15934 * `The value of a string literal containing a multibyte character or
15935 escape sequence not represented in the execution character set
15936 (C90 6.1.4, C99 6.4.5).'
15938 *Note Implementation-defined behavior: (cpp)Implementation-defined
15942 File: gcc.info, Node: Integers implementation, Next: Floating point implementation, Prev: Characters implementation, Up: C Implementation
15947 * `Any extended integer types that exist in the implementation (C99
15950 GCC does not support any extended integer types.
15952 * `Whether signed integer types are represented using sign and
15953 magnitude, two's complement, or one's complement, and whether the
15954 extraordinary value is a trap representation or an ordinary value
15957 GCC supports only two's complement integer types, and all bit
15958 patterns are ordinary values.
15960 * `The rank of any extended integer type relative to another extended
15961 integer type with the same precision (C99 6.3.1.1).'
15963 GCC does not support any extended integer types.
15965 * `The result of, or the signal raised by, converting an integer to a
15966 signed integer type when the value cannot be represented in an
15967 object of that type (C90 6.2.1.2, C99 6.3.1.3).'
15969 For conversion to a type of width N, the value is reduced modulo
15970 2^N to be within range of the type; no signal is raised.
15972 * `The results of some bitwise operations on signed integers (C90
15975 Bitwise operators act on the representation of the value including
15976 both the sign and value bits, where the sign bit is considered
15977 immediately above the highest-value value bit. Signed `>>' acts
15978 on negative numbers by sign extension.
15980 GCC does not use the latitude given in C99 only to treat certain
15981 aspects of signed `<<' as undefined, but this is subject to change.
15983 * `The sign of the remainder on integer division (C90 6.3.5).'
15985 GCC always follows the C99 requirement that the result of division
15986 is truncated towards zero.
15990 File: gcc.info, Node: Floating point implementation, Next: Arrays and pointers implementation, Prev: Integers implementation, Up: C Implementation
15995 * `The accuracy of the floating-point operations and of the library
15996 functions in `<math.h>' and `<complex.h>' that return
15997 floating-point results (C90 and C99 5.2.4.2.2).'
15999 The accuracy is unknown.
16001 * `The rounding behaviors characterized by non-standard values of
16002 `FLT_ROUNDS' (C90 and C99 5.2.4.2.2).'
16004 GCC does not use such values.
16006 * `The evaluation methods characterized by non-standard negative
16007 values of `FLT_EVAL_METHOD' (C99 5.2.4.2.2).'
16009 GCC does not use such values.
16011 * `The direction of rounding when an integer is converted to a
16012 floating-point number that cannot exactly represent the original
16013 value (C90 6.2.1.3, C99 6.3.1.4).'
16015 C99 Annex F is followed.
16017 * `The direction of rounding when a floating-point number is
16018 converted to a narrower floating-point number (C90 6.2.1.4, C99
16021 C99 Annex F is followed.
16023 * `How the nearest representable value or the larger or smaller
16024 representable value immediately adjacent to the nearest
16025 representable value is chosen for certain floating constants (C90
16026 6.1.3.1, C99 6.4.4.2).'
16028 C99 Annex F is followed.
16030 * `Whether and how floating expressions are contracted when not
16031 disallowed by the `FP_CONTRACT' pragma (C99 6.5).'
16033 Expressions are currently only contracted if
16034 `-funsafe-math-optimizations' or `-ffast-math' are used. This is
16037 * `The default state for the `FENV_ACCESS' pragma (C99 7.6.1).'
16039 This pragma is not implemented, but the default is to "off" unless
16040 `-frounding-math' is used in which case it is "on".
16042 * `Additional floating-point exceptions, rounding modes,
16043 environments, and classifications, and their macro names (C99 7.6,
16046 This is dependent on the implementation of the C library, and is
16047 not defined by GCC itself.
16049 * `The default state for the `FP_CONTRACT' pragma (C99 7.12.2).'
16051 This pragma is not implemented. Expressions are currently only
16052 contracted if `-funsafe-math-optimizations' or `-ffast-math' are
16053 used. This is subject to change.
16055 * `Whether the "inexact" floating-point exception can be raised when
16056 the rounded result actually does equal the mathematical result in
16057 an IEC 60559 conformant implementation (C99 F.9).'
16059 This is dependent on the implementation of the C library, and is
16060 not defined by GCC itself.
16062 * `Whether the "underflow" (and "inexact") floating-point exception
16063 can be raised when a result is tiny but not inexact in an IEC
16064 60559 conformant implementation (C99 F.9).'
16066 This is dependent on the implementation of the C library, and is
16067 not defined by GCC itself.
16071 File: gcc.info, Node: Arrays and pointers implementation, Next: Hints implementation, Prev: Floating point implementation, Up: C Implementation
16073 4.7 Arrays and pointers
16074 =======================
16076 * `The result of converting a pointer to an integer or vice versa
16077 (C90 6.3.4, C99 6.3.2.3).'
16079 A cast from pointer to integer discards most-significant bits if
16080 the pointer representation is larger than the integer type,
16081 sign-extends(1) if the pointer representation is smaller than the
16082 integer type, otherwise the bits are unchanged.
16084 A cast from integer to pointer discards most-significant bits if
16085 the pointer representation is smaller than the integer type,
16086 extends according to the signedness of the integer type if the
16087 pointer representation is larger than the integer type, otherwise
16088 the bits are unchanged.
16090 When casting from pointer to integer and back again, the resulting
16091 pointer must reference the same object as the original pointer,
16092 otherwise the behavior is undefined. That is, one may not use
16093 integer arithmetic to avoid the undefined behavior of pointer
16094 arithmetic as proscribed in C99 6.5.6/8.
16096 * `The size of the result of subtracting two pointers to elements of
16097 the same array (C90 6.3.6, C99 6.5.6).'
16099 The value is as specified in the standard and the type is
16100 determined by the ABI.
16103 ---------- Footnotes ----------
16105 (1) Future versions of GCC may zero-extend, or use a target-defined
16106 `ptr_extend' pattern. Do not rely on sign extension.
16109 File: gcc.info, Node: Hints implementation, Next: Structures unions enumerations and bit-fields implementation, Prev: Arrays and pointers implementation, Up: C Implementation
16114 * `The extent to which suggestions made by using the `register'
16115 storage-class specifier are effective (C90 6.5.1, C99 6.7.1).'
16117 The `register' specifier affects code generation only in these
16120 * When used as part of the register variable extension, see
16121 *Note Explicit Reg Vars::.
16123 * When `-O0' is in use, the compiler allocates distinct stack
16124 memory for all variables that do not have the `register'
16125 storage-class specifier; if `register' is specified, the
16126 variable may have a shorter lifespan than the code would
16127 indicate and may never be placed in memory.
16129 * On some rare x86 targets, `setjmp' doesn't save the registers
16130 in all circumstances. In those cases, GCC doesn't allocate
16131 any variables in registers unless they are marked `register'.
16134 * `The extent to which suggestions made by using the inline function
16135 specifier are effective (C99 6.7.4).'
16137 GCC will not inline any functions if the `-fno-inline' option is
16138 used or if `-O0' is used. Otherwise, GCC may still be unable to
16139 inline a function for many reasons; the `-Winline' option may be
16140 used to determine if a function has not been inlined and why not.
16144 File: gcc.info, Node: Structures unions enumerations and bit-fields implementation, Next: Qualifiers implementation, Prev: Hints implementation, Up: C Implementation
16146 4.9 Structures, unions, enumerations, and bit-fields
16147 ====================================================
16149 * `A member of a union object is accessed using a member of a
16150 different type (C90 6.3.2.3).'
16152 The relevant bytes of the representation of the object are treated
16153 as an object of the type used for the access. *Note
16154 Type-punning::. This may be a trap representation.
16156 * `Whether a "plain" `int' bit-field is treated as a `signed int'
16157 bit-field or as an `unsigned int' bit-field (C90 6.5.2, C90
16158 6.5.2.1, C99 6.7.2, C99 6.7.2.1).'
16160 By default it is treated as `signed int' but this may be changed
16161 by the `-funsigned-bitfields' option.
16163 * `Allowable bit-field types other than `_Bool', `signed int', and
16164 `unsigned int' (C99 6.7.2.1).'
16166 No other types are permitted in strictly conforming mode.
16168 * `Whether a bit-field can straddle a storage-unit boundary (C90
16169 6.5.2.1, C99 6.7.2.1).'
16173 * `The order of allocation of bit-fields within a unit (C90 6.5.2.1,
16178 * `The alignment of non-bit-field members of structures (C90
16179 6.5.2.1, C99 6.7.2.1).'
16183 * `The integer type compatible with each enumerated type (C90
16184 6.5.2.2, C99 6.7.2.2).'
16186 Normally, the type is `unsigned int' if there are no negative
16187 values in the enumeration, otherwise `int'. If `-fshort-enums' is
16188 specified, then if there are negative values it is the first of
16189 `signed char', `short' and `int' that can represent all the
16190 values, otherwise it is the first of `unsigned char', `unsigned
16191 short' and `unsigned int' that can represent all the values.
16193 On some targets, `-fshort-enums' is the default; this is
16194 determined by the ABI.
16198 File: gcc.info, Node: Qualifiers implementation, Next: Declarators implementation, Prev: Structures unions enumerations and bit-fields implementation, Up: C Implementation
16203 * `What constitutes an access to an object that has
16204 volatile-qualified type (C90 6.5.3, C99 6.7.3).'
16206 Such an object is normally accessed by pointers and used for
16207 accessing hardware. In most expressions, it is intuitively
16208 obvious what is a read and what is a write. For example
16210 volatile int *dst = SOMEVALUE;
16211 volatile int *src = SOMEOTHERVALUE;
16214 will cause a read of the volatile object pointed to by SRC and
16215 store the value into the volatile object pointed to by DST. There
16216 is no guarantee that these reads and writes are atomic, especially
16217 for objects larger than `int'.
16219 However, if the volatile storage is not being modified, and the
16220 value of the volatile storage is not used, then the situation is
16221 less obvious. For example
16223 volatile int *src = SOMEVALUE;
16226 According to the C standard, such an expression is an rvalue whose
16227 type is the unqualified version of its original type, i.e. `int'.
16228 Whether GCC interprets this as a read of the volatile object being
16229 pointed to or only as a request to evaluate the expression for its
16230 side-effects depends on this type.
16232 If it is a scalar type, or on most targets an aggregate type whose
16233 only member object is of a scalar type, or a union type whose
16234 member objects are of scalar types, the expression is interpreted
16235 by GCC as a read of the volatile object; in the other cases, the
16236 expression is only evaluated for its side-effects.
16240 File: gcc.info, Node: Declarators implementation, Next: Statements implementation, Prev: Qualifiers implementation, Up: C Implementation
16245 * `The maximum number of declarators that may modify an arithmetic,
16246 structure or union type (C90 6.5.4).'
16248 GCC is only limited by available memory.
16252 File: gcc.info, Node: Statements implementation, Next: Preprocessing directives implementation, Prev: Declarators implementation, Up: C Implementation
16257 * `The maximum number of `case' values in a `switch' statement (C90
16260 GCC is only limited by available memory.
16264 File: gcc.info, Node: Preprocessing directives implementation, Next: Library functions implementation, Prev: Statements implementation, Up: C Implementation
16266 4.13 Preprocessing directives
16267 =============================
16269 *Note Implementation-defined behavior: (cpp)Implementation-defined
16270 behavior, for details of these aspects of implementation-defined
16273 * `How sequences in both forms of header names are mapped to headers
16274 or external source file names (C90 6.1.7, C99 6.4.7).'
16276 * `Whether the value of a character constant in a constant expression
16277 that controls conditional inclusion matches the value of the same
16278 character constant in the execution character set (C90 6.8.1, C99
16281 * `Whether the value of a single-character character constant in a
16282 constant expression that controls conditional inclusion may have a
16283 negative value (C90 6.8.1, C99 6.10.1).'
16285 * `The places that are searched for an included `<>' delimited
16286 header, and how the places are specified or the header is
16287 identified (C90 6.8.2, C99 6.10.2).'
16289 * `How the named source file is searched for in an included `""'
16290 delimited header (C90 6.8.2, C99 6.10.2).'
16292 * `The method by which preprocessing tokens (possibly resulting from
16293 macro expansion) in a `#include' directive are combined into a
16294 header name (C90 6.8.2, C99 6.10.2).'
16296 * `The nesting limit for `#include' processing (C90 6.8.2, C99
16299 * `Whether the `#' operator inserts a `\' character before the `\'
16300 character that begins a universal character name in a character
16301 constant or string literal (C99 6.10.3.2).'
16303 * `The behavior on each recognized non-`STDC #pragma' directive (C90
16304 6.8.6, C99 6.10.6).'
16306 *Note Pragmas: (cpp)Pragmas, for details of pragmas accepted by
16307 GCC on all targets. *Note Pragmas Accepted by GCC: Pragmas, for
16308 details of target-specific pragmas.
16310 * `The definitions for `__DATE__' and `__TIME__' when respectively,
16311 the date and time of translation are not available (C90 6.8.8, C99
16316 File: gcc.info, Node: Library functions implementation, Next: Architecture implementation, Prev: Preprocessing directives implementation, Up: C Implementation
16318 4.14 Library functions
16319 ======================
16321 The behavior of most of these points are dependent on the implementation
16322 of the C library, and are not defined by GCC itself.
16324 * `The null pointer constant to which the macro `NULL' expands (C90
16327 In `<stddef.h>', `NULL' expands to `((void *)0)'. GCC does not
16328 provide the other headers which define `NULL' and some library
16329 implementations may use other definitions in those headers.
16333 File: gcc.info, Node: Architecture implementation, Next: Locale-specific behavior implementation, Prev: Library functions implementation, Up: C Implementation
16338 * `The values or expressions assigned to the macros specified in the
16339 headers `<float.h>', `<limits.h>', and `<stdint.h>' (C90 and C99
16340 5.2.4.2, C99 7.18.2, C99 7.18.3).'
16344 * `The number, order, and encoding of bytes in any object (when not
16345 explicitly specified in this International Standard) (C99
16350 * `The value of the result of the `sizeof' operator (C90 6.3.3.4,
16357 File: gcc.info, Node: Locale-specific behavior implementation, Prev: Architecture implementation, Up: C Implementation
16359 4.16 Locale-specific behavior
16360 =============================
16362 The behavior of these points are dependent on the implementation of the
16363 C library, and are not defined by GCC itself.
16366 File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: C Implementation, Up: Top
16368 5 Extensions to the C Language Family
16369 *************************************
16371 GNU C provides several language features not found in ISO standard C.
16372 (The `-pedantic' option directs GCC to print a warning message if any
16373 of these features is used.) To test for the availability of these
16374 features in conditional compilation, check for a predefined macro
16375 `__GNUC__', which is always defined under GCC.
16377 These extensions are available in C and Objective-C. Most of them are
16378 also available in C++. *Note Extensions to the C++ Language: C++
16379 Extensions, for extensions that apply _only_ to C++.
16381 Some features that are in ISO C99 but not C89 or C++ are also, as
16382 extensions, accepted by GCC in C89 mode and in C++.
16386 * Statement Exprs:: Putting statements and declarations inside expressions.
16387 * Local Labels:: Labels local to a block.
16388 * Labels as Values:: Getting pointers to labels, and computed gotos.
16389 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
16390 * Constructing Calls:: Dispatching a call to another function.
16391 * Typeof:: `typeof': referring to the type of an expression.
16392 * Conditionals:: Omitting the middle operand of a `?:' expression.
16393 * Long Long:: Double-word integers---`long long int'.
16394 * Complex:: Data types for complex numbers.
16395 * Floating Types:: Additional Floating Types.
16396 * Decimal Float:: Decimal Floating Types.
16397 * Hex Floats:: Hexadecimal floating-point constants.
16398 * Fixed-Point:: Fixed-Point Types.
16399 * Zero Length:: Zero-length arrays.
16400 * Variable Length:: Arrays whose length is computed at run time.
16401 * Empty Structures:: Structures with no members.
16402 * Variadic Macros:: Macros with a variable number of arguments.
16403 * Escaped Newlines:: Slightly looser rules for escaped newlines.
16404 * Subscripting:: Any array can be subscripted, even if not an lvalue.
16405 * Pointer Arith:: Arithmetic on `void'-pointers and function pointers.
16406 * Initializers:: Non-constant initializers.
16407 * Compound Literals:: Compound literals give structures, unions
16408 or arrays as values.
16409 * Designated Inits:: Labeling elements of initializers.
16410 * Cast to Union:: Casting to union type from any member of the union.
16411 * Case Ranges:: `case 1 ... 9' and such.
16412 * Mixed Declarations:: Mixing declarations and code.
16413 * Function Attributes:: Declaring that functions have no side effects,
16414 or that they can never return.
16415 * Attribute Syntax:: Formal syntax for attributes.
16416 * Function Prototypes:: Prototype declarations and old-style definitions.
16417 * C++ Comments:: C++ comments are recognized.
16418 * Dollar Signs:: Dollar sign is allowed in identifiers.
16419 * Character Escapes:: `\e' stands for the character <ESC>.
16420 * Variable Attributes:: Specifying attributes of variables.
16421 * Type Attributes:: Specifying attributes of types.
16422 * Alignment:: Inquiring about the alignment of a type or variable.
16423 * Inline:: Defining inline functions (as fast as macros).
16424 * Extended Asm:: Assembler instructions with C expressions as operands.
16425 (With them you can define ``built-in'' functions.)
16426 * Constraints:: Constraints for asm operands
16427 * Asm Labels:: Specifying the assembler name to use for a C symbol.
16428 * Explicit Reg Vars:: Defining variables residing in specified registers.
16429 * Alternate Keywords:: `__const__', `__asm__', etc., for header files.
16430 * Incomplete Enums:: `enum foo;', with details to follow.
16431 * Function Names:: Printable strings which are the name of the current
16433 * Return Address:: Getting the return or frame address of a function.
16434 * Vector Extensions:: Using vector instructions through built-in functions.
16435 * Offsetof:: Special syntax for implementing `offsetof'.
16436 * Atomic Builtins:: Built-in functions for atomic memory access.
16437 * Object Size Checking:: Built-in functions for limited buffer overflow
16439 * Other Builtins:: Other built-in functions.
16440 * Target Builtins:: Built-in functions specific to particular targets.
16441 * Target Format Checks:: Format checks specific to particular targets.
16442 * Pragmas:: Pragmas accepted by GCC.
16443 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
16444 * Thread-Local:: Per-thread variables.
16445 * Binary constants:: Binary constants using the `0b' prefix.
16448 File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions
16450 5.1 Statements and Declarations in Expressions
16451 ==============================================
16453 A compound statement enclosed in parentheses may appear as an expression
16454 in GNU C. This allows you to use loops, switches, and local variables
16455 within an expression.
16457 Recall that a compound statement is a sequence of statements surrounded
16458 by braces; in this construct, parentheses go around the braces. For
16461 ({ int y = foo (); int z;
16466 is a valid (though slightly more complex than necessary) expression for
16467 the absolute value of `foo ()'.
16469 The last thing in the compound statement should be an expression
16470 followed by a semicolon; the value of this subexpression serves as the
16471 value of the entire construct. (If you use some other kind of statement
16472 last within the braces, the construct has type `void', and thus
16473 effectively no value.)
16475 This feature is especially useful in making macro definitions "safe"
16476 (so that they evaluate each operand exactly once). For example, the
16477 "maximum" function is commonly defined as a macro in standard C as
16480 #define max(a,b) ((a) > (b) ? (a) : (b))
16482 But this definition computes either A or B twice, with bad results if
16483 the operand has side effects. In GNU C, if you know the type of the
16484 operands (here taken as `int'), you can define the macro safely as
16487 #define maxint(a,b) \
16488 ({int _a = (a), _b = (b); _a > _b ? _a : _b; })
16490 Embedded statements are not allowed in constant expressions, such as
16491 the value of an enumeration constant, the width of a bit-field, or the
16492 initial value of a static variable.
16494 If you don't know the type of the operand, you can still do this, but
16495 you must use `typeof' (*note Typeof::).
16497 In G++, the result value of a statement expression undergoes array and
16498 function pointer decay, and is returned by value to the enclosing
16499 expression. For instance, if `A' is a class, then
16505 will construct a temporary `A' object to hold the result of the
16506 statement expression, and that will be used to invoke `Foo'. Therefore
16507 the `this' pointer observed by `Foo' will not be the address of `a'.
16509 Any temporaries created within a statement within a statement
16510 expression will be destroyed at the statement's end. This makes
16511 statement expressions inside macros slightly different from function
16512 calls. In the latter case temporaries introduced during argument
16513 evaluation will be destroyed at the end of the statement that includes
16514 the function call. In the statement expression case they will be
16515 destroyed during the statement expression. For instance,
16517 #define macro(a) ({__typeof__(a) b = (a); b + 3; })
16518 template<typename T> T function(T a) { T b = a; return b + 3; }
16526 will have different places where temporaries are destroyed. For the
16527 `macro' case, the temporary `X' will be destroyed just after the
16528 initialization of `b'. In the `function' case that temporary will be
16529 destroyed when the function returns.
16531 These considerations mean that it is probably a bad idea to use
16532 statement-expressions of this form in header files that are designed to
16533 work with C++. (Note that some versions of the GNU C Library contained
16534 header files using statement-expression that lead to precisely this
16537 Jumping into a statement expression with `goto' or using a `switch'
16538 statement outside the statement expression with a `case' or `default'
16539 label inside the statement expression is not permitted. Jumping into a
16540 statement expression with a computed `goto' (*note Labels as Values::)
16541 yields undefined behavior. Jumping out of a statement expression is
16542 permitted, but if the statement expression is part of a larger
16543 expression then it is unspecified which other subexpressions of that
16544 expression have been evaluated except where the language definition
16545 requires certain subexpressions to be evaluated before or after the
16546 statement expression. In any case, as with a function call the
16547 evaluation of a statement expression is not interleaved with the
16548 evaluation of other parts of the containing expression. For example,
16550 foo (), (({ bar1 (); goto a; 0; }) + bar2 ()), baz();
16552 will call `foo' and `bar1' and will not call `baz' but may or may not
16553 call `bar2'. If `bar2' is called, it will be called after `foo' and
16557 File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions
16559 5.2 Locally Declared Labels
16560 ===========================
16562 GCC allows you to declare "local labels" in any nested block scope. A
16563 local label is just like an ordinary label, but you can only reference
16564 it (with a `goto' statement, or by taking its address) within the block
16565 in which it was declared.
16567 A local label declaration looks like this:
16573 __label__ LABEL1, LABEL2, /* ... */;
16575 Local label declarations must come at the beginning of the block,
16576 before any ordinary declarations or statements.
16578 The label declaration defines the label _name_, but does not define
16579 the label itself. You must do this in the usual way, with `LABEL:',
16580 within the statements of the statement expression.
16582 The local label feature is useful for complex macros. If a macro
16583 contains nested loops, a `goto' can be useful for breaking out of them.
16584 However, an ordinary label whose scope is the whole function cannot be
16585 used: if the macro can be expanded several times in one function, the
16586 label will be multiply defined in that function. A local label avoids
16587 this problem. For example:
16589 #define SEARCH(value, array, target) \
16592 typeof (target) _SEARCH_target = (target); \
16593 typeof (*(array)) *_SEARCH_array = (array); \
16596 for (i = 0; i < max; i++) \
16597 for (j = 0; j < max; j++) \
16598 if (_SEARCH_array[i][j] == _SEARCH_target) \
16599 { (value) = i; goto found; } \
16604 This could also be written using a statement-expression:
16606 #define SEARCH(array, target) \
16609 typeof (target) _SEARCH_target = (target); \
16610 typeof (*(array)) *_SEARCH_array = (array); \
16613 for (i = 0; i < max; i++) \
16614 for (j = 0; j < max; j++) \
16615 if (_SEARCH_array[i][j] == _SEARCH_target) \
16616 { value = i; goto found; } \
16622 Local label declarations also make the labels they declare visible to
16623 nested functions, if there are any. *Note Nested Functions::, for
16627 File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions
16629 5.3 Labels as Values
16630 ====================
16632 You can get the address of a label defined in the current function (or
16633 a containing function) with the unary operator `&&'. The value has
16634 type `void *'. This value is a constant and can be used wherever a
16635 constant of that type is valid. For example:
16641 To use these values, you need to be able to jump to one. This is done
16642 with the computed goto statement(1), `goto *EXP;'. For example,
16646 Any expression of type `void *' is allowed.
16648 One way of using these constants is in initializing a static array that
16649 will serve as a jump table:
16651 static void *array[] = { &&foo, &&bar, &&hack };
16653 Then you can select a label with indexing, like this:
16657 Note that this does not check whether the subscript is in bounds--array
16658 indexing in C never does that.
16660 Such an array of label values serves a purpose much like that of the
16661 `switch' statement. The `switch' statement is cleaner, so use that
16662 rather than an array unless the problem does not fit a `switch'
16663 statement very well.
16665 Another use of label values is in an interpreter for threaded code.
16666 The labels within the interpreter function can be stored in the
16667 threaded code for super-fast dispatching.
16669 You may not use this mechanism to jump to code in a different function.
16670 If you do that, totally unpredictable things will happen. The best way
16671 to avoid this is to store the label address only in automatic variables
16672 and never pass it as an argument.
16674 An alternate way to write the above example is
16676 static const int array[] = { &&foo - &&foo, &&bar - &&foo,
16678 goto *(&&foo + array[i]);
16680 This is more friendly to code living in shared libraries, as it reduces
16681 the number of dynamic relocations that are needed, and by consequence,
16682 allows the data to be read-only.
16684 The `&&foo' expressions for the same label might have different values
16685 if the containing function is inlined or cloned. If a program relies on
16686 them being always the same, `__attribute__((__noinline__))' should be
16687 used to prevent inlining. If `&&foo' is used in a static variable
16688 initializer, inlining is forbidden.
16690 ---------- Footnotes ----------
16692 (1) The analogous feature in Fortran is called an assigned goto, but
16693 that name seems inappropriate in C, where one can do more than simply
16694 store label addresses in label variables.
16697 File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions
16699 5.4 Nested Functions
16700 ====================
16702 A "nested function" is a function defined inside another function.
16703 (Nested functions are not supported for GNU C++.) The nested function's
16704 name is local to the block where it is defined. For example, here we
16705 define a nested function named `square', and call it twice:
16707 foo (double a, double b)
16709 double square (double z) { return z * z; }
16711 return square (a) + square (b);
16714 The nested function can access all the variables of the containing
16715 function that are visible at the point of its definition. This is
16716 called "lexical scoping". For example, here we show a nested function
16717 which uses an inherited variable named `offset':
16719 bar (int *array, int offset, int size)
16721 int access (int *array, int index)
16722 { return array[index + offset]; }
16725 for (i = 0; i < size; i++)
16726 /* ... */ access (array, i) /* ... */
16729 Nested function definitions are permitted within functions in the
16730 places where variable definitions are allowed; that is, in any block,
16731 mixed with the other declarations and statements in the block.
16733 It is possible to call the nested function from outside the scope of
16734 its name by storing its address or passing the address to another
16737 hack (int *array, int size)
16739 void store (int index, int value)
16740 { array[index] = value; }
16742 intermediate (store, size);
16745 Here, the function `intermediate' receives the address of `store' as
16746 an argument. If `intermediate' calls `store', the arguments given to
16747 `store' are used to store into `array'. But this technique works only
16748 so long as the containing function (`hack', in this example) does not
16751 If you try to call the nested function through its address after the
16752 containing function has exited, all hell will break loose. If you try
16753 to call it after a containing scope level has exited, and if it refers
16754 to some of the variables that are no longer in scope, you may be lucky,
16755 but it's not wise to take the risk. If, however, the nested function
16756 does not refer to anything that has gone out of scope, you should be
16759 GCC implements taking the address of a nested function using a
16760 technique called "trampolines". A paper describing them is available as
16762 `http://people.debian.org/~aaronl/Usenix88-lexic.pdf'.
16764 A nested function can jump to a label inherited from a containing
16765 function, provided the label was explicitly declared in the containing
16766 function (*note Local Labels::). Such a jump returns instantly to the
16767 containing function, exiting the nested function which did the `goto'
16768 and any intermediate functions as well. Here is an example:
16770 bar (int *array, int offset, int size)
16773 int access (int *array, int index)
16777 return array[index + offset];
16781 for (i = 0; i < size; i++)
16782 /* ... */ access (array, i) /* ... */
16786 /* Control comes here from `access'
16787 if it detects an error. */
16792 A nested function always has no linkage. Declaring one with `extern'
16793 or `static' is erroneous. If you need to declare the nested function
16794 before its definition, use `auto' (which is otherwise meaningless for
16795 function declarations).
16797 bar (int *array, int offset, int size)
16800 auto int access (int *, int);
16802 int access (int *array, int index)
16806 return array[index + offset];
16812 File: gcc.info, Node: Constructing Calls, Next: Typeof, Prev: Nested Functions, Up: C Extensions
16814 5.5 Constructing Function Calls
16815 ===============================
16817 Using the built-in functions described below, you can record the
16818 arguments a function received, and call another function with the same
16819 arguments, without knowing the number or types of the arguments.
16821 You can also record the return value of that function call, and later
16822 return that value, without knowing what data type the function tried to
16823 return (as long as your caller expects that data type).
16825 However, these built-in functions may interact badly with some
16826 sophisticated features or other extensions of the language. It is,
16827 therefore, not recommended to use them outside very simple functions
16828 acting as mere forwarders for their arguments.
16830 -- Built-in Function: void * __builtin_apply_args ()
16831 This built-in function returns a pointer to data describing how to
16832 perform a call with the same arguments as were passed to the
16835 The function saves the arg pointer register, structure value
16836 address, and all registers that might be used to pass arguments to
16837 a function into a block of memory allocated on the stack. Then it
16838 returns the address of that block.
16840 -- Built-in Function: void * __builtin_apply (void (*FUNCTION)(), void
16841 *ARGUMENTS, size_t SIZE)
16842 This built-in function invokes FUNCTION with a copy of the
16843 parameters described by ARGUMENTS and SIZE.
16845 The value of ARGUMENTS should be the value returned by
16846 `__builtin_apply_args'. The argument SIZE specifies the size of
16847 the stack argument data, in bytes.
16849 This function returns a pointer to data describing how to return
16850 whatever value was returned by FUNCTION. The data is saved in a
16851 block of memory allocated on the stack.
16853 It is not always simple to compute the proper value for SIZE. The
16854 value is used by `__builtin_apply' to compute the amount of data
16855 that should be pushed on the stack and copied from the incoming
16858 -- Built-in Function: void __builtin_return (void *RESULT)
16859 This built-in function returns the value described by RESULT from
16860 the containing function. You should specify, for RESULT, a value
16861 returned by `__builtin_apply'.
16863 -- Built-in Function: __builtin_va_arg_pack ()
16864 This built-in function represents all anonymous arguments of an
16865 inline function. It can be used only in inline functions which
16866 will be always inlined, never compiled as a separate function,
16867 such as those using `__attribute__ ((__always_inline__))' or
16868 `__attribute__ ((__gnu_inline__))' extern inline functions. It
16869 must be only passed as last argument to some other function with
16870 variable arguments. This is useful for writing small wrapper
16871 inlines for variable argument functions, when using preprocessor
16872 macros is undesirable. For example:
16873 extern int myprintf (FILE *f, const char *format, ...);
16874 extern inline __attribute__ ((__gnu_inline__)) int
16875 myprintf (FILE *f, const char *format, ...)
16877 int r = fprintf (f, "myprintf: ");
16880 int s = fprintf (f, format, __builtin_va_arg_pack ());
16886 -- Built-in Function: __builtin_va_arg_pack_len ()
16887 This built-in function returns the number of anonymous arguments of
16888 an inline function. It can be used only in inline functions which
16889 will be always inlined, never compiled as a separate function, such
16890 as those using `__attribute__ ((__always_inline__))' or
16891 `__attribute__ ((__gnu_inline__))' extern inline functions. For
16892 example following will do link or runtime checking of open
16893 arguments for optimized code:
16894 #ifdef __OPTIMIZE__
16895 extern inline __attribute__((__gnu_inline__)) int
16896 myopen (const char *path, int oflag, ...)
16898 if (__builtin_va_arg_pack_len () > 1)
16899 warn_open_too_many_arguments ();
16901 if (__builtin_constant_p (oflag))
16903 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
16905 warn_open_missing_mode ();
16906 return __open_2 (path, oflag);
16908 return open (path, oflag, __builtin_va_arg_pack ());
16911 if (__builtin_va_arg_pack_len () < 1)
16912 return __open_2 (path, oflag);
16914 return open (path, oflag, __builtin_va_arg_pack ());
16919 File: gcc.info, Node: Typeof, Next: Conditionals, Prev: Constructing Calls, Up: C Extensions
16921 5.6 Referring to a Type with `typeof'
16922 =====================================
16924 Another way to refer to the type of an expression is with `typeof'.
16925 The syntax of using of this keyword looks like `sizeof', but the
16926 construct acts semantically like a type name defined with `typedef'.
16928 There are two ways of writing the argument to `typeof': with an
16929 expression or with a type. Here is an example with an expression:
16933 This assumes that `x' is an array of pointers to functions; the type
16934 described is that of the values of the functions.
16936 Here is an example with a typename as the argument:
16940 Here the type described is that of pointers to `int'.
16942 If you are writing a header file that must work when included in ISO C
16943 programs, write `__typeof__' instead of `typeof'. *Note Alternate
16946 A `typeof'-construct can be used anywhere a typedef name could be
16947 used. For example, you can use it in a declaration, in a cast, or
16948 inside of `sizeof' or `typeof'.
16950 `typeof' is often useful in conjunction with the
16951 statements-within-expressions feature. Here is how the two together can
16952 be used to define a safe "maximum" macro that operates on any
16953 arithmetic type and evaluates each of its arguments exactly once:
16956 ({ typeof (a) _a = (a); \
16957 typeof (b) _b = (b); \
16958 _a > _b ? _a : _b; })
16960 The reason for using names that start with underscores for the local
16961 variables is to avoid conflicts with variable names that occur within
16962 the expressions that are substituted for `a' and `b'. Eventually we
16963 hope to design a new form of declaration syntax that allows you to
16964 declare variables whose scopes start only after their initializers;
16965 this will be a more reliable way to prevent such conflicts.
16967 Some more examples of the use of `typeof':
16969 * This declares `y' with the type of what `x' points to.
16973 * This declares `y' as an array of such values.
16977 * This declares `y' as an array of pointers to characters:
16979 typeof (typeof (char *)[4]) y;
16981 It is equivalent to the following traditional C declaration:
16985 To see the meaning of the declaration using `typeof', and why it
16986 might be a useful way to write, rewrite it with these macros:
16988 #define pointer(T) typeof(T *)
16989 #define array(T, N) typeof(T [N])
16991 Now the declaration can be rewritten this way:
16993 array (pointer (char), 4) y;
16995 Thus, `array (pointer (char), 4)' is the type of arrays of 4
16996 pointers to `char'.
16998 _Compatibility Note:_ In addition to `typeof', GCC 2 supported a more
16999 limited extension which permitted one to write
17003 with the effect of declaring T to have the type of the expression EXPR.
17004 This extension does not work with GCC 3 (versions between 3.0 and 3.2
17005 will crash; 3.2.1 and later give an error). Code which relies on it
17006 should be rewritten to use `typeof':
17008 typedef typeof(EXPR) T;
17010 This will work with all versions of GCC.
17013 File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Typeof, Up: C Extensions
17015 5.7 Conditionals with Omitted Operands
17016 ======================================
17018 The middle operand in a conditional expression may be omitted. Then if
17019 the first operand is nonzero, its value is the value of the conditional
17022 Therefore, the expression
17026 has the value of `x' if that is nonzero; otherwise, the value of `y'.
17028 This example is perfectly equivalent to
17032 In this simple case, the ability to omit the middle operand is not
17033 especially useful. When it becomes useful is when the first operand
17034 does, or may (if it is a macro argument), contain a side effect. Then
17035 repeating the operand in the middle would perform the side effect
17036 twice. Omitting the middle operand uses the value already computed
17037 without the undesirable effects of recomputing it.
17040 File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions
17042 5.8 Double-Word Integers
17043 ========================
17045 ISO C99 supports data types for integers that are at least 64 bits wide,
17046 and as an extension GCC supports them in C89 mode and in C++. Simply
17047 write `long long int' for a signed integer, or `unsigned long long int'
17048 for an unsigned integer. To make an integer constant of type `long
17049 long int', add the suffix `LL' to the integer. To make an integer
17050 constant of type `unsigned long long int', add the suffix `ULL' to the
17053 You can use these types in arithmetic like any other integer types.
17054 Addition, subtraction, and bitwise boolean operations on these types
17055 are open-coded on all types of machines. Multiplication is open-coded
17056 if the machine supports fullword-to-doubleword a widening multiply
17057 instruction. Division and shifts are open-coded only on machines that
17058 provide special support. The operations that are not open-coded use
17059 special library routines that come with GCC.
17061 There may be pitfalls when you use `long long' types for function
17062 arguments, unless you declare function prototypes. If a function
17063 expects type `int' for its argument, and you pass a value of type `long
17064 long int', confusion will result because the caller and the subroutine
17065 will disagree about the number of bytes for the argument. Likewise, if
17066 the function expects `long long int' and you pass `int'. The best way
17067 to avoid such problems is to use prototypes.
17070 File: gcc.info, Node: Complex, Next: Floating Types, Prev: Long Long, Up: C Extensions
17072 5.9 Complex Numbers
17073 ===================
17075 ISO C99 supports complex floating data types, and as an extension GCC
17076 supports them in C89 mode and in C++, and supports complex integer data
17077 types which are not part of ISO C99. You can declare complex types
17078 using the keyword `_Complex'. As an extension, the older GNU keyword
17079 `__complex__' is also supported.
17081 For example, `_Complex double x;' declares `x' as a variable whose
17082 real part and imaginary part are both of type `double'. `_Complex
17083 short int y;' declares `y' to have real and imaginary parts of type
17084 `short int'; this is not likely to be useful, but it shows that the set
17085 of complex types is complete.
17087 To write a constant with a complex data type, use the suffix `i' or
17088 `j' (either one; they are equivalent). For example, `2.5fi' has type
17089 `_Complex float' and `3i' has type `_Complex int'. Such a constant
17090 always has a pure imaginary value, but you can form any complex value
17091 you like by adding one to a real constant. This is a GNU extension; if
17092 you have an ISO C99 conforming C library (such as GNU libc), and want
17093 to construct complex constants of floating type, you should include
17094 `<complex.h>' and use the macros `I' or `_Complex_I' instead.
17096 To extract the real part of a complex-valued expression EXP, write
17097 `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary
17098 part. This is a GNU extension; for values of floating type, you should
17099 use the ISO C99 functions `crealf', `creal', `creall', `cimagf',
17100 `cimag' and `cimagl', declared in `<complex.h>' and also provided as
17101 built-in functions by GCC.
17103 The operator `~' performs complex conjugation when used on a value
17104 with a complex type. This is a GNU extension; for values of floating
17105 type, you should use the ISO C99 functions `conjf', `conj' and `conjl',
17106 declared in `<complex.h>' and also provided as built-in functions by
17109 GCC can allocate complex automatic variables in a noncontiguous
17110 fashion; it's even possible for the real part to be in a register while
17111 the imaginary part is on the stack (or vice-versa). Only the DWARF2
17112 debug info format can represent this, so use of DWARF2 is recommended.
17113 If you are using the stabs debug info format, GCC describes a
17114 noncontiguous complex variable as if it were two separate variables of
17115 noncomplex type. If the variable's actual name is `foo', the two
17116 fictitious variables are named `foo$real' and `foo$imag'. You can
17117 examine and set these two fictitious variables with your debugger.
17120 File: gcc.info, Node: Floating Types, Next: Decimal Float, Prev: Complex, Up: C Extensions
17122 5.10 Additional Floating Types
17123 ==============================
17125 As an extension, the GNU C compiler supports additional floating types,
17126 `__float80' and `__float128' to support 80bit (`XFmode') and 128 bit
17127 (`TFmode') floating types. Support for additional types includes the
17128 arithmetic operators: add, subtract, multiply, divide; unary arithmetic
17129 operators; relational operators; equality operators; and conversions to
17130 and from integer and other floating types. Use a suffix `w' or `W' in
17131 a literal constant of type `__float80' and `q' or `Q' for `_float128'.
17132 You can declare complex types using the corresponding internal complex
17133 type, `XCmode' for `__float80' type and `TCmode' for `__float128' type:
17135 typedef _Complex float __attribute__((mode(TC))) _Complex128;
17136 typedef _Complex float __attribute__((mode(XC))) _Complex80;
17138 Not all targets support additional floating point types. `__float80'
17139 is supported on i386, x86_64 and ia64 targets and target `__float128'
17140 is supported on x86_64 and ia64 targets.
17143 File: gcc.info, Node: Decimal Float, Next: Hex Floats, Prev: Floating Types, Up: C Extensions
17145 5.11 Decimal Floating Types
17146 ===========================
17148 As an extension, the GNU C compiler supports decimal floating types as
17149 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
17150 floating types in GCC will evolve as the draft technical report changes.
17151 Calling conventions for any target might also change. Not all targets
17152 support decimal floating types.
17154 The decimal floating types are `_Decimal32', `_Decimal64', and
17155 `_Decimal128'. They use a radix of ten, unlike the floating types
17156 `float', `double', and `long double' whose radix is not specified by
17157 the C standard but is usually two.
17159 Support for decimal floating types includes the arithmetic operators
17160 add, subtract, multiply, divide; unary arithmetic operators; relational
17161 operators; equality operators; and conversions to and from integer and
17162 other floating types. Use a suffix `df' or `DF' in a literal constant
17163 of type `_Decimal32', `dd' or `DD' for `_Decimal64', and `dl' or `DL'
17166 GCC support of decimal float as specified by the draft technical report
17169 * Pragma `FLOAT_CONST_DECIMAL64' is not supported, nor is the `d'
17170 suffix for literal constants of type `double'.
17172 * When the value of a decimal floating type cannot be represented in
17173 the integer type to which it is being converted, the result is
17174 undefined rather than the result value specified by the draft
17177 * GCC does not provide the C library functionality associated with
17178 `math.h', `fenv.h', `stdio.h', `stdlib.h', and `wchar.h', which
17179 must come from a separate C library implementation. Because of
17180 this the GNU C compiler does not define macro `__STDC_DEC_FP__' to
17181 indicate that the implementation conforms to the technical report.
17183 Types `_Decimal32', `_Decimal64', and `_Decimal128' are supported by
17184 the DWARF2 debug information format.
17187 File: gcc.info, Node: Hex Floats, Next: Fixed-Point, Prev: Decimal Float, Up: C Extensions
17192 ISO C99 supports floating-point numbers written not only in the usual
17193 decimal notation, such as `1.55e1', but also numbers such as `0x1.fp3'
17194 written in hexadecimal format. As a GNU extension, GCC supports this
17195 in C89 mode (except in some cases when strictly conforming) and in C++.
17196 In that format the `0x' hex introducer and the `p' or `P' exponent
17197 field are mandatory. The exponent is a decimal number that indicates
17198 the power of 2 by which the significant part will be multiplied. Thus
17199 `0x1.f' is 1 15/16, `p3' multiplies it by 8, and the value of `0x1.fp3'
17200 is the same as `1.55e1'.
17202 Unlike for floating-point numbers in the decimal notation the exponent
17203 is always required in the hexadecimal notation. Otherwise the compiler
17204 would not be able to resolve the ambiguity of, e.g., `0x1.f'. This
17205 could mean `1.0f' or `1.9375' since `f' is also the extension for
17206 floating-point constants of type `float'.
17209 File: gcc.info, Node: Fixed-Point, Next: Zero Length, Prev: Hex Floats, Up: C Extensions
17211 5.13 Fixed-Point Types
17212 ======================
17214 As an extension, the GNU C compiler supports fixed-point types as
17215 defined in the N1169 draft of ISO/IEC DTR 18037. Support for
17216 fixed-point types in GCC will evolve as the draft technical report
17217 changes. Calling conventions for any target might also change. Not
17218 all targets support fixed-point types.
17220 The fixed-point types are `short _Fract', `_Fract', `long _Fract',
17221 `long long _Fract', `unsigned short _Fract', `unsigned _Fract',
17222 `unsigned long _Fract', `unsigned long long _Fract', `_Sat short
17223 _Fract', `_Sat _Fract', `_Sat long _Fract', `_Sat long long _Fract',
17224 `_Sat unsigned short _Fract', `_Sat unsigned _Fract', `_Sat unsigned
17225 long _Fract', `_Sat unsigned long long _Fract', `short _Accum',
17226 `_Accum', `long _Accum', `long long _Accum', `unsigned short _Accum',
17227 `unsigned _Accum', `unsigned long _Accum', `unsigned long long _Accum',
17228 `_Sat short _Accum', `_Sat _Accum', `_Sat long _Accum', `_Sat long long
17229 _Accum', `_Sat unsigned short _Accum', `_Sat unsigned _Accum', `_Sat
17230 unsigned long _Accum', `_Sat unsigned long long _Accum'.
17232 Fixed-point data values contain fractional and optional integral parts.
17233 The format of fixed-point data varies and depends on the target machine.
17235 Support for fixed-point types includes:
17236 * prefix and postfix increment and decrement operators (`++', `--')
17238 * unary arithmetic operators (`+', `-', `!')
17240 * binary arithmetic operators (`+', `-', `*', `/')
17242 * binary shift operators (`<<', `>>')
17244 * relational operators (`<', `<=', `>=', `>')
17246 * equality operators (`==', `!=')
17248 * assignment operators (`+=', `-=', `*=', `/=', `<<=', `>>=')
17250 * conversions to and from integer, floating-point, or fixed-point
17253 Use a suffix in a fixed-point literal constant:
17254 * `hr' or `HR' for `short _Fract' and `_Sat short _Fract'
17256 * `r' or `R' for `_Fract' and `_Sat _Fract'
17258 * `lr' or `LR' for `long _Fract' and `_Sat long _Fract'
17260 * `llr' or `LLR' for `long long _Fract' and `_Sat long long _Fract'
17262 * `uhr' or `UHR' for `unsigned short _Fract' and `_Sat unsigned
17265 * `ur' or `UR' for `unsigned _Fract' and `_Sat unsigned _Fract'
17267 * `ulr' or `ULR' for `unsigned long _Fract' and `_Sat unsigned long
17270 * `ullr' or `ULLR' for `unsigned long long _Fract' and `_Sat
17271 unsigned long long _Fract'
17273 * `hk' or `HK' for `short _Accum' and `_Sat short _Accum'
17275 * `k' or `K' for `_Accum' and `_Sat _Accum'
17277 * `lk' or `LK' for `long _Accum' and `_Sat long _Accum'
17279 * `llk' or `LLK' for `long long _Accum' and `_Sat long long _Accum'
17281 * `uhk' or `UHK' for `unsigned short _Accum' and `_Sat unsigned
17284 * `uk' or `UK' for `unsigned _Accum' and `_Sat unsigned _Accum'
17286 * `ulk' or `ULK' for `unsigned long _Accum' and `_Sat unsigned long
17289 * `ullk' or `ULLK' for `unsigned long long _Accum' and `_Sat
17290 unsigned long long _Accum'
17292 GCC support of fixed-point types as specified by the draft technical
17293 report is incomplete:
17295 * Pragmas to control overflow and rounding behaviors are not
17298 Fixed-point types are supported by the DWARF2 debug information format.
17301 File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Fixed-Point, Up: C Extensions
17303 5.14 Arrays of Length Zero
17304 ==========================
17306 Zero-length arrays are allowed in GNU C. They are very useful as the
17307 last element of a structure which is really a header for a
17308 variable-length object:
17315 struct line *thisline = (struct line *)
17316 malloc (sizeof (struct line) + this_length);
17317 thisline->length = this_length;
17319 In ISO C90, you would have to give `contents' a length of 1, which
17320 means either you waste space or complicate the argument to `malloc'.
17322 In ISO C99, you would use a "flexible array member", which is slightly
17323 different in syntax and semantics:
17325 * Flexible array members are written as `contents[]' without the `0'.
17327 * Flexible array members have incomplete type, and so the `sizeof'
17328 operator may not be applied. As a quirk of the original
17329 implementation of zero-length arrays, `sizeof' evaluates to zero.
17331 * Flexible array members may only appear as the last member of a
17332 `struct' that is otherwise non-empty.
17334 * A structure containing a flexible array member, or a union
17335 containing such a structure (possibly recursively), may not be a
17336 member of a structure or an element of an array. (However, these
17337 uses are permitted by GCC as extensions.)
17339 GCC versions before 3.0 allowed zero-length arrays to be statically
17340 initialized, as if they were flexible arrays. In addition to those
17341 cases that were useful, it also allowed initializations in situations
17342 that would corrupt later data. Non-empty initialization of zero-length
17343 arrays is now treated like any case where there are more initializer
17344 elements than the array holds, in that a suitable warning about "excess
17345 elements in array" is given, and the excess elements (all of them, in
17346 this case) are ignored.
17348 Instead GCC allows static initialization of flexible array members.
17349 This is equivalent to defining a new structure containing the original
17350 structure followed by an array of sufficient size to contain the data.
17351 I.e. in the following, `f1' is constructed as if it were declared like
17356 } f1 = { 1, { 2, 3, 4 } };
17359 struct f1 f1; int data[3];
17360 } f2 = { { 1 }, { 2, 3, 4 } };
17362 The convenience of this extension is that `f1' has the desired type,
17363 eliminating the need to consistently refer to `f2.f1'.
17365 This has symmetry with normal static arrays, in that an array of
17366 unknown size is also written with `[]'.
17368 Of course, this extension only makes sense if the extra data comes at
17369 the end of a top-level object, as otherwise we would be overwriting
17370 data at subsequent offsets. To avoid undue complication and confusion
17371 with initialization of deeply nested arrays, we simply disallow any
17372 non-empty initialization except when the structure is the top-level
17373 object. For example:
17375 struct foo { int x; int y[]; };
17376 struct bar { struct foo z; };
17378 struct foo a = { 1, { 2, 3, 4 } }; // Valid.
17379 struct bar b = { { 1, { 2, 3, 4 } } }; // Invalid.
17380 struct bar c = { { 1, { } } }; // Valid.
17381 struct foo d[1] = { { 1 { 2, 3, 4 } } }; // Invalid.
17384 File: gcc.info, Node: Empty Structures, Next: Variadic Macros, Prev: Variable Length, Up: C Extensions
17386 5.15 Structures With No Members
17387 ===============================
17389 GCC permits a C structure to have no members:
17394 The structure will have size zero. In C++, empty structures are part
17395 of the language. G++ treats empty structures as if they had a single
17396 member of type `char'.
17399 File: gcc.info, Node: Variable Length, Next: Empty Structures, Prev: Zero Length, Up: C Extensions
17401 5.16 Arrays of Variable Length
17402 ==============================
17404 Variable-length automatic arrays are allowed in ISO C99, and as an
17405 extension GCC accepts them in C89 mode and in C++. (However, GCC's
17406 implementation of variable-length arrays does not yet conform in detail
17407 to the ISO C99 standard.) These arrays are declared like any other
17408 automatic arrays, but with a length that is not a constant expression.
17409 The storage is allocated at the point of declaration and deallocated
17410 when the brace-level is exited. For example:
17413 concat_fopen (char *s1, char *s2, char *mode)
17415 char str[strlen (s1) + strlen (s2) + 1];
17418 return fopen (str, mode);
17421 Jumping or breaking out of the scope of the array name deallocates the
17422 storage. Jumping into the scope is not allowed; you get an error
17425 You can use the function `alloca' to get an effect much like
17426 variable-length arrays. The function `alloca' is available in many
17427 other C implementations (but not in all). On the other hand,
17428 variable-length arrays are more elegant.
17430 There are other differences between these two methods. Space allocated
17431 with `alloca' exists until the containing _function_ returns. The
17432 space for a variable-length array is deallocated as soon as the array
17433 name's scope ends. (If you use both variable-length arrays and
17434 `alloca' in the same function, deallocation of a variable-length array
17435 will also deallocate anything more recently allocated with `alloca'.)
17437 You can also use variable-length arrays as arguments to functions:
17440 tester (int len, char data[len][len])
17445 The length of an array is computed once when the storage is allocated
17446 and is remembered for the scope of the array in case you access it with
17449 If you want to pass the array first and the length afterward, you can
17450 use a forward declaration in the parameter list--another GNU extension.
17453 tester (int len; char data[len][len], int len)
17458 The `int len' before the semicolon is a "parameter forward
17459 declaration", and it serves the purpose of making the name `len' known
17460 when the declaration of `data' is parsed.
17462 You can write any number of such parameter forward declarations in the
17463 parameter list. They can be separated by commas or semicolons, but the
17464 last one must end with a semicolon, which is followed by the "real"
17465 parameter declarations. Each forward declaration must match a "real"
17466 declaration in parameter name and data type. ISO C99 does not support
17467 parameter forward declarations.
17470 File: gcc.info, Node: Variadic Macros, Next: Escaped Newlines, Prev: Empty Structures, Up: C Extensions
17472 5.17 Macros with a Variable Number of Arguments.
17473 ================================================
17475 In the ISO C standard of 1999, a macro can be declared to accept a
17476 variable number of arguments much as a function can. The syntax for
17477 defining the macro is similar to that of a function. Here is an
17480 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
17482 Here `...' is a "variable argument". In the invocation of such a
17483 macro, it represents the zero or more tokens until the closing
17484 parenthesis that ends the invocation, including any commas. This set of
17485 tokens replaces the identifier `__VA_ARGS__' in the macro body wherever
17486 it appears. See the CPP manual for more information.
17488 GCC has long supported variadic macros, and used a different syntax
17489 that allowed you to give a name to the variable arguments just like any
17490 other argument. Here is an example:
17492 #define debug(format, args...) fprintf (stderr, format, args)
17494 This is in all ways equivalent to the ISO C example above, but arguably
17495 more readable and descriptive.
17497 GNU CPP has two further variadic macro extensions, and permits them to
17498 be used with either of the above forms of macro definition.
17500 In standard C, you are not allowed to leave the variable argument out
17501 entirely; but you are allowed to pass an empty argument. For example,
17502 this invocation is invalid in ISO C, because there is no comma after
17505 debug ("A message")
17507 GNU CPP permits you to completely omit the variable arguments in this
17508 way. In the above examples, the compiler would complain, though since
17509 the expansion of the macro still has the extra comma after the format
17512 To help solve this problem, CPP behaves specially for variable
17513 arguments used with the token paste operator, `##'. If instead you
17516 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
17518 and if the variable arguments are omitted or empty, the `##' operator
17519 causes the preprocessor to remove the comma before it. If you do
17520 provide some variable arguments in your macro invocation, GNU CPP does
17521 not complain about the paste operation and instead places the variable
17522 arguments after the comma. Just like any other pasted macro argument,
17523 these arguments are not macro expanded.
17526 File: gcc.info, Node: Escaped Newlines, Next: Subscripting, Prev: Variadic Macros, Up: C Extensions
17528 5.18 Slightly Looser Rules for Escaped Newlines
17529 ===============================================
17531 Recently, the preprocessor has relaxed its treatment of escaped
17532 newlines. Previously, the newline had to immediately follow a
17533 backslash. The current implementation allows whitespace in the form of
17534 spaces, horizontal and vertical tabs, and form feeds between the
17535 backslash and the subsequent newline. The preprocessor issues a
17536 warning, but treats it as a valid escaped newline and combines the two
17537 lines to form a single logical line. This works within comments and
17538 tokens, as well as between tokens. Comments are _not_ treated as
17539 whitespace for the purposes of this relaxation, since they have not yet
17540 been replaced with spaces.
17543 File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Escaped Newlines, Up: C Extensions
17545 5.19 Non-Lvalue Arrays May Have Subscripts
17546 ==========================================
17548 In ISO C99, arrays that are not lvalues still decay to pointers, and
17549 may be subscripted, although they may not be modified or used after the
17550 next sequence point and the unary `&' operator may not be applied to
17551 them. As an extension, GCC allows such arrays to be subscripted in C89
17552 mode, though otherwise they do not decay to pointers outside C99 mode.
17553 For example, this is valid in GNU C though not valid in C89:
17555 struct foo {int a[4];};
17561 return f().a[index];
17565 File: gcc.info, Node: Pointer Arith, Next: Initializers, Prev: Subscripting, Up: C Extensions
17567 5.20 Arithmetic on `void'- and Function-Pointers
17568 ================================================
17570 In GNU C, addition and subtraction operations are supported on pointers
17571 to `void' and on pointers to functions. This is done by treating the
17572 size of a `void' or of a function as 1.
17574 A consequence of this is that `sizeof' is also allowed on `void' and
17575 on function types, and returns 1.
17577 The option `-Wpointer-arith' requests a warning if these extensions
17581 File: gcc.info, Node: Initializers, Next: Compound Literals, Prev: Pointer Arith, Up: C Extensions
17583 5.21 Non-Constant Initializers
17584 ==============================
17586 As in standard C++ and ISO C99, the elements of an aggregate
17587 initializer for an automatic variable are not required to be constant
17588 expressions in GNU C. Here is an example of an initializer with
17589 run-time varying elements:
17591 foo (float f, float g)
17593 float beat_freqs[2] = { f-g, f+g };
17598 File: gcc.info, Node: Compound Literals, Next: Designated Inits, Prev: Initializers, Up: C Extensions
17600 5.22 Compound Literals
17601 ======================
17603 ISO C99 supports compound literals. A compound literal looks like a
17604 cast containing an initializer. Its value is an object of the type
17605 specified in the cast, containing the elements specified in the
17606 initializer; it is an lvalue. As an extension, GCC supports compound
17607 literals in C89 mode and in C++.
17609 Usually, the specified type is a structure. Assume that `struct foo'
17610 and `structure' are declared as shown:
17612 struct foo {int a; char b[2];} structure;
17614 Here is an example of constructing a `struct foo' with a compound
17617 structure = ((struct foo) {x + y, 'a', 0});
17619 This is equivalent to writing the following:
17622 struct foo temp = {x + y, 'a', 0};
17626 You can also construct an array. If all the elements of the compound
17627 literal are (made up of) simple constant expressions, suitable for use
17628 in initializers of objects of static storage duration, then the compound
17629 literal can be coerced to a pointer to its first element and used in
17630 such an initializer, as shown here:
17632 char **foo = (char *[]) { "x", "y", "z" };
17634 Compound literals for scalar types and union types are is also
17635 allowed, but then the compound literal is equivalent to a cast.
17637 As a GNU extension, GCC allows initialization of objects with static
17638 storage duration by compound literals (which is not possible in ISO
17639 C99, because the initializer is not a constant). It is handled as if
17640 the object was initialized only with the bracket enclosed list if the
17641 types of the compound literal and the object match. The initializer
17642 list of the compound literal must be constant. If the object being
17643 initialized has array type of unknown size, the size is determined by
17644 compound literal size.
17646 static struct foo x = (struct foo) {1, 'a', 'b'};
17647 static int y[] = (int []) {1, 2, 3};
17648 static int z[] = (int [3]) {1};
17650 The above lines are equivalent to the following:
17651 static struct foo x = {1, 'a', 'b'};
17652 static int y[] = {1, 2, 3};
17653 static int z[] = {1, 0, 0};
17656 File: gcc.info, Node: Designated Inits, Next: Cast to Union, Prev: Compound Literals, Up: C Extensions
17658 5.23 Designated Initializers
17659 ============================
17661 Standard C89 requires the elements of an initializer to appear in a
17662 fixed order, the same as the order of the elements in the array or
17663 structure being initialized.
17665 In ISO C99 you can give the elements in any order, specifying the array
17666 indices or structure field names they apply to, and GNU C allows this as
17667 an extension in C89 mode as well. This extension is not implemented in
17670 To specify an array index, write `[INDEX] =' before the element value.
17673 int a[6] = { [4] = 29, [2] = 15 };
17677 int a[6] = { 0, 0, 15, 0, 29, 0 };
17679 The index values must be constant expressions, even if the array being
17680 initialized is automatic.
17682 An alternative syntax for this which has been obsolete since GCC 2.5
17683 but GCC still accepts is to write `[INDEX]' before the element value,
17686 To initialize a range of elements to the same value, write `[FIRST ...
17687 LAST] = VALUE'. This is a GNU extension. For example,
17689 int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
17691 If the value in it has side-effects, the side-effects will happen only
17692 once, not for each initialized field by the range initializer.
17694 Note that the length of the array is the highest value specified plus
17697 In a structure initializer, specify the name of a field to initialize
17698 with `.FIELDNAME =' before the element value. For example, given the
17699 following structure,
17701 struct point { int x, y; };
17703 the following initialization
17705 struct point p = { .y = yvalue, .x = xvalue };
17709 struct point p = { xvalue, yvalue };
17711 Another syntax which has the same meaning, obsolete since GCC 2.5, is
17712 `FIELDNAME:', as shown here:
17714 struct point p = { y: yvalue, x: xvalue };
17716 The `[INDEX]' or `.FIELDNAME' is known as a "designator". You can
17717 also use a designator (or the obsolete colon syntax) when initializing
17718 a union, to specify which element of the union should be used. For
17721 union foo { int i; double d; };
17723 union foo f = { .d = 4 };
17725 will convert 4 to a `double' to store it in the union using the second
17726 element. By contrast, casting 4 to type `union foo' would store it
17727 into the union as the integer `i', since it is an integer. (*Note Cast
17730 You can combine this technique of naming elements with ordinary C
17731 initialization of successive elements. Each initializer element that
17732 does not have a designator applies to the next consecutive element of
17733 the array or structure. For example,
17735 int a[6] = { [1] = v1, v2, [4] = v4 };
17739 int a[6] = { 0, v1, v2, 0, v4, 0 };
17741 Labeling the elements of an array initializer is especially useful
17742 when the indices are characters or belong to an `enum' type. For
17745 int whitespace[256]
17746 = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
17747 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
17749 You can also write a series of `.FIELDNAME' and `[INDEX]' designators
17750 before an `=' to specify a nested subobject to initialize; the list is
17751 taken relative to the subobject corresponding to the closest
17752 surrounding brace pair. For example, with the `struct point'
17755 struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
17757 If the same field is initialized multiple times, it will have value from
17758 the last initialization. If any such overridden initialization has
17759 side-effect, it is unspecified whether the side-effect happens or not.
17760 Currently, GCC will discard them and issue a warning.
17763 File: gcc.info, Node: Case Ranges, Next: Mixed Declarations, Prev: Cast to Union, Up: C Extensions
17768 You can specify a range of consecutive values in a single `case' label,
17773 This has the same effect as the proper number of individual `case'
17774 labels, one for each integer value from LOW to HIGH, inclusive.
17776 This feature is especially useful for ranges of ASCII character codes:
17780 *Be careful:* Write spaces around the `...', for otherwise it may be
17781 parsed wrong when you use it with integer values. For example, write
17791 File: gcc.info, Node: Cast to Union, Next: Case Ranges, Prev: Designated Inits, Up: C Extensions
17793 5.25 Cast to a Union Type
17794 =========================
17796 A cast to union type is similar to other casts, except that the type
17797 specified is a union type. You can specify the type either with `union
17798 TAG' or with a typedef name. A cast to union is actually a constructor
17799 though, not a cast, and hence does not yield an lvalue like normal
17800 casts. (*Note Compound Literals::.)
17802 The types that may be cast to the union type are those of the members
17803 of the union. Thus, given the following union and variables:
17805 union foo { int i; double d; };
17809 both `x' and `y' can be cast to type `union foo'.
17811 Using the cast as the right-hand side of an assignment to a variable of
17812 union type is equivalent to storing in a member of the union:
17816 u = (union foo) x == u.i = x
17817 u = (union foo) y == u.d = y
17819 You can also use the union cast as a function argument:
17821 void hack (union foo);
17823 hack ((union foo) x);
17826 File: gcc.info, Node: Mixed Declarations, Next: Function Attributes, Prev: Case Ranges, Up: C Extensions
17828 5.26 Mixed Declarations and Code
17829 ================================
17831 ISO C99 and ISO C++ allow declarations and code to be freely mixed
17832 within compound statements. As an extension, GCC also allows this in
17833 C89 mode. For example, you could do:
17840 Each identifier is visible from where it is declared until the end of
17841 the enclosing block.
17844 File: gcc.info, Node: Function Attributes, Next: Attribute Syntax, Prev: Mixed Declarations, Up: C Extensions
17846 5.27 Declaring Attributes of Functions
17847 ======================================
17849 In GNU C, you declare certain things about functions called in your
17850 program which help the compiler optimize function calls and check your
17851 code more carefully.
17853 The keyword `__attribute__' allows you to specify special attributes
17854 when making a declaration. This keyword is followed by an attribute
17855 specification inside double parentheses. The following attributes are
17856 currently defined for functions on all targets: `aligned',
17857 `alloc_size', `noreturn', `returns_twice', `noinline', `always_inline',
17858 `flatten', `pure', `const', `nothrow', `sentinel', `format',
17859 `format_arg', `no_instrument_function', `section', `constructor',
17860 `destructor', `used', `unused', `deprecated', `weak', `malloc',
17861 `alias', `warn_unused_result', `nonnull', `gnu_inline',
17862 `externally_visible', `hot', `cold', `artificial', `error' and
17863 `warning'. Several other attributes are defined for functions on
17864 particular target systems. Other attributes, including `section' are
17865 supported for variables declarations (*note Variable Attributes::) and
17866 for types (*note Type Attributes::).
17868 You may also specify attributes with `__' preceding and following each
17869 keyword. This allows you to use them in header files without being
17870 concerned about a possible macro of the same name. For example, you
17871 may use `__noreturn__' instead of `noreturn'.
17873 *Note Attribute Syntax::, for details of the exact syntax for using
17877 The `alias' attribute causes the declaration to be emitted as an
17878 alias for another symbol, which must be specified. For instance,
17880 void __f () { /* Do something. */; }
17881 void f () __attribute__ ((weak, alias ("__f")));
17883 defines `f' to be a weak alias for `__f'. In C++, the mangled
17884 name for the target must be used. It is an error if `__f' is not
17885 defined in the same translation unit.
17887 Not all target machines support this attribute.
17889 `aligned (ALIGNMENT)'
17890 This attribute specifies a minimum alignment for the function,
17893 You cannot use this attribute to decrease the alignment of a
17894 function, only to increase it. However, when you explicitly
17895 specify a function alignment this will override the effect of the
17896 `-falign-functions' (*note Optimize Options::) option for this
17899 Note that the effectiveness of `aligned' attributes may be limited
17900 by inherent limitations in your linker. On many systems, the
17901 linker is only able to arrange for functions to be aligned up to a
17902 certain maximum alignment. (For some linkers, the maximum
17903 supported alignment may be very very small.) See your linker
17904 documentation for further information.
17906 The `aligned' attribute can also be used for variables and fields
17907 (*note Variable Attributes::.)
17910 The `alloc_size' attribute is used to tell the compiler that the
17911 function return value points to memory, where the size is given by
17912 one or two of the functions parameters. GCC uses this information
17913 to improve the correctness of `__builtin_object_size'.
17915 The function parameter(s) denoting the allocated size are
17916 specified by one or two integer arguments supplied to the
17917 attribute. The allocated size is either the value of the single
17918 function argument specified or the product of the two function
17919 arguments specified. Argument numbering starts at one.
17923 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
17924 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
17926 declares that my_calloc will return memory of the size given by
17927 the product of parameter 1 and 2 and that my_realloc will return
17928 memory of the size given by parameter 2.
17931 Generally, functions are not inlined unless optimization is
17932 specified. For functions declared inline, this attribute inlines
17933 the function even if no optimization level was specified.
17936 This attribute should be used with a function which is also
17937 declared with the `inline' keyword. It directs GCC to treat the
17938 function as if it were defined in gnu89 mode even when compiling
17939 in C99 or gnu99 mode.
17941 If the function is declared `extern', then this definition of the
17942 function is used only for inlining. In no case is the function
17943 compiled as a standalone function, not even if you take its address
17944 explicitly. Such an address becomes an external reference, as if
17945 you had only declared the function, and had not defined it. This
17946 has almost the effect of a macro. The way to use this is to put a
17947 function definition in a header file with this attribute, and put
17948 another copy of the function, without `extern', in a library file.
17949 The definition in the header file will cause most calls to the
17950 function to be inlined. If any uses of the function remain, they
17951 will refer to the single copy in the library. Note that the two
17952 definitions of the functions need not be precisely the same,
17953 although if they do not have the same effect your program may
17956 In C, if the function is neither `extern' nor `static', then the
17957 function is compiled as a standalone function, as well as being
17958 inlined where possible.
17960 This is how GCC traditionally handled functions declared `inline'.
17961 Since ISO C99 specifies a different semantics for `inline', this
17962 function attribute is provided as a transition measure and as a
17963 useful feature in its own right. This attribute is available in
17964 GCC 4.1.3 and later. It is available if either of the
17965 preprocessor macros `__GNUC_GNU_INLINE__' or
17966 `__GNUC_STDC_INLINE__' are defined. *Note An Inline Function is
17967 As Fast As a Macro: Inline.
17969 In C++, this attribute does not depend on `extern' in any way, but
17970 it still requires the `inline' keyword to enable its special
17974 This attribute is useful for small inline wrappers which if
17975 possible should appear during debugging as a unit, depending on
17976 the debug info format it will either mean marking the function as
17977 artificial or using the caller location for all instructions
17978 within the inlined body.
17981 Generally, inlining into a function is limited. For a function
17982 marked with this attribute, every call inside this function will
17983 be inlined, if possible. Whether the function itself is
17984 considered for inlining depends on its size and the current
17985 inlining parameters.
17987 `error ("MESSAGE")'
17988 If this attribute is used on a function declaration and a call to
17989 such a function is not eliminated through dead code elimination or
17990 other optimizations, an error which will include MESSAGE will be
17991 diagnosed. This is useful for compile time checking, especially
17992 together with `__builtin_constant_p' and inline functions where
17993 checking the inline function arguments is not possible through
17994 `extern char [(condition) ? 1 : -1];' tricks. While it is
17995 possible to leave the function undefined and thus invoke a link
17996 failure, when using this attribute the problem will be diagnosed
17997 earlier and with exact location of the call even in presence of
17998 inline functions or when not emitting debugging information.
18000 `warning ("MESSAGE")'
18001 If this attribute is used on a function declaration and a call to
18002 such a function is not eliminated through dead code elimination or
18003 other optimizations, a warning which will include MESSAGE will be
18004 diagnosed. This is useful for compile time checking, especially
18005 together with `__builtin_constant_p' and inline functions. While
18006 it is possible to define the function with a message in
18007 `.gnu.warning*' section, when using this attribute the problem
18008 will be diagnosed earlier and with exact location of the call even
18009 in presence of inline functions or when not emitting debugging
18013 On the Intel 386, the `cdecl' attribute causes the compiler to
18014 assume that the calling function will pop off the stack space used
18015 to pass arguments. This is useful to override the effects of the
18019 Many functions do not examine any values except their arguments,
18020 and have no effects except the return value. Basically this is
18021 just slightly more strict class than the `pure' attribute below,
18022 since function is not allowed to read global memory.
18024 Note that a function that has pointer arguments and examines the
18025 data pointed to must _not_ be declared `const'. Likewise, a
18026 function that calls a non-`const' function usually must not be
18027 `const'. It does not make sense for a `const' function to return
18030 The attribute `const' is not implemented in GCC versions earlier
18031 than 2.5. An alternative way to declare that a function has no
18032 side effects, which works in the current version and in some older
18033 versions, is as follows:
18035 typedef int intfn ();
18037 extern const intfn square;
18039 This approach does not work in GNU C++ from 2.6.0 on, since the
18040 language specifies that the `const' must be attached to the return
18045 `constructor (PRIORITY)'
18046 `destructor (PRIORITY)'
18047 The `constructor' attribute causes the function to be called
18048 automatically before execution enters `main ()'. Similarly, the
18049 `destructor' attribute causes the function to be called
18050 automatically after `main ()' has completed or `exit ()' has been
18051 called. Functions with these attributes are useful for
18052 initializing data that will be used implicitly during the
18053 execution of the program.
18055 You may provide an optional integer priority to control the order
18056 in which constructor and destructor functions are run. A
18057 constructor with a smaller priority number runs before a
18058 constructor with a larger priority number; the opposite
18059 relationship holds for destructors. So, if you have a constructor
18060 that allocates a resource and a destructor that deallocates the
18061 same resource, both functions typically have the same priority.
18062 The priorities for constructor and destructor functions are the
18063 same as those specified for namespace-scope C++ objects (*note C++
18066 These attributes are not currently implemented for Objective-C.
18069 The `deprecated' attribute results in a warning if the function is
18070 used anywhere in the source file. This is useful when identifying
18071 functions that are expected to be removed in a future version of a
18072 program. The warning also includes the location of the declaration
18073 of the deprecated function, to enable users to easily find further
18074 information about why the function is deprecated, or what they
18075 should do instead. Note that the warnings only occurs for uses:
18077 int old_fn () __attribute__ ((deprecated));
18079 int (*fn_ptr)() = old_fn;
18081 results in a warning on line 3 but not line 2.
18083 The `deprecated' attribute can also be used for variables and
18084 types (*note Variable Attributes::, *note Type Attributes::.)
18087 On Microsoft Windows targets and Symbian OS targets the
18088 `dllexport' attribute causes the compiler to provide a global
18089 pointer to a pointer in a DLL, so that it can be referenced with
18090 the `dllimport' attribute. On Microsoft Windows targets, the
18091 pointer name is formed by combining `_imp__' and the function or
18094 You can use `__declspec(dllexport)' as a synonym for
18095 `__attribute__ ((dllexport))' for compatibility with other
18098 On systems that support the `visibility' attribute, this attribute
18099 also implies "default" visibility. It is an error to explicitly
18100 specify any other visibility.
18102 Currently, the `dllexport' attribute is ignored for inlined
18103 functions, unless the `-fkeep-inline-functions' flag has been
18104 used. The attribute is also ignored for undefined symbols.
18106 When applied to C++ classes, the attribute marks defined
18107 non-inlined member functions and static data members as exports.
18108 Static consts initialized in-class are not marked unless they are
18109 also defined out-of-class.
18111 For Microsoft Windows targets there are alternative methods for
18112 including the symbol in the DLL's export table such as using a
18113 `.def' file with an `EXPORTS' section or, with GNU ld, using the
18114 `--export-all' linker flag.
18117 On Microsoft Windows and Symbian OS targets, the `dllimport'
18118 attribute causes the compiler to reference a function or variable
18119 via a global pointer to a pointer that is set up by the DLL
18120 exporting the symbol. The attribute implies `extern'. On
18121 Microsoft Windows targets, the pointer name is formed by combining
18122 `_imp__' and the function or variable name.
18124 You can use `__declspec(dllimport)' as a synonym for
18125 `__attribute__ ((dllimport))' for compatibility with other
18128 On systems that support the `visibility' attribute, this attribute
18129 also implies "default" visibility. It is an error to explicitly
18130 specify any other visibility.
18132 Currently, the attribute is ignored for inlined functions. If the
18133 attribute is applied to a symbol _definition_, an error is
18134 reported. If a symbol previously declared `dllimport' is later
18135 defined, the attribute is ignored in subsequent references, and a
18136 warning is emitted. The attribute is also overridden by a
18137 subsequent declaration as `dllexport'.
18139 When applied to C++ classes, the attribute marks non-inlined
18140 member functions and static data members as imports. However, the
18141 attribute is ignored for virtual methods to allow creation of
18142 vtables using thunks.
18144 On the SH Symbian OS target the `dllimport' attribute also has
18145 another affect--it can cause the vtable and run-time type
18146 information for a class to be exported. This happens when the
18147 class has a dllimport'ed constructor or a non-inline, non-pure
18148 virtual function and, for either of those two conditions, the
18149 class also has a inline constructor or destructor and has a key
18150 function that is defined in the current translation unit.
18152 For Microsoft Windows based targets the use of the `dllimport'
18153 attribute on functions is not necessary, but provides a small
18154 performance benefit by eliminating a thunk in the DLL. The use of
18155 the `dllimport' attribute on imported variables was required on
18156 older versions of the GNU linker, but can now be avoided by
18157 passing the `--enable-auto-import' switch to the GNU linker. As
18158 with functions, using the attribute for a variable eliminates a
18161 One drawback to using this attribute is that a pointer to a
18162 _variable_ marked as `dllimport' cannot be used as a constant
18163 address. However, a pointer to a _function_ with the `dllimport'
18164 attribute can be used as a constant initializer; in this case, the
18165 address of a stub function in the import lib is referenced. On
18166 Microsoft Windows targets, the attribute can be disabled for
18167 functions by setting the `-mnop-fun-dllimport' flag.
18170 Use this attribute on the H8/300, H8/300H, and H8S to indicate
18171 that the specified variable should be placed into the eight bit
18172 data section. The compiler will generate more efficient code for
18173 certain operations on data in the eight bit data area. Note the
18174 eight bit data area is limited to 256 bytes of data.
18176 You must use GAS and GLD from GNU binutils version 2.7 or later for
18177 this attribute to work correctly.
18179 `exception_handler'
18180 Use this attribute on the Blackfin to indicate that the specified
18181 function is an exception handler. The compiler will generate
18182 function entry and exit sequences suitable for use in an exception
18183 handler when this attribute is present.
18185 `externally_visible'
18186 This attribute, attached to a global variable or function,
18187 nullifies the effect of the `-fwhole-program' command-line option,
18188 so the object remains visible outside the current compilation unit.
18191 On 68HC11 and 68HC12 the `far' attribute causes the compiler to
18192 use a calling convention that takes care of switching memory banks
18193 when entering and leaving a function. This calling convention is
18194 also the default when using the `-mlong-calls' option.
18196 On 68HC12 the compiler will use the `call' and `rtc' instructions
18197 to call and return from a function.
18199 On 68HC11 the compiler will generate a sequence of instructions to
18200 invoke a board-specific routine to switch the memory bank and call
18201 the real function. The board-specific routine simulates a `call'.
18202 At the end of a function, it will jump to a board-specific routine
18203 instead of using `rts'. The board-specific return routine
18204 simulates the `rtc'.
18207 On the Intel 386, the `fastcall' attribute causes the compiler to
18208 pass the first argument (if of integral type) in the register ECX
18209 and the second argument (if of integral type) in the register EDX.
18210 Subsequent and other typed arguments are passed on the stack.
18211 The called function will pop the arguments off the stack. If the
18212 number of arguments is variable all arguments are pushed on the
18215 `format (ARCHETYPE, STRING-INDEX, FIRST-TO-CHECK)'
18216 The `format' attribute specifies that a function takes `printf',
18217 `scanf', `strftime' or `strfmon' style arguments which should be
18218 type-checked against a format string. For example, the
18222 my_printf (void *my_object, const char *my_format, ...)
18223 __attribute__ ((format (printf, 2, 3)));
18225 causes the compiler to check the arguments in calls to `my_printf'
18226 for consistency with the `printf' style format string argument
18229 The parameter ARCHETYPE determines how the format string is
18230 interpreted, and should be `printf', `scanf', `strftime',
18231 `gnu_printf', `gnu_scanf', `gnu_strftime' or `strfmon'. (You can
18232 also use `__printf__', `__scanf__', `__strftime__' or
18233 `__strfmon__'.) On MinGW targets, `ms_printf', `ms_scanf', and
18234 `ms_strftime' are also present. ARCHTYPE values such as `printf'
18235 refer to the formats accepted by the system's C run-time library,
18236 while `gnu_' values always refer to the formats accepted by the
18237 GNU C Library. On Microsoft Windows targets, `ms_' values refer
18238 to the formats accepted by the `msvcrt.dll' library. The
18239 parameter STRING-INDEX specifies which argument is the format
18240 string argument (starting from 1), while FIRST-TO-CHECK is the
18241 number of the first argument to check against the format string.
18242 For functions where the arguments are not available to be checked
18243 (such as `vprintf'), specify the third parameter as zero. In this
18244 case the compiler only checks the format string for consistency.
18245 For `strftime' formats, the third parameter is required to be zero.
18246 Since non-static C++ methods have an implicit `this' argument, the
18247 arguments of such methods should be counted from two, not one, when
18248 giving values for STRING-INDEX and FIRST-TO-CHECK.
18250 In the example above, the format string (`my_format') is the second
18251 argument of the function `my_print', and the arguments to check
18252 start with the third argument, so the correct parameters for the
18253 format attribute are 2 and 3.
18255 The `format' attribute allows you to identify your own functions
18256 which take format strings as arguments, so that GCC can check the
18257 calls to these functions for errors. The compiler always (unless
18258 `-ffreestanding' or `-fno-builtin' is used) checks formats for the
18259 standard library functions `printf', `fprintf', `sprintf',
18260 `scanf', `fscanf', `sscanf', `strftime', `vprintf', `vfprintf' and
18261 `vsprintf' whenever such warnings are requested (using
18262 `-Wformat'), so there is no need to modify the header file
18263 `stdio.h'. In C99 mode, the functions `snprintf', `vsnprintf',
18264 `vscanf', `vfscanf' and `vsscanf' are also checked. Except in
18265 strictly conforming C standard modes, the X/Open function
18266 `strfmon' is also checked as are `printf_unlocked' and
18267 `fprintf_unlocked'. *Note Options Controlling C Dialect: C
18270 The target may provide additional types of format checks. *Note
18271 Format Checks Specific to Particular Target Machines: Target
18274 `format_arg (STRING-INDEX)'
18275 The `format_arg' attribute specifies that a function takes a format
18276 string for a `printf', `scanf', `strftime' or `strfmon' style
18277 function and modifies it (for example, to translate it into
18278 another language), so the result can be passed to a `printf',
18279 `scanf', `strftime' or `strfmon' style function (with the
18280 remaining arguments to the format function the same as they would
18281 have been for the unmodified string). For example, the
18285 my_dgettext (char *my_domain, const char *my_format)
18286 __attribute__ ((format_arg (2)));
18288 causes the compiler to check the arguments in calls to a `printf',
18289 `scanf', `strftime' or `strfmon' type function, whose format
18290 string argument is a call to the `my_dgettext' function, for
18291 consistency with the format string argument `my_format'. If the
18292 `format_arg' attribute had not been specified, all the compiler
18293 could tell in such calls to format functions would be that the
18294 format string argument is not constant; this would generate a
18295 warning when `-Wformat-nonliteral' is used, but the calls could
18296 not be checked without the attribute.
18298 The parameter STRING-INDEX specifies which argument is the format
18299 string argument (starting from one). Since non-static C++ methods
18300 have an implicit `this' argument, the arguments of such methods
18301 should be counted from two.
18303 The `format-arg' attribute allows you to identify your own
18304 functions which modify format strings, so that GCC can check the
18305 calls to `printf', `scanf', `strftime' or `strfmon' type function
18306 whose operands are a call to one of your own function. The
18307 compiler always treats `gettext', `dgettext', and `dcgettext' in
18308 this manner except when strict ISO C support is requested by
18309 `-ansi' or an appropriate `-std' option, or `-ffreestanding' or
18310 `-fno-builtin' is used. *Note Options Controlling C Dialect: C
18314 Use this attribute on the H8/300, H8/300H, and H8S to indicate
18315 that the specified function should be called through the function
18316 vector. Calling a function through the function vector will
18317 reduce code size, however; the function vector has a limited size
18318 (maximum 128 entries on the H8/300 and 64 entries on the H8/300H
18319 and H8S) and shares space with the interrupt vector.
18321 In SH2A target, this attribute declares a function to be called
18322 using the TBR relative addressing mode. The argument to this
18323 attribute is the entry number of the same function in a vector
18324 table containing all the TBR relative addressable functions. For
18325 the successful jump, register TBR should contain the start address
18326 of this TBR relative vector table. In the startup routine of the
18327 user application, user needs to care of this TBR register
18328 initialization. The TBR relative vector table can have at max 256
18329 function entries. The jumps to these functions will be generated
18330 using a SH2A specific, non delayed branch instruction JSR/N
18331 @(disp8,TBR). You must use GAS and GLD from GNU binutils version
18332 2.7 or later for this attribute to work correctly.
18334 Please refer the example of M16C target, to see the use of this
18335 attribute while declaring a function,
18337 In an application, for a function being called once, this
18338 attribute will save at least 8 bytes of code; and if other
18339 successive calls are being made to the same function, it will save
18340 2 bytes of code per each of these calls.
18342 On M16C/M32C targets, the `function_vector' attribute declares a
18343 special page subroutine call function. Use of this attribute
18344 reduces the code size by 2 bytes for each call generated to the
18345 subroutine. The argument to the attribute is the vector number
18346 entry from the special page vector table which contains the 16
18347 low-order bits of the subroutine's entry address. Each vector
18348 table has special page number (18 to 255) which are used in `jsrs'
18349 instruction. Jump addresses of the routines are generated by
18350 adding 0x0F0000 (in case of M16C targets) or 0xFF0000 (in case of
18351 M32C targets), to the 2 byte addresses set in the vector table.
18352 Therefore you need to ensure that all the special page vector
18353 routines should get mapped within the address range 0x0F0000 to
18354 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF (for M32C).
18356 In the following example 2 bytes will be saved for each call to
18359 void foo (void) __attribute__((function_vector(0x18)));
18369 If functions are defined in one file and are called in another
18370 file, then be sure to write this declaration in both files.
18372 This attribute is ignored for R8C target.
18375 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k, and
18376 Xstormy16 ports to indicate that the specified function is an
18377 interrupt handler. The compiler will generate function entry and
18378 exit sequences suitable for use in an interrupt handler when this
18379 attribute is present.
18381 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S,
18382 and SH processors can be specified via the `interrupt_handler'
18385 Note, on the AVR, interrupts will be enabled inside the function.
18387 Note, for the ARM, you can specify the kind of interrupt to be
18388 handled by adding an optional parameter to the interrupt attribute
18391 void f () __attribute__ ((interrupt ("IRQ")));
18393 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT
18396 On ARMv7-M the interrupt type is ignored, and the attribute means
18397 the function may be called with a word aligned stack pointer.
18399 `interrupt_handler'
18400 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S,
18401 and SH to indicate that the specified function is an interrupt
18402 handler. The compiler will generate function entry and exit
18403 sequences suitable for use in an interrupt handler when this
18404 attribute is present.
18407 Use this attribute on fido, a subarchitecture of the m68k, to
18408 indicate that the specified function is an interrupt handler that
18409 is designed to run as a thread. The compiler omits generate
18410 prologue/epilogue sequences and replaces the return instruction
18411 with a `sleep' instruction. This attribute is available only on
18415 Use this attribute on ARM to write Interrupt Service Routines.
18416 This is an alias to the `interrupt' attribute above.
18419 When used together with `interrupt_handler', `exception_handler'
18420 or `nmi_handler', code will be generated to load the stack pointer
18421 from the USP register in the function prologue.
18424 This attribute specifies a function to be placed into L1
18425 Instruction SRAM. The function will be put into a specific section
18426 named `.l1.text'. With `-mfdpic', function calls with a such
18427 function as the callee or caller will use inlined PLT.
18429 `long_call/short_call'
18430 This attribute specifies how a particular function is called on
18431 ARM. Both attributes override the `-mlong-calls' (*note ARM
18432 Options::) command line switch and `#pragma long_calls' settings.
18433 The `long_call' attribute indicates that the function might be far
18434 away from the call site and require a different (more expensive)
18435 calling sequence. The `short_call' attribute always places the
18436 offset to the function from the call site into the `BL'
18437 instruction directly.
18439 `longcall/shortcall'
18440 On the Blackfin, RS/6000 and PowerPC, the `longcall' attribute
18441 indicates that the function might be far away from the call site
18442 and require a different (more expensive) calling sequence. The
18443 `shortcall' attribute indicates that the function is always close
18444 enough for the shorter calling sequence to be used. These
18445 attributes override both the `-mlongcall' switch and, on the
18446 RS/6000 and PowerPC, the `#pragma longcall' setting.
18448 *Note RS/6000 and PowerPC Options::, for more information on
18449 whether long calls are necessary.
18451 `long_call/near/far'
18452 These attributes specify how a particular function is called on
18453 MIPS. The attributes override the `-mlong-calls' (*note MIPS
18454 Options::) command-line switch. The `long_call' and `far'
18455 attributes are synonyms, and cause the compiler to always call the
18456 function by first loading its address into a register, and then
18457 using the contents of that register. The `near' attribute has the
18458 opposite effect; it specifies that non-PIC calls should be made
18459 using the more efficient `jal' instruction.
18462 The `malloc' attribute is used to tell the compiler that a function
18463 may be treated as if any non-`NULL' pointer it returns cannot
18464 alias any other pointer valid when the function returns. This
18465 will often improve optimization. Standard functions with this
18466 property include `malloc' and `calloc'. `realloc'-like functions
18467 have this property as long as the old pointer is never referred to
18468 (including comparing it to the new pointer) after the function
18469 returns a non-`NULL' value.
18472 On MIPS targets, you can use the `mips16' and `nomips16' function
18473 attributes to locally select or turn off MIPS16 code generation.
18474 A function with the `mips16' attribute is emitted as MIPS16 code,
18475 while MIPS16 code generation is disabled for functions with the
18476 `nomips16' attribute. These attributes override the `-mips16' and
18477 `-mno-mips16' options on the command line (*note MIPS Options::).
18479 When compiling files containing mixed MIPS16 and non-MIPS16 code,
18480 the preprocessor symbol `__mips16' reflects the setting on the
18481 command line, not that within individual functions. Mixed MIPS16
18482 and non-MIPS16 code may interact badly with some GCC extensions
18483 such as `__builtin_apply' (*note Constructing Calls::).
18485 `model (MODEL-NAME)'
18486 On the M32R/D, use this attribute to set the addressability of an
18487 object, and of the code generated for a function. The identifier
18488 MODEL-NAME is one of `small', `medium', or `large', representing
18489 each of the code models.
18491 Small model objects live in the lower 16MB of memory (so that their
18492 addresses can be loaded with the `ld24' instruction), and are
18493 callable with the `bl' instruction.
18495 Medium model objects may live anywhere in the 32-bit address space
18496 (the compiler will generate `seth/add3' instructions to load their
18497 addresses), and are callable with the `bl' instruction.
18499 Large model objects may live anywhere in the 32-bit address space
18500 (the compiler will generate `seth/add3' instructions to load their
18501 addresses), and may not be reachable with the `bl' instruction
18502 (the compiler will generate the much slower `seth/add3/jl'
18503 instruction sequence).
18505 On IA-64, use this attribute to set the addressability of an
18506 object. At present, the only supported identifier for MODEL-NAME
18507 is `small', indicating addressability via "small" (22-bit)
18508 addresses (so that their addresses can be loaded with the `addl'
18509 instruction). Caveat: such addressing is by definition not
18510 position independent and hence this attribute must not be used for
18511 objects defined by shared libraries.
18514 On 64-bit x86_64-*-* targets, you can use an ABI attribute to
18515 indicate which calling convention should be used for a function.
18516 The `ms_abi' attribute tells the compiler to use the Microsoft
18517 ABI, while the `sysv_abi' attribute tells the compiler to use the
18518 ABI used on GNU/Linux and other systems. The default is to use
18519 the Microsoft ABI when targeting Windows. On all other systems,
18520 the default is the AMD ABI.
18522 Note, This feature is currently sorried out for Windows targets
18526 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate
18527 that the specified function does not need prologue/epilogue
18528 sequences generated by the compiler. It is up to the programmer
18529 to provide these sequences. The only statements that can be safely
18530 included in naked functions are `asm' statements that do not have
18531 operands. All other statements, including declarations of local
18532 variables, `if' statements, and so forth, should be avoided.
18533 Naked functions should be used to implement the body of an
18534 assembly function, while allowing the compiler to construct the
18535 requisite function declaration for the assembler.
18538 On 68HC11 and 68HC12 the `near' attribute causes the compiler to
18539 use the normal calling convention based on `jsr' and `rts'. This
18540 attribute can be used to cancel the effect of the `-mlong-calls'
18544 Use this attribute together with `interrupt_handler',
18545 `exception_handler' or `nmi_handler' to indicate that the function
18546 entry code should enable nested interrupts or exceptions.
18549 Use this attribute on the Blackfin to indicate that the specified
18550 function is an NMI handler. The compiler will generate function
18551 entry and exit sequences suitable for use in an NMI handler when
18552 this attribute is present.
18554 `no_instrument_function'
18555 If `-finstrument-functions' is given, profiling function calls will
18556 be generated at entry and exit of most user-compiled functions.
18557 Functions with this attribute will not be so instrumented.
18560 This function attribute prevents a function from being considered
18561 for inlining. If the function does not have side-effects, there
18562 are optimizations other than inlining that causes function calls
18563 to be optimized away, although the function call is live. To keep
18564 such calls from being optimized away, put
18566 (*note Extended Asm::) in the called function, to serve as a
18567 special side-effect.
18569 `nonnull (ARG-INDEX, ...)'
18570 The `nonnull' attribute specifies that some function parameters
18571 should be non-null pointers. For instance, the declaration:
18574 my_memcpy (void *dest, const void *src, size_t len)
18575 __attribute__((nonnull (1, 2)));
18577 causes the compiler to check that, in calls to `my_memcpy',
18578 arguments DEST and SRC are non-null. If the compiler determines
18579 that a null pointer is passed in an argument slot marked as
18580 non-null, and the `-Wnonnull' option is enabled, a warning is
18581 issued. The compiler may also choose to make optimizations based
18582 on the knowledge that certain function arguments will not be null.
18584 If no argument index list is given to the `nonnull' attribute, all
18585 pointer arguments are marked as non-null. To illustrate, the
18586 following declaration is equivalent to the previous example:
18589 my_memcpy (void *dest, const void *src, size_t len)
18590 __attribute__((nonnull));
18593 A few standard library functions, such as `abort' and `exit',
18594 cannot return. GCC knows this automatically. Some programs define
18595 their own functions that never return. You can declare them
18596 `noreturn' to tell the compiler this fact. For example,
18598 void fatal () __attribute__ ((noreturn));
18603 /* ... */ /* Print error message. */ /* ... */
18607 The `noreturn' keyword tells the compiler to assume that `fatal'
18608 cannot return. It can then optimize without regard to what would
18609 happen if `fatal' ever did return. This makes slightly better
18610 code. More importantly, it helps avoid spurious warnings of
18611 uninitialized variables.
18613 The `noreturn' keyword does not affect the exceptional path when
18614 that applies: a `noreturn'-marked function may still return to the
18615 caller by throwing an exception or calling `longjmp'.
18617 Do not assume that registers saved by the calling function are
18618 restored before calling the `noreturn' function.
18620 It does not make sense for a `noreturn' function to have a return
18621 type other than `void'.
18623 The attribute `noreturn' is not implemented in GCC versions
18624 earlier than 2.5. An alternative way to declare that a function
18625 does not return, which works in the current version and in some
18626 older versions, is as follows:
18628 typedef void voidfn ();
18630 volatile voidfn fatal;
18632 This approach does not work in GNU C++.
18635 The `nothrow' attribute is used to inform the compiler that a
18636 function cannot throw an exception. For example, most functions in
18637 the standard C library can be guaranteed not to throw an exception
18638 with the notable exceptions of `qsort' and `bsearch' that take
18639 function pointer arguments. The `nothrow' attribute is not
18640 implemented in GCC versions earlier than 3.3.
18643 The `optimize' attribute is used to specify that a function is to
18644 be compiled with different optimization options than specified on
18645 the command line. Arguments can either be numbers or strings.
18646 Numbers are assumed to be an optimization level. Strings that
18647 begin with `O' are assumed to be an optimization option, while
18648 other options are assumed to be used with a `-f' prefix. You can
18649 also use the `#pragma GCC optimize' pragma to set the optimization
18650 options that affect more than one function. *Note Function
18651 Specific Option Pragmas::, for details about the `#pragma GCC
18654 This can be used for instance to have frequently executed functions
18655 compiled with more aggressive optimization options that produce
18656 faster and larger code, while other functions can be called with
18657 less aggressive options.
18660 Many functions have no effects except the return value and their
18661 return value depends only on the parameters and/or global
18662 variables. Such a function can be subject to common subexpression
18663 elimination and loop optimization just as an arithmetic operator
18664 would be. These functions should be declared with the attribute
18665 `pure'. For example,
18667 int square (int) __attribute__ ((pure));
18669 says that the hypothetical function `square' is safe to call fewer
18670 times than the program says.
18672 Some of common examples of pure functions are `strlen' or `memcmp'.
18673 Interesting non-pure functions are functions with infinite loops
18674 or those depending on volatile memory or other system resource,
18675 that may change between two consecutive calls (such as `feof' in a
18676 multithreading environment).
18678 The attribute `pure' is not implemented in GCC versions earlier
18682 The `hot' attribute is used to inform the compiler that a function
18683 is a hot spot of the compiled program. The function is optimized
18684 more aggressively and on many target it is placed into special
18685 subsection of the text section so all hot functions appears close
18686 together improving locality.
18688 When profile feedback is available, via `-fprofile-use', hot
18689 functions are automatically detected and this attribute is ignored.
18691 The `hot' attribute is not implemented in GCC versions earlier
18695 The `cold' attribute is used to inform the compiler that a
18696 function is unlikely executed. The function is optimized for size
18697 rather than speed and on many targets it is placed into special
18698 subsection of the text section so all cold functions appears close
18699 together improving code locality of non-cold parts of program.
18700 The paths leading to call of cold functions within code are marked
18701 as unlikely by the branch prediction mechanism. It is thus useful
18702 to mark functions used to handle unlikely conditions, such as
18703 `perror', as cold to improve optimization of hot functions that do
18704 call marked functions in rare occasions.
18706 When profile feedback is available, via `-fprofile-use', hot
18707 functions are automatically detected and this attribute is ignored.
18709 The `cold' attribute is not implemented in GCC versions earlier
18713 On the Intel 386, the `regparm' attribute causes the compiler to
18714 pass arguments number one to NUMBER if they are of integral type
18715 in registers EAX, EDX, and ECX instead of on the stack. Functions
18716 that take a variable number of arguments will continue to be
18717 passed all of their arguments on the stack.
18719 Beware that on some ELF systems this attribute is unsuitable for
18720 global functions in shared libraries with lazy binding (which is
18721 the default). Lazy binding will send the first call via resolving
18722 code in the loader, which might assume EAX, EDX and ECX can be
18723 clobbered, as per the standard calling conventions. Solaris 8 is
18724 affected by this. GNU systems with GLIBC 2.1 or higher, and
18725 FreeBSD, are believed to be safe since the loaders there save EAX,
18726 EDX and ECX. (Lazy binding can be disabled with the linker or the
18727 loader if desired, to avoid the problem.)
18730 On the Intel 386 with SSE support, the `sseregparm' attribute
18731 causes the compiler to pass up to 3 floating point arguments in
18732 SSE registers instead of on the stack. Functions that take a
18733 variable number of arguments will continue to pass all of their
18734 floating point arguments on the stack.
18736 `force_align_arg_pointer'
18737 On the Intel x86, the `force_align_arg_pointer' attribute may be
18738 applied to individual function definitions, generating an alternate
18739 prologue and epilogue that realigns the runtime stack if necessary.
18740 This supports mixing legacy codes that run with a 4-byte aligned
18741 stack with modern codes that keep a 16-byte stack for SSE
18745 On the SH2A target, this attribute enables the high-speed register
18746 saving and restoration using a register bank for
18747 `interrupt_handler' routines. Saving to the bank is performed
18748 automatically after the CPU accepts an interrupt that uses a
18751 The nineteen 32-bit registers comprising general register R0 to
18752 R14, control register GBR, and system registers MACH, MACL, and PR
18753 and the vector table address offset are saved into a register
18754 bank. Register banks are stacked in first-in last-out (FILO)
18755 sequence. Restoration from the bank is executed by issuing a
18756 RESBANK instruction.
18759 The `returns_twice' attribute tells the compiler that a function
18760 may return more than one time. The compiler will ensure that all
18761 registers are dead before calling such a function and will emit a
18762 warning about the variables that may be clobbered after the second
18763 return from the function. Examples of such functions are `setjmp'
18764 and `vfork'. The `longjmp'-like counterpart of such function, if
18765 any, might need to be marked with the `noreturn' attribute.
18768 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to
18769 indicate that all registers except the stack pointer should be
18770 saved in the prologue regardless of whether they are used or not.
18772 `section ("SECTION-NAME")'
18773 Normally, the compiler places the code it generates in the `text'
18774 section. Sometimes, however, you need additional sections, or you
18775 need certain particular functions to appear in special sections.
18776 The `section' attribute specifies that a function lives in a
18777 particular section. For example, the declaration:
18779 extern void foobar (void) __attribute__ ((section ("bar")));
18781 puts the function `foobar' in the `bar' section.
18783 Some file formats do not support arbitrary sections so the
18784 `section' attribute is not available on all platforms. If you
18785 need to map the entire contents of a module to a particular
18786 section, consider using the facilities of the linker instead.
18789 This function attribute ensures that a parameter in a function
18790 call is an explicit `NULL'. The attribute is only valid on
18791 variadic functions. By default, the sentinel is located at
18792 position zero, the last parameter of the function call. If an
18793 optional integer position argument P is supplied to the attribute,
18794 the sentinel must be located at position P counting backwards from
18795 the end of the argument list.
18797 __attribute__ ((sentinel))
18799 __attribute__ ((sentinel(0)))
18801 The attribute is automatically set with a position of 0 for the
18802 built-in functions `execl' and `execlp'. The built-in function
18803 `execle' has the attribute set with a position of 1.
18805 A valid `NULL' in this context is defined as zero with any pointer
18806 type. If your system defines the `NULL' macro with an integer type
18807 then you need to add an explicit cast. GCC replaces `stddef.h'
18808 with a copy that redefines NULL appropriately.
18810 The warnings for missing or incorrect sentinels are enabled with
18814 See long_call/short_call.
18817 See longcall/shortcall.
18820 Use this attribute on the AVR to indicate that the specified
18821 function is a signal handler. The compiler will generate function
18822 entry and exit sequences suitable for use in a signal handler when
18823 this attribute is present. Interrupts will be disabled inside the
18827 Use this attribute on the SH to indicate an `interrupt_handler'
18828 function should switch to an alternate stack. It expects a string
18829 argument that names a global variable holding the address of the
18833 void f () __attribute__ ((interrupt_handler,
18834 sp_switch ("alt_stack")));
18837 On the Intel 386, the `stdcall' attribute causes the compiler to
18838 assume that the called function will pop off the stack space used
18839 to pass arguments, unless it takes a variable number of arguments.
18842 This attribute is used to modify the IA64 calling convention by
18843 marking all input registers as live at all function exits. This
18844 makes it possible to restart a system call after an interrupt
18845 without having to save/restore the input registers. This also
18846 prevents kernel data from leaking into application code.
18849 The `target' attribute is used to specify that a function is to be
18850 compiled with different target options than specified on the
18851 command line. This can be used for instance to have functions
18852 compiled with a different ISA (instruction set architecture) than
18853 the default. You can also use the `#pragma GCC target' pragma to
18854 set more than one function to be compiled with specific target
18855 options. *Note Function Specific Option Pragmas::, for details
18856 about the `#pragma GCC target' pragma.
18858 For instance on a 386, you could compile one function with
18859 `target("sse4.1,arch=core2")' and another with
18860 `target("sse4a,arch=amdfam10")' that would be equivalent to
18861 compiling the first function with `-msse4.1' and `-march=core2'
18862 options, and the second function with `-msse4a' and
18863 `-march=amdfam10' options. It is up to the user to make sure that
18864 a function is only invoked on a machine that supports the
18865 particular ISA it was compiled for (for example by using `cpuid'
18866 on 386 to determine what feature bits and architecture family are
18869 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
18870 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
18872 On the 386, the following options are allowed:
18876 Enable/disable the generation of the advanced bit
18881 Enable/disable the generation of the AES instructions.
18885 Enable/disable the generation of the MMX instructions.
18889 Enable/disable the generation of the PCLMUL instructions.
18893 Enable/disable the generation of the POPCNT instruction.
18897 Enable/disable the generation of the SSE instructions.
18901 Enable/disable the generation of the SSE2 instructions.
18905 Enable/disable the generation of the SSE3 instructions.
18909 Enable/disable the generation of the SSE4 instructions (both
18910 SSE4.1 and SSE4.2).
18914 Enable/disable the generation of the sse4.1 instructions.
18918 Enable/disable the generation of the sse4.2 instructions.
18922 Enable/disable the generation of the SSE4A instructions.
18926 Enable/disable the generation of the SSE5 instructions.
18930 Enable/disable the generation of the SSSE3 instructions.
18934 Enable/disable the generation of the CLD before string moves.
18937 `no-fancy-math-387'
18938 Enable/disable the generation of the `sin', `cos', and `sqrt'
18939 instructions on the 387 floating point unit.
18943 Enable/disable the generation of the fused multiply/add
18948 Enable/disable the generation of floating point that depends
18949 on IEEE arithmetic.
18951 `inline-all-stringops'
18952 `no-inline-all-stringops'
18953 Enable/disable inlining of string operations.
18955 `inline-stringops-dynamically'
18956 `no-inline-stringops-dynamically'
18957 Enable/disable the generation of the inline code to do small
18958 string operations and calling the library routines for large
18962 `no-align-stringops'
18963 Do/do not align destination of inlined string operations.
18967 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and
18968 RSQRTPS instructions followed an additional Newton-Raphson
18969 step instead of doing a floating point division.
18972 Specify the architecture to generate code for in compiling
18976 Specify the architecture to tune for in compiling the
18980 Specify which floating point unit to use. The
18981 `target("fpmath=sse,387")' option must be specified as
18982 `target("fpmath=sse+387")' because the comma would separate
18985 On the 386, you can use either multiple strings to specify multiple
18986 options, or you can separate the option with a comma (`,').
18988 On the 386, the inliner will not inline a function that has
18989 different target options than the caller, unless the callee has a
18990 subset of the target options of the caller. For example a
18991 function declared with `target("sse5")' can inline a function with
18992 `target("sse2")', since `-msse5' implies `-msse2'.
18994 The `target' attribute is not implemented in GCC versions earlier
18995 than 4.4, and at present only the 386 uses it.
18998 Use this attribute on the H8/300H and H8S to indicate that the
18999 specified variable should be placed into the tiny data section.
19000 The compiler will generate more efficient code for loads and stores
19001 on data in the tiny data section. Note the tiny data area is
19002 limited to slightly under 32kbytes of data.
19005 Use this attribute on the SH for an `interrupt_handler' to return
19006 using `trapa' instead of `rte'. This attribute expects an integer
19007 argument specifying the trap number to be used.
19010 This attribute, attached to a function, means that the function is
19011 meant to be possibly unused. GCC will not produce a warning for
19015 This attribute, attached to a function, means that code must be
19016 emitted for the function even if it appears that the function is
19017 not referenced. This is useful, for example, when the function is
19018 referenced only in inline assembly.
19021 This IA64 HP-UX attribute, attached to a global variable or
19022 function, renames a symbol to contain a version string, thus
19023 allowing for function level versioning. HP-UX system header files
19024 may use version level functioning for some system calls.
19026 extern int foo () __attribute__((version_id ("20040821")));
19028 Calls to FOO will be mapped to calls to FOO{20040821}.
19030 `visibility ("VISIBILITY_TYPE")'
19031 This attribute affects the linkage of the declaration to which it
19032 is attached. There are four supported VISIBILITY_TYPE values:
19033 default, hidden, protected or internal visibility.
19035 void __attribute__ ((visibility ("protected")))
19036 f () { /* Do something. */; }
19037 int i __attribute__ ((visibility ("hidden")));
19039 The possible values of VISIBILITY_TYPE correspond to the
19040 visibility settings in the ELF gABI.
19043 Default visibility is the normal case for the object file
19044 format. This value is available for the visibility attribute
19045 to override other options that may change the assumed
19046 visibility of entities.
19048 On ELF, default visibility means that the declaration is
19049 visible to other modules and, in shared libraries, means that
19050 the declared entity may be overridden.
19052 On Darwin, default visibility means that the declaration is
19053 visible to other modules.
19055 Default visibility corresponds to "external linkage" in the
19059 Hidden visibility indicates that the entity declared will
19060 have a new form of linkage, which we'll call "hidden
19061 linkage". Two declarations of an object with hidden linkage
19062 refer to the same object if they are in the same shared
19066 Internal visibility is like hidden visibility, but with
19067 additional processor specific semantics. Unless otherwise
19068 specified by the psABI, GCC defines internal visibility to
19069 mean that a function is _never_ called from another module.
19070 Compare this with hidden functions which, while they cannot
19071 be referenced directly by other modules, can be referenced
19072 indirectly via function pointers. By indicating that a
19073 function cannot be called from outside the module, GCC may
19074 for instance omit the load of a PIC register since it is known
19075 that the calling function loaded the correct value.
19078 Protected visibility is like default visibility except that it
19079 indicates that references within the defining module will
19080 bind to the definition in that module. That is, the declared
19081 entity cannot be overridden by another module.
19084 All visibilities are supported on many, but not all, ELF targets
19085 (supported when the assembler supports the `.visibility'
19086 pseudo-op). Default visibility is supported everywhere. Hidden
19087 visibility is supported on Darwin targets.
19089 The visibility attribute should be applied only to declarations
19090 which would otherwise have external linkage. The attribute should
19091 be applied consistently, so that the same entity should not be
19092 declared with different settings of the attribute.
19094 In C++, the visibility attribute applies to types as well as
19095 functions and objects, because in C++ types have linkage. A class
19096 must not have greater visibility than its non-static data member
19097 types and bases, and class members default to the visibility of
19098 their class. Also, a declaration without explicit visibility is
19099 limited to the visibility of its type.
19101 In C++, you can mark member functions and static member variables
19102 of a class with the visibility attribute. This is useful if you
19103 know a particular method or static member variable should only be
19104 used from one shared object; then you can mark it hidden while the
19105 rest of the class has default visibility. Care must be taken to
19106 avoid breaking the One Definition Rule; for example, it is usually
19107 not useful to mark an inline method as hidden without marking the
19108 whole class as hidden.
19110 A C++ namespace declaration can also have the visibility attribute.
19111 This attribute applies only to the particular namespace body, not
19112 to other definitions of the same namespace; it is equivalent to
19113 using `#pragma GCC visibility' before and after the namespace
19114 definition (*note Visibility Pragmas::).
19116 In C++, if a template argument has limited visibility, this
19117 restriction is implicitly propagated to the template instantiation.
19118 Otherwise, template instantiations and specializations default to
19119 the visibility of their template.
19121 If both the template and enclosing class have explicit visibility,
19122 the visibility from the template is used.
19124 `warn_unused_result'
19125 The `warn_unused_result' attribute causes a warning to be emitted
19126 if a caller of the function with this attribute does not use its
19127 return value. This is useful for functions where not checking the
19128 result is either a security problem or always a bug, such as
19131 int fn () __attribute__ ((warn_unused_result));
19134 if (fn () < 0) return -1;
19139 results in warning on line 5.
19142 The `weak' attribute causes the declaration to be emitted as a weak
19143 symbol rather than a global. This is primarily useful in defining
19144 library functions which can be overridden in user code, though it
19145 can also be used with non-function declarations. Weak symbols are
19146 supported for ELF targets, and also for a.out targets when using
19147 the GNU assembler and linker.
19150 `weakref ("TARGET")'
19151 The `weakref' attribute marks a declaration as a weak reference.
19152 Without arguments, it should be accompanied by an `alias' attribute
19153 naming the target symbol. Optionally, the TARGET may be given as
19154 an argument to `weakref' itself. In either case, `weakref'
19155 implicitly marks the declaration as `weak'. Without a TARGET,
19156 given as an argument to `weakref' or to `alias', `weakref' is
19157 equivalent to `weak'.
19159 static int x() __attribute__ ((weakref ("y")));
19160 /* is equivalent to... */
19161 static int x() __attribute__ ((weak, weakref, alias ("y")));
19163 static int x() __attribute__ ((weakref));
19164 static int x() __attribute__ ((alias ("y")));
19166 A weak reference is an alias that does not by itself require a
19167 definition to be given for the target symbol. If the target
19168 symbol is only referenced through weak references, then the
19169 becomes a `weak' undefined symbol. If it is directly referenced,
19170 however, then such strong references prevail, and a definition
19171 will be required for the symbol, not necessarily in the same
19174 The effect is equivalent to moving all references to the alias to a
19175 separate translation unit, renaming the alias to the aliased
19176 symbol, declaring it as weak, compiling the two separate
19177 translation units and performing a reloadable link on them.
19179 At present, a declaration to which `weakref' is attached can only
19183 You can specify multiple attributes in a declaration by separating them
19184 by commas within the double parentheses or by immediately following an
19185 attribute declaration with another attribute declaration.
19187 Some people object to the `__attribute__' feature, suggesting that ISO
19188 C's `#pragma' should be used instead. At the time `__attribute__' was
19189 designed, there were two reasons for not doing this.
19191 1. It is impossible to generate `#pragma' commands from a macro.
19193 2. There is no telling what the same `#pragma' might mean in another
19196 These two reasons applied to almost any application that might have
19197 been proposed for `#pragma'. It was basically a mistake to use
19198 `#pragma' for _anything_.
19200 The ISO C99 standard includes `_Pragma', which now allows pragmas to
19201 be generated from macros. In addition, a `#pragma GCC' namespace is
19202 now in use for GCC-specific pragmas. However, it has been found
19203 convenient to use `__attribute__' to achieve a natural attachment of
19204 attributes to their corresponding declarations, whereas `#pragma GCC'
19205 is of use for constructs that do not naturally form part of the
19206 grammar. *Note Miscellaneous Preprocessing Directives: (cpp)Other
19210 File: gcc.info, Node: Attribute Syntax, Next: Function Prototypes, Prev: Function Attributes, Up: C Extensions
19212 5.28 Attribute Syntax
19213 =====================
19215 This section describes the syntax with which `__attribute__' may be
19216 used, and the constructs to which attribute specifiers bind, for the C
19217 language. Some details may vary for C++ and Objective-C. Because of
19218 infelicities in the grammar for attributes, some forms described here
19219 may not be successfully parsed in all cases.
19221 There are some problems with the semantics of attributes in C++. For
19222 example, there are no manglings for attributes, although they may affect
19223 code generation, so problems may arise when attributed types are used in
19224 conjunction with templates or overloading. Similarly, `typeid' does
19225 not distinguish between types with different attributes. Support for
19226 attributes in C++ may be restricted in future to attributes on
19227 declarations only, but not on nested declarators.
19229 *Note Function Attributes::, for details of the semantics of attributes
19230 applying to functions. *Note Variable Attributes::, for details of the
19231 semantics of attributes applying to variables. *Note Type Attributes::,
19232 for details of the semantics of attributes applying to structure, union
19233 and enumerated types.
19235 An "attribute specifier" is of the form `__attribute__
19236 ((ATTRIBUTE-LIST))'. An "attribute list" is a possibly empty
19237 comma-separated sequence of "attributes", where each attribute is one
19240 * Empty. Empty attributes are ignored.
19242 * A word (which may be an identifier such as `unused', or a reserved
19243 word such as `const').
19245 * A word, followed by, in parentheses, parameters for the attribute.
19246 These parameters take one of the following forms:
19248 * An identifier. For example, `mode' attributes use this form.
19250 * An identifier followed by a comma and a non-empty
19251 comma-separated list of expressions. For example, `format'
19252 attributes use this form.
19254 * A possibly empty comma-separated list of expressions. For
19255 example, `format_arg' attributes use this form with the list
19256 being a single integer constant expression, and `alias'
19257 attributes use this form with the list being a single string
19260 An "attribute specifier list" is a sequence of one or more attribute
19261 specifiers, not separated by any other tokens.
19263 In GNU C, an attribute specifier list may appear after the colon
19264 following a label, other than a `case' or `default' label. The only
19265 attribute it makes sense to use after a label is `unused'. This
19266 feature is intended for code generated by programs which contains labels
19267 that may be unused but which is compiled with `-Wall'. It would not
19268 normally be appropriate to use in it human-written code, though it
19269 could be useful in cases where the code that jumps to the label is
19270 contained within an `#ifdef' conditional. GNU C++ does not permit such
19271 placement of attribute lists, as it is permissible for a declaration,
19272 which could begin with an attribute list, to be labelled in C++.
19273 Declarations cannot be labelled in C90 or C99, so the ambiguity does
19276 An attribute specifier list may appear as part of a `struct', `union'
19277 or `enum' specifier. It may go either immediately after the `struct',
19278 `union' or `enum' keyword, or after the closing brace. The former
19279 syntax is preferred. Where attribute specifiers follow the closing
19280 brace, they are considered to relate to the structure, union or
19281 enumerated type defined, not to any enclosing declaration the type
19282 specifier appears in, and the type defined is not complete until after
19283 the attribute specifiers.
19285 Otherwise, an attribute specifier appears as part of a declaration,
19286 counting declarations of unnamed parameters and type names, and relates
19287 to that declaration (which may be nested in another declaration, for
19288 example in the case of a parameter declaration), or to a particular
19289 declarator within a declaration. Where an attribute specifier is
19290 applied to a parameter declared as a function or an array, it should
19291 apply to the function or array rather than the pointer to which the
19292 parameter is implicitly converted, but this is not yet correctly
19295 Any list of specifiers and qualifiers at the start of a declaration may
19296 contain attribute specifiers, whether or not such a list may in that
19297 context contain storage class specifiers. (Some attributes, however,
19298 are essentially in the nature of storage class specifiers, and only make
19299 sense where storage class specifiers may be used; for example,
19300 `section'.) There is one necessary limitation to this syntax: the
19301 first old-style parameter declaration in a function definition cannot
19302 begin with an attribute specifier, because such an attribute applies to
19303 the function instead by syntax described below (which, however, is not
19304 yet implemented in this case). In some other cases, attribute
19305 specifiers are permitted by this grammar but not yet supported by the
19306 compiler. All attribute specifiers in this place relate to the
19307 declaration as a whole. In the obsolescent usage where a type of `int'
19308 is implied by the absence of type specifiers, such a list of specifiers
19309 and qualifiers may be an attribute specifier list with no other
19310 specifiers or qualifiers.
19312 At present, the first parameter in a function prototype must have some
19313 type specifier which is not an attribute specifier; this resolves an
19314 ambiguity in the interpretation of `void f(int (__attribute__((foo))
19315 x))', but is subject to change. At present, if the parentheses of a
19316 function declarator contain only attributes then those attributes are
19317 ignored, rather than yielding an error or warning or implying a single
19318 parameter of type int, but this is subject to change.
19320 An attribute specifier list may appear immediately before a declarator
19321 (other than the first) in a comma-separated list of declarators in a
19322 declaration of more than one identifier using a single list of
19323 specifiers and qualifiers. Such attribute specifiers apply only to the
19324 identifier before whose declarator they appear. For example, in
19326 __attribute__((noreturn)) void d0 (void),
19327 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
19330 the `noreturn' attribute applies to all the functions declared; the
19331 `format' attribute only applies to `d1'.
19333 An attribute specifier list may appear immediately before the comma,
19334 `=' or semicolon terminating the declaration of an identifier other
19335 than a function definition. Such attribute specifiers apply to the
19336 declared object or function. Where an assembler name for an object or
19337 function is specified (*note Asm Labels::), the attribute must follow
19338 the `asm' specification.
19340 An attribute specifier list may, in future, be permitted to appear
19341 after the declarator in a function definition (before any old-style
19342 parameter declarations or the function body).
19344 Attribute specifiers may be mixed with type qualifiers appearing inside
19345 the `[]' of a parameter array declarator, in the C99 construct by which
19346 such qualifiers are applied to the pointer to which the array is
19347 implicitly converted. Such attribute specifiers apply to the pointer,
19348 not to the array, but at present this is not implemented and they are
19351 An attribute specifier list may appear at the start of a nested
19352 declarator. At present, there are some limitations in this usage: the
19353 attributes correctly apply to the declarator, but for most individual
19354 attributes the semantics this implies are not implemented. When
19355 attribute specifiers follow the `*' of a pointer declarator, they may
19356 be mixed with any type qualifiers present. The following describes the
19357 formal semantics of this syntax. It will make the most sense if you
19358 are familiar with the formal specification of declarators in the ISO C
19361 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration `T D1',
19362 where `T' contains declaration specifiers that specify a type TYPE
19363 (such as `int') and `D1' is a declarator that contains an identifier
19364 IDENT. The type specified for IDENT for derived declarators whose type
19365 does not include an attribute specifier is as in the ISO C standard.
19367 If `D1' has the form `( ATTRIBUTE-SPECIFIER-LIST D )', and the
19368 declaration `T D' specifies the type "DERIVED-DECLARATOR-TYPE-LIST
19369 TYPE" for IDENT, then `T D1' specifies the type
19370 "DERIVED-DECLARATOR-TYPE-LIST ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
19372 If `D1' has the form `* TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST
19373 D', and the declaration `T D' specifies the type
19374 "DERIVED-DECLARATOR-TYPE-LIST TYPE" for IDENT, then `T D1' specifies
19375 the type "DERIVED-DECLARATOR-TYPE-LIST
19376 TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
19380 void (__attribute__((noreturn)) ****f) (void);
19382 specifies the type "pointer to pointer to pointer to pointer to
19383 non-returning function returning `void'". As another example,
19385 char *__attribute__((aligned(8))) *f;
19387 specifies the type "pointer to 8-byte-aligned pointer to `char'". Note
19388 again that this does not work with most attributes; for example, the
19389 usage of `aligned' and `noreturn' attributes given above is not yet
19392 For compatibility with existing code written for compiler versions that
19393 did not implement attributes on nested declarators, some laxity is
19394 allowed in the placing of attributes. If an attribute that only applies
19395 to types is applied to a declaration, it will be treated as applying to
19396 the type of that declaration. If an attribute that only applies to
19397 declarations is applied to the type of a declaration, it will be treated
19398 as applying to that declaration; and, for compatibility with code
19399 placing the attributes immediately before the identifier declared, such
19400 an attribute applied to a function return type will be treated as
19401 applying to the function type, and such an attribute applied to an array
19402 element type will be treated as applying to the array type. If an
19403 attribute that only applies to function types is applied to a
19404 pointer-to-function type, it will be treated as applying to the pointer
19405 target type; if such an attribute is applied to a function return type
19406 that is not a pointer-to-function type, it will be treated as applying
19407 to the function type.
19410 File: gcc.info, Node: Function Prototypes, Next: C++ Comments, Prev: Attribute Syntax, Up: C Extensions
19412 5.29 Prototypes and Old-Style Function Definitions
19413 ==================================================
19415 GNU C extends ISO C to allow a function prototype to override a later
19416 old-style non-prototype definition. Consider the following example:
19418 /* Use prototypes unless the compiler is old-fashioned. */
19425 /* Prototype function declaration. */
19426 int isroot P((uid_t));
19428 /* Old-style function definition. */
19430 isroot (x) /* ??? lossage here ??? */
19436 Suppose the type `uid_t' happens to be `short'. ISO C does not allow
19437 this example, because subword arguments in old-style non-prototype
19438 definitions are promoted. Therefore in this example the function
19439 definition's argument is really an `int', which does not match the
19440 prototype argument type of `short'.
19442 This restriction of ISO C makes it hard to write code that is portable
19443 to traditional C compilers, because the programmer does not know
19444 whether the `uid_t' type is `short', `int', or `long'. Therefore, in
19445 cases like these GNU C allows a prototype to override a later old-style
19446 definition. More precisely, in GNU C, a function prototype argument
19447 type overrides the argument type specified by a later old-style
19448 definition if the former type is the same as the latter type before
19449 promotion. Thus in GNU C the above example is equivalent to the
19452 int isroot (uid_t);
19460 GNU C++ does not support old-style function definitions, so this
19461 extension is irrelevant.
19464 File: gcc.info, Node: C++ Comments, Next: Dollar Signs, Prev: Function Prototypes, Up: C Extensions
19466 5.30 C++ Style Comments
19467 =======================
19469 In GNU C, you may use C++ style comments, which start with `//' and
19470 continue until the end of the line. Many other C implementations allow
19471 such comments, and they are included in the 1999 C standard. However,
19472 C++ style comments are not recognized if you specify an `-std' option
19473 specifying a version of ISO C before C99, or `-ansi' (equivalent to
19477 File: gcc.info, Node: Dollar Signs, Next: Character Escapes, Prev: C++ Comments, Up: C Extensions
19479 5.31 Dollar Signs in Identifier Names
19480 =====================================
19482 In GNU C, you may normally use dollar signs in identifier names. This
19483 is because many traditional C implementations allow such identifiers.
19484 However, dollar signs in identifiers are not supported on a few target
19485 machines, typically because the target assembler does not allow them.
19488 File: gcc.info, Node: Character Escapes, Next: Variable Attributes, Prev: Dollar Signs, Up: C Extensions
19490 5.32 The Character <ESC> in Constants
19491 =====================================
19493 You can use the sequence `\e' in a string or character constant to
19494 stand for the ASCII character <ESC>.
19497 File: gcc.info, Node: Alignment, Next: Inline, Prev: Type Attributes, Up: C Extensions
19499 5.33 Inquiring on Alignment of Types or Variables
19500 =================================================
19502 The keyword `__alignof__' allows you to inquire about how an object is
19503 aligned, or the minimum alignment usually required by a type. Its
19504 syntax is just like `sizeof'.
19506 For example, if the target machine requires a `double' value to be
19507 aligned on an 8-byte boundary, then `__alignof__ (double)' is 8. This
19508 is true on many RISC machines. On more traditional machine designs,
19509 `__alignof__ (double)' is 4 or even 2.
19511 Some machines never actually require alignment; they allow reference
19512 to any data type even at an odd address. For these machines,
19513 `__alignof__' reports the smallest alignment that GCC will give the
19514 data type, usually as mandated by the target ABI.
19516 If the operand of `__alignof__' is an lvalue rather than a type, its
19517 value is the required alignment for its type, taking into account any
19518 minimum alignment specified with GCC's `__attribute__' extension (*note
19519 Variable Attributes::). For example, after this declaration:
19521 struct foo { int x; char y; } foo1;
19523 the value of `__alignof__ (foo1.y)' is 1, even though its actual
19524 alignment is probably 2 or 4, the same as `__alignof__ (int)'.
19526 It is an error to ask for the alignment of an incomplete type.
19529 File: gcc.info, Node: Variable Attributes, Next: Type Attributes, Prev: Character Escapes, Up: C Extensions
19531 5.34 Specifying Attributes of Variables
19532 =======================================
19534 The keyword `__attribute__' allows you to specify special attributes of
19535 variables or structure fields. This keyword is followed by an
19536 attribute specification inside double parentheses. Some attributes are
19537 currently defined generically for variables. Other attributes are
19538 defined for variables on particular target systems. Other attributes
19539 are available for functions (*note Function Attributes::) and for types
19540 (*note Type Attributes::). Other front ends might define more
19541 attributes (*note Extensions to the C++ Language: C++ Extensions.).
19543 You may also specify attributes with `__' preceding and following each
19544 keyword. This allows you to use them in header files without being
19545 concerned about a possible macro of the same name. For example, you
19546 may use `__aligned__' instead of `aligned'.
19548 *Note Attribute Syntax::, for details of the exact syntax for using
19551 `aligned (ALIGNMENT)'
19552 This attribute specifies a minimum alignment for the variable or
19553 structure field, measured in bytes. For example, the declaration:
19555 int x __attribute__ ((aligned (16))) = 0;
19557 causes the compiler to allocate the global variable `x' on a
19558 16-byte boundary. On a 68040, this could be used in conjunction
19559 with an `asm' expression to access the `move16' instruction which
19560 requires 16-byte aligned operands.
19562 You can also specify the alignment of structure fields. For
19563 example, to create a double-word aligned `int' pair, you could
19566 struct foo { int x[2] __attribute__ ((aligned (8))); };
19568 This is an alternative to creating a union with a `double' member
19569 that forces the union to be double-word aligned.
19571 As in the preceding examples, you can explicitly specify the
19572 alignment (in bytes) that you wish the compiler to use for a given
19573 variable or structure field. Alternatively, you can leave out the
19574 alignment factor and just ask the compiler to align a variable or
19575 field to the default alignment for the target architecture you are
19576 compiling for. The default alignment is sufficient for all scalar
19577 types, but may not be enough for all vector types on a target
19578 which supports vector operations. The default alignment is fixed
19579 for a particular target ABI.
19581 Gcc also provides a target specific macro `__BIGGEST_ALIGNMENT__',
19582 which is the largest alignment ever used for any data type on the
19583 target machine you are compiling for. For example, you could
19586 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
19588 The compiler automatically sets the alignment for the declared
19589 variable or field to `__BIGGEST_ALIGNMENT__'. Doing this can
19590 often make copy operations more efficient, because the compiler can
19591 use whatever instructions copy the biggest chunks of memory when
19592 performing copies to or from the variables or fields that you have
19593 aligned this way. Note that the value of `__BIGGEST_ALIGNMENT__'
19594 may change depending on command line options.
19596 When used on a struct, or struct member, the `aligned' attribute
19597 can only increase the alignment; in order to decrease it, the
19598 `packed' attribute must be specified as well. When used as part
19599 of a typedef, the `aligned' attribute can both increase and
19600 decrease alignment, and specifying the `packed' attribute will
19601 generate a warning.
19603 Note that the effectiveness of `aligned' attributes may be limited
19604 by inherent limitations in your linker. On many systems, the
19605 linker is only able to arrange for variables to be aligned up to a
19606 certain maximum alignment. (For some linkers, the maximum
19607 supported alignment may be very very small.) If your linker is
19608 only able to align variables up to a maximum of 8 byte alignment,
19609 then specifying `aligned(16)' in an `__attribute__' will still
19610 only provide you with 8 byte alignment. See your linker
19611 documentation for further information.
19613 The `aligned' attribute can also be used for functions (*note
19614 Function Attributes::.)
19616 `cleanup (CLEANUP_FUNCTION)'
19617 The `cleanup' attribute runs a function when the variable goes out
19618 of scope. This attribute can only be applied to auto function
19619 scope variables; it may not be applied to parameters or variables
19620 with static storage duration. The function must take one
19621 parameter, a pointer to a type compatible with the variable. The
19622 return value of the function (if any) is ignored.
19624 If `-fexceptions' is enabled, then CLEANUP_FUNCTION will be run
19625 during the stack unwinding that happens during the processing of
19626 the exception. Note that the `cleanup' attribute does not allow
19627 the exception to be caught, only to perform an action. It is
19628 undefined what happens if CLEANUP_FUNCTION does not return
19633 The `common' attribute requests GCC to place a variable in
19634 "common" storage. The `nocommon' attribute requests the
19635 opposite--to allocate space for it directly.
19637 These attributes override the default chosen by the `-fno-common'
19638 and `-fcommon' flags respectively.
19641 The `deprecated' attribute results in a warning if the variable is
19642 used anywhere in the source file. This is useful when identifying
19643 variables that are expected to be removed in a future version of a
19644 program. The warning also includes the location of the declaration
19645 of the deprecated variable, to enable users to easily find further
19646 information about why the variable is deprecated, or what they
19647 should do instead. Note that the warning only occurs for uses:
19649 extern int old_var __attribute__ ((deprecated));
19650 extern int old_var;
19651 int new_fn () { return old_var; }
19653 results in a warning on line 3 but not line 2.
19655 The `deprecated' attribute can also be used for functions and
19656 types (*note Function Attributes::, *note Type Attributes::.)
19659 This attribute specifies the data type for the
19660 declaration--whichever type corresponds to the mode MODE. This in
19661 effect lets you request an integer or floating point type
19662 according to its width.
19664 You may also specify a mode of `byte' or `__byte__' to indicate
19665 the mode corresponding to a one-byte integer, `word' or `__word__'
19666 for the mode of a one-word integer, and `pointer' or `__pointer__'
19667 for the mode used to represent pointers.
19670 The `packed' attribute specifies that a variable or structure field
19671 should have the smallest possible alignment--one byte for a
19672 variable, and one bit for a field, unless you specify a larger
19673 value with the `aligned' attribute.
19675 Here is a structure in which the field `x' is packed, so that it
19676 immediately follows `a':
19681 int x[2] __attribute__ ((packed));
19684 _Note:_ The 4.1, 4.2 and 4.3 series of GCC ignore the `packed'
19685 attribute on bit-fields of type `char'. This has been fixed in
19686 GCC 4.4 but the change can lead to differences in the structure
19687 layout. See the documentation of `-Wpacked-bitfield-compat' for
19690 `section ("SECTION-NAME")'
19691 Normally, the compiler places the objects it generates in sections
19692 like `data' and `bss'. Sometimes, however, you need additional
19693 sections, or you need certain particular variables to appear in
19694 special sections, for example to map to special hardware. The
19695 `section' attribute specifies that a variable (or function) lives
19696 in a particular section. For example, this small program uses
19697 several specific section names:
19699 struct duart a __attribute__ ((section ("DUART_A"))) = { 0 };
19700 struct duart b __attribute__ ((section ("DUART_B"))) = { 0 };
19701 char stack[10000] __attribute__ ((section ("STACK"))) = { 0 };
19702 int init_data __attribute__ ((section ("INITDATA")));
19706 /* Initialize stack pointer */
19707 init_sp (stack + sizeof (stack));
19709 /* Initialize initialized data */
19710 memcpy (&init_data, &data, &edata - &data);
19712 /* Turn on the serial ports */
19717 Use the `section' attribute with _global_ variables and not
19718 _local_ variables, as shown in the example.
19720 You may use the `section' attribute with initialized or
19721 uninitialized global variables but the linker requires each object
19722 be defined once, with the exception that uninitialized variables
19723 tentatively go in the `common' (or `bss') section and can be
19724 multiply "defined". Using the `section' attribute will change
19725 what section the variable goes into and may cause the linker to
19726 issue an error if an uninitialized variable has multiple
19727 definitions. You can force a variable to be initialized with the
19728 `-fno-common' flag or the `nocommon' attribute.
19730 Some file formats do not support arbitrary sections so the
19731 `section' attribute is not available on all platforms. If you
19732 need to map the entire contents of a module to a particular
19733 section, consider using the facilities of the linker instead.
19736 On Microsoft Windows, in addition to putting variable definitions
19737 in a named section, the section can also be shared among all
19738 running copies of an executable or DLL. For example, this small
19739 program defines shared data by putting it in a named section
19740 `shared' and marking the section shareable:
19742 int foo __attribute__((section ("shared"), shared)) = 0;
19747 /* Read and write foo. All running
19748 copies see the same value. */
19752 You may only use the `shared' attribute along with `section'
19753 attribute with a fully initialized global definition because of
19754 the way linkers work. See `section' attribute for more
19757 The `shared' attribute is only available on Microsoft Windows.
19759 `tls_model ("TLS_MODEL")'
19760 The `tls_model' attribute sets thread-local storage model (*note
19761 Thread-Local::) of a particular `__thread' variable, overriding
19762 `-ftls-model=' command line switch on a per-variable basis. The
19763 TLS_MODEL argument should be one of `global-dynamic',
19764 `local-dynamic', `initial-exec' or `local-exec'.
19766 Not all targets support this attribute.
19769 This attribute, attached to a variable, means that the variable is
19770 meant to be possibly unused. GCC will not produce a warning for
19774 This attribute, attached to a variable, means that the variable
19775 must be emitted even if it appears that the variable is not
19778 `vector_size (BYTES)'
19779 This attribute specifies the vector size for the variable,
19780 measured in bytes. For example, the declaration:
19782 int foo __attribute__ ((vector_size (16)));
19784 causes the compiler to set the mode for `foo', to be 16 bytes,
19785 divided into `int' sized units. Assuming a 32-bit int (a vector of
19786 4 units of 4 bytes), the corresponding mode of `foo' will be V4SI.
19788 This attribute is only applicable to integral and float scalars,
19789 although arrays, pointers, and function return values are allowed
19790 in conjunction with this construct.
19792 Aggregates with this attribute are invalid, even if they are of
19793 the same size as a corresponding scalar. For example, the
19796 struct S { int a; };
19797 struct S __attribute__ ((vector_size (16))) foo;
19799 is invalid even if the size of the structure is the same as the
19803 The `selectany' attribute causes an initialized global variable to
19804 have link-once semantics. When multiple definitions of the
19805 variable are encountered by the linker, the first is selected and
19806 the remainder are discarded. Following usage by the Microsoft
19807 compiler, the linker is told _not_ to warn about size or content
19808 differences of the multiple definitions.
19810 Although the primary usage of this attribute is for POD types, the
19811 attribute can also be applied to global C++ objects that are
19812 initialized by a constructor. In this case, the static
19813 initialization and destruction code for the object is emitted in
19814 each translation defining the object, but the calls to the
19815 constructor and destructor are protected by a link-once guard
19818 The `selectany' attribute is only available on Microsoft Windows
19819 targets. You can use `__declspec (selectany)' as a synonym for
19820 `__attribute__ ((selectany))' for compatibility with other
19824 The `weak' attribute is described in *Note Function Attributes::.
19827 The `dllimport' attribute is described in *Note Function
19831 The `dllexport' attribute is described in *Note Function
19835 5.34.1 Blackfin Variable Attributes
19836 -----------------------------------
19838 Three attributes are currently defined for the Blackfin.
19845 Use these attributes on the Blackfin to place the variable into L1
19846 Data SRAM. Variables with `l1_data' attribute will be put into
19847 the specific section named `.l1.data'. Those with `l1_data_A'
19848 attribute will be put into the specific section named
19849 `.l1.data.A'. Those with `l1_data_B' attribute will be put into
19850 the specific section named `.l1.data.B'.
19852 5.34.2 M32R/D Variable Attributes
19853 ---------------------------------
19855 One attribute is currently defined for the M32R/D.
19857 `model (MODEL-NAME)'
19858 Use this attribute on the M32R/D to set the addressability of an
19859 object. The identifier MODEL-NAME is one of `small', `medium', or
19860 `large', representing each of the code models.
19862 Small model objects live in the lower 16MB of memory (so that their
19863 addresses can be loaded with the `ld24' instruction).
19865 Medium and large model objects may live anywhere in the 32-bit
19866 address space (the compiler will generate `seth/add3' instructions
19867 to load their addresses).
19869 5.34.3 i386 Variable Attributes
19870 -------------------------------
19872 Two attributes are currently defined for i386 configurations:
19873 `ms_struct' and `gcc_struct'
19877 If `packed' is used on a structure, or if bit-fields are used it
19878 may be that the Microsoft ABI packs them differently than GCC
19879 would normally pack them. Particularly when moving packed data
19880 between functions compiled with GCC and the native Microsoft
19881 compiler (either via function call or as data in a file), it may
19882 be necessary to access either format.
19884 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
19885 Windows X86 compilers to match the native Microsoft compiler.
19887 The Microsoft structure layout algorithm is fairly simple with the
19888 exception of the bitfield packing:
19890 The padding and alignment of members of structures and whether a
19891 bit field can straddle a storage-unit boundary
19893 1. Structure members are stored sequentially in the order in
19894 which they are declared: the first member has the lowest
19895 memory address and the last member the highest.
19897 2. Every data object has an alignment-requirement. The
19898 alignment-requirement for all data except structures, unions,
19899 and arrays is either the size of the object or the current
19900 packing size (specified with either the aligned attribute or
19901 the pack pragma), whichever is less. For structures, unions,
19902 and arrays, the alignment-requirement is the largest
19903 alignment-requirement of its members. Every object is
19904 allocated an offset so that:
19906 offset % alignment-requirement == 0
19908 3. Adjacent bit fields are packed into the same 1-, 2-, or
19909 4-byte allocation unit if the integral types are the same
19910 size and if the next bit field fits into the current
19911 allocation unit without crossing the boundary imposed by the
19912 common alignment requirements of the bit fields.
19914 Handling of zero-length bitfields:
19916 MSVC interprets zero-length bitfields in the following ways:
19918 1. If a zero-length bitfield is inserted between two bitfields
19919 that would normally be coalesced, the bitfields will not be
19926 unsigned long bf_1 : 12;
19928 unsigned long bf_2 : 12;
19931 The size of `t1' would be 8 bytes with the zero-length
19932 bitfield. If the zero-length bitfield were removed, `t1''s
19933 size would be 4 bytes.
19935 2. If a zero-length bitfield is inserted after a bitfield,
19936 `foo', and the alignment of the zero-length bitfield is
19937 greater than the member that follows it, `bar', `bar' will be
19938 aligned as the type of the zero-length bitfield.
19956 For `t2', `bar' will be placed at offset 2, rather than
19957 offset 1. Accordingly, the size of `t2' will be 4. For
19958 `t3', the zero-length bitfield will not affect the alignment
19959 of `bar' or, as a result, the size of the structure.
19961 Taking this into account, it is important to note the
19964 1. If a zero-length bitfield follows a normal bitfield, the
19965 type of the zero-length bitfield may affect the
19966 alignment of the structure as whole. For example, `t2'
19967 has a size of 4 bytes, since the zero-length bitfield
19968 follows a normal bitfield, and is of type short.
19970 2. Even if a zero-length bitfield is not followed by a
19971 normal bitfield, it may still affect the alignment of
19980 Here, `t4' will take up 4 bytes.
19982 3. Zero-length bitfields following non-bitfield members are
19992 Here, `t5' will take up 2 bytes.
19994 5.34.4 PowerPC Variable Attributes
19995 ----------------------------------
19997 Three attributes currently are defined for PowerPC configurations:
19998 `altivec', `ms_struct' and `gcc_struct'.
20000 For full documentation of the struct attributes please see the
20001 documentation in *Note i386 Variable Attributes::.
20003 For documentation of `altivec' attribute please see the documentation
20004 in *Note PowerPC Type Attributes::.
20006 5.34.5 SPU Variable Attributes
20007 ------------------------------
20009 The SPU supports the `spu_vector' attribute for variables. For
20010 documentation of this attribute please see the documentation in *Note
20011 SPU Type Attributes::.
20013 5.34.6 Xstormy16 Variable Attributes
20014 ------------------------------------
20016 One attribute is currently defined for xstormy16 configurations:
20020 If a variable has the `below100' attribute (`BELOW100' is allowed
20021 also), GCC will place the variable in the first 0x100 bytes of
20022 memory and use special opcodes to access it. Such variables will
20023 be placed in either the `.bss_below100' section or the
20024 `.data_below100' section.
20027 5.34.7 AVR Variable Attributes
20028 ------------------------------
20031 The `progmem' attribute is used on the AVR to place data in the
20032 Program Memory address space. The AVR is a Harvard Architecture
20033 processor and data normally resides in the Data Memory address
20037 File: gcc.info, Node: Type Attributes, Next: Alignment, Prev: Variable Attributes, Up: C Extensions
20039 5.35 Specifying Attributes of Types
20040 ===================================
20042 The keyword `__attribute__' allows you to specify special attributes of
20043 `struct' and `union' types when you define such types. This keyword is
20044 followed by an attribute specification inside double parentheses.
20045 Seven attributes are currently defined for types: `aligned', `packed',
20046 `transparent_union', `unused', `deprecated', `visibility', and
20047 `may_alias'. Other attributes are defined for functions (*note
20048 Function Attributes::) and for variables (*note Variable Attributes::).
20050 You may also specify any one of these attributes with `__' preceding
20051 and following its keyword. This allows you to use these attributes in
20052 header files without being concerned about a possible macro of the same
20053 name. For example, you may use `__aligned__' instead of `aligned'.
20055 You may specify type attributes in an enum, struct or union type
20056 declaration or definition, or for other types in a `typedef'
20059 For an enum, struct or union type, you may specify attributes either
20060 between the enum, struct or union tag and the name of the type, or just
20061 past the closing curly brace of the _definition_. The former syntax is
20064 *Note Attribute Syntax::, for details of the exact syntax for using
20067 `aligned (ALIGNMENT)'
20068 This attribute specifies a minimum alignment (in bytes) for
20069 variables of the specified type. For example, the declarations:
20071 struct S { short f[3]; } __attribute__ ((aligned (8)));
20072 typedef int more_aligned_int __attribute__ ((aligned (8)));
20074 force the compiler to insure (as far as it can) that each variable
20075 whose type is `struct S' or `more_aligned_int' will be allocated
20076 and aligned _at least_ on a 8-byte boundary. On a SPARC, having
20077 all variables of type `struct S' aligned to 8-byte boundaries
20078 allows the compiler to use the `ldd' and `std' (doubleword load and
20079 store) instructions when copying one variable of type `struct S' to
20080 another, thus improving run-time efficiency.
20082 Note that the alignment of any given `struct' or `union' type is
20083 required by the ISO C standard to be at least a perfect multiple of
20084 the lowest common multiple of the alignments of all of the members
20085 of the `struct' or `union' in question. This means that you _can_
20086 effectively adjust the alignment of a `struct' or `union' type by
20087 attaching an `aligned' attribute to any one of the members of such
20088 a type, but the notation illustrated in the example above is a
20089 more obvious, intuitive, and readable way to request the compiler
20090 to adjust the alignment of an entire `struct' or `union' type.
20092 As in the preceding example, you can explicitly specify the
20093 alignment (in bytes) that you wish the compiler to use for a given
20094 `struct' or `union' type. Alternatively, you can leave out the
20095 alignment factor and just ask the compiler to align a type to the
20096 maximum useful alignment for the target machine you are compiling
20097 for. For example, you could write:
20099 struct S { short f[3]; } __attribute__ ((aligned));
20101 Whenever you leave out the alignment factor in an `aligned'
20102 attribute specification, the compiler automatically sets the
20103 alignment for the type to the largest alignment which is ever used
20104 for any data type on the target machine you are compiling for.
20105 Doing this can often make copy operations more efficient, because
20106 the compiler can use whatever instructions copy the biggest chunks
20107 of memory when performing copies to or from the variables which
20108 have types that you have aligned this way.
20110 In the example above, if the size of each `short' is 2 bytes, then
20111 the size of the entire `struct S' type is 6 bytes. The smallest
20112 power of two which is greater than or equal to that is 8, so the
20113 compiler sets the alignment for the entire `struct S' type to 8
20116 Note that although you can ask the compiler to select a
20117 time-efficient alignment for a given type and then declare only
20118 individual stand-alone objects of that type, the compiler's
20119 ability to select a time-efficient alignment is primarily useful
20120 only when you plan to create arrays of variables having the
20121 relevant (efficiently aligned) type. If you declare or use arrays
20122 of variables of an efficiently-aligned type, then it is likely
20123 that your program will also be doing pointer arithmetic (or
20124 subscripting, which amounts to the same thing) on pointers to the
20125 relevant type, and the code that the compiler generates for these
20126 pointer arithmetic operations will often be more efficient for
20127 efficiently-aligned types than for other types.
20129 The `aligned' attribute can only increase the alignment; but you
20130 can decrease it by specifying `packed' as well. See below.
20132 Note that the effectiveness of `aligned' attributes may be limited
20133 by inherent limitations in your linker. On many systems, the
20134 linker is only able to arrange for variables to be aligned up to a
20135 certain maximum alignment. (For some linkers, the maximum
20136 supported alignment may be very very small.) If your linker is
20137 only able to align variables up to a maximum of 8 byte alignment,
20138 then specifying `aligned(16)' in an `__attribute__' will still
20139 only provide you with 8 byte alignment. See your linker
20140 documentation for further information.
20143 This attribute, attached to `struct' or `union' type definition,
20144 specifies that each member (other than zero-width bitfields) of
20145 the structure or union is placed to minimize the memory required.
20146 When attached to an `enum' definition, it indicates that the
20147 smallest integral type should be used.
20149 Specifying this attribute for `struct' and `union' types is
20150 equivalent to specifying the `packed' attribute on each of the
20151 structure or union members. Specifying the `-fshort-enums' flag
20152 on the line is equivalent to specifying the `packed' attribute on
20153 all `enum' definitions.
20155 In the following example `struct my_packed_struct''s members are
20156 packed closely together, but the internal layout of its `s' member
20157 is not packed--to do that, `struct my_unpacked_struct' would need
20160 struct my_unpacked_struct
20166 struct __attribute__ ((__packed__)) my_packed_struct
20170 struct my_unpacked_struct s;
20173 You may only specify this attribute on the definition of a `enum',
20174 `struct' or `union', not on a `typedef' which does not also define
20175 the enumerated type, structure or union.
20177 `transparent_union'
20178 This attribute, attached to a `union' type definition, indicates
20179 that any function parameter having that union type causes calls to
20180 that function to be treated in a special way.
20182 First, the argument corresponding to a transparent union type can
20183 be of any type in the union; no cast is required. Also, if the
20184 union contains a pointer type, the corresponding argument can be a
20185 null pointer constant or a void pointer expression; and if the
20186 union contains a void pointer type, the corresponding argument can
20187 be any pointer expression. If the union member type is a pointer,
20188 qualifiers like `const' on the referenced type must be respected,
20189 just as with normal pointer conversions.
20191 Second, the argument is passed to the function using the calling
20192 conventions of the first member of the transparent union, not the
20193 calling conventions of the union itself. All members of the union
20194 must have the same machine representation; this is necessary for
20195 this argument passing to work properly.
20197 Transparent unions are designed for library functions that have
20198 multiple interfaces for compatibility reasons. For example,
20199 suppose the `wait' function must accept either a value of type
20200 `int *' to comply with Posix, or a value of type `union wait *' to
20201 comply with the 4.1BSD interface. If `wait''s parameter were
20202 `void *', `wait' would accept both kinds of arguments, but it
20203 would also accept any other pointer type and this would make
20204 argument type checking less useful. Instead, `<sys/wait.h>' might
20205 define the interface as follows:
20207 typedef union __attribute__ ((__transparent_union__))
20211 } wait_status_ptr_t;
20213 pid_t wait (wait_status_ptr_t);
20215 This interface allows either `int *' or `union wait *' arguments
20216 to be passed, using the `int *' calling convention. The program
20217 can call `wait' with arguments of either type:
20219 int w1 () { int w; return wait (&w); }
20220 int w2 () { union wait w; return wait (&w); }
20222 With this interface, `wait''s implementation might look like this:
20224 pid_t wait (wait_status_ptr_t p)
20226 return waitpid (-1, p.__ip, 0);
20230 When attached to a type (including a `union' or a `struct'), this
20231 attribute means that variables of that type are meant to appear
20232 possibly unused. GCC will not produce a warning for any variables
20233 of that type, even if the variable appears to do nothing. This is
20234 often the case with lock or thread classes, which are usually
20235 defined and then not referenced, but contain constructors and
20236 destructors that have nontrivial bookkeeping functions.
20239 The `deprecated' attribute results in a warning if the type is
20240 used anywhere in the source file. This is useful when identifying
20241 types that are expected to be removed in a future version of a
20242 program. If possible, the warning also includes the location of
20243 the declaration of the deprecated type, to enable users to easily
20244 find further information about why the type is deprecated, or what
20245 they should do instead. Note that the warnings only occur for
20246 uses and then only if the type is being applied to an identifier
20247 that itself is not being declared as deprecated.
20249 typedef int T1 __attribute__ ((deprecated));
20253 typedef T1 T3 __attribute__ ((deprecated));
20254 T3 z __attribute__ ((deprecated));
20256 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
20257 warning is issued for line 4 because T2 is not explicitly
20258 deprecated. Line 5 has no warning because T3 is explicitly
20259 deprecated. Similarly for line 6.
20261 The `deprecated' attribute can also be used for functions and
20262 variables (*note Function Attributes::, *note Variable
20266 Accesses through pointers to types with this attribute are not
20267 subject to type-based alias analysis, but are instead assumed to
20268 be able to alias any other type of objects. In the context of
20269 6.5/7 an lvalue expression dereferencing such a pointer is treated
20270 like having a character type. See `-fstrict-aliasing' for more
20271 information on aliasing issues. This extension exists to support
20272 some vector APIs, in which pointers to one vector type are
20273 permitted to alias pointers to a different vector type.
20275 Note that an object of a type with this attribute does not have any
20280 typedef short __attribute__((__may_alias__)) short_a;
20285 int a = 0x12345678;
20286 short_a *b = (short_a *) &a;
20290 if (a == 0x12345678)
20296 If you replaced `short_a' with `short' in the variable
20297 declaration, the above program would abort when compiled with
20298 `-fstrict-aliasing', which is on by default at `-O2' or above in
20299 recent GCC versions.
20302 In C++, attribute visibility (*note Function Attributes::) can
20303 also be applied to class, struct, union and enum types. Unlike
20304 other type attributes, the attribute must appear between the
20305 initial keyword and the name of the type; it cannot appear after
20306 the body of the type.
20308 Note that the type visibility is applied to vague linkage entities
20309 associated with the class (vtable, typeinfo node, etc.). In
20310 particular, if a class is thrown as an exception in one shared
20311 object and caught in another, the class must have default
20312 visibility. Otherwise the two shared objects will be unable to
20313 use the same typeinfo node and exception handling will break.
20316 5.35.1 ARM Type Attributes
20317 --------------------------
20319 On those ARM targets that support `dllimport' (such as Symbian OS), you
20320 can use the `notshared' attribute to indicate that the virtual table
20321 and other similar data for a class should not be exported from a DLL.
20324 class __declspec(notshared) C {
20326 __declspec(dllimport) C();
20330 __declspec(dllexport)
20333 In this code, `C::C' is exported from the current DLL, but the virtual
20334 table for `C' is not exported. (You can use `__attribute__' instead of
20335 `__declspec' if you prefer, but most Symbian OS code uses `__declspec'.)
20337 5.35.2 i386 Type Attributes
20338 ---------------------------
20340 Two attributes are currently defined for i386 configurations:
20341 `ms_struct' and `gcc_struct'.
20345 If `packed' is used on a structure, or if bit-fields are used it
20346 may be that the Microsoft ABI packs them differently than GCC
20347 would normally pack them. Particularly when moving packed data
20348 between functions compiled with GCC and the native Microsoft
20349 compiler (either via function call or as data in a file), it may
20350 be necessary to access either format.
20352 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
20353 Windows X86 compilers to match the native Microsoft compiler.
20355 To specify multiple attributes, separate them by commas within the
20356 double parentheses: for example, `__attribute__ ((aligned (16),
20359 5.35.3 PowerPC Type Attributes
20360 ------------------------------
20362 Three attributes currently are defined for PowerPC configurations:
20363 `altivec', `ms_struct' and `gcc_struct'.
20365 For full documentation of the `ms_struct' and `gcc_struct' attributes
20366 please see the documentation in *Note i386 Type Attributes::.
20368 The `altivec' attribute allows one to declare AltiVec vector data
20369 types supported by the AltiVec Programming Interface Manual. The
20370 attribute requires an argument to specify one of three vector types:
20371 `vector__', `pixel__' (always followed by unsigned short), and `bool__'
20372 (always followed by unsigned).
20374 __attribute__((altivec(vector__)))
20375 __attribute__((altivec(pixel__))) unsigned short
20376 __attribute__((altivec(bool__))) unsigned
20378 These attributes mainly are intended to support the `__vector',
20379 `__pixel', and `__bool' AltiVec keywords.
20381 5.35.4 SPU Type Attributes
20382 --------------------------
20384 The SPU supports the `spu_vector' attribute for types. This attribute
20385 allows one to declare vector data types supported by the
20386 Sony/Toshiba/IBM SPU Language Extensions Specification. It is intended
20387 to support the `__vector' keyword.
20390 File: gcc.info, Node: Inline, Next: Extended Asm, Prev: Alignment, Up: C Extensions
20392 5.36 An Inline Function is As Fast As a Macro
20393 =============================================
20395 By declaring a function inline, you can direct GCC to make calls to
20396 that function faster. One way GCC can achieve this is to integrate
20397 that function's code into the code for its callers. This makes
20398 execution faster by eliminating the function-call overhead; in
20399 addition, if any of the actual argument values are constant, their
20400 known values may permit simplifications at compile time so that not all
20401 of the inline function's code needs to be included. The effect on code
20402 size is less predictable; object code may be larger or smaller with
20403 function inlining, depending on the particular case. You can also
20404 direct GCC to try to integrate all "simple enough" functions into their
20405 callers with the option `-finline-functions'.
20407 GCC implements three different semantics of declaring a function
20408 inline. One is available with `-std=gnu89' or `-fgnu89-inline' or when
20409 `gnu_inline' attribute is present on all inline declarations, another
20410 when `-std=c99' or `-std=gnu99' (without `-fgnu89-inline'), and the
20411 third is used when compiling C++.
20413 To declare a function inline, use the `inline' keyword in its
20414 declaration, like this:
20422 If you are writing a header file to be included in ISO C89 programs,
20423 write `__inline__' instead of `inline'. *Note Alternate Keywords::.
20425 The three types of inlining behave similarly in two important cases:
20426 when the `inline' keyword is used on a `static' function, like the
20427 example above, and when a function is first declared without using the
20428 `inline' keyword and then is defined with `inline', like this:
20430 extern int inc (int *a);
20437 In both of these common cases, the program behaves the same as if you
20438 had not used the `inline' keyword, except for its speed.
20440 When a function is both inline and `static', if all calls to the
20441 function are integrated into the caller, and the function's address is
20442 never used, then the function's own assembler code is never referenced.
20443 In this case, GCC does not actually output assembler code for the
20444 function, unless you specify the option `-fkeep-inline-functions'.
20445 Some calls cannot be integrated for various reasons (in particular,
20446 calls that precede the function's definition cannot be integrated, and
20447 neither can recursive calls within the definition). If there is a
20448 nonintegrated call, then the function is compiled to assembler code as
20449 usual. The function must also be compiled as usual if the program
20450 refers to its address, because that can't be inlined.
20452 Note that certain usages in a function definition can make it
20453 unsuitable for inline substitution. Among these usages are: use of
20454 varargs, use of alloca, use of variable sized data types (*note
20455 Variable Length::), use of computed goto (*note Labels as Values::),
20456 use of nonlocal goto, and nested functions (*note Nested Functions::).
20457 Using `-Winline' will warn when a function marked `inline' could not be
20458 substituted, and will give the reason for the failure.
20460 As required by ISO C++, GCC considers member functions defined within
20461 the body of a class to be marked inline even if they are not explicitly
20462 declared with the `inline' keyword. You can override this with
20463 `-fno-default-inline'; *note Options Controlling C++ Dialect: C++
20466 GCC does not inline any functions when not optimizing unless you
20467 specify the `always_inline' attribute for the function, like this:
20470 inline void foo (const char) __attribute__((always_inline));
20472 The remainder of this section is specific to GNU C89 inlining.
20474 When an inline function is not `static', then the compiler must assume
20475 that there may be calls from other source files; since a global symbol
20476 can be defined only once in any program, the function must not be
20477 defined in the other source files, so the calls therein cannot be
20478 integrated. Therefore, a non-`static' inline function is always
20479 compiled on its own in the usual fashion.
20481 If you specify both `inline' and `extern' in the function definition,
20482 then the definition is used only for inlining. In no case is the
20483 function compiled on its own, not even if you refer to its address
20484 explicitly. Such an address becomes an external reference, as if you
20485 had only declared the function, and had not defined it.
20487 This combination of `inline' and `extern' has almost the effect of a
20488 macro. The way to use it is to put a function definition in a header
20489 file with these keywords, and put another copy of the definition
20490 (lacking `inline' and `extern') in a library file. The definition in
20491 the header file will cause most calls to the function to be inlined.
20492 If any uses of the function remain, they will refer to the single copy
20496 File: gcc.info, Node: Extended Asm, Next: Constraints, Prev: Inline, Up: C Extensions
20498 5.37 Assembler Instructions with C Expression Operands
20499 ======================================================
20501 In an assembler instruction using `asm', you can specify the operands
20502 of the instruction using C expressions. This means you need not guess
20503 which registers or memory locations will contain the data you want to
20506 You must specify an assembler instruction template much like what
20507 appears in a machine description, plus an operand constraint string for
20510 For example, here is how to use the 68881's `fsinx' instruction:
20512 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
20514 Here `angle' is the C expression for the input operand while `result'
20515 is that of the output operand. Each has `"f"' as its operand
20516 constraint, saying that a floating point register is required. The `='
20517 in `=f' indicates that the operand is an output; all output operands'
20518 constraints must use `='. The constraints use the same language used
20519 in the machine description (*note Constraints::).
20521 Each operand is described by an operand-constraint string followed by
20522 the C expression in parentheses. A colon separates the assembler
20523 template from the first output operand and another separates the last
20524 output operand from the first input, if any. Commas separate the
20525 operands within each group. The total number of operands is currently
20526 limited to 30; this limitation may be lifted in some future version of
20529 If there are no output operands but there are input operands, you must
20530 place two consecutive colons surrounding the place where the output
20533 As of GCC version 3.1, it is also possible to specify input and output
20534 operands using symbolic names which can be referenced within the
20535 assembler code. These names are specified inside square brackets
20536 preceding the constraint string, and can be referenced inside the
20537 assembler code using `%[NAME]' instead of a percentage sign followed by
20538 the operand number. Using named operands the above example could look
20541 asm ("fsinx %[angle],%[output]"
20542 : [output] "=f" (result)
20543 : [angle] "f" (angle));
20545 Note that the symbolic operand names have no relation whatsoever to
20546 other C identifiers. You may use any name you like, even those of
20547 existing C symbols, but you must ensure that no two operands within the
20548 same assembler construct use the same symbolic name.
20550 Output operand expressions must be lvalues; the compiler can check
20551 this. The input operands need not be lvalues. The compiler cannot
20552 check whether the operands have data types that are reasonable for the
20553 instruction being executed. It does not parse the assembler instruction
20554 template and does not know what it means or even whether it is valid
20555 assembler input. The extended `asm' feature is most often used for
20556 machine instructions the compiler itself does not know exist. If the
20557 output expression cannot be directly addressed (for example, it is a
20558 bit-field), your constraint must allow a register. In that case, GCC
20559 will use the register as the output of the `asm', and then store that
20560 register into the output.
20562 The ordinary output operands must be write-only; GCC will assume that
20563 the values in these operands before the instruction are dead and need
20564 not be generated. Extended asm supports input-output or read-write
20565 operands. Use the constraint character `+' to indicate such an operand
20566 and list it with the output operands. You should only use read-write
20567 operands when the constraints for the operand (or the operand in which
20568 only some of the bits are to be changed) allow a register.
20570 You may, as an alternative, logically split its function into two
20571 separate operands, one input operand and one write-only output operand.
20572 The connection between them is expressed by constraints which say they
20573 need to be in the same location when the instruction executes. You can
20574 use the same C expression for both operands, or different expressions.
20575 For example, here we write the (fictitious) `combine' instruction with
20576 `bar' as its read-only source operand and `foo' as its read-write
20579 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
20581 The constraint `"0"' for operand 1 says that it must occupy the same
20582 location as operand 0. A number in constraint is allowed only in an
20583 input operand and it must refer to an output operand.
20585 Only a number in the constraint can guarantee that one operand will be
20586 in the same place as another. The mere fact that `foo' is the value of
20587 both operands is not enough to guarantee that they will be in the same
20588 place in the generated assembler code. The following would not work
20591 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
20593 Various optimizations or reloading could cause operands 0 and 1 to be
20594 in different registers; GCC knows no reason not to do so. For example,
20595 the compiler might find a copy of the value of `foo' in one register and
20596 use it for operand 1, but generate the output operand 0 in a different
20597 register (copying it afterward to `foo''s own address). Of course,
20598 since the register for operand 1 is not even mentioned in the assembler
20599 code, the result will not work, but GCC can't tell that.
20601 As of GCC version 3.1, one may write `[NAME]' instead of the operand
20602 number for a matching constraint. For example:
20604 asm ("cmoveq %1,%2,%[result]"
20605 : [result] "=r"(result)
20606 : "r" (test), "r"(new), "[result]"(old));
20608 Sometimes you need to make an `asm' operand be a specific register,
20609 but there's no matching constraint letter for that register _by
20610 itself_. To force the operand into that register, use a local variable
20611 for the operand and specify the register in the variable declaration.
20612 *Note Explicit Reg Vars::. Then for the `asm' operand, use any
20613 register constraint letter that matches the register:
20615 register int *p1 asm ("r0") = ...;
20616 register int *p2 asm ("r1") = ...;
20617 register int *result asm ("r0");
20618 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
20620 In the above example, beware that a register that is call-clobbered by
20621 the target ABI will be overwritten by any function call in the
20622 assignment, including library calls for arithmetic operators. Also a
20623 register may be clobbered when generating some operations, like
20624 variable shift, memory copy or memory move on x86. Assuming it is a
20625 call-clobbered register, this may happen to `r0' above by the
20626 assignment to `p2'. If you have to use such a register, use temporary
20627 variables for expressions between the register assignment and use:
20630 register int *p1 asm ("r0") = ...;
20631 register int *p2 asm ("r1") = t1;
20632 register int *result asm ("r0");
20633 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
20635 Some instructions clobber specific hard registers. To describe this,
20636 write a third colon after the input operands, followed by the names of
20637 the clobbered hard registers (given as strings). Here is a realistic
20638 example for the VAX:
20640 asm volatile ("movc3 %0,%1,%2"
20642 : "g" (from), "g" (to), "g" (count)
20643 : "r0", "r1", "r2", "r3", "r4", "r5");
20645 You may not write a clobber description in a way that overlaps with an
20646 input or output operand. For example, you may not have an operand
20647 describing a register class with one member if you mention that register
20648 in the clobber list. Variables declared to live in specific registers
20649 (*note Explicit Reg Vars::), and used as asm input or output operands
20650 must have no part mentioned in the clobber description. There is no
20651 way for you to specify that an input operand is modified without also
20652 specifying it as an output operand. Note that if all the output
20653 operands you specify are for this purpose (and hence unused), you will
20654 then also need to specify `volatile' for the `asm' construct, as
20655 described below, to prevent GCC from deleting the `asm' statement as
20658 If you refer to a particular hardware register from the assembler code,
20659 you will probably have to list the register after the third colon to
20660 tell the compiler the register's value is modified. In some assemblers,
20661 the register names begin with `%'; to produce one `%' in the assembler
20662 code, you must write `%%' in the input.
20664 If your assembler instruction can alter the condition code register,
20665 add `cc' to the list of clobbered registers. GCC on some machines
20666 represents the condition codes as a specific hardware register; `cc'
20667 serves to name this register. On other machines, the condition code is
20668 handled differently, and specifying `cc' has no effect. But it is
20669 valid no matter what the machine.
20671 If your assembler instructions access memory in an unpredictable
20672 fashion, add `memory' to the list of clobbered registers. This will
20673 cause GCC to not keep memory values cached in registers across the
20674 assembler instruction and not optimize stores or loads to that memory.
20675 You will also want to add the `volatile' keyword if the memory affected
20676 is not listed in the inputs or outputs of the `asm', as the `memory'
20677 clobber does not count as a side-effect of the `asm'. If you know how
20678 large the accessed memory is, you can add it as input or output but if
20679 this is not known, you should add `memory'. As an example, if you
20680 access ten bytes of a string, you can use a memory input like:
20682 {"m"( ({ struct { char x[10]; } *p = (void *)ptr ; *p; }) )}.
20684 Note that in the following example the memory input is necessary,
20685 otherwise GCC might optimize the store to `x' away:
20691 asm ("magic stuff accessing an 'int' pointed to by '%1'"
20692 "=&d" (r) : "a" (y), "m" (*y));
20696 You can put multiple assembler instructions together in a single `asm'
20697 template, separated by the characters normally used in assembly code
20698 for the system. A combination that works in most places is a newline
20699 to break the line, plus a tab character to move to the instruction field
20700 (written as `\n\t'). Sometimes semicolons can be used, if the
20701 assembler allows semicolons as a line-breaking character. Note that
20702 some assembler dialects use semicolons to start a comment. The input
20703 operands are guaranteed not to use any of the clobbered registers, and
20704 neither will the output operands' addresses, so you can read and write
20705 the clobbered registers as many times as you like. Here is an example
20706 of multiple instructions in a template; it assumes the subroutine
20707 `_foo' accepts arguments in registers 9 and 10:
20709 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
20711 : "g" (from), "g" (to)
20714 Unless an output operand has the `&' constraint modifier, GCC may
20715 allocate it in the same register as an unrelated input operand, on the
20716 assumption the inputs are consumed before the outputs are produced.
20717 This assumption may be false if the assembler code actually consists of
20718 more than one instruction. In such a case, use `&' for each output
20719 operand that may not overlap an input. *Note Modifiers::.
20721 If you want to test the condition code produced by an assembler
20722 instruction, you must include a branch and a label in the `asm'
20723 construct, as follows:
20725 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
20729 This assumes your assembler supports local labels, as the GNU assembler
20730 and most Unix assemblers do.
20732 Speaking of labels, jumps from one `asm' to another are not supported.
20733 The compiler's optimizers do not know about these jumps, and therefore
20734 they cannot take account of them when deciding how to optimize.
20736 Usually the most convenient way to use these `asm' instructions is to
20737 encapsulate them in macros that look like functions. For example,
20740 ({ double __value, __arg = (x); \
20741 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
20744 Here the variable `__arg' is used to make sure that the instruction
20745 operates on a proper `double' value, and to accept only those arguments
20746 `x' which can convert automatically to a `double'.
20748 Another way to make sure the instruction operates on the correct data
20749 type is to use a cast in the `asm'. This is different from using a
20750 variable `__arg' in that it converts more different types. For
20751 example, if the desired type were `int', casting the argument to `int'
20752 would accept a pointer with no complaint, while assigning the argument
20753 to an `int' variable named `__arg' would warn about using a pointer
20754 unless the caller explicitly casts it.
20756 If an `asm' has output operands, GCC assumes for optimization purposes
20757 the instruction has no side effects except to change the output
20758 operands. This does not mean instructions with a side effect cannot be
20759 used, but you must be careful, because the compiler may eliminate them
20760 if the output operands aren't used, or move them out of loops, or
20761 replace two with one if they constitute a common subexpression. Also,
20762 if your instruction does have a side effect on a variable that otherwise
20763 appears not to change, the old value of the variable may be reused later
20764 if it happens to be found in a register.
20766 You can prevent an `asm' instruction from being deleted by writing the
20767 keyword `volatile' after the `asm'. For example:
20769 #define get_and_set_priority(new) \
20771 asm volatile ("get_and_set_priority %0, %1" \
20772 : "=g" (__old) : "g" (new)); \
20775 The `volatile' keyword indicates that the instruction has important
20776 side-effects. GCC will not delete a volatile `asm' if it is reachable.
20777 (The instruction can still be deleted if GCC can prove that
20778 control-flow will never reach the location of the instruction.) Note
20779 that even a volatile `asm' instruction can be moved relative to other
20780 code, including across jump instructions. For example, on many targets
20781 there is a system register which can be set to control the rounding
20782 mode of floating point operations. You might try setting it with a
20783 volatile `asm', like this PowerPC example:
20785 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
20788 This will not work reliably, as the compiler may move the addition back
20789 before the volatile `asm'. To make it work you need to add an
20790 artificial dependency to the `asm' referencing a variable in the code
20791 you don't want moved, for example:
20793 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
20796 Similarly, you can't expect a sequence of volatile `asm' instructions
20797 to remain perfectly consecutive. If you want consecutive output, use a
20798 single `asm'. Also, GCC will perform some optimizations across a
20799 volatile `asm' instruction; GCC does not "forget everything" when it
20800 encounters a volatile `asm' instruction the way some other compilers do.
20802 An `asm' instruction without any output operands will be treated
20803 identically to a volatile `asm' instruction.
20805 It is a natural idea to look for a way to give access to the condition
20806 code left by the assembler instruction. However, when we attempted to
20807 implement this, we found no way to make it work reliably. The problem
20808 is that output operands might need reloading, which would result in
20809 additional following "store" instructions. On most machines, these
20810 instructions would alter the condition code before there was time to
20811 test it. This problem doesn't arise for ordinary "test" and "compare"
20812 instructions because they don't have any output operands.
20814 For reasons similar to those described above, it is not possible to
20815 give an assembler instruction access to the condition code left by
20816 previous instructions.
20818 If you are writing a header file that should be includable in ISO C
20819 programs, write `__asm__' instead of `asm'. *Note Alternate Keywords::.
20821 5.37.1 Size of an `asm'
20822 -----------------------
20824 Some targets require that GCC track the size of each instruction used in
20825 order to generate correct code. Because the final length of an `asm'
20826 is only known by the assembler, GCC must make an estimate as to how big
20827 it will be. The estimate is formed by counting the number of
20828 statements in the pattern of the `asm' and multiplying that by the
20829 length of the longest instruction on that processor. Statements in the
20830 `asm' are identified by newline characters and whatever statement
20831 separator characters are supported by the assembler; on most processors
20832 this is the ``;'' character.
20834 Normally, GCC's estimate is perfectly adequate to ensure that correct
20835 code is generated, but it is possible to confuse the compiler if you use
20836 pseudo instructions or assembler macros that expand into multiple real
20837 instructions or if you use assembler directives that expand to more
20838 space in the object file than would be needed for a single instruction.
20839 If this happens then the assembler will produce a diagnostic saying that
20840 a label is unreachable.
20842 5.37.2 i386 floating point asm operands
20843 ---------------------------------------
20845 There are several rules on the usage of stack-like regs in asm_operands
20846 insns. These rules apply only to the operands that are stack-like regs:
20848 1. Given a set of input regs that die in an asm_operands, it is
20849 necessary to know which are implicitly popped by the asm, and
20850 which must be explicitly popped by gcc.
20852 An input reg that is implicitly popped by the asm must be
20853 explicitly clobbered, unless it is constrained to match an output
20856 2. For any input reg that is implicitly popped by an asm, it is
20857 necessary to know how to adjust the stack to compensate for the
20858 pop. If any non-popped input is closer to the top of the
20859 reg-stack than the implicitly popped reg, it would not be possible
20860 to know what the stack looked like--it's not clear how the rest of
20861 the stack "slides up".
20863 All implicitly popped input regs must be closer to the top of the
20864 reg-stack than any input that is not implicitly popped.
20866 It is possible that if an input dies in an insn, reload might use
20867 the input reg for an output reload. Consider this example:
20869 asm ("foo" : "=t" (a) : "f" (b));
20871 This asm says that input B is not popped by the asm, and that the
20872 asm pushes a result onto the reg-stack, i.e., the stack is one
20873 deeper after the asm than it was before. But, it is possible that
20874 reload will think that it can use the same reg for both the input
20875 and the output, if input B dies in this insn.
20877 If any input operand uses the `f' constraint, all output reg
20878 constraints must use the `&' earlyclobber.
20880 The asm above would be written as
20882 asm ("foo" : "=&t" (a) : "f" (b));
20884 3. Some operands need to be in particular places on the stack. All
20885 output operands fall in this category--there is no other way to
20886 know which regs the outputs appear in unless the user indicates
20887 this in the constraints.
20889 Output operands must specifically indicate which reg an output
20890 appears in after an asm. `=f' is not allowed: the operand
20891 constraints must select a class with a single reg.
20893 4. Output operands may not be "inserted" between existing stack regs.
20894 Since no 387 opcode uses a read/write operand, all output operands
20895 are dead before the asm_operands, and are pushed by the
20896 asm_operands. It makes no sense to push anywhere but the top of
20899 Output operands must start at the top of the reg-stack: output
20900 operands may not "skip" a reg.
20902 5. Some asm statements may need extra stack space for internal
20903 calculations. This can be guaranteed by clobbering stack registers
20904 unrelated to the inputs and outputs.
20907 Here are a couple of reasonable asms to want to write. This asm takes
20908 one input, which is internally popped, and produces two outputs.
20910 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
20912 This asm takes two inputs, which are popped by the `fyl2xp1' opcode,
20913 and replaces them with one output. The user must code the `st(1)'
20914 clobber for reg-stack.c to know that `fyl2xp1' pops both inputs.
20916 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
20919 File: gcc.info, Node: Constraints, Next: Asm Labels, Prev: Extended Asm, Up: C Extensions
20921 5.38 Constraints for `asm' Operands
20922 ===================================
20924 Here are specific details on what constraint letters you can use with
20925 `asm' operands. Constraints can say whether an operand may be in a
20926 register, and which kinds of register; whether the operand can be a
20927 memory reference, and which kinds of address; whether the operand may
20928 be an immediate constant, and which possible values it may have.
20929 Constraints can also require two operands to match.
20933 * Simple Constraints:: Basic use of constraints.
20934 * Multi-Alternative:: When an insn has two alternative constraint-patterns.
20935 * Modifiers:: More precise control over effects of constraints.
20936 * Machine Constraints:: Special constraints for some particular machines.
20939 File: gcc.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints
20941 5.38.1 Simple Constraints
20942 -------------------------
20944 The simplest kind of constraint is a string full of letters, each of
20945 which describes one kind of operand that is permitted. Here are the
20946 letters that are allowed:
20949 Whitespace characters are ignored and can be inserted at any
20950 position except the first. This enables each alternative for
20951 different operands to be visually aligned in the machine
20952 description even if they have different number of constraints and
20956 A memory operand is allowed, with any kind of address that the
20957 machine supports in general. Note that the letter used for the
20958 general memory constraint can be re-defined by a back end using
20959 the `TARGET_MEM_CONSTRAINT' macro.
20962 A memory operand is allowed, but only if the address is
20963 "offsettable". This means that adding a small integer (actually,
20964 the width in bytes of the operand, as determined by its machine
20965 mode) may be added to the address and the result is also a valid
20968 For example, an address which is constant is offsettable; so is an
20969 address that is the sum of a register and a constant (as long as a
20970 slightly larger constant is also within the range of
20971 address-offsets supported by the machine); but an autoincrement or
20972 autodecrement address is not offsettable. More complicated
20973 indirect/indexed addresses may or may not be offsettable depending
20974 on the other addressing modes that the machine supports.
20976 Note that in an output operand which can be matched by another
20977 operand, the constraint letter `o' is valid only when accompanied
20978 by both `<' (if the target machine has predecrement addressing)
20979 and `>' (if the target machine has preincrement addressing).
20982 A memory operand that is not offsettable. In other words,
20983 anything that would fit the `m' constraint but not the `o'
20987 A memory operand with autodecrement addressing (either
20988 predecrement or postdecrement) is allowed.
20991 A memory operand with autoincrement addressing (either
20992 preincrement or postincrement) is allowed.
20995 A register operand is allowed provided that it is in a general
20999 An immediate integer operand (one with constant value) is allowed.
21000 This includes symbolic constants whose values will be known only at
21001 assembly time or later.
21004 An immediate integer operand with a known numeric value is allowed.
21005 Many systems cannot support assembly-time constants for operands
21006 less than a word wide. Constraints for these operands should use
21007 `n' rather than `i'.
21009 `I', `J', `K', ... `P'
21010 Other letters in the range `I' through `P' may be defined in a
21011 machine-dependent fashion to permit immediate integer operands with
21012 explicit integer values in specified ranges. For example, on the
21013 68000, `I' is defined to stand for the range of values 1 to 8.
21014 This is the range permitted as a shift count in the shift
21018 An immediate floating operand (expression code `const_double') is
21019 allowed, but only if the target floating point format is the same
21020 as that of the host machine (on which the compiler is running).
21023 An immediate floating operand (expression code `const_double' or
21024 `const_vector') is allowed.
21027 `G' and `H' may be defined in a machine-dependent fashion to
21028 permit immediate floating operands in particular ranges of values.
21031 An immediate integer operand whose value is not an explicit
21032 integer is allowed.
21034 This might appear strange; if an insn allows a constant operand
21035 with a value not known at compile time, it certainly must allow
21036 any known value. So why use `s' instead of `i'? Sometimes it
21037 allows better code to be generated.
21039 For example, on the 68000 in a fullword instruction it is possible
21040 to use an immediate operand; but if the immediate value is between
21041 -128 and 127, better code results from loading the value into a
21042 register and using the register. This is because the load into
21043 the register can be done with a `moveq' instruction. We arrange
21044 for this to happen by defining the letter `K' to mean "any integer
21045 outside the range -128 to 127", and then specifying `Ks' in the
21046 operand constraints.
21049 Any register, memory or immediate integer operand is allowed,
21050 except for registers that are not general registers.
21053 Any operand whatsoever is allowed.
21055 `0', `1', `2', ... `9'
21056 An operand that matches the specified operand number is allowed.
21057 If a digit is used together with letters within the same
21058 alternative, the digit should come last.
21060 This number is allowed to be more than a single digit. If multiple
21061 digits are encountered consecutively, they are interpreted as a
21062 single decimal integer. There is scant chance for ambiguity,
21063 since to-date it has never been desirable that `10' be interpreted
21064 as matching either operand 1 _or_ operand 0. Should this be
21065 desired, one can use multiple alternatives instead.
21067 This is called a "matching constraint" and what it really means is
21068 that the assembler has only a single operand that fills two roles
21069 which `asm' distinguishes. For example, an add instruction uses
21070 two input operands and an output operand, but on most CISC
21071 machines an add instruction really has only two operands, one of
21072 them an input-output operand:
21076 Matching constraints are used in these circumstances. More
21077 precisely, the two operands that match must include one input-only
21078 operand and one output-only operand. Moreover, the digit must be a
21079 smaller number than the number of the operand that uses it in the
21083 An operand that is a valid memory address is allowed. This is for
21084 "load address" and "push address" instructions.
21086 `p' in the constraint must be accompanied by `address_operand' as
21087 the predicate in the `match_operand'. This predicate interprets
21088 the mode specified in the `match_operand' as the mode of the memory
21089 reference for which the address would be valid.
21092 Other letters can be defined in machine-dependent fashion to stand
21093 for particular classes of registers or other arbitrary operand
21094 types. `d', `a' and `f' are defined on the 68000/68020 to stand
21095 for data, address and floating point registers.
21098 File: gcc.info, Node: Multi-Alternative, Next: Modifiers, Prev: Simple Constraints, Up: Constraints
21100 5.38.2 Multiple Alternative Constraints
21101 ---------------------------------------
21103 Sometimes a single instruction has multiple alternative sets of possible
21104 operands. For example, on the 68000, a logical-or instruction can
21105 combine register or an immediate value into memory, or it can combine
21106 any kind of operand into a register; but it cannot combine one memory
21107 location into another.
21109 These constraints are represented as multiple alternatives. An
21110 alternative can be described by a series of letters for each operand.
21111 The overall constraint for an operand is made from the letters for this
21112 operand from the first alternative, a comma, the letters for this
21113 operand from the second alternative, a comma, and so on until the last
21116 If all the operands fit any one alternative, the instruction is valid.
21117 Otherwise, for each alternative, the compiler counts how many
21118 instructions must be added to copy the operands so that that
21119 alternative applies. The alternative requiring the least copying is
21120 chosen. If two alternatives need the same amount of copying, the one
21121 that comes first is chosen. These choices can be altered with the `?'
21122 and `!' characters:
21125 Disparage slightly the alternative that the `?' appears in, as a
21126 choice when no alternative applies exactly. The compiler regards
21127 this alternative as one unit more costly for each `?' that appears
21131 Disparage severely the alternative that the `!' appears in. This
21132 alternative can still be used if it fits without reloading, but if
21133 reloading is needed, some other alternative will be used.
21136 File: gcc.info, Node: Modifiers, Next: Machine Constraints, Prev: Multi-Alternative, Up: Constraints
21138 5.38.3 Constraint Modifier Characters
21139 -------------------------------------
21141 Here are constraint modifier characters.
21144 Means that this operand is write-only for this instruction: the
21145 previous value is discarded and replaced by output data.
21148 Means that this operand is both read and written by the
21151 When the compiler fixes up the operands to satisfy the constraints,
21152 it needs to know which operands are inputs to the instruction and
21153 which are outputs from it. `=' identifies an output; `+'
21154 identifies an operand that is both input and output; all other
21155 operands are assumed to be input only.
21157 If you specify `=' or `+' in a constraint, you put it in the first
21158 character of the constraint string.
21161 Means (in a particular alternative) that this operand is an
21162 "earlyclobber" operand, which is modified before the instruction is
21163 finished using the input operands. Therefore, this operand may
21164 not lie in a register that is used as an input operand or as part
21165 of any memory address.
21167 `&' applies only to the alternative in which it is written. In
21168 constraints with multiple alternatives, sometimes one alternative
21169 requires `&' while others do not. See, for example, the `movdf'
21172 An input operand can be tied to an earlyclobber operand if its only
21173 use as an input occurs before the early result is written. Adding
21174 alternatives of this form often allows GCC to produce better code
21175 when only some of the inputs can be affected by the earlyclobber.
21176 See, for example, the `mulsi3' insn of the ARM.
21178 `&' does not obviate the need to write `='.
21181 Declares the instruction to be commutative for this operand and the
21182 following operand. This means that the compiler may interchange
21183 the two operands if that is the cheapest way to make all operands
21184 fit the constraints. GCC can only handle one commutative pair in
21185 an asm; if you use more, the compiler may fail. Note that you
21186 need not use the modifier if the two alternatives are strictly
21187 identical; this would only waste time in the reload pass. The
21188 modifier is not operational after register allocation, so the
21189 result of `define_peephole2' and `define_split's performed after
21190 reload cannot rely on `%' to make the intended insn match.
21193 Says that all following characters, up to the next comma, are to be
21194 ignored as a constraint. They are significant only for choosing
21195 register preferences.
21198 Says that the following character should be ignored when choosing
21199 register preferences. `*' has no effect on the meaning of the
21200 constraint as a constraint, and no effect on reloading.
21204 File: gcc.info, Node: Machine Constraints, Prev: Modifiers, Up: Constraints
21206 5.38.4 Constraints for Particular Machines
21207 ------------------------------------------
21209 Whenever possible, you should use the general-purpose constraint letters
21210 in `asm' arguments, since they will convey meaning more readily to
21211 people reading your code. Failing that, use the constraint letters
21212 that usually have very similar meanings across architectures. The most
21213 commonly used constraints are `m' and `r' (for memory and
21214 general-purpose registers respectively; *note Simple Constraints::), and
21215 `I', usually the letter indicating the most common immediate-constant
21218 Each architecture defines additional constraints. These constraints
21219 are used by the compiler itself for instruction generation, as well as
21220 for `asm' statements; therefore, some of the constraints are not
21221 particularly useful for `asm'. Here is a summary of some of the
21222 machine-dependent constraints available on some particular machines; it
21223 includes both constraints that are useful for `asm' and constraints
21224 that aren't. The compiler source file mentioned in the table heading
21225 for each architecture is the definitive reference for the meanings of
21226 that architecture's constraints.
21228 _ARM family--`config/arm/arm.h'_
21231 Floating-point register
21234 VFP floating-point register
21237 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0,
21241 Floating-point constant that would satisfy the constraint `F'
21245 Integer that is valid as an immediate operand in a data
21246 processing instruction. That is, an integer in the range 0
21247 to 255 rotated by a multiple of 2
21250 Integer in the range -4095 to 4095
21253 Integer that satisfies constraint `I' when inverted (ones
21257 Integer that satisfies constraint `I' when negated (twos
21261 Integer in the range 0 to 32
21264 A memory reference where the exact address is in a single
21265 register (``m'' is preferable for `asm' statements)
21268 An item in the constant pool
21271 A symbol in the text segment of the current file
21274 A memory reference suitable for VFP load/store insns
21275 (reg+constant offset)
21278 A memory reference suitable for iWMMXt load/store
21282 A memory reference suitable for the ARMv4 ldrsb instruction.
21284 _AVR family--`config/avr/constraints.md'_
21287 Registers from r0 to r15
21290 Registers from r16 to r23
21293 Registers from r16 to r31
21296 Registers from r24 to r31. These registers can be used in
21300 Pointer register (r26-r31)
21303 Base pointer register (r28-r31)
21306 Stack pointer register (SPH:SPL)
21309 Temporary register r0
21312 Register pair X (r27:r26)
21315 Register pair Y (r29:r28)
21318 Register pair Z (r31:r30)
21321 Constant greater than -1, less than 64
21324 Constant greater than -64, less than 1
21333 Constant that fits in 8 bits
21336 Constant integer -1
21339 Constant integer 8, 16, or 24
21345 A floating point constant 0.0
21348 Integer constant in the range -6 ... 5.
21351 A memory address based on Y or Z pointer with displacement.
21353 _CRX Architecture--`config/crx/crx.h'_
21356 Registers from r0 to r14 (registers without stack pointer)
21359 Register r16 (64-bit accumulator lo register)
21362 Register r17 (64-bit accumulator hi register)
21365 Register pair r16-r17. (64-bit accumulator lo-hi pair)
21368 Constant that fits in 3 bits
21371 Constant that fits in 4 bits
21374 Constant that fits in 5 bits
21377 Constant that is one of -1, 4, -4, 7, 8, 12, 16, 20, 32, 48
21380 Floating point constant that is legal for store immediate
21382 _Hewlett-Packard PA-RISC--`config/pa/pa.h'_
21388 Floating point register
21391 Shift amount register
21394 Floating point register (deprecated)
21397 Upper floating point register (32-bit), floating point
21404 Signed 11-bit integer constant
21407 Signed 14-bit integer constant
21410 Integer constant that can be deposited with a `zdepi'
21414 Signed 5-bit integer constant
21420 Integer constant that can be loaded with a `ldil' instruction
21423 Integer constant whose value plus one is a power of 2
21426 Integer constant that can be used for `and' operations in
21427 `depi' and `extru' instructions
21430 Integer constant 31
21433 Integer constant 63
21436 Floating-point constant 0.0
21439 A `lo_sum' data-linkage-table memory operand
21442 A memory operand that can be used as the destination operand
21443 of an integer store instruction
21446 A scaled or unscaled indexed memory operand
21449 A memory operand for floating-point loads and stores
21452 A register indirect memory operand
21454 _picoChip family--`picochip.h'_
21460 Pointer register. A register which can be used to access
21461 memory without supplying an offset. Any other register can
21462 be used to access memory, but will need a constant offset.
21463 In the case of the offset being zero, it is more efficient to
21464 use a pointer register, since this reduces code size.
21467 A twin register. A register which may be paired with an
21468 adjacent register to create a 32-bit register.
21471 Any absolute memory address (e.g., symbolic constant, symbolic
21472 constant + offset).
21475 4-bit signed integer.
21478 4-bit unsigned integer.
21481 8-bit signed integer.
21484 Any constant whose absolute value is no greater than 4-bits.
21487 10-bit signed integer
21490 16-bit signed integer.
21493 _PowerPC and IBM RS6000--`config/rs6000/rs6000.h'_
21496 Address base register
21499 Floating point register
21505 `MQ', `CTR', or `LINK' register
21517 `CR' register (condition register) number 0
21520 `CR' register (condition register)
21523 `FPMEM' stack memory for FPR-GPR transfers
21526 Signed 16-bit constant
21529 Unsigned 16-bit constant shifted left 16 bits (use `L'
21530 instead for `SImode' constants)
21533 Unsigned 16-bit constant
21536 Signed 16-bit constant shifted left 16 bits
21539 Constant larger than 31
21548 Constant whose negation is a signed 16-bit constant
21551 Floating point constant that can be loaded into a register
21552 with one instruction per word
21555 Integer/Floating point constant that can be loaded into a
21556 register using three instructions
21559 Memory operand that is an offset from a register (`m' is
21560 preferable for `asm' statements)
21563 Memory operand that is an indexed or indirect from a register
21564 (`m' is preferable for `asm' statements)
21570 Address operand that is an indexed or indirect from a
21571 register (`p' is preferable for `asm' statements)
21574 Constant suitable as a 64-bit mask operand
21577 Constant suitable as a 32-bit mask operand
21580 System V Release 4 small data area reference
21583 AND masks that can be performed by two rldic{l, r}
21587 Vector constant that does not require memory
21590 _Intel 386--`config/i386/constraints.md'_
21593 Legacy register--the eight integer registers available on all
21594 i386 processors (`a', `b', `c', `d', `si', `di', `bp', `sp').
21597 Any register accessible as `Rl'. In 32-bit mode, `a', `b',
21598 `c', and `d'; in 64-bit mode, any integer register.
21601 Any register accessible as `Rh': `a', `b', `c', and `d'.
21622 The `a' and `d' registers, as a pair (for instructions that
21623 return half the result in one and half in the other).
21626 Any 80387 floating-point (stack) register.
21629 Top of 80387 floating-point stack (`%st(0)').
21632 Second from top of 80387 floating-point stack (`%st(1)').
21641 First SSE register (`%xmm0').
21644 Integer constant in the range 0 ... 31, for 32-bit shifts.
21647 Integer constant in the range 0 ... 63, for 64-bit shifts.
21650 Signed 8-bit integer constant.
21653 `0xFF' or `0xFFFF', for andsi as a zero-extending move.
21656 0, 1, 2, or 3 (shifts for the `lea' instruction).
21659 Unsigned 8-bit integer constant (for `in' and `out'
21663 Standard 80387 floating point constant.
21666 Standard SSE floating point constant.
21669 32-bit signed integer constant, or a symbolic reference known
21670 to fit that range (for immediate operands in sign-extending
21671 x86-64 instructions).
21674 32-bit unsigned integer constant, or a symbolic reference
21675 known to fit that range (for immediate operands in
21676 zero-extending x86-64 instructions).
21679 _Intel IA-64--`config/ia64/ia64.h'_
21682 General register `r0' to `r3' for `addl' instruction
21688 Predicate register (`c' as in "conditional")
21691 Application register residing in M-unit
21694 Application register residing in I-unit
21697 Floating-point register
21700 Memory operand. Remember that `m' allows postincrement and
21701 postdecrement which require printing with `%Pn' on IA-64.
21702 Use `S' to disallow postincrement and postdecrement.
21705 Floating-point constant 0.0 or 1.0
21708 14-bit signed integer constant
21711 22-bit signed integer constant
21714 8-bit signed integer constant for logical instructions
21717 8-bit adjusted signed integer constant for compare pseudo-ops
21720 6-bit unsigned integer constant for shift counts
21723 9-bit signed integer constant for load and store
21730 0 or -1 for `dep' instruction
21733 Non-volatile memory for floating-point loads and stores
21736 Integer constant in the range 1 to 4 for `shladd' instruction
21739 Memory operand except postincrement and postdecrement
21741 _FRV--`config/frv/frv.h'_
21744 Register in the class `ACC_REGS' (`acc0' to `acc7').
21747 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7').
21750 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0'
21754 Register in the class `GPR_REGS' (`gr0' to `gr63').
21757 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd
21758 registers are excluded not in the class but through the use
21759 of a machine mode larger than 4 bytes.
21762 Register in the class `FPR_REGS' (`fr0' to `fr63').
21765 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd
21766 registers are excluded not in the class but through the use
21767 of a machine mode larger than 4 bytes.
21770 Register in the class `LR_REG' (the `lr' register).
21773 Register in the class `QUAD_REGS' (`gr2' to `gr63').
21774 Register numbers not divisible by 4 are excluded not in the
21775 class but through the use of a machine mode larger than 8
21779 Register in the class `ICC_REGS' (`icc0' to `icc3').
21782 Register in the class `FCC_REGS' (`fcc0' to `fcc3').
21785 Register in the class `ICR_REGS' (`cc4' to `cc7').
21788 Register in the class `FCR_REGS' (`cc0' to `cc3').
21791 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63').
21792 Register numbers not divisible by 4 are excluded not in the
21793 class but through the use of a machine mode larger than 8
21797 Register in the class `SPR_REGS' (`lcr' and `lr').
21800 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7').
21803 Register in the class `ACCG_REGS' (`accg0' to `accg7').
21806 Register in the class `CR_REGS' (`cc0' to `cc7').
21809 Floating point constant zero
21812 6-bit signed integer constant
21815 10-bit signed integer constant
21818 16-bit signed integer constant
21821 16-bit unsigned integer constant
21824 12-bit signed integer constant that is negative--i.e. in the
21825 range of -2048 to -1
21831 12-bit signed integer constant that is greater than
21832 zero--i.e. in the range of 1 to 2047.
21835 _Blackfin family--`config/bfin/constraints.md'_
21844 A call clobbered P register.
21847 A single register. If N is in the range 0 to 7, the
21848 corresponding D register. If it is `A', then the register P0.
21851 Even-numbered D register
21854 Odd-numbered D register
21857 Accumulator register.
21860 Even-numbered accumulator register.
21863 Odd-numbered accumulator register.
21875 Registers used for circular buffering, i.e. I, B, or L
21891 Any D, P, B, M, I or L register.
21894 Additional registers typically used only in prologues and
21895 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and
21899 Any register except accumulators or CC.
21902 Signed 16 bit integer (in the range -32768 to 32767)
21905 Unsigned 16 bit integer (in the range 0 to 65535)
21908 Signed 7 bit integer (in the range -64 to 63)
21911 Unsigned 7 bit integer (in the range 0 to 127)
21914 Unsigned 5 bit integer (in the range 0 to 31)
21917 Signed 4 bit integer (in the range -8 to 7)
21920 Signed 3 bit integer (in the range -3 to 4)
21923 Unsigned 3 bit integer (in the range 0 to 7)
21926 Constant N, where N is a single-digit constant in the range 0
21930 An integer equal to one of the MACFLAG_XXX constants that is
21931 suitable for use with either accumulator.
21934 An integer equal to one of the MACFLAG_XXX constants that is
21935 suitable for use only with accumulator A1.
21944 An integer constant with exactly a single bit set.
21947 An integer constant with all bits set except exactly one.
21954 _M32C--`config/m32c/m32c.c'_
21959 `$sp', `$fb', `$sb'.
21962 Any control register, when they're 16 bits wide (nothing if
21963 control registers are 24 bits wide)
21966 Any control register, when they're 24 bits wide.
21972 $r0, $r1, $r2, $r3.
21975 $r0 or $r2, or $r2r0 for 32 bit values.
21978 $r1 or $r3, or $r3r1 for 32 bit values.
21981 A register that can hold a 64 bit value.
21984 $r0 or $r1 (registers with addressable high/low bytes)
21993 Address registers when they're 16 bits wide.
21996 Address registers when they're 24 bits wide.
21999 Registers that can hold QI values.
22002 Registers that can be used with displacements ($a0, $a1, $sb).
22005 Registers that can hold 32 bit values.
22008 Registers that can hold 16 bit values.
22011 Registers chat can hold 16 bit values, including all control
22015 $r0 through R1, plus $a0 and $a1.
22018 The flags register.
22021 The memory-based pseudo-registers $mem0 through $mem15.
22024 Registers that can hold pointers (16 bit registers for r8c,
22025 m16c; 24 bit registers for m32cm, m32c).
22028 Matches multiple registers in a PARALLEL to form a larger
22029 register. Used to match function return values.
22044 -8 ... -1 or 1 ... 8
22047 -16 ... -1 or 1 ... 16
22050 -32 ... -1 or 1 ... 32
22056 An 8 bit value with exactly one bit set.
22059 A 16 bit value with exactly one bit set.
22062 The common src/dest memory addressing modes.
22065 Memory addressed using $a0 or $a1.
22068 Memory addressed with immediate addresses.
22071 Memory addressed using the stack pointer ($sp).
22074 Memory addressed using the frame base register ($fb).
22077 Memory addressed using the small base register ($sb).
22082 _MIPS--`config/mips/constraints.md'_
22085 An address register. This is equivalent to `r' unless
22086 generating MIPS16 code.
22089 A floating-point register (if available).
22092 Formerly the `hi' register. This constraint is no longer
22096 The `lo' register. Use this register to store values that are
22097 no bigger than a word.
22100 The concatenated `hi' and `lo' registers. Use this register
22101 to store doubleword values.
22104 A register suitable for use in an indirect jump. This will
22105 always be `$25' for `-mabicalls'.
22108 Register `$3'. Do not use this constraint in new code; it is
22109 retained only for compatibility with glibc.
22112 Equivalent to `r'; retained for backwards compatibility.
22115 A floating-point condition code register.
22118 A signed 16-bit constant (for arithmetic instructions).
22124 An unsigned 16-bit constant (for logic instructions).
22127 A signed 32-bit constant in which the lower 16 bits are zero.
22128 Such constants can be loaded using `lui'.
22131 A constant that cannot be loaded using `lui', `addiu' or
22135 A constant in the range -65535 to -1 (inclusive).
22138 A signed 15-bit constant.
22141 A constant in the range 1 to 65535 (inclusive).
22144 Floating-point zero.
22147 An address that can be used in a non-macro load or store.
22149 _Motorola 680x0--`config/m68k/constraints.md'_
22158 68881 floating-point register, if available
22161 Integer in the range 1 to 8
22164 16-bit signed number
22167 Signed number whose magnitude is greater than 0x80
22170 Integer in the range -8 to -1
22173 Signed number whose magnitude is greater than 0x100
22176 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate
22179 16 (for rotate using swap)
22182 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate
22185 Numbers that mov3q can handle
22188 Floating point constant that is not a 68881 constant
22191 Operands that satisfy 'm' when -mpcrel is in effect
22194 Operands that satisfy 's' when -mpcrel is not in effect
22197 Address register indirect addressing mode
22200 Register offset addressing
22206 symbol_ref or const
22215 Range of signed numbers that don't fit in 16 bits
22218 Integers valid for mvq
22221 Integers valid for a moveq followed by a swap
22224 Integers valid for mvz
22227 Integers valid for mvs
22233 Non-register operands allowed in clr
22236 _Motorola 68HC11 & 68HC12 families--`config/m68hc11/m68hc11.h'_
22251 Temporary soft register _.tmp
22254 A soft register _.d1 to _.d31
22257 Stack pointer register
22266 Pseudo register `z' (replaced by `x' or `y' at the end)
22269 An address register: x, y or z
22272 An address register: x or y
22275 Register pair (x:d) to form a 32-bit value
22278 Constants in the range -65536 to 65535
22281 Constants whose 16-bit low part is zero
22284 Constant integer 1 or -1
22287 Constant integer 16
22290 Constants in the range -8 to 2
22293 _SPARC--`config/sparc/sparc.h'_
22296 Floating-point register on the SPARC-V8 architecture and
22297 lower floating-point register on the SPARC-V9 architecture.
22300 Floating-point register. It is equivalent to `f' on the
22301 SPARC-V8 architecture and contains both lower and upper
22302 floating-point registers on the SPARC-V9 architecture.
22305 Floating-point condition code register.
22308 Lower floating-point register. It is only valid on the
22309 SPARC-V9 architecture when the Visual Instruction Set is
22313 Floating-point register. It is only valid on the SPARC-V9
22314 architecture when the Visual Instruction Set is available.
22317 64-bit global or out register for the SPARC-V8+ architecture.
22323 Signed 13-bit constant
22329 32-bit constant with the low 12 bits clear (a constant that
22330 can be loaded with the `sethi' instruction)
22333 A constant in the range supported by `movcc' instructions
22336 A constant in the range supported by `movrcc' instructions
22339 Same as `K', except that it verifies that bits that are not
22340 in the lower 32-bit range are all zero. Must be used instead
22341 of `K' for modes wider than `SImode'
22347 Floating-point zero
22350 Signed 13-bit constant, sign-extended to 32 or 64 bits
22353 Floating-point constant whose integral representation can be
22354 moved into an integer register using a single sethi
22358 Floating-point constant whose integral representation can be
22359 moved into an integer register using a single mov instruction
22362 Floating-point constant whose integral representation can be
22363 moved into an integer register using a high/lo_sum
22364 instruction sequence
22367 Memory address aligned to an 8-byte boundary
22373 Memory address for `e' constraint registers
22379 _SPU--`config/spu/spu.h'_
22382 An immediate which can be loaded with the il/ila/ilh/ilhu
22383 instructions. const_int is treated as a 64 bit value.
22386 An immediate for and/xor/or instructions. const_int is
22387 treated as a 64 bit value.
22390 An immediate for the `iohl' instruction. const_int is
22391 treated as a 64 bit value.
22394 An immediate which can be loaded with `fsmbi'.
22397 An immediate which can be loaded with the il/ila/ilh/ilhu
22398 instructions. const_int is treated as a 32 bit value.
22401 An immediate for most arithmetic instructions. const_int is
22402 treated as a 32 bit value.
22405 An immediate for and/xor/or instructions. const_int is
22406 treated as a 32 bit value.
22409 An immediate for the `iohl' instruction. const_int is
22410 treated as a 32 bit value.
22413 A constant in the range [-64, 63] for shift/rotate
22417 An unsigned 7-bit constant for conversion/nop/channel
22421 A signed 10-bit constant for most arithmetic instructions.
22424 A signed 16 bit immediate for `stop'.
22427 An unsigned 16-bit constant for `iohl' and `fsmbi'.
22430 An unsigned 7-bit constant whose 3 least significant bits are
22434 An unsigned 3-bit constant for 16-byte rotates and shifts
22437 Call operand, reg, for indirect calls
22440 Call operand, symbol, for relative calls.
22443 Call operand, const_int, for absolute calls.
22446 An immediate which can be loaded with the il/ila/ilh/ilhu
22447 instructions. const_int is sign extended to 128 bit.
22450 An immediate for shift and rotate instructions. const_int is
22451 treated as a 32 bit value.
22454 An immediate for and/xor/or instructions. const_int is sign
22455 extended as a 128 bit.
22458 An immediate for the `iohl' instruction. const_int is sign
22459 extended to 128 bit.
22462 _S/390 and zSeries--`config/s390/s390.h'_
22465 Address register (general purpose register except r0)
22468 Condition code register
22471 Data register (arbitrary general purpose register)
22474 Floating-point register
22477 Unsigned 8-bit constant (0-255)
22480 Unsigned 12-bit constant (0-4095)
22483 Signed 16-bit constant (-32768-32767)
22486 Value appropriate as displacement.
22488 for short displacement
22490 `(-524288..524287)'
22491 for long displacement
22494 Constant integer with a value of 0x7fffffff.
22497 Multiple letter constraint followed by 4 parameter letters.
22499 number of the part counting from most to least
22506 mode of the containing operand
22509 value of the other parts (F--all bits set)
22510 The constraint matches if the specified part of a constant
22511 has a value different from its other parts.
22514 Memory reference without index register and with short
22518 Memory reference with index register and short displacement.
22521 Memory reference without index register but with long
22525 Memory reference with index register and long displacement.
22528 Pointer with short displacement.
22531 Pointer with long displacement.
22534 Shift count operand.
22537 _Score family--`config/score/score.h'_
22540 Registers from r0 to r32.
22543 Registers from r0 to r16.
22546 r8--r11 or r22--r27 registers.
22567 cnt + lcb + scb register.
22570 cr0--cr15 register.
22582 cp1 + cp2 + cp3 registers.
22585 High 16-bit constant (32-bit constant with 16 LSBs zero).
22588 Unsigned 5 bit integer (in the range 0 to 31).
22591 Unsigned 16 bit integer (in the range 0 to 65535).
22594 Signed 16 bit integer (in the range -32768 to 32767).
22597 Unsigned 14 bit integer (in the range 0 to 16383).
22600 Signed 14 bit integer (in the range -8192 to 8191).
22605 _Xstormy16--`config/stormy16/stormy16.h'_
22620 Registers r0 through r7.
22623 Registers r0 and r1.
22626 The carry register.
22629 Registers r8 and r9.
22632 A constant between 0 and 3 inclusive.
22635 A constant that has exactly one bit set.
22638 A constant that has exactly one bit clear.
22641 A constant between 0 and 255 inclusive.
22644 A constant between -255 and 0 inclusive.
22647 A constant between -3 and 0 inclusive.
22650 A constant between 1 and 4 inclusive.
22653 A constant between -4 and -1 inclusive.
22656 A memory reference that is a stack push.
22659 A memory reference that is a stack pop.
22662 A memory reference that refers to a constant address of known
22666 The register indicated by Rx (not implemented yet).
22669 A constant that is not between 2 and 15 inclusive.
22675 _Xtensa--`config/xtensa/constraints.md'_
22678 General-purpose 32-bit register
22681 One-bit boolean register
22684 MAC16 40-bit accumulator register
22687 Signed 12-bit integer constant, for use in MOVI instructions
22690 Signed 8-bit integer constant, for use in ADDI instructions
22693 Integer constant valid for BccI instructions
22696 Unsigned constant valid for BccUI instructions
22701 File: gcc.info, Node: Asm Labels, Next: Explicit Reg Vars, Prev: Constraints, Up: C Extensions
22703 5.39 Controlling Names Used in Assembler Code
22704 =============================================
22706 You can specify the name to be used in the assembler code for a C
22707 function or variable by writing the `asm' (or `__asm__') keyword after
22708 the declarator as follows:
22710 int foo asm ("myfoo") = 2;
22712 This specifies that the name to be used for the variable `foo' in the
22713 assembler code should be `myfoo' rather than the usual `_foo'.
22715 On systems where an underscore is normally prepended to the name of a C
22716 function or variable, this feature allows you to define names for the
22717 linker that do not start with an underscore.
22719 It does not make sense to use this feature with a non-static local
22720 variable since such variables do not have assembler names. If you are
22721 trying to put the variable in a particular register, see *Note Explicit
22722 Reg Vars::. GCC presently accepts such code with a warning, but will
22723 probably be changed to issue an error, rather than a warning, in the
22726 You cannot use `asm' in this way in a function _definition_; but you
22727 can get the same effect by writing a declaration for the function
22728 before its definition and putting `asm' there, like this:
22730 extern func () asm ("FUNC");
22736 It is up to you to make sure that the assembler names you choose do not
22737 conflict with any other assembler symbols. Also, you must not use a
22738 register name; that would produce completely invalid assembler code.
22739 GCC does not as yet have the ability to store static variables in
22740 registers. Perhaps that will be added.
22743 File: gcc.info, Node: Explicit Reg Vars, Next: Alternate Keywords, Prev: Asm Labels, Up: C Extensions
22745 5.40 Variables in Specified Registers
22746 =====================================
22748 GNU C allows you to put a few global variables into specified hardware
22749 registers. You can also specify the register in which an ordinary
22750 register variable should be allocated.
22752 * Global register variables reserve registers throughout the program.
22753 This may be useful in programs such as programming language
22754 interpreters which have a couple of global variables that are
22755 accessed very often.
22757 * Local register variables in specific registers do not reserve the
22758 registers, except at the point where they are used as input or
22759 output operands in an `asm' statement and the `asm' statement
22760 itself is not deleted. The compiler's data flow analysis is
22761 capable of determining where the specified registers contain live
22762 values, and where they are available for other uses. Stores into
22763 local register variables may be deleted when they appear to be
22764 dead according to dataflow analysis. References to local register
22765 variables may be deleted or moved or simplified.
22767 These local variables are sometimes convenient for use with the
22768 extended `asm' feature (*note Extended Asm::), if you want to
22769 write one output of the assembler instruction directly into a
22770 particular register. (This will work provided the register you
22771 specify fits the constraints specified for that operand in the
22776 * Global Reg Vars::
22780 File: gcc.info, Node: Global Reg Vars, Next: Local Reg Vars, Up: Explicit Reg Vars
22782 5.40.1 Defining Global Register Variables
22783 -----------------------------------------
22785 You can define a global register variable in GNU C like this:
22787 register int *foo asm ("a5");
22789 Here `a5' is the name of the register which should be used. Choose a
22790 register which is normally saved and restored by function calls on your
22791 machine, so that library routines will not clobber it.
22793 Naturally the register name is cpu-dependent, so you would need to
22794 conditionalize your program according to cpu type. The register `a5'
22795 would be a good choice on a 68000 for a variable of pointer type. On
22796 machines with register windows, be sure to choose a "global" register
22797 that is not affected magically by the function call mechanism.
22799 In addition, operating systems on one type of cpu may differ in how
22800 they name the registers; then you would need additional conditionals.
22801 For example, some 68000 operating systems call this register `%a5'.
22803 Eventually there may be a way of asking the compiler to choose a
22804 register automatically, but first we need to figure out how it should
22805 choose and how to enable you to guide the choice. No solution is
22808 Defining a global register variable in a certain register reserves that
22809 register entirely for this use, at least within the current compilation.
22810 The register will not be allocated for any other purpose in the
22811 functions in the current compilation. The register will not be saved
22812 and restored by these functions. Stores into this register are never
22813 deleted even if they would appear to be dead, but references may be
22814 deleted or moved or simplified.
22816 It is not safe to access the global register variables from signal
22817 handlers, or from more than one thread of control, because the system
22818 library routines may temporarily use the register for other things
22819 (unless you recompile them specially for the task at hand).
22821 It is not safe for one function that uses a global register variable to
22822 call another such function `foo' by way of a third function `lose' that
22823 was compiled without knowledge of this variable (i.e. in a different
22824 source file in which the variable wasn't declared). This is because
22825 `lose' might save the register and put some other value there. For
22826 example, you can't expect a global register variable to be available in
22827 the comparison-function that you pass to `qsort', since `qsort' might
22828 have put something else in that register. (If you are prepared to
22829 recompile `qsort' with the same global register variable, you can solve
22832 If you want to recompile `qsort' or other source files which do not
22833 actually use your global register variable, so that they will not use
22834 that register for any other purpose, then it suffices to specify the
22835 compiler option `-ffixed-REG'. You need not actually add a global
22836 register declaration to their source code.
22838 A function which can alter the value of a global register variable
22839 cannot safely be called from a function compiled without this variable,
22840 because it could clobber the value the caller expects to find there on
22841 return. Therefore, the function which is the entry point into the part
22842 of the program that uses the global register variable must explicitly
22843 save and restore the value which belongs to its caller.
22845 On most machines, `longjmp' will restore to each global register
22846 variable the value it had at the time of the `setjmp'. On some
22847 machines, however, `longjmp' will not change the value of global
22848 register variables. To be portable, the function that called `setjmp'
22849 should make other arrangements to save the values of the global register
22850 variables, and to restore them in a `longjmp'. This way, the same
22851 thing will happen regardless of what `longjmp' does.
22853 All global register variable declarations must precede all function
22854 definitions. If such a declaration could appear after function
22855 definitions, the declaration would be too late to prevent the register
22856 from being used for other purposes in the preceding functions.
22858 Global register variables may not have initial values, because an
22859 executable file has no means to supply initial contents for a register.
22861 On the SPARC, there are reports that g3 ... g7 are suitable registers,
22862 but certain library functions, such as `getwd', as well as the
22863 subroutines for division and remainder, modify g3 and g4. g1 and g2
22864 are local temporaries.
22866 On the 68000, a2 ... a5 should be suitable, as should d2 ... d7. Of
22867 course, it will not do to use more than a few of those.
22870 File: gcc.info, Node: Local Reg Vars, Prev: Global Reg Vars, Up: Explicit Reg Vars
22872 5.40.2 Specifying Registers for Local Variables
22873 -----------------------------------------------
22875 You can define a local register variable with a specified register like
22878 register int *foo asm ("a5");
22880 Here `a5' is the name of the register which should be used. Note that
22881 this is the same syntax used for defining global register variables,
22882 but for a local variable it would appear within a function.
22884 Naturally the register name is cpu-dependent, but this is not a
22885 problem, since specific registers are most often useful with explicit
22886 assembler instructions (*note Extended Asm::). Both of these things
22887 generally require that you conditionalize your program according to cpu
22890 In addition, operating systems on one type of cpu may differ in how
22891 they name the registers; then you would need additional conditionals.
22892 For example, some 68000 operating systems call this register `%a5'.
22894 Defining such a register variable does not reserve the register; it
22895 remains available for other uses in places where flow control determines
22896 the variable's value is not live.
22898 This option does not guarantee that GCC will generate code that has
22899 this variable in the register you specify at all times. You may not
22900 code an explicit reference to this register in the _assembler
22901 instruction template_ part of an `asm' statement and assume it will
22902 always refer to this variable. However, using the variable as an `asm'
22903 _operand_ guarantees that the specified register is used for the
22906 Stores into local register variables may be deleted when they appear
22907 to be dead according to dataflow analysis. References to local
22908 register variables may be deleted or moved or simplified.
22910 As for global register variables, it's recommended that you choose a
22911 register which is normally saved and restored by function calls on your
22912 machine, so that library routines will not clobber it. A common
22913 pitfall is to initialize multiple call-clobbered registers with
22914 arbitrary expressions, where a function call or library call for an
22915 arithmetic operator will overwrite a register value from a previous
22916 assignment, for example `r0' below:
22917 register int *p1 asm ("r0") = ...;
22918 register int *p2 asm ("r1") = ...;
22919 In those cases, a solution is to use a temporary variable for each
22920 arbitrary expression. *Note Example of asm with clobbered asm reg::.
22923 File: gcc.info, Node: Alternate Keywords, Next: Incomplete Enums, Prev: Explicit Reg Vars, Up: C Extensions
22925 5.41 Alternate Keywords
22926 =======================
22928 `-ansi' and the various `-std' options disable certain keywords. This
22929 causes trouble when you want to use GNU C extensions, or a
22930 general-purpose header file that should be usable by all programs,
22931 including ISO C programs. The keywords `asm', `typeof' and `inline'
22932 are not available in programs compiled with `-ansi' or `-std' (although
22933 `inline' can be used in a program compiled with `-std=c99'). The ISO
22934 C99 keyword `restrict' is only available when `-std=gnu99' (which will
22935 eventually be the default) or `-std=c99' (or the equivalent
22936 `-std=iso9899:1999') is used.
22938 The way to solve these problems is to put `__' at the beginning and
22939 end of each problematical keyword. For example, use `__asm__' instead
22940 of `asm', and `__inline__' instead of `inline'.
22942 Other C compilers won't accept these alternative keywords; if you want
22943 to compile with another compiler, you can define the alternate keywords
22944 as macros to replace them with the customary keywords. It looks like
22948 #define __asm__ asm
22951 `-pedantic' and other options cause warnings for many GNU C extensions.
22952 You can prevent such warnings within one expression by writing
22953 `__extension__' before the expression. `__extension__' has no effect
22957 File: gcc.info, Node: Incomplete Enums, Next: Function Names, Prev: Alternate Keywords, Up: C Extensions
22959 5.42 Incomplete `enum' Types
22960 ============================
22962 You can define an `enum' tag without specifying its possible values.
22963 This results in an incomplete type, much like what you get if you write
22964 `struct foo' without describing the elements. A later declaration
22965 which does specify the possible values completes the type.
22967 You can't allocate variables or storage using the type while it is
22968 incomplete. However, you can work with pointers to that type.
22970 This extension may not be very useful, but it makes the handling of
22971 `enum' more consistent with the way `struct' and `union' are handled.
22973 This extension is not supported by GNU C++.
22976 File: gcc.info, Node: Function Names, Next: Return Address, Prev: Incomplete Enums, Up: C Extensions
22978 5.43 Function Names as Strings
22979 ==============================
22981 GCC provides three magic variables which hold the name of the current
22982 function, as a string. The first of these is `__func__', which is part
22983 of the C99 standard:
22985 The identifier `__func__' is implicitly declared by the translator as
22986 if, immediately following the opening brace of each function
22987 definition, the declaration
22989 static const char __func__[] = "function-name";
22991 appeared, where function-name is the name of the lexically-enclosing
22992 function. This name is the unadorned name of the function.
22994 `__FUNCTION__' is another name for `__func__'. Older versions of GCC
22995 recognize only this name. However, it is not standardized. For
22996 maximum portability, we recommend you use `__func__', but provide a
22997 fallback definition with the preprocessor:
22999 #if __STDC_VERSION__ < 199901L
23001 # define __func__ __FUNCTION__
23003 # define __func__ "<unknown>"
23007 In C, `__PRETTY_FUNCTION__' is yet another name for `__func__'.
23008 However, in C++, `__PRETTY_FUNCTION__' contains the type signature of
23009 the function as well as its bare name. For example, this program:
23012 extern int printf (char *, ...);
23019 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
23020 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
23035 __PRETTY_FUNCTION__ = void a::sub(int)
23037 These identifiers are not preprocessor macros. In GCC 3.3 and
23038 earlier, in C only, `__FUNCTION__' and `__PRETTY_FUNCTION__' were
23039 treated as string literals; they could be used to initialize `char'
23040 arrays, and they could be concatenated with other string literals. GCC
23041 3.4 and later treat them as variables, like `__func__'. In C++,
23042 `__FUNCTION__' and `__PRETTY_FUNCTION__' have always been variables.
23045 File: gcc.info, Node: Return Address, Next: Vector Extensions, Prev: Function Names, Up: C Extensions
23047 5.44 Getting the Return or Frame Address of a Function
23048 ======================================================
23050 These functions may be used to get information about the callers of a
23053 -- Built-in Function: void * __builtin_return_address (unsigned int
23055 This function returns the return address of the current function,
23056 or of one of its callers. The LEVEL argument is number of frames
23057 to scan up the call stack. A value of `0' yields the return
23058 address of the current function, a value of `1' yields the return
23059 address of the caller of the current function, and so forth. When
23060 inlining the expected behavior is that the function will return
23061 the address of the function that will be returned to. To work
23062 around this behavior use the `noinline' function attribute.
23064 The LEVEL argument must be a constant integer.
23066 On some machines it may be impossible to determine the return
23067 address of any function other than the current one; in such cases,
23068 or when the top of the stack has been reached, this function will
23069 return `0' or a random value. In addition,
23070 `__builtin_frame_address' may be used to determine if the top of
23071 the stack has been reached.
23073 This function should only be used with a nonzero argument for
23074 debugging purposes.
23076 -- Built-in Function: void * __builtin_frame_address (unsigned int
23078 This function is similar to `__builtin_return_address', but it
23079 returns the address of the function frame rather than the return
23080 address of the function. Calling `__builtin_frame_address' with a
23081 value of `0' yields the frame address of the current function, a
23082 value of `1' yields the frame address of the caller of the current
23083 function, and so forth.
23085 The frame is the area on the stack which holds local variables and
23086 saved registers. The frame address is normally the address of the
23087 first word pushed on to the stack by the function. However, the
23088 exact definition depends upon the processor and the calling
23089 convention. If the processor has a dedicated frame pointer
23090 register, and the function has a frame, then
23091 `__builtin_frame_address' will return the value of the frame
23094 On some machines it may be impossible to determine the frame
23095 address of any function other than the current one; in such cases,
23096 or when the top of the stack has been reached, this function will
23097 return `0' if the first frame pointer is properly initialized by
23100 This function should only be used with a nonzero argument for
23101 debugging purposes.
23104 File: gcc.info, Node: Vector Extensions, Next: Offsetof, Prev: Return Address, Up: C Extensions
23106 5.45 Using vector instructions through built-in functions
23107 =========================================================
23109 On some targets, the instruction set contains SIMD vector instructions
23110 that operate on multiple values contained in one large register at the
23111 same time. For example, on the i386 the MMX, 3Dnow! and SSE extensions
23112 can be used this way.
23114 The first step in using these extensions is to provide the necessary
23115 data types. This should be done using an appropriate `typedef':
23117 typedef int v4si __attribute__ ((vector_size (16)));
23119 The `int' type specifies the base type, while the attribute specifies
23120 the vector size for the variable, measured in bytes. For example, the
23121 declaration above causes the compiler to set the mode for the `v4si'
23122 type to be 16 bytes wide and divided into `int' sized units. For a
23123 32-bit `int' this means a vector of 4 units of 4 bytes, and the
23124 corresponding mode of `foo' will be V4SI.
23126 The `vector_size' attribute is only applicable to integral and float
23127 scalars, although arrays, pointers, and function return values are
23128 allowed in conjunction with this construct.
23130 All the basic integer types can be used as base types, both as signed
23131 and as unsigned: `char', `short', `int', `long', `long long'. In
23132 addition, `float' and `double' can be used to build floating-point
23135 Specifying a combination that is not valid for the current architecture
23136 will cause GCC to synthesize the instructions using a narrower mode.
23137 For example, if you specify a variable of type `V4SI' and your
23138 architecture does not allow for this specific SIMD type, GCC will
23139 produce code that uses 4 `SIs'.
23141 The types defined in this manner can be used with a subset of normal C
23142 operations. Currently, GCC will allow using the following operators on
23143 these types: `+, -, *, /, unary minus, ^, |, &, ~'.
23145 The operations behave like C++ `valarrays'. Addition is defined as
23146 the addition of the corresponding elements of the operands. For
23147 example, in the code below, each of the 4 elements in A will be added
23148 to the corresponding 4 elements in B and the resulting vector will be
23151 typedef int v4si __attribute__ ((vector_size (16)));
23157 Subtraction, multiplication, division, and the logical operations
23158 operate in a similar manner. Likewise, the result of using the unary
23159 minus or complement operators on a vector type is a vector whose
23160 elements are the negative or complemented values of the corresponding
23161 elements in the operand.
23163 You can declare variables and use them in function calls and returns,
23164 as well as in assignments and some casts. You can specify a vector
23165 type as a return type for a function. Vector types can also be used as
23166 function arguments. It is possible to cast from one vector type to
23167 another, provided they are of the same size (in fact, you can also cast
23168 vectors to and from other datatypes of the same size).
23170 You cannot operate between vectors of different lengths or different
23171 signedness without a cast.
23173 A port that supports hardware vector operations, usually provides a set
23174 of built-in functions that can be used to operate on vectors. For
23175 example, a function to add two vectors and multiply the result by a
23176 third could look like this:
23178 v4si f (v4si a, v4si b, v4si c)
23180 v4si tmp = __builtin_addv4si (a, b);
23181 return __builtin_mulv4si (tmp, c);
23185 File: gcc.info, Node: Offsetof, Next: Atomic Builtins, Prev: Vector Extensions, Up: C Extensions
23190 GCC implements for both C and C++ a syntactic extension to implement
23191 the `offsetof' macro.
23194 "__builtin_offsetof" "(" `typename' "," offsetof_member_designator ")"
23196 offsetof_member_designator:
23198 | offsetof_member_designator "." `identifier'
23199 | offsetof_member_designator "[" `expr' "]"
23201 This extension is sufficient such that
23203 #define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER)
23205 is a suitable definition of the `offsetof' macro. In C++, TYPE may be
23206 dependent. In either case, MEMBER may consist of a single identifier,
23207 or a sequence of member accesses and array references.
23210 File: gcc.info, Node: Atomic Builtins, Next: Object Size Checking, Prev: Offsetof, Up: C Extensions
23212 5.47 Built-in functions for atomic memory access
23213 ================================================
23215 The following builtins are intended to be compatible with those
23216 described in the `Intel Itanium Processor-specific Application Binary
23217 Interface', section 7.4. As such, they depart from the normal GCC
23218 practice of using the "__builtin_" prefix, and further that they are
23219 overloaded such that they work on multiple types.
23221 The definition given in the Intel documentation allows only for the
23222 use of the types `int', `long', `long long' as well as their unsigned
23223 counterparts. GCC will allow any integral scalar or pointer type that
23224 is 1, 2, 4 or 8 bytes in length.
23226 Not all operations are supported by all target processors. If a
23227 particular operation cannot be implemented on the target processor, a
23228 warning will be generated and a call an external function will be
23229 generated. The external function will carry the same name as the
23230 builtin, with an additional suffix `_N' where N is the size of the data
23233 In most cases, these builtins are considered a "full barrier". That
23234 is, no memory operand will be moved across the operation, either
23235 forward or backward. Further, instructions will be issued as necessary
23236 to prevent the processor from speculating loads across the operation
23237 and from queuing stores after the operation.
23239 All of the routines are described in the Intel documentation to take
23240 "an optional list of variables protected by the memory barrier". It's
23241 not clear what is meant by that; it could mean that _only_ the
23242 following variables are protected, or it could mean that these variables
23243 should in addition be protected. At present GCC ignores this list and
23244 protects all variables which are globally accessible. If in the future
23245 we make some use of this list, an empty list will continue to mean all
23246 globally accessible variables.
23248 `TYPE __sync_fetch_and_add (TYPE *ptr, TYPE value, ...)'
23249 `TYPE __sync_fetch_and_sub (TYPE *ptr, TYPE value, ...)'
23250 `TYPE __sync_fetch_and_or (TYPE *ptr, TYPE value, ...)'
23251 `TYPE __sync_fetch_and_and (TYPE *ptr, TYPE value, ...)'
23252 `TYPE __sync_fetch_and_xor (TYPE *ptr, TYPE value, ...)'
23253 `TYPE __sync_fetch_and_nand (TYPE *ptr, TYPE value, ...)'
23254 These builtins perform the operation suggested by the name, and
23255 returns the value that had previously been in memory. That is,
23257 { tmp = *ptr; *ptr OP= value; return tmp; }
23258 { tmp = *ptr; *ptr = ~(tmp & value); return tmp; } // nand
23260 _Note:_ GCC 4.4 and later implement `__sync_fetch_and_nand'
23261 builtin as `*ptr = ~(tmp & value)' instead of `*ptr = ~tmp &
23264 `TYPE __sync_add_and_fetch (TYPE *ptr, TYPE value, ...)'
23265 `TYPE __sync_sub_and_fetch (TYPE *ptr, TYPE value, ...)'
23266 `TYPE __sync_or_and_fetch (TYPE *ptr, TYPE value, ...)'
23267 `TYPE __sync_and_and_fetch (TYPE *ptr, TYPE value, ...)'
23268 `TYPE __sync_xor_and_fetch (TYPE *ptr, TYPE value, ...)'
23269 `TYPE __sync_nand_and_fetch (TYPE *ptr, TYPE value, ...)'
23270 These builtins perform the operation suggested by the name, and
23271 return the new value. That is,
23273 { *ptr OP= value; return *ptr; }
23274 { *ptr = ~(*ptr & value); return *ptr; } // nand
23276 _Note:_ GCC 4.4 and later implement `__sync_nand_and_fetch'
23277 builtin as `*ptr = ~(*ptr & value)' instead of `*ptr = ~*ptr &
23280 `bool __sync_bool_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
23281 `TYPE __sync_val_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
23282 These builtins perform an atomic compare and swap. That is, if
23283 the current value of `*PTR' is OLDVAL, then write NEWVAL into
23286 The "bool" version returns true if the comparison is successful and
23287 NEWVAL was written. The "val" version returns the contents of
23288 `*PTR' before the operation.
23290 `__sync_synchronize (...)'
23291 This builtin issues a full memory barrier.
23293 `TYPE __sync_lock_test_and_set (TYPE *ptr, TYPE value, ...)'
23294 This builtin, as described by Intel, is not a traditional
23295 test-and-set operation, but rather an atomic exchange operation.
23296 It writes VALUE into `*PTR', and returns the previous contents of
23299 Many targets have only minimal support for such locks, and do not
23300 support a full exchange operation. In this case, a target may
23301 support reduced functionality here by which the _only_ valid value
23302 to store is the immediate constant 1. The exact value actually
23303 stored in `*PTR' is implementation defined.
23305 This builtin is not a full barrier, but rather an "acquire
23306 barrier". This means that references after the builtin cannot
23307 move to (or be speculated to) before the builtin, but previous
23308 memory stores may not be globally visible yet, and previous memory
23309 loads may not yet be satisfied.
23311 `void __sync_lock_release (TYPE *ptr, ...)'
23312 This builtin releases the lock acquired by
23313 `__sync_lock_test_and_set'. Normally this means writing the
23314 constant 0 to `*PTR'.
23316 This builtin is not a full barrier, but rather a "release barrier".
23317 This means that all previous memory stores are globally visible,
23318 and all previous memory loads have been satisfied, but following
23319 memory reads are not prevented from being speculated to before the
23323 File: gcc.info, Node: Object Size Checking, Next: Other Builtins, Prev: Atomic Builtins, Up: C Extensions
23325 5.48 Object Size Checking Builtins
23326 ==================================
23328 GCC implements a limited buffer overflow protection mechanism that can
23329 prevent some buffer overflow attacks.
23331 -- Built-in Function: size_t __builtin_object_size (void * PTR, int
23333 is a built-in construct that returns a constant number of bytes
23334 from PTR to the end of the object PTR pointer points to (if known
23335 at compile time). `__builtin_object_size' never evaluates its
23336 arguments for side-effects. If there are any side-effects in
23337 them, it returns `(size_t) -1' for TYPE 0 or 1 and `(size_t) 0'
23338 for TYPE 2 or 3. If there are multiple objects PTR can point to
23339 and all of them are known at compile time, the returned number is
23340 the maximum of remaining byte counts in those objects if TYPE & 2
23341 is 0 and minimum if nonzero. If it is not possible to determine
23342 which objects PTR points to at compile time,
23343 `__builtin_object_size' should return `(size_t) -1' for TYPE 0 or
23344 1 and `(size_t) 0' for TYPE 2 or 3.
23346 TYPE is an integer constant from 0 to 3. If the least significant
23347 bit is clear, objects are whole variables, if it is set, a closest
23348 surrounding subobject is considered the object a pointer points to.
23349 The second bit determines if maximum or minimum of remaining bytes
23352 struct V { char buf1[10]; int b; char buf2[10]; } var;
23353 char *p = &var.buf1[1], *q = &var.b;
23355 /* Here the object p points to is var. */
23356 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
23357 /* The subobject p points to is var.buf1. */
23358 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
23359 /* The object q points to is var. */
23360 assert (__builtin_object_size (q, 0)
23361 == (char *) (&var + 1) - (char *) &var.b);
23362 /* The subobject q points to is var.b. */
23363 assert (__builtin_object_size (q, 1) == sizeof (var.b));
23365 There are built-in functions added for many common string operation
23366 functions, e.g., for `memcpy' `__builtin___memcpy_chk' built-in is
23367 provided. This built-in has an additional last argument, which is the
23368 number of bytes remaining in object the DEST argument points to or
23369 `(size_t) -1' if the size is not known.
23371 The built-in functions are optimized into the normal string functions
23372 like `memcpy' if the last argument is `(size_t) -1' or if it is known
23373 at compile time that the destination object will not be overflown. If
23374 the compiler can determine at compile time the object will be always
23375 overflown, it issues a warning.
23377 The intended use can be e.g.
23380 #define bos0(dest) __builtin_object_size (dest, 0)
23381 #define memcpy(dest, src, n) \
23382 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
23386 /* It is unknown what object p points to, so this is optimized
23387 into plain memcpy - no checking is possible. */
23388 memcpy (p, "abcde", n);
23389 /* Destination is known and length too. It is known at compile
23390 time there will be no overflow. */
23391 memcpy (&buf[5], "abcde", 5);
23392 /* Destination is known, but the length is not known at compile time.
23393 This will result in __memcpy_chk call that can check for overflow
23395 memcpy (&buf[5], "abcde", n);
23396 /* Destination is known and it is known at compile time there will
23397 be overflow. There will be a warning and __memcpy_chk call that
23398 will abort the program at runtime. */
23399 memcpy (&buf[6], "abcde", 5);
23401 Such built-in functions are provided for `memcpy', `mempcpy',
23402 `memmove', `memset', `strcpy', `stpcpy', `strncpy', `strcat' and
23405 There are also checking built-in functions for formatted output
23407 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
23408 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
23409 const char *fmt, ...);
23410 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
23412 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
23413 const char *fmt, va_list ap);
23415 The added FLAG argument is passed unchanged to `__sprintf_chk' etc.
23416 functions and can contain implementation specific flags on what
23417 additional security measures the checking function might take, such as
23418 handling `%n' differently.
23420 The OS argument is the object size S points to, like in the other
23421 built-in functions. There is a small difference in the behavior
23422 though, if OS is `(size_t) -1', the built-in functions are optimized
23423 into the non-checking functions only if FLAG is 0, otherwise the
23424 checking function is called with OS argument set to `(size_t) -1'.
23426 In addition to this, there are checking built-in functions
23427 `__builtin___printf_chk', `__builtin___vprintf_chk',
23428 `__builtin___fprintf_chk' and `__builtin___vfprintf_chk'. These have
23429 just one additional argument, FLAG, right before format string FMT. If
23430 the compiler is able to optimize them to `fputc' etc. functions, it
23431 will, otherwise the checking function should be called and the FLAG
23432 argument passed to it.
23435 File: gcc.info, Node: Other Builtins, Next: Target Builtins, Prev: Object Size Checking, Up: C Extensions
23437 5.49 Other built-in functions provided by GCC
23438 =============================================
23440 GCC provides a large number of built-in functions other than the ones
23441 mentioned above. Some of these are for internal use in the processing
23442 of exceptions or variable-length argument lists and will not be
23443 documented here because they may change from time to time; we do not
23444 recommend general use of these functions.
23446 The remaining functions are provided for optimization purposes.
23448 GCC includes built-in versions of many of the functions in the standard
23449 C library. The versions prefixed with `__builtin_' will always be
23450 treated as having the same meaning as the C library function even if you
23451 specify the `-fno-builtin' option. (*note C Dialect Options::) Many of
23452 these functions are only optimized in certain cases; if they are not
23453 optimized in a particular case, a call to the library function will be
23456 Outside strict ISO C mode (`-ansi', `-std=c89' or `-std=c99'), the
23457 functions `_exit', `alloca', `bcmp', `bzero', `dcgettext', `dgettext',
23458 `dremf', `dreml', `drem', `exp10f', `exp10l', `exp10', `ffsll', `ffsl',
23459 `ffs', `fprintf_unlocked', `fputs_unlocked', `gammaf', `gammal',
23460 `gamma', `gammaf_r', `gammal_r', `gamma_r', `gettext', `index',
23461 `isascii', `j0f', `j0l', `j0', `j1f', `j1l', `j1', `jnf', `jnl', `jn',
23462 `lgammaf_r', `lgammal_r', `lgamma_r', `mempcpy', `pow10f', `pow10l',
23463 `pow10', `printf_unlocked', `rindex', `scalbf', `scalbl', `scalb',
23464 `signbit', `signbitf', `signbitl', `signbitd32', `signbitd64',
23465 `signbitd128', `significandf', `significandl', `significand', `sincosf',
23466 `sincosl', `sincos', `stpcpy', `stpncpy', `strcasecmp', `strdup',
23467 `strfmon', `strncasecmp', `strndup', `toascii', `y0f', `y0l', `y0',
23468 `y1f', `y1l', `y1', `ynf', `ynl' and `yn' may be handled as built-in
23469 functions. All these functions have corresponding versions prefixed
23470 with `__builtin_', which may be used even in strict C89 mode.
23472 The ISO C99 functions `_Exit', `acoshf', `acoshl', `acosh', `asinhf',
23473 `asinhl', `asinh', `atanhf', `atanhl', `atanh', `cabsf', `cabsl',
23474 `cabs', `cacosf', `cacoshf', `cacoshl', `cacosh', `cacosl', `cacos',
23475 `cargf', `cargl', `carg', `casinf', `casinhf', `casinhl', `casinh',
23476 `casinl', `casin', `catanf', `catanhf', `catanhl', `catanh', `catanl',
23477 `catan', `cbrtf', `cbrtl', `cbrt', `ccosf', `ccoshf', `ccoshl',
23478 `ccosh', `ccosl', `ccos', `cexpf', `cexpl', `cexp', `cimagf', `cimagl',
23479 `cimag', `clogf', `clogl', `clog', `conjf', `conjl', `conj',
23480 `copysignf', `copysignl', `copysign', `cpowf', `cpowl', `cpow',
23481 `cprojf', `cprojl', `cproj', `crealf', `creall', `creal', `csinf',
23482 `csinhf', `csinhl', `csinh', `csinl', `csin', `csqrtf', `csqrtl',
23483 `csqrt', `ctanf', `ctanhf', `ctanhl', `ctanh', `ctanl', `ctan',
23484 `erfcf', `erfcl', `erfc', `erff', `erfl', `erf', `exp2f', `exp2l',
23485 `exp2', `expm1f', `expm1l', `expm1', `fdimf', `fdiml', `fdim', `fmaf',
23486 `fmal', `fmaxf', `fmaxl', `fmax', `fma', `fminf', `fminl', `fmin',
23487 `hypotf', `hypotl', `hypot', `ilogbf', `ilogbl', `ilogb', `imaxabs',
23488 `isblank', `iswblank', `lgammaf', `lgammal', `lgamma', `llabs',
23489 `llrintf', `llrintl', `llrint', `llroundf', `llroundl', `llround',
23490 `log1pf', `log1pl', `log1p', `log2f', `log2l', `log2', `logbf',
23491 `logbl', `logb', `lrintf', `lrintl', `lrint', `lroundf', `lroundl',
23492 `lround', `nearbyintf', `nearbyintl', `nearbyint', `nextafterf',
23493 `nextafterl', `nextafter', `nexttowardf', `nexttowardl', `nexttoward',
23494 `remainderf', `remainderl', `remainder', `remquof', `remquol',
23495 `remquo', `rintf', `rintl', `rint', `roundf', `roundl', `round',
23496 `scalblnf', `scalblnl', `scalbln', `scalbnf', `scalbnl', `scalbn',
23497 `snprintf', `tgammaf', `tgammal', `tgamma', `truncf', `truncl', `trunc',
23498 `vfscanf', `vscanf', `vsnprintf' and `vsscanf' are handled as built-in
23499 functions except in strict ISO C90 mode (`-ansi' or `-std=c89').
23501 There are also built-in versions of the ISO C99 functions `acosf',
23502 `acosl', `asinf', `asinl', `atan2f', `atan2l', `atanf', `atanl',
23503 `ceilf', `ceill', `cosf', `coshf', `coshl', `cosl', `expf', `expl',
23504 `fabsf', `fabsl', `floorf', `floorl', `fmodf', `fmodl', `frexpf',
23505 `frexpl', `ldexpf', `ldexpl', `log10f', `log10l', `logf', `logl',
23506 `modfl', `modf', `powf', `powl', `sinf', `sinhf', `sinhl', `sinl',
23507 `sqrtf', `sqrtl', `tanf', `tanhf', `tanhl' and `tanl' that are
23508 recognized in any mode since ISO C90 reserves these names for the
23509 purpose to which ISO C99 puts them. All these functions have
23510 corresponding versions prefixed with `__builtin_'.
23512 The ISO C94 functions `iswalnum', `iswalpha', `iswcntrl', `iswdigit',
23513 `iswgraph', `iswlower', `iswprint', `iswpunct', `iswspace', `iswupper',
23514 `iswxdigit', `towlower' and `towupper' are handled as built-in functions
23515 except in strict ISO C90 mode (`-ansi' or `-std=c89').
23517 The ISO C90 functions `abort', `abs', `acos', `asin', `atan2', `atan',
23518 `calloc', `ceil', `cosh', `cos', `exit', `exp', `fabs', `floor', `fmod',
23519 `fprintf', `fputs', `frexp', `fscanf', `isalnum', `isalpha', `iscntrl',
23520 `isdigit', `isgraph', `islower', `isprint', `ispunct', `isspace',
23521 `isupper', `isxdigit', `tolower', `toupper', `labs', `ldexp', `log10',
23522 `log', `malloc', `memchr', `memcmp', `memcpy', `memset', `modf', `pow',
23523 `printf', `putchar', `puts', `scanf', `sinh', `sin', `snprintf',
23524 `sprintf', `sqrt', `sscanf', `strcat', `strchr', `strcmp', `strcpy',
23525 `strcspn', `strlen', `strncat', `strncmp', `strncpy', `strpbrk',
23526 `strrchr', `strspn', `strstr', `tanh', `tan', `vfprintf', `vprintf' and
23527 `vsprintf' are all recognized as built-in functions unless
23528 `-fno-builtin' is specified (or `-fno-builtin-FUNCTION' is specified
23529 for an individual function). All of these functions have corresponding
23530 versions prefixed with `__builtin_'.
23532 GCC provides built-in versions of the ISO C99 floating point comparison
23533 macros that avoid raising exceptions for unordered operands. They have
23534 the same names as the standard macros ( `isgreater', `isgreaterequal',
23535 `isless', `islessequal', `islessgreater', and `isunordered') , with
23536 `__builtin_' prefixed. We intend for a library implementor to be able
23537 to simply `#define' each standard macro to its built-in equivalent. In
23538 the same fashion, GCC provides `fpclassify', `isfinite', `isinf_sign'
23539 and `isnormal' built-ins used with `__builtin_' prefixed. The `isinf'
23540 and `isnan' builtins appear both with and without the `__builtin_'
23543 -- Built-in Function: int __builtin_types_compatible_p (TYPE1, TYPE2)
23544 You can use the built-in function `__builtin_types_compatible_p' to
23545 determine whether two types are the same.
23547 This built-in function returns 1 if the unqualified versions of the
23548 types TYPE1 and TYPE2 (which are types, not expressions) are
23549 compatible, 0 otherwise. The result of this built-in function can
23550 be used in integer constant expressions.
23552 This built-in function ignores top level qualifiers (e.g., `const',
23553 `volatile'). For example, `int' is equivalent to `const int'.
23555 The type `int[]' and `int[5]' are compatible. On the other hand,
23556 `int' and `char *' are not compatible, even if the size of their
23557 types, on the particular architecture are the same. Also, the
23558 amount of pointer indirection is taken into account when
23559 determining similarity. Consequently, `short *' is not similar to
23560 `short **'. Furthermore, two types that are typedefed are
23561 considered compatible if their underlying types are compatible.
23563 An `enum' type is not considered to be compatible with another
23564 `enum' type even if both are compatible with the same integer
23565 type; this is what the C standard specifies. For example, `enum
23566 {foo, bar}' is not similar to `enum {hot, dog}'.
23568 You would typically use this function in code whose execution
23569 varies depending on the arguments' types. For example:
23573 typeof (x) tmp = (x); \
23574 if (__builtin_types_compatible_p (typeof (x), long double)) \
23575 tmp = foo_long_double (tmp); \
23576 else if (__builtin_types_compatible_p (typeof (x), double)) \
23577 tmp = foo_double (tmp); \
23578 else if (__builtin_types_compatible_p (typeof (x), float)) \
23579 tmp = foo_float (tmp); \
23585 _Note:_ This construct is only available for C.
23588 -- Built-in Function: TYPE __builtin_choose_expr (CONST_EXP, EXP1,
23590 You can use the built-in function `__builtin_choose_expr' to
23591 evaluate code depending on the value of a constant expression.
23592 This built-in function returns EXP1 if CONST_EXP, which is a
23593 constant expression that must be able to be determined at compile
23594 time, is nonzero. Otherwise it returns 0.
23596 This built-in function is analogous to the `? :' operator in C,
23597 except that the expression returned has its type unaltered by
23598 promotion rules. Also, the built-in function does not evaluate
23599 the expression that was not chosen. For example, if CONST_EXP
23600 evaluates to true, EXP2 is not evaluated even if it has
23603 This built-in function can return an lvalue if the chosen argument
23606 If EXP1 is returned, the return type is the same as EXP1's type.
23607 Similarly, if EXP2 is returned, its return type is the same as
23613 __builtin_choose_expr ( \
23614 __builtin_types_compatible_p (typeof (x), double), \
23616 __builtin_choose_expr ( \
23617 __builtin_types_compatible_p (typeof (x), float), \
23619 /* The void expression results in a compile-time error \
23620 when assigning the result to something. */ \
23623 _Note:_ This construct is only available for C. Furthermore, the
23624 unused expression (EXP1 or EXP2 depending on the value of
23625 CONST_EXP) may still generate syntax errors. This may change in
23629 -- Built-in Function: int __builtin_constant_p (EXP)
23630 You can use the built-in function `__builtin_constant_p' to
23631 determine if a value is known to be constant at compile-time and
23632 hence that GCC can perform constant-folding on expressions
23633 involving that value. The argument of the function is the value
23634 to test. The function returns the integer 1 if the argument is
23635 known to be a compile-time constant and 0 if it is not known to be
23636 a compile-time constant. A return of 0 does not indicate that the
23637 value is _not_ a constant, but merely that GCC cannot prove it is
23638 a constant with the specified value of the `-O' option.
23640 You would typically use this function in an embedded application
23641 where memory was a critical resource. If you have some complex
23642 calculation, you may want it to be folded if it involves
23643 constants, but need to call a function if it does not. For
23646 #define Scale_Value(X) \
23647 (__builtin_constant_p (X) \
23648 ? ((X) * SCALE + OFFSET) : Scale (X))
23650 You may use this built-in function in either a macro or an inline
23651 function. However, if you use it in an inlined function and pass
23652 an argument of the function as the argument to the built-in, GCC
23653 will never return 1 when you call the inline function with a
23654 string constant or compound literal (*note Compound Literals::)
23655 and will not return 1 when you pass a constant numeric value to
23656 the inline function unless you specify the `-O' option.
23658 You may also use `__builtin_constant_p' in initializers for static
23659 data. For instance, you can write
23661 static const int table[] = {
23662 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
23666 This is an acceptable initializer even if EXPRESSION is not a
23667 constant expression. GCC must be more conservative about
23668 evaluating the built-in in this case, because it has no
23669 opportunity to perform optimization.
23671 Previous versions of GCC did not accept this built-in in data
23672 initializers. The earliest version where it is completely safe is
23675 -- Built-in Function: long __builtin_expect (long EXP, long C)
23676 You may use `__builtin_expect' to provide the compiler with branch
23677 prediction information. In general, you should prefer to use
23678 actual profile feedback for this (`-fprofile-arcs'), as
23679 programmers are notoriously bad at predicting how their programs
23680 actually perform. However, there are applications in which this
23681 data is hard to collect.
23683 The return value is the value of EXP, which should be an integral
23684 expression. The semantics of the built-in are that it is expected
23685 that EXP == C. For example:
23687 if (__builtin_expect (x, 0))
23690 would indicate that we do not expect to call `foo', since we
23691 expect `x' to be zero. Since you are limited to integral
23692 expressions for EXP, you should use constructions such as
23694 if (__builtin_expect (ptr != NULL, 1))
23697 when testing pointer or floating-point values.
23699 -- Built-in Function: void __builtin_trap (void)
23700 This function causes the program to exit abnormally. GCC
23701 implements this function by using a target-dependent mechanism
23702 (such as intentionally executing an illegal instruction) or by
23703 calling `abort'. The mechanism used may vary from release to
23704 release so you should not rely on any particular implementation.
23706 -- Built-in Function: void __builtin___clear_cache (char *BEGIN, char
23708 This function is used to flush the processor's instruction cache
23709 for the region of memory between BEGIN inclusive and END
23710 exclusive. Some targets require that the instruction cache be
23711 flushed, after modifying memory containing code, in order to obtain
23712 deterministic behavior.
23714 If the target does not require instruction cache flushes,
23715 `__builtin___clear_cache' has no effect. Otherwise either
23716 instructions are emitted in-line to clear the instruction cache or
23717 a call to the `__clear_cache' function in libgcc is made.
23719 -- Built-in Function: void __builtin_prefetch (const void *ADDR, ...)
23720 This function is used to minimize cache-miss latency by moving
23721 data into a cache before it is accessed. You can insert calls to
23722 `__builtin_prefetch' into code for which you know addresses of
23723 data in memory that is likely to be accessed soon. If the target
23724 supports them, data prefetch instructions will be generated. If
23725 the prefetch is done early enough before the access then the data
23726 will be in the cache by the time it is accessed.
23728 The value of ADDR is the address of the memory to prefetch. There
23729 are two optional arguments, RW and LOCALITY. The value of RW is a
23730 compile-time constant one or zero; one means that the prefetch is
23731 preparing for a write to the memory address and zero, the default,
23732 means that the prefetch is preparing for a read. The value
23733 LOCALITY must be a compile-time constant integer between zero and
23734 three. A value of zero means that the data has no temporal
23735 locality, so it need not be left in the cache after the access. A
23736 value of three means that the data has a high degree of temporal
23737 locality and should be left in all levels of cache possible.
23738 Values of one and two mean, respectively, a low or moderate degree
23739 of temporal locality. The default is three.
23741 for (i = 0; i < n; i++)
23743 a[i] = a[i] + b[i];
23744 __builtin_prefetch (&a[i+j], 1, 1);
23745 __builtin_prefetch (&b[i+j], 0, 1);
23749 Data prefetch does not generate faults if ADDR is invalid, but the
23750 address expression itself must be valid. For example, a prefetch
23751 of `p->next' will not fault if `p->next' is not a valid address,
23752 but evaluation will fault if `p' is not a valid address.
23754 If the target does not support data prefetch, the address
23755 expression is evaluated if it includes side effects but no other
23756 code is generated and GCC does not issue a warning.
23758 -- Built-in Function: double __builtin_huge_val (void)
23759 Returns a positive infinity, if supported by the floating-point
23760 format, else `DBL_MAX'. This function is suitable for
23761 implementing the ISO C macro `HUGE_VAL'.
23763 -- Built-in Function: float __builtin_huge_valf (void)
23764 Similar to `__builtin_huge_val', except the return type is `float'.
23766 -- Built-in Function: long double __builtin_huge_vall (void)
23767 Similar to `__builtin_huge_val', except the return type is `long
23770 -- Built-in Function: int __builtin_fpclassify (int, int, int, int,
23772 This built-in implements the C99 fpclassify functionality. The
23773 first five int arguments should be the target library's notion of
23774 the possible FP classes and are used for return values. They must
23775 be constant values and they must appear in this order: `FP_NAN',
23776 `FP_INFINITE', `FP_NORMAL', `FP_SUBNORMAL' and `FP_ZERO'. The
23777 ellipsis is for exactly one floating point value to classify. GCC
23778 treats the last argument as type-generic, which means it does not
23779 do default promotion from float to double.
23781 -- Built-in Function: double __builtin_inf (void)
23782 Similar to `__builtin_huge_val', except a warning is generated if
23783 the target floating-point format does not support infinities.
23785 -- Built-in Function: _Decimal32 __builtin_infd32 (void)
23786 Similar to `__builtin_inf', except the return type is `_Decimal32'.
23788 -- Built-in Function: _Decimal64 __builtin_infd64 (void)
23789 Similar to `__builtin_inf', except the return type is `_Decimal64'.
23791 -- Built-in Function: _Decimal128 __builtin_infd128 (void)
23792 Similar to `__builtin_inf', except the return type is
23795 -- Built-in Function: float __builtin_inff (void)
23796 Similar to `__builtin_inf', except the return type is `float'.
23797 This function is suitable for implementing the ISO C99 macro
23800 -- Built-in Function: long double __builtin_infl (void)
23801 Similar to `__builtin_inf', except the return type is `long
23804 -- Built-in Function: int __builtin_isinf_sign (...)
23805 Similar to `isinf', except the return value will be negative for
23806 an argument of `-Inf'. Note while the parameter list is an
23807 ellipsis, this function only accepts exactly one floating point
23808 argument. GCC treats this parameter as type-generic, which means
23809 it does not do default promotion from float to double.
23811 -- Built-in Function: double __builtin_nan (const char *str)
23812 This is an implementation of the ISO C99 function `nan'.
23814 Since ISO C99 defines this function in terms of `strtod', which we
23815 do not implement, a description of the parsing is in order. The
23816 string is parsed as by `strtol'; that is, the base is recognized by
23817 leading `0' or `0x' prefixes. The number parsed is placed in the
23818 significand such that the least significant bit of the number is
23819 at the least significant bit of the significand. The number is
23820 truncated to fit the significand field provided. The significand
23821 is forced to be a quiet NaN.
23823 This function, if given a string literal all of which would have
23824 been consumed by strtol, is evaluated early enough that it is
23825 considered a compile-time constant.
23827 -- Built-in Function: _Decimal32 __builtin_nand32 (const char *str)
23828 Similar to `__builtin_nan', except the return type is `_Decimal32'.
23830 -- Built-in Function: _Decimal64 __builtin_nand64 (const char *str)
23831 Similar to `__builtin_nan', except the return type is `_Decimal64'.
23833 -- Built-in Function: _Decimal128 __builtin_nand128 (const char *str)
23834 Similar to `__builtin_nan', except the return type is
23837 -- Built-in Function: float __builtin_nanf (const char *str)
23838 Similar to `__builtin_nan', except the return type is `float'.
23840 -- Built-in Function: long double __builtin_nanl (const char *str)
23841 Similar to `__builtin_nan', except the return type is `long
23844 -- Built-in Function: double __builtin_nans (const char *str)
23845 Similar to `__builtin_nan', except the significand is forced to be
23846 a signaling NaN. The `nans' function is proposed by WG14 N965.
23848 -- Built-in Function: float __builtin_nansf (const char *str)
23849 Similar to `__builtin_nans', except the return type is `float'.
23851 -- Built-in Function: long double __builtin_nansl (const char *str)
23852 Similar to `__builtin_nans', except the return type is `long
23855 -- Built-in Function: int __builtin_ffs (unsigned int x)
23856 Returns one plus the index of the least significant 1-bit of X, or
23857 if X is zero, returns zero.
23859 -- Built-in Function: int __builtin_clz (unsigned int x)
23860 Returns the number of leading 0-bits in X, starting at the most
23861 significant bit position. If X is 0, the result is undefined.
23863 -- Built-in Function: int __builtin_ctz (unsigned int x)
23864 Returns the number of trailing 0-bits in X, starting at the least
23865 significant bit position. If X is 0, the result is undefined.
23867 -- Built-in Function: int __builtin_popcount (unsigned int x)
23868 Returns the number of 1-bits in X.
23870 -- Built-in Function: int __builtin_parity (unsigned int x)
23871 Returns the parity of X, i.e. the number of 1-bits in X modulo 2.
23873 -- Built-in Function: int __builtin_ffsl (unsigned long)
23874 Similar to `__builtin_ffs', except the argument type is `unsigned
23877 -- Built-in Function: int __builtin_clzl (unsigned long)
23878 Similar to `__builtin_clz', except the argument type is `unsigned
23881 -- Built-in Function: int __builtin_ctzl (unsigned long)
23882 Similar to `__builtin_ctz', except the argument type is `unsigned
23885 -- Built-in Function: int __builtin_popcountl (unsigned long)
23886 Similar to `__builtin_popcount', except the argument type is
23889 -- Built-in Function: int __builtin_parityl (unsigned long)
23890 Similar to `__builtin_parity', except the argument type is
23893 -- Built-in Function: int __builtin_ffsll (unsigned long long)
23894 Similar to `__builtin_ffs', except the argument type is `unsigned
23897 -- Built-in Function: int __builtin_clzll (unsigned long long)
23898 Similar to `__builtin_clz', except the argument type is `unsigned
23901 -- Built-in Function: int __builtin_ctzll (unsigned long long)
23902 Similar to `__builtin_ctz', except the argument type is `unsigned
23905 -- Built-in Function: int __builtin_popcountll (unsigned long long)
23906 Similar to `__builtin_popcount', except the argument type is
23907 `unsigned long long'.
23909 -- Built-in Function: int __builtin_parityll (unsigned long long)
23910 Similar to `__builtin_parity', except the argument type is
23911 `unsigned long long'.
23913 -- Built-in Function: double __builtin_powi (double, int)
23914 Returns the first argument raised to the power of the second.
23915 Unlike the `pow' function no guarantees about precision and
23918 -- Built-in Function: float __builtin_powif (float, int)
23919 Similar to `__builtin_powi', except the argument and return types
23922 -- Built-in Function: long double __builtin_powil (long double, int)
23923 Similar to `__builtin_powi', except the argument and return types
23926 -- Built-in Function: int32_t __builtin_bswap32 (int32_t x)
23927 Returns X with the order of the bytes reversed; for example,
23928 `0xaabbccdd' becomes `0xddccbbaa'. Byte here always means exactly
23931 -- Built-in Function: int64_t __builtin_bswap64 (int64_t x)
23932 Similar to `__builtin_bswap32', except the argument and return
23936 File: gcc.info, Node: Target Builtins, Next: Target Format Checks, Prev: Other Builtins, Up: C Extensions
23938 5.50 Built-in Functions Specific to Particular Target Machines
23939 ==============================================================
23941 On some target machines, GCC supports many built-in functions specific
23942 to those machines. Generally these generate calls to specific machine
23943 instructions, but allow the compiler to schedule those calls.
23947 * Alpha Built-in Functions::
23948 * ARM iWMMXt Built-in Functions::
23949 * ARM NEON Intrinsics::
23950 * Blackfin Built-in Functions::
23951 * FR-V Built-in Functions::
23952 * X86 Built-in Functions::
23953 * MIPS DSP Built-in Functions::
23954 * MIPS Paired-Single Support::
23955 * MIPS Loongson Built-in Functions::
23956 * Other MIPS Built-in Functions::
23957 * picoChip Built-in Functions::
23958 * PowerPC AltiVec Built-in Functions::
23959 * SPARC VIS Built-in Functions::
23960 * SPU Built-in Functions::
23963 File: gcc.info, Node: Alpha Built-in Functions, Next: ARM iWMMXt Built-in Functions, Up: Target Builtins
23965 5.50.1 Alpha Built-in Functions
23966 -------------------------------
23968 These built-in functions are available for the Alpha family of
23969 processors, depending on the command-line switches used.
23971 The following built-in functions are always available. They all
23972 generate the machine instruction that is part of the name.
23974 long __builtin_alpha_implver (void)
23975 long __builtin_alpha_rpcc (void)
23976 long __builtin_alpha_amask (long)
23977 long __builtin_alpha_cmpbge (long, long)
23978 long __builtin_alpha_extbl (long, long)
23979 long __builtin_alpha_extwl (long, long)
23980 long __builtin_alpha_extll (long, long)
23981 long __builtin_alpha_extql (long, long)
23982 long __builtin_alpha_extwh (long, long)
23983 long __builtin_alpha_extlh (long, long)
23984 long __builtin_alpha_extqh (long, long)
23985 long __builtin_alpha_insbl (long, long)
23986 long __builtin_alpha_inswl (long, long)
23987 long __builtin_alpha_insll (long, long)
23988 long __builtin_alpha_insql (long, long)
23989 long __builtin_alpha_inswh (long, long)
23990 long __builtin_alpha_inslh (long, long)
23991 long __builtin_alpha_insqh (long, long)
23992 long __builtin_alpha_mskbl (long, long)
23993 long __builtin_alpha_mskwl (long, long)
23994 long __builtin_alpha_mskll (long, long)
23995 long __builtin_alpha_mskql (long, long)
23996 long __builtin_alpha_mskwh (long, long)
23997 long __builtin_alpha_msklh (long, long)
23998 long __builtin_alpha_mskqh (long, long)
23999 long __builtin_alpha_umulh (long, long)
24000 long __builtin_alpha_zap (long, long)
24001 long __builtin_alpha_zapnot (long, long)
24003 The following built-in functions are always with `-mmax' or
24004 `-mcpu=CPU' where CPU is `pca56' or later. They all generate the
24005 machine instruction that is part of the name.
24007 long __builtin_alpha_pklb (long)
24008 long __builtin_alpha_pkwb (long)
24009 long __builtin_alpha_unpkbl (long)
24010 long __builtin_alpha_unpkbw (long)
24011 long __builtin_alpha_minub8 (long, long)
24012 long __builtin_alpha_minsb8 (long, long)
24013 long __builtin_alpha_minuw4 (long, long)
24014 long __builtin_alpha_minsw4 (long, long)
24015 long __builtin_alpha_maxub8 (long, long)
24016 long __builtin_alpha_maxsb8 (long, long)
24017 long __builtin_alpha_maxuw4 (long, long)
24018 long __builtin_alpha_maxsw4 (long, long)
24019 long __builtin_alpha_perr (long, long)
24021 The following built-in functions are always with `-mcix' or
24022 `-mcpu=CPU' where CPU is `ev67' or later. They all generate the
24023 machine instruction that is part of the name.
24025 long __builtin_alpha_cttz (long)
24026 long __builtin_alpha_ctlz (long)
24027 long __builtin_alpha_ctpop (long)
24029 The following builtins are available on systems that use the OSF/1
24030 PALcode. Normally they invoke the `rduniq' and `wruniq' PAL calls, but
24031 when invoked with `-mtls-kernel', they invoke `rdval' and `wrval'.
24033 void *__builtin_thread_pointer (void)
24034 void __builtin_set_thread_pointer (void *)
24037 File: gcc.info, Node: ARM iWMMXt Built-in Functions, Next: ARM NEON Intrinsics, Prev: Alpha Built-in Functions, Up: Target Builtins
24039 5.50.2 ARM iWMMXt Built-in Functions
24040 ------------------------------------
24042 These built-in functions are available for the ARM family of processors
24043 when the `-mcpu=iwmmxt' switch is used:
24045 typedef int v2si __attribute__ ((vector_size (8)));
24046 typedef short v4hi __attribute__ ((vector_size (8)));
24047 typedef char v8qi __attribute__ ((vector_size (8)));
24049 int __builtin_arm_getwcx (int)
24050 void __builtin_arm_setwcx (int, int)
24051 int __builtin_arm_textrmsb (v8qi, int)
24052 int __builtin_arm_textrmsh (v4hi, int)
24053 int __builtin_arm_textrmsw (v2si, int)
24054 int __builtin_arm_textrmub (v8qi, int)
24055 int __builtin_arm_textrmuh (v4hi, int)
24056 int __builtin_arm_textrmuw (v2si, int)
24057 v8qi __builtin_arm_tinsrb (v8qi, int)
24058 v4hi __builtin_arm_tinsrh (v4hi, int)
24059 v2si __builtin_arm_tinsrw (v2si, int)
24060 long long __builtin_arm_tmia (long long, int, int)
24061 long long __builtin_arm_tmiabb (long long, int, int)
24062 long long __builtin_arm_tmiabt (long long, int, int)
24063 long long __builtin_arm_tmiaph (long long, int, int)
24064 long long __builtin_arm_tmiatb (long long, int, int)
24065 long long __builtin_arm_tmiatt (long long, int, int)
24066 int __builtin_arm_tmovmskb (v8qi)
24067 int __builtin_arm_tmovmskh (v4hi)
24068 int __builtin_arm_tmovmskw (v2si)
24069 long long __builtin_arm_waccb (v8qi)
24070 long long __builtin_arm_wacch (v4hi)
24071 long long __builtin_arm_waccw (v2si)
24072 v8qi __builtin_arm_waddb (v8qi, v8qi)
24073 v8qi __builtin_arm_waddbss (v8qi, v8qi)
24074 v8qi __builtin_arm_waddbus (v8qi, v8qi)
24075 v4hi __builtin_arm_waddh (v4hi, v4hi)
24076 v4hi __builtin_arm_waddhss (v4hi, v4hi)
24077 v4hi __builtin_arm_waddhus (v4hi, v4hi)
24078 v2si __builtin_arm_waddw (v2si, v2si)
24079 v2si __builtin_arm_waddwss (v2si, v2si)
24080 v2si __builtin_arm_waddwus (v2si, v2si)
24081 v8qi __builtin_arm_walign (v8qi, v8qi, int)
24082 long long __builtin_arm_wand(long long, long long)
24083 long long __builtin_arm_wandn (long long, long long)
24084 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
24085 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
24086 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
24087 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
24088 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
24089 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
24090 v2si __builtin_arm_wcmpeqw (v2si, v2si)
24091 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
24092 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
24093 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
24094 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
24095 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
24096 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
24097 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
24098 long long __builtin_arm_wmacsz (v4hi, v4hi)
24099 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
24100 long long __builtin_arm_wmacuz (v4hi, v4hi)
24101 v4hi __builtin_arm_wmadds (v4hi, v4hi)
24102 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
24103 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
24104 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
24105 v2si __builtin_arm_wmaxsw (v2si, v2si)
24106 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
24107 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
24108 v2si __builtin_arm_wmaxuw (v2si, v2si)
24109 v8qi __builtin_arm_wminsb (v8qi, v8qi)
24110 v4hi __builtin_arm_wminsh (v4hi, v4hi)
24111 v2si __builtin_arm_wminsw (v2si, v2si)
24112 v8qi __builtin_arm_wminub (v8qi, v8qi)
24113 v4hi __builtin_arm_wminuh (v4hi, v4hi)
24114 v2si __builtin_arm_wminuw (v2si, v2si)
24115 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
24116 v4hi __builtin_arm_wmulul (v4hi, v4hi)
24117 v4hi __builtin_arm_wmulum (v4hi, v4hi)
24118 long long __builtin_arm_wor (long long, long long)
24119 v2si __builtin_arm_wpackdss (long long, long long)
24120 v2si __builtin_arm_wpackdus (long long, long long)
24121 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
24122 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
24123 v4hi __builtin_arm_wpackwss (v2si, v2si)
24124 v4hi __builtin_arm_wpackwus (v2si, v2si)
24125 long long __builtin_arm_wrord (long long, long long)
24126 long long __builtin_arm_wrordi (long long, int)
24127 v4hi __builtin_arm_wrorh (v4hi, long long)
24128 v4hi __builtin_arm_wrorhi (v4hi, int)
24129 v2si __builtin_arm_wrorw (v2si, long long)
24130 v2si __builtin_arm_wrorwi (v2si, int)
24131 v2si __builtin_arm_wsadb (v8qi, v8qi)
24132 v2si __builtin_arm_wsadbz (v8qi, v8qi)
24133 v2si __builtin_arm_wsadh (v4hi, v4hi)
24134 v2si __builtin_arm_wsadhz (v4hi, v4hi)
24135 v4hi __builtin_arm_wshufh (v4hi, int)
24136 long long __builtin_arm_wslld (long long, long long)
24137 long long __builtin_arm_wslldi (long long, int)
24138 v4hi __builtin_arm_wsllh (v4hi, long long)
24139 v4hi __builtin_arm_wsllhi (v4hi, int)
24140 v2si __builtin_arm_wsllw (v2si, long long)
24141 v2si __builtin_arm_wsllwi (v2si, int)
24142 long long __builtin_arm_wsrad (long long, long long)
24143 long long __builtin_arm_wsradi (long long, int)
24144 v4hi __builtin_arm_wsrah (v4hi, long long)
24145 v4hi __builtin_arm_wsrahi (v4hi, int)
24146 v2si __builtin_arm_wsraw (v2si, long long)
24147 v2si __builtin_arm_wsrawi (v2si, int)
24148 long long __builtin_arm_wsrld (long long, long long)
24149 long long __builtin_arm_wsrldi (long long, int)
24150 v4hi __builtin_arm_wsrlh (v4hi, long long)
24151 v4hi __builtin_arm_wsrlhi (v4hi, int)
24152 v2si __builtin_arm_wsrlw (v2si, long long)
24153 v2si __builtin_arm_wsrlwi (v2si, int)
24154 v8qi __builtin_arm_wsubb (v8qi, v8qi)
24155 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
24156 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
24157 v4hi __builtin_arm_wsubh (v4hi, v4hi)
24158 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
24159 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
24160 v2si __builtin_arm_wsubw (v2si, v2si)
24161 v2si __builtin_arm_wsubwss (v2si, v2si)
24162 v2si __builtin_arm_wsubwus (v2si, v2si)
24163 v4hi __builtin_arm_wunpckehsb (v8qi)
24164 v2si __builtin_arm_wunpckehsh (v4hi)
24165 long long __builtin_arm_wunpckehsw (v2si)
24166 v4hi __builtin_arm_wunpckehub (v8qi)
24167 v2si __builtin_arm_wunpckehuh (v4hi)
24168 long long __builtin_arm_wunpckehuw (v2si)
24169 v4hi __builtin_arm_wunpckelsb (v8qi)
24170 v2si __builtin_arm_wunpckelsh (v4hi)
24171 long long __builtin_arm_wunpckelsw (v2si)
24172 v4hi __builtin_arm_wunpckelub (v8qi)
24173 v2si __builtin_arm_wunpckeluh (v4hi)
24174 long long __builtin_arm_wunpckeluw (v2si)
24175 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
24176 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
24177 v2si __builtin_arm_wunpckihw (v2si, v2si)
24178 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
24179 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
24180 v2si __builtin_arm_wunpckilw (v2si, v2si)
24181 long long __builtin_arm_wxor (long long, long long)
24182 long long __builtin_arm_wzero ()
24185 File: gcc.info, Node: ARM NEON Intrinsics, Next: Blackfin Built-in Functions, Prev: ARM iWMMXt Built-in Functions, Up: Target Builtins
24187 5.50.3 ARM NEON Intrinsics
24188 --------------------------
24190 These built-in intrinsics for the ARM Advanced SIMD extension are
24191 available when the `-mfpu=neon' switch is used:
24196 * uint32x2_t vadd_u32 (uint32x2_t, uint32x2_t)
24197 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
24199 * uint16x4_t vadd_u16 (uint16x4_t, uint16x4_t)
24200 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
24202 * uint8x8_t vadd_u8 (uint8x8_t, uint8x8_t)
24203 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
24205 * int32x2_t vadd_s32 (int32x2_t, int32x2_t)
24206 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
24208 * int16x4_t vadd_s16 (int16x4_t, int16x4_t)
24209 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
24211 * int8x8_t vadd_s8 (int8x8_t, int8x8_t)
24212 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
24214 * uint64x1_t vadd_u64 (uint64x1_t, uint64x1_t)
24215 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
24217 * int64x1_t vadd_s64 (int64x1_t, int64x1_t)
24218 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
24220 * float32x2_t vadd_f32 (float32x2_t, float32x2_t)
24221 _Form of expected instruction(s):_ `vadd.f32 D0, D0, D0'
24223 * uint32x4_t vaddq_u32 (uint32x4_t, uint32x4_t)
24224 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
24226 * uint16x8_t vaddq_u16 (uint16x8_t, uint16x8_t)
24227 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
24229 * uint8x16_t vaddq_u8 (uint8x16_t, uint8x16_t)
24230 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
24232 * int32x4_t vaddq_s32 (int32x4_t, int32x4_t)
24233 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
24235 * int16x8_t vaddq_s16 (int16x8_t, int16x8_t)
24236 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
24238 * int8x16_t vaddq_s8 (int8x16_t, int8x16_t)
24239 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
24241 * uint64x2_t vaddq_u64 (uint64x2_t, uint64x2_t)
24242 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
24244 * int64x2_t vaddq_s64 (int64x2_t, int64x2_t)
24245 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
24247 * float32x4_t vaddq_f32 (float32x4_t, float32x4_t)
24248 _Form of expected instruction(s):_ `vadd.f32 Q0, Q0, Q0'
24250 * uint64x2_t vaddl_u32 (uint32x2_t, uint32x2_t)
24251 _Form of expected instruction(s):_ `vaddl.u32 Q0, D0, D0'
24253 * uint32x4_t vaddl_u16 (uint16x4_t, uint16x4_t)
24254 _Form of expected instruction(s):_ `vaddl.u16 Q0, D0, D0'
24256 * uint16x8_t vaddl_u8 (uint8x8_t, uint8x8_t)
24257 _Form of expected instruction(s):_ `vaddl.u8 Q0, D0, D0'
24259 * int64x2_t vaddl_s32 (int32x2_t, int32x2_t)
24260 _Form of expected instruction(s):_ `vaddl.s32 Q0, D0, D0'
24262 * int32x4_t vaddl_s16 (int16x4_t, int16x4_t)
24263 _Form of expected instruction(s):_ `vaddl.s16 Q0, D0, D0'
24265 * int16x8_t vaddl_s8 (int8x8_t, int8x8_t)
24266 _Form of expected instruction(s):_ `vaddl.s8 Q0, D0, D0'
24268 * uint64x2_t vaddw_u32 (uint64x2_t, uint32x2_t)
24269 _Form of expected instruction(s):_ `vaddw.u32 Q0, Q0, D0'
24271 * uint32x4_t vaddw_u16 (uint32x4_t, uint16x4_t)
24272 _Form of expected instruction(s):_ `vaddw.u16 Q0, Q0, D0'
24274 * uint16x8_t vaddw_u8 (uint16x8_t, uint8x8_t)
24275 _Form of expected instruction(s):_ `vaddw.u8 Q0, Q0, D0'
24277 * int64x2_t vaddw_s32 (int64x2_t, int32x2_t)
24278 _Form of expected instruction(s):_ `vaddw.s32 Q0, Q0, D0'
24280 * int32x4_t vaddw_s16 (int32x4_t, int16x4_t)
24281 _Form of expected instruction(s):_ `vaddw.s16 Q0, Q0, D0'
24283 * int16x8_t vaddw_s8 (int16x8_t, int8x8_t)
24284 _Form of expected instruction(s):_ `vaddw.s8 Q0, Q0, D0'
24286 * uint32x2_t vhadd_u32 (uint32x2_t, uint32x2_t)
24287 _Form of expected instruction(s):_ `vhadd.u32 D0, D0, D0'
24289 * uint16x4_t vhadd_u16 (uint16x4_t, uint16x4_t)
24290 _Form of expected instruction(s):_ `vhadd.u16 D0, D0, D0'
24292 * uint8x8_t vhadd_u8 (uint8x8_t, uint8x8_t)
24293 _Form of expected instruction(s):_ `vhadd.u8 D0, D0, D0'
24295 * int32x2_t vhadd_s32 (int32x2_t, int32x2_t)
24296 _Form of expected instruction(s):_ `vhadd.s32 D0, D0, D0'
24298 * int16x4_t vhadd_s16 (int16x4_t, int16x4_t)
24299 _Form of expected instruction(s):_ `vhadd.s16 D0, D0, D0'
24301 * int8x8_t vhadd_s8 (int8x8_t, int8x8_t)
24302 _Form of expected instruction(s):_ `vhadd.s8 D0, D0, D0'
24304 * uint32x4_t vhaddq_u32 (uint32x4_t, uint32x4_t)
24305 _Form of expected instruction(s):_ `vhadd.u32 Q0, Q0, Q0'
24307 * uint16x8_t vhaddq_u16 (uint16x8_t, uint16x8_t)
24308 _Form of expected instruction(s):_ `vhadd.u16 Q0, Q0, Q0'
24310 * uint8x16_t vhaddq_u8 (uint8x16_t, uint8x16_t)
24311 _Form of expected instruction(s):_ `vhadd.u8 Q0, Q0, Q0'
24313 * int32x4_t vhaddq_s32 (int32x4_t, int32x4_t)
24314 _Form of expected instruction(s):_ `vhadd.s32 Q0, Q0, Q0'
24316 * int16x8_t vhaddq_s16 (int16x8_t, int16x8_t)
24317 _Form of expected instruction(s):_ `vhadd.s16 Q0, Q0, Q0'
24319 * int8x16_t vhaddq_s8 (int8x16_t, int8x16_t)
24320 _Form of expected instruction(s):_ `vhadd.s8 Q0, Q0, Q0'
24322 * uint32x2_t vrhadd_u32 (uint32x2_t, uint32x2_t)
24323 _Form of expected instruction(s):_ `vrhadd.u32 D0, D0, D0'
24325 * uint16x4_t vrhadd_u16 (uint16x4_t, uint16x4_t)
24326 _Form of expected instruction(s):_ `vrhadd.u16 D0, D0, D0'
24328 * uint8x8_t vrhadd_u8 (uint8x8_t, uint8x8_t)
24329 _Form of expected instruction(s):_ `vrhadd.u8 D0, D0, D0'
24331 * int32x2_t vrhadd_s32 (int32x2_t, int32x2_t)
24332 _Form of expected instruction(s):_ `vrhadd.s32 D0, D0, D0'
24334 * int16x4_t vrhadd_s16 (int16x4_t, int16x4_t)
24335 _Form of expected instruction(s):_ `vrhadd.s16 D0, D0, D0'
24337 * int8x8_t vrhadd_s8 (int8x8_t, int8x8_t)
24338 _Form of expected instruction(s):_ `vrhadd.s8 D0, D0, D0'
24340 * uint32x4_t vrhaddq_u32 (uint32x4_t, uint32x4_t)
24341 _Form of expected instruction(s):_ `vrhadd.u32 Q0, Q0, Q0'
24343 * uint16x8_t vrhaddq_u16 (uint16x8_t, uint16x8_t)
24344 _Form of expected instruction(s):_ `vrhadd.u16 Q0, Q0, Q0'
24346 * uint8x16_t vrhaddq_u8 (uint8x16_t, uint8x16_t)
24347 _Form of expected instruction(s):_ `vrhadd.u8 Q0, Q0, Q0'
24349 * int32x4_t vrhaddq_s32 (int32x4_t, int32x4_t)
24350 _Form of expected instruction(s):_ `vrhadd.s32 Q0, Q0, Q0'
24352 * int16x8_t vrhaddq_s16 (int16x8_t, int16x8_t)
24353 _Form of expected instruction(s):_ `vrhadd.s16 Q0, Q0, Q0'
24355 * int8x16_t vrhaddq_s8 (int8x16_t, int8x16_t)
24356 _Form of expected instruction(s):_ `vrhadd.s8 Q0, Q0, Q0'
24358 * uint32x2_t vqadd_u32 (uint32x2_t, uint32x2_t)
24359 _Form of expected instruction(s):_ `vqadd.u32 D0, D0, D0'
24361 * uint16x4_t vqadd_u16 (uint16x4_t, uint16x4_t)
24362 _Form of expected instruction(s):_ `vqadd.u16 D0, D0, D0'
24364 * uint8x8_t vqadd_u8 (uint8x8_t, uint8x8_t)
24365 _Form of expected instruction(s):_ `vqadd.u8 D0, D0, D0'
24367 * int32x2_t vqadd_s32 (int32x2_t, int32x2_t)
24368 _Form of expected instruction(s):_ `vqadd.s32 D0, D0, D0'
24370 * int16x4_t vqadd_s16 (int16x4_t, int16x4_t)
24371 _Form of expected instruction(s):_ `vqadd.s16 D0, D0, D0'
24373 * int8x8_t vqadd_s8 (int8x8_t, int8x8_t)
24374 _Form of expected instruction(s):_ `vqadd.s8 D0, D0, D0'
24376 * uint64x1_t vqadd_u64 (uint64x1_t, uint64x1_t)
24377 _Form of expected instruction(s):_ `vqadd.u64 D0, D0, D0'
24379 * int64x1_t vqadd_s64 (int64x1_t, int64x1_t)
24380 _Form of expected instruction(s):_ `vqadd.s64 D0, D0, D0'
24382 * uint32x4_t vqaddq_u32 (uint32x4_t, uint32x4_t)
24383 _Form of expected instruction(s):_ `vqadd.u32 Q0, Q0, Q0'
24385 * uint16x8_t vqaddq_u16 (uint16x8_t, uint16x8_t)
24386 _Form of expected instruction(s):_ `vqadd.u16 Q0, Q0, Q0'
24388 * uint8x16_t vqaddq_u8 (uint8x16_t, uint8x16_t)
24389 _Form of expected instruction(s):_ `vqadd.u8 Q0, Q0, Q0'
24391 * int32x4_t vqaddq_s32 (int32x4_t, int32x4_t)
24392 _Form of expected instruction(s):_ `vqadd.s32 Q0, Q0, Q0'
24394 * int16x8_t vqaddq_s16 (int16x8_t, int16x8_t)
24395 _Form of expected instruction(s):_ `vqadd.s16 Q0, Q0, Q0'
24397 * int8x16_t vqaddq_s8 (int8x16_t, int8x16_t)
24398 _Form of expected instruction(s):_ `vqadd.s8 Q0, Q0, Q0'
24400 * uint64x2_t vqaddq_u64 (uint64x2_t, uint64x2_t)
24401 _Form of expected instruction(s):_ `vqadd.u64 Q0, Q0, Q0'
24403 * int64x2_t vqaddq_s64 (int64x2_t, int64x2_t)
24404 _Form of expected instruction(s):_ `vqadd.s64 Q0, Q0, Q0'
24406 * uint32x2_t vaddhn_u64 (uint64x2_t, uint64x2_t)
24407 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
24409 * uint16x4_t vaddhn_u32 (uint32x4_t, uint32x4_t)
24410 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
24412 * uint8x8_t vaddhn_u16 (uint16x8_t, uint16x8_t)
24413 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
24415 * int32x2_t vaddhn_s64 (int64x2_t, int64x2_t)
24416 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
24418 * int16x4_t vaddhn_s32 (int32x4_t, int32x4_t)
24419 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
24421 * int8x8_t vaddhn_s16 (int16x8_t, int16x8_t)
24422 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
24424 * uint32x2_t vraddhn_u64 (uint64x2_t, uint64x2_t)
24425 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
24427 * uint16x4_t vraddhn_u32 (uint32x4_t, uint32x4_t)
24428 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
24430 * uint8x8_t vraddhn_u16 (uint16x8_t, uint16x8_t)
24431 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
24433 * int32x2_t vraddhn_s64 (int64x2_t, int64x2_t)
24434 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
24436 * int16x4_t vraddhn_s32 (int32x4_t, int32x4_t)
24437 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
24439 * int8x8_t vraddhn_s16 (int16x8_t, int16x8_t)
24440 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
24442 5.50.3.2 Multiplication
24443 .......................
24445 * uint32x2_t vmul_u32 (uint32x2_t, uint32x2_t)
24446 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
24448 * uint16x4_t vmul_u16 (uint16x4_t, uint16x4_t)
24449 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
24451 * uint8x8_t vmul_u8 (uint8x8_t, uint8x8_t)
24452 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
24454 * int32x2_t vmul_s32 (int32x2_t, int32x2_t)
24455 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
24457 * int16x4_t vmul_s16 (int16x4_t, int16x4_t)
24458 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
24460 * int8x8_t vmul_s8 (int8x8_t, int8x8_t)
24461 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
24463 * float32x2_t vmul_f32 (float32x2_t, float32x2_t)
24464 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0'
24466 * poly8x8_t vmul_p8 (poly8x8_t, poly8x8_t)
24467 _Form of expected instruction(s):_ `vmul.p8 D0, D0, D0'
24469 * uint32x4_t vmulq_u32 (uint32x4_t, uint32x4_t)
24470 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
24472 * uint16x8_t vmulq_u16 (uint16x8_t, uint16x8_t)
24473 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
24475 * uint8x16_t vmulq_u8 (uint8x16_t, uint8x16_t)
24476 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
24478 * int32x4_t vmulq_s32 (int32x4_t, int32x4_t)
24479 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
24481 * int16x8_t vmulq_s16 (int16x8_t, int16x8_t)
24482 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
24484 * int8x16_t vmulq_s8 (int8x16_t, int8x16_t)
24485 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
24487 * float32x4_t vmulq_f32 (float32x4_t, float32x4_t)
24488 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, Q0'
24490 * poly8x16_t vmulq_p8 (poly8x16_t, poly8x16_t)
24491 _Form of expected instruction(s):_ `vmul.p8 Q0, Q0, Q0'
24493 * int32x2_t vqdmulh_s32 (int32x2_t, int32x2_t)
24494 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0'
24496 * int16x4_t vqdmulh_s16 (int16x4_t, int16x4_t)
24497 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0'
24499 * int32x4_t vqdmulhq_s32 (int32x4_t, int32x4_t)
24500 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, Q0'
24502 * int16x8_t vqdmulhq_s16 (int16x8_t, int16x8_t)
24503 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, Q0'
24505 * int32x2_t vqrdmulh_s32 (int32x2_t, int32x2_t)
24506 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0'
24508 * int16x4_t vqrdmulh_s16 (int16x4_t, int16x4_t)
24509 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0'
24511 * int32x4_t vqrdmulhq_s32 (int32x4_t, int32x4_t)
24512 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, Q0'
24514 * int16x8_t vqrdmulhq_s16 (int16x8_t, int16x8_t)
24515 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, Q0'
24517 * uint64x2_t vmull_u32 (uint32x2_t, uint32x2_t)
24518 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0'
24520 * uint32x4_t vmull_u16 (uint16x4_t, uint16x4_t)
24521 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0'
24523 * uint16x8_t vmull_u8 (uint8x8_t, uint8x8_t)
24524 _Form of expected instruction(s):_ `vmull.u8 Q0, D0, D0'
24526 * int64x2_t vmull_s32 (int32x2_t, int32x2_t)
24527 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0'
24529 * int32x4_t vmull_s16 (int16x4_t, int16x4_t)
24530 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0'
24532 * int16x8_t vmull_s8 (int8x8_t, int8x8_t)
24533 _Form of expected instruction(s):_ `vmull.s8 Q0, D0, D0'
24535 * poly16x8_t vmull_p8 (poly8x8_t, poly8x8_t)
24536 _Form of expected instruction(s):_ `vmull.p8 Q0, D0, D0'
24538 * int64x2_t vqdmull_s32 (int32x2_t, int32x2_t)
24539 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0'
24541 * int32x4_t vqdmull_s16 (int16x4_t, int16x4_t)
24542 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0'
24544 5.50.3.3 Multiply-accumulate
24545 ............................
24547 * uint32x2_t vmla_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
24548 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
24550 * uint16x4_t vmla_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
24551 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
24553 * uint8x8_t vmla_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
24554 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
24556 * int32x2_t vmla_s32 (int32x2_t, int32x2_t, int32x2_t)
24557 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
24559 * int16x4_t vmla_s16 (int16x4_t, int16x4_t, int16x4_t)
24560 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
24562 * int8x8_t vmla_s8 (int8x8_t, int8x8_t, int8x8_t)
24563 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
24565 * float32x2_t vmla_f32 (float32x2_t, float32x2_t, float32x2_t)
24566 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0'
24568 * uint32x4_t vmlaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
24569 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
24571 * uint16x8_t vmlaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
24572 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
24574 * uint8x16_t vmlaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
24575 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
24577 * int32x4_t vmlaq_s32 (int32x4_t, int32x4_t, int32x4_t)
24578 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
24580 * int16x8_t vmlaq_s16 (int16x8_t, int16x8_t, int16x8_t)
24581 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
24583 * int8x16_t vmlaq_s8 (int8x16_t, int8x16_t, int8x16_t)
24584 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
24586 * float32x4_t vmlaq_f32 (float32x4_t, float32x4_t, float32x4_t)
24587 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, Q0'
24589 * uint64x2_t vmlal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
24590 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0'
24592 * uint32x4_t vmlal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
24593 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0'
24595 * uint16x8_t vmlal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
24596 _Form of expected instruction(s):_ `vmlal.u8 Q0, D0, D0'
24598 * int64x2_t vmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
24599 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0'
24601 * int32x4_t vmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
24602 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0'
24604 * int16x8_t vmlal_s8 (int16x8_t, int8x8_t, int8x8_t)
24605 _Form of expected instruction(s):_ `vmlal.s8 Q0, D0, D0'
24607 * int64x2_t vqdmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
24608 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0'
24610 * int32x4_t vqdmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
24611 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0'
24613 5.50.3.4 Multiply-subtract
24614 ..........................
24616 * uint32x2_t vmls_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
24617 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
24619 * uint16x4_t vmls_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
24620 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
24622 * uint8x8_t vmls_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
24623 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
24625 * int32x2_t vmls_s32 (int32x2_t, int32x2_t, int32x2_t)
24626 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
24628 * int16x4_t vmls_s16 (int16x4_t, int16x4_t, int16x4_t)
24629 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
24631 * int8x8_t vmls_s8 (int8x8_t, int8x8_t, int8x8_t)
24632 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
24634 * float32x2_t vmls_f32 (float32x2_t, float32x2_t, float32x2_t)
24635 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0'
24637 * uint32x4_t vmlsq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
24638 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
24640 * uint16x8_t vmlsq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
24641 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
24643 * uint8x16_t vmlsq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
24644 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
24646 * int32x4_t vmlsq_s32 (int32x4_t, int32x4_t, int32x4_t)
24647 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
24649 * int16x8_t vmlsq_s16 (int16x8_t, int16x8_t, int16x8_t)
24650 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
24652 * int8x16_t vmlsq_s8 (int8x16_t, int8x16_t, int8x16_t)
24653 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
24655 * float32x4_t vmlsq_f32 (float32x4_t, float32x4_t, float32x4_t)
24656 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, Q0'
24658 * uint64x2_t vmlsl_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
24659 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0'
24661 * uint32x4_t vmlsl_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
24662 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0'
24664 * uint16x8_t vmlsl_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
24665 _Form of expected instruction(s):_ `vmlsl.u8 Q0, D0, D0'
24667 * int64x2_t vmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
24668 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0'
24670 * int32x4_t vmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
24671 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0'
24673 * int16x8_t vmlsl_s8 (int16x8_t, int8x8_t, int8x8_t)
24674 _Form of expected instruction(s):_ `vmlsl.s8 Q0, D0, D0'
24676 * int64x2_t vqdmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
24677 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0'
24679 * int32x4_t vqdmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
24680 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0'
24682 5.50.3.5 Subtraction
24683 ....................
24685 * uint32x2_t vsub_u32 (uint32x2_t, uint32x2_t)
24686 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
24688 * uint16x4_t vsub_u16 (uint16x4_t, uint16x4_t)
24689 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
24691 * uint8x8_t vsub_u8 (uint8x8_t, uint8x8_t)
24692 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
24694 * int32x2_t vsub_s32 (int32x2_t, int32x2_t)
24695 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
24697 * int16x4_t vsub_s16 (int16x4_t, int16x4_t)
24698 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
24700 * int8x8_t vsub_s8 (int8x8_t, int8x8_t)
24701 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
24703 * uint64x1_t vsub_u64 (uint64x1_t, uint64x1_t)
24704 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
24706 * int64x1_t vsub_s64 (int64x1_t, int64x1_t)
24707 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
24709 * float32x2_t vsub_f32 (float32x2_t, float32x2_t)
24710 _Form of expected instruction(s):_ `vsub.f32 D0, D0, D0'
24712 * uint32x4_t vsubq_u32 (uint32x4_t, uint32x4_t)
24713 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
24715 * uint16x8_t vsubq_u16 (uint16x8_t, uint16x8_t)
24716 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
24718 * uint8x16_t vsubq_u8 (uint8x16_t, uint8x16_t)
24719 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
24721 * int32x4_t vsubq_s32 (int32x4_t, int32x4_t)
24722 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
24724 * int16x8_t vsubq_s16 (int16x8_t, int16x8_t)
24725 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
24727 * int8x16_t vsubq_s8 (int8x16_t, int8x16_t)
24728 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
24730 * uint64x2_t vsubq_u64 (uint64x2_t, uint64x2_t)
24731 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
24733 * int64x2_t vsubq_s64 (int64x2_t, int64x2_t)
24734 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
24736 * float32x4_t vsubq_f32 (float32x4_t, float32x4_t)
24737 _Form of expected instruction(s):_ `vsub.f32 Q0, Q0, Q0'
24739 * uint64x2_t vsubl_u32 (uint32x2_t, uint32x2_t)
24740 _Form of expected instruction(s):_ `vsubl.u32 Q0, D0, D0'
24742 * uint32x4_t vsubl_u16 (uint16x4_t, uint16x4_t)
24743 _Form of expected instruction(s):_ `vsubl.u16 Q0, D0, D0'
24745 * uint16x8_t vsubl_u8 (uint8x8_t, uint8x8_t)
24746 _Form of expected instruction(s):_ `vsubl.u8 Q0, D0, D0'
24748 * int64x2_t vsubl_s32 (int32x2_t, int32x2_t)
24749 _Form of expected instruction(s):_ `vsubl.s32 Q0, D0, D0'
24751 * int32x4_t vsubl_s16 (int16x4_t, int16x4_t)
24752 _Form of expected instruction(s):_ `vsubl.s16 Q0, D0, D0'
24754 * int16x8_t vsubl_s8 (int8x8_t, int8x8_t)
24755 _Form of expected instruction(s):_ `vsubl.s8 Q0, D0, D0'
24757 * uint64x2_t vsubw_u32 (uint64x2_t, uint32x2_t)
24758 _Form of expected instruction(s):_ `vsubw.u32 Q0, Q0, D0'
24760 * uint32x4_t vsubw_u16 (uint32x4_t, uint16x4_t)
24761 _Form of expected instruction(s):_ `vsubw.u16 Q0, Q0, D0'
24763 * uint16x8_t vsubw_u8 (uint16x8_t, uint8x8_t)
24764 _Form of expected instruction(s):_ `vsubw.u8 Q0, Q0, D0'
24766 * int64x2_t vsubw_s32 (int64x2_t, int32x2_t)
24767 _Form of expected instruction(s):_ `vsubw.s32 Q0, Q0, D0'
24769 * int32x4_t vsubw_s16 (int32x4_t, int16x4_t)
24770 _Form of expected instruction(s):_ `vsubw.s16 Q0, Q0, D0'
24772 * int16x8_t vsubw_s8 (int16x8_t, int8x8_t)
24773 _Form of expected instruction(s):_ `vsubw.s8 Q0, Q0, D0'
24775 * uint32x2_t vhsub_u32 (uint32x2_t, uint32x2_t)
24776 _Form of expected instruction(s):_ `vhsub.u32 D0, D0, D0'
24778 * uint16x4_t vhsub_u16 (uint16x4_t, uint16x4_t)
24779 _Form of expected instruction(s):_ `vhsub.u16 D0, D0, D0'
24781 * uint8x8_t vhsub_u8 (uint8x8_t, uint8x8_t)
24782 _Form of expected instruction(s):_ `vhsub.u8 D0, D0, D0'
24784 * int32x2_t vhsub_s32 (int32x2_t, int32x2_t)
24785 _Form of expected instruction(s):_ `vhsub.s32 D0, D0, D0'
24787 * int16x4_t vhsub_s16 (int16x4_t, int16x4_t)
24788 _Form of expected instruction(s):_ `vhsub.s16 D0, D0, D0'
24790 * int8x8_t vhsub_s8 (int8x8_t, int8x8_t)
24791 _Form of expected instruction(s):_ `vhsub.s8 D0, D0, D0'
24793 * uint32x4_t vhsubq_u32 (uint32x4_t, uint32x4_t)
24794 _Form of expected instruction(s):_ `vhsub.u32 Q0, Q0, Q0'
24796 * uint16x8_t vhsubq_u16 (uint16x8_t, uint16x8_t)
24797 _Form of expected instruction(s):_ `vhsub.u16 Q0, Q0, Q0'
24799 * uint8x16_t vhsubq_u8 (uint8x16_t, uint8x16_t)
24800 _Form of expected instruction(s):_ `vhsub.u8 Q0, Q0, Q0'
24802 * int32x4_t vhsubq_s32 (int32x4_t, int32x4_t)
24803 _Form of expected instruction(s):_ `vhsub.s32 Q0, Q0, Q0'
24805 * int16x8_t vhsubq_s16 (int16x8_t, int16x8_t)
24806 _Form of expected instruction(s):_ `vhsub.s16 Q0, Q0, Q0'
24808 * int8x16_t vhsubq_s8 (int8x16_t, int8x16_t)
24809 _Form of expected instruction(s):_ `vhsub.s8 Q0, Q0, Q0'
24811 * uint32x2_t vqsub_u32 (uint32x2_t, uint32x2_t)
24812 _Form of expected instruction(s):_ `vqsub.u32 D0, D0, D0'
24814 * uint16x4_t vqsub_u16 (uint16x4_t, uint16x4_t)
24815 _Form of expected instruction(s):_ `vqsub.u16 D0, D0, D0'
24817 * uint8x8_t vqsub_u8 (uint8x8_t, uint8x8_t)
24818 _Form of expected instruction(s):_ `vqsub.u8 D0, D0, D0'
24820 * int32x2_t vqsub_s32 (int32x2_t, int32x2_t)
24821 _Form of expected instruction(s):_ `vqsub.s32 D0, D0, D0'
24823 * int16x4_t vqsub_s16 (int16x4_t, int16x4_t)
24824 _Form of expected instruction(s):_ `vqsub.s16 D0, D0, D0'
24826 * int8x8_t vqsub_s8 (int8x8_t, int8x8_t)
24827 _Form of expected instruction(s):_ `vqsub.s8 D0, D0, D0'
24829 * uint64x1_t vqsub_u64 (uint64x1_t, uint64x1_t)
24830 _Form of expected instruction(s):_ `vqsub.u64 D0, D0, D0'
24832 * int64x1_t vqsub_s64 (int64x1_t, int64x1_t)
24833 _Form of expected instruction(s):_ `vqsub.s64 D0, D0, D0'
24835 * uint32x4_t vqsubq_u32 (uint32x4_t, uint32x4_t)
24836 _Form of expected instruction(s):_ `vqsub.u32 Q0, Q0, Q0'
24838 * uint16x8_t vqsubq_u16 (uint16x8_t, uint16x8_t)
24839 _Form of expected instruction(s):_ `vqsub.u16 Q0, Q0, Q0'
24841 * uint8x16_t vqsubq_u8 (uint8x16_t, uint8x16_t)
24842 _Form of expected instruction(s):_ `vqsub.u8 Q0, Q0, Q0'
24844 * int32x4_t vqsubq_s32 (int32x4_t, int32x4_t)
24845 _Form of expected instruction(s):_ `vqsub.s32 Q0, Q0, Q0'
24847 * int16x8_t vqsubq_s16 (int16x8_t, int16x8_t)
24848 _Form of expected instruction(s):_ `vqsub.s16 Q0, Q0, Q0'
24850 * int8x16_t vqsubq_s8 (int8x16_t, int8x16_t)
24851 _Form of expected instruction(s):_ `vqsub.s8 Q0, Q0, Q0'
24853 * uint64x2_t vqsubq_u64 (uint64x2_t, uint64x2_t)
24854 _Form of expected instruction(s):_ `vqsub.u64 Q0, Q0, Q0'
24856 * int64x2_t vqsubq_s64 (int64x2_t, int64x2_t)
24857 _Form of expected instruction(s):_ `vqsub.s64 Q0, Q0, Q0'
24859 * uint32x2_t vsubhn_u64 (uint64x2_t, uint64x2_t)
24860 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
24862 * uint16x4_t vsubhn_u32 (uint32x4_t, uint32x4_t)
24863 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
24865 * uint8x8_t vsubhn_u16 (uint16x8_t, uint16x8_t)
24866 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
24868 * int32x2_t vsubhn_s64 (int64x2_t, int64x2_t)
24869 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
24871 * int16x4_t vsubhn_s32 (int32x4_t, int32x4_t)
24872 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
24874 * int8x8_t vsubhn_s16 (int16x8_t, int16x8_t)
24875 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
24877 * uint32x2_t vrsubhn_u64 (uint64x2_t, uint64x2_t)
24878 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
24880 * uint16x4_t vrsubhn_u32 (uint32x4_t, uint32x4_t)
24881 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
24883 * uint8x8_t vrsubhn_u16 (uint16x8_t, uint16x8_t)
24884 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
24886 * int32x2_t vrsubhn_s64 (int64x2_t, int64x2_t)
24887 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
24889 * int16x4_t vrsubhn_s32 (int32x4_t, int32x4_t)
24890 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
24892 * int8x8_t vrsubhn_s16 (int16x8_t, int16x8_t)
24893 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
24895 5.50.3.6 Comparison (equal-to)
24896 ..............................
24898 * uint32x2_t vceq_u32 (uint32x2_t, uint32x2_t)
24899 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
24901 * uint16x4_t vceq_u16 (uint16x4_t, uint16x4_t)
24902 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
24904 * uint8x8_t vceq_u8 (uint8x8_t, uint8x8_t)
24905 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24907 * uint32x2_t vceq_s32 (int32x2_t, int32x2_t)
24908 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
24910 * uint16x4_t vceq_s16 (int16x4_t, int16x4_t)
24911 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
24913 * uint8x8_t vceq_s8 (int8x8_t, int8x8_t)
24914 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24916 * uint32x2_t vceq_f32 (float32x2_t, float32x2_t)
24917 _Form of expected instruction(s):_ `vceq.f32 D0, D0, D0'
24919 * uint8x8_t vceq_p8 (poly8x8_t, poly8x8_t)
24920 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24922 * uint32x4_t vceqq_u32 (uint32x4_t, uint32x4_t)
24923 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
24925 * uint16x8_t vceqq_u16 (uint16x8_t, uint16x8_t)
24926 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
24928 * uint8x16_t vceqq_u8 (uint8x16_t, uint8x16_t)
24929 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24931 * uint32x4_t vceqq_s32 (int32x4_t, int32x4_t)
24932 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
24934 * uint16x8_t vceqq_s16 (int16x8_t, int16x8_t)
24935 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
24937 * uint8x16_t vceqq_s8 (int8x16_t, int8x16_t)
24938 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24940 * uint32x4_t vceqq_f32 (float32x4_t, float32x4_t)
24941 _Form of expected instruction(s):_ `vceq.f32 Q0, Q0, Q0'
24943 * uint8x16_t vceqq_p8 (poly8x16_t, poly8x16_t)
24944 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24946 5.50.3.7 Comparison (greater-than-or-equal-to)
24947 ..............................................
24949 * uint32x2_t vcge_u32 (uint32x2_t, uint32x2_t)
24950 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
24952 * uint16x4_t vcge_u16 (uint16x4_t, uint16x4_t)
24953 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
24955 * uint8x8_t vcge_u8 (uint8x8_t, uint8x8_t)
24956 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
24958 * uint32x2_t vcge_s32 (int32x2_t, int32x2_t)
24959 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
24961 * uint16x4_t vcge_s16 (int16x4_t, int16x4_t)
24962 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
24964 * uint8x8_t vcge_s8 (int8x8_t, int8x8_t)
24965 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
24967 * uint32x2_t vcge_f32 (float32x2_t, float32x2_t)
24968 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
24970 * uint32x4_t vcgeq_u32 (uint32x4_t, uint32x4_t)
24971 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
24973 * uint16x8_t vcgeq_u16 (uint16x8_t, uint16x8_t)
24974 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
24976 * uint8x16_t vcgeq_u8 (uint8x16_t, uint8x16_t)
24977 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
24979 * uint32x4_t vcgeq_s32 (int32x4_t, int32x4_t)
24980 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
24982 * uint16x8_t vcgeq_s16 (int16x8_t, int16x8_t)
24983 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
24985 * uint8x16_t vcgeq_s8 (int8x16_t, int8x16_t)
24986 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
24988 * uint32x4_t vcgeq_f32 (float32x4_t, float32x4_t)
24989 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
24991 5.50.3.8 Comparison (less-than-or-equal-to)
24992 ...........................................
24994 * uint32x2_t vcle_u32 (uint32x2_t, uint32x2_t)
24995 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
24997 * uint16x4_t vcle_u16 (uint16x4_t, uint16x4_t)
24998 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
25000 * uint8x8_t vcle_u8 (uint8x8_t, uint8x8_t)
25001 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
25003 * uint32x2_t vcle_s32 (int32x2_t, int32x2_t)
25004 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
25006 * uint16x4_t vcle_s16 (int16x4_t, int16x4_t)
25007 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
25009 * uint8x8_t vcle_s8 (int8x8_t, int8x8_t)
25010 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
25012 * uint32x2_t vcle_f32 (float32x2_t, float32x2_t)
25013 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
25015 * uint32x4_t vcleq_u32 (uint32x4_t, uint32x4_t)
25016 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
25018 * uint16x8_t vcleq_u16 (uint16x8_t, uint16x8_t)
25019 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
25021 * uint8x16_t vcleq_u8 (uint8x16_t, uint8x16_t)
25022 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
25024 * uint32x4_t vcleq_s32 (int32x4_t, int32x4_t)
25025 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
25027 * uint16x8_t vcleq_s16 (int16x8_t, int16x8_t)
25028 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
25030 * uint8x16_t vcleq_s8 (int8x16_t, int8x16_t)
25031 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
25033 * uint32x4_t vcleq_f32 (float32x4_t, float32x4_t)
25034 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
25036 5.50.3.9 Comparison (greater-than)
25037 ..................................
25039 * uint32x2_t vcgt_u32 (uint32x2_t, uint32x2_t)
25040 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
25042 * uint16x4_t vcgt_u16 (uint16x4_t, uint16x4_t)
25043 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
25045 * uint8x8_t vcgt_u8 (uint8x8_t, uint8x8_t)
25046 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
25048 * uint32x2_t vcgt_s32 (int32x2_t, int32x2_t)
25049 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
25051 * uint16x4_t vcgt_s16 (int16x4_t, int16x4_t)
25052 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
25054 * uint8x8_t vcgt_s8 (int8x8_t, int8x8_t)
25055 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
25057 * uint32x2_t vcgt_f32 (float32x2_t, float32x2_t)
25058 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
25060 * uint32x4_t vcgtq_u32 (uint32x4_t, uint32x4_t)
25061 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
25063 * uint16x8_t vcgtq_u16 (uint16x8_t, uint16x8_t)
25064 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
25066 * uint8x16_t vcgtq_u8 (uint8x16_t, uint8x16_t)
25067 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
25069 * uint32x4_t vcgtq_s32 (int32x4_t, int32x4_t)
25070 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
25072 * uint16x8_t vcgtq_s16 (int16x8_t, int16x8_t)
25073 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
25075 * uint8x16_t vcgtq_s8 (int8x16_t, int8x16_t)
25076 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
25078 * uint32x4_t vcgtq_f32 (float32x4_t, float32x4_t)
25079 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
25081 5.50.3.10 Comparison (less-than)
25082 ................................
25084 * uint32x2_t vclt_u32 (uint32x2_t, uint32x2_t)
25085 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
25087 * uint16x4_t vclt_u16 (uint16x4_t, uint16x4_t)
25088 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
25090 * uint8x8_t vclt_u8 (uint8x8_t, uint8x8_t)
25091 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
25093 * uint32x2_t vclt_s32 (int32x2_t, int32x2_t)
25094 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
25096 * uint16x4_t vclt_s16 (int16x4_t, int16x4_t)
25097 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
25099 * uint8x8_t vclt_s8 (int8x8_t, int8x8_t)
25100 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
25102 * uint32x2_t vclt_f32 (float32x2_t, float32x2_t)
25103 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
25105 * uint32x4_t vcltq_u32 (uint32x4_t, uint32x4_t)
25106 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
25108 * uint16x8_t vcltq_u16 (uint16x8_t, uint16x8_t)
25109 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
25111 * uint8x16_t vcltq_u8 (uint8x16_t, uint8x16_t)
25112 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
25114 * uint32x4_t vcltq_s32 (int32x4_t, int32x4_t)
25115 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
25117 * uint16x8_t vcltq_s16 (int16x8_t, int16x8_t)
25118 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
25120 * uint8x16_t vcltq_s8 (int8x16_t, int8x16_t)
25121 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
25123 * uint32x4_t vcltq_f32 (float32x4_t, float32x4_t)
25124 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
25126 5.50.3.11 Comparison (absolute greater-than-or-equal-to)
25127 ........................................................
25129 * uint32x2_t vcage_f32 (float32x2_t, float32x2_t)
25130 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
25132 * uint32x4_t vcageq_f32 (float32x4_t, float32x4_t)
25133 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
25135 5.50.3.12 Comparison (absolute less-than-or-equal-to)
25136 .....................................................
25138 * uint32x2_t vcale_f32 (float32x2_t, float32x2_t)
25139 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
25141 * uint32x4_t vcaleq_f32 (float32x4_t, float32x4_t)
25142 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
25144 5.50.3.13 Comparison (absolute greater-than)
25145 ............................................
25147 * uint32x2_t vcagt_f32 (float32x2_t, float32x2_t)
25148 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
25150 * uint32x4_t vcagtq_f32 (float32x4_t, float32x4_t)
25151 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
25153 5.50.3.14 Comparison (absolute less-than)
25154 .........................................
25156 * uint32x2_t vcalt_f32 (float32x2_t, float32x2_t)
25157 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
25159 * uint32x4_t vcaltq_f32 (float32x4_t, float32x4_t)
25160 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
25162 5.50.3.15 Test bits
25163 ...................
25165 * uint32x2_t vtst_u32 (uint32x2_t, uint32x2_t)
25166 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
25168 * uint16x4_t vtst_u16 (uint16x4_t, uint16x4_t)
25169 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
25171 * uint8x8_t vtst_u8 (uint8x8_t, uint8x8_t)
25172 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25174 * uint32x2_t vtst_s32 (int32x2_t, int32x2_t)
25175 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
25177 * uint16x4_t vtst_s16 (int16x4_t, int16x4_t)
25178 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
25180 * uint8x8_t vtst_s8 (int8x8_t, int8x8_t)
25181 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25183 * uint8x8_t vtst_p8 (poly8x8_t, poly8x8_t)
25184 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25186 * uint32x4_t vtstq_u32 (uint32x4_t, uint32x4_t)
25187 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
25189 * uint16x8_t vtstq_u16 (uint16x8_t, uint16x8_t)
25190 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
25192 * uint8x16_t vtstq_u8 (uint8x16_t, uint8x16_t)
25193 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25195 * uint32x4_t vtstq_s32 (int32x4_t, int32x4_t)
25196 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
25198 * uint16x8_t vtstq_s16 (int16x8_t, int16x8_t)
25199 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
25201 * uint8x16_t vtstq_s8 (int8x16_t, int8x16_t)
25202 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25204 * uint8x16_t vtstq_p8 (poly8x16_t, poly8x16_t)
25205 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25207 5.50.3.16 Absolute difference
25208 .............................
25210 * uint32x2_t vabd_u32 (uint32x2_t, uint32x2_t)
25211 _Form of expected instruction(s):_ `vabd.u32 D0, D0, D0'
25213 * uint16x4_t vabd_u16 (uint16x4_t, uint16x4_t)
25214 _Form of expected instruction(s):_ `vabd.u16 D0, D0, D0'
25216 * uint8x8_t vabd_u8 (uint8x8_t, uint8x8_t)
25217 _Form of expected instruction(s):_ `vabd.u8 D0, D0, D0'
25219 * int32x2_t vabd_s32 (int32x2_t, int32x2_t)
25220 _Form of expected instruction(s):_ `vabd.s32 D0, D0, D0'
25222 * int16x4_t vabd_s16 (int16x4_t, int16x4_t)
25223 _Form of expected instruction(s):_ `vabd.s16 D0, D0, D0'
25225 * int8x8_t vabd_s8 (int8x8_t, int8x8_t)
25226 _Form of expected instruction(s):_ `vabd.s8 D0, D0, D0'
25228 * float32x2_t vabd_f32 (float32x2_t, float32x2_t)
25229 _Form of expected instruction(s):_ `vabd.f32 D0, D0, D0'
25231 * uint32x4_t vabdq_u32 (uint32x4_t, uint32x4_t)
25232 _Form of expected instruction(s):_ `vabd.u32 Q0, Q0, Q0'
25234 * uint16x8_t vabdq_u16 (uint16x8_t, uint16x8_t)
25235 _Form of expected instruction(s):_ `vabd.u16 Q0, Q0, Q0'
25237 * uint8x16_t vabdq_u8 (uint8x16_t, uint8x16_t)
25238 _Form of expected instruction(s):_ `vabd.u8 Q0, Q0, Q0'
25240 * int32x4_t vabdq_s32 (int32x4_t, int32x4_t)
25241 _Form of expected instruction(s):_ `vabd.s32 Q0, Q0, Q0'
25243 * int16x8_t vabdq_s16 (int16x8_t, int16x8_t)
25244 _Form of expected instruction(s):_ `vabd.s16 Q0, Q0, Q0'
25246 * int8x16_t vabdq_s8 (int8x16_t, int8x16_t)
25247 _Form of expected instruction(s):_ `vabd.s8 Q0, Q0, Q0'
25249 * float32x4_t vabdq_f32 (float32x4_t, float32x4_t)
25250 _Form of expected instruction(s):_ `vabd.f32 Q0, Q0, Q0'
25252 * uint64x2_t vabdl_u32 (uint32x2_t, uint32x2_t)
25253 _Form of expected instruction(s):_ `vabdl.u32 Q0, D0, D0'
25255 * uint32x4_t vabdl_u16 (uint16x4_t, uint16x4_t)
25256 _Form of expected instruction(s):_ `vabdl.u16 Q0, D0, D0'
25258 * uint16x8_t vabdl_u8 (uint8x8_t, uint8x8_t)
25259 _Form of expected instruction(s):_ `vabdl.u8 Q0, D0, D0'
25261 * int64x2_t vabdl_s32 (int32x2_t, int32x2_t)
25262 _Form of expected instruction(s):_ `vabdl.s32 Q0, D0, D0'
25264 * int32x4_t vabdl_s16 (int16x4_t, int16x4_t)
25265 _Form of expected instruction(s):_ `vabdl.s16 Q0, D0, D0'
25267 * int16x8_t vabdl_s8 (int8x8_t, int8x8_t)
25268 _Form of expected instruction(s):_ `vabdl.s8 Q0, D0, D0'
25270 5.50.3.17 Absolute difference and accumulate
25271 ............................................
25273 * uint32x2_t vaba_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
25274 _Form of expected instruction(s):_ `vaba.u32 D0, D0, D0'
25276 * uint16x4_t vaba_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
25277 _Form of expected instruction(s):_ `vaba.u16 D0, D0, D0'
25279 * uint8x8_t vaba_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
25280 _Form of expected instruction(s):_ `vaba.u8 D0, D0, D0'
25282 * int32x2_t vaba_s32 (int32x2_t, int32x2_t, int32x2_t)
25283 _Form of expected instruction(s):_ `vaba.s32 D0, D0, D0'
25285 * int16x4_t vaba_s16 (int16x4_t, int16x4_t, int16x4_t)
25286 _Form of expected instruction(s):_ `vaba.s16 D0, D0, D0'
25288 * int8x8_t vaba_s8 (int8x8_t, int8x8_t, int8x8_t)
25289 _Form of expected instruction(s):_ `vaba.s8 D0, D0, D0'
25291 * uint32x4_t vabaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
25292 _Form of expected instruction(s):_ `vaba.u32 Q0, Q0, Q0'
25294 * uint16x8_t vabaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
25295 _Form of expected instruction(s):_ `vaba.u16 Q0, Q0, Q0'
25297 * uint8x16_t vabaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
25298 _Form of expected instruction(s):_ `vaba.u8 Q0, Q0, Q0'
25300 * int32x4_t vabaq_s32 (int32x4_t, int32x4_t, int32x4_t)
25301 _Form of expected instruction(s):_ `vaba.s32 Q0, Q0, Q0'
25303 * int16x8_t vabaq_s16 (int16x8_t, int16x8_t, int16x8_t)
25304 _Form of expected instruction(s):_ `vaba.s16 Q0, Q0, Q0'
25306 * int8x16_t vabaq_s8 (int8x16_t, int8x16_t, int8x16_t)
25307 _Form of expected instruction(s):_ `vaba.s8 Q0, Q0, Q0'
25309 * uint64x2_t vabal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
25310 _Form of expected instruction(s):_ `vabal.u32 Q0, D0, D0'
25312 * uint32x4_t vabal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
25313 _Form of expected instruction(s):_ `vabal.u16 Q0, D0, D0'
25315 * uint16x8_t vabal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
25316 _Form of expected instruction(s):_ `vabal.u8 Q0, D0, D0'
25318 * int64x2_t vabal_s32 (int64x2_t, int32x2_t, int32x2_t)
25319 _Form of expected instruction(s):_ `vabal.s32 Q0, D0, D0'
25321 * int32x4_t vabal_s16 (int32x4_t, int16x4_t, int16x4_t)
25322 _Form of expected instruction(s):_ `vabal.s16 Q0, D0, D0'
25324 * int16x8_t vabal_s8 (int16x8_t, int8x8_t, int8x8_t)
25325 _Form of expected instruction(s):_ `vabal.s8 Q0, D0, D0'
25330 * uint32x2_t vmax_u32 (uint32x2_t, uint32x2_t)
25331 _Form of expected instruction(s):_ `vmax.u32 D0, D0, D0'
25333 * uint16x4_t vmax_u16 (uint16x4_t, uint16x4_t)
25334 _Form of expected instruction(s):_ `vmax.u16 D0, D0, D0'
25336 * uint8x8_t vmax_u8 (uint8x8_t, uint8x8_t)
25337 _Form of expected instruction(s):_ `vmax.u8 D0, D0, D0'
25339 * int32x2_t vmax_s32 (int32x2_t, int32x2_t)
25340 _Form of expected instruction(s):_ `vmax.s32 D0, D0, D0'
25342 * int16x4_t vmax_s16 (int16x4_t, int16x4_t)
25343 _Form of expected instruction(s):_ `vmax.s16 D0, D0, D0'
25345 * int8x8_t vmax_s8 (int8x8_t, int8x8_t)
25346 _Form of expected instruction(s):_ `vmax.s8 D0, D0, D0'
25348 * float32x2_t vmax_f32 (float32x2_t, float32x2_t)
25349 _Form of expected instruction(s):_ `vmax.f32 D0, D0, D0'
25351 * uint32x4_t vmaxq_u32 (uint32x4_t, uint32x4_t)
25352 _Form of expected instruction(s):_ `vmax.u32 Q0, Q0, Q0'
25354 * uint16x8_t vmaxq_u16 (uint16x8_t, uint16x8_t)
25355 _Form of expected instruction(s):_ `vmax.u16 Q0, Q0, Q0'
25357 * uint8x16_t vmaxq_u8 (uint8x16_t, uint8x16_t)
25358 _Form of expected instruction(s):_ `vmax.u8 Q0, Q0, Q0'
25360 * int32x4_t vmaxq_s32 (int32x4_t, int32x4_t)
25361 _Form of expected instruction(s):_ `vmax.s32 Q0, Q0, Q0'
25363 * int16x8_t vmaxq_s16 (int16x8_t, int16x8_t)
25364 _Form of expected instruction(s):_ `vmax.s16 Q0, Q0, Q0'
25366 * int8x16_t vmaxq_s8 (int8x16_t, int8x16_t)
25367 _Form of expected instruction(s):_ `vmax.s8 Q0, Q0, Q0'
25369 * float32x4_t vmaxq_f32 (float32x4_t, float32x4_t)
25370 _Form of expected instruction(s):_ `vmax.f32 Q0, Q0, Q0'
25375 * uint32x2_t vmin_u32 (uint32x2_t, uint32x2_t)
25376 _Form of expected instruction(s):_ `vmin.u32 D0, D0, D0'
25378 * uint16x4_t vmin_u16 (uint16x4_t, uint16x4_t)
25379 _Form of expected instruction(s):_ `vmin.u16 D0, D0, D0'
25381 * uint8x8_t vmin_u8 (uint8x8_t, uint8x8_t)
25382 _Form of expected instruction(s):_ `vmin.u8 D0, D0, D0'
25384 * int32x2_t vmin_s32 (int32x2_t, int32x2_t)
25385 _Form of expected instruction(s):_ `vmin.s32 D0, D0, D0'
25387 * int16x4_t vmin_s16 (int16x4_t, int16x4_t)
25388 _Form of expected instruction(s):_ `vmin.s16 D0, D0, D0'
25390 * int8x8_t vmin_s8 (int8x8_t, int8x8_t)
25391 _Form of expected instruction(s):_ `vmin.s8 D0, D0, D0'
25393 * float32x2_t vmin_f32 (float32x2_t, float32x2_t)
25394 _Form of expected instruction(s):_ `vmin.f32 D0, D0, D0'
25396 * uint32x4_t vminq_u32 (uint32x4_t, uint32x4_t)
25397 _Form of expected instruction(s):_ `vmin.u32 Q0, Q0, Q0'
25399 * uint16x8_t vminq_u16 (uint16x8_t, uint16x8_t)
25400 _Form of expected instruction(s):_ `vmin.u16 Q0, Q0, Q0'
25402 * uint8x16_t vminq_u8 (uint8x16_t, uint8x16_t)
25403 _Form of expected instruction(s):_ `vmin.u8 Q0, Q0, Q0'
25405 * int32x4_t vminq_s32 (int32x4_t, int32x4_t)
25406 _Form of expected instruction(s):_ `vmin.s32 Q0, Q0, Q0'
25408 * int16x8_t vminq_s16 (int16x8_t, int16x8_t)
25409 _Form of expected instruction(s):_ `vmin.s16 Q0, Q0, Q0'
25411 * int8x16_t vminq_s8 (int8x16_t, int8x16_t)
25412 _Form of expected instruction(s):_ `vmin.s8 Q0, Q0, Q0'
25414 * float32x4_t vminq_f32 (float32x4_t, float32x4_t)
25415 _Form of expected instruction(s):_ `vmin.f32 Q0, Q0, Q0'
25417 5.50.3.20 Pairwise add
25418 ......................
25420 * uint32x2_t vpadd_u32 (uint32x2_t, uint32x2_t)
25421 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
25423 * uint16x4_t vpadd_u16 (uint16x4_t, uint16x4_t)
25424 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
25426 * uint8x8_t vpadd_u8 (uint8x8_t, uint8x8_t)
25427 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
25429 * int32x2_t vpadd_s32 (int32x2_t, int32x2_t)
25430 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
25432 * int16x4_t vpadd_s16 (int16x4_t, int16x4_t)
25433 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
25435 * int8x8_t vpadd_s8 (int8x8_t, int8x8_t)
25436 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
25438 * float32x2_t vpadd_f32 (float32x2_t, float32x2_t)
25439 _Form of expected instruction(s):_ `vpadd.f32 D0, D0, D0'
25441 * uint64x1_t vpaddl_u32 (uint32x2_t)
25442 _Form of expected instruction(s):_ `vpaddl.u32 D0, D0'
25444 * uint32x2_t vpaddl_u16 (uint16x4_t)
25445 _Form of expected instruction(s):_ `vpaddl.u16 D0, D0'
25447 * uint16x4_t vpaddl_u8 (uint8x8_t)
25448 _Form of expected instruction(s):_ `vpaddl.u8 D0, D0'
25450 * int64x1_t vpaddl_s32 (int32x2_t)
25451 _Form of expected instruction(s):_ `vpaddl.s32 D0, D0'
25453 * int32x2_t vpaddl_s16 (int16x4_t)
25454 _Form of expected instruction(s):_ `vpaddl.s16 D0, D0'
25456 * int16x4_t vpaddl_s8 (int8x8_t)
25457 _Form of expected instruction(s):_ `vpaddl.s8 D0, D0'
25459 * uint64x2_t vpaddlq_u32 (uint32x4_t)
25460 _Form of expected instruction(s):_ `vpaddl.u32 Q0, Q0'
25462 * uint32x4_t vpaddlq_u16 (uint16x8_t)
25463 _Form of expected instruction(s):_ `vpaddl.u16 Q0, Q0'
25465 * uint16x8_t vpaddlq_u8 (uint8x16_t)
25466 _Form of expected instruction(s):_ `vpaddl.u8 Q0, Q0'
25468 * int64x2_t vpaddlq_s32 (int32x4_t)
25469 _Form of expected instruction(s):_ `vpaddl.s32 Q0, Q0'
25471 * int32x4_t vpaddlq_s16 (int16x8_t)
25472 _Form of expected instruction(s):_ `vpaddl.s16 Q0, Q0'
25474 * int16x8_t vpaddlq_s8 (int8x16_t)
25475 _Form of expected instruction(s):_ `vpaddl.s8 Q0, Q0'
25477 5.50.3.21 Pairwise add, single_opcode widen and accumulate
25478 ..........................................................
25480 * uint64x1_t vpadal_u32 (uint64x1_t, uint32x2_t)
25481 _Form of expected instruction(s):_ `vpadal.u32 D0, D0'
25483 * uint32x2_t vpadal_u16 (uint32x2_t, uint16x4_t)
25484 _Form of expected instruction(s):_ `vpadal.u16 D0, D0'
25486 * uint16x4_t vpadal_u8 (uint16x4_t, uint8x8_t)
25487 _Form of expected instruction(s):_ `vpadal.u8 D0, D0'
25489 * int64x1_t vpadal_s32 (int64x1_t, int32x2_t)
25490 _Form of expected instruction(s):_ `vpadal.s32 D0, D0'
25492 * int32x2_t vpadal_s16 (int32x2_t, int16x4_t)
25493 _Form of expected instruction(s):_ `vpadal.s16 D0, D0'
25495 * int16x4_t vpadal_s8 (int16x4_t, int8x8_t)
25496 _Form of expected instruction(s):_ `vpadal.s8 D0, D0'
25498 * uint64x2_t vpadalq_u32 (uint64x2_t, uint32x4_t)
25499 _Form of expected instruction(s):_ `vpadal.u32 Q0, Q0'
25501 * uint32x4_t vpadalq_u16 (uint32x4_t, uint16x8_t)
25502 _Form of expected instruction(s):_ `vpadal.u16 Q0, Q0'
25504 * uint16x8_t vpadalq_u8 (uint16x8_t, uint8x16_t)
25505 _Form of expected instruction(s):_ `vpadal.u8 Q0, Q0'
25507 * int64x2_t vpadalq_s32 (int64x2_t, int32x4_t)
25508 _Form of expected instruction(s):_ `vpadal.s32 Q0, Q0'
25510 * int32x4_t vpadalq_s16 (int32x4_t, int16x8_t)
25511 _Form of expected instruction(s):_ `vpadal.s16 Q0, Q0'
25513 * int16x8_t vpadalq_s8 (int16x8_t, int8x16_t)
25514 _Form of expected instruction(s):_ `vpadal.s8 Q0, Q0'
25516 5.50.3.22 Folding maximum
25517 .........................
25519 * uint32x2_t vpmax_u32 (uint32x2_t, uint32x2_t)
25520 _Form of expected instruction(s):_ `vpmax.u32 D0, D0, D0'
25522 * uint16x4_t vpmax_u16 (uint16x4_t, uint16x4_t)
25523 _Form of expected instruction(s):_ `vpmax.u16 D0, D0, D0'
25525 * uint8x8_t vpmax_u8 (uint8x8_t, uint8x8_t)
25526 _Form of expected instruction(s):_ `vpmax.u8 D0, D0, D0'
25528 * int32x2_t vpmax_s32 (int32x2_t, int32x2_t)
25529 _Form of expected instruction(s):_ `vpmax.s32 D0, D0, D0'
25531 * int16x4_t vpmax_s16 (int16x4_t, int16x4_t)
25532 _Form of expected instruction(s):_ `vpmax.s16 D0, D0, D0'
25534 * int8x8_t vpmax_s8 (int8x8_t, int8x8_t)
25535 _Form of expected instruction(s):_ `vpmax.s8 D0, D0, D0'
25537 * float32x2_t vpmax_f32 (float32x2_t, float32x2_t)
25538 _Form of expected instruction(s):_ `vpmax.f32 D0, D0, D0'
25540 5.50.3.23 Folding minimum
25541 .........................
25543 * uint32x2_t vpmin_u32 (uint32x2_t, uint32x2_t)
25544 _Form of expected instruction(s):_ `vpmin.u32 D0, D0, D0'
25546 * uint16x4_t vpmin_u16 (uint16x4_t, uint16x4_t)
25547 _Form of expected instruction(s):_ `vpmin.u16 D0, D0, D0'
25549 * uint8x8_t vpmin_u8 (uint8x8_t, uint8x8_t)
25550 _Form of expected instruction(s):_ `vpmin.u8 D0, D0, D0'
25552 * int32x2_t vpmin_s32 (int32x2_t, int32x2_t)
25553 _Form of expected instruction(s):_ `vpmin.s32 D0, D0, D0'
25555 * int16x4_t vpmin_s16 (int16x4_t, int16x4_t)
25556 _Form of expected instruction(s):_ `vpmin.s16 D0, D0, D0'
25558 * int8x8_t vpmin_s8 (int8x8_t, int8x8_t)
25559 _Form of expected instruction(s):_ `vpmin.s8 D0, D0, D0'
25561 * float32x2_t vpmin_f32 (float32x2_t, float32x2_t)
25562 _Form of expected instruction(s):_ `vpmin.f32 D0, D0, D0'
25564 5.50.3.24 Reciprocal step
25565 .........................
25567 * float32x2_t vrecps_f32 (float32x2_t, float32x2_t)
25568 _Form of expected instruction(s):_ `vrecps.f32 D0, D0, D0'
25570 * float32x4_t vrecpsq_f32 (float32x4_t, float32x4_t)
25571 _Form of expected instruction(s):_ `vrecps.f32 Q0, Q0, Q0'
25573 * float32x2_t vrsqrts_f32 (float32x2_t, float32x2_t)
25574 _Form of expected instruction(s):_ `vrsqrts.f32 D0, D0, D0'
25576 * float32x4_t vrsqrtsq_f32 (float32x4_t, float32x4_t)
25577 _Form of expected instruction(s):_ `vrsqrts.f32 Q0, Q0, Q0'
25579 5.50.3.25 Vector shift left
25580 ...........................
25582 * uint32x2_t vshl_u32 (uint32x2_t, int32x2_t)
25583 _Form of expected instruction(s):_ `vshl.u32 D0, D0, D0'
25585 * uint16x4_t vshl_u16 (uint16x4_t, int16x4_t)
25586 _Form of expected instruction(s):_ `vshl.u16 D0, D0, D0'
25588 * uint8x8_t vshl_u8 (uint8x8_t, int8x8_t)
25589 _Form of expected instruction(s):_ `vshl.u8 D0, D0, D0'
25591 * int32x2_t vshl_s32 (int32x2_t, int32x2_t)
25592 _Form of expected instruction(s):_ `vshl.s32 D0, D0, D0'
25594 * int16x4_t vshl_s16 (int16x4_t, int16x4_t)
25595 _Form of expected instruction(s):_ `vshl.s16 D0, D0, D0'
25597 * int8x8_t vshl_s8 (int8x8_t, int8x8_t)
25598 _Form of expected instruction(s):_ `vshl.s8 D0, D0, D0'
25600 * uint64x1_t vshl_u64 (uint64x1_t, int64x1_t)
25601 _Form of expected instruction(s):_ `vshl.u64 D0, D0, D0'
25603 * int64x1_t vshl_s64 (int64x1_t, int64x1_t)
25604 _Form of expected instruction(s):_ `vshl.s64 D0, D0, D0'
25606 * uint32x4_t vshlq_u32 (uint32x4_t, int32x4_t)
25607 _Form of expected instruction(s):_ `vshl.u32 Q0, Q0, Q0'
25609 * uint16x8_t vshlq_u16 (uint16x8_t, int16x8_t)
25610 _Form of expected instruction(s):_ `vshl.u16 Q0, Q0, Q0'
25612 * uint8x16_t vshlq_u8 (uint8x16_t, int8x16_t)
25613 _Form of expected instruction(s):_ `vshl.u8 Q0, Q0, Q0'
25615 * int32x4_t vshlq_s32 (int32x4_t, int32x4_t)
25616 _Form of expected instruction(s):_ `vshl.s32 Q0, Q0, Q0'
25618 * int16x8_t vshlq_s16 (int16x8_t, int16x8_t)
25619 _Form of expected instruction(s):_ `vshl.s16 Q0, Q0, Q0'
25621 * int8x16_t vshlq_s8 (int8x16_t, int8x16_t)
25622 _Form of expected instruction(s):_ `vshl.s8 Q0, Q0, Q0'
25624 * uint64x2_t vshlq_u64 (uint64x2_t, int64x2_t)
25625 _Form of expected instruction(s):_ `vshl.u64 Q0, Q0, Q0'
25627 * int64x2_t vshlq_s64 (int64x2_t, int64x2_t)
25628 _Form of expected instruction(s):_ `vshl.s64 Q0, Q0, Q0'
25630 * uint32x2_t vrshl_u32 (uint32x2_t, int32x2_t)
25631 _Form of expected instruction(s):_ `vrshl.u32 D0, D0, D0'
25633 * uint16x4_t vrshl_u16 (uint16x4_t, int16x4_t)
25634 _Form of expected instruction(s):_ `vrshl.u16 D0, D0, D0'
25636 * uint8x8_t vrshl_u8 (uint8x8_t, int8x8_t)
25637 _Form of expected instruction(s):_ `vrshl.u8 D0, D0, D0'
25639 * int32x2_t vrshl_s32 (int32x2_t, int32x2_t)
25640 _Form of expected instruction(s):_ `vrshl.s32 D0, D0, D0'
25642 * int16x4_t vrshl_s16 (int16x4_t, int16x4_t)
25643 _Form of expected instruction(s):_ `vrshl.s16 D0, D0, D0'
25645 * int8x8_t vrshl_s8 (int8x8_t, int8x8_t)
25646 _Form of expected instruction(s):_ `vrshl.s8 D0, D0, D0'
25648 * uint64x1_t vrshl_u64 (uint64x1_t, int64x1_t)
25649 _Form of expected instruction(s):_ `vrshl.u64 D0, D0, D0'
25651 * int64x1_t vrshl_s64 (int64x1_t, int64x1_t)
25652 _Form of expected instruction(s):_ `vrshl.s64 D0, D0, D0'
25654 * uint32x4_t vrshlq_u32 (uint32x4_t, int32x4_t)
25655 _Form of expected instruction(s):_ `vrshl.u32 Q0, Q0, Q0'
25657 * uint16x8_t vrshlq_u16 (uint16x8_t, int16x8_t)
25658 _Form of expected instruction(s):_ `vrshl.u16 Q0, Q0, Q0'
25660 * uint8x16_t vrshlq_u8 (uint8x16_t, int8x16_t)
25661 _Form of expected instruction(s):_ `vrshl.u8 Q0, Q0, Q0'
25663 * int32x4_t vrshlq_s32 (int32x4_t, int32x4_t)
25664 _Form of expected instruction(s):_ `vrshl.s32 Q0, Q0, Q0'
25666 * int16x8_t vrshlq_s16 (int16x8_t, int16x8_t)
25667 _Form of expected instruction(s):_ `vrshl.s16 Q0, Q0, Q0'
25669 * int8x16_t vrshlq_s8 (int8x16_t, int8x16_t)
25670 _Form of expected instruction(s):_ `vrshl.s8 Q0, Q0, Q0'
25672 * uint64x2_t vrshlq_u64 (uint64x2_t, int64x2_t)
25673 _Form of expected instruction(s):_ `vrshl.u64 Q0, Q0, Q0'
25675 * int64x2_t vrshlq_s64 (int64x2_t, int64x2_t)
25676 _Form of expected instruction(s):_ `vrshl.s64 Q0, Q0, Q0'
25678 * uint32x2_t vqshl_u32 (uint32x2_t, int32x2_t)
25679 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, D0'
25681 * uint16x4_t vqshl_u16 (uint16x4_t, int16x4_t)
25682 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, D0'
25684 * uint8x8_t vqshl_u8 (uint8x8_t, int8x8_t)
25685 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, D0'
25687 * int32x2_t vqshl_s32 (int32x2_t, int32x2_t)
25688 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, D0'
25690 * int16x4_t vqshl_s16 (int16x4_t, int16x4_t)
25691 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, D0'
25693 * int8x8_t vqshl_s8 (int8x8_t, int8x8_t)
25694 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, D0'
25696 * uint64x1_t vqshl_u64 (uint64x1_t, int64x1_t)
25697 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, D0'
25699 * int64x1_t vqshl_s64 (int64x1_t, int64x1_t)
25700 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, D0'
25702 * uint32x4_t vqshlq_u32 (uint32x4_t, int32x4_t)
25703 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, Q0'
25705 * uint16x8_t vqshlq_u16 (uint16x8_t, int16x8_t)
25706 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, Q0'
25708 * uint8x16_t vqshlq_u8 (uint8x16_t, int8x16_t)
25709 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, Q0'
25711 * int32x4_t vqshlq_s32 (int32x4_t, int32x4_t)
25712 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, Q0'
25714 * int16x8_t vqshlq_s16 (int16x8_t, int16x8_t)
25715 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, Q0'
25717 * int8x16_t vqshlq_s8 (int8x16_t, int8x16_t)
25718 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, Q0'
25720 * uint64x2_t vqshlq_u64 (uint64x2_t, int64x2_t)
25721 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, Q0'
25723 * int64x2_t vqshlq_s64 (int64x2_t, int64x2_t)
25724 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, Q0'
25726 * uint32x2_t vqrshl_u32 (uint32x2_t, int32x2_t)
25727 _Form of expected instruction(s):_ `vqrshl.u32 D0, D0, D0'
25729 * uint16x4_t vqrshl_u16 (uint16x4_t, int16x4_t)
25730 _Form of expected instruction(s):_ `vqrshl.u16 D0, D0, D0'
25732 * uint8x8_t vqrshl_u8 (uint8x8_t, int8x8_t)
25733 _Form of expected instruction(s):_ `vqrshl.u8 D0, D0, D0'
25735 * int32x2_t vqrshl_s32 (int32x2_t, int32x2_t)
25736 _Form of expected instruction(s):_ `vqrshl.s32 D0, D0, D0'
25738 * int16x4_t vqrshl_s16 (int16x4_t, int16x4_t)
25739 _Form of expected instruction(s):_ `vqrshl.s16 D0, D0, D0'
25741 * int8x8_t vqrshl_s8 (int8x8_t, int8x8_t)
25742 _Form of expected instruction(s):_ `vqrshl.s8 D0, D0, D0'
25744 * uint64x1_t vqrshl_u64 (uint64x1_t, int64x1_t)
25745 _Form of expected instruction(s):_ `vqrshl.u64 D0, D0, D0'
25747 * int64x1_t vqrshl_s64 (int64x1_t, int64x1_t)
25748 _Form of expected instruction(s):_ `vqrshl.s64 D0, D0, D0'
25750 * uint32x4_t vqrshlq_u32 (uint32x4_t, int32x4_t)
25751 _Form of expected instruction(s):_ `vqrshl.u32 Q0, Q0, Q0'
25753 * uint16x8_t vqrshlq_u16 (uint16x8_t, int16x8_t)
25754 _Form of expected instruction(s):_ `vqrshl.u16 Q0, Q0, Q0'
25756 * uint8x16_t vqrshlq_u8 (uint8x16_t, int8x16_t)
25757 _Form of expected instruction(s):_ `vqrshl.u8 Q0, Q0, Q0'
25759 * int32x4_t vqrshlq_s32 (int32x4_t, int32x4_t)
25760 _Form of expected instruction(s):_ `vqrshl.s32 Q0, Q0, Q0'
25762 * int16x8_t vqrshlq_s16 (int16x8_t, int16x8_t)
25763 _Form of expected instruction(s):_ `vqrshl.s16 Q0, Q0, Q0'
25765 * int8x16_t vqrshlq_s8 (int8x16_t, int8x16_t)
25766 _Form of expected instruction(s):_ `vqrshl.s8 Q0, Q0, Q0'
25768 * uint64x2_t vqrshlq_u64 (uint64x2_t, int64x2_t)
25769 _Form of expected instruction(s):_ `vqrshl.u64 Q0, Q0, Q0'
25771 * int64x2_t vqrshlq_s64 (int64x2_t, int64x2_t)
25772 _Form of expected instruction(s):_ `vqrshl.s64 Q0, Q0, Q0'
25774 5.50.3.26 Vector shift left by constant
25775 .......................................
25777 * uint32x2_t vshl_n_u32 (uint32x2_t, const int)
25778 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
25780 * uint16x4_t vshl_n_u16 (uint16x4_t, const int)
25781 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
25783 * uint8x8_t vshl_n_u8 (uint8x8_t, const int)
25784 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
25786 * int32x2_t vshl_n_s32 (int32x2_t, const int)
25787 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
25789 * int16x4_t vshl_n_s16 (int16x4_t, const int)
25790 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
25792 * int8x8_t vshl_n_s8 (int8x8_t, const int)
25793 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
25795 * uint64x1_t vshl_n_u64 (uint64x1_t, const int)
25796 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
25798 * int64x1_t vshl_n_s64 (int64x1_t, const int)
25799 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
25801 * uint32x4_t vshlq_n_u32 (uint32x4_t, const int)
25802 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
25804 * uint16x8_t vshlq_n_u16 (uint16x8_t, const int)
25805 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
25807 * uint8x16_t vshlq_n_u8 (uint8x16_t, const int)
25808 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
25810 * int32x4_t vshlq_n_s32 (int32x4_t, const int)
25811 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
25813 * int16x8_t vshlq_n_s16 (int16x8_t, const int)
25814 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
25816 * int8x16_t vshlq_n_s8 (int8x16_t, const int)
25817 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
25819 * uint64x2_t vshlq_n_u64 (uint64x2_t, const int)
25820 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
25822 * int64x2_t vshlq_n_s64 (int64x2_t, const int)
25823 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
25825 * uint32x2_t vqshl_n_u32 (uint32x2_t, const int)
25826 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, #0'
25828 * uint16x4_t vqshl_n_u16 (uint16x4_t, const int)
25829 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, #0'
25831 * uint8x8_t vqshl_n_u8 (uint8x8_t, const int)
25832 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, #0'
25834 * int32x2_t vqshl_n_s32 (int32x2_t, const int)
25835 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, #0'
25837 * int16x4_t vqshl_n_s16 (int16x4_t, const int)
25838 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, #0'
25840 * int8x8_t vqshl_n_s8 (int8x8_t, const int)
25841 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, #0'
25843 * uint64x1_t vqshl_n_u64 (uint64x1_t, const int)
25844 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, #0'
25846 * int64x1_t vqshl_n_s64 (int64x1_t, const int)
25847 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, #0'
25849 * uint32x4_t vqshlq_n_u32 (uint32x4_t, const int)
25850 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, #0'
25852 * uint16x8_t vqshlq_n_u16 (uint16x8_t, const int)
25853 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, #0'
25855 * uint8x16_t vqshlq_n_u8 (uint8x16_t, const int)
25856 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, #0'
25858 * int32x4_t vqshlq_n_s32 (int32x4_t, const int)
25859 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, #0'
25861 * int16x8_t vqshlq_n_s16 (int16x8_t, const int)
25862 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, #0'
25864 * int8x16_t vqshlq_n_s8 (int8x16_t, const int)
25865 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, #0'
25867 * uint64x2_t vqshlq_n_u64 (uint64x2_t, const int)
25868 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, #0'
25870 * int64x2_t vqshlq_n_s64 (int64x2_t, const int)
25871 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, #0'
25873 * uint64x1_t vqshlu_n_s64 (int64x1_t, const int)
25874 _Form of expected instruction(s):_ `vqshlu.s64 D0, D0, #0'
25876 * uint32x2_t vqshlu_n_s32 (int32x2_t, const int)
25877 _Form of expected instruction(s):_ `vqshlu.s32 D0, D0, #0'
25879 * uint16x4_t vqshlu_n_s16 (int16x4_t, const int)
25880 _Form of expected instruction(s):_ `vqshlu.s16 D0, D0, #0'
25882 * uint8x8_t vqshlu_n_s8 (int8x8_t, const int)
25883 _Form of expected instruction(s):_ `vqshlu.s8 D0, D0, #0'
25885 * uint64x2_t vqshluq_n_s64 (int64x2_t, const int)
25886 _Form of expected instruction(s):_ `vqshlu.s64 Q0, Q0, #0'
25888 * uint32x4_t vqshluq_n_s32 (int32x4_t, const int)
25889 _Form of expected instruction(s):_ `vqshlu.s32 Q0, Q0, #0'
25891 * uint16x8_t vqshluq_n_s16 (int16x8_t, const int)
25892 _Form of expected instruction(s):_ `vqshlu.s16 Q0, Q0, #0'
25894 * uint8x16_t vqshluq_n_s8 (int8x16_t, const int)
25895 _Form of expected instruction(s):_ `vqshlu.s8 Q0, Q0, #0'
25897 * uint64x2_t vshll_n_u32 (uint32x2_t, const int)
25898 _Form of expected instruction(s):_ `vshll.u32 Q0, D0, #0'
25900 * uint32x4_t vshll_n_u16 (uint16x4_t, const int)
25901 _Form of expected instruction(s):_ `vshll.u16 Q0, D0, #0'
25903 * uint16x8_t vshll_n_u8 (uint8x8_t, const int)
25904 _Form of expected instruction(s):_ `vshll.u8 Q0, D0, #0'
25906 * int64x2_t vshll_n_s32 (int32x2_t, const int)
25907 _Form of expected instruction(s):_ `vshll.s32 Q0, D0, #0'
25909 * int32x4_t vshll_n_s16 (int16x4_t, const int)
25910 _Form of expected instruction(s):_ `vshll.s16 Q0, D0, #0'
25912 * int16x8_t vshll_n_s8 (int8x8_t, const int)
25913 _Form of expected instruction(s):_ `vshll.s8 Q0, D0, #0'
25915 5.50.3.27 Vector shift right by constant
25916 ........................................
25918 * uint32x2_t vshr_n_u32 (uint32x2_t, const int)
25919 _Form of expected instruction(s):_ `vshr.u32 D0, D0, #0'
25921 * uint16x4_t vshr_n_u16 (uint16x4_t, const int)
25922 _Form of expected instruction(s):_ `vshr.u16 D0, D0, #0'
25924 * uint8x8_t vshr_n_u8 (uint8x8_t, const int)
25925 _Form of expected instruction(s):_ `vshr.u8 D0, D0, #0'
25927 * int32x2_t vshr_n_s32 (int32x2_t, const int)
25928 _Form of expected instruction(s):_ `vshr.s32 D0, D0, #0'
25930 * int16x4_t vshr_n_s16 (int16x4_t, const int)
25931 _Form of expected instruction(s):_ `vshr.s16 D0, D0, #0'
25933 * int8x8_t vshr_n_s8 (int8x8_t, const int)
25934 _Form of expected instruction(s):_ `vshr.s8 D0, D0, #0'
25936 * uint64x1_t vshr_n_u64 (uint64x1_t, const int)
25937 _Form of expected instruction(s):_ `vshr.u64 D0, D0, #0'
25939 * int64x1_t vshr_n_s64 (int64x1_t, const int)
25940 _Form of expected instruction(s):_ `vshr.s64 D0, D0, #0'
25942 * uint32x4_t vshrq_n_u32 (uint32x4_t, const int)
25943 _Form of expected instruction(s):_ `vshr.u32 Q0, Q0, #0'
25945 * uint16x8_t vshrq_n_u16 (uint16x8_t, const int)
25946 _Form of expected instruction(s):_ `vshr.u16 Q0, Q0, #0'
25948 * uint8x16_t vshrq_n_u8 (uint8x16_t, const int)
25949 _Form of expected instruction(s):_ `vshr.u8 Q0, Q0, #0'
25951 * int32x4_t vshrq_n_s32 (int32x4_t, const int)
25952 _Form of expected instruction(s):_ `vshr.s32 Q0, Q0, #0'
25954 * int16x8_t vshrq_n_s16 (int16x8_t, const int)
25955 _Form of expected instruction(s):_ `vshr.s16 Q0, Q0, #0'
25957 * int8x16_t vshrq_n_s8 (int8x16_t, const int)
25958 _Form of expected instruction(s):_ `vshr.s8 Q0, Q0, #0'
25960 * uint64x2_t vshrq_n_u64 (uint64x2_t, const int)
25961 _Form of expected instruction(s):_ `vshr.u64 Q0, Q0, #0'
25963 * int64x2_t vshrq_n_s64 (int64x2_t, const int)
25964 _Form of expected instruction(s):_ `vshr.s64 Q0, Q0, #0'
25966 * uint32x2_t vrshr_n_u32 (uint32x2_t, const int)
25967 _Form of expected instruction(s):_ `vrshr.u32 D0, D0, #0'
25969 * uint16x4_t vrshr_n_u16 (uint16x4_t, const int)
25970 _Form of expected instruction(s):_ `vrshr.u16 D0, D0, #0'
25972 * uint8x8_t vrshr_n_u8 (uint8x8_t, const int)
25973 _Form of expected instruction(s):_ `vrshr.u8 D0, D0, #0'
25975 * int32x2_t vrshr_n_s32 (int32x2_t, const int)
25976 _Form of expected instruction(s):_ `vrshr.s32 D0, D0, #0'
25978 * int16x4_t vrshr_n_s16 (int16x4_t, const int)
25979 _Form of expected instruction(s):_ `vrshr.s16 D0, D0, #0'
25981 * int8x8_t vrshr_n_s8 (int8x8_t, const int)
25982 _Form of expected instruction(s):_ `vrshr.s8 D0, D0, #0'
25984 * uint64x1_t vrshr_n_u64 (uint64x1_t, const int)
25985 _Form of expected instruction(s):_ `vrshr.u64 D0, D0, #0'
25987 * int64x1_t vrshr_n_s64 (int64x1_t, const int)
25988 _Form of expected instruction(s):_ `vrshr.s64 D0, D0, #0'
25990 * uint32x4_t vrshrq_n_u32 (uint32x4_t, const int)
25991 _Form of expected instruction(s):_ `vrshr.u32 Q0, Q0, #0'
25993 * uint16x8_t vrshrq_n_u16 (uint16x8_t, const int)
25994 _Form of expected instruction(s):_ `vrshr.u16 Q0, Q0, #0'
25996 * uint8x16_t vrshrq_n_u8 (uint8x16_t, const int)
25997 _Form of expected instruction(s):_ `vrshr.u8 Q0, Q0, #0'
25999 * int32x4_t vrshrq_n_s32 (int32x4_t, const int)
26000 _Form of expected instruction(s):_ `vrshr.s32 Q0, Q0, #0'
26002 * int16x8_t vrshrq_n_s16 (int16x8_t, const int)
26003 _Form of expected instruction(s):_ `vrshr.s16 Q0, Q0, #0'
26005 * int8x16_t vrshrq_n_s8 (int8x16_t, const int)
26006 _Form of expected instruction(s):_ `vrshr.s8 Q0, Q0, #0'
26008 * uint64x2_t vrshrq_n_u64 (uint64x2_t, const int)
26009 _Form of expected instruction(s):_ `vrshr.u64 Q0, Q0, #0'
26011 * int64x2_t vrshrq_n_s64 (int64x2_t, const int)
26012 _Form of expected instruction(s):_ `vrshr.s64 Q0, Q0, #0'
26014 * uint32x2_t vshrn_n_u64 (uint64x2_t, const int)
26015 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
26017 * uint16x4_t vshrn_n_u32 (uint32x4_t, const int)
26018 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
26020 * uint8x8_t vshrn_n_u16 (uint16x8_t, const int)
26021 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
26023 * int32x2_t vshrn_n_s64 (int64x2_t, const int)
26024 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
26026 * int16x4_t vshrn_n_s32 (int32x4_t, const int)
26027 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
26029 * int8x8_t vshrn_n_s16 (int16x8_t, const int)
26030 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
26032 * uint32x2_t vrshrn_n_u64 (uint64x2_t, const int)
26033 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
26035 * uint16x4_t vrshrn_n_u32 (uint32x4_t, const int)
26036 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
26038 * uint8x8_t vrshrn_n_u16 (uint16x8_t, const int)
26039 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
26041 * int32x2_t vrshrn_n_s64 (int64x2_t, const int)
26042 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
26044 * int16x4_t vrshrn_n_s32 (int32x4_t, const int)
26045 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
26047 * int8x8_t vrshrn_n_s16 (int16x8_t, const int)
26048 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
26050 * uint32x2_t vqshrn_n_u64 (uint64x2_t, const int)
26051 _Form of expected instruction(s):_ `vqshrn.u64 D0, Q0, #0'
26053 * uint16x4_t vqshrn_n_u32 (uint32x4_t, const int)
26054 _Form of expected instruction(s):_ `vqshrn.u32 D0, Q0, #0'
26056 * uint8x8_t vqshrn_n_u16 (uint16x8_t, const int)
26057 _Form of expected instruction(s):_ `vqshrn.u16 D0, Q0, #0'
26059 * int32x2_t vqshrn_n_s64 (int64x2_t, const int)
26060 _Form of expected instruction(s):_ `vqshrn.s64 D0, Q0, #0'
26062 * int16x4_t vqshrn_n_s32 (int32x4_t, const int)
26063 _Form of expected instruction(s):_ `vqshrn.s32 D0, Q0, #0'
26065 * int8x8_t vqshrn_n_s16 (int16x8_t, const int)
26066 _Form of expected instruction(s):_ `vqshrn.s16 D0, Q0, #0'
26068 * uint32x2_t vqrshrn_n_u64 (uint64x2_t, const int)
26069 _Form of expected instruction(s):_ `vqrshrn.u64 D0, Q0, #0'
26071 * uint16x4_t vqrshrn_n_u32 (uint32x4_t, const int)
26072 _Form of expected instruction(s):_ `vqrshrn.u32 D0, Q0, #0'
26074 * uint8x8_t vqrshrn_n_u16 (uint16x8_t, const int)
26075 _Form of expected instruction(s):_ `vqrshrn.u16 D0, Q0, #0'
26077 * int32x2_t vqrshrn_n_s64 (int64x2_t, const int)
26078 _Form of expected instruction(s):_ `vqrshrn.s64 D0, Q0, #0'
26080 * int16x4_t vqrshrn_n_s32 (int32x4_t, const int)
26081 _Form of expected instruction(s):_ `vqrshrn.s32 D0, Q0, #0'
26083 * int8x8_t vqrshrn_n_s16 (int16x8_t, const int)
26084 _Form of expected instruction(s):_ `vqrshrn.s16 D0, Q0, #0'
26086 * uint32x2_t vqshrun_n_s64 (int64x2_t, const int)
26087 _Form of expected instruction(s):_ `vqshrun.s64 D0, Q0, #0'
26089 * uint16x4_t vqshrun_n_s32 (int32x4_t, const int)
26090 _Form of expected instruction(s):_ `vqshrun.s32 D0, Q0, #0'
26092 * uint8x8_t vqshrun_n_s16 (int16x8_t, const int)
26093 _Form of expected instruction(s):_ `vqshrun.s16 D0, Q0, #0'
26095 * uint32x2_t vqrshrun_n_s64 (int64x2_t, const int)
26096 _Form of expected instruction(s):_ `vqrshrun.s64 D0, Q0, #0'
26098 * uint16x4_t vqrshrun_n_s32 (int32x4_t, const int)
26099 _Form of expected instruction(s):_ `vqrshrun.s32 D0, Q0, #0'
26101 * uint8x8_t vqrshrun_n_s16 (int16x8_t, const int)
26102 _Form of expected instruction(s):_ `vqrshrun.s16 D0, Q0, #0'
26104 5.50.3.28 Vector shift right by constant and accumulate
26105 .......................................................
26107 * uint32x2_t vsra_n_u32 (uint32x2_t, uint32x2_t, const int)
26108 _Form of expected instruction(s):_ `vsra.u32 D0, D0, #0'
26110 * uint16x4_t vsra_n_u16 (uint16x4_t, uint16x4_t, const int)
26111 _Form of expected instruction(s):_ `vsra.u16 D0, D0, #0'
26113 * uint8x8_t vsra_n_u8 (uint8x8_t, uint8x8_t, const int)
26114 _Form of expected instruction(s):_ `vsra.u8 D0, D0, #0'
26116 * int32x2_t vsra_n_s32 (int32x2_t, int32x2_t, const int)
26117 _Form of expected instruction(s):_ `vsra.s32 D0, D0, #0'
26119 * int16x4_t vsra_n_s16 (int16x4_t, int16x4_t, const int)
26120 _Form of expected instruction(s):_ `vsra.s16 D0, D0, #0'
26122 * int8x8_t vsra_n_s8 (int8x8_t, int8x8_t, const int)
26123 _Form of expected instruction(s):_ `vsra.s8 D0, D0, #0'
26125 * uint64x1_t vsra_n_u64 (uint64x1_t, uint64x1_t, const int)
26126 _Form of expected instruction(s):_ `vsra.u64 D0, D0, #0'
26128 * int64x1_t vsra_n_s64 (int64x1_t, int64x1_t, const int)
26129 _Form of expected instruction(s):_ `vsra.s64 D0, D0, #0'
26131 * uint32x4_t vsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
26132 _Form of expected instruction(s):_ `vsra.u32 Q0, Q0, #0'
26134 * uint16x8_t vsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
26135 _Form of expected instruction(s):_ `vsra.u16 Q0, Q0, #0'
26137 * uint8x16_t vsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
26138 _Form of expected instruction(s):_ `vsra.u8 Q0, Q0, #0'
26140 * int32x4_t vsraq_n_s32 (int32x4_t, int32x4_t, const int)
26141 _Form of expected instruction(s):_ `vsra.s32 Q0, Q0, #0'
26143 * int16x8_t vsraq_n_s16 (int16x8_t, int16x8_t, const int)
26144 _Form of expected instruction(s):_ `vsra.s16 Q0, Q0, #0'
26146 * int8x16_t vsraq_n_s8 (int8x16_t, int8x16_t, const int)
26147 _Form of expected instruction(s):_ `vsra.s8 Q0, Q0, #0'
26149 * uint64x2_t vsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
26150 _Form of expected instruction(s):_ `vsra.u64 Q0, Q0, #0'
26152 * int64x2_t vsraq_n_s64 (int64x2_t, int64x2_t, const int)
26153 _Form of expected instruction(s):_ `vsra.s64 Q0, Q0, #0'
26155 * uint32x2_t vrsra_n_u32 (uint32x2_t, uint32x2_t, const int)
26156 _Form of expected instruction(s):_ `vrsra.u32 D0, D0, #0'
26158 * uint16x4_t vrsra_n_u16 (uint16x4_t, uint16x4_t, const int)
26159 _Form of expected instruction(s):_ `vrsra.u16 D0, D0, #0'
26161 * uint8x8_t vrsra_n_u8 (uint8x8_t, uint8x8_t, const int)
26162 _Form of expected instruction(s):_ `vrsra.u8 D0, D0, #0'
26164 * int32x2_t vrsra_n_s32 (int32x2_t, int32x2_t, const int)
26165 _Form of expected instruction(s):_ `vrsra.s32 D0, D0, #0'
26167 * int16x4_t vrsra_n_s16 (int16x4_t, int16x4_t, const int)
26168 _Form of expected instruction(s):_ `vrsra.s16 D0, D0, #0'
26170 * int8x8_t vrsra_n_s8 (int8x8_t, int8x8_t, const int)
26171 _Form of expected instruction(s):_ `vrsra.s8 D0, D0, #0'
26173 * uint64x1_t vrsra_n_u64 (uint64x1_t, uint64x1_t, const int)
26174 _Form of expected instruction(s):_ `vrsra.u64 D0, D0, #0'
26176 * int64x1_t vrsra_n_s64 (int64x1_t, int64x1_t, const int)
26177 _Form of expected instruction(s):_ `vrsra.s64 D0, D0, #0'
26179 * uint32x4_t vrsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
26180 _Form of expected instruction(s):_ `vrsra.u32 Q0, Q0, #0'
26182 * uint16x8_t vrsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
26183 _Form of expected instruction(s):_ `vrsra.u16 Q0, Q0, #0'
26185 * uint8x16_t vrsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
26186 _Form of expected instruction(s):_ `vrsra.u8 Q0, Q0, #0'
26188 * int32x4_t vrsraq_n_s32 (int32x4_t, int32x4_t, const int)
26189 _Form of expected instruction(s):_ `vrsra.s32 Q0, Q0, #0'
26191 * int16x8_t vrsraq_n_s16 (int16x8_t, int16x8_t, const int)
26192 _Form of expected instruction(s):_ `vrsra.s16 Q0, Q0, #0'
26194 * int8x16_t vrsraq_n_s8 (int8x16_t, int8x16_t, const int)
26195 _Form of expected instruction(s):_ `vrsra.s8 Q0, Q0, #0'
26197 * uint64x2_t vrsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
26198 _Form of expected instruction(s):_ `vrsra.u64 Q0, Q0, #0'
26200 * int64x2_t vrsraq_n_s64 (int64x2_t, int64x2_t, const int)
26201 _Form of expected instruction(s):_ `vrsra.s64 Q0, Q0, #0'
26203 5.50.3.29 Vector shift right and insert
26204 .......................................
26206 * uint32x2_t vsri_n_u32 (uint32x2_t, uint32x2_t, const int)
26207 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
26209 * uint16x4_t vsri_n_u16 (uint16x4_t, uint16x4_t, const int)
26210 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26212 * uint8x8_t vsri_n_u8 (uint8x8_t, uint8x8_t, const int)
26213 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26215 * int32x2_t vsri_n_s32 (int32x2_t, int32x2_t, const int)
26216 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
26218 * int16x4_t vsri_n_s16 (int16x4_t, int16x4_t, const int)
26219 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26221 * int8x8_t vsri_n_s8 (int8x8_t, int8x8_t, const int)
26222 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26224 * uint64x1_t vsri_n_u64 (uint64x1_t, uint64x1_t, const int)
26225 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
26227 * int64x1_t vsri_n_s64 (int64x1_t, int64x1_t, const int)
26228 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
26230 * poly16x4_t vsri_n_p16 (poly16x4_t, poly16x4_t, const int)
26231 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26233 * poly8x8_t vsri_n_p8 (poly8x8_t, poly8x8_t, const int)
26234 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26236 * uint32x4_t vsriq_n_u32 (uint32x4_t, uint32x4_t, const int)
26237 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
26239 * uint16x8_t vsriq_n_u16 (uint16x8_t, uint16x8_t, const int)
26240 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26242 * uint8x16_t vsriq_n_u8 (uint8x16_t, uint8x16_t, const int)
26243 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26245 * int32x4_t vsriq_n_s32 (int32x4_t, int32x4_t, const int)
26246 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
26248 * int16x8_t vsriq_n_s16 (int16x8_t, int16x8_t, const int)
26249 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26251 * int8x16_t vsriq_n_s8 (int8x16_t, int8x16_t, const int)
26252 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26254 * uint64x2_t vsriq_n_u64 (uint64x2_t, uint64x2_t, const int)
26255 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
26257 * int64x2_t vsriq_n_s64 (int64x2_t, int64x2_t, const int)
26258 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
26260 * poly16x8_t vsriq_n_p16 (poly16x8_t, poly16x8_t, const int)
26261 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26263 * poly8x16_t vsriq_n_p8 (poly8x16_t, poly8x16_t, const int)
26264 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26266 5.50.3.30 Vector shift left and insert
26267 ......................................
26269 * uint32x2_t vsli_n_u32 (uint32x2_t, uint32x2_t, const int)
26270 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
26272 * uint16x4_t vsli_n_u16 (uint16x4_t, uint16x4_t, const int)
26273 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26275 * uint8x8_t vsli_n_u8 (uint8x8_t, uint8x8_t, const int)
26276 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26278 * int32x2_t vsli_n_s32 (int32x2_t, int32x2_t, const int)
26279 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
26281 * int16x4_t vsli_n_s16 (int16x4_t, int16x4_t, const int)
26282 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26284 * int8x8_t vsli_n_s8 (int8x8_t, int8x8_t, const int)
26285 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26287 * uint64x1_t vsli_n_u64 (uint64x1_t, uint64x1_t, const int)
26288 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
26290 * int64x1_t vsli_n_s64 (int64x1_t, int64x1_t, const int)
26291 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
26293 * poly16x4_t vsli_n_p16 (poly16x4_t, poly16x4_t, const int)
26294 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26296 * poly8x8_t vsli_n_p8 (poly8x8_t, poly8x8_t, const int)
26297 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26299 * uint32x4_t vsliq_n_u32 (uint32x4_t, uint32x4_t, const int)
26300 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
26302 * uint16x8_t vsliq_n_u16 (uint16x8_t, uint16x8_t, const int)
26303 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26305 * uint8x16_t vsliq_n_u8 (uint8x16_t, uint8x16_t, const int)
26306 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26308 * int32x4_t vsliq_n_s32 (int32x4_t, int32x4_t, const int)
26309 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
26311 * int16x8_t vsliq_n_s16 (int16x8_t, int16x8_t, const int)
26312 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26314 * int8x16_t vsliq_n_s8 (int8x16_t, int8x16_t, const int)
26315 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26317 * uint64x2_t vsliq_n_u64 (uint64x2_t, uint64x2_t, const int)
26318 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
26320 * int64x2_t vsliq_n_s64 (int64x2_t, int64x2_t, const int)
26321 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
26323 * poly16x8_t vsliq_n_p16 (poly16x8_t, poly16x8_t, const int)
26324 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26326 * poly8x16_t vsliq_n_p8 (poly8x16_t, poly8x16_t, const int)
26327 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26329 5.50.3.31 Absolute value
26330 ........................
26332 * float32x2_t vabs_f32 (float32x2_t)
26333 _Form of expected instruction(s):_ `vabs.f32 D0, D0'
26335 * int32x2_t vabs_s32 (int32x2_t)
26336 _Form of expected instruction(s):_ `vabs.s32 D0, D0'
26338 * int16x4_t vabs_s16 (int16x4_t)
26339 _Form of expected instruction(s):_ `vabs.s16 D0, D0'
26341 * int8x8_t vabs_s8 (int8x8_t)
26342 _Form of expected instruction(s):_ `vabs.s8 D0, D0'
26344 * float32x4_t vabsq_f32 (float32x4_t)
26345 _Form of expected instruction(s):_ `vabs.f32 Q0, Q0'
26347 * int32x4_t vabsq_s32 (int32x4_t)
26348 _Form of expected instruction(s):_ `vabs.s32 Q0, Q0'
26350 * int16x8_t vabsq_s16 (int16x8_t)
26351 _Form of expected instruction(s):_ `vabs.s16 Q0, Q0'
26353 * int8x16_t vabsq_s8 (int8x16_t)
26354 _Form of expected instruction(s):_ `vabs.s8 Q0, Q0'
26356 * int32x2_t vqabs_s32 (int32x2_t)
26357 _Form of expected instruction(s):_ `vqabs.s32 D0, D0'
26359 * int16x4_t vqabs_s16 (int16x4_t)
26360 _Form of expected instruction(s):_ `vqabs.s16 D0, D0'
26362 * int8x8_t vqabs_s8 (int8x8_t)
26363 _Form of expected instruction(s):_ `vqabs.s8 D0, D0'
26365 * int32x4_t vqabsq_s32 (int32x4_t)
26366 _Form of expected instruction(s):_ `vqabs.s32 Q0, Q0'
26368 * int16x8_t vqabsq_s16 (int16x8_t)
26369 _Form of expected instruction(s):_ `vqabs.s16 Q0, Q0'
26371 * int8x16_t vqabsq_s8 (int8x16_t)
26372 _Form of expected instruction(s):_ `vqabs.s8 Q0, Q0'
26377 * float32x2_t vneg_f32 (float32x2_t)
26378 _Form of expected instruction(s):_ `vneg.f32 D0, D0'
26380 * int32x2_t vneg_s32 (int32x2_t)
26381 _Form of expected instruction(s):_ `vneg.s32 D0, D0'
26383 * int16x4_t vneg_s16 (int16x4_t)
26384 _Form of expected instruction(s):_ `vneg.s16 D0, D0'
26386 * int8x8_t vneg_s8 (int8x8_t)
26387 _Form of expected instruction(s):_ `vneg.s8 D0, D0'
26389 * float32x4_t vnegq_f32 (float32x4_t)
26390 _Form of expected instruction(s):_ `vneg.f32 Q0, Q0'
26392 * int32x4_t vnegq_s32 (int32x4_t)
26393 _Form of expected instruction(s):_ `vneg.s32 Q0, Q0'
26395 * int16x8_t vnegq_s16 (int16x8_t)
26396 _Form of expected instruction(s):_ `vneg.s16 Q0, Q0'
26398 * int8x16_t vnegq_s8 (int8x16_t)
26399 _Form of expected instruction(s):_ `vneg.s8 Q0, Q0'
26401 * int32x2_t vqneg_s32 (int32x2_t)
26402 _Form of expected instruction(s):_ `vqneg.s32 D0, D0'
26404 * int16x4_t vqneg_s16 (int16x4_t)
26405 _Form of expected instruction(s):_ `vqneg.s16 D0, D0'
26407 * int8x8_t vqneg_s8 (int8x8_t)
26408 _Form of expected instruction(s):_ `vqneg.s8 D0, D0'
26410 * int32x4_t vqnegq_s32 (int32x4_t)
26411 _Form of expected instruction(s):_ `vqneg.s32 Q0, Q0'
26413 * int16x8_t vqnegq_s16 (int16x8_t)
26414 _Form of expected instruction(s):_ `vqneg.s16 Q0, Q0'
26416 * int8x16_t vqnegq_s8 (int8x16_t)
26417 _Form of expected instruction(s):_ `vqneg.s8 Q0, Q0'
26419 5.50.3.33 Bitwise not
26420 .....................
26422 * uint32x2_t vmvn_u32 (uint32x2_t)
26423 _Form of expected instruction(s):_ `vmvn D0, D0'
26425 * uint16x4_t vmvn_u16 (uint16x4_t)
26426 _Form of expected instruction(s):_ `vmvn D0, D0'
26428 * uint8x8_t vmvn_u8 (uint8x8_t)
26429 _Form of expected instruction(s):_ `vmvn D0, D0'
26431 * int32x2_t vmvn_s32 (int32x2_t)
26432 _Form of expected instruction(s):_ `vmvn D0, D0'
26434 * int16x4_t vmvn_s16 (int16x4_t)
26435 _Form of expected instruction(s):_ `vmvn D0, D0'
26437 * int8x8_t vmvn_s8 (int8x8_t)
26438 _Form of expected instruction(s):_ `vmvn D0, D0'
26440 * poly8x8_t vmvn_p8 (poly8x8_t)
26441 _Form of expected instruction(s):_ `vmvn D0, D0'
26443 * uint32x4_t vmvnq_u32 (uint32x4_t)
26444 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26446 * uint16x8_t vmvnq_u16 (uint16x8_t)
26447 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26449 * uint8x16_t vmvnq_u8 (uint8x16_t)
26450 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26452 * int32x4_t vmvnq_s32 (int32x4_t)
26453 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26455 * int16x8_t vmvnq_s16 (int16x8_t)
26456 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26458 * int8x16_t vmvnq_s8 (int8x16_t)
26459 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26461 * poly8x16_t vmvnq_p8 (poly8x16_t)
26462 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26464 5.50.3.34 Count leading sign bits
26465 .................................
26467 * int32x2_t vcls_s32 (int32x2_t)
26468 _Form of expected instruction(s):_ `vcls.s32 D0, D0'
26470 * int16x4_t vcls_s16 (int16x4_t)
26471 _Form of expected instruction(s):_ `vcls.s16 D0, D0'
26473 * int8x8_t vcls_s8 (int8x8_t)
26474 _Form of expected instruction(s):_ `vcls.s8 D0, D0'
26476 * int32x4_t vclsq_s32 (int32x4_t)
26477 _Form of expected instruction(s):_ `vcls.s32 Q0, Q0'
26479 * int16x8_t vclsq_s16 (int16x8_t)
26480 _Form of expected instruction(s):_ `vcls.s16 Q0, Q0'
26482 * int8x16_t vclsq_s8 (int8x16_t)
26483 _Form of expected instruction(s):_ `vcls.s8 Q0, Q0'
26485 5.50.3.35 Count leading zeros
26486 .............................
26488 * uint32x2_t vclz_u32 (uint32x2_t)
26489 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
26491 * uint16x4_t vclz_u16 (uint16x4_t)
26492 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
26494 * uint8x8_t vclz_u8 (uint8x8_t)
26495 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
26497 * int32x2_t vclz_s32 (int32x2_t)
26498 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
26500 * int16x4_t vclz_s16 (int16x4_t)
26501 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
26503 * int8x8_t vclz_s8 (int8x8_t)
26504 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
26506 * uint32x4_t vclzq_u32 (uint32x4_t)
26507 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
26509 * uint16x8_t vclzq_u16 (uint16x8_t)
26510 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
26512 * uint8x16_t vclzq_u8 (uint8x16_t)
26513 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
26515 * int32x4_t vclzq_s32 (int32x4_t)
26516 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
26518 * int16x8_t vclzq_s16 (int16x8_t)
26519 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
26521 * int8x16_t vclzq_s8 (int8x16_t)
26522 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
26524 5.50.3.36 Count number of set bits
26525 ..................................
26527 * uint8x8_t vcnt_u8 (uint8x8_t)
26528 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26530 * int8x8_t vcnt_s8 (int8x8_t)
26531 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26533 * poly8x8_t vcnt_p8 (poly8x8_t)
26534 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26536 * uint8x16_t vcntq_u8 (uint8x16_t)
26537 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26539 * int8x16_t vcntq_s8 (int8x16_t)
26540 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26542 * poly8x16_t vcntq_p8 (poly8x16_t)
26543 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26545 5.50.3.37 Reciprocal estimate
26546 .............................
26548 * float32x2_t vrecpe_f32 (float32x2_t)
26549 _Form of expected instruction(s):_ `vrecpe.f32 D0, D0'
26551 * uint32x2_t vrecpe_u32 (uint32x2_t)
26552 _Form of expected instruction(s):_ `vrecpe.u32 D0, D0'
26554 * float32x4_t vrecpeq_f32 (float32x4_t)
26555 _Form of expected instruction(s):_ `vrecpe.f32 Q0, Q0'
26557 * uint32x4_t vrecpeq_u32 (uint32x4_t)
26558 _Form of expected instruction(s):_ `vrecpe.u32 Q0, Q0'
26560 5.50.3.38 Reciprocal square-root estimate
26561 .........................................
26563 * float32x2_t vrsqrte_f32 (float32x2_t)
26564 _Form of expected instruction(s):_ `vrsqrte.f32 D0, D0'
26566 * uint32x2_t vrsqrte_u32 (uint32x2_t)
26567 _Form of expected instruction(s):_ `vrsqrte.u32 D0, D0'
26569 * float32x4_t vrsqrteq_f32 (float32x4_t)
26570 _Form of expected instruction(s):_ `vrsqrte.f32 Q0, Q0'
26572 * uint32x4_t vrsqrteq_u32 (uint32x4_t)
26573 _Form of expected instruction(s):_ `vrsqrte.u32 Q0, Q0'
26575 5.50.3.39 Get lanes from a vector
26576 .................................
26578 * uint32_t vget_lane_u32 (uint32x2_t, const int)
26579 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
26581 * uint16_t vget_lane_u16 (uint16x4_t, const int)
26582 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26584 * uint8_t vget_lane_u8 (uint8x8_t, const int)
26585 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26587 * int32_t vget_lane_s32 (int32x2_t, const int)
26588 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
26590 * int16_t vget_lane_s16 (int16x4_t, const int)
26591 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
26593 * int8_t vget_lane_s8 (int8x8_t, const int)
26594 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
26596 * float32_t vget_lane_f32 (float32x2_t, const int)
26597 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
26599 * poly16_t vget_lane_p16 (poly16x4_t, const int)
26600 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26602 * poly8_t vget_lane_p8 (poly8x8_t, const int)
26603 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26605 * uint64_t vget_lane_u64 (uint64x1_t, const int)
26606 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26608 * int64_t vget_lane_s64 (int64x1_t, const int)
26609 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26611 * uint32_t vgetq_lane_u32 (uint32x4_t, const int)
26612 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
26614 * uint16_t vgetq_lane_u16 (uint16x8_t, const int)
26615 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26617 * uint8_t vgetq_lane_u8 (uint8x16_t, const int)
26618 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26620 * int32_t vgetq_lane_s32 (int32x4_t, const int)
26621 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
26623 * int16_t vgetq_lane_s16 (int16x8_t, const int)
26624 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
26626 * int8_t vgetq_lane_s8 (int8x16_t, const int)
26627 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
26629 * float32_t vgetq_lane_f32 (float32x4_t, const int)
26630 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
26632 * poly16_t vgetq_lane_p16 (poly16x8_t, const int)
26633 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26635 * poly8_t vgetq_lane_p8 (poly8x16_t, const int)
26636 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26638 * uint64_t vgetq_lane_u64 (uint64x2_t, const int)
26639 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26641 * int64_t vgetq_lane_s64 (int64x2_t, const int)
26642 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26644 5.50.3.40 Set lanes in a vector
26645 ...............................
26647 * uint32x2_t vset_lane_u32 (uint32_t, uint32x2_t, const int)
26648 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26650 * uint16x4_t vset_lane_u16 (uint16_t, uint16x4_t, const int)
26651 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26653 * uint8x8_t vset_lane_u8 (uint8_t, uint8x8_t, const int)
26654 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26656 * int32x2_t vset_lane_s32 (int32_t, int32x2_t, const int)
26657 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26659 * int16x4_t vset_lane_s16 (int16_t, int16x4_t, const int)
26660 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26662 * int8x8_t vset_lane_s8 (int8_t, int8x8_t, const int)
26663 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26665 * float32x2_t vset_lane_f32 (float32_t, float32x2_t, const int)
26666 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26668 * poly16x4_t vset_lane_p16 (poly16_t, poly16x4_t, const int)
26669 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26671 * poly8x8_t vset_lane_p8 (poly8_t, poly8x8_t, const int)
26672 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26674 * uint64x1_t vset_lane_u64 (uint64_t, uint64x1_t, const int)
26675 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26677 * int64x1_t vset_lane_s64 (int64_t, int64x1_t, const int)
26678 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26680 * uint32x4_t vsetq_lane_u32 (uint32_t, uint32x4_t, const int)
26681 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26683 * uint16x8_t vsetq_lane_u16 (uint16_t, uint16x8_t, const int)
26684 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26686 * uint8x16_t vsetq_lane_u8 (uint8_t, uint8x16_t, const int)
26687 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26689 * int32x4_t vsetq_lane_s32 (int32_t, int32x4_t, const int)
26690 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26692 * int16x8_t vsetq_lane_s16 (int16_t, int16x8_t, const int)
26693 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26695 * int8x16_t vsetq_lane_s8 (int8_t, int8x16_t, const int)
26696 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26698 * float32x4_t vsetq_lane_f32 (float32_t, float32x4_t, const int)
26699 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26701 * poly16x8_t vsetq_lane_p16 (poly16_t, poly16x8_t, const int)
26702 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26704 * poly8x16_t vsetq_lane_p8 (poly8_t, poly8x16_t, const int)
26705 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26707 * uint64x2_t vsetq_lane_u64 (uint64_t, uint64x2_t, const int)
26708 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26710 * int64x2_t vsetq_lane_s64 (int64_t, int64x2_t, const int)
26711 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26713 5.50.3.41 Create vector from literal bit pattern
26714 ................................................
26716 * uint32x2_t vcreate_u32 (uint64_t)
26718 * uint16x4_t vcreate_u16 (uint64_t)
26720 * uint8x8_t vcreate_u8 (uint64_t)
26722 * int32x2_t vcreate_s32 (uint64_t)
26724 * int16x4_t vcreate_s16 (uint64_t)
26726 * int8x8_t vcreate_s8 (uint64_t)
26728 * uint64x1_t vcreate_u64 (uint64_t)
26730 * int64x1_t vcreate_s64 (uint64_t)
26732 * float32x2_t vcreate_f32 (uint64_t)
26734 * poly16x4_t vcreate_p16 (uint64_t)
26736 * poly8x8_t vcreate_p8 (uint64_t)
26738 5.50.3.42 Set all lanes to the same value
26739 .........................................
26741 * uint32x2_t vdup_n_u32 (uint32_t)
26742 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26744 * uint16x4_t vdup_n_u16 (uint16_t)
26745 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26747 * uint8x8_t vdup_n_u8 (uint8_t)
26748 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26750 * int32x2_t vdup_n_s32 (int32_t)
26751 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26753 * int16x4_t vdup_n_s16 (int16_t)
26754 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26756 * int8x8_t vdup_n_s8 (int8_t)
26757 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26759 * float32x2_t vdup_n_f32 (float32_t)
26760 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26762 * poly16x4_t vdup_n_p16 (poly16_t)
26763 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26765 * poly8x8_t vdup_n_p8 (poly8_t)
26766 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26768 * uint64x1_t vdup_n_u64 (uint64_t)
26769 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26771 * int64x1_t vdup_n_s64 (int64_t)
26772 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26774 * uint32x4_t vdupq_n_u32 (uint32_t)
26775 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26777 * uint16x8_t vdupq_n_u16 (uint16_t)
26778 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26780 * uint8x16_t vdupq_n_u8 (uint8_t)
26781 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26783 * int32x4_t vdupq_n_s32 (int32_t)
26784 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26786 * int16x8_t vdupq_n_s16 (int16_t)
26787 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26789 * int8x16_t vdupq_n_s8 (int8_t)
26790 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26792 * float32x4_t vdupq_n_f32 (float32_t)
26793 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26795 * poly16x8_t vdupq_n_p16 (poly16_t)
26796 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26798 * poly8x16_t vdupq_n_p8 (poly8_t)
26799 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26801 * uint64x2_t vdupq_n_u64 (uint64_t)
26802 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26804 * int64x2_t vdupq_n_s64 (int64_t)
26805 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26807 * uint32x2_t vmov_n_u32 (uint32_t)
26808 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26810 * uint16x4_t vmov_n_u16 (uint16_t)
26811 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26813 * uint8x8_t vmov_n_u8 (uint8_t)
26814 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26816 * int32x2_t vmov_n_s32 (int32_t)
26817 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26819 * int16x4_t vmov_n_s16 (int16_t)
26820 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26822 * int8x8_t vmov_n_s8 (int8_t)
26823 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26825 * float32x2_t vmov_n_f32 (float32_t)
26826 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26828 * poly16x4_t vmov_n_p16 (poly16_t)
26829 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26831 * poly8x8_t vmov_n_p8 (poly8_t)
26832 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26834 * uint64x1_t vmov_n_u64 (uint64_t)
26835 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26837 * int64x1_t vmov_n_s64 (int64_t)
26838 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26840 * uint32x4_t vmovq_n_u32 (uint32_t)
26841 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26843 * uint16x8_t vmovq_n_u16 (uint16_t)
26844 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26846 * uint8x16_t vmovq_n_u8 (uint8_t)
26847 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26849 * int32x4_t vmovq_n_s32 (int32_t)
26850 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26852 * int16x8_t vmovq_n_s16 (int16_t)
26853 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26855 * int8x16_t vmovq_n_s8 (int8_t)
26856 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26858 * float32x4_t vmovq_n_f32 (float32_t)
26859 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26861 * poly16x8_t vmovq_n_p16 (poly16_t)
26862 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26864 * poly8x16_t vmovq_n_p8 (poly8_t)
26865 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26867 * uint64x2_t vmovq_n_u64 (uint64_t)
26868 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26870 * int64x2_t vmovq_n_s64 (int64_t)
26871 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26873 * uint32x2_t vdup_lane_u32 (uint32x2_t, const int)
26874 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26876 * uint16x4_t vdup_lane_u16 (uint16x4_t, const int)
26877 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26879 * uint8x8_t vdup_lane_u8 (uint8x8_t, const int)
26880 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26882 * int32x2_t vdup_lane_s32 (int32x2_t, const int)
26883 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26885 * int16x4_t vdup_lane_s16 (int16x4_t, const int)
26886 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26888 * int8x8_t vdup_lane_s8 (int8x8_t, const int)
26889 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26891 * float32x2_t vdup_lane_f32 (float32x2_t, const int)
26892 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26894 * poly16x4_t vdup_lane_p16 (poly16x4_t, const int)
26895 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26897 * poly8x8_t vdup_lane_p8 (poly8x8_t, const int)
26898 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26900 * uint64x1_t vdup_lane_u64 (uint64x1_t, const int)
26902 * int64x1_t vdup_lane_s64 (int64x1_t, const int)
26904 * uint32x4_t vdupq_lane_u32 (uint32x2_t, const int)
26905 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26907 * uint16x8_t vdupq_lane_u16 (uint16x4_t, const int)
26908 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26910 * uint8x16_t vdupq_lane_u8 (uint8x8_t, const int)
26911 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26913 * int32x4_t vdupq_lane_s32 (int32x2_t, const int)
26914 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26916 * int16x8_t vdupq_lane_s16 (int16x4_t, const int)
26917 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26919 * int8x16_t vdupq_lane_s8 (int8x8_t, const int)
26920 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26922 * float32x4_t vdupq_lane_f32 (float32x2_t, const int)
26923 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26925 * poly16x8_t vdupq_lane_p16 (poly16x4_t, const int)
26926 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26928 * poly8x16_t vdupq_lane_p8 (poly8x8_t, const int)
26929 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26931 * uint64x2_t vdupq_lane_u64 (uint64x1_t, const int)
26933 * int64x2_t vdupq_lane_s64 (int64x1_t, const int)
26935 5.50.3.43 Combining vectors
26936 ...........................
26938 * uint32x4_t vcombine_u32 (uint32x2_t, uint32x2_t)
26940 * uint16x8_t vcombine_u16 (uint16x4_t, uint16x4_t)
26942 * uint8x16_t vcombine_u8 (uint8x8_t, uint8x8_t)
26944 * int32x4_t vcombine_s32 (int32x2_t, int32x2_t)
26946 * int16x8_t vcombine_s16 (int16x4_t, int16x4_t)
26948 * int8x16_t vcombine_s8 (int8x8_t, int8x8_t)
26950 * uint64x2_t vcombine_u64 (uint64x1_t, uint64x1_t)
26952 * int64x2_t vcombine_s64 (int64x1_t, int64x1_t)
26954 * float32x4_t vcombine_f32 (float32x2_t, float32x2_t)
26956 * poly16x8_t vcombine_p16 (poly16x4_t, poly16x4_t)
26958 * poly8x16_t vcombine_p8 (poly8x8_t, poly8x8_t)
26960 5.50.3.44 Splitting vectors
26961 ...........................
26963 * uint32x2_t vget_high_u32 (uint32x4_t)
26965 * uint16x4_t vget_high_u16 (uint16x8_t)
26967 * uint8x8_t vget_high_u8 (uint8x16_t)
26969 * int32x2_t vget_high_s32 (int32x4_t)
26971 * int16x4_t vget_high_s16 (int16x8_t)
26973 * int8x8_t vget_high_s8 (int8x16_t)
26975 * uint64x1_t vget_high_u64 (uint64x2_t)
26977 * int64x1_t vget_high_s64 (int64x2_t)
26979 * float32x2_t vget_high_f32 (float32x4_t)
26981 * poly16x4_t vget_high_p16 (poly16x8_t)
26983 * poly8x8_t vget_high_p8 (poly8x16_t)
26985 * uint32x2_t vget_low_u32 (uint32x4_t)
26986 _Form of expected instruction(s):_ `vmov D0, D0'
26988 * uint16x4_t vget_low_u16 (uint16x8_t)
26989 _Form of expected instruction(s):_ `vmov D0, D0'
26991 * uint8x8_t vget_low_u8 (uint8x16_t)
26992 _Form of expected instruction(s):_ `vmov D0, D0'
26994 * int32x2_t vget_low_s32 (int32x4_t)
26995 _Form of expected instruction(s):_ `vmov D0, D0'
26997 * int16x4_t vget_low_s16 (int16x8_t)
26998 _Form of expected instruction(s):_ `vmov D0, D0'
27000 * int8x8_t vget_low_s8 (int8x16_t)
27001 _Form of expected instruction(s):_ `vmov D0, D0'
27003 * uint64x1_t vget_low_u64 (uint64x2_t)
27004 _Form of expected instruction(s):_ `vmov D0, D0'
27006 * int64x1_t vget_low_s64 (int64x2_t)
27007 _Form of expected instruction(s):_ `vmov D0, D0'
27009 * float32x2_t vget_low_f32 (float32x4_t)
27010 _Form of expected instruction(s):_ `vmov D0, D0'
27012 * poly16x4_t vget_low_p16 (poly16x8_t)
27013 _Form of expected instruction(s):_ `vmov D0, D0'
27015 * poly8x8_t vget_low_p8 (poly8x16_t)
27016 _Form of expected instruction(s):_ `vmov D0, D0'
27018 5.50.3.45 Conversions
27019 .....................
27021 * float32x2_t vcvt_f32_u32 (uint32x2_t)
27022 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0'
27024 * float32x2_t vcvt_f32_s32 (int32x2_t)
27025 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0'
27027 * uint32x2_t vcvt_u32_f32 (float32x2_t)
27028 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0'
27030 * int32x2_t vcvt_s32_f32 (float32x2_t)
27031 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0'
27033 * float32x4_t vcvtq_f32_u32 (uint32x4_t)
27034 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0'
27036 * float32x4_t vcvtq_f32_s32 (int32x4_t)
27037 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0'
27039 * uint32x4_t vcvtq_u32_f32 (float32x4_t)
27040 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0'
27042 * int32x4_t vcvtq_s32_f32 (float32x4_t)
27043 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0'
27045 * float32x2_t vcvt_n_f32_u32 (uint32x2_t, const int)
27046 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0, #0'
27048 * float32x2_t vcvt_n_f32_s32 (int32x2_t, const int)
27049 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0, #0'
27051 * uint32x2_t vcvt_n_u32_f32 (float32x2_t, const int)
27052 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0, #0'
27054 * int32x2_t vcvt_n_s32_f32 (float32x2_t, const int)
27055 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0, #0'
27057 * float32x4_t vcvtq_n_f32_u32 (uint32x4_t, const int)
27058 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0, #0'
27060 * float32x4_t vcvtq_n_f32_s32 (int32x4_t, const int)
27061 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0, #0'
27063 * uint32x4_t vcvtq_n_u32_f32 (float32x4_t, const int)
27064 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0, #0'
27066 * int32x4_t vcvtq_n_s32_f32 (float32x4_t, const int)
27067 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0, #0'
27069 5.50.3.46 Move, single_opcode narrowing
27070 .......................................
27072 * uint32x2_t vmovn_u64 (uint64x2_t)
27073 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
27075 * uint16x4_t vmovn_u32 (uint32x4_t)
27076 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
27078 * uint8x8_t vmovn_u16 (uint16x8_t)
27079 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
27081 * int32x2_t vmovn_s64 (int64x2_t)
27082 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
27084 * int16x4_t vmovn_s32 (int32x4_t)
27085 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
27087 * int8x8_t vmovn_s16 (int16x8_t)
27088 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
27090 * uint32x2_t vqmovn_u64 (uint64x2_t)
27091 _Form of expected instruction(s):_ `vqmovn.u64 D0, Q0'
27093 * uint16x4_t vqmovn_u32 (uint32x4_t)
27094 _Form of expected instruction(s):_ `vqmovn.u32 D0, Q0'
27096 * uint8x8_t vqmovn_u16 (uint16x8_t)
27097 _Form of expected instruction(s):_ `vqmovn.u16 D0, Q0'
27099 * int32x2_t vqmovn_s64 (int64x2_t)
27100 _Form of expected instruction(s):_ `vqmovn.s64 D0, Q0'
27102 * int16x4_t vqmovn_s32 (int32x4_t)
27103 _Form of expected instruction(s):_ `vqmovn.s32 D0, Q0'
27105 * int8x8_t vqmovn_s16 (int16x8_t)
27106 _Form of expected instruction(s):_ `vqmovn.s16 D0, Q0'
27108 * uint32x2_t vqmovun_s64 (int64x2_t)
27109 _Form of expected instruction(s):_ `vqmovun.s64 D0, Q0'
27111 * uint16x4_t vqmovun_s32 (int32x4_t)
27112 _Form of expected instruction(s):_ `vqmovun.s32 D0, Q0'
27114 * uint8x8_t vqmovun_s16 (int16x8_t)
27115 _Form of expected instruction(s):_ `vqmovun.s16 D0, Q0'
27117 5.50.3.47 Move, single_opcode long
27118 ..................................
27120 * uint64x2_t vmovl_u32 (uint32x2_t)
27121 _Form of expected instruction(s):_ `vmovl.u32 Q0, D0'
27123 * uint32x4_t vmovl_u16 (uint16x4_t)
27124 _Form of expected instruction(s):_ `vmovl.u16 Q0, D0'
27126 * uint16x8_t vmovl_u8 (uint8x8_t)
27127 _Form of expected instruction(s):_ `vmovl.u8 Q0, D0'
27129 * int64x2_t vmovl_s32 (int32x2_t)
27130 _Form of expected instruction(s):_ `vmovl.s32 Q0, D0'
27132 * int32x4_t vmovl_s16 (int16x4_t)
27133 _Form of expected instruction(s):_ `vmovl.s16 Q0, D0'
27135 * int16x8_t vmovl_s8 (int8x8_t)
27136 _Form of expected instruction(s):_ `vmovl.s8 Q0, D0'
27138 5.50.3.48 Table lookup
27139 ......................
27141 * poly8x8_t vtbl1_p8 (poly8x8_t, uint8x8_t)
27142 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27144 * int8x8_t vtbl1_s8 (int8x8_t, int8x8_t)
27145 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27147 * uint8x8_t vtbl1_u8 (uint8x8_t, uint8x8_t)
27148 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27150 * poly8x8_t vtbl2_p8 (poly8x8x2_t, uint8x8_t)
27151 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27153 * int8x8_t vtbl2_s8 (int8x8x2_t, int8x8_t)
27154 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27156 * uint8x8_t vtbl2_u8 (uint8x8x2_t, uint8x8_t)
27157 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27159 * poly8x8_t vtbl3_p8 (poly8x8x3_t, uint8x8_t)
27160 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27162 * int8x8_t vtbl3_s8 (int8x8x3_t, int8x8_t)
27163 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27165 * uint8x8_t vtbl3_u8 (uint8x8x3_t, uint8x8_t)
27166 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27168 * poly8x8_t vtbl4_p8 (poly8x8x4_t, uint8x8_t)
27169 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27172 * int8x8_t vtbl4_s8 (int8x8x4_t, int8x8_t)
27173 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27176 * uint8x8_t vtbl4_u8 (uint8x8x4_t, uint8x8_t)
27177 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27180 5.50.3.49 Extended table lookup
27181 ...............................
27183 * poly8x8_t vtbx1_p8 (poly8x8_t, poly8x8_t, uint8x8_t)
27184 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27186 * int8x8_t vtbx1_s8 (int8x8_t, int8x8_t, int8x8_t)
27187 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27189 * uint8x8_t vtbx1_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
27190 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27192 * poly8x8_t vtbx2_p8 (poly8x8_t, poly8x8x2_t, uint8x8_t)
27193 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27195 * int8x8_t vtbx2_s8 (int8x8_t, int8x8x2_t, int8x8_t)
27196 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27198 * uint8x8_t vtbx2_u8 (uint8x8_t, uint8x8x2_t, uint8x8_t)
27199 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27201 * poly8x8_t vtbx3_p8 (poly8x8_t, poly8x8x3_t, uint8x8_t)
27202 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27204 * int8x8_t vtbx3_s8 (int8x8_t, int8x8x3_t, int8x8_t)
27205 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27207 * uint8x8_t vtbx3_u8 (uint8x8_t, uint8x8x3_t, uint8x8_t)
27208 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27210 * poly8x8_t vtbx4_p8 (poly8x8_t, poly8x8x4_t, uint8x8_t)
27211 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27214 * int8x8_t vtbx4_s8 (int8x8_t, int8x8x4_t, int8x8_t)
27215 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27218 * uint8x8_t vtbx4_u8 (uint8x8_t, uint8x8x4_t, uint8x8_t)
27219 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27222 5.50.3.50 Multiply, lane
27223 ........................
27225 * float32x2_t vmul_lane_f32 (float32x2_t, float32x2_t, const int)
27226 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
27228 * uint32x2_t vmul_lane_u32 (uint32x2_t, uint32x2_t, const int)
27229 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27231 * uint16x4_t vmul_lane_u16 (uint16x4_t, uint16x4_t, const int)
27232 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27234 * int32x2_t vmul_lane_s32 (int32x2_t, int32x2_t, const int)
27235 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27237 * int16x4_t vmul_lane_s16 (int16x4_t, int16x4_t, const int)
27238 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27240 * float32x4_t vmulq_lane_f32 (float32x4_t, float32x2_t, const int)
27241 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
27243 * uint32x4_t vmulq_lane_u32 (uint32x4_t, uint32x2_t, const int)
27244 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27246 * uint16x8_t vmulq_lane_u16 (uint16x8_t, uint16x4_t, const int)
27247 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27249 * int32x4_t vmulq_lane_s32 (int32x4_t, int32x2_t, const int)
27250 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27252 * int16x8_t vmulq_lane_s16 (int16x8_t, int16x4_t, const int)
27253 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27255 5.50.3.51 Long multiply, lane
27256 .............................
27258 * uint64x2_t vmull_lane_u32 (uint32x2_t, uint32x2_t, const int)
27259 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
27261 * uint32x4_t vmull_lane_u16 (uint16x4_t, uint16x4_t, const int)
27262 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
27264 * int64x2_t vmull_lane_s32 (int32x2_t, int32x2_t, const int)
27265 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
27267 * int32x4_t vmull_lane_s16 (int16x4_t, int16x4_t, const int)
27268 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
27270 5.50.3.52 Saturating doubling long multiply, lane
27271 .................................................
27273 * int64x2_t vqdmull_lane_s32 (int32x2_t, int32x2_t, const int)
27274 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
27276 * int32x4_t vqdmull_lane_s16 (int16x4_t, int16x4_t, const int)
27277 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
27279 5.50.3.53 Saturating doubling multiply high, lane
27280 .................................................
27282 * int32x4_t vqdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
27283 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
27285 * int16x8_t vqdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
27286 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
27288 * int32x2_t vqdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
27289 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
27291 * int16x4_t vqdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
27292 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
27294 * int32x4_t vqrdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
27295 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
27297 * int16x8_t vqrdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
27298 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
27300 * int32x2_t vqrdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
27301 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
27303 * int16x4_t vqrdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
27304 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
27306 5.50.3.54 Multiply-accumulate, lane
27307 ...................................
27309 * float32x2_t vmla_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
27311 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
27313 * uint32x2_t vmla_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
27315 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27317 * uint16x4_t vmla_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
27319 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27321 * int32x2_t vmla_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
27323 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27325 * int16x4_t vmla_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
27327 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27329 * float32x4_t vmlaq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
27331 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
27333 * uint32x4_t vmlaq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
27335 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27337 * uint16x8_t vmlaq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
27339 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27341 * int32x4_t vmlaq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
27343 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27345 * int16x8_t vmlaq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
27347 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27349 * uint64x2_t vmlal_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
27351 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
27353 * uint32x4_t vmlal_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
27355 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
27357 * int64x2_t vmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27359 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
27361 * int32x4_t vmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27363 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
27365 * int64x2_t vqdmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27367 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
27369 * int32x4_t vqdmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27371 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
27373 5.50.3.55 Multiply-subtract, lane
27374 .................................
27376 * float32x2_t vmls_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
27378 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
27380 * uint32x2_t vmls_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
27382 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27384 * uint16x4_t vmls_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
27386 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27388 * int32x2_t vmls_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
27390 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27392 * int16x4_t vmls_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
27394 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27396 * float32x4_t vmlsq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
27398 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
27400 * uint32x4_t vmlsq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
27402 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27404 * uint16x8_t vmlsq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
27406 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27408 * int32x4_t vmlsq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
27410 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27412 * int16x8_t vmlsq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
27414 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27416 * uint64x2_t vmlsl_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
27418 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
27420 * uint32x4_t vmlsl_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
27422 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
27424 * int64x2_t vmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27426 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
27428 * int32x4_t vmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27430 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
27432 * int64x2_t vqdmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27434 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
27436 * int32x4_t vqdmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27438 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
27440 5.50.3.56 Vector multiply by scalar
27441 ...................................
27443 * float32x2_t vmul_n_f32 (float32x2_t, float32_t)
27444 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
27446 * uint32x2_t vmul_n_u32 (uint32x2_t, uint32_t)
27447 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27449 * uint16x4_t vmul_n_u16 (uint16x4_t, uint16_t)
27450 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27452 * int32x2_t vmul_n_s32 (int32x2_t, int32_t)
27453 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27455 * int16x4_t vmul_n_s16 (int16x4_t, int16_t)
27456 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27458 * float32x4_t vmulq_n_f32 (float32x4_t, float32_t)
27459 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
27461 * uint32x4_t vmulq_n_u32 (uint32x4_t, uint32_t)
27462 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27464 * uint16x8_t vmulq_n_u16 (uint16x8_t, uint16_t)
27465 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27467 * int32x4_t vmulq_n_s32 (int32x4_t, int32_t)
27468 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27470 * int16x8_t vmulq_n_s16 (int16x8_t, int16_t)
27471 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27473 5.50.3.57 Vector long multiply by scalar
27474 ........................................
27476 * uint64x2_t vmull_n_u32 (uint32x2_t, uint32_t)
27477 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
27479 * uint32x4_t vmull_n_u16 (uint16x4_t, uint16_t)
27480 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
27482 * int64x2_t vmull_n_s32 (int32x2_t, int32_t)
27483 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
27485 * int32x4_t vmull_n_s16 (int16x4_t, int16_t)
27486 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
27488 5.50.3.58 Vector saturating doubling long multiply by scalar
27489 ............................................................
27491 * int64x2_t vqdmull_n_s32 (int32x2_t, int32_t)
27492 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
27494 * int32x4_t vqdmull_n_s16 (int16x4_t, int16_t)
27495 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
27497 5.50.3.59 Vector saturating doubling multiply high by scalar
27498 ............................................................
27500 * int32x4_t vqdmulhq_n_s32 (int32x4_t, int32_t)
27501 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
27503 * int16x8_t vqdmulhq_n_s16 (int16x8_t, int16_t)
27504 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
27506 * int32x2_t vqdmulh_n_s32 (int32x2_t, int32_t)
27507 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
27509 * int16x4_t vqdmulh_n_s16 (int16x4_t, int16_t)
27510 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
27512 * int32x4_t vqrdmulhq_n_s32 (int32x4_t, int32_t)
27513 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
27515 * int16x8_t vqrdmulhq_n_s16 (int16x8_t, int16_t)
27516 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
27518 * int32x2_t vqrdmulh_n_s32 (int32x2_t, int32_t)
27519 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
27521 * int16x4_t vqrdmulh_n_s16 (int16x4_t, int16_t)
27522 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
27524 5.50.3.60 Vector multiply-accumulate by scalar
27525 ..............................................
27527 * float32x2_t vmla_n_f32 (float32x2_t, float32x2_t, float32_t)
27528 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
27530 * uint32x2_t vmla_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
27531 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27533 * uint16x4_t vmla_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
27534 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27536 * int32x2_t vmla_n_s32 (int32x2_t, int32x2_t, int32_t)
27537 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27539 * int16x4_t vmla_n_s16 (int16x4_t, int16x4_t, int16_t)
27540 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27542 * float32x4_t vmlaq_n_f32 (float32x4_t, float32x4_t, float32_t)
27543 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
27545 * uint32x4_t vmlaq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
27546 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27548 * uint16x8_t vmlaq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
27549 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27551 * int32x4_t vmlaq_n_s32 (int32x4_t, int32x4_t, int32_t)
27552 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27554 * int16x8_t vmlaq_n_s16 (int16x8_t, int16x8_t, int16_t)
27555 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27557 * uint64x2_t vmlal_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
27558 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
27560 * uint32x4_t vmlal_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
27561 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
27563 * int64x2_t vmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
27564 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
27566 * int32x4_t vmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
27567 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
27569 * int64x2_t vqdmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
27570 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
27572 * int32x4_t vqdmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
27573 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
27575 5.50.3.61 Vector multiply-subtract by scalar
27576 ............................................
27578 * float32x2_t vmls_n_f32 (float32x2_t, float32x2_t, float32_t)
27579 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
27581 * uint32x2_t vmls_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
27582 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27584 * uint16x4_t vmls_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
27585 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27587 * int32x2_t vmls_n_s32 (int32x2_t, int32x2_t, int32_t)
27588 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27590 * int16x4_t vmls_n_s16 (int16x4_t, int16x4_t, int16_t)
27591 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27593 * float32x4_t vmlsq_n_f32 (float32x4_t, float32x4_t, float32_t)
27594 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
27596 * uint32x4_t vmlsq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
27597 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27599 * uint16x8_t vmlsq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
27600 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27602 * int32x4_t vmlsq_n_s32 (int32x4_t, int32x4_t, int32_t)
27603 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27605 * int16x8_t vmlsq_n_s16 (int16x8_t, int16x8_t, int16_t)
27606 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27608 * uint64x2_t vmlsl_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
27609 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
27611 * uint32x4_t vmlsl_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
27612 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
27614 * int64x2_t vmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
27615 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
27617 * int32x4_t vmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
27618 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
27620 * int64x2_t vqdmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
27621 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
27623 * int32x4_t vqdmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
27624 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
27626 5.50.3.62 Vector extract
27627 ........................
27629 * uint32x2_t vext_u32 (uint32x2_t, uint32x2_t, const int)
27630 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27632 * uint16x4_t vext_u16 (uint16x4_t, uint16x4_t, const int)
27633 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27635 * uint8x8_t vext_u8 (uint8x8_t, uint8x8_t, const int)
27636 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27638 * int32x2_t vext_s32 (int32x2_t, int32x2_t, const int)
27639 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27641 * int16x4_t vext_s16 (int16x4_t, int16x4_t, const int)
27642 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27644 * int8x8_t vext_s8 (int8x8_t, int8x8_t, const int)
27645 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27647 * uint64x1_t vext_u64 (uint64x1_t, uint64x1_t, const int)
27648 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
27650 * int64x1_t vext_s64 (int64x1_t, int64x1_t, const int)
27651 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
27653 * float32x2_t vext_f32 (float32x2_t, float32x2_t, const int)
27654 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27656 * poly16x4_t vext_p16 (poly16x4_t, poly16x4_t, const int)
27657 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27659 * poly8x8_t vext_p8 (poly8x8_t, poly8x8_t, const int)
27660 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27662 * uint32x4_t vextq_u32 (uint32x4_t, uint32x4_t, const int)
27663 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27665 * uint16x8_t vextq_u16 (uint16x8_t, uint16x8_t, const int)
27666 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27668 * uint8x16_t vextq_u8 (uint8x16_t, uint8x16_t, const int)
27669 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27671 * int32x4_t vextq_s32 (int32x4_t, int32x4_t, const int)
27672 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27674 * int16x8_t vextq_s16 (int16x8_t, int16x8_t, const int)
27675 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27677 * int8x16_t vextq_s8 (int8x16_t, int8x16_t, const int)
27678 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27680 * uint64x2_t vextq_u64 (uint64x2_t, uint64x2_t, const int)
27681 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
27683 * int64x2_t vextq_s64 (int64x2_t, int64x2_t, const int)
27684 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
27686 * float32x4_t vextq_f32 (float32x4_t, float32x4_t, const int)
27687 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27689 * poly16x8_t vextq_p16 (poly16x8_t, poly16x8_t, const int)
27690 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27692 * poly8x16_t vextq_p8 (poly8x16_t, poly8x16_t, const int)
27693 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27695 5.50.3.63 Reverse elements
27696 ..........................
27698 * uint32x2_t vrev64_u32 (uint32x2_t)
27699 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27701 * uint16x4_t vrev64_u16 (uint16x4_t)
27702 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27704 * uint8x8_t vrev64_u8 (uint8x8_t)
27705 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27707 * int32x2_t vrev64_s32 (int32x2_t)
27708 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27710 * int16x4_t vrev64_s16 (int16x4_t)
27711 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27713 * int8x8_t vrev64_s8 (int8x8_t)
27714 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27716 * float32x2_t vrev64_f32 (float32x2_t)
27717 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27719 * poly16x4_t vrev64_p16 (poly16x4_t)
27720 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27722 * poly8x8_t vrev64_p8 (poly8x8_t)
27723 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27725 * uint32x4_t vrev64q_u32 (uint32x4_t)
27726 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27728 * uint16x8_t vrev64q_u16 (uint16x8_t)
27729 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27731 * uint8x16_t vrev64q_u8 (uint8x16_t)
27732 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27734 * int32x4_t vrev64q_s32 (int32x4_t)
27735 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27737 * int16x8_t vrev64q_s16 (int16x8_t)
27738 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27740 * int8x16_t vrev64q_s8 (int8x16_t)
27741 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27743 * float32x4_t vrev64q_f32 (float32x4_t)
27744 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27746 * poly16x8_t vrev64q_p16 (poly16x8_t)
27747 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27749 * poly8x16_t vrev64q_p8 (poly8x16_t)
27750 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27752 * uint16x4_t vrev32_u16 (uint16x4_t)
27753 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27755 * int16x4_t vrev32_s16 (int16x4_t)
27756 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27758 * uint8x8_t vrev32_u8 (uint8x8_t)
27759 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27761 * int8x8_t vrev32_s8 (int8x8_t)
27762 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27764 * poly16x4_t vrev32_p16 (poly16x4_t)
27765 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27767 * poly8x8_t vrev32_p8 (poly8x8_t)
27768 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27770 * uint16x8_t vrev32q_u16 (uint16x8_t)
27771 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27773 * int16x8_t vrev32q_s16 (int16x8_t)
27774 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27776 * uint8x16_t vrev32q_u8 (uint8x16_t)
27777 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27779 * int8x16_t vrev32q_s8 (int8x16_t)
27780 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27782 * poly16x8_t vrev32q_p16 (poly16x8_t)
27783 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27785 * poly8x16_t vrev32q_p8 (poly8x16_t)
27786 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27788 * uint8x8_t vrev16_u8 (uint8x8_t)
27789 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27791 * int8x8_t vrev16_s8 (int8x8_t)
27792 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27794 * poly8x8_t vrev16_p8 (poly8x8_t)
27795 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27797 * uint8x16_t vrev16q_u8 (uint8x16_t)
27798 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27800 * int8x16_t vrev16q_s8 (int8x16_t)
27801 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27803 * poly8x16_t vrev16q_p8 (poly8x16_t)
27804 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27806 5.50.3.64 Bit selection
27807 .......................
27809 * uint32x2_t vbsl_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
27810 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27811 D0, D0, D0' _or_ `vbif D0, D0, D0'
27813 * uint16x4_t vbsl_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
27814 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27815 D0, D0, D0' _or_ `vbif D0, D0, D0'
27817 * uint8x8_t vbsl_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
27818 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27819 D0, D0, D0' _or_ `vbif D0, D0, D0'
27821 * int32x2_t vbsl_s32 (uint32x2_t, int32x2_t, int32x2_t)
27822 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27823 D0, D0, D0' _or_ `vbif D0, D0, D0'
27825 * int16x4_t vbsl_s16 (uint16x4_t, int16x4_t, int16x4_t)
27826 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27827 D0, D0, D0' _or_ `vbif D0, D0, D0'
27829 * int8x8_t vbsl_s8 (uint8x8_t, int8x8_t, int8x8_t)
27830 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27831 D0, D0, D0' _or_ `vbif D0, D0, D0'
27833 * uint64x1_t vbsl_u64 (uint64x1_t, uint64x1_t, uint64x1_t)
27834 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27835 D0, D0, D0' _or_ `vbif D0, D0, D0'
27837 * int64x1_t vbsl_s64 (uint64x1_t, int64x1_t, int64x1_t)
27838 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27839 D0, D0, D0' _or_ `vbif D0, D0, D0'
27841 * float32x2_t vbsl_f32 (uint32x2_t, float32x2_t, float32x2_t)
27842 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27843 D0, D0, D0' _or_ `vbif D0, D0, D0'
27845 * poly16x4_t vbsl_p16 (uint16x4_t, poly16x4_t, poly16x4_t)
27846 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27847 D0, D0, D0' _or_ `vbif D0, D0, D0'
27849 * poly8x8_t vbsl_p8 (uint8x8_t, poly8x8_t, poly8x8_t)
27850 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27851 D0, D0, D0' _or_ `vbif D0, D0, D0'
27853 * uint32x4_t vbslq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
27854 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27855 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27857 * uint16x8_t vbslq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
27858 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27859 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27861 * uint8x16_t vbslq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
27862 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27863 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27865 * int32x4_t vbslq_s32 (uint32x4_t, int32x4_t, int32x4_t)
27866 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27867 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27869 * int16x8_t vbslq_s16 (uint16x8_t, int16x8_t, int16x8_t)
27870 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27871 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27873 * int8x16_t vbslq_s8 (uint8x16_t, int8x16_t, int8x16_t)
27874 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27875 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27877 * uint64x2_t vbslq_u64 (uint64x2_t, uint64x2_t, uint64x2_t)
27878 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27879 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27881 * int64x2_t vbslq_s64 (uint64x2_t, int64x2_t, int64x2_t)
27882 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27883 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27885 * float32x4_t vbslq_f32 (uint32x4_t, float32x4_t, float32x4_t)
27886 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27887 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27889 * poly16x8_t vbslq_p16 (uint16x8_t, poly16x8_t, poly16x8_t)
27890 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27891 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27893 * poly8x16_t vbslq_p8 (uint8x16_t, poly8x16_t, poly8x16_t)
27894 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27895 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27897 5.50.3.65 Transpose elements
27898 ............................
27900 * uint32x2x2_t vtrn_u32 (uint32x2_t, uint32x2_t)
27901 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27903 * uint16x4x2_t vtrn_u16 (uint16x4_t, uint16x4_t)
27904 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27906 * uint8x8x2_t vtrn_u8 (uint8x8_t, uint8x8_t)
27907 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27909 * int32x2x2_t vtrn_s32 (int32x2_t, int32x2_t)
27910 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27912 * int16x4x2_t vtrn_s16 (int16x4_t, int16x4_t)
27913 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27915 * int8x8x2_t vtrn_s8 (int8x8_t, int8x8_t)
27916 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27918 * float32x2x2_t vtrn_f32 (float32x2_t, float32x2_t)
27919 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27921 * poly16x4x2_t vtrn_p16 (poly16x4_t, poly16x4_t)
27922 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27924 * poly8x8x2_t vtrn_p8 (poly8x8_t, poly8x8_t)
27925 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27927 * uint32x4x2_t vtrnq_u32 (uint32x4_t, uint32x4_t)
27928 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27930 * uint16x8x2_t vtrnq_u16 (uint16x8_t, uint16x8_t)
27931 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27933 * uint8x16x2_t vtrnq_u8 (uint8x16_t, uint8x16_t)
27934 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27936 * int32x4x2_t vtrnq_s32 (int32x4_t, int32x4_t)
27937 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27939 * int16x8x2_t vtrnq_s16 (int16x8_t, int16x8_t)
27940 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27942 * int8x16x2_t vtrnq_s8 (int8x16_t, int8x16_t)
27943 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27945 * float32x4x2_t vtrnq_f32 (float32x4_t, float32x4_t)
27946 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27948 * poly16x8x2_t vtrnq_p16 (poly16x8_t, poly16x8_t)
27949 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27951 * poly8x16x2_t vtrnq_p8 (poly8x16_t, poly8x16_t)
27952 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27954 5.50.3.66 Zip elements
27955 ......................
27957 * uint32x2x2_t vzip_u32 (uint32x2_t, uint32x2_t)
27958 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27960 * uint16x4x2_t vzip_u16 (uint16x4_t, uint16x4_t)
27961 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27963 * uint8x8x2_t vzip_u8 (uint8x8_t, uint8x8_t)
27964 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27966 * int32x2x2_t vzip_s32 (int32x2_t, int32x2_t)
27967 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27969 * int16x4x2_t vzip_s16 (int16x4_t, int16x4_t)
27970 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27972 * int8x8x2_t vzip_s8 (int8x8_t, int8x8_t)
27973 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27975 * float32x2x2_t vzip_f32 (float32x2_t, float32x2_t)
27976 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27978 * poly16x4x2_t vzip_p16 (poly16x4_t, poly16x4_t)
27979 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27981 * poly8x8x2_t vzip_p8 (poly8x8_t, poly8x8_t)
27982 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27984 * uint32x4x2_t vzipq_u32 (uint32x4_t, uint32x4_t)
27985 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
27987 * uint16x8x2_t vzipq_u16 (uint16x8_t, uint16x8_t)
27988 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
27990 * uint8x16x2_t vzipq_u8 (uint8x16_t, uint8x16_t)
27991 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
27993 * int32x4x2_t vzipq_s32 (int32x4_t, int32x4_t)
27994 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
27996 * int16x8x2_t vzipq_s16 (int16x8_t, int16x8_t)
27997 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
27999 * int8x16x2_t vzipq_s8 (int8x16_t, int8x16_t)
28000 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
28002 * float32x4x2_t vzipq_f32 (float32x4_t, float32x4_t)
28003 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
28005 * poly16x8x2_t vzipq_p16 (poly16x8_t, poly16x8_t)
28006 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
28008 * poly8x16x2_t vzipq_p8 (poly8x16_t, poly8x16_t)
28009 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
28011 5.50.3.67 Unzip elements
28012 ........................
28014 * uint32x2x2_t vuzp_u32 (uint32x2_t, uint32x2_t)
28015 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
28017 * uint16x4x2_t vuzp_u16 (uint16x4_t, uint16x4_t)
28018 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
28020 * uint8x8x2_t vuzp_u8 (uint8x8_t, uint8x8_t)
28021 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
28023 * int32x2x2_t vuzp_s32 (int32x2_t, int32x2_t)
28024 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
28026 * int16x4x2_t vuzp_s16 (int16x4_t, int16x4_t)
28027 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
28029 * int8x8x2_t vuzp_s8 (int8x8_t, int8x8_t)
28030 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
28032 * float32x2x2_t vuzp_f32 (float32x2_t, float32x2_t)
28033 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
28035 * poly16x4x2_t vuzp_p16 (poly16x4_t, poly16x4_t)
28036 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
28038 * poly8x8x2_t vuzp_p8 (poly8x8_t, poly8x8_t)
28039 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
28041 * uint32x4x2_t vuzpq_u32 (uint32x4_t, uint32x4_t)
28042 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28044 * uint16x8x2_t vuzpq_u16 (uint16x8_t, uint16x8_t)
28045 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28047 * uint8x16x2_t vuzpq_u8 (uint8x16_t, uint8x16_t)
28048 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28050 * int32x4x2_t vuzpq_s32 (int32x4_t, int32x4_t)
28051 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28053 * int16x8x2_t vuzpq_s16 (int16x8_t, int16x8_t)
28054 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28056 * int8x16x2_t vuzpq_s8 (int8x16_t, int8x16_t)
28057 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28059 * float32x4x2_t vuzpq_f32 (float32x4_t, float32x4_t)
28060 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28062 * poly16x8x2_t vuzpq_p16 (poly16x8_t, poly16x8_t)
28063 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28065 * poly8x16x2_t vuzpq_p8 (poly8x16_t, poly8x16_t)
28066 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28068 5.50.3.68 Element/structure loads, VLD1 variants
28069 ................................................
28071 * uint32x2_t vld1_u32 (const uint32_t *)
28072 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28074 * uint16x4_t vld1_u16 (const uint16_t *)
28075 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28077 * uint8x8_t vld1_u8 (const uint8_t *)
28078 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28080 * int32x2_t vld1_s32 (const int32_t *)
28081 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28083 * int16x4_t vld1_s16 (const int16_t *)
28084 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28086 * int8x8_t vld1_s8 (const int8_t *)
28087 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28089 * uint64x1_t vld1_u64 (const uint64_t *)
28090 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28092 * int64x1_t vld1_s64 (const int64_t *)
28093 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28095 * float32x2_t vld1_f32 (const float32_t *)
28096 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28098 * poly16x4_t vld1_p16 (const poly16_t *)
28099 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28101 * poly8x8_t vld1_p8 (const poly8_t *)
28102 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28104 * uint32x4_t vld1q_u32 (const uint32_t *)
28105 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28107 * uint16x8_t vld1q_u16 (const uint16_t *)
28108 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28110 * uint8x16_t vld1q_u8 (const uint8_t *)
28111 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28113 * int32x4_t vld1q_s32 (const int32_t *)
28114 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28116 * int16x8_t vld1q_s16 (const int16_t *)
28117 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28119 * int8x16_t vld1q_s8 (const int8_t *)
28120 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28122 * uint64x2_t vld1q_u64 (const uint64_t *)
28123 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28125 * int64x2_t vld1q_s64 (const int64_t *)
28126 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28128 * float32x4_t vld1q_f32 (const float32_t *)
28129 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28131 * poly16x8_t vld1q_p16 (const poly16_t *)
28132 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28134 * poly8x16_t vld1q_p8 (const poly8_t *)
28135 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28137 * uint32x2_t vld1_lane_u32 (const uint32_t *, uint32x2_t, const int)
28138 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28140 * uint16x4_t vld1_lane_u16 (const uint16_t *, uint16x4_t, const int)
28141 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28143 * uint8x8_t vld1_lane_u8 (const uint8_t *, uint8x8_t, const int)
28144 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28146 * int32x2_t vld1_lane_s32 (const int32_t *, int32x2_t, const int)
28147 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28149 * int16x4_t vld1_lane_s16 (const int16_t *, int16x4_t, const int)
28150 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28152 * int8x8_t vld1_lane_s8 (const int8_t *, int8x8_t, const int)
28153 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28155 * float32x2_t vld1_lane_f32 (const float32_t *, float32x2_t, const
28157 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28159 * poly16x4_t vld1_lane_p16 (const poly16_t *, poly16x4_t, const int)
28160 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28162 * poly8x8_t vld1_lane_p8 (const poly8_t *, poly8x8_t, const int)
28163 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28165 * uint64x1_t vld1_lane_u64 (const uint64_t *, uint64x1_t, const int)
28166 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28168 * int64x1_t vld1_lane_s64 (const int64_t *, int64x1_t, const int)
28169 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28171 * uint32x4_t vld1q_lane_u32 (const uint32_t *, uint32x4_t, const int)
28172 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28174 * uint16x8_t vld1q_lane_u16 (const uint16_t *, uint16x8_t, const int)
28175 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28177 * uint8x16_t vld1q_lane_u8 (const uint8_t *, uint8x16_t, const int)
28178 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28180 * int32x4_t vld1q_lane_s32 (const int32_t *, int32x4_t, const int)
28181 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28183 * int16x8_t vld1q_lane_s16 (const int16_t *, int16x8_t, const int)
28184 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28186 * int8x16_t vld1q_lane_s8 (const int8_t *, int8x16_t, const int)
28187 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28189 * float32x4_t vld1q_lane_f32 (const float32_t *, float32x4_t, const
28191 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28193 * poly16x8_t vld1q_lane_p16 (const poly16_t *, poly16x8_t, const int)
28194 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28196 * poly8x16_t vld1q_lane_p8 (const poly8_t *, poly8x16_t, const int)
28197 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28199 * uint64x2_t vld1q_lane_u64 (const uint64_t *, uint64x2_t, const int)
28200 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28202 * int64x2_t vld1q_lane_s64 (const int64_t *, int64x2_t, const int)
28203 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28205 * uint32x2_t vld1_dup_u32 (const uint32_t *)
28206 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28208 * uint16x4_t vld1_dup_u16 (const uint16_t *)
28209 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28211 * uint8x8_t vld1_dup_u8 (const uint8_t *)
28212 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28214 * int32x2_t vld1_dup_s32 (const int32_t *)
28215 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28217 * int16x4_t vld1_dup_s16 (const int16_t *)
28218 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28220 * int8x8_t vld1_dup_s8 (const int8_t *)
28221 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28223 * float32x2_t vld1_dup_f32 (const float32_t *)
28224 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28226 * poly16x4_t vld1_dup_p16 (const poly16_t *)
28227 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28229 * poly8x8_t vld1_dup_p8 (const poly8_t *)
28230 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28232 * uint64x1_t vld1_dup_u64 (const uint64_t *)
28233 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28235 * int64x1_t vld1_dup_s64 (const int64_t *)
28236 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28238 * uint32x4_t vld1q_dup_u32 (const uint32_t *)
28239 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28241 * uint16x8_t vld1q_dup_u16 (const uint16_t *)
28242 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28244 * uint8x16_t vld1q_dup_u8 (const uint8_t *)
28245 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28247 * int32x4_t vld1q_dup_s32 (const int32_t *)
28248 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28250 * int16x8_t vld1q_dup_s16 (const int16_t *)
28251 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28253 * int8x16_t vld1q_dup_s8 (const int8_t *)
28254 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28256 * float32x4_t vld1q_dup_f32 (const float32_t *)
28257 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28259 * poly16x8_t vld1q_dup_p16 (const poly16_t *)
28260 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28262 * poly8x16_t vld1q_dup_p8 (const poly8_t *)
28263 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28265 * uint64x2_t vld1q_dup_u64 (const uint64_t *)
28266 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28268 * int64x2_t vld1q_dup_s64 (const int64_t *)
28269 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28271 5.50.3.69 Element/structure stores, VST1 variants
28272 .................................................
28274 * void vst1_u32 (uint32_t *, uint32x2_t)
28275 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28277 * void vst1_u16 (uint16_t *, uint16x4_t)
28278 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28280 * void vst1_u8 (uint8_t *, uint8x8_t)
28281 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28283 * void vst1_s32 (int32_t *, int32x2_t)
28284 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28286 * void vst1_s16 (int16_t *, int16x4_t)
28287 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28289 * void vst1_s8 (int8_t *, int8x8_t)
28290 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28292 * void vst1_u64 (uint64_t *, uint64x1_t)
28293 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28295 * void vst1_s64 (int64_t *, int64x1_t)
28296 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28298 * void vst1_f32 (float32_t *, float32x2_t)
28299 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28301 * void vst1_p16 (poly16_t *, poly16x4_t)
28302 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28304 * void vst1_p8 (poly8_t *, poly8x8_t)
28305 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28307 * void vst1q_u32 (uint32_t *, uint32x4_t)
28308 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28310 * void vst1q_u16 (uint16_t *, uint16x8_t)
28311 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28313 * void vst1q_u8 (uint8_t *, uint8x16_t)
28314 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28316 * void vst1q_s32 (int32_t *, int32x4_t)
28317 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28319 * void vst1q_s16 (int16_t *, int16x8_t)
28320 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28322 * void vst1q_s8 (int8_t *, int8x16_t)
28323 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28325 * void vst1q_u64 (uint64_t *, uint64x2_t)
28326 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28328 * void vst1q_s64 (int64_t *, int64x2_t)
28329 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28331 * void vst1q_f32 (float32_t *, float32x4_t)
28332 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28334 * void vst1q_p16 (poly16_t *, poly16x8_t)
28335 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28337 * void vst1q_p8 (poly8_t *, poly8x16_t)
28338 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28340 * void vst1_lane_u32 (uint32_t *, uint32x2_t, const int)
28341 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28343 * void vst1_lane_u16 (uint16_t *, uint16x4_t, const int)
28344 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28346 * void vst1_lane_u8 (uint8_t *, uint8x8_t, const int)
28347 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28349 * void vst1_lane_s32 (int32_t *, int32x2_t, const int)
28350 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28352 * void vst1_lane_s16 (int16_t *, int16x4_t, const int)
28353 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28355 * void vst1_lane_s8 (int8_t *, int8x8_t, const int)
28356 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28358 * void vst1_lane_f32 (float32_t *, float32x2_t, const int)
28359 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28361 * void vst1_lane_p16 (poly16_t *, poly16x4_t, const int)
28362 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28364 * void vst1_lane_p8 (poly8_t *, poly8x8_t, const int)
28365 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28367 * void vst1_lane_s64 (int64_t *, int64x1_t, const int)
28368 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28370 * void vst1_lane_u64 (uint64_t *, uint64x1_t, const int)
28371 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28373 * void vst1q_lane_u32 (uint32_t *, uint32x4_t, const int)
28374 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28376 * void vst1q_lane_u16 (uint16_t *, uint16x8_t, const int)
28377 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28379 * void vst1q_lane_u8 (uint8_t *, uint8x16_t, const int)
28380 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28382 * void vst1q_lane_s32 (int32_t *, int32x4_t, const int)
28383 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28385 * void vst1q_lane_s16 (int16_t *, int16x8_t, const int)
28386 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28388 * void vst1q_lane_s8 (int8_t *, int8x16_t, const int)
28389 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28391 * void vst1q_lane_f32 (float32_t *, float32x4_t, const int)
28392 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28394 * void vst1q_lane_p16 (poly16_t *, poly16x8_t, const int)
28395 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28397 * void vst1q_lane_p8 (poly8_t *, poly8x16_t, const int)
28398 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28400 * void vst1q_lane_s64 (int64_t *, int64x2_t, const int)
28401 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28403 * void vst1q_lane_u64 (uint64_t *, uint64x2_t, const int)
28404 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28406 5.50.3.70 Element/structure loads, VLD2 variants
28407 ................................................
28409 * uint32x2x2_t vld2_u32 (const uint32_t *)
28410 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28412 * uint16x4x2_t vld2_u16 (const uint16_t *)
28413 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28415 * uint8x8x2_t vld2_u8 (const uint8_t *)
28416 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28418 * int32x2x2_t vld2_s32 (const int32_t *)
28419 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28421 * int16x4x2_t vld2_s16 (const int16_t *)
28422 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28424 * int8x8x2_t vld2_s8 (const int8_t *)
28425 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28427 * float32x2x2_t vld2_f32 (const float32_t *)
28428 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28430 * poly16x4x2_t vld2_p16 (const poly16_t *)
28431 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28433 * poly8x8x2_t vld2_p8 (const poly8_t *)
28434 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28436 * uint64x1x2_t vld2_u64 (const uint64_t *)
28437 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28439 * int64x1x2_t vld2_s64 (const int64_t *)
28440 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28442 * uint32x4x2_t vld2q_u32 (const uint32_t *)
28443 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28445 * uint16x8x2_t vld2q_u16 (const uint16_t *)
28446 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28448 * uint8x16x2_t vld2q_u8 (const uint8_t *)
28449 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28451 * int32x4x2_t vld2q_s32 (const int32_t *)
28452 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28454 * int16x8x2_t vld2q_s16 (const int16_t *)
28455 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28457 * int8x16x2_t vld2q_s8 (const int8_t *)
28458 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28460 * float32x4x2_t vld2q_f32 (const float32_t *)
28461 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28463 * poly16x8x2_t vld2q_p16 (const poly16_t *)
28464 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28466 * poly8x16x2_t vld2q_p8 (const poly8_t *)
28467 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28469 * uint32x2x2_t vld2_lane_u32 (const uint32_t *, uint32x2x2_t, const
28471 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28473 * uint16x4x2_t vld2_lane_u16 (const uint16_t *, uint16x4x2_t, const
28475 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28477 * uint8x8x2_t vld2_lane_u8 (const uint8_t *, uint8x8x2_t, const int)
28478 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28480 * int32x2x2_t vld2_lane_s32 (const int32_t *, int32x2x2_t, const int)
28481 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28483 * int16x4x2_t vld2_lane_s16 (const int16_t *, int16x4x2_t, const int)
28484 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28486 * int8x8x2_t vld2_lane_s8 (const int8_t *, int8x8x2_t, const int)
28487 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28489 * float32x2x2_t vld2_lane_f32 (const float32_t *, float32x2x2_t,
28491 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28493 * poly16x4x2_t vld2_lane_p16 (const poly16_t *, poly16x4x2_t, const
28495 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28497 * poly8x8x2_t vld2_lane_p8 (const poly8_t *, poly8x8x2_t, const int)
28498 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28500 * int32x4x2_t vld2q_lane_s32 (const int32_t *, int32x4x2_t, const
28502 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28504 * int16x8x2_t vld2q_lane_s16 (const int16_t *, int16x8x2_t, const
28506 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28508 * uint32x4x2_t vld2q_lane_u32 (const uint32_t *, uint32x4x2_t, const
28510 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28512 * uint16x8x2_t vld2q_lane_u16 (const uint16_t *, uint16x8x2_t, const
28514 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28516 * float32x4x2_t vld2q_lane_f32 (const float32_t *, float32x4x2_t,
28518 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28520 * poly16x8x2_t vld2q_lane_p16 (const poly16_t *, poly16x8x2_t, const
28522 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28524 * uint32x2x2_t vld2_dup_u32 (const uint32_t *)
28525 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28527 * uint16x4x2_t vld2_dup_u16 (const uint16_t *)
28528 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28530 * uint8x8x2_t vld2_dup_u8 (const uint8_t *)
28531 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28533 * int32x2x2_t vld2_dup_s32 (const int32_t *)
28534 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28536 * int16x4x2_t vld2_dup_s16 (const int16_t *)
28537 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28539 * int8x8x2_t vld2_dup_s8 (const int8_t *)
28540 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28542 * float32x2x2_t vld2_dup_f32 (const float32_t *)
28543 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28545 * poly16x4x2_t vld2_dup_p16 (const poly16_t *)
28546 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28548 * poly8x8x2_t vld2_dup_p8 (const poly8_t *)
28549 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28551 * uint64x1x2_t vld2_dup_u64 (const uint64_t *)
28552 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28554 * int64x1x2_t vld2_dup_s64 (const int64_t *)
28555 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28557 5.50.3.71 Element/structure stores, VST2 variants
28558 .................................................
28560 * void vst2_u32 (uint32_t *, uint32x2x2_t)
28561 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28563 * void vst2_u16 (uint16_t *, uint16x4x2_t)
28564 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28566 * void vst2_u8 (uint8_t *, uint8x8x2_t)
28567 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28569 * void vst2_s32 (int32_t *, int32x2x2_t)
28570 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28572 * void vst2_s16 (int16_t *, int16x4x2_t)
28573 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28575 * void vst2_s8 (int8_t *, int8x8x2_t)
28576 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28578 * void vst2_f32 (float32_t *, float32x2x2_t)
28579 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28581 * void vst2_p16 (poly16_t *, poly16x4x2_t)
28582 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28584 * void vst2_p8 (poly8_t *, poly8x8x2_t)
28585 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28587 * void vst2_u64 (uint64_t *, uint64x1x2_t)
28588 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28590 * void vst2_s64 (int64_t *, int64x1x2_t)
28591 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28593 * void vst2q_u32 (uint32_t *, uint32x4x2_t)
28594 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28596 * void vst2q_u16 (uint16_t *, uint16x8x2_t)
28597 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28599 * void vst2q_u8 (uint8_t *, uint8x16x2_t)
28600 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28602 * void vst2q_s32 (int32_t *, int32x4x2_t)
28603 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28605 * void vst2q_s16 (int16_t *, int16x8x2_t)
28606 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28608 * void vst2q_s8 (int8_t *, int8x16x2_t)
28609 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28611 * void vst2q_f32 (float32_t *, float32x4x2_t)
28612 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28614 * void vst2q_p16 (poly16_t *, poly16x8x2_t)
28615 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28617 * void vst2q_p8 (poly8_t *, poly8x16x2_t)
28618 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28620 * void vst2_lane_u32 (uint32_t *, uint32x2x2_t, const int)
28621 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28623 * void vst2_lane_u16 (uint16_t *, uint16x4x2_t, const int)
28624 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28626 * void vst2_lane_u8 (uint8_t *, uint8x8x2_t, const int)
28627 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28629 * void vst2_lane_s32 (int32_t *, int32x2x2_t, const int)
28630 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28632 * void vst2_lane_s16 (int16_t *, int16x4x2_t, const int)
28633 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28635 * void vst2_lane_s8 (int8_t *, int8x8x2_t, const int)
28636 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28638 * void vst2_lane_f32 (float32_t *, float32x2x2_t, const int)
28639 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28641 * void vst2_lane_p16 (poly16_t *, poly16x4x2_t, const int)
28642 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28644 * void vst2_lane_p8 (poly8_t *, poly8x8x2_t, const int)
28645 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28647 * void vst2q_lane_s32 (int32_t *, int32x4x2_t, const int)
28648 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28650 * void vst2q_lane_s16 (int16_t *, int16x8x2_t, const int)
28651 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28653 * void vst2q_lane_u32 (uint32_t *, uint32x4x2_t, const int)
28654 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28656 * void vst2q_lane_u16 (uint16_t *, uint16x8x2_t, const int)
28657 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28659 * void vst2q_lane_f32 (float32_t *, float32x4x2_t, const int)
28660 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28662 * void vst2q_lane_p16 (poly16_t *, poly16x8x2_t, const int)
28663 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28665 5.50.3.72 Element/structure loads, VLD3 variants
28666 ................................................
28668 * uint32x2x3_t vld3_u32 (const uint32_t *)
28669 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28671 * uint16x4x3_t vld3_u16 (const uint16_t *)
28672 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28674 * uint8x8x3_t vld3_u8 (const uint8_t *)
28675 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28677 * int32x2x3_t vld3_s32 (const int32_t *)
28678 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28680 * int16x4x3_t vld3_s16 (const int16_t *)
28681 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28683 * int8x8x3_t vld3_s8 (const int8_t *)
28684 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28686 * float32x2x3_t vld3_f32 (const float32_t *)
28687 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28689 * poly16x4x3_t vld3_p16 (const poly16_t *)
28690 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28692 * poly8x8x3_t vld3_p8 (const poly8_t *)
28693 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28695 * uint64x1x3_t vld3_u64 (const uint64_t *)
28696 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28698 * int64x1x3_t vld3_s64 (const int64_t *)
28699 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28701 * uint32x4x3_t vld3q_u32 (const uint32_t *)
28702 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28704 * uint16x8x3_t vld3q_u16 (const uint16_t *)
28705 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28707 * uint8x16x3_t vld3q_u8 (const uint8_t *)
28708 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28710 * int32x4x3_t vld3q_s32 (const int32_t *)
28711 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28713 * int16x8x3_t vld3q_s16 (const int16_t *)
28714 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28716 * int8x16x3_t vld3q_s8 (const int8_t *)
28717 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28719 * float32x4x3_t vld3q_f32 (const float32_t *)
28720 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28722 * poly16x8x3_t vld3q_p16 (const poly16_t *)
28723 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28725 * poly8x16x3_t vld3q_p8 (const poly8_t *)
28726 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28728 * uint32x2x3_t vld3_lane_u32 (const uint32_t *, uint32x2x3_t, const
28730 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28733 * uint16x4x3_t vld3_lane_u16 (const uint16_t *, uint16x4x3_t, const
28735 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28738 * uint8x8x3_t vld3_lane_u8 (const uint8_t *, uint8x8x3_t, const int)
28739 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28742 * int32x2x3_t vld3_lane_s32 (const int32_t *, int32x2x3_t, const int)
28743 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28746 * int16x4x3_t vld3_lane_s16 (const int16_t *, int16x4x3_t, const int)
28747 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28750 * int8x8x3_t vld3_lane_s8 (const int8_t *, int8x8x3_t, const int)
28751 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28754 * float32x2x3_t vld3_lane_f32 (const float32_t *, float32x2x3_t,
28756 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28759 * poly16x4x3_t vld3_lane_p16 (const poly16_t *, poly16x4x3_t, const
28761 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28764 * poly8x8x3_t vld3_lane_p8 (const poly8_t *, poly8x8x3_t, const int)
28765 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28768 * int32x4x3_t vld3q_lane_s32 (const int32_t *, int32x4x3_t, const
28770 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28773 * int16x8x3_t vld3q_lane_s16 (const int16_t *, int16x8x3_t, const
28775 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28778 * uint32x4x3_t vld3q_lane_u32 (const uint32_t *, uint32x4x3_t, const
28780 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28783 * uint16x8x3_t vld3q_lane_u16 (const uint16_t *, uint16x8x3_t, const
28785 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28788 * float32x4x3_t vld3q_lane_f32 (const float32_t *, float32x4x3_t,
28790 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28793 * poly16x8x3_t vld3q_lane_p16 (const poly16_t *, poly16x8x3_t, const
28795 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28798 * uint32x2x3_t vld3_dup_u32 (const uint32_t *)
28799 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28802 * uint16x4x3_t vld3_dup_u16 (const uint16_t *)
28803 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28806 * uint8x8x3_t vld3_dup_u8 (const uint8_t *)
28807 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28810 * int32x2x3_t vld3_dup_s32 (const int32_t *)
28811 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28814 * int16x4x3_t vld3_dup_s16 (const int16_t *)
28815 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28818 * int8x8x3_t vld3_dup_s8 (const int8_t *)
28819 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28822 * float32x2x3_t vld3_dup_f32 (const float32_t *)
28823 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28826 * poly16x4x3_t vld3_dup_p16 (const poly16_t *)
28827 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28830 * poly8x8x3_t vld3_dup_p8 (const poly8_t *)
28831 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28834 * uint64x1x3_t vld3_dup_u64 (const uint64_t *)
28835 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28837 * int64x1x3_t vld3_dup_s64 (const int64_t *)
28838 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28840 5.50.3.73 Element/structure stores, VST3 variants
28841 .................................................
28843 * void vst3_u32 (uint32_t *, uint32x2x3_t)
28844 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28846 * void vst3_u16 (uint16_t *, uint16x4x3_t)
28847 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28849 * void vst3_u8 (uint8_t *, uint8x8x3_t)
28850 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28852 * void vst3_s32 (int32_t *, int32x2x3_t)
28853 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28855 * void vst3_s16 (int16_t *, int16x4x3_t)
28856 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28858 * void vst3_s8 (int8_t *, int8x8x3_t)
28859 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28861 * void vst3_f32 (float32_t *, float32x2x3_t)
28862 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28864 * void vst3_p16 (poly16_t *, poly16x4x3_t)
28865 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28867 * void vst3_p8 (poly8_t *, poly8x8x3_t)
28868 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28870 * void vst3_u64 (uint64_t *, uint64x1x3_t)
28871 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
28873 * void vst3_s64 (int64_t *, int64x1x3_t)
28874 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
28876 * void vst3q_u32 (uint32_t *, uint32x4x3_t)
28877 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28879 * void vst3q_u16 (uint16_t *, uint16x8x3_t)
28880 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28882 * void vst3q_u8 (uint8_t *, uint8x16x3_t)
28883 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28885 * void vst3q_s32 (int32_t *, int32x4x3_t)
28886 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28888 * void vst3q_s16 (int16_t *, int16x8x3_t)
28889 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28891 * void vst3q_s8 (int8_t *, int8x16x3_t)
28892 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28894 * void vst3q_f32 (float32_t *, float32x4x3_t)
28895 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28897 * void vst3q_p16 (poly16_t *, poly16x8x3_t)
28898 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28900 * void vst3q_p8 (poly8_t *, poly8x16x3_t)
28901 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28903 * void vst3_lane_u32 (uint32_t *, uint32x2x3_t, const int)
28904 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28907 * void vst3_lane_u16 (uint16_t *, uint16x4x3_t, const int)
28908 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28911 * void vst3_lane_u8 (uint8_t *, uint8x8x3_t, const int)
28912 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28915 * void vst3_lane_s32 (int32_t *, int32x2x3_t, const int)
28916 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28919 * void vst3_lane_s16 (int16_t *, int16x4x3_t, const int)
28920 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28923 * void vst3_lane_s8 (int8_t *, int8x8x3_t, const int)
28924 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28927 * void vst3_lane_f32 (float32_t *, float32x2x3_t, const int)
28928 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28931 * void vst3_lane_p16 (poly16_t *, poly16x4x3_t, const int)
28932 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28935 * void vst3_lane_p8 (poly8_t *, poly8x8x3_t, const int)
28936 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28939 * void vst3q_lane_s32 (int32_t *, int32x4x3_t, const int)
28940 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28943 * void vst3q_lane_s16 (int16_t *, int16x8x3_t, const int)
28944 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28947 * void vst3q_lane_u32 (uint32_t *, uint32x4x3_t, const int)
28948 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28951 * void vst3q_lane_u16 (uint16_t *, uint16x8x3_t, const int)
28952 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28955 * void vst3q_lane_f32 (float32_t *, float32x4x3_t, const int)
28956 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28959 * void vst3q_lane_p16 (poly16_t *, poly16x8x3_t, const int)
28960 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28963 5.50.3.74 Element/structure loads, VLD4 variants
28964 ................................................
28966 * uint32x2x4_t vld4_u32 (const uint32_t *)
28967 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28969 * uint16x4x4_t vld4_u16 (const uint16_t *)
28970 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28972 * uint8x8x4_t vld4_u8 (const uint8_t *)
28973 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28975 * int32x2x4_t vld4_s32 (const int32_t *)
28976 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28978 * int16x4x4_t vld4_s16 (const int16_t *)
28979 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28981 * int8x8x4_t vld4_s8 (const int8_t *)
28982 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28984 * float32x2x4_t vld4_f32 (const float32_t *)
28985 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28987 * poly16x4x4_t vld4_p16 (const poly16_t *)
28988 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28990 * poly8x8x4_t vld4_p8 (const poly8_t *)
28991 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28993 * uint64x1x4_t vld4_u64 (const uint64_t *)
28994 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
28996 * int64x1x4_t vld4_s64 (const int64_t *)
28997 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
28999 * uint32x4x4_t vld4q_u32 (const uint32_t *)
29000 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
29002 * uint16x8x4_t vld4q_u16 (const uint16_t *)
29003 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
29005 * uint8x16x4_t vld4q_u8 (const uint8_t *)
29006 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
29008 * int32x4x4_t vld4q_s32 (const int32_t *)
29009 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
29011 * int16x8x4_t vld4q_s16 (const int16_t *)
29012 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
29014 * int8x16x4_t vld4q_s8 (const int8_t *)
29015 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
29017 * float32x4x4_t vld4q_f32 (const float32_t *)
29018 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
29020 * poly16x8x4_t vld4q_p16 (const poly16_t *)
29021 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
29023 * poly8x16x4_t vld4q_p8 (const poly8_t *)
29024 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
29026 * uint32x2x4_t vld4_lane_u32 (const uint32_t *, uint32x2x4_t, const
29028 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29031 * uint16x4x4_t vld4_lane_u16 (const uint16_t *, uint16x4x4_t, const
29033 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29036 * uint8x8x4_t vld4_lane_u8 (const uint8_t *, uint8x8x4_t, const int)
29037 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
29040 * int32x2x4_t vld4_lane_s32 (const int32_t *, int32x2x4_t, const int)
29041 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29044 * int16x4x4_t vld4_lane_s16 (const int16_t *, int16x4x4_t, const int)
29045 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29048 * int8x8x4_t vld4_lane_s8 (const int8_t *, int8x8x4_t, const int)
29049 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
29052 * float32x2x4_t vld4_lane_f32 (const float32_t *, float32x2x4_t,
29054 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29057 * poly16x4x4_t vld4_lane_p16 (const poly16_t *, poly16x4x4_t, const
29059 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29062 * poly8x8x4_t vld4_lane_p8 (const poly8_t *, poly8x8x4_t, const int)
29063 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
29066 * int32x4x4_t vld4q_lane_s32 (const int32_t *, int32x4x4_t, const
29068 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29071 * int16x8x4_t vld4q_lane_s16 (const int16_t *, int16x8x4_t, const
29073 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29076 * uint32x4x4_t vld4q_lane_u32 (const uint32_t *, uint32x4x4_t, const
29078 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29081 * uint16x8x4_t vld4q_lane_u16 (const uint16_t *, uint16x8x4_t, const
29083 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29086 * float32x4x4_t vld4q_lane_f32 (const float32_t *, float32x4x4_t,
29088 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29091 * poly16x8x4_t vld4q_lane_p16 (const poly16_t *, poly16x8x4_t, const
29093 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29096 * uint32x2x4_t vld4_dup_u32 (const uint32_t *)
29097 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29100 * uint16x4x4_t vld4_dup_u16 (const uint16_t *)
29101 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29104 * uint8x8x4_t vld4_dup_u8 (const uint8_t *)
29105 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29108 * int32x2x4_t vld4_dup_s32 (const int32_t *)
29109 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29112 * int16x4x4_t vld4_dup_s16 (const int16_t *)
29113 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29116 * int8x8x4_t vld4_dup_s8 (const int8_t *)
29117 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29120 * float32x2x4_t vld4_dup_f32 (const float32_t *)
29121 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29124 * poly16x4x4_t vld4_dup_p16 (const poly16_t *)
29125 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29128 * poly8x8x4_t vld4_dup_p8 (const poly8_t *)
29129 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29132 * uint64x1x4_t vld4_dup_u64 (const uint64_t *)
29133 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
29135 * int64x1x4_t vld4_dup_s64 (const int64_t *)
29136 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
29138 5.50.3.75 Element/structure stores, VST4 variants
29139 .................................................
29141 * void vst4_u32 (uint32_t *, uint32x2x4_t)
29142 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29144 * void vst4_u16 (uint16_t *, uint16x4x4_t)
29145 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29147 * void vst4_u8 (uint8_t *, uint8x8x4_t)
29148 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29150 * void vst4_s32 (int32_t *, int32x2x4_t)
29151 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29153 * void vst4_s16 (int16_t *, int16x4x4_t)
29154 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29156 * void vst4_s8 (int8_t *, int8x8x4_t)
29157 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29159 * void vst4_f32 (float32_t *, float32x2x4_t)
29160 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29162 * void vst4_p16 (poly16_t *, poly16x4x4_t)
29163 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29165 * void vst4_p8 (poly8_t *, poly8x8x4_t)
29166 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29168 * void vst4_u64 (uint64_t *, uint64x1x4_t)
29169 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
29171 * void vst4_s64 (int64_t *, int64x1x4_t)
29172 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
29174 * void vst4q_u32 (uint32_t *, uint32x4x4_t)
29175 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29177 * void vst4q_u16 (uint16_t *, uint16x8x4_t)
29178 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29180 * void vst4q_u8 (uint8_t *, uint8x16x4_t)
29181 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29183 * void vst4q_s32 (int32_t *, int32x4x4_t)
29184 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29186 * void vst4q_s16 (int16_t *, int16x8x4_t)
29187 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29189 * void vst4q_s8 (int8_t *, int8x16x4_t)
29190 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29192 * void vst4q_f32 (float32_t *, float32x4x4_t)
29193 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29195 * void vst4q_p16 (poly16_t *, poly16x8x4_t)
29196 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29198 * void vst4q_p8 (poly8_t *, poly8x16x4_t)
29199 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29201 * void vst4_lane_u32 (uint32_t *, uint32x2x4_t, const int)
29202 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29205 * void vst4_lane_u16 (uint16_t *, uint16x4x4_t, const int)
29206 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29209 * void vst4_lane_u8 (uint8_t *, uint8x8x4_t, const int)
29210 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29213 * void vst4_lane_s32 (int32_t *, int32x2x4_t, const int)
29214 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29217 * void vst4_lane_s16 (int16_t *, int16x4x4_t, const int)
29218 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29221 * void vst4_lane_s8 (int8_t *, int8x8x4_t, const int)
29222 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29225 * void vst4_lane_f32 (float32_t *, float32x2x4_t, const int)
29226 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29229 * void vst4_lane_p16 (poly16_t *, poly16x4x4_t, const int)
29230 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29233 * void vst4_lane_p8 (poly8_t *, poly8x8x4_t, const int)
29234 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29237 * void vst4q_lane_s32 (int32_t *, int32x4x4_t, const int)
29238 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29241 * void vst4q_lane_s16 (int16_t *, int16x8x4_t, const int)
29242 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29245 * void vst4q_lane_u32 (uint32_t *, uint32x4x4_t, const int)
29246 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29249 * void vst4q_lane_u16 (uint16_t *, uint16x8x4_t, const int)
29250 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29253 * void vst4q_lane_f32 (float32_t *, float32x4x4_t, const int)
29254 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29257 * void vst4q_lane_p16 (poly16_t *, poly16x8x4_t, const int)
29258 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29261 5.50.3.76 Logical operations (AND)
29262 ..................................
29264 * uint32x2_t vand_u32 (uint32x2_t, uint32x2_t)
29265 _Form of expected instruction(s):_ `vand D0, D0, D0'
29267 * uint16x4_t vand_u16 (uint16x4_t, uint16x4_t)
29268 _Form of expected instruction(s):_ `vand D0, D0, D0'
29270 * uint8x8_t vand_u8 (uint8x8_t, uint8x8_t)
29271 _Form of expected instruction(s):_ `vand D0, D0, D0'
29273 * int32x2_t vand_s32 (int32x2_t, int32x2_t)
29274 _Form of expected instruction(s):_ `vand D0, D0, D0'
29276 * int16x4_t vand_s16 (int16x4_t, int16x4_t)
29277 _Form of expected instruction(s):_ `vand D0, D0, D0'
29279 * int8x8_t vand_s8 (int8x8_t, int8x8_t)
29280 _Form of expected instruction(s):_ `vand D0, D0, D0'
29282 * uint64x1_t vand_u64 (uint64x1_t, uint64x1_t)
29283 _Form of expected instruction(s):_ `vand D0, D0, D0'
29285 * int64x1_t vand_s64 (int64x1_t, int64x1_t)
29286 _Form of expected instruction(s):_ `vand D0, D0, D0'
29288 * uint32x4_t vandq_u32 (uint32x4_t, uint32x4_t)
29289 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29291 * uint16x8_t vandq_u16 (uint16x8_t, uint16x8_t)
29292 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29294 * uint8x16_t vandq_u8 (uint8x16_t, uint8x16_t)
29295 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29297 * int32x4_t vandq_s32 (int32x4_t, int32x4_t)
29298 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29300 * int16x8_t vandq_s16 (int16x8_t, int16x8_t)
29301 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29303 * int8x16_t vandq_s8 (int8x16_t, int8x16_t)
29304 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29306 * uint64x2_t vandq_u64 (uint64x2_t, uint64x2_t)
29307 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29309 * int64x2_t vandq_s64 (int64x2_t, int64x2_t)
29310 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29312 5.50.3.77 Logical operations (OR)
29313 .................................
29315 * uint32x2_t vorr_u32 (uint32x2_t, uint32x2_t)
29316 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29318 * uint16x4_t vorr_u16 (uint16x4_t, uint16x4_t)
29319 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29321 * uint8x8_t vorr_u8 (uint8x8_t, uint8x8_t)
29322 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29324 * int32x2_t vorr_s32 (int32x2_t, int32x2_t)
29325 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29327 * int16x4_t vorr_s16 (int16x4_t, int16x4_t)
29328 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29330 * int8x8_t vorr_s8 (int8x8_t, int8x8_t)
29331 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29333 * uint64x1_t vorr_u64 (uint64x1_t, uint64x1_t)
29334 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29336 * int64x1_t vorr_s64 (int64x1_t, int64x1_t)
29337 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29339 * uint32x4_t vorrq_u32 (uint32x4_t, uint32x4_t)
29340 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29342 * uint16x8_t vorrq_u16 (uint16x8_t, uint16x8_t)
29343 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29345 * uint8x16_t vorrq_u8 (uint8x16_t, uint8x16_t)
29346 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29348 * int32x4_t vorrq_s32 (int32x4_t, int32x4_t)
29349 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29351 * int16x8_t vorrq_s16 (int16x8_t, int16x8_t)
29352 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29354 * int8x16_t vorrq_s8 (int8x16_t, int8x16_t)
29355 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29357 * uint64x2_t vorrq_u64 (uint64x2_t, uint64x2_t)
29358 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29360 * int64x2_t vorrq_s64 (int64x2_t, int64x2_t)
29361 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29363 5.50.3.78 Logical operations (exclusive OR)
29364 ...........................................
29366 * uint32x2_t veor_u32 (uint32x2_t, uint32x2_t)
29367 _Form of expected instruction(s):_ `veor D0, D0, D0'
29369 * uint16x4_t veor_u16 (uint16x4_t, uint16x4_t)
29370 _Form of expected instruction(s):_ `veor D0, D0, D0'
29372 * uint8x8_t veor_u8 (uint8x8_t, uint8x8_t)
29373 _Form of expected instruction(s):_ `veor D0, D0, D0'
29375 * int32x2_t veor_s32 (int32x2_t, int32x2_t)
29376 _Form of expected instruction(s):_ `veor D0, D0, D0'
29378 * int16x4_t veor_s16 (int16x4_t, int16x4_t)
29379 _Form of expected instruction(s):_ `veor D0, D0, D0'
29381 * int8x8_t veor_s8 (int8x8_t, int8x8_t)
29382 _Form of expected instruction(s):_ `veor D0, D0, D0'
29384 * uint64x1_t veor_u64 (uint64x1_t, uint64x1_t)
29385 _Form of expected instruction(s):_ `veor D0, D0, D0'
29387 * int64x1_t veor_s64 (int64x1_t, int64x1_t)
29388 _Form of expected instruction(s):_ `veor D0, D0, D0'
29390 * uint32x4_t veorq_u32 (uint32x4_t, uint32x4_t)
29391 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29393 * uint16x8_t veorq_u16 (uint16x8_t, uint16x8_t)
29394 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29396 * uint8x16_t veorq_u8 (uint8x16_t, uint8x16_t)
29397 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29399 * int32x4_t veorq_s32 (int32x4_t, int32x4_t)
29400 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29402 * int16x8_t veorq_s16 (int16x8_t, int16x8_t)
29403 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29405 * int8x16_t veorq_s8 (int8x16_t, int8x16_t)
29406 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29408 * uint64x2_t veorq_u64 (uint64x2_t, uint64x2_t)
29409 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29411 * int64x2_t veorq_s64 (int64x2_t, int64x2_t)
29412 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29414 5.50.3.79 Logical operations (AND-NOT)
29415 ......................................
29417 * uint32x2_t vbic_u32 (uint32x2_t, uint32x2_t)
29418 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29420 * uint16x4_t vbic_u16 (uint16x4_t, uint16x4_t)
29421 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29423 * uint8x8_t vbic_u8 (uint8x8_t, uint8x8_t)
29424 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29426 * int32x2_t vbic_s32 (int32x2_t, int32x2_t)
29427 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29429 * int16x4_t vbic_s16 (int16x4_t, int16x4_t)
29430 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29432 * int8x8_t vbic_s8 (int8x8_t, int8x8_t)
29433 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29435 * uint64x1_t vbic_u64 (uint64x1_t, uint64x1_t)
29436 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29438 * int64x1_t vbic_s64 (int64x1_t, int64x1_t)
29439 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29441 * uint32x4_t vbicq_u32 (uint32x4_t, uint32x4_t)
29442 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29444 * uint16x8_t vbicq_u16 (uint16x8_t, uint16x8_t)
29445 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29447 * uint8x16_t vbicq_u8 (uint8x16_t, uint8x16_t)
29448 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29450 * int32x4_t vbicq_s32 (int32x4_t, int32x4_t)
29451 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29453 * int16x8_t vbicq_s16 (int16x8_t, int16x8_t)
29454 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29456 * int8x16_t vbicq_s8 (int8x16_t, int8x16_t)
29457 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29459 * uint64x2_t vbicq_u64 (uint64x2_t, uint64x2_t)
29460 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29462 * int64x2_t vbicq_s64 (int64x2_t, int64x2_t)
29463 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29465 5.50.3.80 Logical operations (OR-NOT)
29466 .....................................
29468 * uint32x2_t vorn_u32 (uint32x2_t, uint32x2_t)
29469 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29471 * uint16x4_t vorn_u16 (uint16x4_t, uint16x4_t)
29472 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29474 * uint8x8_t vorn_u8 (uint8x8_t, uint8x8_t)
29475 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29477 * int32x2_t vorn_s32 (int32x2_t, int32x2_t)
29478 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29480 * int16x4_t vorn_s16 (int16x4_t, int16x4_t)
29481 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29483 * int8x8_t vorn_s8 (int8x8_t, int8x8_t)
29484 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29486 * uint64x1_t vorn_u64 (uint64x1_t, uint64x1_t)
29487 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29489 * int64x1_t vorn_s64 (int64x1_t, int64x1_t)
29490 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29492 * uint32x4_t vornq_u32 (uint32x4_t, uint32x4_t)
29493 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29495 * uint16x8_t vornq_u16 (uint16x8_t, uint16x8_t)
29496 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29498 * uint8x16_t vornq_u8 (uint8x16_t, uint8x16_t)
29499 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29501 * int32x4_t vornq_s32 (int32x4_t, int32x4_t)
29502 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29504 * int16x8_t vornq_s16 (int16x8_t, int16x8_t)
29505 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29507 * int8x16_t vornq_s8 (int8x16_t, int8x16_t)
29508 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29510 * uint64x2_t vornq_u64 (uint64x2_t, uint64x2_t)
29511 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29513 * int64x2_t vornq_s64 (int64x2_t, int64x2_t)
29514 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29516 5.50.3.81 Reinterpret casts
29517 ...........................
29519 * poly8x8_t vreinterpret_p8_u32 (uint32x2_t)
29521 * poly8x8_t vreinterpret_p8_u16 (uint16x4_t)
29523 * poly8x8_t vreinterpret_p8_u8 (uint8x8_t)
29525 * poly8x8_t vreinterpret_p8_s32 (int32x2_t)
29527 * poly8x8_t vreinterpret_p8_s16 (int16x4_t)
29529 * poly8x8_t vreinterpret_p8_s8 (int8x8_t)
29531 * poly8x8_t vreinterpret_p8_u64 (uint64x1_t)
29533 * poly8x8_t vreinterpret_p8_s64 (int64x1_t)
29535 * poly8x8_t vreinterpret_p8_f32 (float32x2_t)
29537 * poly8x8_t vreinterpret_p8_p16 (poly16x4_t)
29539 * poly8x16_t vreinterpretq_p8_u32 (uint32x4_t)
29541 * poly8x16_t vreinterpretq_p8_u16 (uint16x8_t)
29543 * poly8x16_t vreinterpretq_p8_u8 (uint8x16_t)
29545 * poly8x16_t vreinterpretq_p8_s32 (int32x4_t)
29547 * poly8x16_t vreinterpretq_p8_s16 (int16x8_t)
29549 * poly8x16_t vreinterpretq_p8_s8 (int8x16_t)
29551 * poly8x16_t vreinterpretq_p8_u64 (uint64x2_t)
29553 * poly8x16_t vreinterpretq_p8_s64 (int64x2_t)
29555 * poly8x16_t vreinterpretq_p8_f32 (float32x4_t)
29557 * poly8x16_t vreinterpretq_p8_p16 (poly16x8_t)
29559 * poly16x4_t vreinterpret_p16_u32 (uint32x2_t)
29561 * poly16x4_t vreinterpret_p16_u16 (uint16x4_t)
29563 * poly16x4_t vreinterpret_p16_u8 (uint8x8_t)
29565 * poly16x4_t vreinterpret_p16_s32 (int32x2_t)
29567 * poly16x4_t vreinterpret_p16_s16 (int16x4_t)
29569 * poly16x4_t vreinterpret_p16_s8 (int8x8_t)
29571 * poly16x4_t vreinterpret_p16_u64 (uint64x1_t)
29573 * poly16x4_t vreinterpret_p16_s64 (int64x1_t)
29575 * poly16x4_t vreinterpret_p16_f32 (float32x2_t)
29577 * poly16x4_t vreinterpret_p16_p8 (poly8x8_t)
29579 * poly16x8_t vreinterpretq_p16_u32 (uint32x4_t)
29581 * poly16x8_t vreinterpretq_p16_u16 (uint16x8_t)
29583 * poly16x8_t vreinterpretq_p16_u8 (uint8x16_t)
29585 * poly16x8_t vreinterpretq_p16_s32 (int32x4_t)
29587 * poly16x8_t vreinterpretq_p16_s16 (int16x8_t)
29589 * poly16x8_t vreinterpretq_p16_s8 (int8x16_t)
29591 * poly16x8_t vreinterpretq_p16_u64 (uint64x2_t)
29593 * poly16x8_t vreinterpretq_p16_s64 (int64x2_t)
29595 * poly16x8_t vreinterpretq_p16_f32 (float32x4_t)
29597 * poly16x8_t vreinterpretq_p16_p8 (poly8x16_t)
29599 * float32x2_t vreinterpret_f32_u32 (uint32x2_t)
29601 * float32x2_t vreinterpret_f32_u16 (uint16x4_t)
29603 * float32x2_t vreinterpret_f32_u8 (uint8x8_t)
29605 * float32x2_t vreinterpret_f32_s32 (int32x2_t)
29607 * float32x2_t vreinterpret_f32_s16 (int16x4_t)
29609 * float32x2_t vreinterpret_f32_s8 (int8x8_t)
29611 * float32x2_t vreinterpret_f32_u64 (uint64x1_t)
29613 * float32x2_t vreinterpret_f32_s64 (int64x1_t)
29615 * float32x2_t vreinterpret_f32_p16 (poly16x4_t)
29617 * float32x2_t vreinterpret_f32_p8 (poly8x8_t)
29619 * float32x4_t vreinterpretq_f32_u32 (uint32x4_t)
29621 * float32x4_t vreinterpretq_f32_u16 (uint16x8_t)
29623 * float32x4_t vreinterpretq_f32_u8 (uint8x16_t)
29625 * float32x4_t vreinterpretq_f32_s32 (int32x4_t)
29627 * float32x4_t vreinterpretq_f32_s16 (int16x8_t)
29629 * float32x4_t vreinterpretq_f32_s8 (int8x16_t)
29631 * float32x4_t vreinterpretq_f32_u64 (uint64x2_t)
29633 * float32x4_t vreinterpretq_f32_s64 (int64x2_t)
29635 * float32x4_t vreinterpretq_f32_p16 (poly16x8_t)
29637 * float32x4_t vreinterpretq_f32_p8 (poly8x16_t)
29639 * int64x1_t vreinterpret_s64_u32 (uint32x2_t)
29641 * int64x1_t vreinterpret_s64_u16 (uint16x4_t)
29643 * int64x1_t vreinterpret_s64_u8 (uint8x8_t)
29645 * int64x1_t vreinterpret_s64_s32 (int32x2_t)
29647 * int64x1_t vreinterpret_s64_s16 (int16x4_t)
29649 * int64x1_t vreinterpret_s64_s8 (int8x8_t)
29651 * int64x1_t vreinterpret_s64_u64 (uint64x1_t)
29653 * int64x1_t vreinterpret_s64_f32 (float32x2_t)
29655 * int64x1_t vreinterpret_s64_p16 (poly16x4_t)
29657 * int64x1_t vreinterpret_s64_p8 (poly8x8_t)
29659 * int64x2_t vreinterpretq_s64_u32 (uint32x4_t)
29661 * int64x2_t vreinterpretq_s64_u16 (uint16x8_t)
29663 * int64x2_t vreinterpretq_s64_u8 (uint8x16_t)
29665 * int64x2_t vreinterpretq_s64_s32 (int32x4_t)
29667 * int64x2_t vreinterpretq_s64_s16 (int16x8_t)
29669 * int64x2_t vreinterpretq_s64_s8 (int8x16_t)
29671 * int64x2_t vreinterpretq_s64_u64 (uint64x2_t)
29673 * int64x2_t vreinterpretq_s64_f32 (float32x4_t)
29675 * int64x2_t vreinterpretq_s64_p16 (poly16x8_t)
29677 * int64x2_t vreinterpretq_s64_p8 (poly8x16_t)
29679 * uint64x1_t vreinterpret_u64_u32 (uint32x2_t)
29681 * uint64x1_t vreinterpret_u64_u16 (uint16x4_t)
29683 * uint64x1_t vreinterpret_u64_u8 (uint8x8_t)
29685 * uint64x1_t vreinterpret_u64_s32 (int32x2_t)
29687 * uint64x1_t vreinterpret_u64_s16 (int16x4_t)
29689 * uint64x1_t vreinterpret_u64_s8 (int8x8_t)
29691 * uint64x1_t vreinterpret_u64_s64 (int64x1_t)
29693 * uint64x1_t vreinterpret_u64_f32 (float32x2_t)
29695 * uint64x1_t vreinterpret_u64_p16 (poly16x4_t)
29697 * uint64x1_t vreinterpret_u64_p8 (poly8x8_t)
29699 * uint64x2_t vreinterpretq_u64_u32 (uint32x4_t)
29701 * uint64x2_t vreinterpretq_u64_u16 (uint16x8_t)
29703 * uint64x2_t vreinterpretq_u64_u8 (uint8x16_t)
29705 * uint64x2_t vreinterpretq_u64_s32 (int32x4_t)
29707 * uint64x2_t vreinterpretq_u64_s16 (int16x8_t)
29709 * uint64x2_t vreinterpretq_u64_s8 (int8x16_t)
29711 * uint64x2_t vreinterpretq_u64_s64 (int64x2_t)
29713 * uint64x2_t vreinterpretq_u64_f32 (float32x4_t)
29715 * uint64x2_t vreinterpretq_u64_p16 (poly16x8_t)
29717 * uint64x2_t vreinterpretq_u64_p8 (poly8x16_t)
29719 * int8x8_t vreinterpret_s8_u32 (uint32x2_t)
29721 * int8x8_t vreinterpret_s8_u16 (uint16x4_t)
29723 * int8x8_t vreinterpret_s8_u8 (uint8x8_t)
29725 * int8x8_t vreinterpret_s8_s32 (int32x2_t)
29727 * int8x8_t vreinterpret_s8_s16 (int16x4_t)
29729 * int8x8_t vreinterpret_s8_u64 (uint64x1_t)
29731 * int8x8_t vreinterpret_s8_s64 (int64x1_t)
29733 * int8x8_t vreinterpret_s8_f32 (float32x2_t)
29735 * int8x8_t vreinterpret_s8_p16 (poly16x4_t)
29737 * int8x8_t vreinterpret_s8_p8 (poly8x8_t)
29739 * int8x16_t vreinterpretq_s8_u32 (uint32x4_t)
29741 * int8x16_t vreinterpretq_s8_u16 (uint16x8_t)
29743 * int8x16_t vreinterpretq_s8_u8 (uint8x16_t)
29745 * int8x16_t vreinterpretq_s8_s32 (int32x4_t)
29747 * int8x16_t vreinterpretq_s8_s16 (int16x8_t)
29749 * int8x16_t vreinterpretq_s8_u64 (uint64x2_t)
29751 * int8x16_t vreinterpretq_s8_s64 (int64x2_t)
29753 * int8x16_t vreinterpretq_s8_f32 (float32x4_t)
29755 * int8x16_t vreinterpretq_s8_p16 (poly16x8_t)
29757 * int8x16_t vreinterpretq_s8_p8 (poly8x16_t)
29759 * int16x4_t vreinterpret_s16_u32 (uint32x2_t)
29761 * int16x4_t vreinterpret_s16_u16 (uint16x4_t)
29763 * int16x4_t vreinterpret_s16_u8 (uint8x8_t)
29765 * int16x4_t vreinterpret_s16_s32 (int32x2_t)
29767 * int16x4_t vreinterpret_s16_s8 (int8x8_t)
29769 * int16x4_t vreinterpret_s16_u64 (uint64x1_t)
29771 * int16x4_t vreinterpret_s16_s64 (int64x1_t)
29773 * int16x4_t vreinterpret_s16_f32 (float32x2_t)
29775 * int16x4_t vreinterpret_s16_p16 (poly16x4_t)
29777 * int16x4_t vreinterpret_s16_p8 (poly8x8_t)
29779 * int16x8_t vreinterpretq_s16_u32 (uint32x4_t)
29781 * int16x8_t vreinterpretq_s16_u16 (uint16x8_t)
29783 * int16x8_t vreinterpretq_s16_u8 (uint8x16_t)
29785 * int16x8_t vreinterpretq_s16_s32 (int32x4_t)
29787 * int16x8_t vreinterpretq_s16_s8 (int8x16_t)
29789 * int16x8_t vreinterpretq_s16_u64 (uint64x2_t)
29791 * int16x8_t vreinterpretq_s16_s64 (int64x2_t)
29793 * int16x8_t vreinterpretq_s16_f32 (float32x4_t)
29795 * int16x8_t vreinterpretq_s16_p16 (poly16x8_t)
29797 * int16x8_t vreinterpretq_s16_p8 (poly8x16_t)
29799 * int32x2_t vreinterpret_s32_u32 (uint32x2_t)
29801 * int32x2_t vreinterpret_s32_u16 (uint16x4_t)
29803 * int32x2_t vreinterpret_s32_u8 (uint8x8_t)
29805 * int32x2_t vreinterpret_s32_s16 (int16x4_t)
29807 * int32x2_t vreinterpret_s32_s8 (int8x8_t)
29809 * int32x2_t vreinterpret_s32_u64 (uint64x1_t)
29811 * int32x2_t vreinterpret_s32_s64 (int64x1_t)
29813 * int32x2_t vreinterpret_s32_f32 (float32x2_t)
29815 * int32x2_t vreinterpret_s32_p16 (poly16x4_t)
29817 * int32x2_t vreinterpret_s32_p8 (poly8x8_t)
29819 * int32x4_t vreinterpretq_s32_u32 (uint32x4_t)
29821 * int32x4_t vreinterpretq_s32_u16 (uint16x8_t)
29823 * int32x4_t vreinterpretq_s32_u8 (uint8x16_t)
29825 * int32x4_t vreinterpretq_s32_s16 (int16x8_t)
29827 * int32x4_t vreinterpretq_s32_s8 (int8x16_t)
29829 * int32x4_t vreinterpretq_s32_u64 (uint64x2_t)
29831 * int32x4_t vreinterpretq_s32_s64 (int64x2_t)
29833 * int32x4_t vreinterpretq_s32_f32 (float32x4_t)
29835 * int32x4_t vreinterpretq_s32_p16 (poly16x8_t)
29837 * int32x4_t vreinterpretq_s32_p8 (poly8x16_t)
29839 * uint8x8_t vreinterpret_u8_u32 (uint32x2_t)
29841 * uint8x8_t vreinterpret_u8_u16 (uint16x4_t)
29843 * uint8x8_t vreinterpret_u8_s32 (int32x2_t)
29845 * uint8x8_t vreinterpret_u8_s16 (int16x4_t)
29847 * uint8x8_t vreinterpret_u8_s8 (int8x8_t)
29849 * uint8x8_t vreinterpret_u8_u64 (uint64x1_t)
29851 * uint8x8_t vreinterpret_u8_s64 (int64x1_t)
29853 * uint8x8_t vreinterpret_u8_f32 (float32x2_t)
29855 * uint8x8_t vreinterpret_u8_p16 (poly16x4_t)
29857 * uint8x8_t vreinterpret_u8_p8 (poly8x8_t)
29859 * uint8x16_t vreinterpretq_u8_u32 (uint32x4_t)
29861 * uint8x16_t vreinterpretq_u8_u16 (uint16x8_t)
29863 * uint8x16_t vreinterpretq_u8_s32 (int32x4_t)
29865 * uint8x16_t vreinterpretq_u8_s16 (int16x8_t)
29867 * uint8x16_t vreinterpretq_u8_s8 (int8x16_t)
29869 * uint8x16_t vreinterpretq_u8_u64 (uint64x2_t)
29871 * uint8x16_t vreinterpretq_u8_s64 (int64x2_t)
29873 * uint8x16_t vreinterpretq_u8_f32 (float32x4_t)
29875 * uint8x16_t vreinterpretq_u8_p16 (poly16x8_t)
29877 * uint8x16_t vreinterpretq_u8_p8 (poly8x16_t)
29879 * uint16x4_t vreinterpret_u16_u32 (uint32x2_t)
29881 * uint16x4_t vreinterpret_u16_u8 (uint8x8_t)
29883 * uint16x4_t vreinterpret_u16_s32 (int32x2_t)
29885 * uint16x4_t vreinterpret_u16_s16 (int16x4_t)
29887 * uint16x4_t vreinterpret_u16_s8 (int8x8_t)
29889 * uint16x4_t vreinterpret_u16_u64 (uint64x1_t)
29891 * uint16x4_t vreinterpret_u16_s64 (int64x1_t)
29893 * uint16x4_t vreinterpret_u16_f32 (float32x2_t)
29895 * uint16x4_t vreinterpret_u16_p16 (poly16x4_t)
29897 * uint16x4_t vreinterpret_u16_p8 (poly8x8_t)
29899 * uint16x8_t vreinterpretq_u16_u32 (uint32x4_t)
29901 * uint16x8_t vreinterpretq_u16_u8 (uint8x16_t)
29903 * uint16x8_t vreinterpretq_u16_s32 (int32x4_t)
29905 * uint16x8_t vreinterpretq_u16_s16 (int16x8_t)
29907 * uint16x8_t vreinterpretq_u16_s8 (int8x16_t)
29909 * uint16x8_t vreinterpretq_u16_u64 (uint64x2_t)
29911 * uint16x8_t vreinterpretq_u16_s64 (int64x2_t)
29913 * uint16x8_t vreinterpretq_u16_f32 (float32x4_t)
29915 * uint16x8_t vreinterpretq_u16_p16 (poly16x8_t)
29917 * uint16x8_t vreinterpretq_u16_p8 (poly8x16_t)
29919 * uint32x2_t vreinterpret_u32_u16 (uint16x4_t)
29921 * uint32x2_t vreinterpret_u32_u8 (uint8x8_t)
29923 * uint32x2_t vreinterpret_u32_s32 (int32x2_t)
29925 * uint32x2_t vreinterpret_u32_s16 (int16x4_t)
29927 * uint32x2_t vreinterpret_u32_s8 (int8x8_t)
29929 * uint32x2_t vreinterpret_u32_u64 (uint64x1_t)
29931 * uint32x2_t vreinterpret_u32_s64 (int64x1_t)
29933 * uint32x2_t vreinterpret_u32_f32 (float32x2_t)
29935 * uint32x2_t vreinterpret_u32_p16 (poly16x4_t)
29937 * uint32x2_t vreinterpret_u32_p8 (poly8x8_t)
29939 * uint32x4_t vreinterpretq_u32_u16 (uint16x8_t)
29941 * uint32x4_t vreinterpretq_u32_u8 (uint8x16_t)
29943 * uint32x4_t vreinterpretq_u32_s32 (int32x4_t)
29945 * uint32x4_t vreinterpretq_u32_s16 (int16x8_t)
29947 * uint32x4_t vreinterpretq_u32_s8 (int8x16_t)
29949 * uint32x4_t vreinterpretq_u32_u64 (uint64x2_t)
29951 * uint32x4_t vreinterpretq_u32_s64 (int64x2_t)
29953 * uint32x4_t vreinterpretq_u32_f32 (float32x4_t)
29955 * uint32x4_t vreinterpretq_u32_p16 (poly16x8_t)
29957 * uint32x4_t vreinterpretq_u32_p8 (poly8x16_t)
29960 File: gcc.info, Node: Blackfin Built-in Functions, Next: FR-V Built-in Functions, Prev: ARM NEON Intrinsics, Up: Target Builtins
29962 5.50.4 Blackfin Built-in Functions
29963 ----------------------------------
29965 Currently, there are two Blackfin-specific built-in functions. These
29966 are used for generating `CSYNC' and `SSYNC' machine insns without using
29967 inline assembly; by using these built-in functions the compiler can
29968 automatically add workarounds for hardware errata involving these
29969 instructions. These functions are named as follows:
29971 void __builtin_bfin_csync (void)
29972 void __builtin_bfin_ssync (void)
29975 File: gcc.info, Node: FR-V Built-in Functions, Next: X86 Built-in Functions, Prev: Blackfin Built-in Functions, Up: Target Builtins
29977 5.50.5 FR-V Built-in Functions
29978 ------------------------------
29980 GCC provides many FR-V-specific built-in functions. In general, these
29981 functions are intended to be compatible with those described by `FR-V
29982 Family, Softune C/C++ Compiler Manual (V6), Fujitsu Semiconductor'.
29983 The two exceptions are `__MDUNPACKH' and `__MBTOHE', the gcc forms of
29984 which pass 128-bit values by pointer rather than by value.
29986 Most of the functions are named after specific FR-V instructions.
29987 Such functions are said to be "directly mapped" and are summarized here
29993 * Directly-mapped Integer Functions::
29994 * Directly-mapped Media Functions::
29995 * Raw read/write Functions::
29996 * Other Built-in Functions::
29999 File: gcc.info, Node: Argument Types, Next: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
30001 5.50.5.1 Argument Types
30002 .......................
30004 The arguments to the built-in functions can be divided into three
30005 groups: register numbers, compile-time constants and run-time values.
30006 In order to make this classification clear at a glance, the arguments
30007 and return values are given the following pseudo types:
30009 Pseudo type Real C type Constant? Description
30010 `uh' `unsigned short' No an unsigned halfword
30011 `uw1' `unsigned int' No an unsigned word
30012 `sw1' `int' No a signed word
30013 `uw2' `unsigned long long' No an unsigned doubleword
30014 `sw2' `long long' No a signed doubleword
30015 `const' `int' Yes an integer constant
30016 `acc' `int' Yes an ACC register number
30017 `iacc' `int' Yes an IACC register number
30019 These pseudo types are not defined by GCC, they are simply a notational
30020 convenience used in this manual.
30022 Arguments of type `uh', `uw1', `sw1', `uw2' and `sw2' are evaluated at
30023 run time. They correspond to register operands in the underlying FR-V
30026 `const' arguments represent immediate operands in the underlying FR-V
30027 instructions. They must be compile-time constants.
30029 `acc' arguments are evaluated at compile time and specify the number
30030 of an accumulator register. For example, an `acc' argument of 2 will
30031 select the ACC2 register.
30033 `iacc' arguments are similar to `acc' arguments but specify the number
30034 of an IACC register. See *note Other Built-in Functions:: for more
30038 File: gcc.info, Node: Directly-mapped Integer Functions, Next: Directly-mapped Media Functions, Prev: Argument Types, Up: FR-V Built-in Functions
30040 5.50.5.2 Directly-mapped Integer Functions
30041 ..........................................
30043 The functions listed below map directly to FR-V I-type instructions.
30045 Function prototype Example usage Assembly output
30046 `sw1 __ADDSS (sw1, sw1)' `C = __ADDSS (A, B)' `ADDSS A,B,C'
30047 `sw1 __SCAN (sw1, sw1)' `C = __SCAN (A, B)' `SCAN A,B,C'
30048 `sw1 __SCUTSS (sw1)' `B = __SCUTSS (A)' `SCUTSS A,B'
30049 `sw1 __SLASS (sw1, sw1)' `C = __SLASS (A, B)' `SLASS A,B,C'
30050 `void __SMASS (sw1, sw1)' `__SMASS (A, B)' `SMASS A,B'
30051 `void __SMSSS (sw1, sw1)' `__SMSSS (A, B)' `SMSSS A,B'
30052 `void __SMU (sw1, sw1)' `__SMU (A, B)' `SMU A,B'
30053 `sw2 __SMUL (sw1, sw1)' `C = __SMUL (A, B)' `SMUL A,B,C'
30054 `sw1 __SUBSS (sw1, sw1)' `C = __SUBSS (A, B)' `SUBSS A,B,C'
30055 `uw2 __UMUL (uw1, uw1)' `C = __UMUL (A, B)' `UMUL A,B,C'
30058 File: gcc.info, Node: Directly-mapped Media Functions, Next: Raw read/write Functions, Prev: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
30060 5.50.5.3 Directly-mapped Media Functions
30061 ........................................
30063 The functions listed below map directly to FR-V M-type instructions.
30065 Function prototype Example usage Assembly output
30066 `uw1 __MABSHS (sw1)' `B = __MABSHS (A)' `MABSHS A,B'
30067 `void __MADDACCS (acc, acc)' `__MADDACCS (B, A)' `MADDACCS A,B'
30068 `sw1 __MADDHSS (sw1, sw1)' `C = __MADDHSS (A, B)' `MADDHSS A,B,C'
30069 `uw1 __MADDHUS (uw1, uw1)' `C = __MADDHUS (A, B)' `MADDHUS A,B,C'
30070 `uw1 __MAND (uw1, uw1)' `C = __MAND (A, B)' `MAND A,B,C'
30071 `void __MASACCS (acc, acc)' `__MASACCS (B, A)' `MASACCS A,B'
30072 `uw1 __MAVEH (uw1, uw1)' `C = __MAVEH (A, B)' `MAVEH A,B,C'
30073 `uw2 __MBTOH (uw1)' `B = __MBTOH (A)' `MBTOH A,B'
30074 `void __MBTOHE (uw1 *, uw1)' `__MBTOHE (&B, A)' `MBTOHE A,B'
30075 `void __MCLRACC (acc)' `__MCLRACC (A)' `MCLRACC A'
30076 `void __MCLRACCA (void)' `__MCLRACCA ()' `MCLRACCA'
30077 `uw1 __Mcop1 (uw1, uw1)' `C = __Mcop1 (A, B)' `Mcop1 A,B,C'
30078 `uw1 __Mcop2 (uw1, uw1)' `C = __Mcop2 (A, B)' `Mcop2 A,B,C'
30079 `uw1 __MCPLHI (uw2, const)' `C = __MCPLHI (A, B)' `MCPLHI A,#B,C'
30080 `uw1 __MCPLI (uw2, const)' `C = __MCPLI (A, B)' `MCPLI A,#B,C'
30081 `void __MCPXIS (acc, sw1, sw1)' `__MCPXIS (C, A, B)' `MCPXIS A,B,C'
30082 `void __MCPXIU (acc, uw1, uw1)' `__MCPXIU (C, A, B)' `MCPXIU A,B,C'
30083 `void __MCPXRS (acc, sw1, sw1)' `__MCPXRS (C, A, B)' `MCPXRS A,B,C'
30084 `void __MCPXRU (acc, uw1, uw1)' `__MCPXRU (C, A, B)' `MCPXRU A,B,C'
30085 `uw1 __MCUT (acc, uw1)' `C = __MCUT (A, B)' `MCUT A,B,C'
30086 `uw1 __MCUTSS (acc, sw1)' `C = __MCUTSS (A, B)' `MCUTSS A,B,C'
30087 `void __MDADDACCS (acc, acc)' `__MDADDACCS (B, A)' `MDADDACCS A,B'
30088 `void __MDASACCS (acc, acc)' `__MDASACCS (B, A)' `MDASACCS A,B'
30089 `uw2 __MDCUTSSI (acc, const)' `C = __MDCUTSSI (A, B)' `MDCUTSSI A,#B,C'
30090 `uw2 __MDPACKH (uw2, uw2)' `C = __MDPACKH (A, B)' `MDPACKH A,B,C'
30091 `uw2 __MDROTLI (uw2, const)' `C = __MDROTLI (A, B)' `MDROTLI A,#B,C'
30092 `void __MDSUBACCS (acc, acc)' `__MDSUBACCS (B, A)' `MDSUBACCS A,B'
30093 `void __MDUNPACKH (uw1 *, uw2)' `__MDUNPACKH (&B, A)' `MDUNPACKH A,B'
30094 `uw2 __MEXPDHD (uw1, const)' `C = __MEXPDHD (A, B)' `MEXPDHD A,#B,C'
30095 `uw1 __MEXPDHW (uw1, const)' `C = __MEXPDHW (A, B)' `MEXPDHW A,#B,C'
30096 `uw1 __MHDSETH (uw1, const)' `C = __MHDSETH (A, B)' `MHDSETH A,#B,C'
30097 `sw1 __MHDSETS (const)' `B = __MHDSETS (A)' `MHDSETS #A,B'
30098 `uw1 __MHSETHIH (uw1, const)' `B = __MHSETHIH (B, A)' `MHSETHIH #A,B'
30099 `sw1 __MHSETHIS (sw1, const)' `B = __MHSETHIS (B, A)' `MHSETHIS #A,B'
30100 `uw1 __MHSETLOH (uw1, const)' `B = __MHSETLOH (B, A)' `MHSETLOH #A,B'
30101 `sw1 __MHSETLOS (sw1, const)' `B = __MHSETLOS (B, A)' `MHSETLOS #A,B'
30102 `uw1 __MHTOB (uw2)' `B = __MHTOB (A)' `MHTOB A,B'
30103 `void __MMACHS (acc, sw1, sw1)' `__MMACHS (C, A, B)' `MMACHS A,B,C'
30104 `void __MMACHU (acc, uw1, uw1)' `__MMACHU (C, A, B)' `MMACHU A,B,C'
30105 `void __MMRDHS (acc, sw1, sw1)' `__MMRDHS (C, A, B)' `MMRDHS A,B,C'
30106 `void __MMRDHU (acc, uw1, uw1)' `__MMRDHU (C, A, B)' `MMRDHU A,B,C'
30107 `void __MMULHS (acc, sw1, sw1)' `__MMULHS (C, A, B)' `MMULHS A,B,C'
30108 `void __MMULHU (acc, uw1, uw1)' `__MMULHU (C, A, B)' `MMULHU A,B,C'
30109 `void __MMULXHS (acc, sw1, sw1)' `__MMULXHS (C, A, B)' `MMULXHS A,B,C'
30110 `void __MMULXHU (acc, uw1, uw1)' `__MMULXHU (C, A, B)' `MMULXHU A,B,C'
30111 `uw1 __MNOT (uw1)' `B = __MNOT (A)' `MNOT A,B'
30112 `uw1 __MOR (uw1, uw1)' `C = __MOR (A, B)' `MOR A,B,C'
30113 `uw1 __MPACKH (uh, uh)' `C = __MPACKH (A, B)' `MPACKH A,B,C'
30114 `sw2 __MQADDHSS (sw2, sw2)' `C = __MQADDHSS (A, B)' `MQADDHSS A,B,C'
30115 `uw2 __MQADDHUS (uw2, uw2)' `C = __MQADDHUS (A, B)' `MQADDHUS A,B,C'
30116 `void __MQCPXIS (acc, sw2, sw2)' `__MQCPXIS (C, A, B)' `MQCPXIS A,B,C'
30117 `void __MQCPXIU (acc, uw2, uw2)' `__MQCPXIU (C, A, B)' `MQCPXIU A,B,C'
30118 `void __MQCPXRS (acc, sw2, sw2)' `__MQCPXRS (C, A, B)' `MQCPXRS A,B,C'
30119 `void __MQCPXRU (acc, uw2, uw2)' `__MQCPXRU (C, A, B)' `MQCPXRU A,B,C'
30120 `sw2 __MQLCLRHS (sw2, sw2)' `C = __MQLCLRHS (A, B)' `MQLCLRHS A,B,C'
30121 `sw2 __MQLMTHS (sw2, sw2)' `C = __MQLMTHS (A, B)' `MQLMTHS A,B,C'
30122 `void __MQMACHS (acc, sw2, sw2)' `__MQMACHS (C, A, B)' `MQMACHS A,B,C'
30123 `void __MQMACHU (acc, uw2, uw2)' `__MQMACHU (C, A, B)' `MQMACHU A,B,C'
30124 `void __MQMACXHS (acc, sw2, `__MQMACXHS (C, A, B)' `MQMACXHS A,B,C'
30126 `void __MQMULHS (acc, sw2, sw2)' `__MQMULHS (C, A, B)' `MQMULHS A,B,C'
30127 `void __MQMULHU (acc, uw2, uw2)' `__MQMULHU (C, A, B)' `MQMULHU A,B,C'
30128 `void __MQMULXHS (acc, sw2, `__MQMULXHS (C, A, B)' `MQMULXHS A,B,C'
30130 `void __MQMULXHU (acc, uw2, `__MQMULXHU (C, A, B)' `MQMULXHU A,B,C'
30132 `sw2 __MQSATHS (sw2, sw2)' `C = __MQSATHS (A, B)' `MQSATHS A,B,C'
30133 `uw2 __MQSLLHI (uw2, int)' `C = __MQSLLHI (A, B)' `MQSLLHI A,B,C'
30134 `sw2 __MQSRAHI (sw2, int)' `C = __MQSRAHI (A, B)' `MQSRAHI A,B,C'
30135 `sw2 __MQSUBHSS (sw2, sw2)' `C = __MQSUBHSS (A, B)' `MQSUBHSS A,B,C'
30136 `uw2 __MQSUBHUS (uw2, uw2)' `C = __MQSUBHUS (A, B)' `MQSUBHUS A,B,C'
30137 `void __MQXMACHS (acc, sw2, `__MQXMACHS (C, A, B)' `MQXMACHS A,B,C'
30139 `void __MQXMACXHS (acc, sw2, `__MQXMACXHS (C, A, B)' `MQXMACXHS A,B,C'
30141 `uw1 __MRDACC (acc)' `B = __MRDACC (A)' `MRDACC A,B'
30142 `uw1 __MRDACCG (acc)' `B = __MRDACCG (A)' `MRDACCG A,B'
30143 `uw1 __MROTLI (uw1, const)' `C = __MROTLI (A, B)' `MROTLI A,#B,C'
30144 `uw1 __MROTRI (uw1, const)' `C = __MROTRI (A, B)' `MROTRI A,#B,C'
30145 `sw1 __MSATHS (sw1, sw1)' `C = __MSATHS (A, B)' `MSATHS A,B,C'
30146 `uw1 __MSATHU (uw1, uw1)' `C = __MSATHU (A, B)' `MSATHU A,B,C'
30147 `uw1 __MSLLHI (uw1, const)' `C = __MSLLHI (A, B)' `MSLLHI A,#B,C'
30148 `sw1 __MSRAHI (sw1, const)' `C = __MSRAHI (A, B)' `MSRAHI A,#B,C'
30149 `uw1 __MSRLHI (uw1, const)' `C = __MSRLHI (A, B)' `MSRLHI A,#B,C'
30150 `void __MSUBACCS (acc, acc)' `__MSUBACCS (B, A)' `MSUBACCS A,B'
30151 `sw1 __MSUBHSS (sw1, sw1)' `C = __MSUBHSS (A, B)' `MSUBHSS A,B,C'
30152 `uw1 __MSUBHUS (uw1, uw1)' `C = __MSUBHUS (A, B)' `MSUBHUS A,B,C'
30153 `void __MTRAP (void)' `__MTRAP ()' `MTRAP'
30154 `uw2 __MUNPACKH (uw1)' `B = __MUNPACKH (A)' `MUNPACKH A,B'
30155 `uw1 __MWCUT (uw2, uw1)' `C = __MWCUT (A, B)' `MWCUT A,B,C'
30156 `void __MWTACC (acc, uw1)' `__MWTACC (B, A)' `MWTACC A,B'
30157 `void __MWTACCG (acc, uw1)' `__MWTACCG (B, A)' `MWTACCG A,B'
30158 `uw1 __MXOR (uw1, uw1)' `C = __MXOR (A, B)' `MXOR A,B,C'
30161 File: gcc.info, Node: Raw read/write Functions, Next: Other Built-in Functions, Prev: Directly-mapped Media Functions, Up: FR-V Built-in Functions
30163 5.50.5.4 Raw read/write Functions
30164 .................................
30166 This sections describes built-in functions related to read and write
30167 instructions to access memory. These functions generate `membar'
30168 instructions to flush the I/O load and stores where appropriate, as
30169 described in Fujitsu's manual described above.
30171 `unsigned char __builtin_read8 (void *DATA)'
30173 `unsigned short __builtin_read16 (void *DATA)'
30175 `unsigned long __builtin_read32 (void *DATA)'
30177 `unsigned long long __builtin_read64 (void *DATA)'
30179 `void __builtin_write8 (void *DATA, unsigned char DATUM)'
30181 `void __builtin_write16 (void *DATA, unsigned short DATUM)'
30183 `void __builtin_write32 (void *DATA, unsigned long DATUM)'
30185 `void __builtin_write64 (void *DATA, unsigned long long DATUM)'
30188 File: gcc.info, Node: Other Built-in Functions, Prev: Raw read/write Functions, Up: FR-V Built-in Functions
30190 5.50.5.5 Other Built-in Functions
30191 .................................
30193 This section describes built-in functions that are not named after a
30194 specific FR-V instruction.
30196 `sw2 __IACCreadll (iacc REG)'
30197 Return the full 64-bit value of IACC0. The REG argument is
30198 reserved for future expansion and must be 0.
30200 `sw1 __IACCreadl (iacc REG)'
30201 Return the value of IACC0H if REG is 0 and IACC0L if REG is 1.
30202 Other values of REG are rejected as invalid.
30204 `void __IACCsetll (iacc REG, sw2 X)'
30205 Set the full 64-bit value of IACC0 to X. The REG argument is
30206 reserved for future expansion and must be 0.
30208 `void __IACCsetl (iacc REG, sw1 X)'
30209 Set IACC0H to X if REG is 0 and IACC0L to X if REG is 1. Other
30210 values of REG are rejected as invalid.
30212 `void __data_prefetch0 (const void *X)'
30213 Use the `dcpl' instruction to load the contents of address X into
30216 `void __data_prefetch (const void *X)'
30217 Use the `nldub' instruction to load the contents of address X into
30218 the data cache. The instruction will be issued in slot I1.
30221 File: gcc.info, Node: X86 Built-in Functions, Next: MIPS DSP Built-in Functions, Prev: FR-V Built-in Functions, Up: Target Builtins
30223 5.50.6 X86 Built-in Functions
30224 -----------------------------
30226 These built-in functions are available for the i386 and x86-64 family
30227 of computers, depending on the command-line switches used.
30229 Note that, if you specify command-line switches such as `-msse', the
30230 compiler could use the extended instruction sets even if the built-ins
30231 are not used explicitly in the program. For this reason, applications
30232 which perform runtime CPU detection must compile separate files for each
30233 supported architecture, using the appropriate flags. In particular,
30234 the file containing the CPU detection code should be compiled without
30237 The following machine modes are available for use with MMX built-in
30238 functions (*note Vector Extensions::): `V2SI' for a vector of two
30239 32-bit integers, `V4HI' for a vector of four 16-bit integers, and
30240 `V8QI' for a vector of eight 8-bit integers. Some of the built-in
30241 functions operate on MMX registers as a whole 64-bit entity, these use
30242 `V1DI' as their mode.
30244 If 3Dnow extensions are enabled, `V2SF' is used as a mode for a vector
30245 of two 32-bit floating point values.
30247 If SSE extensions are enabled, `V4SF' is used for a vector of four
30248 32-bit floating point values. Some instructions use a vector of four
30249 32-bit integers, these use `V4SI'. Finally, some instructions operate
30250 on an entire vector register, interpreting it as a 128-bit integer,
30251 these use mode `TI'.
30253 In 64-bit mode, the x86-64 family of processors uses additional
30254 built-in functions for efficient use of `TF' (`__float128') 128-bit
30255 floating point and `TC' 128-bit complex floating point values.
30257 The following floating point built-in functions are available in 64-bit
30258 mode. All of them implement the function that is part of the name.
30260 __float128 __builtin_fabsq (__float128)
30261 __float128 __builtin_copysignq (__float128, __float128)
30263 The following floating point built-in functions are made available in
30266 `__float128 __builtin_infq (void)'
30267 Similar to `__builtin_inf', except the return type is `__float128'.
30269 The following built-in functions are made available by `-mmmx'. All
30270 of them generate the machine instruction that is part of the name.
30272 v8qi __builtin_ia32_paddb (v8qi, v8qi)
30273 v4hi __builtin_ia32_paddw (v4hi, v4hi)
30274 v2si __builtin_ia32_paddd (v2si, v2si)
30275 v8qi __builtin_ia32_psubb (v8qi, v8qi)
30276 v4hi __builtin_ia32_psubw (v4hi, v4hi)
30277 v2si __builtin_ia32_psubd (v2si, v2si)
30278 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
30279 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
30280 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
30281 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
30282 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
30283 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
30284 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
30285 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
30286 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
30287 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
30288 di __builtin_ia32_pand (di, di)
30289 di __builtin_ia32_pandn (di,di)
30290 di __builtin_ia32_por (di, di)
30291 di __builtin_ia32_pxor (di, di)
30292 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
30293 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
30294 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
30295 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
30296 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
30297 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
30298 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
30299 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
30300 v2si __builtin_ia32_punpckhdq (v2si, v2si)
30301 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
30302 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
30303 v2si __builtin_ia32_punpckldq (v2si, v2si)
30304 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
30305 v4hi __builtin_ia32_packssdw (v2si, v2si)
30306 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
30308 v4hi __builtin_ia32_psllw (v4hi, v4hi)
30309 v2si __builtin_ia32_pslld (v2si, v2si)
30310 v1di __builtin_ia32_psllq (v1di, v1di)
30311 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
30312 v2si __builtin_ia32_psrld (v2si, v2si)
30313 v1di __builtin_ia32_psrlq (v1di, v1di)
30314 v4hi __builtin_ia32_psraw (v4hi, v4hi)
30315 v2si __builtin_ia32_psrad (v2si, v2si)
30316 v4hi __builtin_ia32_psllwi (v4hi, int)
30317 v2si __builtin_ia32_pslldi (v2si, int)
30318 v1di __builtin_ia32_psllqi (v1di, int)
30319 v4hi __builtin_ia32_psrlwi (v4hi, int)
30320 v2si __builtin_ia32_psrldi (v2si, int)
30321 v1di __builtin_ia32_psrlqi (v1di, int)
30322 v4hi __builtin_ia32_psrawi (v4hi, int)
30323 v2si __builtin_ia32_psradi (v2si, int)
30325 The following built-in functions are made available either with
30326 `-msse', or with a combination of `-m3dnow' and `-march=athlon'. All
30327 of them generate the machine instruction that is part of the name.
30329 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
30330 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
30331 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
30332 v1di __builtin_ia32_psadbw (v8qi, v8qi)
30333 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
30334 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
30335 v8qi __builtin_ia32_pminub (v8qi, v8qi)
30336 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
30337 int __builtin_ia32_pextrw (v4hi, int)
30338 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
30339 int __builtin_ia32_pmovmskb (v8qi)
30340 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
30341 void __builtin_ia32_movntq (di *, di)
30342 void __builtin_ia32_sfence (void)
30344 The following built-in functions are available when `-msse' is used.
30345 All of them generate the machine instruction that is part of the name.
30347 int __builtin_ia32_comieq (v4sf, v4sf)
30348 int __builtin_ia32_comineq (v4sf, v4sf)
30349 int __builtin_ia32_comilt (v4sf, v4sf)
30350 int __builtin_ia32_comile (v4sf, v4sf)
30351 int __builtin_ia32_comigt (v4sf, v4sf)
30352 int __builtin_ia32_comige (v4sf, v4sf)
30353 int __builtin_ia32_ucomieq (v4sf, v4sf)
30354 int __builtin_ia32_ucomineq (v4sf, v4sf)
30355 int __builtin_ia32_ucomilt (v4sf, v4sf)
30356 int __builtin_ia32_ucomile (v4sf, v4sf)
30357 int __builtin_ia32_ucomigt (v4sf, v4sf)
30358 int __builtin_ia32_ucomige (v4sf, v4sf)
30359 v4sf __builtin_ia32_addps (v4sf, v4sf)
30360 v4sf __builtin_ia32_subps (v4sf, v4sf)
30361 v4sf __builtin_ia32_mulps (v4sf, v4sf)
30362 v4sf __builtin_ia32_divps (v4sf, v4sf)
30363 v4sf __builtin_ia32_addss (v4sf, v4sf)
30364 v4sf __builtin_ia32_subss (v4sf, v4sf)
30365 v4sf __builtin_ia32_mulss (v4sf, v4sf)
30366 v4sf __builtin_ia32_divss (v4sf, v4sf)
30367 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
30368 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
30369 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
30370 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
30371 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
30372 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
30373 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
30374 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
30375 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
30376 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
30377 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
30378 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
30379 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
30380 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
30381 v4si __builtin_ia32_cmpless (v4sf, v4sf)
30382 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
30383 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
30384 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
30385 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
30386 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
30387 v4sf __builtin_ia32_maxps (v4sf, v4sf)
30388 v4sf __builtin_ia32_maxss (v4sf, v4sf)
30389 v4sf __builtin_ia32_minps (v4sf, v4sf)
30390 v4sf __builtin_ia32_minss (v4sf, v4sf)
30391 v4sf __builtin_ia32_andps (v4sf, v4sf)
30392 v4sf __builtin_ia32_andnps (v4sf, v4sf)
30393 v4sf __builtin_ia32_orps (v4sf, v4sf)
30394 v4sf __builtin_ia32_xorps (v4sf, v4sf)
30395 v4sf __builtin_ia32_movss (v4sf, v4sf)
30396 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
30397 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
30398 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
30399 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
30400 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
30401 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
30402 v2si __builtin_ia32_cvtps2pi (v4sf)
30403 int __builtin_ia32_cvtss2si (v4sf)
30404 v2si __builtin_ia32_cvttps2pi (v4sf)
30405 int __builtin_ia32_cvttss2si (v4sf)
30406 v4sf __builtin_ia32_rcpps (v4sf)
30407 v4sf __builtin_ia32_rsqrtps (v4sf)
30408 v4sf __builtin_ia32_sqrtps (v4sf)
30409 v4sf __builtin_ia32_rcpss (v4sf)
30410 v4sf __builtin_ia32_rsqrtss (v4sf)
30411 v4sf __builtin_ia32_sqrtss (v4sf)
30412 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
30413 void __builtin_ia32_movntps (float *, v4sf)
30414 int __builtin_ia32_movmskps (v4sf)
30416 The following built-in functions are available when `-msse' is used.
30418 `v4sf __builtin_ia32_loadaps (float *)'
30419 Generates the `movaps' machine instruction as a load from memory.
30421 `void __builtin_ia32_storeaps (float *, v4sf)'
30422 Generates the `movaps' machine instruction as a store to memory.
30424 `v4sf __builtin_ia32_loadups (float *)'
30425 Generates the `movups' machine instruction as a load from memory.
30427 `void __builtin_ia32_storeups (float *, v4sf)'
30428 Generates the `movups' machine instruction as a store to memory.
30430 `v4sf __builtin_ia32_loadsss (float *)'
30431 Generates the `movss' machine instruction as a load from memory.
30433 `void __builtin_ia32_storess (float *, v4sf)'
30434 Generates the `movss' machine instruction as a store to memory.
30436 `v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)'
30437 Generates the `movhps' machine instruction as a load from memory.
30439 `v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)'
30440 Generates the `movlps' machine instruction as a load from memory
30442 `void __builtin_ia32_storehps (v2sf *, v4sf)'
30443 Generates the `movhps' machine instruction as a store to memory.
30445 `void __builtin_ia32_storelps (v2sf *, v4sf)'
30446 Generates the `movlps' machine instruction as a store to memory.
30448 The following built-in functions are available when `-msse2' is used.
30449 All of them generate the machine instruction that is part of the name.
30451 int __builtin_ia32_comisdeq (v2df, v2df)
30452 int __builtin_ia32_comisdlt (v2df, v2df)
30453 int __builtin_ia32_comisdle (v2df, v2df)
30454 int __builtin_ia32_comisdgt (v2df, v2df)
30455 int __builtin_ia32_comisdge (v2df, v2df)
30456 int __builtin_ia32_comisdneq (v2df, v2df)
30457 int __builtin_ia32_ucomisdeq (v2df, v2df)
30458 int __builtin_ia32_ucomisdlt (v2df, v2df)
30459 int __builtin_ia32_ucomisdle (v2df, v2df)
30460 int __builtin_ia32_ucomisdgt (v2df, v2df)
30461 int __builtin_ia32_ucomisdge (v2df, v2df)
30462 int __builtin_ia32_ucomisdneq (v2df, v2df)
30463 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
30464 v2df __builtin_ia32_cmpltpd (v2df, v2df)
30465 v2df __builtin_ia32_cmplepd (v2df, v2df)
30466 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
30467 v2df __builtin_ia32_cmpgepd (v2df, v2df)
30468 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
30469 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
30470 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
30471 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
30472 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
30473 v2df __builtin_ia32_cmpngepd (v2df, v2df)
30474 v2df __builtin_ia32_cmpordpd (v2df, v2df)
30475 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
30476 v2df __builtin_ia32_cmpltsd (v2df, v2df)
30477 v2df __builtin_ia32_cmplesd (v2df, v2df)
30478 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
30479 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
30480 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
30481 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
30482 v2df __builtin_ia32_cmpordsd (v2df, v2df)
30483 v2di __builtin_ia32_paddq (v2di, v2di)
30484 v2di __builtin_ia32_psubq (v2di, v2di)
30485 v2df __builtin_ia32_addpd (v2df, v2df)
30486 v2df __builtin_ia32_subpd (v2df, v2df)
30487 v2df __builtin_ia32_mulpd (v2df, v2df)
30488 v2df __builtin_ia32_divpd (v2df, v2df)
30489 v2df __builtin_ia32_addsd (v2df, v2df)
30490 v2df __builtin_ia32_subsd (v2df, v2df)
30491 v2df __builtin_ia32_mulsd (v2df, v2df)
30492 v2df __builtin_ia32_divsd (v2df, v2df)
30493 v2df __builtin_ia32_minpd (v2df, v2df)
30494 v2df __builtin_ia32_maxpd (v2df, v2df)
30495 v2df __builtin_ia32_minsd (v2df, v2df)
30496 v2df __builtin_ia32_maxsd (v2df, v2df)
30497 v2df __builtin_ia32_andpd (v2df, v2df)
30498 v2df __builtin_ia32_andnpd (v2df, v2df)
30499 v2df __builtin_ia32_orpd (v2df, v2df)
30500 v2df __builtin_ia32_xorpd (v2df, v2df)
30501 v2df __builtin_ia32_movsd (v2df, v2df)
30502 v2df __builtin_ia32_unpckhpd (v2df, v2df)
30503 v2df __builtin_ia32_unpcklpd (v2df, v2df)
30504 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
30505 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
30506 v4si __builtin_ia32_paddd128 (v4si, v4si)
30507 v2di __builtin_ia32_paddq128 (v2di, v2di)
30508 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
30509 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
30510 v4si __builtin_ia32_psubd128 (v4si, v4si)
30511 v2di __builtin_ia32_psubq128 (v2di, v2di)
30512 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
30513 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
30514 v2di __builtin_ia32_pand128 (v2di, v2di)
30515 v2di __builtin_ia32_pandn128 (v2di, v2di)
30516 v2di __builtin_ia32_por128 (v2di, v2di)
30517 v2di __builtin_ia32_pxor128 (v2di, v2di)
30518 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
30519 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
30520 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
30521 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
30522 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
30523 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
30524 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
30525 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
30526 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
30527 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
30528 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
30529 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
30530 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
30531 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
30532 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
30533 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
30534 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
30535 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
30536 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
30537 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
30538 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
30539 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
30540 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
30541 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
30542 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
30543 v2df __builtin_ia32_loadupd (double *)
30544 void __builtin_ia32_storeupd (double *, v2df)
30545 v2df __builtin_ia32_loadhpd (v2df, double const *)
30546 v2df __builtin_ia32_loadlpd (v2df, double const *)
30547 int __builtin_ia32_movmskpd (v2df)
30548 int __builtin_ia32_pmovmskb128 (v16qi)
30549 void __builtin_ia32_movnti (int *, int)
30550 void __builtin_ia32_movntpd (double *, v2df)
30551 void __builtin_ia32_movntdq (v2df *, v2df)
30552 v4si __builtin_ia32_pshufd (v4si, int)
30553 v8hi __builtin_ia32_pshuflw (v8hi, int)
30554 v8hi __builtin_ia32_pshufhw (v8hi, int)
30555 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
30556 v2df __builtin_ia32_sqrtpd (v2df)
30557 v2df __builtin_ia32_sqrtsd (v2df)
30558 v2df __builtin_ia32_shufpd (v2df, v2df, int)
30559 v2df __builtin_ia32_cvtdq2pd (v4si)
30560 v4sf __builtin_ia32_cvtdq2ps (v4si)
30561 v4si __builtin_ia32_cvtpd2dq (v2df)
30562 v2si __builtin_ia32_cvtpd2pi (v2df)
30563 v4sf __builtin_ia32_cvtpd2ps (v2df)
30564 v4si __builtin_ia32_cvttpd2dq (v2df)
30565 v2si __builtin_ia32_cvttpd2pi (v2df)
30566 v2df __builtin_ia32_cvtpi2pd (v2si)
30567 int __builtin_ia32_cvtsd2si (v2df)
30568 int __builtin_ia32_cvttsd2si (v2df)
30569 long long __builtin_ia32_cvtsd2si64 (v2df)
30570 long long __builtin_ia32_cvttsd2si64 (v2df)
30571 v4si __builtin_ia32_cvtps2dq (v4sf)
30572 v2df __builtin_ia32_cvtps2pd (v4sf)
30573 v4si __builtin_ia32_cvttps2dq (v4sf)
30574 v2df __builtin_ia32_cvtsi2sd (v2df, int)
30575 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
30576 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
30577 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
30578 void __builtin_ia32_clflush (const void *)
30579 void __builtin_ia32_lfence (void)
30580 void __builtin_ia32_mfence (void)
30581 v16qi __builtin_ia32_loaddqu (const char *)
30582 void __builtin_ia32_storedqu (char *, v16qi)
30583 v1di __builtin_ia32_pmuludq (v2si, v2si)
30584 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
30585 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
30586 v4si __builtin_ia32_pslld128 (v4si, v4si)
30587 v2di __builtin_ia32_psllq128 (v2di, v2di)
30588 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
30589 v4si __builtin_ia32_psrld128 (v4si, v4si)
30590 v2di __builtin_ia32_psrlq128 (v2di, v2di)
30591 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
30592 v4si __builtin_ia32_psrad128 (v4si, v4si)
30593 v2di __builtin_ia32_pslldqi128 (v2di, int)
30594 v8hi __builtin_ia32_psllwi128 (v8hi, int)
30595 v4si __builtin_ia32_pslldi128 (v4si, int)
30596 v2di __builtin_ia32_psllqi128 (v2di, int)
30597 v2di __builtin_ia32_psrldqi128 (v2di, int)
30598 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
30599 v4si __builtin_ia32_psrldi128 (v4si, int)
30600 v2di __builtin_ia32_psrlqi128 (v2di, int)
30601 v8hi __builtin_ia32_psrawi128 (v8hi, int)
30602 v4si __builtin_ia32_psradi128 (v4si, int)
30603 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
30604 v2di __builtin_ia32_movq128 (v2di)
30606 The following built-in functions are available when `-msse3' is used.
30607 All of them generate the machine instruction that is part of the name.
30609 v2df __builtin_ia32_addsubpd (v2df, v2df)
30610 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
30611 v2df __builtin_ia32_haddpd (v2df, v2df)
30612 v4sf __builtin_ia32_haddps (v4sf, v4sf)
30613 v2df __builtin_ia32_hsubpd (v2df, v2df)
30614 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
30615 v16qi __builtin_ia32_lddqu (char const *)
30616 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
30617 v2df __builtin_ia32_movddup (v2df)
30618 v4sf __builtin_ia32_movshdup (v4sf)
30619 v4sf __builtin_ia32_movsldup (v4sf)
30620 void __builtin_ia32_mwait (unsigned int, unsigned int)
30622 The following built-in functions are available when `-msse3' is used.
30624 `v2df __builtin_ia32_loadddup (double const *)'
30625 Generates the `movddup' machine instruction as a load from memory.
30627 The following built-in functions are available when `-mssse3' is used.
30628 All of them generate the machine instruction that is part of the name
30629 with MMX registers.
30631 v2si __builtin_ia32_phaddd (v2si, v2si)
30632 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
30633 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
30634 v2si __builtin_ia32_phsubd (v2si, v2si)
30635 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
30636 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
30637 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
30638 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
30639 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
30640 v8qi __builtin_ia32_psignb (v8qi, v8qi)
30641 v2si __builtin_ia32_psignd (v2si, v2si)
30642 v4hi __builtin_ia32_psignw (v4hi, v4hi)
30643 v1di __builtin_ia32_palignr (v1di, v1di, int)
30644 v8qi __builtin_ia32_pabsb (v8qi)
30645 v2si __builtin_ia32_pabsd (v2si)
30646 v4hi __builtin_ia32_pabsw (v4hi)
30648 The following built-in functions are available when `-mssse3' is used.
30649 All of them generate the machine instruction that is part of the name
30650 with SSE registers.
30652 v4si __builtin_ia32_phaddd128 (v4si, v4si)
30653 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
30654 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
30655 v4si __builtin_ia32_phsubd128 (v4si, v4si)
30656 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
30657 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
30658 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
30659 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
30660 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
30661 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
30662 v4si __builtin_ia32_psignd128 (v4si, v4si)
30663 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
30664 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
30665 v16qi __builtin_ia32_pabsb128 (v16qi)
30666 v4si __builtin_ia32_pabsd128 (v4si)
30667 v8hi __builtin_ia32_pabsw128 (v8hi)
30669 The following built-in functions are available when `-msse4.1' is
30670 used. All of them generate the machine instruction that is part of the
30673 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
30674 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
30675 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
30676 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
30677 v2df __builtin_ia32_dppd (v2df, v2df, const int)
30678 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
30679 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
30680 v2di __builtin_ia32_movntdqa (v2di *);
30681 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
30682 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
30683 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
30684 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
30685 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
30686 v8hi __builtin_ia32_phminposuw128 (v8hi)
30687 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
30688 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
30689 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
30690 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
30691 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
30692 v4si __builtin_ia32_pminsd128 (v4si, v4si)
30693 v4si __builtin_ia32_pminud128 (v4si, v4si)
30694 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
30695 v4si __builtin_ia32_pmovsxbd128 (v16qi)
30696 v2di __builtin_ia32_pmovsxbq128 (v16qi)
30697 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
30698 v2di __builtin_ia32_pmovsxdq128 (v4si)
30699 v4si __builtin_ia32_pmovsxwd128 (v8hi)
30700 v2di __builtin_ia32_pmovsxwq128 (v8hi)
30701 v4si __builtin_ia32_pmovzxbd128 (v16qi)
30702 v2di __builtin_ia32_pmovzxbq128 (v16qi)
30703 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
30704 v2di __builtin_ia32_pmovzxdq128 (v4si)
30705 v4si __builtin_ia32_pmovzxwd128 (v8hi)
30706 v2di __builtin_ia32_pmovzxwq128 (v8hi)
30707 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
30708 v4si __builtin_ia32_pmulld128 (v4si, v4si)
30709 int __builtin_ia32_ptestc128 (v2di, v2di)
30710 int __builtin_ia32_ptestnzc128 (v2di, v2di)
30711 int __builtin_ia32_ptestz128 (v2di, v2di)
30712 v2df __builtin_ia32_roundpd (v2df, const int)
30713 v4sf __builtin_ia32_roundps (v4sf, const int)
30714 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
30715 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
30717 The following built-in functions are available when `-msse4.1' is used.
30719 `v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)'
30720 Generates the `insertps' machine instruction.
30722 `int __builtin_ia32_vec_ext_v16qi (v16qi, const int)'
30723 Generates the `pextrb' machine instruction.
30725 `v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)'
30726 Generates the `pinsrb' machine instruction.
30728 `v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)'
30729 Generates the `pinsrd' machine instruction.
30731 `v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)'
30732 Generates the `pinsrq' machine instruction in 64bit mode.
30734 The following built-in functions are changed to generate new SSE4.1
30735 instructions when `-msse4.1' is used.
30737 `float __builtin_ia32_vec_ext_v4sf (v4sf, const int)'
30738 Generates the `extractps' machine instruction.
30740 `int __builtin_ia32_vec_ext_v4si (v4si, const int)'
30741 Generates the `pextrd' machine instruction.
30743 `long long __builtin_ia32_vec_ext_v2di (v2di, const int)'
30744 Generates the `pextrq' machine instruction in 64bit mode.
30746 The following built-in functions are available when `-msse4.2' is
30747 used. All of them generate the machine instruction that is part of the
30750 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
30751 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
30752 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
30753 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
30754 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
30755 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
30756 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
30757 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
30758 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
30759 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
30760 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
30761 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
30762 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
30763 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
30764 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
30766 The following built-in functions are available when `-msse4.2' is used.
30768 `unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)'
30769 Generates the `crc32b' machine instruction.
30771 `unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)'
30772 Generates the `crc32w' machine instruction.
30774 `unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)'
30775 Generates the `crc32l' machine instruction.
30777 `unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)'
30779 The following built-in functions are changed to generate new SSE4.2
30780 instructions when `-msse4.2' is used.
30782 `int __builtin_popcount (unsigned int)'
30783 Generates the `popcntl' machine instruction.
30785 `int __builtin_popcountl (unsigned long)'
30786 Generates the `popcntl' or `popcntq' machine instruction,
30787 depending on the size of `unsigned long'.
30789 `int __builtin_popcountll (unsigned long long)'
30790 Generates the `popcntq' machine instruction.
30792 The following built-in functions are available when `-mavx' is used.
30793 All of them generate the machine instruction that is part of the name.
30795 v4df __builtin_ia32_addpd256 (v4df,v4df)
30796 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
30797 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
30798 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
30799 v4df __builtin_ia32_andnpd256 (v4df,v4df)
30800 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
30801 v4df __builtin_ia32_andpd256 (v4df,v4df)
30802 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
30803 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
30804 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
30805 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
30806 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
30807 v2df __builtin_ia32_cmppd (v2df,v2df,int)
30808 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
30809 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
30810 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
30811 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
30812 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
30813 v4df __builtin_ia32_cvtdq2pd256 (v4si)
30814 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
30815 v4si __builtin_ia32_cvtpd2dq256 (v4df)
30816 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
30817 v8si __builtin_ia32_cvtps2dq256 (v8sf)
30818 v4df __builtin_ia32_cvtps2pd256 (v4sf)
30819 v4si __builtin_ia32_cvttpd2dq256 (v4df)
30820 v8si __builtin_ia32_cvttps2dq256 (v8sf)
30821 v4df __builtin_ia32_divpd256 (v4df,v4df)
30822 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
30823 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
30824 v4df __builtin_ia32_haddpd256 (v4df,v4df)
30825 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
30826 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
30827 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
30828 v32qi __builtin_ia32_lddqu256 (pcchar)
30829 v32qi __builtin_ia32_loaddqu256 (pcchar)
30830 v4df __builtin_ia32_loadupd256 (pcdouble)
30831 v8sf __builtin_ia32_loadups256 (pcfloat)
30832 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
30833 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
30834 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
30835 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
30836 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
30837 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
30838 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
30839 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
30840 v4df __builtin_ia32_maxpd256 (v4df,v4df)
30841 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
30842 v4df __builtin_ia32_minpd256 (v4df,v4df)
30843 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
30844 v4df __builtin_ia32_movddup256 (v4df)
30845 int __builtin_ia32_movmskpd256 (v4df)
30846 int __builtin_ia32_movmskps256 (v8sf)
30847 v8sf __builtin_ia32_movshdup256 (v8sf)
30848 v8sf __builtin_ia32_movsldup256 (v8sf)
30849 v4df __builtin_ia32_mulpd256 (v4df,v4df)
30850 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
30851 v4df __builtin_ia32_orpd256 (v4df,v4df)
30852 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
30853 v2df __builtin_ia32_pd_pd256 (v4df)
30854 v4df __builtin_ia32_pd256_pd (v2df)
30855 v4sf __builtin_ia32_ps_ps256 (v8sf)
30856 v8sf __builtin_ia32_ps256_ps (v4sf)
30857 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
30858 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
30859 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
30860 v8sf __builtin_ia32_rcpps256 (v8sf)
30861 v4df __builtin_ia32_roundpd256 (v4df,int)
30862 v8sf __builtin_ia32_roundps256 (v8sf,int)
30863 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
30864 v8sf __builtin_ia32_rsqrtps256 (v8sf)
30865 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
30866 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
30867 v4si __builtin_ia32_si_si256 (v8si)
30868 v8si __builtin_ia32_si256_si (v4si)
30869 v4df __builtin_ia32_sqrtpd256 (v4df)
30870 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
30871 v8sf __builtin_ia32_sqrtps256 (v8sf)
30872 void __builtin_ia32_storedqu256 (pchar,v32qi)
30873 void __builtin_ia32_storeupd256 (pdouble,v4df)
30874 void __builtin_ia32_storeups256 (pfloat,v8sf)
30875 v4df __builtin_ia32_subpd256 (v4df,v4df)
30876 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
30877 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
30878 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
30879 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
30880 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
30881 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
30882 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
30883 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
30884 v4sf __builtin_ia32_vbroadcastss (pcfloat)
30885 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
30886 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
30887 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
30888 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
30889 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
30890 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
30891 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
30892 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
30893 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
30894 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
30895 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
30896 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
30897 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
30898 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
30899 v2df __builtin_ia32_vpermilpd (v2df,int)
30900 v4df __builtin_ia32_vpermilpd256 (v4df,int)
30901 v4sf __builtin_ia32_vpermilps (v4sf,int)
30902 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
30903 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
30904 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
30905 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
30906 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
30907 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
30908 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
30909 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
30910 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
30911 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
30912 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
30913 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
30914 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
30915 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
30916 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
30917 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
30918 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
30919 void __builtin_ia32_vzeroall (void)
30920 void __builtin_ia32_vzeroupper (void)
30921 v4df __builtin_ia32_xorpd256 (v4df,v4df)
30922 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
30924 The following built-in functions are available when `-maes' is used.
30925 All of them generate the machine instruction that is part of the name.
30927 v2di __builtin_ia32_aesenc128 (v2di, v2di)
30928 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
30929 v2di __builtin_ia32_aesdec128 (v2di, v2di)
30930 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
30931 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
30932 v2di __builtin_ia32_aesimc128 (v2di)
30934 The following built-in function is available when `-mpclmul' is used.
30936 `v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)'
30937 Generates the `pclmulqdq' machine instruction.
30939 The following built-in functions are available when `-msse4a' is used.
30940 All of them generate the machine instruction that is part of the name.
30942 void __builtin_ia32_movntsd (double *, v2df)
30943 void __builtin_ia32_movntss (float *, v4sf)
30944 v2di __builtin_ia32_extrq (v2di, v16qi)
30945 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
30946 v2di __builtin_ia32_insertq (v2di, v2di)
30947 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
30949 The following built-in functions are available when `-msse5' is used.
30950 All of them generate the machine instruction that is part of the name
30951 with MMX registers.
30953 v2df __builtin_ia32_comeqpd (v2df, v2df)
30954 v2df __builtin_ia32_comeqps (v2df, v2df)
30955 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
30956 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
30957 v2df __builtin_ia32_comfalsepd (v2df, v2df)
30958 v2df __builtin_ia32_comfalseps (v2df, v2df)
30959 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
30960 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
30961 v2df __builtin_ia32_comgepd (v2df, v2df)
30962 v2df __builtin_ia32_comgeps (v2df, v2df)
30963 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
30964 v4sf __builtin_ia32_comgess (v4sf, v4sf)
30965 v2df __builtin_ia32_comgtpd (v2df, v2df)
30966 v2df __builtin_ia32_comgtps (v2df, v2df)
30967 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
30968 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
30969 v2df __builtin_ia32_comlepd (v2df, v2df)
30970 v2df __builtin_ia32_comleps (v2df, v2df)
30971 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
30972 v4sf __builtin_ia32_comless (v4sf, v4sf)
30973 v2df __builtin_ia32_comltpd (v2df, v2df)
30974 v2df __builtin_ia32_comltps (v2df, v2df)
30975 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
30976 v4sf __builtin_ia32_comltss (v4sf, v4sf)
30977 v2df __builtin_ia32_comnepd (v2df, v2df)
30978 v2df __builtin_ia32_comneps (v2df, v2df)
30979 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
30980 v4sf __builtin_ia32_comness (v4sf, v4sf)
30981 v2df __builtin_ia32_comordpd (v2df, v2df)
30982 v2df __builtin_ia32_comordps (v2df, v2df)
30983 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
30984 v4sf __builtin_ia32_comordss (v4sf, v4sf)
30985 v2df __builtin_ia32_comtruepd (v2df, v2df)
30986 v2df __builtin_ia32_comtrueps (v2df, v2df)
30987 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
30988 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
30989 v2df __builtin_ia32_comueqpd (v2df, v2df)
30990 v2df __builtin_ia32_comueqps (v2df, v2df)
30991 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
30992 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
30993 v2df __builtin_ia32_comugepd (v2df, v2df)
30994 v2df __builtin_ia32_comugeps (v2df, v2df)
30995 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
30996 v4sf __builtin_ia32_comugess (v4sf, v4sf)
30997 v2df __builtin_ia32_comugtpd (v2df, v2df)
30998 v2df __builtin_ia32_comugtps (v2df, v2df)
30999 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
31000 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
31001 v2df __builtin_ia32_comulepd (v2df, v2df)
31002 v2df __builtin_ia32_comuleps (v2df, v2df)
31003 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
31004 v4sf __builtin_ia32_comuless (v4sf, v4sf)
31005 v2df __builtin_ia32_comultpd (v2df, v2df)
31006 v2df __builtin_ia32_comultps (v2df, v2df)
31007 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
31008 v4sf __builtin_ia32_comultss (v4sf, v4sf)
31009 v2df __builtin_ia32_comunepd (v2df, v2df)
31010 v2df __builtin_ia32_comuneps (v2df, v2df)
31011 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
31012 v4sf __builtin_ia32_comuness (v4sf, v4sf)
31013 v2df __builtin_ia32_comunordpd (v2df, v2df)
31014 v2df __builtin_ia32_comunordps (v2df, v2df)
31015 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
31016 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
31017 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
31018 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
31019 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
31020 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
31021 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
31022 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
31023 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
31024 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
31025 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
31026 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
31027 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
31028 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
31029 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
31030 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
31031 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
31032 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
31033 v2df __builtin_ia32_frczpd (v2df)
31034 v4sf __builtin_ia32_frczps (v4sf)
31035 v2df __builtin_ia32_frczsd (v2df, v2df)
31036 v4sf __builtin_ia32_frczss (v4sf, v4sf)
31037 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
31038 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
31039 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
31040 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
31041 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
31042 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
31043 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
31044 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
31045 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
31046 v4si __builtin_ia32_pcomeqd (v4si, v4si)
31047 v2di __builtin_ia32_pcomeqq (v2di, v2di)
31048 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
31049 v4si __builtin_ia32_pcomequd (v4si, v4si)
31050 v2di __builtin_ia32_pcomequq (v2di, v2di)
31051 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
31052 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
31053 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
31054 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
31055 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
31056 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
31057 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
31058 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
31059 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
31060 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
31061 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
31062 v4si __builtin_ia32_pcomged (v4si, v4si)
31063 v2di __builtin_ia32_pcomgeq (v2di, v2di)
31064 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
31065 v4si __builtin_ia32_pcomgeud (v4si, v4si)
31066 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
31067 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
31068 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
31069 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
31070 v4si __builtin_ia32_pcomgtd (v4si, v4si)
31071 v2di __builtin_ia32_pcomgtq (v2di, v2di)
31072 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
31073 v4si __builtin_ia32_pcomgtud (v4si, v4si)
31074 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
31075 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
31076 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
31077 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
31078 v4si __builtin_ia32_pcomled (v4si, v4si)
31079 v2di __builtin_ia32_pcomleq (v2di, v2di)
31080 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
31081 v4si __builtin_ia32_pcomleud (v4si, v4si)
31082 v2di __builtin_ia32_pcomleuq (v2di, v2di)
31083 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
31084 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
31085 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
31086 v4si __builtin_ia32_pcomltd (v4si, v4si)
31087 v2di __builtin_ia32_pcomltq (v2di, v2di)
31088 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
31089 v4si __builtin_ia32_pcomltud (v4si, v4si)
31090 v2di __builtin_ia32_pcomltuq (v2di, v2di)
31091 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
31092 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
31093 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
31094 v4si __builtin_ia32_pcomned (v4si, v4si)
31095 v2di __builtin_ia32_pcomneq (v2di, v2di)
31096 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
31097 v4si __builtin_ia32_pcomneud (v4si, v4si)
31098 v2di __builtin_ia32_pcomneuq (v2di, v2di)
31099 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
31100 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
31101 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
31102 v4si __builtin_ia32_pcomtrued (v4si, v4si)
31103 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
31104 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
31105 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
31106 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
31107 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
31108 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
31109 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
31110 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
31111 v4si __builtin_ia32_phaddbd (v16qi)
31112 v2di __builtin_ia32_phaddbq (v16qi)
31113 v8hi __builtin_ia32_phaddbw (v16qi)
31114 v2di __builtin_ia32_phadddq (v4si)
31115 v4si __builtin_ia32_phaddubd (v16qi)
31116 v2di __builtin_ia32_phaddubq (v16qi)
31117 v8hi __builtin_ia32_phaddubw (v16qi)
31118 v2di __builtin_ia32_phaddudq (v4si)
31119 v4si __builtin_ia32_phadduwd (v8hi)
31120 v2di __builtin_ia32_phadduwq (v8hi)
31121 v4si __builtin_ia32_phaddwd (v8hi)
31122 v2di __builtin_ia32_phaddwq (v8hi)
31123 v8hi __builtin_ia32_phsubbw (v16qi)
31124 v2di __builtin_ia32_phsubdq (v4si)
31125 v4si __builtin_ia32_phsubwd (v8hi)
31126 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
31127 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
31128 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
31129 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
31130 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
31131 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
31132 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
31133 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
31134 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
31135 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
31136 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
31137 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
31138 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
31139 v16qi __builtin_ia32_protb (v16qi, v16qi)
31140 v4si __builtin_ia32_protd (v4si, v4si)
31141 v2di __builtin_ia32_protq (v2di, v2di)
31142 v8hi __builtin_ia32_protw (v8hi, v8hi)
31143 v16qi __builtin_ia32_pshab (v16qi, v16qi)
31144 v4si __builtin_ia32_pshad (v4si, v4si)
31145 v2di __builtin_ia32_pshaq (v2di, v2di)
31146 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
31147 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
31148 v4si __builtin_ia32_pshld (v4si, v4si)
31149 v2di __builtin_ia32_pshlq (v2di, v2di)
31150 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
31152 The following builtin-in functions are available when `-msse5' is
31153 used. The second argument must be an integer constant and generate the
31154 machine instruction that is part of the name with the `_imm' suffix
31157 v16qi __builtin_ia32_protb_imm (v16qi, int)
31158 v4si __builtin_ia32_protd_imm (v4si, int)
31159 v2di __builtin_ia32_protq_imm (v2di, int)
31160 v8hi __builtin_ia32_protw_imm (v8hi, int)
31162 The following built-in functions are available when `-m3dnow' is used.
31163 All of them generate the machine instruction that is part of the name.
31165 void __builtin_ia32_femms (void)
31166 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
31167 v2si __builtin_ia32_pf2id (v2sf)
31168 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
31169 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
31170 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
31171 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
31172 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
31173 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
31174 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
31175 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
31176 v2sf __builtin_ia32_pfrcp (v2sf)
31177 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
31178 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
31179 v2sf __builtin_ia32_pfrsqrt (v2sf)
31180 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
31181 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
31182 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
31183 v2sf __builtin_ia32_pi2fd (v2si)
31184 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
31186 The following built-in functions are available when both `-m3dnow' and
31187 `-march=athlon' are used. All of them generate the machine instruction
31188 that is part of the name.
31190 v2si __builtin_ia32_pf2iw (v2sf)
31191 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
31192 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
31193 v2sf __builtin_ia32_pi2fw (v2si)
31194 v2sf __builtin_ia32_pswapdsf (v2sf)
31195 v2si __builtin_ia32_pswapdsi (v2si)
31198 File: gcc.info, Node: MIPS DSP Built-in Functions, Next: MIPS Paired-Single Support, Prev: X86 Built-in Functions, Up: Target Builtins
31200 5.50.7 MIPS DSP Built-in Functions
31201 ----------------------------------
31203 The MIPS DSP Application-Specific Extension (ASE) includes new
31204 instructions that are designed to improve the performance of DSP and
31205 media applications. It provides instructions that operate on packed
31206 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
31208 GCC supports MIPS DSP operations using both the generic vector
31209 extensions (*note Vector Extensions::) and a collection of
31210 MIPS-specific built-in functions. Both kinds of support are enabled by
31211 the `-mdsp' command-line option.
31213 Revision 2 of the ASE was introduced in the second half of 2006. This
31214 revision adds extra instructions to the original ASE, but is otherwise
31215 backwards-compatible with it. You can select revision 2 using the
31216 command-line option `-mdspr2'; this option implies `-mdsp'.
31218 The SCOUNT and POS bits of the DSP control register are global. The
31219 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
31220 POS bits. During optimization, the compiler will not delete these
31221 instructions and it will not delete calls to functions containing these
31224 At present, GCC only provides support for operations on 32-bit
31225 vectors. The vector type associated with 8-bit integer data is usually
31226 called `v4i8', the vector type associated with Q7 is usually called
31227 `v4q7', the vector type associated with 16-bit integer data is usually
31228 called `v2i16', and the vector type associated with Q15 is usually
31229 called `v2q15'. They can be defined in C as follows:
31231 typedef signed char v4i8 __attribute__ ((vector_size(4)));
31232 typedef signed char v4q7 __attribute__ ((vector_size(4)));
31233 typedef short v2i16 __attribute__ ((vector_size(4)));
31234 typedef short v2q15 __attribute__ ((vector_size(4)));
31236 `v4i8', `v4q7', `v2i16' and `v2q15' values are initialized in the same
31237 way as aggregates. For example:
31239 v4i8 a = {1, 2, 3, 4};
31241 b = (v4i8) {5, 6, 7, 8};
31243 v2q15 c = {0x0fcb, 0x3a75};
31245 d = (v2q15) {0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15};
31247 _Note:_ The CPU's endianness determines the order in which values are
31248 packed. On little-endian targets, the first value is the least
31249 significant and the last value is the most significant. The opposite
31250 order applies to big-endian targets. For example, the code above will
31251 set the lowest byte of `a' to `1' on little-endian targets and `4' on
31252 big-endian targets.
31254 _Note:_ Q7, Q15 and Q31 values must be initialized with their integer
31255 representation. As shown in this example, the integer representation
31256 of a Q7 value can be obtained by multiplying the fractional value by
31257 `0x1.0p7'. The equivalent for Q15 values is to multiply by `0x1.0p15'.
31258 The equivalent for Q31 values is to multiply by `0x1.0p31'.
31260 The table below lists the `v4i8' and `v2q15' operations for which
31261 hardware support exists. `a' and `b' are `v4i8' values, and `c' and
31262 `d' are `v2q15' values.
31264 C code MIPS instruction
31270 The table below lists the `v2i16' operation for which hardware support
31271 exists for the DSP ASE REV 2. `e' and `f' are `v2i16' values.
31273 C code MIPS instruction
31276 It is easier to describe the DSP built-in functions if we first define
31277 the following types:
31281 typedef unsigned int ui32;
31282 typedef long long a64;
31284 `q31' and `i32' are actually the same as `int', but we use `q31' to
31285 indicate a Q31 fractional value and `i32' to indicate a 32-bit integer
31286 value. Similarly, `a64' is the same as `long long', but we use `a64'
31287 to indicate values that will be placed in one of the four DSP
31288 accumulators (`$ac0', `$ac1', `$ac2' or `$ac3').
31290 Also, some built-in functions prefer or require immediate numbers as
31291 parameters, because the corresponding DSP instructions accept both
31292 immediate numbers and register operands, or accept immediate numbers
31293 only. The immediate parameters are listed as follows.
31300 imm0_255: 0 to 255.
31301 imm_n32_31: -32 to 31.
31302 imm_n512_511: -512 to 511.
31304 The following built-in functions map directly to a particular MIPS DSP
31305 instruction. Please refer to the architecture specification for
31306 details on what each instruction does.
31308 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
31309 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
31310 q31 __builtin_mips_addq_s_w (q31, q31)
31311 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
31312 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
31313 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
31314 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
31315 q31 __builtin_mips_subq_s_w (q31, q31)
31316 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
31317 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
31318 i32 __builtin_mips_addsc (i32, i32)
31319 i32 __builtin_mips_addwc (i32, i32)
31320 i32 __builtin_mips_modsub (i32, i32)
31321 i32 __builtin_mips_raddu_w_qb (v4i8)
31322 v2q15 __builtin_mips_absq_s_ph (v2q15)
31323 q31 __builtin_mips_absq_s_w (q31)
31324 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
31325 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
31326 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
31327 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
31328 q31 __builtin_mips_preceq_w_phl (v2q15)
31329 q31 __builtin_mips_preceq_w_phr (v2q15)
31330 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
31331 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
31332 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
31333 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
31334 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
31335 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
31336 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
31337 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
31338 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
31339 v4i8 __builtin_mips_shll_qb (v4i8, i32)
31340 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
31341 v2q15 __builtin_mips_shll_ph (v2q15, i32)
31342 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
31343 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
31344 q31 __builtin_mips_shll_s_w (q31, imm0_31)
31345 q31 __builtin_mips_shll_s_w (q31, i32)
31346 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
31347 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
31348 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
31349 v2q15 __builtin_mips_shra_ph (v2q15, i32)
31350 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
31351 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
31352 q31 __builtin_mips_shra_r_w (q31, imm0_31)
31353 q31 __builtin_mips_shra_r_w (q31, i32)
31354 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
31355 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
31356 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
31357 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
31358 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
31359 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
31360 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
31361 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
31362 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
31363 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
31364 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
31365 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
31366 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
31367 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
31368 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
31369 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
31370 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
31371 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
31372 i32 __builtin_mips_bitrev (i32)
31373 i32 __builtin_mips_insv (i32, i32)
31374 v4i8 __builtin_mips_repl_qb (imm0_255)
31375 v4i8 __builtin_mips_repl_qb (i32)
31376 v2q15 __builtin_mips_repl_ph (imm_n512_511)
31377 v2q15 __builtin_mips_repl_ph (i32)
31378 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
31379 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
31380 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
31381 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
31382 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
31383 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
31384 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
31385 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
31386 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
31387 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
31388 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
31389 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
31390 i32 __builtin_mips_extr_w (a64, imm0_31)
31391 i32 __builtin_mips_extr_w (a64, i32)
31392 i32 __builtin_mips_extr_r_w (a64, imm0_31)
31393 i32 __builtin_mips_extr_s_h (a64, i32)
31394 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
31395 i32 __builtin_mips_extr_rs_w (a64, i32)
31396 i32 __builtin_mips_extr_s_h (a64, imm0_31)
31397 i32 __builtin_mips_extr_r_w (a64, i32)
31398 i32 __builtin_mips_extp (a64, imm0_31)
31399 i32 __builtin_mips_extp (a64, i32)
31400 i32 __builtin_mips_extpdp (a64, imm0_31)
31401 i32 __builtin_mips_extpdp (a64, i32)
31402 a64 __builtin_mips_shilo (a64, imm_n32_31)
31403 a64 __builtin_mips_shilo (a64, i32)
31404 a64 __builtin_mips_mthlip (a64, i32)
31405 void __builtin_mips_wrdsp (i32, imm0_63)
31406 i32 __builtin_mips_rddsp (imm0_63)
31407 i32 __builtin_mips_lbux (void *, i32)
31408 i32 __builtin_mips_lhx (void *, i32)
31409 i32 __builtin_mips_lwx (void *, i32)
31410 i32 __builtin_mips_bposge32 (void)
31412 The following built-in functions map directly to a particular MIPS DSP
31413 REV 2 instruction. Please refer to the architecture specification for
31414 details on what each instruction does.
31416 v4q7 __builtin_mips_absq_s_qb (v4q7);
31417 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
31418 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
31419 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
31420 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
31421 i32 __builtin_mips_append (i32, i32, imm0_31);
31422 i32 __builtin_mips_balign (i32, i32, imm0_3);
31423 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
31424 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
31425 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
31426 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
31427 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
31428 a64 __builtin_mips_madd (a64, i32, i32);
31429 a64 __builtin_mips_maddu (a64, ui32, ui32);
31430 a64 __builtin_mips_msub (a64, i32, i32);
31431 a64 __builtin_mips_msubu (a64, ui32, ui32);
31432 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
31433 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
31434 q31 __builtin_mips_mulq_rs_w (q31, q31);
31435 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
31436 q31 __builtin_mips_mulq_s_w (q31, q31);
31437 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
31438 a64 __builtin_mips_mult (i32, i32);
31439 a64 __builtin_mips_multu (ui32, ui32);
31440 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
31441 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
31442 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
31443 i32 __builtin_mips_prepend (i32, i32, imm0_31);
31444 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
31445 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
31446 v4i8 __builtin_mips_shra_qb (v4i8, i32);
31447 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
31448 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
31449 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
31450 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
31451 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
31452 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
31453 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
31454 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
31455 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
31456 q31 __builtin_mips_addqh_w (q31, q31);
31457 q31 __builtin_mips_addqh_r_w (q31, q31);
31458 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
31459 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
31460 q31 __builtin_mips_subqh_w (q31, q31);
31461 q31 __builtin_mips_subqh_r_w (q31, q31);
31462 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
31463 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
31464 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
31465 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
31466 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
31467 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
31470 File: gcc.info, Node: MIPS Paired-Single Support, Next: MIPS Loongson Built-in Functions, Prev: MIPS DSP Built-in Functions, Up: Target Builtins
31472 5.50.8 MIPS Paired-Single Support
31473 ---------------------------------
31475 The MIPS64 architecture includes a number of instructions that operate
31476 on pairs of single-precision floating-point values. Each pair is
31477 packed into a 64-bit floating-point register, with one element being
31478 designated the "upper half" and the other being designated the "lower
31481 GCC supports paired-single operations using both the generic vector
31482 extensions (*note Vector Extensions::) and a collection of
31483 MIPS-specific built-in functions. Both kinds of support are enabled by
31484 the `-mpaired-single' command-line option.
31486 The vector type associated with paired-single values is usually called
31487 `v2sf'. It can be defined in C as follows:
31489 typedef float v2sf __attribute__ ((vector_size (8)));
31491 `v2sf' values are initialized in the same way as aggregates. For
31494 v2sf a = {1.5, 9.1};
31499 _Note:_ The CPU's endianness determines which value is stored in the
31500 upper half of a register and which value is stored in the lower half.
31501 On little-endian targets, the first value is the lower one and the
31502 second value is the upper one. The opposite order applies to
31503 big-endian targets. For example, the code above will set the lower
31504 half of `a' to `1.5' on little-endian targets and `9.1' on big-endian
31508 File: gcc.info, Node: MIPS Loongson Built-in Functions, Next: Other MIPS Built-in Functions, Prev: MIPS Paired-Single Support, Up: Target Builtins
31510 5.50.9 MIPS Loongson Built-in Functions
31511 ---------------------------------------
31513 GCC provides intrinsics to access the SIMD instructions provided by the
31514 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
31515 available after inclusion of the `loongson.h' header file, operate on
31516 the following 64-bit vector types:
31518 * `uint8x8_t', a vector of eight unsigned 8-bit integers;
31520 * `uint16x4_t', a vector of four unsigned 16-bit integers;
31522 * `uint32x2_t', a vector of two unsigned 32-bit integers;
31524 * `int8x8_t', a vector of eight signed 8-bit integers;
31526 * `int16x4_t', a vector of four signed 16-bit integers;
31528 * `int32x2_t', a vector of two signed 32-bit integers.
31530 The intrinsics provided are listed below; each is named after the
31531 machine instruction to which it corresponds, with suffixes added as
31532 appropriate to distinguish intrinsics that expand to the same machine
31533 instruction yet have different argument types. Refer to the
31534 architecture documentation for a description of the functionality of
31537 int16x4_t packsswh (int32x2_t s, int32x2_t t);
31538 int8x8_t packsshb (int16x4_t s, int16x4_t t);
31539 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
31540 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
31541 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
31542 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
31543 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
31544 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
31545 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
31546 uint64_t paddd_u (uint64_t s, uint64_t t);
31547 int64_t paddd_s (int64_t s, int64_t t);
31548 int16x4_t paddsh (int16x4_t s, int16x4_t t);
31549 int8x8_t paddsb (int8x8_t s, int8x8_t t);
31550 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
31551 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
31552 uint64_t pandn_ud (uint64_t s, uint64_t t);
31553 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
31554 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
31555 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
31556 int64_t pandn_sd (int64_t s, int64_t t);
31557 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
31558 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
31559 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
31560 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
31561 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
31562 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
31563 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
31564 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
31565 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
31566 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
31567 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
31568 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
31569 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
31570 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
31571 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
31572 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
31573 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
31574 uint16x4_t pextrh_u (uint16x4_t s, int field);
31575 int16x4_t pextrh_s (int16x4_t s, int field);
31576 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
31577 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
31578 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
31579 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
31580 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
31581 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
31582 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
31583 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
31584 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
31585 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
31586 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
31587 int16x4_t pminsh (int16x4_t s, int16x4_t t);
31588 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
31589 uint8x8_t pmovmskb_u (uint8x8_t s);
31590 int8x8_t pmovmskb_s (int8x8_t s);
31591 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
31592 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
31593 int16x4_t pmullh (int16x4_t s, int16x4_t t);
31594 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
31595 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
31596 uint16x4_t biadd (uint8x8_t s);
31597 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
31598 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
31599 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
31600 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
31601 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
31602 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
31603 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
31604 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
31605 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
31606 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
31607 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
31608 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
31609 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
31610 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
31611 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
31612 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
31613 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
31614 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
31615 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
31616 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
31617 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
31618 uint64_t psubd_u (uint64_t s, uint64_t t);
31619 int64_t psubd_s (int64_t s, int64_t t);
31620 int16x4_t psubsh (int16x4_t s, int16x4_t t);
31621 int8x8_t psubsb (int8x8_t s, int8x8_t t);
31622 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
31623 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
31624 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
31625 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
31626 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
31627 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
31628 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
31629 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
31630 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
31631 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
31632 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
31633 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
31634 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
31635 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
31639 * Paired-Single Arithmetic::
31640 * Paired-Single Built-in Functions::
31641 * MIPS-3D Built-in Functions::
31644 File: gcc.info, Node: Paired-Single Arithmetic, Next: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions
31646 5.50.9.1 Paired-Single Arithmetic
31647 .................................
31649 The table below lists the `v2sf' operations for which hardware support
31650 exists. `a', `b' and `c' are `v2sf' values and `x' is an integral
31653 C code MIPS instruction
31658 `a * b + c' `madd.ps'
31659 `a * b - c' `msub.ps'
31660 `-(a * b + c)' `nmadd.ps'
31661 `-(a * b - c)' `nmsub.ps'
31662 `x ? a : b' `movn.ps'/`movz.ps'
31664 Note that the multiply-accumulate instructions can be disabled using
31665 the command-line option `-mno-fused-madd'.
31668 File: gcc.info, Node: Paired-Single Built-in Functions, Next: MIPS-3D Built-in Functions, Prev: Paired-Single Arithmetic, Up: MIPS Loongson Built-in Functions
31670 5.50.9.2 Paired-Single Built-in Functions
31671 .........................................
31673 The following paired-single functions map directly to a particular MIPS
31674 instruction. Please refer to the architecture specification for
31675 details on what each instruction does.
31677 `v2sf __builtin_mips_pll_ps (v2sf, v2sf)'
31678 Pair lower lower (`pll.ps').
31680 `v2sf __builtin_mips_pul_ps (v2sf, v2sf)'
31681 Pair upper lower (`pul.ps').
31683 `v2sf __builtin_mips_plu_ps (v2sf, v2sf)'
31684 Pair lower upper (`plu.ps').
31686 `v2sf __builtin_mips_puu_ps (v2sf, v2sf)'
31687 Pair upper upper (`puu.ps').
31689 `v2sf __builtin_mips_cvt_ps_s (float, float)'
31690 Convert pair to paired single (`cvt.ps.s').
31692 `float __builtin_mips_cvt_s_pl (v2sf)'
31693 Convert pair lower to single (`cvt.s.pl').
31695 `float __builtin_mips_cvt_s_pu (v2sf)'
31696 Convert pair upper to single (`cvt.s.pu').
31698 `v2sf __builtin_mips_abs_ps (v2sf)'
31699 Absolute value (`abs.ps').
31701 `v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)'
31702 Align variable (`alnv.ps').
31704 _Note:_ The value of the third parameter must be 0 or 4 modulo 8,
31705 otherwise the result will be unpredictable. Please read the
31706 instruction description for details.
31708 The following multi-instruction functions are also available. In each
31709 case, COND can be any of the 16 floating-point conditions: `f', `un',
31710 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
31711 `lt', `nge', `le' or `ngt'.
31713 `v2sf __builtin_mips_movt_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31714 `v2sf __builtin_mips_movf_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31715 Conditional move based on floating point comparison (`c.COND.ps',
31716 `movt.ps'/`movf.ps').
31718 The `movt' functions return the value X computed by:
31724 The `movf' functions are similar but use `movf.ps' instead of
31727 `int __builtin_mips_upper_c_COND_ps (v2sf A, v2sf B)'
31728 `int __builtin_mips_lower_c_COND_ps (v2sf A, v2sf B)'
31729 Comparison of two paired-single values (`c.COND.ps',
31732 These functions compare A and B using `c.COND.ps' and return
31733 either the upper or lower half of the result. For example:
31736 if (__builtin_mips_upper_c_eq_ps (a, b))
31737 upper_halves_are_equal ();
31739 upper_halves_are_unequal ();
31741 if (__builtin_mips_lower_c_eq_ps (a, b))
31742 lower_halves_are_equal ();
31744 lower_halves_are_unequal ();
31747 File: gcc.info, Node: MIPS-3D Built-in Functions, Prev: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions
31749 5.50.9.3 MIPS-3D Built-in Functions
31750 ...................................
31752 The MIPS-3D Application-Specific Extension (ASE) includes additional
31753 paired-single instructions that are designed to improve the performance
31754 of 3D graphics operations. Support for these instructions is controlled
31755 by the `-mips3d' command-line option.
31757 The functions listed below map directly to a particular MIPS-3D
31758 instruction. Please refer to the architecture specification for more
31759 details on what each instruction does.
31761 `v2sf __builtin_mips_addr_ps (v2sf, v2sf)'
31762 Reduction add (`addr.ps').
31764 `v2sf __builtin_mips_mulr_ps (v2sf, v2sf)'
31765 Reduction multiply (`mulr.ps').
31767 `v2sf __builtin_mips_cvt_pw_ps (v2sf)'
31768 Convert paired single to paired word (`cvt.pw.ps').
31770 `v2sf __builtin_mips_cvt_ps_pw (v2sf)'
31771 Convert paired word to paired single (`cvt.ps.pw').
31773 `float __builtin_mips_recip1_s (float)'
31774 `double __builtin_mips_recip1_d (double)'
31775 `v2sf __builtin_mips_recip1_ps (v2sf)'
31776 Reduced precision reciprocal (sequence step 1) (`recip1.FMT').
31778 `float __builtin_mips_recip2_s (float, float)'
31779 `double __builtin_mips_recip2_d (double, double)'
31780 `v2sf __builtin_mips_recip2_ps (v2sf, v2sf)'
31781 Reduced precision reciprocal (sequence step 2) (`recip2.FMT').
31783 `float __builtin_mips_rsqrt1_s (float)'
31784 `double __builtin_mips_rsqrt1_d (double)'
31785 `v2sf __builtin_mips_rsqrt1_ps (v2sf)'
31786 Reduced precision reciprocal square root (sequence step 1)
31789 `float __builtin_mips_rsqrt2_s (float, float)'
31790 `double __builtin_mips_rsqrt2_d (double, double)'
31791 `v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)'
31792 Reduced precision reciprocal square root (sequence step 2)
31795 The following multi-instruction functions are also available. In each
31796 case, COND can be any of the 16 floating-point conditions: `f', `un',
31797 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
31798 `lt', `nge', `le' or `ngt'.
31800 `int __builtin_mips_cabs_COND_s (float A, float B)'
31801 `int __builtin_mips_cabs_COND_d (double A, double B)'
31802 Absolute comparison of two scalar values (`cabs.COND.FMT',
31805 These functions compare A and B using `cabs.COND.s' or
31806 `cabs.COND.d' and return the result as a boolean value. For
31810 if (__builtin_mips_cabs_eq_s (a, b))
31815 `int __builtin_mips_upper_cabs_COND_ps (v2sf A, v2sf B)'
31816 `int __builtin_mips_lower_cabs_COND_ps (v2sf A, v2sf B)'
31817 Absolute comparison of two paired-single values (`cabs.COND.ps',
31820 These functions compare A and B using `cabs.COND.ps' and return
31821 either the upper or lower half of the result. For example:
31824 if (__builtin_mips_upper_cabs_eq_ps (a, b))
31825 upper_halves_are_equal ();
31827 upper_halves_are_unequal ();
31829 if (__builtin_mips_lower_cabs_eq_ps (a, b))
31830 lower_halves_are_equal ();
31832 lower_halves_are_unequal ();
31834 `v2sf __builtin_mips_movt_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31835 `v2sf __builtin_mips_movf_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31836 Conditional move based on absolute comparison (`cabs.COND.ps',
31837 `movt.ps'/`movf.ps').
31839 The `movt' functions return the value X computed by:
31841 cabs.COND.ps CC,A,B
31845 The `movf' functions are similar but use `movf.ps' instead of
31848 `int __builtin_mips_any_c_COND_ps (v2sf A, v2sf B)'
31849 `int __builtin_mips_all_c_COND_ps (v2sf A, v2sf B)'
31850 `int __builtin_mips_any_cabs_COND_ps (v2sf A, v2sf B)'
31851 `int __builtin_mips_all_cabs_COND_ps (v2sf A, v2sf B)'
31852 Comparison of two paired-single values (`c.COND.ps'/`cabs.COND.ps',
31853 `bc1any2t'/`bc1any2f').
31855 These functions compare A and B using `c.COND.ps' or
31856 `cabs.COND.ps'. The `any' forms return true if either result is
31857 true and the `all' forms return true if both results are true.
31861 if (__builtin_mips_any_c_eq_ps (a, b))
31866 if (__builtin_mips_all_c_eq_ps (a, b))
31871 `int __builtin_mips_any_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31872 `int __builtin_mips_all_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31873 `int __builtin_mips_any_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31874 `int __builtin_mips_all_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31875 Comparison of four paired-single values
31876 (`c.COND.ps'/`cabs.COND.ps', `bc1any4t'/`bc1any4f').
31878 These functions use `c.COND.ps' or `cabs.COND.ps' to compare A
31879 with B and to compare C with D. The `any' forms return true if
31880 any of the four results are true and the `all' forms return true
31881 if all four results are true. For example:
31884 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
31889 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
31895 File: gcc.info, Node: picoChip Built-in Functions, Next: PowerPC AltiVec Built-in Functions, Prev: Other MIPS Built-in Functions, Up: Target Builtins
31897 5.50.10 picoChip Built-in Functions
31898 -----------------------------------
31900 GCC provides an interface to selected machine instructions from the
31901 picoChip instruction set.
31903 `int __builtin_sbc (int VALUE)'
31904 Sign bit count. Return the number of consecutive bits in VALUE
31905 which have the same value as the sign-bit. The result is the
31906 number of leading sign bits minus one, giving the number of
31907 redundant sign bits in VALUE.
31909 `int __builtin_byteswap (int VALUE)'
31910 Byte swap. Return the result of swapping the upper and lower
31913 `int __builtin_brev (int VALUE)'
31914 Bit reversal. Return the result of reversing the bits in VALUE.
31915 Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1, and so
31918 `int __builtin_adds (int X, int Y)'
31919 Saturating addition. Return the result of adding X and Y, storing
31920 the value 32767 if the result overflows.
31922 `int __builtin_subs (int X, int Y)'
31923 Saturating subtraction. Return the result of subtracting Y from
31924 X, storing the value -32768 if the result overflows.
31926 `void __builtin_halt (void)'
31927 Halt. The processor will stop execution. This built-in is useful
31928 for implementing assertions.
31932 File: gcc.info, Node: Other MIPS Built-in Functions, Next: picoChip Built-in Functions, Prev: MIPS Loongson Built-in Functions, Up: Target Builtins
31934 5.50.11 Other MIPS Built-in Functions
31935 -------------------------------------
31937 GCC provides other MIPS-specific built-in functions:
31939 `void __builtin_mips_cache (int OP, const volatile void *ADDR)'
31940 Insert a `cache' instruction with operands OP and ADDR. GCC
31941 defines the preprocessor macro `___GCC_HAVE_BUILTIN_MIPS_CACHE'
31942 when this function is available.
31945 File: gcc.info, Node: PowerPC AltiVec Built-in Functions, Next: SPARC VIS Built-in Functions, Prev: picoChip Built-in Functions, Up: Target Builtins
31947 5.50.12 PowerPC AltiVec Built-in Functions
31948 ------------------------------------------
31950 GCC provides an interface for the PowerPC family of processors to access
31951 the AltiVec operations described in Motorola's AltiVec Programming
31952 Interface Manual. The interface is made available by including
31953 `<altivec.h>' and using `-maltivec' and `-mabi=altivec'. The interface
31954 supports the following vector types.
31956 vector unsigned char
31960 vector unsigned short
31961 vector signed short
31965 vector unsigned int
31970 GCC's implementation of the high-level language interface available
31971 from C and C++ code differs from Motorola's documentation in several
31974 * A vector constant is a list of constant expressions within curly
31977 * A vector initializer requires no cast if the vector constant is of
31978 the same type as the variable it is initializing.
31980 * If `signed' or `unsigned' is omitted, the signedness of the vector
31981 type is the default signedness of the base type. The default
31982 varies depending on the operating system, so a portable program
31983 should always specify the signedness.
31985 * Compiling with `-maltivec' adds keywords `__vector', `vector',
31986 `__pixel', `pixel', `__bool' and `bool'. When compiling ISO C,
31987 the context-sensitive substitution of the keywords `vector',
31988 `pixel' and `bool' is disabled. To use them, you must include
31989 `<altivec.h>' instead.
31991 * GCC allows using a `typedef' name as the type specifier for a
31994 * For C, overloaded functions are implemented with macros so the
31995 following does not work:
31997 vec_add ((vector signed int){1, 2, 3, 4}, foo);
31999 Since `vec_add' is a macro, the vector constant in the example is
32000 treated as four separate arguments. Wrap the entire argument in
32001 parentheses for this to work.
32003 _Note:_ Only the `<altivec.h>' interface is supported. Internally,
32004 GCC uses built-in functions to achieve the functionality in the
32005 aforementioned header file, but they are not supported and are subject
32006 to change without notice.
32008 The following interfaces are supported for the generic and specific
32009 AltiVec operations and the AltiVec predicates. In cases where there is
32010 a direct mapping between generic and specific operations, only the
32011 generic names are shown here, although the specific operations can also
32014 Arguments that are documented as `const int' require literal integral
32015 values within the range required for that operation.
32017 vector signed char vec_abs (vector signed char);
32018 vector signed short vec_abs (vector signed short);
32019 vector signed int vec_abs (vector signed int);
32020 vector float vec_abs (vector float);
32022 vector signed char vec_abss (vector signed char);
32023 vector signed short vec_abss (vector signed short);
32024 vector signed int vec_abss (vector signed int);
32026 vector signed char vec_add (vector bool char, vector signed char);
32027 vector signed char vec_add (vector signed char, vector bool char);
32028 vector signed char vec_add (vector signed char, vector signed char);
32029 vector unsigned char vec_add (vector bool char, vector unsigned char);
32030 vector unsigned char vec_add (vector unsigned char, vector bool char);
32031 vector unsigned char vec_add (vector unsigned char,
32032 vector unsigned char);
32033 vector signed short vec_add (vector bool short, vector signed short);
32034 vector signed short vec_add (vector signed short, vector bool short);
32035 vector signed short vec_add (vector signed short, vector signed short);
32036 vector unsigned short vec_add (vector bool short,
32037 vector unsigned short);
32038 vector unsigned short vec_add (vector unsigned short,
32039 vector bool short);
32040 vector unsigned short vec_add (vector unsigned short,
32041 vector unsigned short);
32042 vector signed int vec_add (vector bool int, vector signed int);
32043 vector signed int vec_add (vector signed int, vector bool int);
32044 vector signed int vec_add (vector signed int, vector signed int);
32045 vector unsigned int vec_add (vector bool int, vector unsigned int);
32046 vector unsigned int vec_add (vector unsigned int, vector bool int);
32047 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
32048 vector float vec_add (vector float, vector float);
32050 vector float vec_vaddfp (vector float, vector float);
32052 vector signed int vec_vadduwm (vector bool int, vector signed int);
32053 vector signed int vec_vadduwm (vector signed int, vector bool int);
32054 vector signed int vec_vadduwm (vector signed int, vector signed int);
32055 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
32056 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
32057 vector unsigned int vec_vadduwm (vector unsigned int,
32058 vector unsigned int);
32060 vector signed short vec_vadduhm (vector bool short,
32061 vector signed short);
32062 vector signed short vec_vadduhm (vector signed short,
32063 vector bool short);
32064 vector signed short vec_vadduhm (vector signed short,
32065 vector signed short);
32066 vector unsigned short vec_vadduhm (vector bool short,
32067 vector unsigned short);
32068 vector unsigned short vec_vadduhm (vector unsigned short,
32069 vector bool short);
32070 vector unsigned short vec_vadduhm (vector unsigned short,
32071 vector unsigned short);
32073 vector signed char vec_vaddubm (vector bool char, vector signed char);
32074 vector signed char vec_vaddubm (vector signed char, vector bool char);
32075 vector signed char vec_vaddubm (vector signed char, vector signed char);
32076 vector unsigned char vec_vaddubm (vector bool char,
32077 vector unsigned char);
32078 vector unsigned char vec_vaddubm (vector unsigned char,
32080 vector unsigned char vec_vaddubm (vector unsigned char,
32081 vector unsigned char);
32083 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
32085 vector unsigned char vec_adds (vector bool char, vector unsigned char);
32086 vector unsigned char vec_adds (vector unsigned char, vector bool char);
32087 vector unsigned char vec_adds (vector unsigned char,
32088 vector unsigned char);
32089 vector signed char vec_adds (vector bool char, vector signed char);
32090 vector signed char vec_adds (vector signed char, vector bool char);
32091 vector signed char vec_adds (vector signed char, vector signed char);
32092 vector unsigned short vec_adds (vector bool short,
32093 vector unsigned short);
32094 vector unsigned short vec_adds (vector unsigned short,
32095 vector bool short);
32096 vector unsigned short vec_adds (vector unsigned short,
32097 vector unsigned short);
32098 vector signed short vec_adds (vector bool short, vector signed short);
32099 vector signed short vec_adds (vector signed short, vector bool short);
32100 vector signed short vec_adds (vector signed short, vector signed short);
32101 vector unsigned int vec_adds (vector bool int, vector unsigned int);
32102 vector unsigned int vec_adds (vector unsigned int, vector bool int);
32103 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
32104 vector signed int vec_adds (vector bool int, vector signed int);
32105 vector signed int vec_adds (vector signed int, vector bool int);
32106 vector signed int vec_adds (vector signed int, vector signed int);
32108 vector signed int vec_vaddsws (vector bool int, vector signed int);
32109 vector signed int vec_vaddsws (vector signed int, vector bool int);
32110 vector signed int vec_vaddsws (vector signed int, vector signed int);
32112 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
32113 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
32114 vector unsigned int vec_vadduws (vector unsigned int,
32115 vector unsigned int);
32117 vector signed short vec_vaddshs (vector bool short,
32118 vector signed short);
32119 vector signed short vec_vaddshs (vector signed short,
32120 vector bool short);
32121 vector signed short vec_vaddshs (vector signed short,
32122 vector signed short);
32124 vector unsigned short vec_vadduhs (vector bool short,
32125 vector unsigned short);
32126 vector unsigned short vec_vadduhs (vector unsigned short,
32127 vector bool short);
32128 vector unsigned short vec_vadduhs (vector unsigned short,
32129 vector unsigned short);
32131 vector signed char vec_vaddsbs (vector bool char, vector signed char);
32132 vector signed char vec_vaddsbs (vector signed char, vector bool char);
32133 vector signed char vec_vaddsbs (vector signed char, vector signed char);
32135 vector unsigned char vec_vaddubs (vector bool char,
32136 vector unsigned char);
32137 vector unsigned char vec_vaddubs (vector unsigned char,
32139 vector unsigned char vec_vaddubs (vector unsigned char,
32140 vector unsigned char);
32142 vector float vec_and (vector float, vector float);
32143 vector float vec_and (vector float, vector bool int);
32144 vector float vec_and (vector bool int, vector float);
32145 vector bool int vec_and (vector bool int, vector bool int);
32146 vector signed int vec_and (vector bool int, vector signed int);
32147 vector signed int vec_and (vector signed int, vector bool int);
32148 vector signed int vec_and (vector signed int, vector signed int);
32149 vector unsigned int vec_and (vector bool int, vector unsigned int);
32150 vector unsigned int vec_and (vector unsigned int, vector bool int);
32151 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
32152 vector bool short vec_and (vector bool short, vector bool short);
32153 vector signed short vec_and (vector bool short, vector signed short);
32154 vector signed short vec_and (vector signed short, vector bool short);
32155 vector signed short vec_and (vector signed short, vector signed short);
32156 vector unsigned short vec_and (vector bool short,
32157 vector unsigned short);
32158 vector unsigned short vec_and (vector unsigned short,
32159 vector bool short);
32160 vector unsigned short vec_and (vector unsigned short,
32161 vector unsigned short);
32162 vector signed char vec_and (vector bool char, vector signed char);
32163 vector bool char vec_and (vector bool char, vector bool char);
32164 vector signed char vec_and (vector signed char, vector bool char);
32165 vector signed char vec_and (vector signed char, vector signed char);
32166 vector unsigned char vec_and (vector bool char, vector unsigned char);
32167 vector unsigned char vec_and (vector unsigned char, vector bool char);
32168 vector unsigned char vec_and (vector unsigned char,
32169 vector unsigned char);
32171 vector float vec_andc (vector float, vector float);
32172 vector float vec_andc (vector float, vector bool int);
32173 vector float vec_andc (vector bool int, vector float);
32174 vector bool int vec_andc (vector bool int, vector bool int);
32175 vector signed int vec_andc (vector bool int, vector signed int);
32176 vector signed int vec_andc (vector signed int, vector bool int);
32177 vector signed int vec_andc (vector signed int, vector signed int);
32178 vector unsigned int vec_andc (vector bool int, vector unsigned int);
32179 vector unsigned int vec_andc (vector unsigned int, vector bool int);
32180 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
32181 vector bool short vec_andc (vector bool short, vector bool short);
32182 vector signed short vec_andc (vector bool short, vector signed short);
32183 vector signed short vec_andc (vector signed short, vector bool short);
32184 vector signed short vec_andc (vector signed short, vector signed short);
32185 vector unsigned short vec_andc (vector bool short,
32186 vector unsigned short);
32187 vector unsigned short vec_andc (vector unsigned short,
32188 vector bool short);
32189 vector unsigned short vec_andc (vector unsigned short,
32190 vector unsigned short);
32191 vector signed char vec_andc (vector bool char, vector signed char);
32192 vector bool char vec_andc (vector bool char, vector bool char);
32193 vector signed char vec_andc (vector signed char, vector bool char);
32194 vector signed char vec_andc (vector signed char, vector signed char);
32195 vector unsigned char vec_andc (vector bool char, vector unsigned char);
32196 vector unsigned char vec_andc (vector unsigned char, vector bool char);
32197 vector unsigned char vec_andc (vector unsigned char,
32198 vector unsigned char);
32200 vector unsigned char vec_avg (vector unsigned char,
32201 vector unsigned char);
32202 vector signed char vec_avg (vector signed char, vector signed char);
32203 vector unsigned short vec_avg (vector unsigned short,
32204 vector unsigned short);
32205 vector signed short vec_avg (vector signed short, vector signed short);
32206 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
32207 vector signed int vec_avg (vector signed int, vector signed int);
32209 vector signed int vec_vavgsw (vector signed int, vector signed int);
32211 vector unsigned int vec_vavguw (vector unsigned int,
32212 vector unsigned int);
32214 vector signed short vec_vavgsh (vector signed short,
32215 vector signed short);
32217 vector unsigned short vec_vavguh (vector unsigned short,
32218 vector unsigned short);
32220 vector signed char vec_vavgsb (vector signed char, vector signed char);
32222 vector unsigned char vec_vavgub (vector unsigned char,
32223 vector unsigned char);
32225 vector float vec_ceil (vector float);
32227 vector signed int vec_cmpb (vector float, vector float);
32229 vector bool char vec_cmpeq (vector signed char, vector signed char);
32230 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
32231 vector bool short vec_cmpeq (vector signed short, vector signed short);
32232 vector bool short vec_cmpeq (vector unsigned short,
32233 vector unsigned short);
32234 vector bool int vec_cmpeq (vector signed int, vector signed int);
32235 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
32236 vector bool int vec_cmpeq (vector float, vector float);
32238 vector bool int vec_vcmpeqfp (vector float, vector float);
32240 vector bool int vec_vcmpequw (vector signed int, vector signed int);
32241 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
32243 vector bool short vec_vcmpequh (vector signed short,
32244 vector signed short);
32245 vector bool short vec_vcmpequh (vector unsigned short,
32246 vector unsigned short);
32248 vector bool char vec_vcmpequb (vector signed char, vector signed char);
32249 vector bool char vec_vcmpequb (vector unsigned char,
32250 vector unsigned char);
32252 vector bool int vec_cmpge (vector float, vector float);
32254 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
32255 vector bool char vec_cmpgt (vector signed char, vector signed char);
32256 vector bool short vec_cmpgt (vector unsigned short,
32257 vector unsigned short);
32258 vector bool short vec_cmpgt (vector signed short, vector signed short);
32259 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
32260 vector bool int vec_cmpgt (vector signed int, vector signed int);
32261 vector bool int vec_cmpgt (vector float, vector float);
32263 vector bool int vec_vcmpgtfp (vector float, vector float);
32265 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
32267 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
32269 vector bool short vec_vcmpgtsh (vector signed short,
32270 vector signed short);
32272 vector bool short vec_vcmpgtuh (vector unsigned short,
32273 vector unsigned short);
32275 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
32277 vector bool char vec_vcmpgtub (vector unsigned char,
32278 vector unsigned char);
32280 vector bool int vec_cmple (vector float, vector float);
32282 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
32283 vector bool char vec_cmplt (vector signed char, vector signed char);
32284 vector bool short vec_cmplt (vector unsigned short,
32285 vector unsigned short);
32286 vector bool short vec_cmplt (vector signed short, vector signed short);
32287 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
32288 vector bool int vec_cmplt (vector signed int, vector signed int);
32289 vector bool int vec_cmplt (vector float, vector float);
32291 vector float vec_ctf (vector unsigned int, const int);
32292 vector float vec_ctf (vector signed int, const int);
32294 vector float vec_vcfsx (vector signed int, const int);
32296 vector float vec_vcfux (vector unsigned int, const int);
32298 vector signed int vec_cts (vector float, const int);
32300 vector unsigned int vec_ctu (vector float, const int);
32302 void vec_dss (const int);
32304 void vec_dssall (void);
32306 void vec_dst (const vector unsigned char *, int, const int);
32307 void vec_dst (const vector signed char *, int, const int);
32308 void vec_dst (const vector bool char *, int, const int);
32309 void vec_dst (const vector unsigned short *, int, const int);
32310 void vec_dst (const vector signed short *, int, const int);
32311 void vec_dst (const vector bool short *, int, const int);
32312 void vec_dst (const vector pixel *, int, const int);
32313 void vec_dst (const vector unsigned int *, int, const int);
32314 void vec_dst (const vector signed int *, int, const int);
32315 void vec_dst (const vector bool int *, int, const int);
32316 void vec_dst (const vector float *, int, const int);
32317 void vec_dst (const unsigned char *, int, const int);
32318 void vec_dst (const signed char *, int, const int);
32319 void vec_dst (const unsigned short *, int, const int);
32320 void vec_dst (const short *, int, const int);
32321 void vec_dst (const unsigned int *, int, const int);
32322 void vec_dst (const int *, int, const int);
32323 void vec_dst (const unsigned long *, int, const int);
32324 void vec_dst (const long *, int, const int);
32325 void vec_dst (const float *, int, const int);
32327 void vec_dstst (const vector unsigned char *, int, const int);
32328 void vec_dstst (const vector signed char *, int, const int);
32329 void vec_dstst (const vector bool char *, int, const int);
32330 void vec_dstst (const vector unsigned short *, int, const int);
32331 void vec_dstst (const vector signed short *, int, const int);
32332 void vec_dstst (const vector bool short *, int, const int);
32333 void vec_dstst (const vector pixel *, int, const int);
32334 void vec_dstst (const vector unsigned int *, int, const int);
32335 void vec_dstst (const vector signed int *, int, const int);
32336 void vec_dstst (const vector bool int *, int, const int);
32337 void vec_dstst (const vector float *, int, const int);
32338 void vec_dstst (const unsigned char *, int, const int);
32339 void vec_dstst (const signed char *, int, const int);
32340 void vec_dstst (const unsigned short *, int, const int);
32341 void vec_dstst (const short *, int, const int);
32342 void vec_dstst (const unsigned int *, int, const int);
32343 void vec_dstst (const int *, int, const int);
32344 void vec_dstst (const unsigned long *, int, const int);
32345 void vec_dstst (const long *, int, const int);
32346 void vec_dstst (const float *, int, const int);
32348 void vec_dststt (const vector unsigned char *, int, const int);
32349 void vec_dststt (const vector signed char *, int, const int);
32350 void vec_dststt (const vector bool char *, int, const int);
32351 void vec_dststt (const vector unsigned short *, int, const int);
32352 void vec_dststt (const vector signed short *, int, const int);
32353 void vec_dststt (const vector bool short *, int, const int);
32354 void vec_dststt (const vector pixel *, int, const int);
32355 void vec_dststt (const vector unsigned int *, int, const int);
32356 void vec_dststt (const vector signed int *, int, const int);
32357 void vec_dststt (const vector bool int *, int, const int);
32358 void vec_dststt (const vector float *, int, const int);
32359 void vec_dststt (const unsigned char *, int, const int);
32360 void vec_dststt (const signed char *, int, const int);
32361 void vec_dststt (const unsigned short *, int, const int);
32362 void vec_dststt (const short *, int, const int);
32363 void vec_dststt (const unsigned int *, int, const int);
32364 void vec_dststt (const int *, int, const int);
32365 void vec_dststt (const unsigned long *, int, const int);
32366 void vec_dststt (const long *, int, const int);
32367 void vec_dststt (const float *, int, const int);
32369 void vec_dstt (const vector unsigned char *, int, const int);
32370 void vec_dstt (const vector signed char *, int, const int);
32371 void vec_dstt (const vector bool char *, int, const int);
32372 void vec_dstt (const vector unsigned short *, int, const int);
32373 void vec_dstt (const vector signed short *, int, const int);
32374 void vec_dstt (const vector bool short *, int, const int);
32375 void vec_dstt (const vector pixel *, int, const int);
32376 void vec_dstt (const vector unsigned int *, int, const int);
32377 void vec_dstt (const vector signed int *, int, const int);
32378 void vec_dstt (const vector bool int *, int, const int);
32379 void vec_dstt (const vector float *, int, const int);
32380 void vec_dstt (const unsigned char *, int, const int);
32381 void vec_dstt (const signed char *, int, const int);
32382 void vec_dstt (const unsigned short *, int, const int);
32383 void vec_dstt (const short *, int, const int);
32384 void vec_dstt (const unsigned int *, int, const int);
32385 void vec_dstt (const int *, int, const int);
32386 void vec_dstt (const unsigned long *, int, const int);
32387 void vec_dstt (const long *, int, const int);
32388 void vec_dstt (const float *, int, const int);
32390 vector float vec_expte (vector float);
32392 vector float vec_floor (vector float);
32394 vector float vec_ld (int, const vector float *);
32395 vector float vec_ld (int, const float *);
32396 vector bool int vec_ld (int, const vector bool int *);
32397 vector signed int vec_ld (int, const vector signed int *);
32398 vector signed int vec_ld (int, const int *);
32399 vector signed int vec_ld (int, const long *);
32400 vector unsigned int vec_ld (int, const vector unsigned int *);
32401 vector unsigned int vec_ld (int, const unsigned int *);
32402 vector unsigned int vec_ld (int, const unsigned long *);
32403 vector bool short vec_ld (int, const vector bool short *);
32404 vector pixel vec_ld (int, const vector pixel *);
32405 vector signed short vec_ld (int, const vector signed short *);
32406 vector signed short vec_ld (int, const short *);
32407 vector unsigned short vec_ld (int, const vector unsigned short *);
32408 vector unsigned short vec_ld (int, const unsigned short *);
32409 vector bool char vec_ld (int, const vector bool char *);
32410 vector signed char vec_ld (int, const vector signed char *);
32411 vector signed char vec_ld (int, const signed char *);
32412 vector unsigned char vec_ld (int, const vector unsigned char *);
32413 vector unsigned char vec_ld (int, const unsigned char *);
32415 vector signed char vec_lde (int, const signed char *);
32416 vector unsigned char vec_lde (int, const unsigned char *);
32417 vector signed short vec_lde (int, const short *);
32418 vector unsigned short vec_lde (int, const unsigned short *);
32419 vector float vec_lde (int, const float *);
32420 vector signed int vec_lde (int, const int *);
32421 vector unsigned int vec_lde (int, const unsigned int *);
32422 vector signed int vec_lde (int, const long *);
32423 vector unsigned int vec_lde (int, const unsigned long *);
32425 vector float vec_lvewx (int, float *);
32426 vector signed int vec_lvewx (int, int *);
32427 vector unsigned int vec_lvewx (int, unsigned int *);
32428 vector signed int vec_lvewx (int, long *);
32429 vector unsigned int vec_lvewx (int, unsigned long *);
32431 vector signed short vec_lvehx (int, short *);
32432 vector unsigned short vec_lvehx (int, unsigned short *);
32434 vector signed char vec_lvebx (int, char *);
32435 vector unsigned char vec_lvebx (int, unsigned char *);
32437 vector float vec_ldl (int, const vector float *);
32438 vector float vec_ldl (int, const float *);
32439 vector bool int vec_ldl (int, const vector bool int *);
32440 vector signed int vec_ldl (int, const vector signed int *);
32441 vector signed int vec_ldl (int, const int *);
32442 vector signed int vec_ldl (int, const long *);
32443 vector unsigned int vec_ldl (int, const vector unsigned int *);
32444 vector unsigned int vec_ldl (int, const unsigned int *);
32445 vector unsigned int vec_ldl (int, const unsigned long *);
32446 vector bool short vec_ldl (int, const vector bool short *);
32447 vector pixel vec_ldl (int, const vector pixel *);
32448 vector signed short vec_ldl (int, const vector signed short *);
32449 vector signed short vec_ldl (int, const short *);
32450 vector unsigned short vec_ldl (int, const vector unsigned short *);
32451 vector unsigned short vec_ldl (int, const unsigned short *);
32452 vector bool char vec_ldl (int, const vector bool char *);
32453 vector signed char vec_ldl (int, const vector signed char *);
32454 vector signed char vec_ldl (int, const signed char *);
32455 vector unsigned char vec_ldl (int, const vector unsigned char *);
32456 vector unsigned char vec_ldl (int, const unsigned char *);
32458 vector float vec_loge (vector float);
32460 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
32461 vector unsigned char vec_lvsl (int, const volatile signed char *);
32462 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
32463 vector unsigned char vec_lvsl (int, const volatile short *);
32464 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
32465 vector unsigned char vec_lvsl (int, const volatile int *);
32466 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
32467 vector unsigned char vec_lvsl (int, const volatile long *);
32468 vector unsigned char vec_lvsl (int, const volatile float *);
32470 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
32471 vector unsigned char vec_lvsr (int, const volatile signed char *);
32472 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
32473 vector unsigned char vec_lvsr (int, const volatile short *);
32474 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
32475 vector unsigned char vec_lvsr (int, const volatile int *);
32476 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
32477 vector unsigned char vec_lvsr (int, const volatile long *);
32478 vector unsigned char vec_lvsr (int, const volatile float *);
32480 vector float vec_madd (vector float, vector float, vector float);
32482 vector signed short vec_madds (vector signed short,
32483 vector signed short,
32484 vector signed short);
32486 vector unsigned char vec_max (vector bool char, vector unsigned char);
32487 vector unsigned char vec_max (vector unsigned char, vector bool char);
32488 vector unsigned char vec_max (vector unsigned char,
32489 vector unsigned char);
32490 vector signed char vec_max (vector bool char, vector signed char);
32491 vector signed char vec_max (vector signed char, vector bool char);
32492 vector signed char vec_max (vector signed char, vector signed char);
32493 vector unsigned short vec_max (vector bool short,
32494 vector unsigned short);
32495 vector unsigned short vec_max (vector unsigned short,
32496 vector bool short);
32497 vector unsigned short vec_max (vector unsigned short,
32498 vector unsigned short);
32499 vector signed short vec_max (vector bool short, vector signed short);
32500 vector signed short vec_max (vector signed short, vector bool short);
32501 vector signed short vec_max (vector signed short, vector signed short);
32502 vector unsigned int vec_max (vector bool int, vector unsigned int);
32503 vector unsigned int vec_max (vector unsigned int, vector bool int);
32504 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
32505 vector signed int vec_max (vector bool int, vector signed int);
32506 vector signed int vec_max (vector signed int, vector bool int);
32507 vector signed int vec_max (vector signed int, vector signed int);
32508 vector float vec_max (vector float, vector float);
32510 vector float vec_vmaxfp (vector float, vector float);
32512 vector signed int vec_vmaxsw (vector bool int, vector signed int);
32513 vector signed int vec_vmaxsw (vector signed int, vector bool int);
32514 vector signed int vec_vmaxsw (vector signed int, vector signed int);
32516 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
32517 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
32518 vector unsigned int vec_vmaxuw (vector unsigned int,
32519 vector unsigned int);
32521 vector signed short vec_vmaxsh (vector bool short, vector signed short);
32522 vector signed short vec_vmaxsh (vector signed short, vector bool short);
32523 vector signed short vec_vmaxsh (vector signed short,
32524 vector signed short);
32526 vector unsigned short vec_vmaxuh (vector bool short,
32527 vector unsigned short);
32528 vector unsigned short vec_vmaxuh (vector unsigned short,
32529 vector bool short);
32530 vector unsigned short vec_vmaxuh (vector unsigned short,
32531 vector unsigned short);
32533 vector signed char vec_vmaxsb (vector bool char, vector signed char);
32534 vector signed char vec_vmaxsb (vector signed char, vector bool char);
32535 vector signed char vec_vmaxsb (vector signed char, vector signed char);
32537 vector unsigned char vec_vmaxub (vector bool char,
32538 vector unsigned char);
32539 vector unsigned char vec_vmaxub (vector unsigned char,
32541 vector unsigned char vec_vmaxub (vector unsigned char,
32542 vector unsigned char);
32544 vector bool char vec_mergeh (vector bool char, vector bool char);
32545 vector signed char vec_mergeh (vector signed char, vector signed char);
32546 vector unsigned char vec_mergeh (vector unsigned char,
32547 vector unsigned char);
32548 vector bool short vec_mergeh (vector bool short, vector bool short);
32549 vector pixel vec_mergeh (vector pixel, vector pixel);
32550 vector signed short vec_mergeh (vector signed short,
32551 vector signed short);
32552 vector unsigned short vec_mergeh (vector unsigned short,
32553 vector unsigned short);
32554 vector float vec_mergeh (vector float, vector float);
32555 vector bool int vec_mergeh (vector bool int, vector bool int);
32556 vector signed int vec_mergeh (vector signed int, vector signed int);
32557 vector unsigned int vec_mergeh (vector unsigned int,
32558 vector unsigned int);
32560 vector float vec_vmrghw (vector float, vector float);
32561 vector bool int vec_vmrghw (vector bool int, vector bool int);
32562 vector signed int vec_vmrghw (vector signed int, vector signed int);
32563 vector unsigned int vec_vmrghw (vector unsigned int,
32564 vector unsigned int);
32566 vector bool short vec_vmrghh (vector bool short, vector bool short);
32567 vector signed short vec_vmrghh (vector signed short,
32568 vector signed short);
32569 vector unsigned short vec_vmrghh (vector unsigned short,
32570 vector unsigned short);
32571 vector pixel vec_vmrghh (vector pixel, vector pixel);
32573 vector bool char vec_vmrghb (vector bool char, vector bool char);
32574 vector signed char vec_vmrghb (vector signed char, vector signed char);
32575 vector unsigned char vec_vmrghb (vector unsigned char,
32576 vector unsigned char);
32578 vector bool char vec_mergel (vector bool char, vector bool char);
32579 vector signed char vec_mergel (vector signed char, vector signed char);
32580 vector unsigned char vec_mergel (vector unsigned char,
32581 vector unsigned char);
32582 vector bool short vec_mergel (vector bool short, vector bool short);
32583 vector pixel vec_mergel (vector pixel, vector pixel);
32584 vector signed short vec_mergel (vector signed short,
32585 vector signed short);
32586 vector unsigned short vec_mergel (vector unsigned short,
32587 vector unsigned short);
32588 vector float vec_mergel (vector float, vector float);
32589 vector bool int vec_mergel (vector bool int, vector bool int);
32590 vector signed int vec_mergel (vector signed int, vector signed int);
32591 vector unsigned int vec_mergel (vector unsigned int,
32592 vector unsigned int);
32594 vector float vec_vmrglw (vector float, vector float);
32595 vector signed int vec_vmrglw (vector signed int, vector signed int);
32596 vector unsigned int vec_vmrglw (vector unsigned int,
32597 vector unsigned int);
32598 vector bool int vec_vmrglw (vector bool int, vector bool int);
32600 vector bool short vec_vmrglh (vector bool short, vector bool short);
32601 vector signed short vec_vmrglh (vector signed short,
32602 vector signed short);
32603 vector unsigned short vec_vmrglh (vector unsigned short,
32604 vector unsigned short);
32605 vector pixel vec_vmrglh (vector pixel, vector pixel);
32607 vector bool char vec_vmrglb (vector bool char, vector bool char);
32608 vector signed char vec_vmrglb (vector signed char, vector signed char);
32609 vector unsigned char vec_vmrglb (vector unsigned char,
32610 vector unsigned char);
32612 vector unsigned short vec_mfvscr (void);
32614 vector unsigned char vec_min (vector bool char, vector unsigned char);
32615 vector unsigned char vec_min (vector unsigned char, vector bool char);
32616 vector unsigned char vec_min (vector unsigned char,
32617 vector unsigned char);
32618 vector signed char vec_min (vector bool char, vector signed char);
32619 vector signed char vec_min (vector signed char, vector bool char);
32620 vector signed char vec_min (vector signed char, vector signed char);
32621 vector unsigned short vec_min (vector bool short,
32622 vector unsigned short);
32623 vector unsigned short vec_min (vector unsigned short,
32624 vector bool short);
32625 vector unsigned short vec_min (vector unsigned short,
32626 vector unsigned short);
32627 vector signed short vec_min (vector bool short, vector signed short);
32628 vector signed short vec_min (vector signed short, vector bool short);
32629 vector signed short vec_min (vector signed short, vector signed short);
32630 vector unsigned int vec_min (vector bool int, vector unsigned int);
32631 vector unsigned int vec_min (vector unsigned int, vector bool int);
32632 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
32633 vector signed int vec_min (vector bool int, vector signed int);
32634 vector signed int vec_min (vector signed int, vector bool int);
32635 vector signed int vec_min (vector signed int, vector signed int);
32636 vector float vec_min (vector float, vector float);
32638 vector float vec_vminfp (vector float, vector float);
32640 vector signed int vec_vminsw (vector bool int, vector signed int);
32641 vector signed int vec_vminsw (vector signed int, vector bool int);
32642 vector signed int vec_vminsw (vector signed int, vector signed int);
32644 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
32645 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
32646 vector unsigned int vec_vminuw (vector unsigned int,
32647 vector unsigned int);
32649 vector signed short vec_vminsh (vector bool short, vector signed short);
32650 vector signed short vec_vminsh (vector signed short, vector bool short);
32651 vector signed short vec_vminsh (vector signed short,
32652 vector signed short);
32654 vector unsigned short vec_vminuh (vector bool short,
32655 vector unsigned short);
32656 vector unsigned short vec_vminuh (vector unsigned short,
32657 vector bool short);
32658 vector unsigned short vec_vminuh (vector unsigned short,
32659 vector unsigned short);
32661 vector signed char vec_vminsb (vector bool char, vector signed char);
32662 vector signed char vec_vminsb (vector signed char, vector bool char);
32663 vector signed char vec_vminsb (vector signed char, vector signed char);
32665 vector unsigned char vec_vminub (vector bool char,
32666 vector unsigned char);
32667 vector unsigned char vec_vminub (vector unsigned char,
32669 vector unsigned char vec_vminub (vector unsigned char,
32670 vector unsigned char);
32672 vector signed short vec_mladd (vector signed short,
32673 vector signed short,
32674 vector signed short);
32675 vector signed short vec_mladd (vector signed short,
32676 vector unsigned short,
32677 vector unsigned short);
32678 vector signed short vec_mladd (vector unsigned short,
32679 vector signed short,
32680 vector signed short);
32681 vector unsigned short vec_mladd (vector unsigned short,
32682 vector unsigned short,
32683 vector unsigned short);
32685 vector signed short vec_mradds (vector signed short,
32686 vector signed short,
32687 vector signed short);
32689 vector unsigned int vec_msum (vector unsigned char,
32690 vector unsigned char,
32691 vector unsigned int);
32692 vector signed int vec_msum (vector signed char,
32693 vector unsigned char,
32694 vector signed int);
32695 vector unsigned int vec_msum (vector unsigned short,
32696 vector unsigned short,
32697 vector unsigned int);
32698 vector signed int vec_msum (vector signed short,
32699 vector signed short,
32700 vector signed int);
32702 vector signed int vec_vmsumshm (vector signed short,
32703 vector signed short,
32704 vector signed int);
32706 vector unsigned int vec_vmsumuhm (vector unsigned short,
32707 vector unsigned short,
32708 vector unsigned int);
32710 vector signed int vec_vmsummbm (vector signed char,
32711 vector unsigned char,
32712 vector signed int);
32714 vector unsigned int vec_vmsumubm (vector unsigned char,
32715 vector unsigned char,
32716 vector unsigned int);
32718 vector unsigned int vec_msums (vector unsigned short,
32719 vector unsigned short,
32720 vector unsigned int);
32721 vector signed int vec_msums (vector signed short,
32722 vector signed short,
32723 vector signed int);
32725 vector signed int vec_vmsumshs (vector signed short,
32726 vector signed short,
32727 vector signed int);
32729 vector unsigned int vec_vmsumuhs (vector unsigned short,
32730 vector unsigned short,
32731 vector unsigned int);
32733 void vec_mtvscr (vector signed int);
32734 void vec_mtvscr (vector unsigned int);
32735 void vec_mtvscr (vector bool int);
32736 void vec_mtvscr (vector signed short);
32737 void vec_mtvscr (vector unsigned short);
32738 void vec_mtvscr (vector bool short);
32739 void vec_mtvscr (vector pixel);
32740 void vec_mtvscr (vector signed char);
32741 void vec_mtvscr (vector unsigned char);
32742 void vec_mtvscr (vector bool char);
32744 vector unsigned short vec_mule (vector unsigned char,
32745 vector unsigned char);
32746 vector signed short vec_mule (vector signed char,
32747 vector signed char);
32748 vector unsigned int vec_mule (vector unsigned short,
32749 vector unsigned short);
32750 vector signed int vec_mule (vector signed short, vector signed short);
32752 vector signed int vec_vmulesh (vector signed short,
32753 vector signed short);
32755 vector unsigned int vec_vmuleuh (vector unsigned short,
32756 vector unsigned short);
32758 vector signed short vec_vmulesb (vector signed char,
32759 vector signed char);
32761 vector unsigned short vec_vmuleub (vector unsigned char,
32762 vector unsigned char);
32764 vector unsigned short vec_mulo (vector unsigned char,
32765 vector unsigned char);
32766 vector signed short vec_mulo (vector signed char, vector signed char);
32767 vector unsigned int vec_mulo (vector unsigned short,
32768 vector unsigned short);
32769 vector signed int vec_mulo (vector signed short, vector signed short);
32771 vector signed int vec_vmulosh (vector signed short,
32772 vector signed short);
32774 vector unsigned int vec_vmulouh (vector unsigned short,
32775 vector unsigned short);
32777 vector signed short vec_vmulosb (vector signed char,
32778 vector signed char);
32780 vector unsigned short vec_vmuloub (vector unsigned char,
32781 vector unsigned char);
32783 vector float vec_nmsub (vector float, vector float, vector float);
32785 vector float vec_nor (vector float, vector float);
32786 vector signed int vec_nor (vector signed int, vector signed int);
32787 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
32788 vector bool int vec_nor (vector bool int, vector bool int);
32789 vector signed short vec_nor (vector signed short, vector signed short);
32790 vector unsigned short vec_nor (vector unsigned short,
32791 vector unsigned short);
32792 vector bool short vec_nor (vector bool short, vector bool short);
32793 vector signed char vec_nor (vector signed char, vector signed char);
32794 vector unsigned char vec_nor (vector unsigned char,
32795 vector unsigned char);
32796 vector bool char vec_nor (vector bool char, vector bool char);
32798 vector float vec_or (vector float, vector float);
32799 vector float vec_or (vector float, vector bool int);
32800 vector float vec_or (vector bool int, vector float);
32801 vector bool int vec_or (vector bool int, vector bool int);
32802 vector signed int vec_or (vector bool int, vector signed int);
32803 vector signed int vec_or (vector signed int, vector bool int);
32804 vector signed int vec_or (vector signed int, vector signed int);
32805 vector unsigned int vec_or (vector bool int, vector unsigned int);
32806 vector unsigned int vec_or (vector unsigned int, vector bool int);
32807 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
32808 vector bool short vec_or (vector bool short, vector bool short);
32809 vector signed short vec_or (vector bool short, vector signed short);
32810 vector signed short vec_or (vector signed short, vector bool short);
32811 vector signed short vec_or (vector signed short, vector signed short);
32812 vector unsigned short vec_or (vector bool short, vector unsigned short);
32813 vector unsigned short vec_or (vector unsigned short, vector bool short);
32814 vector unsigned short vec_or (vector unsigned short,
32815 vector unsigned short);
32816 vector signed char vec_or (vector bool char, vector signed char);
32817 vector bool char vec_or (vector bool char, vector bool char);
32818 vector signed char vec_or (vector signed char, vector bool char);
32819 vector signed char vec_or (vector signed char, vector signed char);
32820 vector unsigned char vec_or (vector bool char, vector unsigned char);
32821 vector unsigned char vec_or (vector unsigned char, vector bool char);
32822 vector unsigned char vec_or (vector unsigned char,
32823 vector unsigned char);
32825 vector signed char vec_pack (vector signed short, vector signed short);
32826 vector unsigned char vec_pack (vector unsigned short,
32827 vector unsigned short);
32828 vector bool char vec_pack (vector bool short, vector bool short);
32829 vector signed short vec_pack (vector signed int, vector signed int);
32830 vector unsigned short vec_pack (vector unsigned int,
32831 vector unsigned int);
32832 vector bool short vec_pack (vector bool int, vector bool int);
32834 vector bool short vec_vpkuwum (vector bool int, vector bool int);
32835 vector signed short vec_vpkuwum (vector signed int, vector signed int);
32836 vector unsigned short vec_vpkuwum (vector unsigned int,
32837 vector unsigned int);
32839 vector bool char vec_vpkuhum (vector bool short, vector bool short);
32840 vector signed char vec_vpkuhum (vector signed short,
32841 vector signed short);
32842 vector unsigned char vec_vpkuhum (vector unsigned short,
32843 vector unsigned short);
32845 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
32847 vector unsigned char vec_packs (vector unsigned short,
32848 vector unsigned short);
32849 vector signed char vec_packs (vector signed short, vector signed short);
32850 vector unsigned short vec_packs (vector unsigned int,
32851 vector unsigned int);
32852 vector signed short vec_packs (vector signed int, vector signed int);
32854 vector signed short vec_vpkswss (vector signed int, vector signed int);
32856 vector unsigned short vec_vpkuwus (vector unsigned int,
32857 vector unsigned int);
32859 vector signed char vec_vpkshss (vector signed short,
32860 vector signed short);
32862 vector unsigned char vec_vpkuhus (vector unsigned short,
32863 vector unsigned short);
32865 vector unsigned char vec_packsu (vector unsigned short,
32866 vector unsigned short);
32867 vector unsigned char vec_packsu (vector signed short,
32868 vector signed short);
32869 vector unsigned short vec_packsu (vector unsigned int,
32870 vector unsigned int);
32871 vector unsigned short vec_packsu (vector signed int, vector signed int);
32873 vector unsigned short vec_vpkswus (vector signed int,
32874 vector signed int);
32876 vector unsigned char vec_vpkshus (vector signed short,
32877 vector signed short);
32879 vector float vec_perm (vector float,
32881 vector unsigned char);
32882 vector signed int vec_perm (vector signed int,
32884 vector unsigned char);
32885 vector unsigned int vec_perm (vector unsigned int,
32886 vector unsigned int,
32887 vector unsigned char);
32888 vector bool int vec_perm (vector bool int,
32890 vector unsigned char);
32891 vector signed short vec_perm (vector signed short,
32892 vector signed short,
32893 vector unsigned char);
32894 vector unsigned short vec_perm (vector unsigned short,
32895 vector unsigned short,
32896 vector unsigned char);
32897 vector bool short vec_perm (vector bool short,
32899 vector unsigned char);
32900 vector pixel vec_perm (vector pixel,
32902 vector unsigned char);
32903 vector signed char vec_perm (vector signed char,
32904 vector signed char,
32905 vector unsigned char);
32906 vector unsigned char vec_perm (vector unsigned char,
32907 vector unsigned char,
32908 vector unsigned char);
32909 vector bool char vec_perm (vector bool char,
32911 vector unsigned char);
32913 vector float vec_re (vector float);
32915 vector signed char vec_rl (vector signed char,
32916 vector unsigned char);
32917 vector unsigned char vec_rl (vector unsigned char,
32918 vector unsigned char);
32919 vector signed short vec_rl (vector signed short, vector unsigned short);
32920 vector unsigned short vec_rl (vector unsigned short,
32921 vector unsigned short);
32922 vector signed int vec_rl (vector signed int, vector unsigned int);
32923 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
32925 vector signed int vec_vrlw (vector signed int, vector unsigned int);
32926 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
32928 vector signed short vec_vrlh (vector signed short,
32929 vector unsigned short);
32930 vector unsigned short vec_vrlh (vector unsigned short,
32931 vector unsigned short);
32933 vector signed char vec_vrlb (vector signed char, vector unsigned char);
32934 vector unsigned char vec_vrlb (vector unsigned char,
32935 vector unsigned char);
32937 vector float vec_round (vector float);
32939 vector float vec_rsqrte (vector float);
32941 vector float vec_sel (vector float, vector float, vector bool int);
32942 vector float vec_sel (vector float, vector float, vector unsigned int);
32943 vector signed int vec_sel (vector signed int,
32946 vector signed int vec_sel (vector signed int,
32948 vector unsigned int);
32949 vector unsigned int vec_sel (vector unsigned int,
32950 vector unsigned int,
32952 vector unsigned int vec_sel (vector unsigned int,
32953 vector unsigned int,
32954 vector unsigned int);
32955 vector bool int vec_sel (vector bool int,
32958 vector bool int vec_sel (vector bool int,
32960 vector unsigned int);
32961 vector signed short vec_sel (vector signed short,
32962 vector signed short,
32963 vector bool short);
32964 vector signed short vec_sel (vector signed short,
32965 vector signed short,
32966 vector unsigned short);
32967 vector unsigned short vec_sel (vector unsigned short,
32968 vector unsigned short,
32969 vector bool short);
32970 vector unsigned short vec_sel (vector unsigned short,
32971 vector unsigned short,
32972 vector unsigned short);
32973 vector bool short vec_sel (vector bool short,
32975 vector bool short);
32976 vector bool short vec_sel (vector bool short,
32978 vector unsigned short);
32979 vector signed char vec_sel (vector signed char,
32980 vector signed char,
32982 vector signed char vec_sel (vector signed char,
32983 vector signed char,
32984 vector unsigned char);
32985 vector unsigned char vec_sel (vector unsigned char,
32986 vector unsigned char,
32988 vector unsigned char vec_sel (vector unsigned char,
32989 vector unsigned char,
32990 vector unsigned char);
32991 vector bool char vec_sel (vector bool char,
32994 vector bool char vec_sel (vector bool char,
32996 vector unsigned char);
32998 vector signed char vec_sl (vector signed char,
32999 vector unsigned char);
33000 vector unsigned char vec_sl (vector unsigned char,
33001 vector unsigned char);
33002 vector signed short vec_sl (vector signed short, vector unsigned short);
33003 vector unsigned short vec_sl (vector unsigned short,
33004 vector unsigned short);
33005 vector signed int vec_sl (vector signed int, vector unsigned int);
33006 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
33008 vector signed int vec_vslw (vector signed int, vector unsigned int);
33009 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
33011 vector signed short vec_vslh (vector signed short,
33012 vector unsigned short);
33013 vector unsigned short vec_vslh (vector unsigned short,
33014 vector unsigned short);
33016 vector signed char vec_vslb (vector signed char, vector unsigned char);
33017 vector unsigned char vec_vslb (vector unsigned char,
33018 vector unsigned char);
33020 vector float vec_sld (vector float, vector float, const int);
33021 vector signed int vec_sld (vector signed int,
33024 vector unsigned int vec_sld (vector unsigned int,
33025 vector unsigned int,
33027 vector bool int vec_sld (vector bool int,
33030 vector signed short vec_sld (vector signed short,
33031 vector signed short,
33033 vector unsigned short vec_sld (vector unsigned short,
33034 vector unsigned short,
33036 vector bool short vec_sld (vector bool short,
33039 vector pixel vec_sld (vector pixel,
33042 vector signed char vec_sld (vector signed char,
33043 vector signed char,
33045 vector unsigned char vec_sld (vector unsigned char,
33046 vector unsigned char,
33048 vector bool char vec_sld (vector bool char,
33052 vector signed int vec_sll (vector signed int,
33053 vector unsigned int);
33054 vector signed int vec_sll (vector signed int,
33055 vector unsigned short);
33056 vector signed int vec_sll (vector signed int,
33057 vector unsigned char);
33058 vector unsigned int vec_sll (vector unsigned int,
33059 vector unsigned int);
33060 vector unsigned int vec_sll (vector unsigned int,
33061 vector unsigned short);
33062 vector unsigned int vec_sll (vector unsigned int,
33063 vector unsigned char);
33064 vector bool int vec_sll (vector bool int,
33065 vector unsigned int);
33066 vector bool int vec_sll (vector bool int,
33067 vector unsigned short);
33068 vector bool int vec_sll (vector bool int,
33069 vector unsigned char);
33070 vector signed short vec_sll (vector signed short,
33071 vector unsigned int);
33072 vector signed short vec_sll (vector signed short,
33073 vector unsigned short);
33074 vector signed short vec_sll (vector signed short,
33075 vector unsigned char);
33076 vector unsigned short vec_sll (vector unsigned short,
33077 vector unsigned int);
33078 vector unsigned short vec_sll (vector unsigned short,
33079 vector unsigned short);
33080 vector unsigned short vec_sll (vector unsigned short,
33081 vector unsigned char);
33082 vector bool short vec_sll (vector bool short, vector unsigned int);
33083 vector bool short vec_sll (vector bool short, vector unsigned short);
33084 vector bool short vec_sll (vector bool short, vector unsigned char);
33085 vector pixel vec_sll (vector pixel, vector unsigned int);
33086 vector pixel vec_sll (vector pixel, vector unsigned short);
33087 vector pixel vec_sll (vector pixel, vector unsigned char);
33088 vector signed char vec_sll (vector signed char, vector unsigned int);
33089 vector signed char vec_sll (vector signed char, vector unsigned short);
33090 vector signed char vec_sll (vector signed char, vector unsigned char);
33091 vector unsigned char vec_sll (vector unsigned char,
33092 vector unsigned int);
33093 vector unsigned char vec_sll (vector unsigned char,
33094 vector unsigned short);
33095 vector unsigned char vec_sll (vector unsigned char,
33096 vector unsigned char);
33097 vector bool char vec_sll (vector bool char, vector unsigned int);
33098 vector bool char vec_sll (vector bool char, vector unsigned short);
33099 vector bool char vec_sll (vector bool char, vector unsigned char);
33101 vector float vec_slo (vector float, vector signed char);
33102 vector float vec_slo (vector float, vector unsigned char);
33103 vector signed int vec_slo (vector signed int, vector signed char);
33104 vector signed int vec_slo (vector signed int, vector unsigned char);
33105 vector unsigned int vec_slo (vector unsigned int, vector signed char);
33106 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
33107 vector signed short vec_slo (vector signed short, vector signed char);
33108 vector signed short vec_slo (vector signed short, vector unsigned char);
33109 vector unsigned short vec_slo (vector unsigned short,
33110 vector signed char);
33111 vector unsigned short vec_slo (vector unsigned short,
33112 vector unsigned char);
33113 vector pixel vec_slo (vector pixel, vector signed char);
33114 vector pixel vec_slo (vector pixel, vector unsigned char);
33115 vector signed char vec_slo (vector signed char, vector signed char);
33116 vector signed char vec_slo (vector signed char, vector unsigned char);
33117 vector unsigned char vec_slo (vector unsigned char, vector signed char);
33118 vector unsigned char vec_slo (vector unsigned char,
33119 vector unsigned char);
33121 vector signed char vec_splat (vector signed char, const int);
33122 vector unsigned char vec_splat (vector unsigned char, const int);
33123 vector bool char vec_splat (vector bool char, const int);
33124 vector signed short vec_splat (vector signed short, const int);
33125 vector unsigned short vec_splat (vector unsigned short, const int);
33126 vector bool short vec_splat (vector bool short, const int);
33127 vector pixel vec_splat (vector pixel, const int);
33128 vector float vec_splat (vector float, const int);
33129 vector signed int vec_splat (vector signed int, const int);
33130 vector unsigned int vec_splat (vector unsigned int, const int);
33131 vector bool int vec_splat (vector bool int, const int);
33133 vector float vec_vspltw (vector float, const int);
33134 vector signed int vec_vspltw (vector signed int, const int);
33135 vector unsigned int vec_vspltw (vector unsigned int, const int);
33136 vector bool int vec_vspltw (vector bool int, const int);
33138 vector bool short vec_vsplth (vector bool short, const int);
33139 vector signed short vec_vsplth (vector signed short, const int);
33140 vector unsigned short vec_vsplth (vector unsigned short, const int);
33141 vector pixel vec_vsplth (vector pixel, const int);
33143 vector signed char vec_vspltb (vector signed char, const int);
33144 vector unsigned char vec_vspltb (vector unsigned char, const int);
33145 vector bool char vec_vspltb (vector bool char, const int);
33147 vector signed char vec_splat_s8 (const int);
33149 vector signed short vec_splat_s16 (const int);
33151 vector signed int vec_splat_s32 (const int);
33153 vector unsigned char vec_splat_u8 (const int);
33155 vector unsigned short vec_splat_u16 (const int);
33157 vector unsigned int vec_splat_u32 (const int);
33159 vector signed char vec_sr (vector signed char, vector unsigned char);
33160 vector unsigned char vec_sr (vector unsigned char,
33161 vector unsigned char);
33162 vector signed short vec_sr (vector signed short,
33163 vector unsigned short);
33164 vector unsigned short vec_sr (vector unsigned short,
33165 vector unsigned short);
33166 vector signed int vec_sr (vector signed int, vector unsigned int);
33167 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
33169 vector signed int vec_vsrw (vector signed int, vector unsigned int);
33170 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
33172 vector signed short vec_vsrh (vector signed short,
33173 vector unsigned short);
33174 vector unsigned short vec_vsrh (vector unsigned short,
33175 vector unsigned short);
33177 vector signed char vec_vsrb (vector signed char, vector unsigned char);
33178 vector unsigned char vec_vsrb (vector unsigned char,
33179 vector unsigned char);
33181 vector signed char vec_sra (vector signed char, vector unsigned char);
33182 vector unsigned char vec_sra (vector unsigned char,
33183 vector unsigned char);
33184 vector signed short vec_sra (vector signed short,
33185 vector unsigned short);
33186 vector unsigned short vec_sra (vector unsigned short,
33187 vector unsigned short);
33188 vector signed int vec_sra (vector signed int, vector unsigned int);
33189 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
33191 vector signed int vec_vsraw (vector signed int, vector unsigned int);
33192 vector unsigned int vec_vsraw (vector unsigned int,
33193 vector unsigned int);
33195 vector signed short vec_vsrah (vector signed short,
33196 vector unsigned short);
33197 vector unsigned short vec_vsrah (vector unsigned short,
33198 vector unsigned short);
33200 vector signed char vec_vsrab (vector signed char, vector unsigned char);
33201 vector unsigned char vec_vsrab (vector unsigned char,
33202 vector unsigned char);
33204 vector signed int vec_srl (vector signed int, vector unsigned int);
33205 vector signed int vec_srl (vector signed int, vector unsigned short);
33206 vector signed int vec_srl (vector signed int, vector unsigned char);
33207 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
33208 vector unsigned int vec_srl (vector unsigned int,
33209 vector unsigned short);
33210 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
33211 vector bool int vec_srl (vector bool int, vector unsigned int);
33212 vector bool int vec_srl (vector bool int, vector unsigned short);
33213 vector bool int vec_srl (vector bool int, vector unsigned char);
33214 vector signed short vec_srl (vector signed short, vector unsigned int);
33215 vector signed short vec_srl (vector signed short,
33216 vector unsigned short);
33217 vector signed short vec_srl (vector signed short, vector unsigned char);
33218 vector unsigned short vec_srl (vector unsigned short,
33219 vector unsigned int);
33220 vector unsigned short vec_srl (vector unsigned short,
33221 vector unsigned short);
33222 vector unsigned short vec_srl (vector unsigned short,
33223 vector unsigned char);
33224 vector bool short vec_srl (vector bool short, vector unsigned int);
33225 vector bool short vec_srl (vector bool short, vector unsigned short);
33226 vector bool short vec_srl (vector bool short, vector unsigned char);
33227 vector pixel vec_srl (vector pixel, vector unsigned int);
33228 vector pixel vec_srl (vector pixel, vector unsigned short);
33229 vector pixel vec_srl (vector pixel, vector unsigned char);
33230 vector signed char vec_srl (vector signed char, vector unsigned int);
33231 vector signed char vec_srl (vector signed char, vector unsigned short);
33232 vector signed char vec_srl (vector signed char, vector unsigned char);
33233 vector unsigned char vec_srl (vector unsigned char,
33234 vector unsigned int);
33235 vector unsigned char vec_srl (vector unsigned char,
33236 vector unsigned short);
33237 vector unsigned char vec_srl (vector unsigned char,
33238 vector unsigned char);
33239 vector bool char vec_srl (vector bool char, vector unsigned int);
33240 vector bool char vec_srl (vector bool char, vector unsigned short);
33241 vector bool char vec_srl (vector bool char, vector unsigned char);
33243 vector float vec_sro (vector float, vector signed char);
33244 vector float vec_sro (vector float, vector unsigned char);
33245 vector signed int vec_sro (vector signed int, vector signed char);
33246 vector signed int vec_sro (vector signed int, vector unsigned char);
33247 vector unsigned int vec_sro (vector unsigned int, vector signed char);
33248 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
33249 vector signed short vec_sro (vector signed short, vector signed char);
33250 vector signed short vec_sro (vector signed short, vector unsigned char);
33251 vector unsigned short vec_sro (vector unsigned short,
33252 vector signed char);
33253 vector unsigned short vec_sro (vector unsigned short,
33254 vector unsigned char);
33255 vector pixel vec_sro (vector pixel, vector signed char);
33256 vector pixel vec_sro (vector pixel, vector unsigned char);
33257 vector signed char vec_sro (vector signed char, vector signed char);
33258 vector signed char vec_sro (vector signed char, vector unsigned char);
33259 vector unsigned char vec_sro (vector unsigned char, vector signed char);
33260 vector unsigned char vec_sro (vector unsigned char,
33261 vector unsigned char);
33263 void vec_st (vector float, int, vector float *);
33264 void vec_st (vector float, int, float *);
33265 void vec_st (vector signed int, int, vector signed int *);
33266 void vec_st (vector signed int, int, int *);
33267 void vec_st (vector unsigned int, int, vector unsigned int *);
33268 void vec_st (vector unsigned int, int, unsigned int *);
33269 void vec_st (vector bool int, int, vector bool int *);
33270 void vec_st (vector bool int, int, unsigned int *);
33271 void vec_st (vector bool int, int, int *);
33272 void vec_st (vector signed short, int, vector signed short *);
33273 void vec_st (vector signed short, int, short *);
33274 void vec_st (vector unsigned short, int, vector unsigned short *);
33275 void vec_st (vector unsigned short, int, unsigned short *);
33276 void vec_st (vector bool short, int, vector bool short *);
33277 void vec_st (vector bool short, int, unsigned short *);
33278 void vec_st (vector pixel, int, vector pixel *);
33279 void vec_st (vector pixel, int, unsigned short *);
33280 void vec_st (vector pixel, int, short *);
33281 void vec_st (vector bool short, int, short *);
33282 void vec_st (vector signed char, int, vector signed char *);
33283 void vec_st (vector signed char, int, signed char *);
33284 void vec_st (vector unsigned char, int, vector unsigned char *);
33285 void vec_st (vector unsigned char, int, unsigned char *);
33286 void vec_st (vector bool char, int, vector bool char *);
33287 void vec_st (vector bool char, int, unsigned char *);
33288 void vec_st (vector bool char, int, signed char *);
33290 void vec_ste (vector signed char, int, signed char *);
33291 void vec_ste (vector unsigned char, int, unsigned char *);
33292 void vec_ste (vector bool char, int, signed char *);
33293 void vec_ste (vector bool char, int, unsigned char *);
33294 void vec_ste (vector signed short, int, short *);
33295 void vec_ste (vector unsigned short, int, unsigned short *);
33296 void vec_ste (vector bool short, int, short *);
33297 void vec_ste (vector bool short, int, unsigned short *);
33298 void vec_ste (vector pixel, int, short *);
33299 void vec_ste (vector pixel, int, unsigned short *);
33300 void vec_ste (vector float, int, float *);
33301 void vec_ste (vector signed int, int, int *);
33302 void vec_ste (vector unsigned int, int, unsigned int *);
33303 void vec_ste (vector bool int, int, int *);
33304 void vec_ste (vector bool int, int, unsigned int *);
33306 void vec_stvewx (vector float, int, float *);
33307 void vec_stvewx (vector signed int, int, int *);
33308 void vec_stvewx (vector unsigned int, int, unsigned int *);
33309 void vec_stvewx (vector bool int, int, int *);
33310 void vec_stvewx (vector bool int, int, unsigned int *);
33312 void vec_stvehx (vector signed short, int, short *);
33313 void vec_stvehx (vector unsigned short, int, unsigned short *);
33314 void vec_stvehx (vector bool short, int, short *);
33315 void vec_stvehx (vector bool short, int, unsigned short *);
33316 void vec_stvehx (vector pixel, int, short *);
33317 void vec_stvehx (vector pixel, int, unsigned short *);
33319 void vec_stvebx (vector signed char, int, signed char *);
33320 void vec_stvebx (vector unsigned char, int, unsigned char *);
33321 void vec_stvebx (vector bool char, int, signed char *);
33322 void vec_stvebx (vector bool char, int, unsigned char *);
33324 void vec_stl (vector float, int, vector float *);
33325 void vec_stl (vector float, int, float *);
33326 void vec_stl (vector signed int, int, vector signed int *);
33327 void vec_stl (vector signed int, int, int *);
33328 void vec_stl (vector unsigned int, int, vector unsigned int *);
33329 void vec_stl (vector unsigned int, int, unsigned int *);
33330 void vec_stl (vector bool int, int, vector bool int *);
33331 void vec_stl (vector bool int, int, unsigned int *);
33332 void vec_stl (vector bool int, int, int *);
33333 void vec_stl (vector signed short, int, vector signed short *);
33334 void vec_stl (vector signed short, int, short *);
33335 void vec_stl (vector unsigned short, int, vector unsigned short *);
33336 void vec_stl (vector unsigned short, int, unsigned short *);
33337 void vec_stl (vector bool short, int, vector bool short *);
33338 void vec_stl (vector bool short, int, unsigned short *);
33339 void vec_stl (vector bool short, int, short *);
33340 void vec_stl (vector pixel, int, vector pixel *);
33341 void vec_stl (vector pixel, int, unsigned short *);
33342 void vec_stl (vector pixel, int, short *);
33343 void vec_stl (vector signed char, int, vector signed char *);
33344 void vec_stl (vector signed char, int, signed char *);
33345 void vec_stl (vector unsigned char, int, vector unsigned char *);
33346 void vec_stl (vector unsigned char, int, unsigned char *);
33347 void vec_stl (vector bool char, int, vector bool char *);
33348 void vec_stl (vector bool char, int, unsigned char *);
33349 void vec_stl (vector bool char, int, signed char *);
33351 vector signed char vec_sub (vector bool char, vector signed char);
33352 vector signed char vec_sub (vector signed char, vector bool char);
33353 vector signed char vec_sub (vector signed char, vector signed char);
33354 vector unsigned char vec_sub (vector bool char, vector unsigned char);
33355 vector unsigned char vec_sub (vector unsigned char, vector bool char);
33356 vector unsigned char vec_sub (vector unsigned char,
33357 vector unsigned char);
33358 vector signed short vec_sub (vector bool short, vector signed short);
33359 vector signed short vec_sub (vector signed short, vector bool short);
33360 vector signed short vec_sub (vector signed short, vector signed short);
33361 vector unsigned short vec_sub (vector bool short,
33362 vector unsigned short);
33363 vector unsigned short vec_sub (vector unsigned short,
33364 vector bool short);
33365 vector unsigned short vec_sub (vector unsigned short,
33366 vector unsigned short);
33367 vector signed int vec_sub (vector bool int, vector signed int);
33368 vector signed int vec_sub (vector signed int, vector bool int);
33369 vector signed int vec_sub (vector signed int, vector signed int);
33370 vector unsigned int vec_sub (vector bool int, vector unsigned int);
33371 vector unsigned int vec_sub (vector unsigned int, vector bool int);
33372 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
33373 vector float vec_sub (vector float, vector float);
33375 vector float vec_vsubfp (vector float, vector float);
33377 vector signed int vec_vsubuwm (vector bool int, vector signed int);
33378 vector signed int vec_vsubuwm (vector signed int, vector bool int);
33379 vector signed int vec_vsubuwm (vector signed int, vector signed int);
33380 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
33381 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
33382 vector unsigned int vec_vsubuwm (vector unsigned int,
33383 vector unsigned int);
33385 vector signed short vec_vsubuhm (vector bool short,
33386 vector signed short);
33387 vector signed short vec_vsubuhm (vector signed short,
33388 vector bool short);
33389 vector signed short vec_vsubuhm (vector signed short,
33390 vector signed short);
33391 vector unsigned short vec_vsubuhm (vector bool short,
33392 vector unsigned short);
33393 vector unsigned short vec_vsubuhm (vector unsigned short,
33394 vector bool short);
33395 vector unsigned short vec_vsubuhm (vector unsigned short,
33396 vector unsigned short);
33398 vector signed char vec_vsububm (vector bool char, vector signed char);
33399 vector signed char vec_vsububm (vector signed char, vector bool char);
33400 vector signed char vec_vsububm (vector signed char, vector signed char);
33401 vector unsigned char vec_vsububm (vector bool char,
33402 vector unsigned char);
33403 vector unsigned char vec_vsububm (vector unsigned char,
33405 vector unsigned char vec_vsububm (vector unsigned char,
33406 vector unsigned char);
33408 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
33410 vector unsigned char vec_subs (vector bool char, vector unsigned char);
33411 vector unsigned char vec_subs (vector unsigned char, vector bool char);
33412 vector unsigned char vec_subs (vector unsigned char,
33413 vector unsigned char);
33414 vector signed char vec_subs (vector bool char, vector signed char);
33415 vector signed char vec_subs (vector signed char, vector bool char);
33416 vector signed char vec_subs (vector signed char, vector signed char);
33417 vector unsigned short vec_subs (vector bool short,
33418 vector unsigned short);
33419 vector unsigned short vec_subs (vector unsigned short,
33420 vector bool short);
33421 vector unsigned short vec_subs (vector unsigned short,
33422 vector unsigned short);
33423 vector signed short vec_subs (vector bool short, vector signed short);
33424 vector signed short vec_subs (vector signed short, vector bool short);
33425 vector signed short vec_subs (vector signed short, vector signed short);
33426 vector unsigned int vec_subs (vector bool int, vector unsigned int);
33427 vector unsigned int vec_subs (vector unsigned int, vector bool int);
33428 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
33429 vector signed int vec_subs (vector bool int, vector signed int);
33430 vector signed int vec_subs (vector signed int, vector bool int);
33431 vector signed int vec_subs (vector signed int, vector signed int);
33433 vector signed int vec_vsubsws (vector bool int, vector signed int);
33434 vector signed int vec_vsubsws (vector signed int, vector bool int);
33435 vector signed int vec_vsubsws (vector signed int, vector signed int);
33437 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
33438 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
33439 vector unsigned int vec_vsubuws (vector unsigned int,
33440 vector unsigned int);
33442 vector signed short vec_vsubshs (vector bool short,
33443 vector signed short);
33444 vector signed short vec_vsubshs (vector signed short,
33445 vector bool short);
33446 vector signed short vec_vsubshs (vector signed short,
33447 vector signed short);
33449 vector unsigned short vec_vsubuhs (vector bool short,
33450 vector unsigned short);
33451 vector unsigned short vec_vsubuhs (vector unsigned short,
33452 vector bool short);
33453 vector unsigned short vec_vsubuhs (vector unsigned short,
33454 vector unsigned short);
33456 vector signed char vec_vsubsbs (vector bool char, vector signed char);
33457 vector signed char vec_vsubsbs (vector signed char, vector bool char);
33458 vector signed char vec_vsubsbs (vector signed char, vector signed char);
33460 vector unsigned char vec_vsububs (vector bool char,
33461 vector unsigned char);
33462 vector unsigned char vec_vsububs (vector unsigned char,
33464 vector unsigned char vec_vsububs (vector unsigned char,
33465 vector unsigned char);
33467 vector unsigned int vec_sum4s (vector unsigned char,
33468 vector unsigned int);
33469 vector signed int vec_sum4s (vector signed char, vector signed int);
33470 vector signed int vec_sum4s (vector signed short, vector signed int);
33472 vector signed int vec_vsum4shs (vector signed short, vector signed int);
33474 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
33476 vector unsigned int vec_vsum4ubs (vector unsigned char,
33477 vector unsigned int);
33479 vector signed int vec_sum2s (vector signed int, vector signed int);
33481 vector signed int vec_sums (vector signed int, vector signed int);
33483 vector float vec_trunc (vector float);
33485 vector signed short vec_unpackh (vector signed char);
33486 vector bool short vec_unpackh (vector bool char);
33487 vector signed int vec_unpackh (vector signed short);
33488 vector bool int vec_unpackh (vector bool short);
33489 vector unsigned int vec_unpackh (vector pixel);
33491 vector bool int vec_vupkhsh (vector bool short);
33492 vector signed int vec_vupkhsh (vector signed short);
33494 vector unsigned int vec_vupkhpx (vector pixel);
33496 vector bool short vec_vupkhsb (vector bool char);
33497 vector signed short vec_vupkhsb (vector signed char);
33499 vector signed short vec_unpackl (vector signed char);
33500 vector bool short vec_unpackl (vector bool char);
33501 vector unsigned int vec_unpackl (vector pixel);
33502 vector signed int vec_unpackl (vector signed short);
33503 vector bool int vec_unpackl (vector bool short);
33505 vector unsigned int vec_vupklpx (vector pixel);
33507 vector bool int vec_vupklsh (vector bool short);
33508 vector signed int vec_vupklsh (vector signed short);
33510 vector bool short vec_vupklsb (vector bool char);
33511 vector signed short vec_vupklsb (vector signed char);
33513 vector float vec_xor (vector float, vector float);
33514 vector float vec_xor (vector float, vector bool int);
33515 vector float vec_xor (vector bool int, vector float);
33516 vector bool int vec_xor (vector bool int, vector bool int);
33517 vector signed int vec_xor (vector bool int, vector signed int);
33518 vector signed int vec_xor (vector signed int, vector bool int);
33519 vector signed int vec_xor (vector signed int, vector signed int);
33520 vector unsigned int vec_xor (vector bool int, vector unsigned int);
33521 vector unsigned int vec_xor (vector unsigned int, vector bool int);
33522 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
33523 vector bool short vec_xor (vector bool short, vector bool short);
33524 vector signed short vec_xor (vector bool short, vector signed short);
33525 vector signed short vec_xor (vector signed short, vector bool short);
33526 vector signed short vec_xor (vector signed short, vector signed short);
33527 vector unsigned short vec_xor (vector bool short,
33528 vector unsigned short);
33529 vector unsigned short vec_xor (vector unsigned short,
33530 vector bool short);
33531 vector unsigned short vec_xor (vector unsigned short,
33532 vector unsigned short);
33533 vector signed char vec_xor (vector bool char, vector signed char);
33534 vector bool char vec_xor (vector bool char, vector bool char);
33535 vector signed char vec_xor (vector signed char, vector bool char);
33536 vector signed char vec_xor (vector signed char, vector signed char);
33537 vector unsigned char vec_xor (vector bool char, vector unsigned char);
33538 vector unsigned char vec_xor (vector unsigned char, vector bool char);
33539 vector unsigned char vec_xor (vector unsigned char,
33540 vector unsigned char);
33542 int vec_all_eq (vector signed char, vector bool char);
33543 int vec_all_eq (vector signed char, vector signed char);
33544 int vec_all_eq (vector unsigned char, vector bool char);
33545 int vec_all_eq (vector unsigned char, vector unsigned char);
33546 int vec_all_eq (vector bool char, vector bool char);
33547 int vec_all_eq (vector bool char, vector unsigned char);
33548 int vec_all_eq (vector bool char, vector signed char);
33549 int vec_all_eq (vector signed short, vector bool short);
33550 int vec_all_eq (vector signed short, vector signed short);
33551 int vec_all_eq (vector unsigned short, vector bool short);
33552 int vec_all_eq (vector unsigned short, vector unsigned short);
33553 int vec_all_eq (vector bool short, vector bool short);
33554 int vec_all_eq (vector bool short, vector unsigned short);
33555 int vec_all_eq (vector bool short, vector signed short);
33556 int vec_all_eq (vector pixel, vector pixel);
33557 int vec_all_eq (vector signed int, vector bool int);
33558 int vec_all_eq (vector signed int, vector signed int);
33559 int vec_all_eq (vector unsigned int, vector bool int);
33560 int vec_all_eq (vector unsigned int, vector unsigned int);
33561 int vec_all_eq (vector bool int, vector bool int);
33562 int vec_all_eq (vector bool int, vector unsigned int);
33563 int vec_all_eq (vector bool int, vector signed int);
33564 int vec_all_eq (vector float, vector float);
33566 int vec_all_ge (vector bool char, vector unsigned char);
33567 int vec_all_ge (vector unsigned char, vector bool char);
33568 int vec_all_ge (vector unsigned char, vector unsigned char);
33569 int vec_all_ge (vector bool char, vector signed char);
33570 int vec_all_ge (vector signed char, vector bool char);
33571 int vec_all_ge (vector signed char, vector signed char);
33572 int vec_all_ge (vector bool short, vector unsigned short);
33573 int vec_all_ge (vector unsigned short, vector bool short);
33574 int vec_all_ge (vector unsigned short, vector unsigned short);
33575 int vec_all_ge (vector signed short, vector signed short);
33576 int vec_all_ge (vector bool short, vector signed short);
33577 int vec_all_ge (vector signed short, vector bool short);
33578 int vec_all_ge (vector bool int, vector unsigned int);
33579 int vec_all_ge (vector unsigned int, vector bool int);
33580 int vec_all_ge (vector unsigned int, vector unsigned int);
33581 int vec_all_ge (vector bool int, vector signed int);
33582 int vec_all_ge (vector signed int, vector bool int);
33583 int vec_all_ge (vector signed int, vector signed int);
33584 int vec_all_ge (vector float, vector float);
33586 int vec_all_gt (vector bool char, vector unsigned char);
33587 int vec_all_gt (vector unsigned char, vector bool char);
33588 int vec_all_gt (vector unsigned char, vector unsigned char);
33589 int vec_all_gt (vector bool char, vector signed char);
33590 int vec_all_gt (vector signed char, vector bool char);
33591 int vec_all_gt (vector signed char, vector signed char);
33592 int vec_all_gt (vector bool short, vector unsigned short);
33593 int vec_all_gt (vector unsigned short, vector bool short);
33594 int vec_all_gt (vector unsigned short, vector unsigned short);
33595 int vec_all_gt (vector bool short, vector signed short);
33596 int vec_all_gt (vector signed short, vector bool short);
33597 int vec_all_gt (vector signed short, vector signed short);
33598 int vec_all_gt (vector bool int, vector unsigned int);
33599 int vec_all_gt (vector unsigned int, vector bool int);
33600 int vec_all_gt (vector unsigned int, vector unsigned int);
33601 int vec_all_gt (vector bool int, vector signed int);
33602 int vec_all_gt (vector signed int, vector bool int);
33603 int vec_all_gt (vector signed int, vector signed int);
33604 int vec_all_gt (vector float, vector float);
33606 int vec_all_in (vector float, vector float);
33608 int vec_all_le (vector bool char, vector unsigned char);
33609 int vec_all_le (vector unsigned char, vector bool char);
33610 int vec_all_le (vector unsigned char, vector unsigned char);
33611 int vec_all_le (vector bool char, vector signed char);
33612 int vec_all_le (vector signed char, vector bool char);
33613 int vec_all_le (vector signed char, vector signed char);
33614 int vec_all_le (vector bool short, vector unsigned short);
33615 int vec_all_le (vector unsigned short, vector bool short);
33616 int vec_all_le (vector unsigned short, vector unsigned short);
33617 int vec_all_le (vector bool short, vector signed short);
33618 int vec_all_le (vector signed short, vector bool short);
33619 int vec_all_le (vector signed short, vector signed short);
33620 int vec_all_le (vector bool int, vector unsigned int);
33621 int vec_all_le (vector unsigned int, vector bool int);
33622 int vec_all_le (vector unsigned int, vector unsigned int);
33623 int vec_all_le (vector bool int, vector signed int);
33624 int vec_all_le (vector signed int, vector bool int);
33625 int vec_all_le (vector signed int, vector signed int);
33626 int vec_all_le (vector float, vector float);
33628 int vec_all_lt (vector bool char, vector unsigned char);
33629 int vec_all_lt (vector unsigned char, vector bool char);
33630 int vec_all_lt (vector unsigned char, vector unsigned char);
33631 int vec_all_lt (vector bool char, vector signed char);
33632 int vec_all_lt (vector signed char, vector bool char);
33633 int vec_all_lt (vector signed char, vector signed char);
33634 int vec_all_lt (vector bool short, vector unsigned short);
33635 int vec_all_lt (vector unsigned short, vector bool short);
33636 int vec_all_lt (vector unsigned short, vector unsigned short);
33637 int vec_all_lt (vector bool short, vector signed short);
33638 int vec_all_lt (vector signed short, vector bool short);
33639 int vec_all_lt (vector signed short, vector signed short);
33640 int vec_all_lt (vector bool int, vector unsigned int);
33641 int vec_all_lt (vector unsigned int, vector bool int);
33642 int vec_all_lt (vector unsigned int, vector unsigned int);
33643 int vec_all_lt (vector bool int, vector signed int);
33644 int vec_all_lt (vector signed int, vector bool int);
33645 int vec_all_lt (vector signed int, vector signed int);
33646 int vec_all_lt (vector float, vector float);
33648 int vec_all_nan (vector float);
33650 int vec_all_ne (vector signed char, vector bool char);
33651 int vec_all_ne (vector signed char, vector signed char);
33652 int vec_all_ne (vector unsigned char, vector bool char);
33653 int vec_all_ne (vector unsigned char, vector unsigned char);
33654 int vec_all_ne (vector bool char, vector bool char);
33655 int vec_all_ne (vector bool char, vector unsigned char);
33656 int vec_all_ne (vector bool char, vector signed char);
33657 int vec_all_ne (vector signed short, vector bool short);
33658 int vec_all_ne (vector signed short, vector signed short);
33659 int vec_all_ne (vector unsigned short, vector bool short);
33660 int vec_all_ne (vector unsigned short, vector unsigned short);
33661 int vec_all_ne (vector bool short, vector bool short);
33662 int vec_all_ne (vector bool short, vector unsigned short);
33663 int vec_all_ne (vector bool short, vector signed short);
33664 int vec_all_ne (vector pixel, vector pixel);
33665 int vec_all_ne (vector signed int, vector bool int);
33666 int vec_all_ne (vector signed int, vector signed int);
33667 int vec_all_ne (vector unsigned int, vector bool int);
33668 int vec_all_ne (vector unsigned int, vector unsigned int);
33669 int vec_all_ne (vector bool int, vector bool int);
33670 int vec_all_ne (vector bool int, vector unsigned int);
33671 int vec_all_ne (vector bool int, vector signed int);
33672 int vec_all_ne (vector float, vector float);
33674 int vec_all_nge (vector float, vector float);
33676 int vec_all_ngt (vector float, vector float);
33678 int vec_all_nle (vector float, vector float);
33680 int vec_all_nlt (vector float, vector float);
33682 int vec_all_numeric (vector float);
33684 int vec_any_eq (vector signed char, vector bool char);
33685 int vec_any_eq (vector signed char, vector signed char);
33686 int vec_any_eq (vector unsigned char, vector bool char);
33687 int vec_any_eq (vector unsigned char, vector unsigned char);
33688 int vec_any_eq (vector bool char, vector bool char);
33689 int vec_any_eq (vector bool char, vector unsigned char);
33690 int vec_any_eq (vector bool char, vector signed char);
33691 int vec_any_eq (vector signed short, vector bool short);
33692 int vec_any_eq (vector signed short, vector signed short);
33693 int vec_any_eq (vector unsigned short, vector bool short);
33694 int vec_any_eq (vector unsigned short, vector unsigned short);
33695 int vec_any_eq (vector bool short, vector bool short);
33696 int vec_any_eq (vector bool short, vector unsigned short);
33697 int vec_any_eq (vector bool short, vector signed short);
33698 int vec_any_eq (vector pixel, vector pixel);
33699 int vec_any_eq (vector signed int, vector bool int);
33700 int vec_any_eq (vector signed int, vector signed int);
33701 int vec_any_eq (vector unsigned int, vector bool int);
33702 int vec_any_eq (vector unsigned int, vector unsigned int);
33703 int vec_any_eq (vector bool int, vector bool int);
33704 int vec_any_eq (vector bool int, vector unsigned int);
33705 int vec_any_eq (vector bool int, vector signed int);
33706 int vec_any_eq (vector float, vector float);
33708 int vec_any_ge (vector signed char, vector bool char);
33709 int vec_any_ge (vector unsigned char, vector bool char);
33710 int vec_any_ge (vector unsigned char, vector unsigned char);
33711 int vec_any_ge (vector signed char, vector signed char);
33712 int vec_any_ge (vector bool char, vector unsigned char);
33713 int vec_any_ge (vector bool char, vector signed char);
33714 int vec_any_ge (vector unsigned short, vector bool short);
33715 int vec_any_ge (vector unsigned short, vector unsigned short);
33716 int vec_any_ge (vector signed short, vector signed short);
33717 int vec_any_ge (vector signed short, vector bool short);
33718 int vec_any_ge (vector bool short, vector unsigned short);
33719 int vec_any_ge (vector bool short, vector signed short);
33720 int vec_any_ge (vector signed int, vector bool int);
33721 int vec_any_ge (vector unsigned int, vector bool int);
33722 int vec_any_ge (vector unsigned int, vector unsigned int);
33723 int vec_any_ge (vector signed int, vector signed int);
33724 int vec_any_ge (vector bool int, vector unsigned int);
33725 int vec_any_ge (vector bool int, vector signed int);
33726 int vec_any_ge (vector float, vector float);
33728 int vec_any_gt (vector bool char, vector unsigned char);
33729 int vec_any_gt (vector unsigned char, vector bool char);
33730 int vec_any_gt (vector unsigned char, vector unsigned char);
33731 int vec_any_gt (vector bool char, vector signed char);
33732 int vec_any_gt (vector signed char, vector bool char);
33733 int vec_any_gt (vector signed char, vector signed char);
33734 int vec_any_gt (vector bool short, vector unsigned short);
33735 int vec_any_gt (vector unsigned short, vector bool short);
33736 int vec_any_gt (vector unsigned short, vector unsigned short);
33737 int vec_any_gt (vector bool short, vector signed short);
33738 int vec_any_gt (vector signed short, vector bool short);
33739 int vec_any_gt (vector signed short, vector signed short);
33740 int vec_any_gt (vector bool int, vector unsigned int);
33741 int vec_any_gt (vector unsigned int, vector bool int);
33742 int vec_any_gt (vector unsigned int, vector unsigned int);
33743 int vec_any_gt (vector bool int, vector signed int);
33744 int vec_any_gt (vector signed int, vector bool int);
33745 int vec_any_gt (vector signed int, vector signed int);
33746 int vec_any_gt (vector float, vector float);
33748 int vec_any_le (vector bool char, vector unsigned char);
33749 int vec_any_le (vector unsigned char, vector bool char);
33750 int vec_any_le (vector unsigned char, vector unsigned char);
33751 int vec_any_le (vector bool char, vector signed char);
33752 int vec_any_le (vector signed char, vector bool char);
33753 int vec_any_le (vector signed char, vector signed char);
33754 int vec_any_le (vector bool short, vector unsigned short);
33755 int vec_any_le (vector unsigned short, vector bool short);
33756 int vec_any_le (vector unsigned short, vector unsigned short);
33757 int vec_any_le (vector bool short, vector signed short);
33758 int vec_any_le (vector signed short, vector bool short);
33759 int vec_any_le (vector signed short, vector signed short);
33760 int vec_any_le (vector bool int, vector unsigned int);
33761 int vec_any_le (vector unsigned int, vector bool int);
33762 int vec_any_le (vector unsigned int, vector unsigned int);
33763 int vec_any_le (vector bool int, vector signed int);
33764 int vec_any_le (vector signed int, vector bool int);
33765 int vec_any_le (vector signed int, vector signed int);
33766 int vec_any_le (vector float, vector float);
33768 int vec_any_lt (vector bool char, vector unsigned char);
33769 int vec_any_lt (vector unsigned char, vector bool char);
33770 int vec_any_lt (vector unsigned char, vector unsigned char);
33771 int vec_any_lt (vector bool char, vector signed char);
33772 int vec_any_lt (vector signed char, vector bool char);
33773 int vec_any_lt (vector signed char, vector signed char);
33774 int vec_any_lt (vector bool short, vector unsigned short);
33775 int vec_any_lt (vector unsigned short, vector bool short);
33776 int vec_any_lt (vector unsigned short, vector unsigned short);
33777 int vec_any_lt (vector bool short, vector signed short);
33778 int vec_any_lt (vector signed short, vector bool short);
33779 int vec_any_lt (vector signed short, vector signed short);
33780 int vec_any_lt (vector bool int, vector unsigned int);
33781 int vec_any_lt (vector unsigned int, vector bool int);
33782 int vec_any_lt (vector unsigned int, vector unsigned int);
33783 int vec_any_lt (vector bool int, vector signed int);
33784 int vec_any_lt (vector signed int, vector bool int);
33785 int vec_any_lt (vector signed int, vector signed int);
33786 int vec_any_lt (vector float, vector float);
33788 int vec_any_nan (vector float);
33790 int vec_any_ne (vector signed char, vector bool char);
33791 int vec_any_ne (vector signed char, vector signed char);
33792 int vec_any_ne (vector unsigned char, vector bool char);
33793 int vec_any_ne (vector unsigned char, vector unsigned char);
33794 int vec_any_ne (vector bool char, vector bool char);
33795 int vec_any_ne (vector bool char, vector unsigned char);
33796 int vec_any_ne (vector bool char, vector signed char);
33797 int vec_any_ne (vector signed short, vector bool short);
33798 int vec_any_ne (vector signed short, vector signed short);
33799 int vec_any_ne (vector unsigned short, vector bool short);
33800 int vec_any_ne (vector unsigned short, vector unsigned short);
33801 int vec_any_ne (vector bool short, vector bool short);
33802 int vec_any_ne (vector bool short, vector unsigned short);
33803 int vec_any_ne (vector bool short, vector signed short);
33804 int vec_any_ne (vector pixel, vector pixel);
33805 int vec_any_ne (vector signed int, vector bool int);
33806 int vec_any_ne (vector signed int, vector signed int);
33807 int vec_any_ne (vector unsigned int, vector bool int);
33808 int vec_any_ne (vector unsigned int, vector unsigned int);
33809 int vec_any_ne (vector bool int, vector bool int);
33810 int vec_any_ne (vector bool int, vector unsigned int);
33811 int vec_any_ne (vector bool int, vector signed int);
33812 int vec_any_ne (vector float, vector float);
33814 int vec_any_nge (vector float, vector float);
33816 int vec_any_ngt (vector float, vector float);
33818 int vec_any_nle (vector float, vector float);
33820 int vec_any_nlt (vector float, vector float);
33822 int vec_any_numeric (vector float);
33824 int vec_any_out (vector float, vector float);
33827 File: gcc.info, Node: SPARC VIS Built-in Functions, Next: SPU Built-in Functions, Prev: PowerPC AltiVec Built-in Functions, Up: Target Builtins
33829 5.50.13 SPARC VIS Built-in Functions
33830 ------------------------------------
33832 GCC supports SIMD operations on the SPARC using both the generic vector
33833 extensions (*note Vector Extensions::) as well as built-in functions for
33834 the SPARC Visual Instruction Set (VIS). When you use the `-mvis'
33835 switch, the VIS extension is exposed as the following built-in
33838 typedef int v2si __attribute__ ((vector_size (8)));
33839 typedef short v4hi __attribute__ ((vector_size (8)));
33840 typedef short v2hi __attribute__ ((vector_size (4)));
33841 typedef char v8qi __attribute__ ((vector_size (8)));
33842 typedef char v4qi __attribute__ ((vector_size (4)));
33844 void * __builtin_vis_alignaddr (void *, long);
33845 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
33846 v2si __builtin_vis_faligndatav2si (v2si, v2si);
33847 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
33848 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
33850 v4hi __builtin_vis_fexpand (v4qi);
33852 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
33853 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
33854 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
33855 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
33856 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
33857 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
33858 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
33860 v4qi __builtin_vis_fpack16 (v4hi);
33861 v8qi __builtin_vis_fpack32 (v2si, v2si);
33862 v2hi __builtin_vis_fpackfix (v2si);
33863 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
33865 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
33868 File: gcc.info, Node: SPU Built-in Functions, Prev: SPARC VIS Built-in Functions, Up: Target Builtins
33870 5.50.14 SPU Built-in Functions
33871 ------------------------------
33873 GCC provides extensions for the SPU processor as described in the
33874 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
33875 found at `http://cell.scei.co.jp/' or
33876 `http://www.ibm.com/developerworks/power/cell/'. GCC's implementation
33877 differs in several ways.
33879 * The optional extension of specifying vector constants in
33880 parentheses is not supported.
33882 * A vector initializer requires no cast if the vector constant is of
33883 the same type as the variable it is initializing.
33885 * If `signed' or `unsigned' is omitted, the signedness of the vector
33886 type is the default signedness of the base type. The default
33887 varies depending on the operating system, so a portable program
33888 should always specify the signedness.
33890 * By default, the keyword `__vector' is added. The macro `vector' is
33891 defined in `<spu_intrinsics.h>' and can be undefined.
33893 * GCC allows using a `typedef' name as the type specifier for a
33896 * For C, overloaded functions are implemented with macros so the
33897 following does not work:
33899 spu_add ((vector signed int){1, 2, 3, 4}, foo);
33901 Since `spu_add' is a macro, the vector constant in the example is
33902 treated as four separate arguments. Wrap the entire argument in
33903 parentheses for this to work.
33905 * The extended version of `__builtin_expect' is not supported.
33908 _Note:_ Only the interface described in the aforementioned
33909 specification is supported. Internally, GCC uses built-in functions to
33910 implement the required functionality, but these are not supported and
33911 are subject to change without notice.
33914 File: gcc.info, Node: Target Format Checks, Next: Pragmas, Prev: Target Builtins, Up: C Extensions
33916 5.51 Format Checks Specific to Particular Target Machines
33917 =========================================================
33919 For some target machines, GCC supports additional options to the format
33920 attribute (*note Declaring Attributes of Functions: Function
33925 * Solaris Format Checks::
33928 File: gcc.info, Node: Solaris Format Checks, Up: Target Format Checks
33930 5.51.1 Solaris Format Checks
33931 ----------------------------
33933 Solaris targets support the `cmn_err' (or `__cmn_err__') format check.
33934 `cmn_err' accepts a subset of the standard `printf' conversions, and
33935 the two-argument `%b' conversion for displaying bit-fields. See the
33936 Solaris man page for `cmn_err' for more information.
33939 File: gcc.info, Node: Pragmas, Next: Unnamed Fields, Prev: Target Format Checks, Up: C Extensions
33941 5.52 Pragmas Accepted by GCC
33942 ============================
33944 GCC supports several types of pragmas, primarily in order to compile
33945 code originally written for other compilers. Note that in general we
33946 do not recommend the use of pragmas; *Note Function Attributes::, for
33947 further explanation.
33953 * RS/6000 and PowerPC Pragmas::
33955 * Solaris Pragmas::
33956 * Symbol-Renaming Pragmas::
33957 * Structure-Packing Pragmas::
33959 * Diagnostic Pragmas::
33960 * Visibility Pragmas::
33961 * Push/Pop Macro Pragmas::
33962 * Function Specific Option Pragmas::
33965 File: gcc.info, Node: ARM Pragmas, Next: M32C Pragmas, Up: Pragmas
33970 The ARM target defines pragmas for controlling the default addition of
33971 `long_call' and `short_call' attributes to functions. *Note Function
33972 Attributes::, for information about the effects of these attributes.
33975 Set all subsequent functions to have the `long_call' attribute.
33978 Set all subsequent functions to have the `short_call' attribute.
33981 Do not affect the `long_call' or `short_call' attributes of
33982 subsequent functions.
33985 File: gcc.info, Node: M32C Pragmas, Next: RS/6000 and PowerPC Pragmas, Prev: ARM Pragmas, Up: Pragmas
33987 5.52.2 M32C Pragmas
33988 -------------------
33991 Overrides the command line option `-memregs=' for the current
33992 file. Use with care! This pragma must be before any function in
33993 the file, and mixing different memregs values in different objects
33994 may make them incompatible. This pragma is useful when a
33995 performance-critical function uses a memreg for temporary values,
33996 as it may allow you to reduce the number of memregs used.
34000 File: gcc.info, Node: RS/6000 and PowerPC Pragmas, Next: Darwin Pragmas, Prev: M32C Pragmas, Up: Pragmas
34002 5.52.3 RS/6000 and PowerPC Pragmas
34003 ----------------------------------
34005 The RS/6000 and PowerPC targets define one pragma for controlling
34006 whether or not the `longcall' attribute is added to function
34007 declarations by default. This pragma overrides the `-mlongcall'
34008 option, but not the `longcall' and `shortcall' attributes. *Note
34009 RS/6000 and PowerPC Options::, for more information about when long
34010 calls are and are not necessary.
34013 Apply the `longcall' attribute to all subsequent function
34017 Do not apply the `longcall' attribute to subsequent function
34021 File: gcc.info, Node: Darwin Pragmas, Next: Solaris Pragmas, Prev: RS/6000 and PowerPC Pragmas, Up: Pragmas
34023 5.52.4 Darwin Pragmas
34024 ---------------------
34026 The following pragmas are available for all architectures running the
34027 Darwin operating system. These are useful for compatibility with other
34031 This pragma is accepted, but has no effect.
34033 `options align=ALIGNMENT'
34034 This pragma sets the alignment of fields in structures. The
34035 values of ALIGNMENT may be `mac68k', to emulate m68k alignment, or
34036 `power', to emulate PowerPC alignment. Uses of this pragma nest
34037 properly; to restore the previous setting, use `reset' for the
34040 `segment TOKENS...'
34041 This pragma is accepted, but has no effect.
34043 `unused (VAR [, VAR]...)'
34044 This pragma declares variables to be possibly unused. GCC will not
34045 produce warnings for the listed variables. The effect is similar
34046 to that of the `unused' attribute, except that this pragma may
34047 appear anywhere within the variables' scopes.
34050 File: gcc.info, Node: Solaris Pragmas, Next: Symbol-Renaming Pragmas, Prev: Darwin Pragmas, Up: Pragmas
34052 5.52.5 Solaris Pragmas
34053 ----------------------
34055 The Solaris target supports `#pragma redefine_extname' (*note
34056 Symbol-Renaming Pragmas::). It also supports additional `#pragma'
34057 directives for compatibility with the system compiler.
34059 `align ALIGNMENT (VARIABLE [, VARIABLE]...)'
34060 Increase the minimum alignment of each VARIABLE to ALIGNMENT.
34061 This is the same as GCC's `aligned' attribute *note Variable
34062 Attributes::). Macro expansion occurs on the arguments to this
34063 pragma when compiling C and Objective-C. It does not currently
34064 occur when compiling C++, but this is a bug which may be fixed in
34067 `fini (FUNCTION [, FUNCTION]...)'
34068 This pragma causes each listed FUNCTION to be called after main,
34069 or during shared module unloading, by adding a call to the `.fini'
34072 `init (FUNCTION [, FUNCTION]...)'
34073 This pragma causes each listed FUNCTION to be called during
34074 initialization (before `main') or during shared module loading, by
34075 adding a call to the `.init' section.
34079 File: gcc.info, Node: Symbol-Renaming Pragmas, Next: Structure-Packing Pragmas, Prev: Solaris Pragmas, Up: Pragmas
34081 5.52.6 Symbol-Renaming Pragmas
34082 ------------------------------
34084 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
34085 supports two `#pragma' directives which change the name used in
34086 assembly for a given declaration. These pragmas are only available on
34087 platforms whose system headers need them. To get this effect on all
34088 platforms supported by GCC, use the asm labels extension (*note Asm
34091 `redefine_extname OLDNAME NEWNAME'
34092 This pragma gives the C function OLDNAME the assembly symbol
34093 NEWNAME. The preprocessor macro `__PRAGMA_REDEFINE_EXTNAME' will
34094 be defined if this pragma is available (currently only on Solaris).
34096 `extern_prefix STRING'
34097 This pragma causes all subsequent external function and variable
34098 declarations to have STRING prepended to their assembly symbols.
34099 This effect may be terminated with another `extern_prefix' pragma
34100 whose argument is an empty string. The preprocessor macro
34101 `__PRAGMA_EXTERN_PREFIX' will be defined if this pragma is
34102 available (currently only on Tru64 UNIX).
34104 These pragmas and the asm labels extension interact in a complicated
34105 manner. Here are some corner cases you may want to be aware of.
34107 1. Both pragmas silently apply only to declarations with external
34108 linkage. Asm labels do not have this restriction.
34110 2. In C++, both pragmas silently apply only to declarations with "C"
34111 linkage. Again, asm labels do not have this restriction.
34113 3. If any of the three ways of changing the assembly name of a
34114 declaration is applied to a declaration whose assembly name has
34115 already been determined (either by a previous use of one of these
34116 features, or because the compiler needed the assembly name in
34117 order to generate code), and the new name is different, a warning
34118 issues and the name does not change.
34120 4. The OLDNAME used by `#pragma redefine_extname' is always the
34123 5. If `#pragma extern_prefix' is in effect, and a declaration occurs
34124 with an asm label attached, the prefix is silently ignored for
34127 6. If `#pragma extern_prefix' and `#pragma redefine_extname' apply to
34128 the same declaration, whichever triggered first wins, and a
34129 warning issues if they contradict each other. (We would like to
34130 have `#pragma redefine_extname' always win, for consistency with
34131 asm labels, but if `#pragma extern_prefix' triggers first we have
34132 no way of knowing that that happened.)
34135 File: gcc.info, Node: Structure-Packing Pragmas, Next: Weak Pragmas, Prev: Symbol-Renaming Pragmas, Up: Pragmas
34137 5.52.7 Structure-Packing Pragmas
34138 --------------------------------
34140 For compatibility with Microsoft Windows compilers, GCC supports a set
34141 of `#pragma' directives which change the maximum alignment of members
34142 of structures (other than zero-width bitfields), unions, and classes
34143 subsequently defined. The N value below always is required to be a
34144 small power of two and specifies the new alignment in bytes.
34146 1. `#pragma pack(N)' simply sets the new alignment.
34148 2. `#pragma pack()' sets the alignment to the one that was in effect
34149 when compilation started (see also command line option
34150 `-fpack-struct[=<n>]' *note Code Gen Options::).
34152 3. `#pragma pack(push[,N])' pushes the current alignment setting on
34153 an internal stack and then optionally sets the new alignment.
34155 4. `#pragma pack(pop)' restores the alignment setting to the one
34156 saved at the top of the internal stack (and removes that stack
34157 entry). Note that `#pragma pack([N])' does not influence this
34158 internal stack; thus it is possible to have `#pragma pack(push)'
34159 followed by multiple `#pragma pack(N)' instances and finalized by
34160 a single `#pragma pack(pop)'.
34162 Some targets, e.g. i386 and powerpc, support the `ms_struct' `#pragma'
34163 which lays out a structure as the documented `__attribute__
34165 1. `#pragma ms_struct on' turns on the layout for structures declared.
34167 2. `#pragma ms_struct off' turns off the layout for structures
34170 3. `#pragma ms_struct reset' goes back to the default layout.
34173 File: gcc.info, Node: Weak Pragmas, Next: Diagnostic Pragmas, Prev: Structure-Packing Pragmas, Up: Pragmas
34175 5.52.8 Weak Pragmas
34176 -------------------
34178 For compatibility with SVR4, GCC supports a set of `#pragma' directives
34179 for declaring symbols to be weak, and defining weak aliases.
34181 `#pragma weak SYMBOL'
34182 This pragma declares SYMBOL to be weak, as if the declaration had
34183 the attribute of the same name. The pragma may appear before or
34184 after the declaration of SYMBOL, but must appear before either its
34185 first use or its definition. It is not an error for SYMBOL to
34186 never be defined at all.
34188 `#pragma weak SYMBOL1 = SYMBOL2'
34189 This pragma declares SYMBOL1 to be a weak alias of SYMBOL2. It is
34190 an error if SYMBOL2 is not defined in the current translation unit.
34193 File: gcc.info, Node: Diagnostic Pragmas, Next: Visibility Pragmas, Prev: Weak Pragmas, Up: Pragmas
34195 5.52.9 Diagnostic Pragmas
34196 -------------------------
34198 GCC allows the user to selectively enable or disable certain types of
34199 diagnostics, and change the kind of the diagnostic. For example, a
34200 project's policy might require that all sources compile with `-Werror'
34201 but certain files might have exceptions allowing specific types of
34202 warnings. Or, a project might selectively enable diagnostics and treat
34203 them as errors depending on which preprocessor macros are defined.
34205 `#pragma GCC diagnostic KIND OPTION'
34206 Modifies the disposition of a diagnostic. Note that not all
34207 diagnostics are modifiable; at the moment only warnings (normally
34208 controlled by `-W...') can be controlled, and not all of them.
34209 Use `-fdiagnostics-show-option' to determine which diagnostics are
34210 controllable and which option controls them.
34212 KIND is `error' to treat this diagnostic as an error, `warning' to
34213 treat it like a warning (even if `-Werror' is in effect), or
34214 `ignored' if the diagnostic is to be ignored. OPTION is a double
34215 quoted string which matches the command line option.
34217 #pragma GCC diagnostic warning "-Wformat"
34218 #pragma GCC diagnostic error "-Wformat"
34219 #pragma GCC diagnostic ignored "-Wformat"
34221 Note that these pragmas override any command line options. Also,
34222 while it is syntactically valid to put these pragmas anywhere in
34223 your sources, the only supported location for them is before any
34224 data or functions are defined. Doing otherwise may result in
34225 unpredictable results depending on how the optimizer manages your
34226 sources. If the same option is listed multiple times, the last
34227 one specified is the one that is in effect. This pragma is not
34228 intended to be a general purpose replacement for command line
34229 options, but for implementing strict control over project policies.
34232 GCC also offers a simple mechanism for printing messages during
34235 `#pragma message STRING'
34236 Prints STRING as a compiler message on compilation. The message
34237 is informational only, and is neither a compilation warning nor an
34240 #pragma message "Compiling " __FILE__ "..."
34242 STRING may be parenthesized, and is printed with location
34243 information. For example,
34245 #define DO_PRAGMA(x) _Pragma (#x)
34246 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
34248 TODO(Remember to fix this)
34250 prints `/tmp/file.c:4: note: #pragma message: TODO - Remember to
34255 File: gcc.info, Node: Visibility Pragmas, Next: Push/Pop Macro Pragmas, Prev: Diagnostic Pragmas, Up: Pragmas
34257 5.52.10 Visibility Pragmas
34258 --------------------------
34260 `#pragma GCC visibility push(VISIBILITY)'
34261 `#pragma GCC visibility pop'
34262 This pragma allows the user to set the visibility for multiple
34263 declarations without having to give each a visibility attribute
34264 *Note Function Attributes::, for more information about visibility
34265 and the attribute syntax.
34267 In C++, `#pragma GCC visibility' affects only namespace-scope
34268 declarations. Class members and template specializations are not
34269 affected; if you want to override the visibility for a particular
34270 member or instantiation, you must use an attribute.
34274 File: gcc.info, Node: Push/Pop Macro Pragmas, Next: Function Specific Option Pragmas, Prev: Visibility Pragmas, Up: Pragmas
34276 5.52.11 Push/Pop Macro Pragmas
34277 ------------------------------
34279 For compatibility with Microsoft Windows compilers, GCC supports
34280 `#pragma push_macro("MACRO_NAME")' and `#pragma
34281 pop_macro("MACRO_NAME")'.
34283 `#pragma push_macro("MACRO_NAME")'
34284 This pragma saves the value of the macro named as MACRO_NAME to
34285 the top of the stack for this macro.
34287 `#pragma pop_macro("MACRO_NAME")'
34288 This pragma sets the value of the macro named as MACRO_NAME to the
34289 value on top of the stack for this macro. If the stack for
34290 MACRO_NAME is empty, the value of the macro remains unchanged.
34295 #pragma push_macro("X")
34298 #pragma pop_macro("X")
34301 In this example, the definition of X as 1 is saved by `#pragma
34302 push_macro' and restored by `#pragma pop_macro'.
34305 File: gcc.info, Node: Function Specific Option Pragmas, Prev: Push/Pop Macro Pragmas, Up: Pragmas
34307 5.52.12 Function Specific Option Pragmas
34308 ----------------------------------------
34310 `#pragma GCC target ("STRING"...)'
34311 This pragma allows you to set target specific options for functions
34312 defined later in the source file. One or more strings can be
34313 specified. Each function that is defined after this point will be
34314 as if `attribute((target("STRING")))' was specified for that
34315 function. The parenthesis around the options is optional. *Note
34316 Function Attributes::, for more information about the `target'
34317 attribute and the attribute syntax.
34319 The `#pragma GCC target' pragma is not implemented in GCC versions
34320 earlier than 4.4, and is currently only implemented for the 386
34321 and x86_64 backends.
34323 `#pragma GCC optimize ("STRING"...)'
34324 This pragma allows you to set global optimization options for
34325 functions defined later in the source file. One or more strings
34326 can be specified. Each function that is defined after this point
34327 will be as if `attribute((optimize("STRING")))' was specified for
34328 that function. The parenthesis around the options is optional.
34329 *Note Function Attributes::, for more information about the
34330 `optimize' attribute and the attribute syntax.
34332 The `#pragma GCC optimize' pragma is not implemented in GCC
34333 versions earlier than 4.4.
34335 `#pragma GCC push_options'
34336 `#pragma GCC pop_options'
34337 These pragmas maintain a stack of the current target and
34338 optimization options. It is intended for include files where you
34339 temporarily want to switch to using a different `#pragma GCC
34340 target' or `#pragma GCC optimize' and then to pop back to the
34343 The `#pragma GCC push_options' and `#pragma GCC pop_options'
34344 pragmas are not implemented in GCC versions earlier than 4.4.
34346 `#pragma GCC reset_options'
34347 This pragma clears the current `#pragma GCC target' and `#pragma
34348 GCC optimize' to use the default switches as specified on the
34351 The `#pragma GCC reset_options' pragma is not implemented in GCC
34352 versions earlier than 4.4.
34355 File: gcc.info, Node: Unnamed Fields, Next: Thread-Local, Prev: Pragmas, Up: C Extensions
34357 5.53 Unnamed struct/union fields within structs/unions
34358 ======================================================
34360 For compatibility with other compilers, GCC allows you to define a
34361 structure or union that contains, as fields, structures and unions
34362 without names. For example:
34373 In this example, the user would be able to access members of the
34374 unnamed union with code like `foo.b'. Note that only unnamed structs
34375 and unions are allowed, you may not have, for example, an unnamed `int'.
34377 You must never create such structures that cause ambiguous field
34378 definitions. For example, this structure:
34387 It is ambiguous which `a' is being referred to with `foo.a'. Such
34388 constructs are not supported and must be avoided. In the future, such
34389 constructs may be detected and treated as compilation errors.
34391 Unless `-fms-extensions' is used, the unnamed field must be a
34392 structure or union definition without a tag (for example, `struct { int
34393 a; };'). If `-fms-extensions' is used, the field may also be a
34394 definition with a tag such as `struct foo { int a; };', a reference to
34395 a previously defined structure or union such as `struct foo;', or a
34396 reference to a `typedef' name for a previously defined structure or
34400 File: gcc.info, Node: Thread-Local, Next: Binary constants, Prev: Unnamed Fields, Up: C Extensions
34402 5.54 Thread-Local Storage
34403 =========================
34405 Thread-local storage (TLS) is a mechanism by which variables are
34406 allocated such that there is one instance of the variable per extant
34407 thread. The run-time model GCC uses to implement this originates in
34408 the IA-64 processor-specific ABI, but has since been migrated to other
34409 processors as well. It requires significant support from the linker
34410 (`ld'), dynamic linker (`ld.so'), and system libraries (`libc.so' and
34411 `libpthread.so'), so it is not available everywhere.
34413 At the user level, the extension is visible with a new storage class
34414 keyword: `__thread'. For example:
34417 extern __thread struct state s;
34418 static __thread char *p;
34420 The `__thread' specifier may be used alone, with the `extern' or
34421 `static' specifiers, but with no other storage class specifier. When
34422 used with `extern' or `static', `__thread' must appear immediately
34423 after the other storage class specifier.
34425 The `__thread' specifier may be applied to any global, file-scoped
34426 static, function-scoped static, or static data member of a class. It
34427 may not be applied to block-scoped automatic or non-static data member.
34429 When the address-of operator is applied to a thread-local variable, it
34430 is evaluated at run-time and returns the address of the current thread's
34431 instance of that variable. An address so obtained may be used by any
34432 thread. When a thread terminates, any pointers to thread-local
34433 variables in that thread become invalid.
34435 No static initialization may refer to the address of a thread-local
34438 In C++, if an initializer is present for a thread-local variable, it
34439 must be a CONSTANT-EXPRESSION, as defined in 5.19.2 of the ANSI/ISO C++
34442 See ELF Handling For Thread-Local Storage
34443 (http://people.redhat.com/drepper/tls.pdf) for a detailed explanation of
34444 the four thread-local storage addressing models, and how the run-time
34445 is expected to function.
34449 * C99 Thread-Local Edits::
34450 * C++98 Thread-Local Edits::
34453 File: gcc.info, Node: C99 Thread-Local Edits, Next: C++98 Thread-Local Edits, Up: Thread-Local
34455 5.54.1 ISO/IEC 9899:1999 Edits for Thread-Local Storage
34456 -------------------------------------------------------
34458 The following are a set of changes to ISO/IEC 9899:1999 (aka C99) that
34459 document the exact semantics of the language extension.
34461 * `5.1.2 Execution environments'
34463 Add new text after paragraph 1
34465 Within either execution environment, a "thread" is a flow of
34466 control within a program. It is implementation defined
34467 whether or not there may be more than one thread associated
34468 with a program. It is implementation defined how threads
34469 beyond the first are created, the name and type of the
34470 function called at thread startup, and how threads may be
34471 terminated. However, objects with thread storage duration
34472 shall be initialized before thread startup.
34474 * `6.2.4 Storage durations of objects'
34476 Add new text before paragraph 3
34478 An object whose identifier is declared with the storage-class
34479 specifier `__thread' has "thread storage duration". Its
34480 lifetime is the entire execution of the thread, and its
34481 stored value is initialized only once, prior to thread
34488 * `6.7.1 Storage-class specifiers'
34490 Add `__thread' to the list of storage class specifiers in
34493 Change paragraph 2 to
34495 With the exception of `__thread', at most one storage-class
34496 specifier may be given [...]. The `__thread' specifier may
34497 be used alone, or immediately following `extern' or `static'.
34499 Add new text after paragraph 6
34501 The declaration of an identifier for a variable that has
34502 block scope that specifies `__thread' shall also specify
34503 either `extern' or `static'.
34505 The `__thread' specifier shall be used only with variables.
34508 File: gcc.info, Node: C++98 Thread-Local Edits, Prev: C99 Thread-Local Edits, Up: Thread-Local
34510 5.54.2 ISO/IEC 14882:1998 Edits for Thread-Local Storage
34511 --------------------------------------------------------
34513 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
34514 that document the exact semantics of the language extension.
34516 * [intro.execution]
34518 New text after paragraph 4
34520 A "thread" is a flow of control within the abstract machine.
34521 It is implementation defined whether or not there may be more
34524 New text after paragraph 7
34526 It is unspecified whether additional action must be taken to
34527 ensure when and whether side effects are visible to other
34534 * [basic.start.main]
34536 Add after paragraph 5
34538 The thread that begins execution at the `main' function is
34539 called the "main thread". It is implementation defined how
34540 functions beginning threads other than the main thread are
34541 designated or typed. A function so designated, as well as
34542 the `main' function, is called a "thread startup function".
34543 It is implementation defined what happens if a thread startup
34544 function returns. It is implementation defined what happens
34545 to other threads when any thread calls `exit'.
34547 * [basic.start.init]
34549 Add after paragraph 4
34551 The storage for an object of thread storage duration shall be
34552 statically initialized before the first statement of the
34553 thread startup function. An object of thread storage
34554 duration shall not require dynamic initialization.
34556 * [basic.start.term]
34558 Add after paragraph 3
34560 The type of an object with thread storage duration shall not
34561 have a non-trivial destructor, nor shall it be an array type
34562 whose elements (directly or indirectly) have non-trivial
34567 Add "thread storage duration" to the list in paragraph 1.
34571 Thread, static, and automatic storage durations are
34572 associated with objects introduced by declarations [...].
34574 Add `__thread' to the list of specifiers in paragraph 3.
34576 * [basic.stc.thread]
34578 New section before [basic.stc.static]
34580 The keyword `__thread' applied to a non-local object gives the
34581 object thread storage duration.
34583 A local variable or class data member declared both `static'
34584 and `__thread' gives the variable or member thread storage
34587 * [basic.stc.static]
34591 All objects which have neither thread storage duration,
34592 dynamic storage duration nor are local [...].
34596 Add `__thread' to the list in paragraph 1.
34600 With the exception of `__thread', at most one
34601 STORAGE-CLASS-SPECIFIER shall appear in a given
34602 DECL-SPECIFIER-SEQ. The `__thread' specifier may be used
34603 alone, or immediately following the `extern' or `static'
34606 Add after paragraph 5
34608 The `__thread' specifier can be applied only to the names of
34609 objects and to anonymous unions.
34613 Add after paragraph 6
34615 Non-`static' members shall not be `__thread'.
34618 File: gcc.info, Node: Binary constants, Prev: Thread-Local, Up: C Extensions
34620 5.55 Binary constants using the `0b' prefix
34621 ===========================================
34623 Integer constants can be written as binary constants, consisting of a
34624 sequence of `0' and `1' digits, prefixed by `0b' or `0B'. This is
34625 particularly useful in environments that operate a lot on the bit-level
34626 (like microcontrollers).
34628 The following statements are identical:
34635 The type of these constants follows the same rules as for octal or
34636 hexadecimal integer constants, so suffixes like `L' or `UL' can be
34640 File: gcc.info, Node: C++ Extensions, Next: Objective-C, Prev: C Extensions, Up: Top
34642 6 Extensions to the C++ Language
34643 ********************************
34645 The GNU compiler provides these extensions to the C++ language (and you
34646 can also use most of the C language extensions in your C++ programs).
34647 If you want to write code that checks whether these features are
34648 available, you can test for the GNU compiler the same way as for C
34649 programs: check for a predefined macro `__GNUC__'. You can also use
34650 `__GNUG__' to test specifically for GNU C++ (*note Predefined Macros:
34651 (cpp)Common Predefined Macros.).
34655 * Volatiles:: What constitutes an access to a volatile object.
34656 * Restricted Pointers:: C99 restricted pointers and references.
34657 * Vague Linkage:: Where G++ puts inlines, vtables and such.
34658 * C++ Interface:: You can use a single C++ header file for both
34659 declarations and definitions.
34660 * Template Instantiation:: Methods for ensuring that exactly one copy of
34661 each needed template instantiation is emitted.
34662 * Bound member functions:: You can extract a function pointer to the
34663 method denoted by a `->*' or `.*' expression.
34664 * C++ Attributes:: Variable, function, and type attributes for C++ only.
34665 * Namespace Association:: Strong using-directives for namespace association.
34666 * Type Traits:: Compiler support for type traits
34667 * Java Exceptions:: Tweaking exception handling to work with Java.
34668 * Deprecated Features:: Things will disappear from g++.
34669 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
34672 File: gcc.info, Node: Volatiles, Next: Restricted Pointers, Up: C++ Extensions
34674 6.1 When is a Volatile Object Accessed?
34675 =======================================
34677 Both the C and C++ standard have the concept of volatile objects. These
34678 are normally accessed by pointers and used for accessing hardware. The
34679 standards encourage compilers to refrain from optimizations concerning
34680 accesses to volatile objects. The C standard leaves it implementation
34681 defined as to what constitutes a volatile access. The C++ standard
34682 omits to specify this, except to say that C++ should behave in a
34683 similar manner to C with respect to volatiles, where possible. The
34684 minimum either standard specifies is that at a sequence point all
34685 previous accesses to volatile objects have stabilized and no subsequent
34686 accesses have occurred. Thus an implementation is free to reorder and
34687 combine volatile accesses which occur between sequence points, but
34688 cannot do so for accesses across a sequence point. The use of
34689 volatiles does not allow you to violate the restriction on updating
34690 objects multiple times within a sequence point.
34692 *Note Volatile qualifier and the C compiler: Qualifiers implementation.
34694 The behavior differs slightly between C and C++ in the non-obvious
34697 volatile int *src = SOMEVALUE;
34700 With C, such expressions are rvalues, and GCC interprets this either
34701 as a read of the volatile object being pointed to or only as request to
34702 evaluate the side-effects. The C++ standard specifies that such
34703 expressions do not undergo lvalue to rvalue conversion, and that the
34704 type of the dereferenced object may be incomplete. The C++ standard
34705 does not specify explicitly that it is this lvalue to rvalue conversion
34706 which may be responsible for causing an access. However, there is
34707 reason to believe that it is, because otherwise certain simple
34708 expressions become undefined. However, because it would surprise most
34709 programmers, G++ treats dereferencing a pointer to volatile object of
34710 complete type when the value is unused as GCC would do for an
34711 equivalent type in C. When the object has incomplete type, G++ issues
34712 a warning; if you wish to force an error, you must force a conversion
34713 to rvalue with, for instance, a static cast.
34715 When using a reference to volatile, G++ does not treat equivalent
34716 expressions as accesses to volatiles, but instead issues a warning that
34717 no volatile is accessed. The rationale for this is that otherwise it
34718 becomes difficult to determine where volatile access occur, and not
34719 possible to ignore the return value from functions returning volatile
34720 references. Again, if you wish to force a read, cast the reference to
34724 File: gcc.info, Node: Restricted Pointers, Next: Vague Linkage, Prev: Volatiles, Up: C++ Extensions
34726 6.2 Restricting Pointer Aliasing
34727 ================================
34729 As with the C front end, G++ understands the C99 feature of restricted
34730 pointers, specified with the `__restrict__', or `__restrict' type
34731 qualifier. Because you cannot compile C++ by specifying the `-std=c99'
34732 language flag, `restrict' is not a keyword in C++.
34734 In addition to allowing restricted pointers, you can specify restricted
34735 references, which indicate that the reference is not aliased in the
34738 void fn (int *__restrict__ rptr, int &__restrict__ rref)
34743 In the body of `fn', RPTR points to an unaliased integer and RREF
34744 refers to a (different) unaliased integer.
34746 You may also specify whether a member function's THIS pointer is
34747 unaliased by using `__restrict__' as a member function qualifier.
34749 void T::fn () __restrict__
34754 Within the body of `T::fn', THIS will have the effective definition `T
34755 *__restrict__ const this'. Notice that the interpretation of a
34756 `__restrict__' member function qualifier is different to that of
34757 `const' or `volatile' qualifier, in that it is applied to the pointer
34758 rather than the object. This is consistent with other compilers which
34759 implement restricted pointers.
34761 As with all outermost parameter qualifiers, `__restrict__' is ignored
34762 in function definition matching. This means you only need to specify
34763 `__restrict__' in a function definition, rather than in a function
34767 File: gcc.info, Node: Vague Linkage, Next: C++ Interface, Prev: Restricted Pointers, Up: C++ Extensions
34772 There are several constructs in C++ which require space in the object
34773 file but are not clearly tied to a single translation unit. We say that
34774 these constructs have "vague linkage". Typically such constructs are
34775 emitted wherever they are needed, though sometimes we can be more
34779 Inline functions are typically defined in a header file which can
34780 be included in many different compilations. Hopefully they can
34781 usually be inlined, but sometimes an out-of-line copy is
34782 necessary, if the address of the function is taken or if inlining
34783 fails. In general, we emit an out-of-line copy in all translation
34784 units where one is needed. As an exception, we only emit inline
34785 virtual functions with the vtable, since it will always require a
34788 Local static variables and string constants used in an inline
34789 function are also considered to have vague linkage, since they
34790 must be shared between all inlined and out-of-line instances of
34794 C++ virtual functions are implemented in most compilers using a
34795 lookup table, known as a vtable. The vtable contains pointers to
34796 the virtual functions provided by a class, and each object of the
34797 class contains a pointer to its vtable (or vtables, in some
34798 multiple-inheritance situations). If the class declares any
34799 non-inline, non-pure virtual functions, the first one is chosen as
34800 the "key method" for the class, and the vtable is only emitted in
34801 the translation unit where the key method is defined.
34803 _Note:_ If the chosen key method is later defined as inline, the
34804 vtable will still be emitted in every translation unit which
34805 defines it. Make sure that any inline virtuals are declared
34806 inline in the class body, even if they are not defined there.
34809 C++ requires information about types to be written out in order to
34810 implement `dynamic_cast', `typeid' and exception handling. For
34811 polymorphic classes (classes with virtual functions), the type_info
34812 object is written out along with the vtable so that `dynamic_cast'
34813 can determine the dynamic type of a class object at runtime. For
34814 all other types, we write out the type_info object when it is
34815 used: when applying `typeid' to an expression, throwing an object,
34816 or referring to a type in a catch clause or exception
34819 Template Instantiations
34820 Most everything in this section also applies to template
34821 instantiations, but there are other options as well. *Note
34822 Where's the Template?: Template Instantiation.
34825 When used with GNU ld version 2.8 or later on an ELF system such as
34826 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
34827 these constructs will be discarded at link time. This is known as
34830 On targets that don't support COMDAT, but do support weak symbols, GCC
34831 will use them. This way one copy will override all the others, but the
34832 unused copies will still take up space in the executable.
34834 For targets which do not support either COMDAT or weak symbols, most
34835 entities with vague linkage will be emitted as local symbols to avoid
34836 duplicate definition errors from the linker. This will not happen for
34837 local statics in inlines, however, as having multiple copies will
34838 almost certainly break things.
34840 *Note Declarations and Definitions in One Header: C++ Interface, for
34841 another way to control placement of these constructs.
34844 File: gcc.info, Node: C++ Interface, Next: Template Instantiation, Prev: Vague Linkage, Up: C++ Extensions
34846 6.4 #pragma interface and implementation
34847 ========================================
34849 `#pragma interface' and `#pragma implementation' provide the user with
34850 a way of explicitly directing the compiler to emit entities with vague
34851 linkage (and debugging information) in a particular translation unit.
34853 _Note:_ As of GCC 2.7.2, these `#pragma's are not useful in most
34854 cases, because of COMDAT support and the "key method" heuristic
34855 mentioned in *Note Vague Linkage::. Using them can actually cause your
34856 program to grow due to unnecessary out-of-line copies of inline
34857 functions. Currently (3.4) the only benefit of these `#pragma's is
34858 reduced duplication of debugging information, and that should be
34859 addressed soon on DWARF 2 targets with the use of COMDAT groups.
34861 `#pragma interface'
34862 `#pragma interface "SUBDIR/OBJECTS.h"'
34863 Use this directive in _header files_ that define object classes,
34864 to save space in most of the object files that use those classes.
34865 Normally, local copies of certain information (backup copies of
34866 inline member functions, debugging information, and the internal
34867 tables that implement virtual functions) must be kept in each
34868 object file that includes class definitions. You can use this
34869 pragma to avoid such duplication. When a header file containing
34870 `#pragma interface' is included in a compilation, this auxiliary
34871 information will not be generated (unless the main input source
34872 file itself uses `#pragma implementation'). Instead, the object
34873 files will contain references to be resolved at link time.
34875 The second form of this directive is useful for the case where you
34876 have multiple headers with the same name in different directories.
34877 If you use this form, you must specify the same string to `#pragma
34880 `#pragma implementation'
34881 `#pragma implementation "OBJECTS.h"'
34882 Use this pragma in a _main input file_, when you want full output
34883 from included header files to be generated (and made globally
34884 visible). The included header file, in turn, should use `#pragma
34885 interface'. Backup copies of inline member functions, debugging
34886 information, and the internal tables used to implement virtual
34887 functions are all generated in implementation files.
34889 If you use `#pragma implementation' with no argument, it applies to
34890 an include file with the same basename(1) as your source file.
34891 For example, in `allclass.cc', giving just `#pragma implementation'
34892 by itself is equivalent to `#pragma implementation "allclass.h"'.
34894 In versions of GNU C++ prior to 2.6.0 `allclass.h' was treated as
34895 an implementation file whenever you would include it from
34896 `allclass.cc' even if you never specified `#pragma
34897 implementation'. This was deemed to be more trouble than it was
34898 worth, however, and disabled.
34900 Use the string argument if you want a single implementation file to
34901 include code from multiple header files. (You must also use
34902 `#include' to include the header file; `#pragma implementation'
34903 only specifies how to use the file--it doesn't actually include
34906 There is no way to split up the contents of a single header file
34907 into multiple implementation files.
34909 `#pragma implementation' and `#pragma interface' also have an effect
34910 on function inlining.
34912 If you define a class in a header file marked with `#pragma
34913 interface', the effect on an inline function defined in that class is
34914 similar to an explicit `extern' declaration--the compiler emits no code
34915 at all to define an independent version of the function. Its
34916 definition is used only for inlining with its callers.
34918 Conversely, when you include the same header file in a main source file
34919 that declares it as `#pragma implementation', the compiler emits code
34920 for the function itself; this defines a version of the function that
34921 can be found via pointers (or by callers compiled without inlining).
34922 If all calls to the function can be inlined, you can avoid emitting the
34923 function by compiling with `-fno-implement-inlines'. If any calls were
34924 not inlined, you will get linker errors.
34926 ---------- Footnotes ----------
34928 (1) A file's "basename" was the name stripped of all leading path
34929 information and of trailing suffixes, such as `.h' or `.C' or `.cc'.
34932 File: gcc.info, Node: Template Instantiation, Next: Bound member functions, Prev: C++ Interface, Up: C++ Extensions
34934 6.5 Where's the Template?
34935 =========================
34937 C++ templates are the first language feature to require more
34938 intelligence from the environment than one usually finds on a UNIX
34939 system. Somehow the compiler and linker have to make sure that each
34940 template instance occurs exactly once in the executable if it is needed,
34941 and not at all otherwise. There are two basic approaches to this
34942 problem, which are referred to as the Borland model and the Cfront
34946 Borland C++ solved the template instantiation problem by adding
34947 the code equivalent of common blocks to their linker; the compiler
34948 emits template instances in each translation unit that uses them,
34949 and the linker collapses them together. The advantage of this
34950 model is that the linker only has to consider the object files
34951 themselves; there is no external complexity to worry about. This
34952 disadvantage is that compilation time is increased because the
34953 template code is being compiled repeatedly. Code written for this
34954 model tends to include definitions of all templates in the header
34955 file, since they must be seen to be instantiated.
34958 The AT&T C++ translator, Cfront, solved the template instantiation
34959 problem by creating the notion of a template repository, an
34960 automatically maintained place where template instances are
34961 stored. A more modern version of the repository works as follows:
34962 As individual object files are built, the compiler places any
34963 template definitions and instantiations encountered in the
34964 repository. At link time, the link wrapper adds in the objects in
34965 the repository and compiles any needed instances that were not
34966 previously emitted. The advantages of this model are more optimal
34967 compilation speed and the ability to use the system linker; to
34968 implement the Borland model a compiler vendor also needs to
34969 replace the linker. The disadvantages are vastly increased
34970 complexity, and thus potential for error; for some code this can be
34971 just as transparent, but in practice it can been very difficult to
34972 build multiple programs in one directory and one program in
34973 multiple directories. Code written for this model tends to
34974 separate definitions of non-inline member templates into a
34975 separate file, which should be compiled separately.
34977 When used with GNU ld version 2.8 or later on an ELF system such as
34978 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
34979 Borland model. On other systems, G++ implements neither automatic
34982 A future version of G++ will support a hybrid model whereby the
34983 compiler will emit any instantiations for which the template definition
34984 is included in the compile, and store template definitions and
34985 instantiation context information into the object file for the rest.
34986 The link wrapper will extract that information as necessary and invoke
34987 the compiler to produce the remaining instantiations. The linker will
34988 then combine duplicate instantiations.
34990 In the mean time, you have the following options for dealing with
34991 template instantiations:
34993 1. Compile your template-using code with `-frepo'. The compiler will
34994 generate files with the extension `.rpo' listing all of the
34995 template instantiations used in the corresponding object files
34996 which could be instantiated there; the link wrapper, `collect2',
34997 will then update the `.rpo' files to tell the compiler where to
34998 place those instantiations and rebuild any affected object files.
34999 The link-time overhead is negligible after the first pass, as the
35000 compiler will continue to place the instantiations in the same
35003 This is your best option for application code written for the
35004 Borland model, as it will just work. Code written for the Cfront
35005 model will need to be modified so that the template definitions
35006 are available at one or more points of instantiation; usually this
35007 is as simple as adding `#include <tmethods.cc>' to the end of each
35010 For library code, if you want the library to provide all of the
35011 template instantiations it needs, just try to link all of its
35012 object files together; the link will fail, but cause the
35013 instantiations to be generated as a side effect. Be warned,
35014 however, that this may cause conflicts if multiple libraries try
35015 to provide the same instantiations. For greater control, use
35016 explicit instantiation as described in the next option.
35018 2. Compile your code with `-fno-implicit-templates' to disable the
35019 implicit generation of template instances, and explicitly
35020 instantiate all the ones you use. This approach requires more
35021 knowledge of exactly which instances you need than do the others,
35022 but it's less mysterious and allows greater control. You can
35023 scatter the explicit instantiations throughout your program,
35024 perhaps putting them in the translation units where the instances
35025 are used or the translation units that define the templates
35026 themselves; you can put all of the explicit instantiations you
35027 need into one big file; or you can create small files like
35032 template class Foo<int>;
35033 template ostream& operator <<
35034 (ostream&, const Foo<int>&);
35036 for each of the instances you need, and create a template
35037 instantiation library from those.
35039 If you are using Cfront-model code, you can probably get away with
35040 not using `-fno-implicit-templates' when compiling files that don't
35041 `#include' the member template definitions.
35043 If you use one big file to do the instantiations, you may want to
35044 compile it without `-fno-implicit-templates' so you get all of the
35045 instances required by your explicit instantiations (but not by any
35046 other files) without having to specify them as well.
35048 G++ has extended the template instantiation syntax given in the ISO
35049 standard to allow forward declaration of explicit instantiations
35050 (with `extern'), instantiation of the compiler support data for a
35051 template class (i.e. the vtable) without instantiating any of its
35052 members (with `inline'), and instantiation of only the static data
35053 members of a template class, without the support data or member
35054 functions (with (`static'):
35056 extern template int max (int, int);
35057 inline template class Foo<int>;
35058 static template class Foo<int>;
35060 3. Do nothing. Pretend G++ does implement automatic instantiation
35061 management. Code written for the Borland model will work fine, but
35062 each translation unit will contain instances of each of the
35063 templates it uses. In a large program, this can lead to an
35064 unacceptable amount of code duplication.
35067 File: gcc.info, Node: Bound member functions, Next: C++ Attributes, Prev: Template Instantiation, Up: C++ Extensions
35069 6.6 Extracting the function pointer from a bound pointer to member function
35070 ===========================================================================
35072 In C++, pointer to member functions (PMFs) are implemented using a wide
35073 pointer of sorts to handle all the possible call mechanisms; the PMF
35074 needs to store information about how to adjust the `this' pointer, and
35075 if the function pointed to is virtual, where to find the vtable, and
35076 where in the vtable to look for the member function. If you are using
35077 PMFs in an inner loop, you should really reconsider that decision. If
35078 that is not an option, you can extract the pointer to the function that
35079 would be called for a given object/PMF pair and call it directly inside
35080 the inner loop, to save a bit of time.
35082 Note that you will still be paying the penalty for the call through a
35083 function pointer; on most modern architectures, such a call defeats the
35084 branch prediction features of the CPU. This is also true of normal
35085 virtual function calls.
35087 The syntax for this extension is
35090 extern int (A::*fp)();
35091 typedef int (*fptr)(A *);
35093 fptr p = (fptr)(a.*fp);
35095 For PMF constants (i.e. expressions of the form `&Klasse::Member'), no
35096 object is needed to obtain the address of the function. They can be
35097 converted to function pointers directly:
35099 fptr p1 = (fptr)(&A::foo);
35101 You must specify `-Wno-pmf-conversions' to use this extension.
35104 File: gcc.info, Node: C++ Attributes, Next: Namespace Association, Prev: Bound member functions, Up: C++ Extensions
35106 6.7 C++-Specific Variable, Function, and Type Attributes
35107 ========================================================
35109 Some attributes only make sense for C++ programs.
35111 `init_priority (PRIORITY)'
35112 In Standard C++, objects defined at namespace scope are guaranteed
35113 to be initialized in an order in strict accordance with that of
35114 their definitions _in a given translation unit_. No guarantee is
35115 made for initializations across translation units. However, GNU
35116 C++ allows users to control the order of initialization of objects
35117 defined at namespace scope with the `init_priority' attribute by
35118 specifying a relative PRIORITY, a constant integral expression
35119 currently bounded between 101 and 65535 inclusive. Lower numbers
35120 indicate a higher priority.
35122 In the following example, `A' would normally be created before
35123 `B', but the `init_priority' attribute has reversed that order:
35125 Some_Class A __attribute__ ((init_priority (2000)));
35126 Some_Class B __attribute__ ((init_priority (543)));
35128 Note that the particular values of PRIORITY do not matter; only
35129 their relative ordering.
35132 This type attribute informs C++ that the class is a Java
35133 interface. It may only be applied to classes declared within an
35134 `extern "Java"' block. Calls to methods declared in this
35135 interface will be dispatched using GCJ's interface table
35136 mechanism, instead of regular virtual table dispatch.
35139 See also *Note Namespace Association::.
35142 File: gcc.info, Node: Namespace Association, Next: Type Traits, Prev: C++ Attributes, Up: C++ Extensions
35144 6.8 Namespace Association
35145 =========================
35147 *Caution:* The semantics of this extension are not fully defined.
35148 Users should refrain from using this extension as its semantics may
35149 change subtly over time. It is possible that this extension will be
35150 removed in future versions of G++.
35152 A using-directive with `__attribute ((strong))' is stronger than a
35153 normal using-directive in two ways:
35155 * Templates from the used namespace can be specialized and explicitly
35156 instantiated as though they were members of the using namespace.
35158 * The using namespace is considered an associated namespace of all
35159 templates in the used namespace for purposes of argument-dependent
35162 The used namespace must be nested within the using namespace so that
35163 normal unqualified lookup works properly.
35165 This is useful for composing a namespace transparently from
35166 implementation namespaces. For example:
35170 template <class T> struct A { };
35172 using namespace debug __attribute ((__strong__));
35173 template <> struct A<int> { }; // ok to specialize
35175 template <class T> void f (A<T>);
35180 f (std::A<float>()); // lookup finds std::f
35185 File: gcc.info, Node: Type Traits, Next: Java Exceptions, Prev: Namespace Association, Up: C++ Extensions
35190 The C++ front-end implements syntactic extensions that allow to
35191 determine at compile time various characteristics of a type (or of a
35194 `__has_nothrow_assign (type)'
35195 If `type' is const qualified or is a reference type then the trait
35196 is false. Otherwise if `__has_trivial_assign (type)' is true then
35197 the trait is true, else if `type' is a cv class or union type with
35198 copy assignment operators that are known not to throw an exception
35199 then the trait is true, else it is false. Requires: `type' shall
35200 be a complete type, an array type of unknown bound, or is a `void'
35203 `__has_nothrow_copy (type)'
35204 If `__has_trivial_copy (type)' is true then the trait is true,
35205 else if `type' is a cv class or union type with copy constructors
35206 that are known not to throw an exception then the trait is true,
35207 else it is false. Requires: `type' shall be a complete type, an
35208 array type of unknown bound, or is a `void' type.
35210 `__has_nothrow_constructor (type)'
35211 If `__has_trivial_constructor (type)' is true then the trait is
35212 true, else if `type' is a cv class or union type (or array
35213 thereof) with a default constructor that is known not to throw an
35214 exception then the trait is true, else it is false. Requires:
35215 `type' shall be a complete type, an array type of unknown bound,
35216 or is a `void' type.
35218 `__has_trivial_assign (type)'
35219 If `type' is const qualified or is a reference type then the trait
35220 is false. Otherwise if `__is_pod (type)' is true then the trait is
35221 true, else if `type' is a cv class or union type with a trivial
35222 copy assignment ([class.copy]) then the trait is true, else it is
35223 false. Requires: `type' shall be a complete type, an array type
35224 of unknown bound, or is a `void' type.
35226 `__has_trivial_copy (type)'
35227 If `__is_pod (type)' is true or `type' is a reference type then
35228 the trait is true, else if `type' is a cv class or union type with
35229 a trivial copy constructor ([class.copy]) then the trait is true,
35230 else it is false. Requires: `type' shall be a complete type, an
35231 array type of unknown bound, or is a `void' type.
35233 `__has_trivial_constructor (type)'
35234 If `__is_pod (type)' is true then the trait is true, else if
35235 `type' is a cv class or union type (or array thereof) with a
35236 trivial default constructor ([class.ctor]) then the trait is true,
35237 else it is false. Requires: `type' shall be a complete type, an
35238 array type of unknown bound, or is a `void' type.
35240 `__has_trivial_destructor (type)'
35241 If `__is_pod (type)' is true or `type' is a reference type then
35242 the trait is true, else if `type' is a cv class or union type (or
35243 array thereof) with a trivial destructor ([class.dtor]) then the
35244 trait is true, else it is false. Requires: `type' shall be a
35245 complete type, an array type of unknown bound, or is a `void' type.
35247 `__has_virtual_destructor (type)'
35248 If `type' is a class type with a virtual destructor ([class.dtor])
35249 then the trait is true, else it is false. Requires: `type' shall
35250 be a complete type, an array type of unknown bound, or is a `void'
35253 `__is_abstract (type)'
35254 If `type' is an abstract class ([class.abstract]) then the trait
35255 is true, else it is false. Requires: `type' shall be a complete
35256 type, an array type of unknown bound, or is a `void' type.
35258 `__is_base_of (base_type, derived_type)'
35259 If `base_type' is a base class of `derived_type' ([class.derived])
35260 then the trait is true, otherwise it is false. Top-level cv
35261 qualifications of `base_type' and `derived_type' are ignored. For
35262 the purposes of this trait, a class type is considered is own
35263 base. Requires: if `__is_class (base_type)' and `__is_class
35264 (derived_type)' are true and `base_type' and `derived_type' are
35265 not the same type (disregarding cv-qualifiers), `derived_type'
35266 shall be a complete type. Diagnostic is produced if this
35267 requirement is not met.
35269 `__is_class (type)'
35270 If `type' is a cv class type, and not a union type
35271 ([basic.compound]) the trait is true, else it is false.
35273 `__is_empty (type)'
35274 If `__is_class (type)' is false then the trait is false.
35275 Otherwise `type' is considered empty if and only if: `type' has no
35276 non-static data members, or all non-static data members, if any,
35277 are bit-fields of length 0, and `type' has no virtual members, and
35278 `type' has no virtual base classes, and `type' has no base classes
35279 `base_type' for which `__is_empty (base_type)' is false.
35280 Requires: `type' shall be a complete type, an array type of
35281 unknown bound, or is a `void' type.
35284 If `type' is a cv enumeration type ([basic.compound]) the trait is
35285 true, else it is false.
35288 If `type' is a cv POD type ([basic.types]) then the trait is true,
35289 else it is false. Requires: `type' shall be a complete type, an
35290 array type of unknown bound, or is a `void' type.
35292 `__is_polymorphic (type)'
35293 If `type' is a polymorphic class ([class.virtual]) then the trait
35294 is true, else it is false. Requires: `type' shall be a complete
35295 type, an array type of unknown bound, or is a `void' type.
35297 `__is_union (type)'
35298 If `type' is a cv union type ([basic.compound]) the trait is true,
35303 File: gcc.info, Node: Java Exceptions, Next: Deprecated Features, Prev: Type Traits, Up: C++ Extensions
35305 6.10 Java Exceptions
35306 ====================
35308 The Java language uses a slightly different exception handling model
35309 from C++. Normally, GNU C++ will automatically detect when you are
35310 writing C++ code that uses Java exceptions, and handle them
35311 appropriately. However, if C++ code only needs to execute destructors
35312 when Java exceptions are thrown through it, GCC will guess incorrectly.
35313 Sample problematic code is:
35315 struct S { ~S(); };
35316 extern void bar(); // is written in Java, and may throw exceptions
35323 The usual effect of an incorrect guess is a link failure, complaining of
35324 a missing routine called `__gxx_personality_v0'.
35326 You can inform the compiler that Java exceptions are to be used in a
35327 translation unit, irrespective of what it might think, by writing
35328 `#pragma GCC java_exceptions' at the head of the file. This `#pragma'
35329 must appear before any functions that throw or catch exceptions, or run
35330 destructors when exceptions are thrown through them.
35332 You cannot mix Java and C++ exceptions in the same translation unit.
35333 It is believed to be safe to throw a C++ exception from one file through
35334 another file compiled for the Java exception model, or vice versa, but
35335 there may be bugs in this area.
35338 File: gcc.info, Node: Deprecated Features, Next: Backwards Compatibility, Prev: Java Exceptions, Up: C++ Extensions
35340 6.11 Deprecated Features
35341 ========================
35343 In the past, the GNU C++ compiler was extended to experiment with new
35344 features, at a time when the C++ language was still evolving. Now that
35345 the C++ standard is complete, some of those features are superseded by
35346 superior alternatives. Using the old features might cause a warning in
35347 some cases that the feature will be dropped in the future. In other
35348 cases, the feature might be gone already.
35350 While the list below is not exhaustive, it documents some of the
35351 options that are now deprecated:
35353 `-fexternal-templates'
35354 `-falt-external-templates'
35355 These are two of the many ways for G++ to implement template
35356 instantiation. *Note Template Instantiation::. The C++ standard
35357 clearly defines how template definitions have to be organized
35358 across implementation units. G++ has an implicit instantiation
35359 mechanism that should work just fine for standard-conforming code.
35361 `-fstrict-prototype'
35362 `-fno-strict-prototype'
35363 Previously it was possible to use an empty prototype parameter
35364 list to indicate an unspecified number of parameters (like C),
35365 rather than no parameters, as C++ demands. This feature has been
35366 removed, except where it is required for backwards compatibility.
35367 *Note Backwards Compatibility::.
35369 G++ allows a virtual function returning `void *' to be overridden by
35370 one returning a different pointer type. This extension to the
35371 covariant return type rules is now deprecated and will be removed from a
35374 The G++ minimum and maximum operators (`<?' and `>?') and their
35375 compound forms (`<?=') and `>?=') have been deprecated and are now
35376 removed from G++. Code using these operators should be modified to use
35377 `std::min' and `std::max' instead.
35379 The named return value extension has been deprecated, and is now
35382 The use of initializer lists with new expressions has been deprecated,
35383 and is now removed from G++.
35385 Floating and complex non-type template parameters have been deprecated,
35386 and are now removed from G++.
35388 The implicit typename extension has been deprecated and is now removed
35391 The use of default arguments in function pointers, function typedefs
35392 and other places where they are not permitted by the standard is
35393 deprecated and will be removed from a future version of G++.
35395 G++ allows floating-point literals to appear in integral constant
35396 expressions, e.g. ` enum E { e = int(2.2 * 3.7) } ' This extension is
35397 deprecated and will be removed from a future version.
35399 G++ allows static data members of const floating-point type to be
35400 declared with an initializer in a class definition. The standard only
35401 allows initializers for static members of const integral types and const
35402 enumeration types so this extension has been deprecated and will be
35403 removed from a future version.
35406 File: gcc.info, Node: Backwards Compatibility, Prev: Deprecated Features, Up: C++ Extensions
35408 6.12 Backwards Compatibility
35409 ============================
35411 Now that there is a definitive ISO standard C++, G++ has a specification
35412 to adhere to. The C++ language evolved over time, and features that
35413 used to be acceptable in previous drafts of the standard, such as the
35414 ARM [Annotated C++ Reference Manual], are no longer accepted. In order
35415 to allow compilation of C++ written to such drafts, G++ contains some
35416 backwards compatibilities. _All such backwards compatibility features
35417 are liable to disappear in future versions of G++._ They should be
35418 considered deprecated. *Note Deprecated Features::.
35421 If a variable is declared at for scope, it used to remain in scope
35422 until the end of the scope which contained the for statement
35423 (rather than just within the for scope). G++ retains this, but
35424 issues a warning, if such a variable is accessed outside the for
35427 `Implicit C language'
35428 Old C system header files did not contain an `extern "C" {...}'
35429 scope to set the language. On such systems, all header files are
35430 implicitly scoped inside a C language scope. Also, an empty
35431 prototype `()' will be treated as an unspecified number of
35432 arguments, rather than no arguments, as C++ demands.
35435 File: gcc.info, Node: Objective-C, Next: Compatibility, Prev: C++ Extensions, Up: Top
35437 7 GNU Objective-C runtime features
35438 **********************************
35440 This document is meant to describe some of the GNU Objective-C runtime
35441 features. It is not intended to teach you Objective-C, there are
35442 several resources on the Internet that present the language. Questions
35443 and comments about this document to Ovidiu Predescu <ovidiu@cup.hp.com>.
35447 * Executing code before main::
35449 * Garbage Collection::
35450 * Constant string objects::
35451 * compatibility_alias::
35454 File: gcc.info, Node: Executing code before main, Next: Type encoding, Prev: Objective-C, Up: Objective-C
35456 7.1 `+load': Executing code before main
35457 =======================================
35459 The GNU Objective-C runtime provides a way that allows you to execute
35460 code before the execution of the program enters the `main' function.
35461 The code is executed on a per-class and a per-category basis, through a
35462 special class method `+load'.
35464 This facility is very useful if you want to initialize global variables
35465 which can be accessed by the program directly, without sending a message
35466 to the class first. The usual way to initialize global variables, in
35467 the `+initialize' method, might not be useful because `+initialize' is
35468 only called when the first message is sent to a class object, which in
35469 some cases could be too late.
35471 Suppose for example you have a `FileStream' class that declares
35472 `Stdin', `Stdout' and `Stderr' as global variables, like below:
35475 FileStream *Stdin = nil;
35476 FileStream *Stdout = nil;
35477 FileStream *Stderr = nil;
35479 @implementation FileStream
35483 Stdin = [[FileStream new] initWithFd:0];
35484 Stdout = [[FileStream new] initWithFd:1];
35485 Stderr = [[FileStream new] initWithFd:2];
35488 /* Other methods here */
35491 In this example, the initialization of `Stdin', `Stdout' and `Stderr'
35492 in `+initialize' occurs too late. The programmer can send a message to
35493 one of these objects before the variables are actually initialized,
35494 thus sending messages to the `nil' object. The `+initialize' method
35495 which actually initializes the global variables is not invoked until
35496 the first message is sent to the class object. The solution would
35497 require these variables to be initialized just before entering `main'.
35499 The correct solution of the above problem is to use the `+load' method
35500 instead of `+initialize':
35503 @implementation FileStream
35507 Stdin = [[FileStream new] initWithFd:0];
35508 Stdout = [[FileStream new] initWithFd:1];
35509 Stderr = [[FileStream new] initWithFd:2];
35512 /* Other methods here */
35515 The `+load' is a method that is not overridden by categories. If a
35516 class and a category of it both implement `+load', both methods are
35517 invoked. This allows some additional initializations to be performed in
35520 This mechanism is not intended to be a replacement for `+initialize'.
35521 You should be aware of its limitations when you decide to use it
35522 instead of `+initialize'.
35526 * What you can and what you cannot do in +load::
35529 File: gcc.info, Node: What you can and what you cannot do in +load, Prev: Executing code before main, Up: Executing code before main
35531 7.1.1 What you can and what you cannot do in `+load'
35532 ----------------------------------------------------
35534 The `+load' implementation in the GNU runtime guarantees you the
35537 * you can write whatever C code you like;
35539 * you can send messages to Objective-C constant strings (`@"this is a
35540 constant string"');
35542 * you can allocate and send messages to objects whose class is
35543 implemented in the same file;
35545 * the `+load' implementation of all super classes of a class are
35546 executed before the `+load' of that class is executed;
35548 * the `+load' implementation of a class is executed before the
35549 `+load' implementation of any category.
35552 In particular, the following things, even if they can work in a
35553 particular case, are not guaranteed:
35555 * allocation of or sending messages to arbitrary objects;
35557 * allocation of or sending messages to objects whose classes have a
35558 category implemented in the same file;
35561 You should make no assumptions about receiving `+load' in sibling
35562 classes when you write `+load' of a class. The order in which sibling
35563 classes receive `+load' is not guaranteed.
35565 The order in which `+load' and `+initialize' are called could be
35566 problematic if this matters. If you don't allocate objects inside
35567 `+load', it is guaranteed that `+load' is called before `+initialize'.
35568 If you create an object inside `+load' the `+initialize' method of
35569 object's class is invoked even if `+load' was not invoked. Note if you
35570 explicitly call `+load' on a class, `+initialize' will be called first.
35571 To avoid possible problems try to implement only one of these methods.
35573 The `+load' method is also invoked when a bundle is dynamically loaded
35574 into your running program. This happens automatically without any
35575 intervening operation from you. When you write bundles and you need to
35576 write `+load' you can safely create and send messages to objects whose
35577 classes already exist in the running program. The same restrictions as
35578 above apply to classes defined in bundle.
35581 File: gcc.info, Node: Type encoding, Next: Garbage Collection, Prev: Executing code before main, Up: Objective-C
35586 The Objective-C compiler generates type encodings for all the types.
35587 These type encodings are used at runtime to find out information about
35588 selectors and methods and about objects and classes.
35590 The types are encoded in the following way:
35594 `unsigned char' `C'
35596 `unsigned short' `S'
35600 `unsigned long' `L'
35612 Complex types `j' followed by the inner type. For example
35613 `_Complex double' is encoded as "jd".
35614 bit-fields `b' followed by the starting position of the
35615 bit-field, the type of the bit-field and the size of
35616 the bit-field (the bit-fields encoding was changed
35617 from the NeXT's compiler encoding, see below)
35619 The encoding of bit-fields has changed to allow bit-fields to be
35620 properly handled by the runtime functions that compute sizes and
35621 alignments of types that contain bit-fields. The previous encoding
35622 contained only the size of the bit-field. Using only this information
35623 it is not possible to reliably compute the size occupied by the
35624 bit-field. This is very important in the presence of the Boehm's
35625 garbage collector because the objects are allocated using the typed
35626 memory facility available in this collector. The typed memory
35627 allocation requires information about where the pointers are located
35630 The position in the bit-field is the position, counting in bits, of the
35631 bit closest to the beginning of the structure.
35633 The non-atomic types are encoded as follows:
35635 pointers `^' followed by the pointed type.
35636 arrays `[' followed by the number of elements in the array
35637 followed by the type of the elements followed by `]'
35638 structures `{' followed by the name of the structure (or `?' if the
35639 structure is unnamed), the `=' sign, the type of the
35641 unions `(' followed by the name of the structure (or `?' if the
35642 union is unnamed), the `=' sign, the type of the members
35645 Here are some types and their encodings, as they are generated by the
35646 compiler on an i386 machine:
35649 Objective-C type Compiler encoding
35651 struct { `{?=i[3f]b128i3b131i2c}'
35660 In addition to the types the compiler also encodes the type
35661 specifiers. The table below describes the encoding of the current
35662 Objective-C type specifiers:
35674 The type specifiers are encoded just before the type. Unlike types
35675 however, the type specifiers are only encoded when they appear in method
35679 File: gcc.info, Node: Garbage Collection, Next: Constant string objects, Prev: Type encoding, Up: Objective-C
35681 7.3 Garbage Collection
35682 ======================
35684 Support for a new memory management policy has been added by using a
35685 powerful conservative garbage collector, known as the
35686 Boehm-Demers-Weiser conservative garbage collector. It is available
35687 from `http://www.hpl.hp.com/personal/Hans_Boehm/gc/'.
35689 To enable the support for it you have to configure the compiler using
35690 an additional argument, `--enable-objc-gc'. You need to have garbage
35691 collector installed before building the compiler. This will build an
35692 additional runtime library which has several enhancements to support
35693 the garbage collector. The new library has a new name, `libobjc_gc.a'
35694 to not conflict with the non-garbage-collected library.
35696 When the garbage collector is used, the objects are allocated using the
35697 so-called typed memory allocation mechanism available in the
35698 Boehm-Demers-Weiser collector. This mode requires precise information
35699 on where pointers are located inside objects. This information is
35700 computed once per class, immediately after the class has been
35703 There is a new runtime function `class_ivar_set_gcinvisible()' which
35704 can be used to declare a so-called "weak pointer" reference. Such a
35705 pointer is basically hidden for the garbage collector; this can be
35706 useful in certain situations, especially when you want to keep track of
35707 the allocated objects, yet allow them to be collected. This kind of
35708 pointers can only be members of objects, you cannot declare a global
35709 pointer as a weak reference. Every type which is a pointer type can be
35710 declared a weak pointer, including `id', `Class' and `SEL'.
35712 Here is an example of how to use this feature. Suppose you want to
35713 implement a class whose instances hold a weak pointer reference; the
35714 following class does this:
35717 @interface WeakPointer : Object
35719 const void* weakPointer;
35722 - initWithPointer:(const void*)p;
35723 - (const void*)weakPointer;
35727 @implementation WeakPointer
35731 class_ivar_set_gcinvisible (self, "weakPointer", YES);
35734 - initWithPointer:(const void*)p
35740 - (const void*)weakPointer
35742 return weakPointer;
35747 Weak pointers are supported through a new type character specifier
35748 represented by the `!' character. The `class_ivar_set_gcinvisible()'
35749 function adds or removes this specifier to the string type description
35750 of the instance variable named as argument.
35753 File: gcc.info, Node: Constant string objects, Next: compatibility_alias, Prev: Garbage Collection, Up: Objective-C
35755 7.4 Constant string objects
35756 ===========================
35758 GNU Objective-C provides constant string objects that are generated
35759 directly by the compiler. You declare a constant string object by
35760 prefixing a C constant string with the character `@':
35762 id myString = @"this is a constant string object";
35764 The constant string objects are by default instances of the
35765 `NXConstantString' class which is provided by the GNU Objective-C
35766 runtime. To get the definition of this class you must include the
35767 `objc/NXConstStr.h' header file.
35769 User defined libraries may want to implement their own constant string
35770 class. To be able to support them, the GNU Objective-C compiler
35771 provides a new command line options
35772 `-fconstant-string-class=CLASS-NAME'. The provided class should adhere
35773 to a strict structure, the same as `NXConstantString''s structure:
35776 @interface MyConstantStringClass
35784 `NXConstantString' inherits from `Object'; user class libraries may
35785 choose to inherit the customized constant string class from a different
35786 class than `Object'. There is no requirement in the methods the
35787 constant string class has to implement, but the final ivar layout of
35788 the class must be the compatible with the given structure.
35790 When the compiler creates the statically allocated constant string
35791 object, the `c_string' field will be filled by the compiler with the
35792 string; the `length' field will be filled by the compiler with the
35793 string length; the `isa' pointer will be filled with `NULL' by the
35794 compiler, and it will later be fixed up automatically at runtime by the
35795 GNU Objective-C runtime library to point to the class which was set by
35796 the `-fconstant-string-class' option when the object file is loaded (if
35797 you wonder how it works behind the scenes, the name of the class to
35798 use, and the list of static objects to fixup, are stored by the
35799 compiler in the object file in a place where the GNU runtime library
35800 will find them at runtime).
35802 As a result, when a file is compiled with the
35803 `-fconstant-string-class' option, all the constant string objects will
35804 be instances of the class specified as argument to this option. It is
35805 possible to have multiple compilation units referring to different
35806 constant string classes, neither the compiler nor the linker impose any
35807 restrictions in doing this.
35810 File: gcc.info, Node: compatibility_alias, Prev: Constant string objects, Up: Objective-C
35812 7.5 compatibility_alias
35813 =======================
35815 This is a feature of the Objective-C compiler rather than of the
35816 runtime, anyway since it is documented nowhere and its existence was
35817 forgotten, we are documenting it here.
35819 The keyword `@compatibility_alias' allows you to define a class name
35820 as equivalent to another class name. For example:
35822 @compatibility_alias WOApplication GSWApplication;
35824 tells the compiler that each time it encounters `WOApplication' as a
35825 class name, it should replace it with `GSWApplication' (that is,
35826 `WOApplication' is just an alias for `GSWApplication').
35828 There are some constraints on how this can be used--
35830 * `WOApplication' (the alias) must not be an existing class;
35832 * `GSWApplication' (the real class) must be an existing class.
35836 File: gcc.info, Node: Compatibility, Next: Gcov, Prev: Objective-C, Up: Top
35838 8 Binary Compatibility
35839 **********************
35841 Binary compatibility encompasses several related concepts:
35843 "application binary interface (ABI)"
35844 The set of runtime conventions followed by all of the tools that
35845 deal with binary representations of a program, including
35846 compilers, assemblers, linkers, and language runtime support.
35847 Some ABIs are formal with a written specification, possibly
35848 designed by multiple interested parties. Others are simply the
35849 way things are actually done by a particular set of tools.
35852 A compiler conforms to an ABI if it generates code that follows
35853 all of the specifications enumerated by that ABI. A library
35854 conforms to an ABI if it is implemented according to that ABI. An
35855 application conforms to an ABI if it is built using tools that
35856 conform to that ABI and does not contain source code that
35857 specifically changes behavior specified by the ABI.
35859 "calling conventions"
35860 Calling conventions are a subset of an ABI that specify of how
35861 arguments are passed and function results are returned.
35864 Different sets of tools are interoperable if they generate files
35865 that can be used in the same program. The set of tools includes
35866 compilers, assemblers, linkers, libraries, header files, startup
35867 files, and debuggers. Binaries produced by different sets of
35868 tools are not interoperable unless they implement the same ABI.
35869 This applies to different versions of the same tools as well as
35870 tools from different vendors.
35873 Whether a function in a binary built by one set of tools can call a
35874 function in a binary built by a different set of tools is a subset
35875 of interoperability.
35877 "implementation-defined features"
35878 Language standards include lists of implementation-defined
35879 features whose behavior can vary from one implementation to
35880 another. Some of these features are normally covered by a
35881 platform's ABI and others are not. The features that are not
35882 covered by an ABI generally affect how a program behaves, but not
35886 Conformance to the same ABI and the same behavior of
35887 implementation-defined features are both relevant for
35890 The application binary interface implemented by a C or C++ compiler
35891 affects code generation and runtime support for:
35893 * size and alignment of data types
35895 * layout of structured types
35897 * calling conventions
35899 * register usage conventions
35901 * interfaces for runtime arithmetic support
35903 * object file formats
35905 In addition, the application binary interface implemented by a C++
35906 compiler affects code generation and runtime support for:
35909 * exception handling
35911 * invoking constructors and destructors
35913 * layout, alignment, and padding of classes
35915 * layout and alignment of virtual tables
35917 Some GCC compilation options cause the compiler to generate code that
35918 does not conform to the platform's default ABI. Other options cause
35919 different program behavior for implementation-defined features that are
35920 not covered by an ABI. These options are provided for consistency with
35921 other compilers that do not follow the platform's default ABI or the
35922 usual behavior of implementation-defined features for the platform. Be
35923 very careful about using such options.
35925 Most platforms have a well-defined ABI that covers C code, but ABIs
35926 that cover C++ functionality are not yet common.
35928 Starting with GCC 3.2, GCC binary conventions for C++ are based on a
35929 written, vendor-neutral C++ ABI that was designed to be specific to
35930 64-bit Itanium but also includes generic specifications that apply to
35931 any platform. This C++ ABI is also implemented by other compiler
35932 vendors on some platforms, notably GNU/Linux and BSD systems. We have
35933 tried hard to provide a stable ABI that will be compatible with future
35934 GCC releases, but it is possible that we will encounter problems that
35935 make this difficult. Such problems could include different
35936 interpretations of the C++ ABI by different vendors, bugs in the ABI, or
35937 bugs in the implementation of the ABI in different compilers. GCC's
35938 `-Wabi' switch warns when G++ generates code that is probably not
35939 compatible with the C++ ABI.
35941 The C++ library used with a C++ compiler includes the Standard C++
35942 Library, with functionality defined in the C++ Standard, plus language
35943 runtime support. The runtime support is included in a C++ ABI, but
35944 there is no formal ABI for the Standard C++ Library. Two
35945 implementations of that library are interoperable if one follows the
35946 de-facto ABI of the other and if they are both built with the same
35947 compiler, or with compilers that conform to the same ABI for C++
35948 compiler and runtime support.
35950 When G++ and another C++ compiler conform to the same C++ ABI, but the
35951 implementations of the Standard C++ Library that they normally use do
35952 not follow the same ABI for the Standard C++ Library, object files
35953 built with those compilers can be used in the same program only if they
35954 use the same C++ library. This requires specifying the location of the
35955 C++ library header files when invoking the compiler whose usual library
35956 is not being used. The location of GCC's C++ header files depends on
35957 how the GCC build was configured, but can be seen by using the G++ `-v'
35958 option. With default configuration options for G++ 3.3 the compile
35959 line for a different C++ compiler needs to include
35961 -IGCC_INSTALL_DIRECTORY/include/c++/3.3
35963 Similarly, compiling code with G++ that must use a C++ library other
35964 than the GNU C++ library requires specifying the location of the header
35965 files for that other library.
35967 The most straightforward way to link a program to use a particular C++
35968 library is to use a C++ driver that specifies that C++ library by
35969 default. The `g++' driver, for example, tells the linker where to find
35970 GCC's C++ library (`libstdc++') plus the other libraries and startup
35971 files it needs, in the proper order.
35973 If a program must use a different C++ library and it's not possible to
35974 do the final link using a C++ driver that uses that library by default,
35975 it is necessary to tell `g++' the location and name of that library.
35976 It might also be necessary to specify different startup files and other
35977 runtime support libraries, and to suppress the use of GCC's support
35978 libraries with one or more of the options `-nostdlib', `-nostartfiles',
35979 and `-nodefaultlibs'.
35982 File: gcc.info, Node: Gcov, Next: Trouble, Prev: Compatibility, Up: Top
35984 9 `gcov'--a Test Coverage Program
35985 *********************************
35987 `gcov' is a tool you can use in conjunction with GCC to test code
35988 coverage in your programs.
35992 * Gcov Intro:: Introduction to gcov.
35993 * Invoking Gcov:: How to use gcov.
35994 * Gcov and Optimization:: Using gcov with GCC optimization.
35995 * Gcov Data Files:: The files used by gcov.
35996 * Cross-profiling:: Data file relocation.
35999 File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov
36001 9.1 Introduction to `gcov'
36002 ==========================
36004 `gcov' is a test coverage program. Use it in concert with GCC to
36005 analyze your programs to help create more efficient, faster running
36006 code and to discover untested parts of your program. You can use
36007 `gcov' as a profiling tool to help discover where your optimization
36008 efforts will best affect your code. You can also use `gcov' along with
36009 the other profiling tool, `gprof', to assess which parts of your code
36010 use the greatest amount of computing time.
36012 Profiling tools help you analyze your code's performance. Using a
36013 profiler such as `gcov' or `gprof', you can find out some basic
36014 performance statistics, such as:
36016 * how often each line of code executes
36018 * what lines of code are actually executed
36020 * how much computing time each section of code uses
36022 Once you know these things about how your code works when compiled, you
36023 can look at each module to see which modules should be optimized.
36024 `gcov' helps you determine where to work on optimization.
36026 Software developers also use coverage testing in concert with
36027 testsuites, to make sure software is actually good enough for a release.
36028 Testsuites can verify that a program works as expected; a coverage
36029 program tests to see how much of the program is exercised by the
36030 testsuite. Developers can then determine what kinds of test cases need
36031 to be added to the testsuites to create both better testing and a better
36034 You should compile your code without optimization if you plan to use
36035 `gcov' because the optimization, by combining some lines of code into
36036 one function, may not give you as much information as you need to look
36037 for `hot spots' where the code is using a great deal of computer time.
36038 Likewise, because `gcov' accumulates statistics by line (at the lowest
36039 resolution), it works best with a programming style that places only
36040 one statement on each line. If you use complicated macros that expand
36041 to loops or to other control structures, the statistics are less
36042 helpful--they only report on the line where the macro call appears. If
36043 your complex macros behave like functions, you can replace them with
36044 inline functions to solve this problem.
36046 `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates how
36047 many times each line of a source file `SOURCEFILE.c' has executed. You
36048 can use these logfiles along with `gprof' to aid in fine-tuning the
36049 performance of your programs. `gprof' gives timing information you can
36050 use along with the information you get from `gcov'.
36052 `gcov' works only on code compiled with GCC. It is not compatible
36053 with any other profiling or test coverage mechanism.
36056 File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov
36058 9.2 Invoking `gcov'
36059 ===================
36061 gcov [OPTIONS] SOURCEFILES
36063 `gcov' accepts the following options:
36067 Display help about using `gcov' (on the standard output), and exit
36068 without doing any further processing.
36072 Display the `gcov' version number (on the standard output), and
36073 exit without doing any further processing.
36077 Write individual execution counts for every basic block. Normally
36078 gcov outputs execution counts only for the main blocks of a line.
36079 With this option you can determine if blocks within a single line
36080 are not being executed.
36083 `--branch-probabilities'
36084 Write branch frequencies to the output file, and write branch
36085 summary info to the standard output. This option allows you to
36086 see how often each branch in your program was taken.
36087 Unconditional branches will not be shown, unless the `-u' option
36092 Write branch frequencies as the number of branches taken, rather
36093 than the percentage of branches taken.
36097 Do not create the `gcov' output file.
36100 `--long-file-names'
36101 Create long file names for included source files. For example, if
36102 the header file `x.h' contains code, and was included in the file
36103 `a.c', then running `gcov' on the file `a.c' will produce an
36104 output file called `a.c##x.h.gcov' instead of `x.h.gcov'. This
36105 can be useful if `x.h' is included in multiple source files. If
36106 you use the `-p' option, both the including and included file
36107 names will be complete path names.
36111 Preserve complete path information in the names of generated
36112 `.gcov' files. Without this option, just the filename component is
36113 used. With this option, all directories are used, with `/'
36114 characters translated to `#' characters, `.' directory components
36115 removed and `..' components renamed to `^'. This is useful if
36116 sourcefiles are in several different directories. It also affects
36120 `--function-summaries'
36121 Output summaries for each function in addition to the file level
36124 `-o DIRECTORY|FILE'
36125 `--object-directory DIRECTORY'
36126 `--object-file FILE'
36127 Specify either the directory containing the gcov data files, or the
36128 object path name. The `.gcno', and `.gcda' data files are
36129 searched for using this option. If a directory is specified, the
36130 data files are in that directory and named after the source file
36131 name, without its extension. If a file is specified here, the
36132 data files are named after that file, without its extension. If
36133 this option is not supplied, it defaults to the current directory.
36136 `--unconditional-branches'
36137 When branch probabilities are given, include those of
36138 unconditional branches. Unconditional branches are normally not
36142 `--intermediate-format'
36143 Output gcov file in an intermediate text format that can be used by
36144 `lcov' or other applications. It will output a single *.gcov file
36145 per *gcda file. No source code required.
36148 `gcov' should be run with the current directory the same as that when
36149 you invoked the compiler. Otherwise it will not be able to locate the
36150 source files. `gcov' produces files called `MANGLEDNAME.gcov' in the
36151 current directory. These contain the coverage information of the
36152 source file they correspond to. One `.gcov' file is produced for each
36153 source file containing code, which was compiled to produce the data
36154 files. The MANGLEDNAME part of the output file name is usually simply
36155 the source file name, but can be something more complicated if the `-l'
36156 or `-p' options are given. Refer to those options for details.
36158 The `.gcov' files contain the `:' separated fields along with program
36159 source code. The format is
36161 EXECUTION_COUNT:LINE_NUMBER:SOURCE LINE TEXT
36163 Additional block information may succeed each line, when requested by
36164 command line option. The EXECUTION_COUNT is `-' for lines containing
36165 no code and `#####' for lines which were never executed. Some lines of
36166 information at the start have LINE_NUMBER of zero.
36168 The preamble lines are of the form
36172 The ordering and number of these preamble lines will be augmented as
36173 `gcov' development progresses -- do not rely on them remaining
36174 unchanged. Use TAG to locate a particular preamble line.
36176 The additional block information is of the form
36180 The INFORMATION is human readable, but designed to be simple enough
36181 for machine parsing too.
36183 When printing percentages, 0% and 100% are only printed when the values
36184 are _exactly_ 0% and 100% respectively. Other values which would
36185 conventionally be rounded to 0% or 100% are instead printed as the
36186 nearest non-boundary value.
36188 When using `gcov', you must first compile your program with two
36189 special GCC options: `-fprofile-arcs -ftest-coverage'. This tells the
36190 compiler to generate additional information needed by gcov (basically a
36191 flow graph of the program) and also includes additional code in the
36192 object files for generating the extra profiling information needed by
36193 gcov. These additional files are placed in the directory where the
36194 object file is located.
36196 Running the program will cause profile output to be generated. For
36197 each source file compiled with `-fprofile-arcs', an accompanying
36198 `.gcda' file will be placed in the object file directory.
36200 Running `gcov' with your program's source file names as arguments will
36201 now produce a listing of the code along with frequency of execution for
36202 each line. For example, if your program is called `tmp.c', this is
36203 what you see when you use the basic `gcov' facility:
36205 $ gcc -fprofile-arcs -ftest-coverage tmp.c
36208 90.00% of 10 source lines executed in file tmp.c
36209 Creating tmp.c.gcov.
36211 The file `tmp.c.gcov' contains output from `gcov'. Here is a sample:
36214 -: 0:Graph:tmp.gcno
36218 -: 1:#include <stdio.h>
36220 -: 3:int main (void)
36222 1: 5: int i, total;
36226 11: 9: for (i = 0; i < 10; i++)
36227 10: 10: total += i;
36229 1: 12: if (total != 45)
36230 #####: 13: printf ("Failure\n");
36232 1: 15: printf ("Success\n");
36236 When you use the `-a' option, you will get individual block counts,
36237 and the output looks like this:
36240 -: 0:Graph:tmp.gcno
36244 -: 1:#include <stdio.h>
36246 -: 3:int main (void)
36249 1: 5: int i, total;
36253 11: 9: for (i = 0; i < 10; i++)
36255 10: 10: total += i;
36258 1: 12: if (total != 45)
36260 #####: 13: printf ("Failure\n");
36263 1: 15: printf ("Success\n");
36269 In this mode, each basic block is only shown on one line - the last
36270 line of the block. A multi-line block will only contribute to the
36271 execution count of that last line, and other lines will not be shown to
36272 contain code, unless previous blocks end on those lines. The total
36273 execution count of a line is shown and subsequent lines show the
36274 execution counts for individual blocks that end on that line. After
36275 each block, the branch and call counts of the block will be shown, if
36276 the `-b' option is given.
36278 Because of the way GCC instruments calls, a call count can be shown
36279 after a line with no individual blocks. As you can see, line 13
36280 contains a basic block that was not executed.
36282 When you use the `-b' option, your output looks like this:
36285 90.00% of 10 source lines executed in file tmp.c
36286 80.00% of 5 branches executed in file tmp.c
36287 80.00% of 5 branches taken at least once in file tmp.c
36288 50.00% of 2 calls executed in file tmp.c
36289 Creating tmp.c.gcov.
36291 Here is a sample of a resulting `tmp.c.gcov' file:
36294 -: 0:Graph:tmp.gcno
36298 -: 1:#include <stdio.h>
36300 -: 3:int main (void)
36301 function main called 1 returned 1 blocks executed 75%
36303 1: 5: int i, total;
36307 11: 9: for (i = 0; i < 10; i++)
36308 branch 0 taken 91% (fallthrough)
36310 10: 10: total += i;
36312 1: 12: if (total != 45)
36313 branch 0 taken 0% (fallthrough)
36314 branch 1 taken 100%
36315 #####: 13: printf ("Failure\n");
36316 call 0 never executed
36318 1: 15: printf ("Success\n");
36319 call 0 called 1 returned 100%
36323 For each function, a line is printed showing how many times the
36324 function is called, how many times it returns and what percentage of the
36325 function's blocks were executed.
36327 For each basic block, a line is printed after the last line of the
36328 basic block describing the branch or call that ends the basic block.
36329 There can be multiple branches and calls listed for a single source
36330 line if there are multiple basic blocks that end on that line. In this
36331 case, the branches and calls are each given a number. There is no
36332 simple way to map these branches and calls back to source constructs.
36333 In general, though, the lowest numbered branch or call will correspond
36334 to the leftmost construct on the source line.
36336 For a branch, if it was executed at least once, then a percentage
36337 indicating the number of times the branch was taken divided by the
36338 number of times the branch was executed will be printed. Otherwise, the
36339 message "never executed" is printed.
36341 For a call, if it was executed at least once, then a percentage
36342 indicating the number of times the call returned divided by the number
36343 of times the call was executed will be printed. This will usually be
36344 100%, but may be less for functions that call `exit' or `longjmp', and
36345 thus may not return every time they are called.
36347 The execution counts are cumulative. If the example program were
36348 executed again without removing the `.gcda' file, the count for the
36349 number of times each line in the source was executed would be added to
36350 the results of the previous run(s). This is potentially useful in
36351 several ways. For example, it could be used to accumulate data over a
36352 number of program runs as part of a test verification suite, or to
36353 provide more accurate long-term information over a large number of
36356 The data in the `.gcda' files is saved immediately before the program
36357 exits. For each source file compiled with `-fprofile-arcs', the
36358 profiling code first attempts to read in an existing `.gcda' file; if
36359 the file doesn't match the executable (differing number of basic block
36360 counts) it will ignore the contents of the file. It then adds in the
36361 new execution counts and finally writes the data to the file.
36364 File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov
36366 9.3 Using `gcov' with GCC Optimization
36367 ======================================
36369 If you plan to use `gcov' to help optimize your code, you must first
36370 compile your program with two special GCC options: `-fprofile-arcs
36371 -ftest-coverage'. Aside from that, you can use any other GCC options;
36372 but if you want to prove that every single line in your program was
36373 executed, you should not compile with optimization at the same time.
36374 On some machines the optimizer can eliminate some simple code lines by
36375 combining them with other lines. For example, code like this:
36382 can be compiled into one instruction on some machines. In this case,
36383 there is no way for `gcov' to calculate separate execution counts for
36384 each line because there isn't separate code for each line. Hence the
36385 `gcov' output looks like this if you compiled the program with
36388 100: 12:if (a != b)
36393 The output shows that this block of code, combined by optimization,
36394 executed 100 times. In one sense this result is correct, because there
36395 was only one instruction representing all four of these lines. However,
36396 the output does not indicate how many times the result was 0 and how
36397 many times the result was 1.
36399 Inlineable functions can create unexpected line counts. Line counts
36400 are shown for the source code of the inlineable function, but what is
36401 shown depends on where the function is inlined, or if it is not inlined
36404 If the function is not inlined, the compiler must emit an out of line
36405 copy of the function, in any object file that needs it. If `fileA.o'
36406 and `fileB.o' both contain out of line bodies of a particular
36407 inlineable function, they will also both contain coverage counts for
36408 that function. When `fileA.o' and `fileB.o' are linked together, the
36409 linker will, on many systems, select one of those out of line bodies
36410 for all calls to that function, and remove or ignore the other.
36411 Unfortunately, it will not remove the coverage counters for the unused
36412 function body. Hence when instrumented, all but one use of that
36413 function will show zero counts.
36415 If the function is inlined in several places, the block structure in
36416 each location might not be the same. For instance, a condition might
36417 now be calculable at compile time in some instances. Because the
36418 coverage of all the uses of the inline function will be shown for the
36419 same source lines, the line counts themselves might seem inconsistent.
36422 File: gcc.info, Node: Gcov Data Files, Next: Cross-profiling, Prev: Gcov and Optimization, Up: Gcov
36424 9.4 Brief description of `gcov' data files
36425 ==========================================
36427 `gcov' uses two files for profiling. The names of these files are
36428 derived from the original _object_ file by substituting the file suffix
36429 with either `.gcno', or `.gcda'. All of these files are placed in the
36430 same directory as the object file, and contain data stored in a
36431 platform-independent format.
36433 The `.gcno' file is generated when the source file is compiled with
36434 the GCC `-ftest-coverage' option. It contains information to
36435 reconstruct the basic block graphs and assign source line numbers to
36438 The `.gcda' file is generated when a program containing object files
36439 built with the GCC `-fprofile-arcs' option is executed. A separate
36440 `.gcda' file is created for each object file compiled with this option.
36441 It contains arc transition counts, and some summary information.
36443 The full details of the file format is specified in `gcov-io.h', and
36444 functions provided in that header file should be used to access the
36448 File: gcc.info, Node: Cross-profiling, Prev: Gcov Data Files, Up: Gcov
36450 9.5 Data file relocation to support cross-profiling
36451 ===================================================
36453 Running the program will cause profile output to be generated. For each
36454 source file compiled with `-fprofile-arcs', an accompanying `.gcda'
36455 file will be placed in the object file directory. That implicitly
36456 requires running the program on the same system as it was built or
36457 having the same absolute directory structure on the target system. The
36458 program will try to create the needed directory structure, if it is not
36461 To support cross-profiling, a program compiled with `-fprofile-arcs'
36462 can relocate the data files based on two environment variables:
36464 * GCOV_PREFIX contains the prefix to add to the absolute paths in
36465 the object file. Prefix must be absolute as well, otherwise its
36466 value is ignored. The default is no prefix.
36468 * GCOV_PREFIX_STRIP indicates the how many initial directory names
36469 to strip off the hardwired absolute paths. Default value is 0.
36471 _Note:_ GCOV_PREFIX_STRIP has no effect if GCOV_PREFIX is
36472 undefined, empty or non-absolute.
36474 For example, if the object file `/user/build/foo.o' was built with
36475 `-fprofile-arcs', the final executable will try to create the data file
36476 `/user/build/foo.gcda' when running on the target system. This will
36477 fail if the corresponding directory does not exist and it is unable to
36478 create it. This can be overcome by, for example, setting the
36479 environment as `GCOV_PREFIX=/target/run' and `GCOV_PREFIX_STRIP=1'.
36480 Such a setting will name the data file `/target/run/build/foo.gcda'.
36482 You must move the data files to the expected directory tree in order to
36483 use them for profile directed optimizations (`--use-profile'), or to
36484 use the `gcov' tool.
36487 File: gcc.info, Node: Trouble, Next: Bugs, Prev: Gcov, Up: Top
36489 10 Known Causes of Trouble with GCC
36490 ***********************************
36492 This section describes known problems that affect users of GCC. Most
36493 of these are not GCC bugs per se--if they were, we would fix them. But
36494 the result for a user may be like the result of a bug.
36496 Some of these problems are due to bugs in other software, some are
36497 missing features that are too much work to add, and some are places
36498 where people's opinions differ as to what is best.
36502 * Actual Bugs:: Bugs we will fix later.
36503 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
36504 * Interoperation:: Problems using GCC with other compilers,
36505 and with certain linkers, assemblers and debuggers.
36506 * Incompatibilities:: GCC is incompatible with traditional C.
36507 * Fixed Headers:: GCC uses corrected versions of system header files.
36508 This is necessary, but doesn't always work smoothly.
36509 * Standard Libraries:: GCC uses the system C library, which might not be
36510 compliant with the ISO C standard.
36511 * Disappointments:: Regrettable things we can't change, but not quite bugs.
36512 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
36513 * Protoize Caveats:: Things to watch out for when using `protoize'.
36514 * Non-bugs:: Things we think are right, but some others disagree.
36515 * Warnings and Errors:: Which problems in your code get warnings,
36516 and which get errors.
36519 File: gcc.info, Node: Actual Bugs, Next: Cross-Compiler Problems, Up: Trouble
36521 10.1 Actual Bugs We Haven't Fixed Yet
36522 =====================================
36524 * The `fixincludes' script interacts badly with automounters; if the
36525 directory of system header files is automounted, it tends to be
36526 unmounted while `fixincludes' is running. This would seem to be a
36527 bug in the automounter. We don't know any good way to work around
36530 * The `fixproto' script will sometimes add prototypes for the
36531 `sigsetjmp' and `siglongjmp' functions that reference the
36532 `jmp_buf' type before that type is defined. To work around this,
36533 edit the offending file and place the typedef in front of the
36537 File: gcc.info, Node: Cross-Compiler Problems, Next: Interoperation, Prev: Actual Bugs, Up: Trouble
36539 10.2 Cross-Compiler Problems
36540 ============================
36542 You may run into problems with cross compilation on certain machines,
36543 for several reasons.
36545 * At present, the program `mips-tfile' which adds debug support to
36546 object files on MIPS systems does not work in a cross compile
36550 File: gcc.info, Node: Interoperation, Next: Incompatibilities, Prev: Cross-Compiler Problems, Up: Trouble
36552 10.3 Interoperation
36553 ===================
36555 This section lists various difficulties encountered in using GCC
36556 together with other compilers or with the assemblers, linkers,
36557 libraries and debuggers on certain systems.
36559 * On many platforms, GCC supports a different ABI for C++ than do
36560 other compilers, so the object files compiled by GCC cannot be
36561 used with object files generated by another C++ compiler.
36563 An area where the difference is most apparent is name mangling.
36564 The use of different name mangling is intentional, to protect you
36565 from more subtle problems. Compilers differ as to many internal
36566 details of C++ implementation, including: how class instances are
36567 laid out, how multiple inheritance is implemented, and how virtual
36568 function calls are handled. If the name encoding were made the
36569 same, your programs would link against libraries provided from
36570 other compilers--but the programs would then crash when run.
36571 Incompatible libraries are then detected at link time, rather than
36574 * On some BSD systems, including some versions of Ultrix, use of
36575 profiling causes static variable destructors (currently used only
36576 in C++) not to be run.
36578 * On some SGI systems, when you use `-lgl_s' as an option, it gets
36579 translated magically to `-lgl_s -lX11_s -lc_s'. Naturally, this
36580 does not happen when you use GCC. You must specify all three
36581 options explicitly.
36583 * On a SPARC, GCC aligns all values of type `double' on an 8-byte
36584 boundary, and it expects every `double' to be so aligned. The Sun
36585 compiler usually gives `double' values 8-byte alignment, with one
36586 exception: function arguments of type `double' may not be aligned.
36588 As a result, if a function compiled with Sun CC takes the address
36589 of an argument of type `double' and passes this pointer of type
36590 `double *' to a function compiled with GCC, dereferencing the
36591 pointer may cause a fatal signal.
36593 One way to solve this problem is to compile your entire program
36594 with GCC. Another solution is to modify the function that is
36595 compiled with Sun CC to copy the argument into a local variable;
36596 local variables are always properly aligned. A third solution is
36597 to modify the function that uses the pointer to dereference it via
36598 the following function `access_double' instead of directly with
36602 access_double (double *unaligned_ptr)
36604 union d2i { double d; int i[2]; };
36606 union d2i *p = (union d2i *) unaligned_ptr;
36615 Storing into the pointer can be done likewise with the same union.
36617 * On Solaris, the `malloc' function in the `libmalloc.a' library may
36618 allocate memory that is only 4 byte aligned. Since GCC on the
36619 SPARC assumes that doubles are 8 byte aligned, this may result in a
36620 fatal signal if doubles are stored in memory allocated by the
36621 `libmalloc.a' library.
36623 The solution is to not use the `libmalloc.a' library. Use instead
36624 `malloc' and related functions from `libc.a'; they do not have
36627 * On the HP PA machine, ADB sometimes fails to work on functions
36628 compiled with GCC. Specifically, it fails to work on functions
36629 that use `alloca' or variable-size arrays. This is because GCC
36630 doesn't generate HP-UX unwind descriptors for such functions. It
36631 may even be impossible to generate them.
36633 * Debugging (`-g') is not supported on the HP PA machine, unless you
36634 use the preliminary GNU tools.
36636 * Taking the address of a label may generate errors from the HP-UX
36637 PA assembler. GAS for the PA does not have this problem.
36639 * Using floating point parameters for indirect calls to static
36640 functions will not work when using the HP assembler. There simply
36641 is no way for GCC to specify what registers hold arguments for
36642 static functions when using the HP assembler. GAS for the PA does
36643 not have this problem.
36645 * In extremely rare cases involving some very large functions you may
36646 receive errors from the HP linker complaining about an out of
36647 bounds unconditional branch offset. This used to occur more often
36648 in previous versions of GCC, but is now exceptionally rare. If
36649 you should run into it, you can work around by making your
36652 * GCC compiled code sometimes emits warnings from the HP-UX
36653 assembler of the form:
36655 (warning) Use of GR3 when
36656 frame >= 8192 may cause conflict.
36658 These warnings are harmless and can be safely ignored.
36660 * In extremely rare cases involving some very large functions you may
36661 receive errors from the AIX Assembler complaining about a
36662 displacement that is too large. If you should run into it, you
36663 can work around by making your function smaller.
36665 * The `libstdc++.a' library in GCC relies on the SVR4 dynamic linker
36666 semantics which merges global symbols between libraries and
36667 applications, especially necessary for C++ streams functionality.
36668 This is not the default behavior of AIX shared libraries and
36669 dynamic linking. `libstdc++.a' is built on AIX with
36670 "runtime-linking" enabled so that symbol merging can occur. To
36671 utilize this feature, the application linked with `libstdc++.a'
36672 must include the `-Wl,-brtl' flag on the link line. G++ cannot
36673 impose this because this option may interfere with the semantics
36674 of the user program and users may not always use `g++' to link his
36675 or her application. Applications are not required to use the
36676 `-Wl,-brtl' flag on the link line--the rest of the `libstdc++.a'
36677 library which is not dependent on the symbol merging semantics
36678 will continue to function correctly.
36680 * An application can interpose its own definition of functions for
36681 functions invoked by `libstdc++.a' with "runtime-linking" enabled
36682 on AIX. To accomplish this the application must be linked with
36683 "runtime-linking" option and the functions explicitly must be
36684 exported by the application (`-Wl,-brtl,-bE:exportfile').
36686 * AIX on the RS/6000 provides support (NLS) for environments outside
36687 of the United States. Compilers and assemblers use NLS to support
36688 locale-specific representations of various objects including
36689 floating-point numbers (`.' vs `,' for separating decimal
36690 fractions). There have been problems reported where the library
36691 linked with GCC does not produce the same floating-point formats
36692 that the assembler accepts. If you have this problem, set the
36693 `LANG' environment variable to `C' or `En_US'.
36695 * Even if you specify `-fdollars-in-identifiers', you cannot
36696 successfully use `$' in identifiers on the RS/6000 due to a
36697 restriction in the IBM assembler. GAS supports these identifiers.
36701 File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: Interoperation, Up: Trouble
36703 10.4 Incompatibilities of GCC
36704 =============================
36706 There are several noteworthy incompatibilities between GNU C and K&R
36707 (non-ISO) versions of C.
36709 * GCC normally makes string constants read-only. If several
36710 identical-looking string constants are used, GCC stores only one
36711 copy of the string.
36713 One consequence is that you cannot call `mktemp' with a string
36714 constant argument. The function `mktemp' always alters the string
36715 its argument points to.
36717 Another consequence is that `sscanf' does not work on some very
36718 old systems when passed a string constant as its format control
36719 string or input. This is because `sscanf' incorrectly tries to
36720 write into the string constant. Likewise `fscanf' and `scanf'.
36722 The solution to these problems is to change the program to use
36723 `char'-array variables with initialization strings for these
36724 purposes instead of string constants.
36726 * `-2147483648' is positive.
36728 This is because 2147483648 cannot fit in the type `int', so
36729 (following the ISO C rules) its data type is `unsigned long int'.
36730 Negating this value yields 2147483648 again.
36732 * GCC does not substitute macro arguments when they appear inside of
36733 string constants. For example, the following macro in GCC
36737 will produce output `"a"' regardless of what the argument A is.
36739 * When you use `setjmp' and `longjmp', the only automatic variables
36740 guaranteed to remain valid are those declared `volatile'. This is
36741 a consequence of automatic register allocation. Consider this
36755 /* `longjmp (j)' may occur in `fun3'. */
36756 return a + fun3 ();
36759 Here `a' may or may not be restored to its first value when the
36760 `longjmp' occurs. If `a' is allocated in a register, then its
36761 first value is restored; otherwise, it keeps the last value stored
36764 If you use the `-W' option with the `-O' option, you will get a
36765 warning when GCC thinks such a problem might be possible.
36767 * Programs that use preprocessing directives in the middle of macro
36768 arguments do not work with GCC. For example, a program like this
36775 ISO C does not permit such a construct.
36777 * K&R compilers allow comments to cross over an inclusion boundary
36778 (i.e. started in an include file and ended in the including file).
36780 * Declarations of external variables and functions within a block
36781 apply only to the block containing the declaration. In other
36782 words, they have the same scope as any other declaration in the
36785 In some other C compilers, a `extern' declaration affects all the
36786 rest of the file even if it happens within a block.
36788 * In traditional C, you can combine `long', etc., with a typedef
36789 name, as shown here:
36792 typedef long foo bar;
36794 In ISO C, this is not allowed: `long' and other type modifiers
36795 require an explicit `int'.
36797 * PCC allows typedef names to be used as function parameters.
36799 * Traditional C allows the following erroneous pair of declarations
36800 to appear together in a given scope:
36805 * GCC treats all characters of identifiers as significant.
36806 According to K&R-1 (2.2), "No more than the first eight characters
36807 are significant, although more may be used.". Also according to
36808 K&R-1 (2.2), "An identifier is a sequence of letters and digits;
36809 the first character must be a letter. The underscore _ counts as
36810 a letter.", but GCC also allows dollar signs in identifiers.
36812 * PCC allows whitespace in the middle of compound assignment
36813 operators such as `+='. GCC, following the ISO standard, does not
36816 * GCC complains about unterminated character constants inside of
36817 preprocessing conditionals that fail. Some programs have English
36818 comments enclosed in conditionals that are guaranteed to fail; if
36819 these comments contain apostrophes, GCC will probably report an
36820 error. For example, this code would produce an error:
36823 You can't expect this to work.
36826 The best solution to such a problem is to put the text into an
36827 actual C comment delimited by `/*...*/'.
36829 * Many user programs contain the declaration `long time ();'. In the
36830 past, the system header files on many systems did not actually
36831 declare `time', so it did not matter what type your program
36832 declared it to return. But in systems with ISO C headers, `time'
36833 is declared to return `time_t', and if that is not the same as
36834 `long', then `long time ();' is erroneous.
36836 The solution is to change your program to use appropriate system
36837 headers (`<time.h>' on systems with ISO C headers) and not to
36838 declare `time' if the system header files declare it, or failing
36839 that to use `time_t' as the return type of `time'.
36841 * When compiling functions that return `float', PCC converts it to a
36842 double. GCC actually returns a `float'. If you are concerned
36843 with PCC compatibility, you should declare your functions to return
36844 `double'; you might as well say what you mean.
36846 * When compiling functions that return structures or unions, GCC
36847 output code normally uses a method different from that used on most
36848 versions of Unix. As a result, code compiled with GCC cannot call
36849 a structure-returning function compiled with PCC, and vice versa.
36851 The method used by GCC is as follows: a structure or union which is
36852 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or
36853 union with any other size is stored into an address supplied by
36854 the caller (usually in a special, fixed register, but on some
36855 machines it is passed on the stack). The target hook
36856 `TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address.
36858 By contrast, PCC on most target machines returns structures and
36859 unions of any size by copying the data into an area of static
36860 storage, and then returning the address of that storage as if it
36861 were a pointer value. The caller must copy the data from that
36862 memory area to the place where the value is wanted. GCC does not
36863 use this method because it is slower and nonreentrant.
36865 On some newer machines, PCC uses a reentrant convention for all
36866 structure and union returning. GCC on most of these machines uses
36867 a compatible convention when returning structures and unions in
36868 memory, but still returns small structures and unions in registers.
36870 You can tell GCC to use a compatible convention for all structure
36871 and union returning with the option `-fpcc-struct-return'.
36873 * GCC complains about program fragments such as `0x74ae-0x4000'
36874 which appear to be two hexadecimal constants separated by the minus
36875 operator. Actually, this string is a single "preprocessing token".
36876 Each such token must correspond to one token in C. Since this
36877 does not, GCC prints an error message. Although it may appear
36878 obvious that what is meant is an operator and two values, the ISO
36879 C standard specifically requires that this be treated as erroneous.
36881 A "preprocessing token" is a "preprocessing number" if it begins
36882 with a digit and is followed by letters, underscores, digits,
36883 periods and `e+', `e-', `E+', `E-', `p+', `p-', `P+', or `P-'
36884 character sequences. (In strict C89 mode, the sequences `p+',
36885 `p-', `P+' and `P-' cannot appear in preprocessing numbers.)
36887 To make the above program fragment valid, place whitespace in
36888 front of the minus sign. This whitespace will end the
36889 preprocessing number.
36892 File: gcc.info, Node: Fixed Headers, Next: Standard Libraries, Prev: Incompatibilities, Up: Trouble
36894 10.5 Fixed Header Files
36895 =======================
36897 GCC needs to install corrected versions of some system header files.
36898 This is because most target systems have some header files that won't
36899 work with GCC unless they are changed. Some have bugs, some are
36900 incompatible with ISO C, and some depend on special features of other
36903 Installing GCC automatically creates and installs the fixed header
36904 files, by running a program called `fixincludes'. Normally, you don't
36905 need to pay attention to this. But there are cases where it doesn't do
36906 the right thing automatically.
36908 * If you update the system's header files, such as by installing a
36909 new system version, the fixed header files of GCC are not
36910 automatically updated. They can be updated using the `mkheaders'
36911 script installed in `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
36913 * On some systems, header file directories contain machine-specific
36914 symbolic links in certain places. This makes it possible to share
36915 most of the header files among hosts running the same version of
36916 the system on different machine models.
36918 The programs that fix the header files do not understand this
36919 special way of using symbolic links; therefore, the directory of
36920 fixed header files is good only for the machine model used to
36923 It is possible to make separate sets of fixed header files for the
36924 different machine models, and arrange a structure of symbolic
36925 links so as to use the proper set, but you'll have to do this by
36929 File: gcc.info, Node: Standard Libraries, Next: Disappointments, Prev: Fixed Headers, Up: Trouble
36931 10.6 Standard Libraries
36932 =======================
36934 GCC by itself attempts to be a conforming freestanding implementation.
36935 *Note Language Standards Supported by GCC: Standards, for details of
36936 what this means. Beyond the library facilities required of such an
36937 implementation, the rest of the C library is supplied by the vendor of
36938 the operating system. If that C library doesn't conform to the C
36939 standards, then your programs might get warnings (especially when using
36940 `-Wall') that you don't expect.
36942 For example, the `sprintf' function on SunOS 4.1.3 returns `char *'
36943 while the C standard says that `sprintf' returns an `int'. The
36944 `fixincludes' program could make the prototype for this function match
36945 the Standard, but that would be wrong, since the function will still
36948 If you need a Standard compliant library, then you need to find one, as
36949 GCC does not provide one. The GNU C library (called `glibc') provides
36950 ISO C, POSIX, BSD, SystemV and X/Open compatibility for GNU/Linux and
36951 HURD-based GNU systems; no recent version of it supports other systems,
36952 though some very old versions did. Version 2.2 of the GNU C library
36953 includes nearly complete C99 support. You could also ask your
36954 operating system vendor if newer libraries are available.
36957 File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Standard Libraries, Up: Trouble
36959 10.7 Disappointments and Misunderstandings
36960 ==========================================
36962 These problems are perhaps regrettable, but we don't know any practical
36965 * Certain local variables aren't recognized by debuggers when you
36966 compile with optimization.
36968 This occurs because sometimes GCC optimizes the variable out of
36969 existence. There is no way to tell the debugger how to compute the
36970 value such a variable "would have had", and it is not clear that
36971 would be desirable anyway. So GCC simply does not mention the
36972 eliminated variable when it writes debugging information.
36974 You have to expect a certain amount of disagreement between the
36975 executable and your source code, when you use optimization.
36977 * Users often think it is a bug when GCC reports an error for code
36980 int foo (struct mumble *);
36982 struct mumble { ... };
36984 int foo (struct mumble *x)
36987 This code really is erroneous, because the scope of `struct
36988 mumble' in the prototype is limited to the argument list
36989 containing it. It does not refer to the `struct mumble' defined
36990 with file scope immediately below--they are two unrelated types
36991 with similar names in different scopes.
36993 But in the definition of `foo', the file-scope type is used
36994 because that is available to be inherited. Thus, the definition
36995 and the prototype do not match, and you get an error.
36997 This behavior may seem silly, but it's what the ISO standard
36998 specifies. It is easy enough for you to make your code work by
36999 moving the definition of `struct mumble' above the prototype.
37000 It's not worth being incompatible with ISO C just to avoid an
37001 error for the example shown above.
37003 * Accesses to bit-fields even in volatile objects works by accessing
37004 larger objects, such as a byte or a word. You cannot rely on what
37005 size of object is accessed in order to read or write the
37006 bit-field; it may even vary for a given bit-field according to the
37009 If you care about controlling the amount of memory that is
37010 accessed, use volatile but do not use bit-fields.
37012 * GCC comes with shell scripts to fix certain known problems in
37013 system header files. They install corrected copies of various
37014 header files in a special directory where only GCC will normally
37015 look for them. The scripts adapt to various systems by searching
37016 all the system header files for the problem cases that we know
37019 If new system header files are installed, nothing automatically
37020 arranges to update the corrected header files. They can be
37021 updated using the `mkheaders' script installed in
37022 `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
37024 * On 68000 and x86 systems, for instance, you can get paradoxical
37025 results if you test the precise values of floating point numbers.
37026 For example, you can find that a floating point value which is not
37027 a NaN is not equal to itself. This results from the fact that the
37028 floating point registers hold a few more bits of precision than
37029 fit in a `double' in memory. Compiled code moves values between
37030 memory and floating point registers at its convenience, and moving
37031 them into memory truncates them.
37033 You can partially avoid this problem by using the `-ffloat-store'
37034 option (*note Optimize Options::).
37036 * On AIX and other platforms without weak symbol support, templates
37037 need to be instantiated explicitly and symbols for static members
37038 of templates will not be generated.
37040 * On AIX, GCC scans object files and library archives for static
37041 constructors and destructors when linking an application before the
37042 linker prunes unreferenced symbols. This is necessary to prevent
37043 the AIX linker from mistakenly assuming that static constructor or
37044 destructor are unused and removing them before the scanning can
37045 occur. All static constructors and destructors found will be
37046 referenced even though the modules in which they occur may not be
37047 used by the program. This may lead to both increased executable
37048 size and unexpected symbol references.
37051 File: gcc.info, Node: C++ Misunderstandings, Next: Protoize Caveats, Prev: Disappointments, Up: Trouble
37053 10.8 Common Misunderstandings with GNU C++
37054 ==========================================
37056 C++ is a complex language and an evolving one, and its standard
37057 definition (the ISO C++ standard) was only recently completed. As a
37058 result, your C++ compiler may occasionally surprise you, even when its
37059 behavior is correct. This section discusses some areas that frequently
37060 give rise to questions of this sort.
37064 * Static Definitions:: Static member declarations are not definitions
37065 * Name lookup:: Name lookup, templates, and accessing members of base classes
37066 * Temporaries:: Temporaries may vanish before you expect
37067 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
37070 File: gcc.info, Node: Static Definitions, Next: Name lookup, Up: C++ Misunderstandings
37072 10.8.1 Declare _and_ Define Static Members
37073 ------------------------------------------
37075 When a class has static data members, it is not enough to _declare_ the
37076 static member; you must also _define_ it. For example:
37085 This declaration only establishes that the class `Foo' has an `int'
37086 named `Foo::bar', and a member function named `Foo::method'. But you
37087 still need to define _both_ `method' and `bar' elsewhere. According to
37088 the ISO standard, you must supply an initializer in one (and only one)
37089 source file, such as:
37093 Other C++ compilers may not correctly implement the standard behavior.
37094 As a result, when you switch to `g++' from one of these compilers, you
37095 may discover that a program that appeared to work correctly in fact
37096 does not conform to the standard: `g++' reports as undefined symbols
37097 any static data members that lack definitions.
37100 File: gcc.info, Node: Name lookup, Next: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings
37102 10.8.2 Name lookup, templates, and accessing members of base classes
37103 --------------------------------------------------------------------
37105 The C++ standard prescribes that all names that are not dependent on
37106 template parameters are bound to their present definitions when parsing
37107 a template function or class.(1) Only names that are dependent are
37108 looked up at the point of instantiation. For example, consider
37113 template <typename T>
37122 static const int N;
37125 Here, the names `foo' and `N' appear in a context that does not depend
37126 on the type of `T'. The compiler will thus require that they are
37127 defined in the context of use in the template, not only before the
37128 point of instantiation, and will here use `::foo(double)' and `A::N',
37129 respectively. In particular, it will convert the integer value to a
37130 `double' when passing it to `::foo(double)'.
37132 Conversely, `bar' and the call to `foo' in the fourth marked line are
37133 used in contexts that do depend on the type of `T', so they are only
37134 looked up at the point of instantiation, and you can provide
37135 declarations for them after declaring the template, but before
37136 instantiating it. In particular, if you instantiate `A::f<int>', the
37137 last line will call an overloaded `::foo(int)' if one was provided,
37138 even if after the declaration of `struct A'.
37140 This distinction between lookup of dependent and non-dependent names is
37141 called two-stage (or dependent) name lookup. G++ implements it since
37144 Two-stage name lookup sometimes leads to situations with behavior
37145 different from non-template codes. The most common is probably this:
37147 template <typename T> struct Base {
37151 template <typename T> struct Derived : public Base<T> {
37152 int get_i() { return i; }
37155 In `get_i()', `i' is not used in a dependent context, so the compiler
37156 will look for a name declared at the enclosing namespace scope (which
37157 is the global scope here). It will not look into the base class, since
37158 that is dependent and you may declare specializations of `Base' even
37159 after declaring `Derived', so the compiler can't really know what `i'
37160 would refer to. If there is no global variable `i', then you will get
37163 In order to make it clear that you want the member of the base class,
37164 you need to defer lookup until instantiation time, at which the base
37165 class is known. For this, you need to access `i' in a dependent
37166 context, by either using `this->i' (remember that `this' is of type
37167 `Derived<T>*', so is obviously dependent), or using `Base<T>::i'.
37168 Alternatively, `Base<T>::i' might be brought into scope by a
37169 `using'-declaration.
37171 Another, similar example involves calling member functions of a base
37174 template <typename T> struct Base {
37178 template <typename T> struct Derived : Base<T> {
37179 int g() { return f(); };
37182 Again, the call to `f()' is not dependent on template arguments (there
37183 are no arguments that depend on the type `T', and it is also not
37184 otherwise specified that the call should be in a dependent context).
37185 Thus a global declaration of such a function must be available, since
37186 the one in the base class is not visible until instantiation time. The
37187 compiler will consequently produce the following error message:
37189 x.cc: In member function `int Derived<T>::g()':
37190 x.cc:6: error: there are no arguments to `f' that depend on a template
37191 parameter, so a declaration of `f' must be available
37192 x.cc:6: error: (if you use `-fpermissive', G++ will accept your code, but
37193 allowing the use of an undeclared name is deprecated)
37195 To make the code valid either use `this->f()', or `Base<T>::f()'.
37196 Using the `-fpermissive' flag will also let the compiler accept the
37197 code, by marking all function calls for which no declaration is visible
37198 at the time of definition of the template for later lookup at
37199 instantiation time, as if it were a dependent call. We do not
37200 recommend using `-fpermissive' to work around invalid code, and it will
37201 also only catch cases where functions in base classes are called, not
37202 where variables in base classes are used (as in the example above).
37204 Note that some compilers (including G++ versions prior to 3.4) get
37205 these examples wrong and accept above code without an error. Those
37206 compilers do not implement two-stage name lookup correctly.
37208 ---------- Footnotes ----------
37210 (1) The C++ standard just uses the term "dependent" for names that
37211 depend on the type or value of template parameters. This shorter term
37212 will also be used in the rest of this section.
37215 File: gcc.info, Node: Temporaries, Next: Copy Assignment, Prev: Name lookup, Up: C++ Misunderstandings
37217 10.8.3 Temporaries May Vanish Before You Expect
37218 -----------------------------------------------
37220 It is dangerous to use pointers or references to _portions_ of a
37221 temporary object. The compiler may very well delete the object before
37222 you expect it to, leaving a pointer to garbage. The most common place
37223 where this problem crops up is in classes like string classes,
37224 especially ones that define a conversion function to type `char *' or
37225 `const char *'--which is one reason why the standard `string' class
37226 requires you to call the `c_str' member function. However, any class
37227 that returns a pointer to some internal structure is potentially
37228 subject to this problem.
37230 For example, a program may use a function `strfunc' that returns
37231 `string' objects, and another function `charfunc' that operates on
37232 pointers to `char':
37235 void charfunc (const char *);
37240 const char *p = strfunc().c_str();
37247 In this situation, it may seem reasonable to save a pointer to the C
37248 string returned by the `c_str' member function and use that rather than
37249 call `c_str' repeatedly. However, the temporary string created by the
37250 call to `strfunc' is destroyed after `p' is initialized, at which point
37251 `p' is left pointing to freed memory.
37253 Code like this may run successfully under some other compilers,
37254 particularly obsolete cfront-based compilers that delete temporaries
37255 along with normal local variables. However, the GNU C++ behavior is
37256 standard-conforming, so if your program depends on late destruction of
37257 temporaries it is not portable.
37259 The safe way to write such code is to give the temporary a name, which
37260 forces it to remain until the end of the scope of the name. For
37263 const string& tmp = strfunc ();
37264 charfunc (tmp.c_str ());
37267 File: gcc.info, Node: Copy Assignment, Prev: Temporaries, Up: C++ Misunderstandings
37269 10.8.4 Implicit Copy-Assignment for Virtual Bases
37270 -------------------------------------------------
37272 When a base class is virtual, only one subobject of the base class
37273 belongs to each full object. Also, the constructors and destructors are
37274 invoked only once, and called from the most-derived class. However,
37275 such objects behave unspecified when being assigned. For example:
37279 Base(char *n) : name(strdup(n)){}
37280 Base& operator= (const Base& other){
37282 name = strdup (other.name);
37286 struct A:virtual Base{
37291 struct B:virtual Base{
37296 struct Derived:public A, public B{
37297 Derived():Base("Derived"){}
37300 void func(Derived &d1, Derived &d2)
37305 The C++ standard specifies that `Base::Base' is only called once when
37306 constructing or copy-constructing a Derived object. It is unspecified
37307 whether `Base::operator=' is called more than once when the implicit
37308 copy-assignment for Derived objects is invoked (as it is inside `func'
37311 G++ implements the "intuitive" algorithm for copy-assignment: assign
37312 all direct bases, then assign all members. In that algorithm, the
37313 virtual base subobject can be encountered more than once. In the
37314 example, copying proceeds in the following order: `val', `name' (via
37315 `strdup'), `bval', and `name' again.
37317 If application code relies on copy-assignment, a user-defined
37318 copy-assignment operator removes any uncertainties. With such an
37319 operator, the application can define whether and how the virtual base
37320 subobject is assigned.
37323 File: gcc.info, Node: Protoize Caveats, Next: Non-bugs, Prev: C++ Misunderstandings, Up: Trouble
37325 10.9 Caveats of using `protoize'
37326 ================================
37328 The conversion programs `protoize' and `unprotoize' can sometimes
37329 change a source file in a way that won't work unless you rearrange it.
37331 * `protoize' can insert references to a type name or type tag before
37332 the definition, or in a file where they are not defined.
37334 If this happens, compiler error messages should show you where the
37335 new references are, so fixing the file by hand is straightforward.
37337 * There are some C constructs which `protoize' cannot figure out.
37338 For example, it can't determine argument types for declaring a
37339 pointer-to-function variable; this you must do by hand. `protoize'
37340 inserts a comment containing `???' each time it finds such a
37341 variable; so you can find all such variables by searching for this
37342 string. ISO C does not require declaring the argument types of
37343 pointer-to-function types.
37345 * Using `unprotoize' can easily introduce bugs. If the program
37346 relied on prototypes to bring about conversion of arguments, these
37347 conversions will not take place in the program without prototypes.
37348 One case in which you can be sure `unprotoize' is safe is when you
37349 are removing prototypes that were made with `protoize'; if the
37350 program worked before without any prototypes, it will work again
37353 You can find all the places where this problem might occur by
37354 compiling the program with the `-Wtraditional-conversion' option.
37355 It prints a warning whenever an argument is converted.
37357 * Both conversion programs can be confused if there are macro calls
37358 in and around the text to be converted. In other words, the
37359 standard syntax for a declaration or definition must not result
37360 from expanding a macro. This problem is inherent in the design of
37361 C and cannot be fixed. If only a few functions have confusing
37362 macro calls, you can easily convert them manually.
37364 * `protoize' cannot get the argument types for a function whose
37365 definition was not actually compiled due to preprocessing
37366 conditionals. When this happens, `protoize' changes nothing in
37367 regard to such a function. `protoize' tries to detect such
37368 instances and warn about them.
37370 You can generally work around this problem by using `protoize' step
37371 by step, each time specifying a different set of `-D' options for
37372 compilation, until all of the functions have been converted.
37373 There is no automatic way to verify that you have got them all,
37376 * Confusion may result if there is an occasion to convert a function
37377 declaration or definition in a region of source code where there
37378 is more than one formal parameter list present. Thus, attempts to
37379 convert code containing multiple (conditionally compiled) versions
37380 of a single function header (in the same vicinity) may not produce
37381 the desired (or expected) results.
37383 If you plan on converting source files which contain such code, it
37384 is recommended that you first make sure that each conditionally
37385 compiled region of source code which contains an alternative
37386 function header also contains at least one additional follower
37387 token (past the final right parenthesis of the function header).
37388 This should circumvent the problem.
37390 * `unprotoize' can become confused when trying to convert a function
37391 definition or declaration which contains a declaration for a
37392 pointer-to-function formal argument which has the same name as the
37393 function being defined or declared. We recommend you avoid such
37394 choices of formal parameter names.
37396 * You might also want to correct some of the indentation by hand and
37397 break long lines. (The conversion programs don't write lines
37398 longer than eighty characters in any case.)
37401 File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: Protoize Caveats, Up: Trouble
37403 10.10 Certain Changes We Don't Want to Make
37404 ===========================================
37406 This section lists changes that people frequently request, but which we
37407 do not make because we think GCC is better without them.
37409 * Checking the number and type of arguments to a function which has
37410 an old-fashioned definition and no prototype.
37412 Such a feature would work only occasionally--only for calls that
37413 appear in the same file as the called function, following the
37414 definition. The only way to check all calls reliably is to add a
37415 prototype for the function. But adding a prototype eliminates the
37416 motivation for this feature. So the feature is not worthwhile.
37418 * Warning about using an expression whose type is signed as a shift
37421 Shift count operands are probably signed more often than unsigned.
37422 Warning about this would cause far more annoyance than good.
37424 * Warning about assigning a signed value to an unsigned variable.
37426 Such assignments must be very common; warning about them would
37427 cause more annoyance than good.
37429 * Warning when a non-void function value is ignored.
37431 C contains many standard functions that return a value that most
37432 programs choose to ignore. One obvious example is `printf'.
37433 Warning about this practice only leads the defensive programmer to
37434 clutter programs with dozens of casts to `void'. Such casts are
37435 required so frequently that they become visual noise. Writing
37436 those casts becomes so automatic that they no longer convey useful
37437 information about the intentions of the programmer. For functions
37438 where the return value should never be ignored, use the
37439 `warn_unused_result' function attribute (*note Function
37442 * Making `-fshort-enums' the default.
37444 This would cause storage layout to be incompatible with most other
37445 C compilers. And it doesn't seem very important, given that you
37446 can get the same result in other ways. The case where it matters
37447 most is when the enumeration-valued object is inside a structure,
37448 and in that case you can specify a field width explicitly.
37450 * Making bit-fields unsigned by default on particular machines where
37451 "the ABI standard" says to do so.
37453 The ISO C standard leaves it up to the implementation whether a
37454 bit-field declared plain `int' is signed or not. This in effect
37455 creates two alternative dialects of C.
37457 The GNU C compiler supports both dialects; you can specify the
37458 signed dialect with `-fsigned-bitfields' and the unsigned dialect
37459 with `-funsigned-bitfields'. However, this leaves open the
37460 question of which dialect to use by default.
37462 Currently, the preferred dialect makes plain bit-fields signed,
37463 because this is simplest. Since `int' is the same as `signed int'
37464 in every other context, it is cleanest for them to be the same in
37465 bit-fields as well.
37467 Some computer manufacturers have published Application Binary
37468 Interface standards which specify that plain bit-fields should be
37469 unsigned. It is a mistake, however, to say anything about this
37470 issue in an ABI. This is because the handling of plain bit-fields
37471 distinguishes two dialects of C. Both dialects are meaningful on
37472 every type of machine. Whether a particular object file was
37473 compiled using signed bit-fields or unsigned is of no concern to
37474 other object files, even if they access the same bit-fields in the
37475 same data structures.
37477 A given program is written in one or the other of these two
37478 dialects. The program stands a chance to work on most any machine
37479 if it is compiled with the proper dialect. It is unlikely to work
37480 at all if compiled with the wrong dialect.
37482 Many users appreciate the GNU C compiler because it provides an
37483 environment that is uniform across machines. These users would be
37484 inconvenienced if the compiler treated plain bit-fields
37485 differently on certain machines.
37487 Occasionally users write programs intended only for a particular
37488 machine type. On these occasions, the users would benefit if the
37489 GNU C compiler were to support by default the same dialect as the
37490 other compilers on that machine. But such applications are rare.
37491 And users writing a program to run on more than one type of
37492 machine cannot possibly benefit from this kind of compatibility.
37494 This is why GCC does and will treat plain bit-fields in the same
37495 fashion on all types of machines (by default).
37497 There are some arguments for making bit-fields unsigned by default
37498 on all machines. If, for example, this becomes a universal de
37499 facto standard, it would make sense for GCC to go along with it.
37500 This is something to be considered in the future.
37502 (Of course, users strongly concerned about portability should
37503 indicate explicitly in each bit-field whether it is signed or not.
37504 In this way, they write programs which have the same meaning in
37507 * Undefining `__STDC__' when `-ansi' is not used.
37509 Currently, GCC defines `__STDC__' unconditionally. This provides
37510 good results in practice.
37512 Programmers normally use conditionals on `__STDC__' to ask whether
37513 it is safe to use certain features of ISO C, such as function
37514 prototypes or ISO token concatenation. Since plain `gcc' supports
37515 all the features of ISO C, the correct answer to these questions is
37518 Some users try to use `__STDC__' to check for the availability of
37519 certain library facilities. This is actually incorrect usage in
37520 an ISO C program, because the ISO C standard says that a conforming
37521 freestanding implementation should define `__STDC__' even though it
37522 does not have the library facilities. `gcc -ansi -pedantic' is a
37523 conforming freestanding implementation, and it is therefore
37524 required to define `__STDC__', even though it does not come with
37527 Sometimes people say that defining `__STDC__' in a compiler that
37528 does not completely conform to the ISO C standard somehow violates
37529 the standard. This is illogical. The standard is a standard for
37530 compilers that claim to support ISO C, such as `gcc -ansi'--not
37531 for other compilers such as plain `gcc'. Whatever the ISO C
37532 standard says is relevant to the design of plain `gcc' without
37533 `-ansi' only for pragmatic reasons, not as a requirement.
37535 GCC normally defines `__STDC__' to be 1, and in addition defines
37536 `__STRICT_ANSI__' if you specify the `-ansi' option, or a `-std'
37537 option for strict conformance to some version of ISO C. On some
37538 hosts, system include files use a different convention, where
37539 `__STDC__' is normally 0, but is 1 if the user specifies strict
37540 conformance to the C Standard. GCC follows the host convention
37541 when processing system include files, but when processing user
37542 files it follows the usual GNU C convention.
37544 * Undefining `__STDC__' in C++.
37546 Programs written to compile with C++-to-C translators get the
37547 value of `__STDC__' that goes with the C compiler that is
37548 subsequently used. These programs must test `__STDC__' to
37549 determine what kind of C preprocessor that compiler uses: whether
37550 they should concatenate tokens in the ISO C fashion or in the
37551 traditional fashion.
37553 These programs work properly with GNU C++ if `__STDC__' is defined.
37554 They would not work otherwise.
37556 In addition, many header files are written to provide prototypes
37557 in ISO C but not in traditional C. Many of these header files can
37558 work without change in C++ provided `__STDC__' is defined. If
37559 `__STDC__' is not defined, they will all fail, and will all need
37560 to be changed to test explicitly for C++ as well.
37562 * Deleting "empty" loops.
37564 Historically, GCC has not deleted "empty" loops under the
37565 assumption that the most likely reason you would put one in a
37566 program is to have a delay, so deleting them will not make real
37567 programs run any faster.
37569 However, the rationale here is that optimization of a nonempty loop
37570 cannot produce an empty one. This held for carefully written C
37571 compiled with less powerful optimizers but is not always the case
37572 for carefully written C++ or with more powerful optimizers. Thus
37573 GCC will remove operations from loops whenever it can determine
37574 those operations are not externally visible (apart from the time
37575 taken to execute them, of course). In case the loop can be proved
37576 to be finite, GCC will also remove the loop itself.
37578 Be aware of this when performing timing tests, for instance the
37579 following loop can be completely removed, provided
37580 `some_expression' can provably not change any global state.
37586 for (ix = 0; ix != 10000; ix++)
37587 sum += some_expression;
37590 Even though `sum' is accumulated in the loop, no use is made of
37591 that summation, so the accumulation can be removed.
37593 * Making side effects happen in the same order as in some other
37596 It is never safe to depend on the order of evaluation of side
37597 effects. For example, a function call like this may very well
37598 behave differently from one compiler to another:
37600 void func (int, int);
37605 There is no guarantee (in either the C or the C++ standard language
37606 definitions) that the increments will be evaluated in any
37607 particular order. Either increment might happen first. `func'
37608 might get the arguments `2, 3', or it might get `3, 2', or even
37611 * Making certain warnings into errors by default.
37613 Some ISO C testsuites report failure when the compiler does not
37614 produce an error message for a certain program.
37616 ISO C requires a "diagnostic" message for certain kinds of invalid
37617 programs, but a warning is defined by GCC to count as a
37618 diagnostic. If GCC produces a warning but not an error, that is
37619 correct ISO C support. If testsuites call this "failure", they
37620 should be run with the GCC option `-pedantic-errors', which will
37621 turn these warnings into errors.
37625 File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble
37627 10.11 Warning Messages and Error Messages
37628 =========================================
37630 The GNU compiler can produce two kinds of diagnostics: errors and
37631 warnings. Each kind has a different purpose:
37633 "Errors" report problems that make it impossible to compile your
37634 program. GCC reports errors with the source file name and line
37635 number where the problem is apparent.
37637 "Warnings" report other unusual conditions in your code that _may_
37638 indicate a problem, although compilation can (and does) proceed.
37639 Warning messages also report the source file name and line number,
37640 but include the text `warning:' to distinguish them from error
37643 Warnings may indicate danger points where you should check to make sure
37644 that your program really does what you intend; or the use of obsolete
37645 features; or the use of nonstandard features of GNU C or C++. Many
37646 warnings are issued only if you ask for them, with one of the `-W'
37647 options (for instance, `-Wall' requests a variety of useful warnings).
37649 GCC always tries to compile your program if possible; it never
37650 gratuitously rejects a program whose meaning is clear merely because
37651 (for instance) it fails to conform to a standard. In some cases,
37652 however, the C and C++ standards specify that certain extensions are
37653 forbidden, and a diagnostic _must_ be issued by a conforming compiler.
37654 The `-pedantic' option tells GCC to issue warnings in such cases;
37655 `-pedantic-errors' says to make them errors instead. This does not
37656 mean that _all_ non-ISO constructs get warnings or errors.
37658 *Note Options to Request or Suppress Warnings: Warning Options, for
37659 more detail on these and related command-line options.
37662 File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top
37667 Your bug reports play an essential role in making GCC reliable.
37669 When you encounter a problem, the first thing to do is to see if it is
37670 already known. *Note Trouble::. If it isn't known, then you should
37671 report the problem.
37675 * Criteria: Bug Criteria. Have you really found a bug?
37676 * Reporting: Bug Reporting. How to report a bug effectively.
37677 * Known: Trouble. Known problems.
37678 * Help: Service. Where to ask for help.
37681 File: gcc.info, Node: Bug Criteria, Next: Bug Reporting, Up: Bugs
37683 11.1 Have You Found a Bug?
37684 ==========================
37686 If you are not sure whether you have found a bug, here are some
37689 * If the compiler gets a fatal signal, for any input whatever, that
37690 is a compiler bug. Reliable compilers never crash.
37692 * If the compiler produces invalid assembly code, for any input
37693 whatever (except an `asm' statement), that is a compiler bug,
37694 unless the compiler reports errors (not just warnings) which would
37695 ordinarily prevent the assembler from being run.
37697 * If the compiler produces valid assembly code that does not
37698 correctly execute the input source code, that is a compiler bug.
37700 However, you must double-check to make sure, because you may have a
37701 program whose behavior is undefined, which happened by chance to
37702 give the desired results with another C or C++ compiler.
37704 For example, in many nonoptimizing compilers, you can write `x;'
37705 at the end of a function instead of `return x;', with the same
37706 results. But the value of the function is undefined if `return'
37707 is omitted; it is not a bug when GCC produces different results.
37709 Problems often result from expressions with two increment
37710 operators, as in `f (*p++, *p++)'. Your previous compiler might
37711 have interpreted that expression the way you intended; GCC might
37712 interpret it another way. Neither compiler is wrong. The bug is
37715 After you have localized the error to a single source line, it
37716 should be easy to check for these things. If your program is
37717 correct and well defined, you have found a compiler bug.
37719 * If the compiler produces an error message for valid input, that is
37722 * If the compiler does not produce an error message for invalid
37723 input, that is a compiler bug. However, you should note that your
37724 idea of "invalid input" might be someone else's idea of "an
37725 extension" or "support for traditional practice".
37727 * If you are an experienced user of one of the languages GCC
37728 supports, your suggestions for improvement of GCC are welcome in
37732 File: gcc.info, Node: Bug Reporting, Prev: Bug Criteria, Up: Bugs
37734 11.2 How and where to Report Bugs
37735 =================================
37737 Bugs should be reported to the bug database at
37738 `http://gcc.gnu.org/bugs.html'.
37741 File: gcc.info, Node: Service, Next: Contributing, Prev: Bugs, Up: Top
37743 12 How To Get Help with GCC
37744 ***************************
37746 If you need help installing, using or changing GCC, there are two ways
37749 * Send a message to a suitable network mailing list. First try
37750 <gcc-help@gcc.gnu.org> (for help installing or using GCC), and if
37751 that brings no response, try <gcc@gcc.gnu.org>. For help changing
37752 GCC, ask <gcc@gcc.gnu.org>. If you think you have found a bug in
37753 GCC, please report it following the instructions at *note Bug
37756 * Look in the service directory for someone who might help you for a
37757 fee. The service directory is found at
37758 `http://www.gnu.org/prep/service.html'.
37760 For further information, see `http://gcc.gnu.org/faq.html#support'.
37763 File: gcc.info, Node: Contributing, Next: Funding, Prev: Service, Up: Top
37765 13 Contributing to GCC Development
37766 **********************************
37768 If you would like to help pretest GCC releases to assure they work well,
37769 current development sources are available by SVN (see
37770 `http://gcc.gnu.org/svn.html'). Source and binary snapshots are also
37771 available for FTP; see `http://gcc.gnu.org/snapshots.html'.
37773 If you would like to work on improvements to GCC, please read the
37774 advice at these URLs:
37776 `http://gcc.gnu.org/contribute.html'
37777 `http://gcc.gnu.org/contributewhy.html'
37779 for information on how to make useful contributions and avoid
37780 duplication of effort. Suggested projects are listed at
37781 `http://gcc.gnu.org/projects/'.
37784 File: gcc.info, Node: Funding, Next: GNU Project, Prev: Contributing, Up: Top
37786 Funding Free Software
37787 *********************
37789 If you want to have more free software a few years from now, it makes
37790 sense for you to help encourage people to contribute funds for its
37791 development. The most effective approach known is to encourage
37792 commercial redistributors to donate.
37794 Users of free software systems can boost the pace of development by
37795 encouraging for-a-fee distributors to donate part of their selling price
37796 to free software developers--the Free Software Foundation, and others.
37798 The way to convince distributors to do this is to demand it and expect
37799 it from them. So when you compare distributors, judge them partly by
37800 how much they give to free software development. Show distributors
37801 they must compete to be the one who gives the most.
37803 To make this approach work, you must insist on numbers that you can
37804 compare, such as, "We will donate ten dollars to the Frobnitz project
37805 for each disk sold." Don't be satisfied with a vague promise, such as
37806 "A portion of the profits are donated," since it doesn't give a basis
37809 Even a precise fraction "of the profits from this disk" is not very
37810 meaningful, since creative accounting and unrelated business decisions
37811 can greatly alter what fraction of the sales price counts as profit.
37812 If the price you pay is $50, ten percent of the profit is probably less
37813 than a dollar; it might be a few cents, or nothing at all.
37815 Some redistributors do development work themselves. This is useful
37816 too; but to keep everyone honest, you need to inquire how much they do,
37817 and what kind. Some kinds of development make much more long-term
37818 difference than others. For example, maintaining a separate version of
37819 a program contributes very little; maintaining the standard version of a
37820 program for the whole community contributes much. Easy new ports
37821 contribute little, since someone else would surely do them; difficult
37822 ports such as adding a new CPU to the GNU Compiler Collection
37823 contribute more; major new features or packages contribute the most.
37825 By establishing the idea that supporting further development is "the
37826 proper thing to do" when distributing free software for a fee, we can
37827 assure a steady flow of resources into making more free software.
37829 Copyright (C) 1994 Free Software Foundation, Inc.
37830 Verbatim copying and redistribution of this section is permitted
37831 without royalty; alteration is not permitted.
37834 File: gcc.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top
37836 The GNU Project and GNU/Linux
37837 *****************************
37839 The GNU Project was launched in 1984 to develop a complete Unix-like
37840 operating system which is free software: the GNU system. (GNU is a
37841 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".)
37842 Variants of the GNU operating system, which use the kernel Linux, are
37843 now widely used; though these systems are often referred to as "Linux",
37844 they are more accurately called GNU/Linux systems.
37846 For more information, see:
37847 `http://www.gnu.org/'
37848 `http://www.gnu.org/gnu/linux-and-gnu.html'
37851 File: gcc.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top
37853 GNU General Public License
37854 **************************
37856 Version 3, 29 June 2007
37858 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/'
37860 Everyone is permitted to copy and distribute verbatim copies of this
37861 license document, but changing it is not allowed.
37866 The GNU General Public License is a free, copyleft license for software
37867 and other kinds of works.
37869 The licenses for most software and other practical works are designed
37870 to take away your freedom to share and change the works. By contrast,
37871 the GNU General Public License is intended to guarantee your freedom to
37872 share and change all versions of a program-to make sure it remains free
37873 software for all its users. We, the Free Software Foundation, use the
37874 GNU General Public License for most of our software; it applies also to
37875 any other work released this way by its authors. You can apply it to
37876 your programs, too.
37878 When we speak of free software, we are referring to freedom, not
37879 price. Our General Public Licenses are designed to make sure that you
37880 have the freedom to distribute copies of free software (and charge for
37881 them if you wish), that you receive source code or can get it if you
37882 want it, that you can change the software or use pieces of it in new
37883 free programs, and that you know you can do these things.
37885 To protect your rights, we need to prevent others from denying you
37886 these rights or asking you to surrender the rights. Therefore, you
37887 have certain responsibilities if you distribute copies of the software,
37888 or if you modify it: responsibilities to respect the freedom of others.
37890 For example, if you distribute copies of such a program, whether
37891 gratis or for a fee, you must pass on to the recipients the same
37892 freedoms that you received. You must make sure that they, too, receive
37893 or can get the source code. And you must show them these terms so they
37896 Developers that use the GNU GPL protect your rights with two steps:
37897 (1) assert copyright on the software, and (2) offer you this License
37898 giving you legal permission to copy, distribute and/or modify it.
37900 For the developers' and authors' protection, the GPL clearly explains
37901 that there is no warranty for this free software. For both users' and
37902 authors' sake, the GPL requires that modified versions be marked as
37903 changed, so that their problems will not be attributed erroneously to
37904 authors of previous versions.
37906 Some devices are designed to deny users access to install or run
37907 modified versions of the software inside them, although the
37908 manufacturer can do so. This is fundamentally incompatible with the
37909 aim of protecting users' freedom to change the software. The
37910 systematic pattern of such abuse occurs in the area of products for
37911 individuals to use, which is precisely where it is most unacceptable.
37912 Therefore, we have designed this version of the GPL to prohibit the
37913 practice for those products. If such problems arise substantially in
37914 other domains, we stand ready to extend this provision to those domains
37915 in future versions of the GPL, as needed to protect the freedom of
37918 Finally, every program is threatened constantly by software patents.
37919 States should not allow patents to restrict development and use of
37920 software on general-purpose computers, but in those that do, we wish to
37921 avoid the special danger that patents applied to a free program could
37922 make it effectively proprietary. To prevent this, the GPL assures that
37923 patents cannot be used to render the program non-free.
37925 The precise terms and conditions for copying, distribution and
37926 modification follow.
37928 TERMS AND CONDITIONS
37929 ====================
37933 "This License" refers to version 3 of the GNU General Public
37936 "Copyright" also means copyright-like laws that apply to other
37937 kinds of works, such as semiconductor masks.
37939 "The Program" refers to any copyrightable work licensed under this
37940 License. Each licensee is addressed as "you". "Licensees" and
37941 "recipients" may be individuals or organizations.
37943 To "modify" a work means to copy from or adapt all or part of the
37944 work in a fashion requiring copyright permission, other than the
37945 making of an exact copy. The resulting work is called a "modified
37946 version" of the earlier work or a work "based on" the earlier work.
37948 A "covered work" means either the unmodified Program or a work
37949 based on the Program.
37951 To "propagate" a work means to do anything with it that, without
37952 permission, would make you directly or secondarily liable for
37953 infringement under applicable copyright law, except executing it
37954 on a computer or modifying a private copy. Propagation includes
37955 copying, distribution (with or without modification), making
37956 available to the public, and in some countries other activities as
37959 To "convey" a work means any kind of propagation that enables other
37960 parties to make or receive copies. Mere interaction with a user
37961 through a computer network, with no transfer of a copy, is not
37964 An interactive user interface displays "Appropriate Legal Notices"
37965 to the extent that it includes a convenient and prominently visible
37966 feature that (1) displays an appropriate copyright notice, and (2)
37967 tells the user that there is no warranty for the work (except to
37968 the extent that warranties are provided), that licensees may
37969 convey the work under this License, and how to view a copy of this
37970 License. If the interface presents a list of user commands or
37971 options, such as a menu, a prominent item in the list meets this
37976 The "source code" for a work means the preferred form of the work
37977 for making modifications to it. "Object code" means any
37978 non-source form of a work.
37980 A "Standard Interface" means an interface that either is an
37981 official standard defined by a recognized standards body, or, in
37982 the case of interfaces specified for a particular programming
37983 language, one that is widely used among developers working in that
37986 The "System Libraries" of an executable work include anything,
37987 other than the work as a whole, that (a) is included in the normal
37988 form of packaging a Major Component, but which is not part of that
37989 Major Component, and (b) serves only to enable use of the work
37990 with that Major Component, or to implement a Standard Interface
37991 for which an implementation is available to the public in source
37992 code form. A "Major Component", in this context, means a major
37993 essential component (kernel, window system, and so on) of the
37994 specific operating system (if any) on which the executable work
37995 runs, or a compiler used to produce the work, or an object code
37996 interpreter used to run it.
37998 The "Corresponding Source" for a work in object code form means all
37999 the source code needed to generate, install, and (for an executable
38000 work) run the object code and to modify the work, including
38001 scripts to control those activities. However, it does not include
38002 the work's System Libraries, or general-purpose tools or generally
38003 available free programs which are used unmodified in performing
38004 those activities but which are not part of the work. For example,
38005 Corresponding Source includes interface definition files
38006 associated with source files for the work, and the source code for
38007 shared libraries and dynamically linked subprograms that the work
38008 is specifically designed to require, such as by intimate data
38009 communication or control flow between those subprograms and other
38012 The Corresponding Source need not include anything that users can
38013 regenerate automatically from other parts of the Corresponding
38016 The Corresponding Source for a work in source code form is that
38019 2. Basic Permissions.
38021 All rights granted under this License are granted for the term of
38022 copyright on the Program, and are irrevocable provided the stated
38023 conditions are met. This License explicitly affirms your unlimited
38024 permission to run the unmodified Program. The output from running
38025 a covered work is covered by this License only if the output,
38026 given its content, constitutes a covered work. This License
38027 acknowledges your rights of fair use or other equivalent, as
38028 provided by copyright law.
38030 You may make, run and propagate covered works that you do not
38031 convey, without conditions so long as your license otherwise
38032 remains in force. You may convey covered works to others for the
38033 sole purpose of having them make modifications exclusively for
38034 you, or provide you with facilities for running those works,
38035 provided that you comply with the terms of this License in
38036 conveying all material for which you do not control copyright.
38037 Those thus making or running the covered works for you must do so
38038 exclusively on your behalf, under your direction and control, on
38039 terms that prohibit them from making any copies of your
38040 copyrighted material outside their relationship with you.
38042 Conveying under any other circumstances is permitted solely under
38043 the conditions stated below. Sublicensing is not allowed; section
38044 10 makes it unnecessary.
38046 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
38048 No covered work shall be deemed part of an effective technological
38049 measure under any applicable law fulfilling obligations under
38050 article 11 of the WIPO copyright treaty adopted on 20 December
38051 1996, or similar laws prohibiting or restricting circumvention of
38054 When you convey a covered work, you waive any legal power to forbid
38055 circumvention of technological measures to the extent such
38056 circumvention is effected by exercising rights under this License
38057 with respect to the covered work, and you disclaim any intention
38058 to limit operation or modification of the work as a means of
38059 enforcing, against the work's users, your or third parties' legal
38060 rights to forbid circumvention of technological measures.
38062 4. Conveying Verbatim Copies.
38064 You may convey verbatim copies of the Program's source code as you
38065 receive it, in any medium, provided that you conspicuously and
38066 appropriately publish on each copy an appropriate copyright notice;
38067 keep intact all notices stating that this License and any
38068 non-permissive terms added in accord with section 7 apply to the
38069 code; keep intact all notices of the absence of any warranty; and
38070 give all recipients a copy of this License along with the Program.
38072 You may charge any price or no price for each copy that you convey,
38073 and you may offer support or warranty protection for a fee.
38075 5. Conveying Modified Source Versions.
38077 You may convey a work based on the Program, or the modifications to
38078 produce it from the Program, in the form of source code under the
38079 terms of section 4, provided that you also meet all of these
38082 a. The work must carry prominent notices stating that you
38083 modified it, and giving a relevant date.
38085 b. The work must carry prominent notices stating that it is
38086 released under this License and any conditions added under
38087 section 7. This requirement modifies the requirement in
38088 section 4 to "keep intact all notices".
38090 c. You must license the entire work, as a whole, under this
38091 License to anyone who comes into possession of a copy. This
38092 License will therefore apply, along with any applicable
38093 section 7 additional terms, to the whole of the work, and all
38094 its parts, regardless of how they are packaged. This License
38095 gives no permission to license the work in any other way, but
38096 it does not invalidate such permission if you have separately
38099 d. If the work has interactive user interfaces, each must display
38100 Appropriate Legal Notices; however, if the Program has
38101 interactive interfaces that do not display Appropriate Legal
38102 Notices, your work need not make them do so.
38104 A compilation of a covered work with other separate and independent
38105 works, which are not by their nature extensions of the covered
38106 work, and which are not combined with it such as to form a larger
38107 program, in or on a volume of a storage or distribution medium, is
38108 called an "aggregate" if the compilation and its resulting
38109 copyright are not used to limit the access or legal rights of the
38110 compilation's users beyond what the individual works permit.
38111 Inclusion of a covered work in an aggregate does not cause this
38112 License to apply to the other parts of the aggregate.
38114 6. Conveying Non-Source Forms.
38116 You may convey a covered work in object code form under the terms
38117 of sections 4 and 5, provided that you also convey the
38118 machine-readable Corresponding Source under the terms of this
38119 License, in one of these ways:
38121 a. Convey the object code in, or embodied in, a physical product
38122 (including a physical distribution medium), accompanied by the
38123 Corresponding Source fixed on a durable physical medium
38124 customarily used for software interchange.
38126 b. Convey the object code in, or embodied in, a physical product
38127 (including a physical distribution medium), accompanied by a
38128 written offer, valid for at least three years and valid for
38129 as long as you offer spare parts or customer support for that
38130 product model, to give anyone who possesses the object code
38131 either (1) a copy of the Corresponding Source for all the
38132 software in the product that is covered by this License, on a
38133 durable physical medium customarily used for software
38134 interchange, for a price no more than your reasonable cost of
38135 physically performing this conveying of source, or (2) access
38136 to copy the Corresponding Source from a network server at no
38139 c. Convey individual copies of the object code with a copy of
38140 the written offer to provide the Corresponding Source. This
38141 alternative is allowed only occasionally and noncommercially,
38142 and only if you received the object code with such an offer,
38143 in accord with subsection 6b.
38145 d. Convey the object code by offering access from a designated
38146 place (gratis or for a charge), and offer equivalent access
38147 to the Corresponding Source in the same way through the same
38148 place at no further charge. You need not require recipients
38149 to copy the Corresponding Source along with the object code.
38150 If the place to copy the object code is a network server, the
38151 Corresponding Source may be on a different server (operated
38152 by you or a third party) that supports equivalent copying
38153 facilities, provided you maintain clear directions next to
38154 the object code saying where to find the Corresponding Source.
38155 Regardless of what server hosts the Corresponding Source, you
38156 remain obligated to ensure that it is available for as long
38157 as needed to satisfy these requirements.
38159 e. Convey the object code using peer-to-peer transmission,
38160 provided you inform other peers where the object code and
38161 Corresponding Source of the work are being offered to the
38162 general public at no charge under subsection 6d.
38165 A separable portion of the object code, whose source code is
38166 excluded from the Corresponding Source as a System Library, need
38167 not be included in conveying the object code work.
38169 A "User Product" is either (1) a "consumer product", which means
38170 any tangible personal property which is normally used for personal,
38171 family, or household purposes, or (2) anything designed or sold for
38172 incorporation into a dwelling. In determining whether a product
38173 is a consumer product, doubtful cases shall be resolved in favor of
38174 coverage. For a particular product received by a particular user,
38175 "normally used" refers to a typical or common use of that class of
38176 product, regardless of the status of the particular user or of the
38177 way in which the particular user actually uses, or expects or is
38178 expected to use, the product. A product is a consumer product
38179 regardless of whether the product has substantial commercial,
38180 industrial or non-consumer uses, unless such uses represent the
38181 only significant mode of use of the product.
38183 "Installation Information" for a User Product means any methods,
38184 procedures, authorization keys, or other information required to
38185 install and execute modified versions of a covered work in that
38186 User Product from a modified version of its Corresponding Source.
38187 The information must suffice to ensure that the continued
38188 functioning of the modified object code is in no case prevented or
38189 interfered with solely because modification has been made.
38191 If you convey an object code work under this section in, or with,
38192 or specifically for use in, a User Product, and the conveying
38193 occurs as part of a transaction in which the right of possession
38194 and use of the User Product is transferred to the recipient in
38195 perpetuity or for a fixed term (regardless of how the transaction
38196 is characterized), the Corresponding Source conveyed under this
38197 section must be accompanied by the Installation Information. But
38198 this requirement does not apply if neither you nor any third party
38199 retains the ability to install modified object code on the User
38200 Product (for example, the work has been installed in ROM).
38202 The requirement to provide Installation Information does not
38203 include a requirement to continue to provide support service,
38204 warranty, or updates for a work that has been modified or
38205 installed by the recipient, or for the User Product in which it
38206 has been modified or installed. Access to a network may be denied
38207 when the modification itself materially and adversely affects the
38208 operation of the network or violates the rules and protocols for
38209 communication across the network.
38211 Corresponding Source conveyed, and Installation Information
38212 provided, in accord with this section must be in a format that is
38213 publicly documented (and with an implementation available to the
38214 public in source code form), and must require no special password
38215 or key for unpacking, reading or copying.
38217 7. Additional Terms.
38219 "Additional permissions" are terms that supplement the terms of
38220 this License by making exceptions from one or more of its
38221 conditions. Additional permissions that are applicable to the
38222 entire Program shall be treated as though they were included in
38223 this License, to the extent that they are valid under applicable
38224 law. If additional permissions apply only to part of the Program,
38225 that part may be used separately under those permissions, but the
38226 entire Program remains governed by this License without regard to
38227 the additional permissions.
38229 When you convey a copy of a covered work, you may at your option
38230 remove any additional permissions from that copy, or from any part
38231 of it. (Additional permissions may be written to require their own
38232 removal in certain cases when you modify the work.) You may place
38233 additional permissions on material, added by you to a covered work,
38234 for which you have or can give appropriate copyright permission.
38236 Notwithstanding any other provision of this License, for material
38237 you add to a covered work, you may (if authorized by the copyright
38238 holders of that material) supplement the terms of this License
38241 a. Disclaiming warranty or limiting liability differently from
38242 the terms of sections 15 and 16 of this License; or
38244 b. Requiring preservation of specified reasonable legal notices
38245 or author attributions in that material or in the Appropriate
38246 Legal Notices displayed by works containing it; or
38248 c. Prohibiting misrepresentation of the origin of that material,
38249 or requiring that modified versions of such material be
38250 marked in reasonable ways as different from the original
38253 d. Limiting the use for publicity purposes of names of licensors
38254 or authors of the material; or
38256 e. Declining to grant rights under trademark law for use of some
38257 trade names, trademarks, or service marks; or
38259 f. Requiring indemnification of licensors and authors of that
38260 material by anyone who conveys the material (or modified
38261 versions of it) with contractual assumptions of liability to
38262 the recipient, for any liability that these contractual
38263 assumptions directly impose on those licensors and authors.
38265 All other non-permissive additional terms are considered "further
38266 restrictions" within the meaning of section 10. If the Program as
38267 you received it, or any part of it, contains a notice stating that
38268 it is governed by this License along with a term that is a further
38269 restriction, you may remove that term. If a license document
38270 contains a further restriction but permits relicensing or
38271 conveying under this License, you may add to a covered work
38272 material governed by the terms of that license document, provided
38273 that the further restriction does not survive such relicensing or
38276 If you add terms to a covered work in accord with this section, you
38277 must place, in the relevant source files, a statement of the
38278 additional terms that apply to those files, or a notice indicating
38279 where to find the applicable terms.
38281 Additional terms, permissive or non-permissive, may be stated in
38282 the form of a separately written license, or stated as exceptions;
38283 the above requirements apply either way.
38287 You may not propagate or modify a covered work except as expressly
38288 provided under this License. Any attempt otherwise to propagate or
38289 modify it is void, and will automatically terminate your rights
38290 under this License (including any patent licenses granted under
38291 the third paragraph of section 11).
38293 However, if you cease all violation of this License, then your
38294 license from a particular copyright holder is reinstated (a)
38295 provisionally, unless and until the copyright holder explicitly
38296 and finally terminates your license, and (b) permanently, if the
38297 copyright holder fails to notify you of the violation by some
38298 reasonable means prior to 60 days after the cessation.
38300 Moreover, your license from a particular copyright holder is
38301 reinstated permanently if the copyright holder notifies you of the
38302 violation by some reasonable means, this is the first time you have
38303 received notice of violation of this License (for any work) from
38304 that copyright holder, and you cure the violation prior to 30 days
38305 after your receipt of the notice.
38307 Termination of your rights under this section does not terminate
38308 the licenses of parties who have received copies or rights from
38309 you under this License. If your rights have been terminated and
38310 not permanently reinstated, you do not qualify to receive new
38311 licenses for the same material under section 10.
38313 9. Acceptance Not Required for Having Copies.
38315 You are not required to accept this License in order to receive or
38316 run a copy of the Program. Ancillary propagation of a covered work
38317 occurring solely as a consequence of using peer-to-peer
38318 transmission to receive a copy likewise does not require
38319 acceptance. However, nothing other than this License grants you
38320 permission to propagate or modify any covered work. These actions
38321 infringe copyright if you do not accept this License. Therefore,
38322 by modifying or propagating a covered work, you indicate your
38323 acceptance of this License to do so.
38325 10. Automatic Licensing of Downstream Recipients.
38327 Each time you convey a covered work, the recipient automatically
38328 receives a license from the original licensors, to run, modify and
38329 propagate that work, subject to this License. You are not
38330 responsible for enforcing compliance by third parties with this
38333 An "entity transaction" is a transaction transferring control of an
38334 organization, or substantially all assets of one, or subdividing an
38335 organization, or merging organizations. If propagation of a
38336 covered work results from an entity transaction, each party to that
38337 transaction who receives a copy of the work also receives whatever
38338 licenses to the work the party's predecessor in interest had or
38339 could give under the previous paragraph, plus a right to
38340 possession of the Corresponding Source of the work from the
38341 predecessor in interest, if the predecessor has it or can get it
38342 with reasonable efforts.
38344 You may not impose any further restrictions on the exercise of the
38345 rights granted or affirmed under this License. For example, you
38346 may not impose a license fee, royalty, or other charge for
38347 exercise of rights granted under this License, and you may not
38348 initiate litigation (including a cross-claim or counterclaim in a
38349 lawsuit) alleging that any patent claim is infringed by making,
38350 using, selling, offering for sale, or importing the Program or any
38355 A "contributor" is a copyright holder who authorizes use under this
38356 License of the Program or a work on which the Program is based.
38357 The work thus licensed is called the contributor's "contributor
38360 A contributor's "essential patent claims" are all patent claims
38361 owned or controlled by the contributor, whether already acquired or
38362 hereafter acquired, that would be infringed by some manner,
38363 permitted by this License, of making, using, or selling its
38364 contributor version, but do not include claims that would be
38365 infringed only as a consequence of further modification of the
38366 contributor version. For purposes of this definition, "control"
38367 includes the right to grant patent sublicenses in a manner
38368 consistent with the requirements of this License.
38370 Each contributor grants you a non-exclusive, worldwide,
38371 royalty-free patent license under the contributor's essential
38372 patent claims, to make, use, sell, offer for sale, import and
38373 otherwise run, modify and propagate the contents of its
38374 contributor version.
38376 In the following three paragraphs, a "patent license" is any
38377 express agreement or commitment, however denominated, not to
38378 enforce a patent (such as an express permission to practice a
38379 patent or covenant not to sue for patent infringement). To
38380 "grant" such a patent license to a party means to make such an
38381 agreement or commitment not to enforce a patent against the party.
38383 If you convey a covered work, knowingly relying on a patent
38384 license, and the Corresponding Source of the work is not available
38385 for anyone to copy, free of charge and under the terms of this
38386 License, through a publicly available network server or other
38387 readily accessible means, then you must either (1) cause the
38388 Corresponding Source to be so available, or (2) arrange to deprive
38389 yourself of the benefit of the patent license for this particular
38390 work, or (3) arrange, in a manner consistent with the requirements
38391 of this License, to extend the patent license to downstream
38392 recipients. "Knowingly relying" means you have actual knowledge
38393 that, but for the patent license, your conveying the covered work
38394 in a country, or your recipient's use of the covered work in a
38395 country, would infringe one or more identifiable patents in that
38396 country that you have reason to believe are valid.
38398 If, pursuant to or in connection with a single transaction or
38399 arrangement, you convey, or propagate by procuring conveyance of, a
38400 covered work, and grant a patent license to some of the parties
38401 receiving the covered work authorizing them to use, propagate,
38402 modify or convey a specific copy of the covered work, then the
38403 patent license you grant is automatically extended to all
38404 recipients of the covered work and works based on it.
38406 A patent license is "discriminatory" if it does not include within
38407 the scope of its coverage, prohibits the exercise of, or is
38408 conditioned on the non-exercise of one or more of the rights that
38409 are specifically granted under this License. You may not convey a
38410 covered work if you are a party to an arrangement with a third
38411 party that is in the business of distributing software, under
38412 which you make payment to the third party based on the extent of
38413 your activity of conveying the work, and under which the third
38414 party grants, to any of the parties who would receive the covered
38415 work from you, a discriminatory patent license (a) in connection
38416 with copies of the covered work conveyed by you (or copies made
38417 from those copies), or (b) primarily for and in connection with
38418 specific products or compilations that contain the covered work,
38419 unless you entered into that arrangement, or that patent license
38420 was granted, prior to 28 March 2007.
38422 Nothing in this License shall be construed as excluding or limiting
38423 any implied license or other defenses to infringement that may
38424 otherwise be available to you under applicable patent law.
38426 12. No Surrender of Others' Freedom.
38428 If conditions are imposed on you (whether by court order,
38429 agreement or otherwise) that contradict the conditions of this
38430 License, they do not excuse you from the conditions of this
38431 License. If you cannot convey a covered work so as to satisfy
38432 simultaneously your obligations under this License and any other
38433 pertinent obligations, then as a consequence you may not convey it
38434 at all. For example, if you agree to terms that obligate you to
38435 collect a royalty for further conveying from those to whom you
38436 convey the Program, the only way you could satisfy both those
38437 terms and this License would be to refrain entirely from conveying
38440 13. Use with the GNU Affero General Public License.
38442 Notwithstanding any other provision of this License, you have
38443 permission to link or combine any covered work with a work licensed
38444 under version 3 of the GNU Affero General Public License into a
38445 single combined work, and to convey the resulting work. The terms
38446 of this License will continue to apply to the part which is the
38447 covered work, but the special requirements of the GNU Affero
38448 General Public License, section 13, concerning interaction through
38449 a network will apply to the combination as such.
38451 14. Revised Versions of this License.
38453 The Free Software Foundation may publish revised and/or new
38454 versions of the GNU General Public License from time to time.
38455 Such new versions will be similar in spirit to the present
38456 version, but may differ in detail to address new problems or
38459 Each version is given a distinguishing version number. If the
38460 Program specifies that a certain numbered version of the GNU
38461 General Public License "or any later version" applies to it, you
38462 have the option of following the terms and conditions either of
38463 that numbered version or of any later version published by the
38464 Free Software Foundation. If the Program does not specify a
38465 version number of the GNU General Public License, you may choose
38466 any version ever published by the Free Software Foundation.
38468 If the Program specifies that a proxy can decide which future
38469 versions of the GNU General Public License can be used, that
38470 proxy's public statement of acceptance of a version permanently
38471 authorizes you to choose that version for the Program.
38473 Later license versions may give you additional or different
38474 permissions. However, no additional obligations are imposed on any
38475 author or copyright holder as a result of your choosing to follow a
38478 15. Disclaimer of Warranty.
38480 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
38481 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE
38482 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS"
38483 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED,
38484 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
38485 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE
38486 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.
38487 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
38488 NECESSARY SERVICING, REPAIR OR CORRECTION.
38490 16. Limitation of Liability.
38492 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
38493 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES
38494 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU
38495 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
38496 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
38497 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA
38498 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
38499 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
38500 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF
38501 THE POSSIBILITY OF SUCH DAMAGES.
38503 17. Interpretation of Sections 15 and 16.
38505 If the disclaimer of warranty and limitation of liability provided
38506 above cannot be given local legal effect according to their terms,
38507 reviewing courts shall apply local law that most closely
38508 approximates an absolute waiver of all civil liability in
38509 connection with the Program, unless a warranty or assumption of
38510 liability accompanies a copy of the Program in return for a fee.
38513 END OF TERMS AND CONDITIONS
38514 ===========================
38516 How to Apply These Terms to Your New Programs
38517 =============================================
38519 If you develop a new program, and you want it to be of the greatest
38520 possible use to the public, the best way to achieve this is to make it
38521 free software which everyone can redistribute and change under these
38524 To do so, attach the following notices to the program. It is safest
38525 to attach them to the start of each source file to most effectively
38526 state the exclusion of warranty; and each file should have at least the
38527 "copyright" line and a pointer to where the full notice is found.
38529 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES.
38530 Copyright (C) YEAR NAME OF AUTHOR
38532 This program is free software: you can redistribute it and/or modify
38533 it under the terms of the GNU General Public License as published by
38534 the Free Software Foundation, either version 3 of the License, or (at
38535 your option) any later version.
38537 This program is distributed in the hope that it will be useful, but
38538 WITHOUT ANY WARRANTY; without even the implied warranty of
38539 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
38540 General Public License for more details.
38542 You should have received a copy of the GNU General Public License
38543 along with this program. If not, see `http://www.gnu.org/licenses/'.
38545 Also add information on how to contact you by electronic and paper
38548 If the program does terminal interaction, make it output a short
38549 notice like this when it starts in an interactive mode:
38551 PROGRAM Copyright (C) YEAR NAME OF AUTHOR
38552 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
38553 This is free software, and you are welcome to redistribute it
38554 under certain conditions; type `show c' for details.
38556 The hypothetical commands `show w' and `show c' should show the
38557 appropriate parts of the General Public License. Of course, your
38558 program's commands might be different; for a GUI interface, you would
38559 use an "about box".
38561 You should also get your employer (if you work as a programmer) or
38562 school, if any, to sign a "copyright disclaimer" for the program, if
38563 necessary. For more information on this, and how to apply and follow
38564 the GNU GPL, see `http://www.gnu.org/licenses/'.
38566 The GNU General Public License does not permit incorporating your
38567 program into proprietary programs. If your program is a subroutine
38568 library, you may consider it more useful to permit linking proprietary
38569 applications with the library. If this is what you want to do, use the
38570 GNU Lesser General Public License instead of this License. But first,
38571 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'.
38574 File: gcc.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top
38576 GNU Free Documentation License
38577 ******************************
38579 Version 1.2, November 2002
38581 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc.
38582 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
38584 Everyone is permitted to copy and distribute verbatim copies
38585 of this license document, but changing it is not allowed.
38589 The purpose of this License is to make a manual, textbook, or other
38590 functional and useful document "free" in the sense of freedom: to
38591 assure everyone the effective freedom to copy and redistribute it,
38592 with or without modifying it, either commercially or
38593 noncommercially. Secondarily, this License preserves for the
38594 author and publisher a way to get credit for their work, while not
38595 being considered responsible for modifications made by others.
38597 This License is a kind of "copyleft", which means that derivative
38598 works of the document must themselves be free in the same sense.
38599 It complements the GNU General Public License, which is a copyleft
38600 license designed for free software.
38602 We have designed this License in order to use it for manuals for
38603 free software, because free software needs free documentation: a
38604 free program should come with manuals providing the same freedoms
38605 that the software does. But this License is not limited to
38606 software manuals; it can be used for any textual work, regardless
38607 of subject matter or whether it is published as a printed book.
38608 We recommend this License principally for works whose purpose is
38609 instruction or reference.
38611 1. APPLICABILITY AND DEFINITIONS
38613 This License applies to any manual or other work, in any medium,
38614 that contains a notice placed by the copyright holder saying it
38615 can be distributed under the terms of this License. Such a notice
38616 grants a world-wide, royalty-free license, unlimited in duration,
38617 to use that work under the conditions stated herein. The
38618 "Document", below, refers to any such manual or work. Any member
38619 of the public is a licensee, and is addressed as "you". You
38620 accept the license if you copy, modify or distribute the work in a
38621 way requiring permission under copyright law.
38623 A "Modified Version" of the Document means any work containing the
38624 Document or a portion of it, either copied verbatim, or with
38625 modifications and/or translated into another language.
38627 A "Secondary Section" is a named appendix or a front-matter section
38628 of the Document that deals exclusively with the relationship of the
38629 publishers or authors of the Document to the Document's overall
38630 subject (or to related matters) and contains nothing that could
38631 fall directly within that overall subject. (Thus, if the Document
38632 is in part a textbook of mathematics, a Secondary Section may not
38633 explain any mathematics.) The relationship could be a matter of
38634 historical connection with the subject or with related matters, or
38635 of legal, commercial, philosophical, ethical or political position
38638 The "Invariant Sections" are certain Secondary Sections whose
38639 titles are designated, as being those of Invariant Sections, in
38640 the notice that says that the Document is released under this
38641 License. If a section does not fit the above definition of
38642 Secondary then it is not allowed to be designated as Invariant.
38643 The Document may contain zero Invariant Sections. If the Document
38644 does not identify any Invariant Sections then there are none.
38646 The "Cover Texts" are certain short passages of text that are
38647 listed, as Front-Cover Texts or Back-Cover Texts, in the notice
38648 that says that the Document is released under this License. A
38649 Front-Cover Text may be at most 5 words, and a Back-Cover Text may
38650 be at most 25 words.
38652 A "Transparent" copy of the Document means a machine-readable copy,
38653 represented in a format whose specification is available to the
38654 general public, that is suitable for revising the document
38655 straightforwardly with generic text editors or (for images
38656 composed of pixels) generic paint programs or (for drawings) some
38657 widely available drawing editor, and that is suitable for input to
38658 text formatters or for automatic translation to a variety of
38659 formats suitable for input to text formatters. A copy made in an
38660 otherwise Transparent file format whose markup, or absence of
38661 markup, has been arranged to thwart or discourage subsequent
38662 modification by readers is not Transparent. An image format is
38663 not Transparent if used for any substantial amount of text. A
38664 copy that is not "Transparent" is called "Opaque".
38666 Examples of suitable formats for Transparent copies include plain
38667 ASCII without markup, Texinfo input format, LaTeX input format,
38668 SGML or XML using a publicly available DTD, and
38669 standard-conforming simple HTML, PostScript or PDF designed for
38670 human modification. Examples of transparent image formats include
38671 PNG, XCF and JPG. Opaque formats include proprietary formats that
38672 can be read and edited only by proprietary word processors, SGML or
38673 XML for which the DTD and/or processing tools are not generally
38674 available, and the machine-generated HTML, PostScript or PDF
38675 produced by some word processors for output purposes only.
38677 The "Title Page" means, for a printed book, the title page itself,
38678 plus such following pages as are needed to hold, legibly, the
38679 material this License requires to appear in the title page. For
38680 works in formats which do not have any title page as such, "Title
38681 Page" means the text near the most prominent appearance of the
38682 work's title, preceding the beginning of the body of the text.
38684 A section "Entitled XYZ" means a named subunit of the Document
38685 whose title either is precisely XYZ or contains XYZ in parentheses
38686 following text that translates XYZ in another language. (Here XYZ
38687 stands for a specific section name mentioned below, such as
38688 "Acknowledgements", "Dedications", "Endorsements", or "History".)
38689 To "Preserve the Title" of such a section when you modify the
38690 Document means that it remains a section "Entitled XYZ" according
38691 to this definition.
38693 The Document may include Warranty Disclaimers next to the notice
38694 which states that this License applies to the Document. These
38695 Warranty Disclaimers are considered to be included by reference in
38696 this License, but only as regards disclaiming warranties: any other
38697 implication that these Warranty Disclaimers may have is void and
38698 has no effect on the meaning of this License.
38700 2. VERBATIM COPYING
38702 You may copy and distribute the Document in any medium, either
38703 commercially or noncommercially, provided that this License, the
38704 copyright notices, and the license notice saying this License
38705 applies to the Document are reproduced in all copies, and that you
38706 add no other conditions whatsoever to those of this License. You
38707 may not use technical measures to obstruct or control the reading
38708 or further copying of the copies you make or distribute. However,
38709 you may accept compensation in exchange for copies. If you
38710 distribute a large enough number of copies you must also follow
38711 the conditions in section 3.
38713 You may also lend copies, under the same conditions stated above,
38714 and you may publicly display copies.
38716 3. COPYING IN QUANTITY
38718 If you publish printed copies (or copies in media that commonly
38719 have printed covers) of the Document, numbering more than 100, and
38720 the Document's license notice requires Cover Texts, you must
38721 enclose the copies in covers that carry, clearly and legibly, all
38722 these Cover Texts: Front-Cover Texts on the front cover, and
38723 Back-Cover Texts on the back cover. Both covers must also clearly
38724 and legibly identify you as the publisher of these copies. The
38725 front cover must present the full title with all words of the
38726 title equally prominent and visible. You may add other material
38727 on the covers in addition. Copying with changes limited to the
38728 covers, as long as they preserve the title of the Document and
38729 satisfy these conditions, can be treated as verbatim copying in
38732 If the required texts for either cover are too voluminous to fit
38733 legibly, you should put the first ones listed (as many as fit
38734 reasonably) on the actual cover, and continue the rest onto
38737 If you publish or distribute Opaque copies of the Document
38738 numbering more than 100, you must either include a
38739 machine-readable Transparent copy along with each Opaque copy, or
38740 state in or with each Opaque copy a computer-network location from
38741 which the general network-using public has access to download
38742 using public-standard network protocols a complete Transparent
38743 copy of the Document, free of added material. If you use the
38744 latter option, you must take reasonably prudent steps, when you
38745 begin distribution of Opaque copies in quantity, to ensure that
38746 this Transparent copy will remain thus accessible at the stated
38747 location until at least one year after the last time you
38748 distribute an Opaque copy (directly or through your agents or
38749 retailers) of that edition to the public.
38751 It is requested, but not required, that you contact the authors of
38752 the Document well before redistributing any large number of
38753 copies, to give them a chance to provide you with an updated
38754 version of the Document.
38758 You may copy and distribute a Modified Version of the Document
38759 under the conditions of sections 2 and 3 above, provided that you
38760 release the Modified Version under precisely this License, with
38761 the Modified Version filling the role of the Document, thus
38762 licensing distribution and modification of the Modified Version to
38763 whoever possesses a copy of it. In addition, you must do these
38764 things in the Modified Version:
38766 A. Use in the Title Page (and on the covers, if any) a title
38767 distinct from that of the Document, and from those of
38768 previous versions (which should, if there were any, be listed
38769 in the History section of the Document). You may use the
38770 same title as a previous version if the original publisher of
38771 that version gives permission.
38773 B. List on the Title Page, as authors, one or more persons or
38774 entities responsible for authorship of the modifications in
38775 the Modified Version, together with at least five of the
38776 principal authors of the Document (all of its principal
38777 authors, if it has fewer than five), unless they release you
38778 from this requirement.
38780 C. State on the Title page the name of the publisher of the
38781 Modified Version, as the publisher.
38783 D. Preserve all the copyright notices of the Document.
38785 E. Add an appropriate copyright notice for your modifications
38786 adjacent to the other copyright notices.
38788 F. Include, immediately after the copyright notices, a license
38789 notice giving the public permission to use the Modified
38790 Version under the terms of this License, in the form shown in
38791 the Addendum below.
38793 G. Preserve in that license notice the full lists of Invariant
38794 Sections and required Cover Texts given in the Document's
38797 H. Include an unaltered copy of this License.
38799 I. Preserve the section Entitled "History", Preserve its Title,
38800 and add to it an item stating at least the title, year, new
38801 authors, and publisher of the Modified Version as given on
38802 the Title Page. If there is no section Entitled "History" in
38803 the Document, create one stating the title, year, authors,
38804 and publisher of the Document as given on its Title Page,
38805 then add an item describing the Modified Version as stated in
38806 the previous sentence.
38808 J. Preserve the network location, if any, given in the Document
38809 for public access to a Transparent copy of the Document, and
38810 likewise the network locations given in the Document for
38811 previous versions it was based on. These may be placed in
38812 the "History" section. You may omit a network location for a
38813 work that was published at least four years before the
38814 Document itself, or if the original publisher of the version
38815 it refers to gives permission.
38817 K. For any section Entitled "Acknowledgements" or "Dedications",
38818 Preserve the Title of the section, and preserve in the
38819 section all the substance and tone of each of the contributor
38820 acknowledgements and/or dedications given therein.
38822 L. Preserve all the Invariant Sections of the Document,
38823 unaltered in their text and in their titles. Section numbers
38824 or the equivalent are not considered part of the section
38827 M. Delete any section Entitled "Endorsements". Such a section
38828 may not be included in the Modified Version.
38830 N. Do not retitle any existing section to be Entitled
38831 "Endorsements" or to conflict in title with any Invariant
38834 O. Preserve any Warranty Disclaimers.
38836 If the Modified Version includes new front-matter sections or
38837 appendices that qualify as Secondary Sections and contain no
38838 material copied from the Document, you may at your option
38839 designate some or all of these sections as invariant. To do this,
38840 add their titles to the list of Invariant Sections in the Modified
38841 Version's license notice. These titles must be distinct from any
38842 other section titles.
38844 You may add a section Entitled "Endorsements", provided it contains
38845 nothing but endorsements of your Modified Version by various
38846 parties--for example, statements of peer review or that the text
38847 has been approved by an organization as the authoritative
38848 definition of a standard.
38850 You may add a passage of up to five words as a Front-Cover Text,
38851 and a passage of up to 25 words as a Back-Cover Text, to the end
38852 of the list of Cover Texts in the Modified Version. Only one
38853 passage of Front-Cover Text and one of Back-Cover Text may be
38854 added by (or through arrangements made by) any one entity. If the
38855 Document already includes a cover text for the same cover,
38856 previously added by you or by arrangement made by the same entity
38857 you are acting on behalf of, you may not add another; but you may
38858 replace the old one, on explicit permission from the previous
38859 publisher that added the old one.
38861 The author(s) and publisher(s) of the Document do not by this
38862 License give permission to use their names for publicity for or to
38863 assert or imply endorsement of any Modified Version.
38865 5. COMBINING DOCUMENTS
38867 You may combine the Document with other documents released under
38868 this License, under the terms defined in section 4 above for
38869 modified versions, provided that you include in the combination
38870 all of the Invariant Sections of all of the original documents,
38871 unmodified, and list them all as Invariant Sections of your
38872 combined work in its license notice, and that you preserve all
38873 their Warranty Disclaimers.
38875 The combined work need only contain one copy of this License, and
38876 multiple identical Invariant Sections may be replaced with a single
38877 copy. If there are multiple Invariant Sections with the same name
38878 but different contents, make the title of each such section unique
38879 by adding at the end of it, in parentheses, the name of the
38880 original author or publisher of that section if known, or else a
38881 unique number. Make the same adjustment to the section titles in
38882 the list of Invariant Sections in the license notice of the
38885 In the combination, you must combine any sections Entitled
38886 "History" in the various original documents, forming one section
38887 Entitled "History"; likewise combine any sections Entitled
38888 "Acknowledgements", and any sections Entitled "Dedications". You
38889 must delete all sections Entitled "Endorsements."
38891 6. COLLECTIONS OF DOCUMENTS
38893 You may make a collection consisting of the Document and other
38894 documents released under this License, and replace the individual
38895 copies of this License in the various documents with a single copy
38896 that is included in the collection, provided that you follow the
38897 rules of this License for verbatim copying of each of the
38898 documents in all other respects.
38900 You may extract a single document from such a collection, and
38901 distribute it individually under this License, provided you insert
38902 a copy of this License into the extracted document, and follow
38903 this License in all other respects regarding verbatim copying of
38906 7. AGGREGATION WITH INDEPENDENT WORKS
38908 A compilation of the Document or its derivatives with other
38909 separate and independent documents or works, in or on a volume of
38910 a storage or distribution medium, is called an "aggregate" if the
38911 copyright resulting from the compilation is not used to limit the
38912 legal rights of the compilation's users beyond what the individual
38913 works permit. When the Document is included in an aggregate, this
38914 License does not apply to the other works in the aggregate which
38915 are not themselves derivative works of the Document.
38917 If the Cover Text requirement of section 3 is applicable to these
38918 copies of the Document, then if the Document is less than one half
38919 of the entire aggregate, the Document's Cover Texts may be placed
38920 on covers that bracket the Document within the aggregate, or the
38921 electronic equivalent of covers if the Document is in electronic
38922 form. Otherwise they must appear on printed covers that bracket
38923 the whole aggregate.
38927 Translation is considered a kind of modification, so you may
38928 distribute translations of the Document under the terms of section
38929 4. Replacing Invariant Sections with translations requires special
38930 permission from their copyright holders, but you may include
38931 translations of some or all Invariant Sections in addition to the
38932 original versions of these Invariant Sections. You may include a
38933 translation of this License, and all the license notices in the
38934 Document, and any Warranty Disclaimers, provided that you also
38935 include the original English version of this License and the
38936 original versions of those notices and disclaimers. In case of a
38937 disagreement between the translation and the original version of
38938 this License or a notice or disclaimer, the original version will
38941 If a section in the Document is Entitled "Acknowledgements",
38942 "Dedications", or "History", the requirement (section 4) to
38943 Preserve its Title (section 1) will typically require changing the
38948 You may not copy, modify, sublicense, or distribute the Document
38949 except as expressly provided for under this License. Any other
38950 attempt to copy, modify, sublicense or distribute the Document is
38951 void, and will automatically terminate your rights under this
38952 License. However, parties who have received copies, or rights,
38953 from you under this License will not have their licenses
38954 terminated so long as such parties remain in full compliance.
38956 10. FUTURE REVISIONS OF THIS LICENSE
38958 The Free Software Foundation may publish new, revised versions of
38959 the GNU Free Documentation License from time to time. Such new
38960 versions will be similar in spirit to the present version, but may
38961 differ in detail to address new problems or concerns. See
38962 `http://www.gnu.org/copyleft/'.
38964 Each version of the License is given a distinguishing version
38965 number. If the Document specifies that a particular numbered
38966 version of this License "or any later version" applies to it, you
38967 have the option of following the terms and conditions either of
38968 that specified version or of any later version that has been
38969 published (not as a draft) by the Free Software Foundation. If
38970 the Document does not specify a version number of this License,
38971 you may choose any version ever published (not as a draft) by the
38972 Free Software Foundation.
38974 ADDENDUM: How to use this License for your documents
38975 ====================================================
38977 To use this License in a document you have written, include a copy of
38978 the License in the document and put the following copyright and license
38979 notices just after the title page:
38981 Copyright (C) YEAR YOUR NAME.
38982 Permission is granted to copy, distribute and/or modify this document
38983 under the terms of the GNU Free Documentation License, Version 1.2
38984 or any later version published by the Free Software Foundation;
38985 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
38986 Texts. A copy of the license is included in the section entitled ``GNU
38987 Free Documentation License''.
38989 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts,
38990 replace the "with...Texts." line with this:
38992 with the Invariant Sections being LIST THEIR TITLES, with
38993 the Front-Cover Texts being LIST, and with the Back-Cover Texts
38996 If you have Invariant Sections without Cover Texts, or some other
38997 combination of the three, merge those two alternatives to suit the
39000 If your document contains nontrivial examples of program code, we
39001 recommend releasing these examples in parallel under your choice of
39002 free software license, such as the GNU General Public License, to
39003 permit their use in free software.
39006 File: gcc.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top
39008 Contributors to GCC
39009 *******************
39011 The GCC project would like to thank its many contributors. Without
39012 them the project would not have been nearly as successful as it has
39013 been. Any omissions in this list are accidental. Feel free to contact
39014 <law@redhat.com> or <gerald@pfeifer.com> if you have been left out or
39015 some of your contributions are not listed. Please keep this list in
39016 alphabetical order.
39018 * Analog Devices helped implement the support for complex data types
39021 * John David Anglin for threading-related fixes and improvements to
39022 libstdc++-v3, and the HP-UX port.
39024 * James van Artsdalen wrote the code that makes efficient use of the
39025 Intel 80387 register stack.
39027 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta
39030 * Alasdair Baird for various bug fixes.
39032 * Giovanni Bajo for analyzing lots of complicated C++ problem
39035 * Peter Barada for his work to improve code generation for new
39038 * Gerald Baumgartner added the signature extension to the C++ front
39041 * Godmar Back for his Java improvements and encouragement.
39043 * Scott Bambrough for help porting the Java compiler.
39045 * Wolfgang Bangerth for processing tons of bug reports.
39047 * Jon Beniston for his Microsoft Windows port of Java.
39049 * Daniel Berlin for better DWARF2 support, faster/better
39050 optimizations, improved alias analysis, plus migrating GCC to
39053 * Geoff Berry for his Java object serialization work and various
39056 * Uros Bizjak for the implementation of x87 math built-in functions
39057 and for various middle end and i386 back end improvements and bug
39060 * Eric Blake for helping to make GCJ and libgcj conform to the
39063 * Janne Blomqvist for contributions to GNU Fortran.
39065 * Segher Boessenkool for various fixes.
39067 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and
39070 * Neil Booth for work on cpplib, lang hooks, debug hooks and other
39071 miscellaneous clean-ups.
39073 * Steven Bosscher for integrating the GNU Fortran front end into GCC
39074 and for contributing to the tree-ssa branch.
39076 * Eric Botcazou for fixing middle- and backend bugs left and right.
39078 * Per Bothner for his direction via the steering committee and
39079 various improvements to the infrastructure for supporting new
39080 languages. Chill front end implementation. Initial
39081 implementations of cpplib, fix-header, config.guess, libio, and
39082 past C++ library (libg++) maintainer. Dreaming up, designing and
39083 implementing much of GCJ.
39085 * Devon Bowen helped port GCC to the Tahoe.
39087 * Don Bowman for mips-vxworks contributions.
39089 * Dave Brolley for work on cpplib and Chill.
39091 * Paul Brook for work on the ARM architecture and maintaining GNU
39094 * Robert Brown implemented the support for Encore 32000 systems.
39096 * Christian Bruel for improvements to local store elimination.
39098 * Herman A.J. ten Brugge for various fixes.
39100 * Joerg Brunsmann for Java compiler hacking and help with the GCJ
39103 * Joe Buck for his direction via the steering committee.
39105 * Craig Burley for leadership of the G77 Fortran effort.
39107 * Stephan Buys for contributing Doxygen notes for libstdc++.
39109 * Paolo Carlini for libstdc++ work: lots of efficiency improvements
39110 to the C++ strings, streambufs and formatted I/O, hard detective
39111 work on the frustrating localization issues, and keeping up with
39112 the problem reports.
39114 * John Carr for his alias work, SPARC hacking, infrastructure
39115 improvements, previous contributions to the steering committee,
39116 loop optimizations, etc.
39118 * Stephane Carrez for 68HC11 and 68HC12 ports.
39120 * Steve Chamberlain for support for the Renesas SH and H8 processors
39121 and the PicoJava processor, and for GCJ config fixes.
39123 * Glenn Chambers for help with the GCJ FAQ.
39125 * John-Marc Chandonia for various libgcj patches.
39127 * Scott Christley for his Objective-C contributions.
39129 * Eric Christopher for his Java porting help and clean-ups.
39131 * Branko Cibej for more warning contributions.
39133 * The GNU Classpath project for all of their merged runtime code.
39135 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and
39136 other random hacking.
39138 * Michael Cook for libstdc++ cleanup patches to reduce warnings.
39140 * R. Kelley Cook for making GCC buildable from a read-only directory
39141 as well as other miscellaneous build process and documentation
39144 * Ralf Corsepius for SH testing and minor bug fixing.
39146 * Stan Cox for care and feeding of the x86 port and lots of behind
39147 the scenes hacking.
39149 * Alex Crain provided changes for the 3b1.
39151 * Ian Dall for major improvements to the NS32k port.
39153 * Paul Dale for his work to add uClinux platform support to the m68k
39156 * Dario Dariol contributed the four varieties of sample programs
39157 that print a copy of their source.
39159 * Russell Davidson for fstream and stringstream fixes in libstdc++.
39161 * Bud Davis for work on the G77 and GNU Fortran compilers.
39163 * Mo DeJong for GCJ and libgcj bug fixes.
39165 * DJ Delorie for the DJGPP port, build and libiberty maintenance,
39166 various bug fixes, and the M32C port.
39168 * Arnaud Desitter for helping to debug GNU Fortran.
39170 * Gabriel Dos Reis for contributions to G++, contributions and
39171 maintenance of GCC diagnostics infrastructure, libstdc++-v3,
39172 including `valarray<>', `complex<>', maintaining the numerics
39173 library (including that pesky `<limits>' :-) and keeping
39174 up-to-date anything to do with numbers.
39176 * Ulrich Drepper for his work on glibc, testing of GCC using glibc,
39177 ISO C99 support, CFG dumping support, etc., plus support of the
39178 C++ runtime libraries including for all kinds of C interface
39179 issues, contributing and maintaining `complex<>', sanity checking
39180 and disbursement, configuration architecture, libio maintenance,
39181 and early math work.
39183 * Zdenek Dvorak for a new loop unroller and various fixes.
39185 * Richard Earnshaw for his ongoing work with the ARM.
39187 * David Edelsohn for his direction via the steering committee,
39188 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa
39189 loop changes, doing the entire AIX port of libstdc++ with his bare
39190 hands, and for ensuring GCC properly keeps working on AIX.
39192 * Kevin Ediger for the floating point formatting of num_put::do_put
39195 * Phil Edwards for libstdc++ work including configuration hackery,
39196 documentation maintainer, chief breaker of the web pages, the
39197 occasional iostream bug fix, and work on shared library symbol
39200 * Paul Eggert for random hacking all over GCC.
39202 * Mark Elbrecht for various DJGPP improvements, and for libstdc++
39203 configuration support for locales and fstream-related fixes.
39205 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and
39208 * Christian Ehrhardt for dealing with bug reports.
39210 * Ben Elliston for his work to move the Objective-C runtime into its
39211 own subdirectory and for his work on autoconf.
39213 * Revital Eres for work on the PowerPC 750CL port.
39215 * Marc Espie for OpenBSD support.
39217 * Doug Evans for much of the global optimization framework, arc,
39218 m32r, and SPARC work.
39220 * Christopher Faylor for his work on the Cygwin port and for caring
39221 and feeding the gcc.gnu.org box and saving its users tons of spam.
39223 * Fred Fish for BeOS support and Ada fixes.
39225 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ.
39227 * Peter Gerwinski for various bug fixes and the Pascal front end.
39229 * Kaveh R. Ghazi for his direction via the steering committee,
39230 amazing work to make `-W -Wall -W* -Werror' useful, and
39231 continuously testing GCC on a plethora of platforms. Kaveh
39232 extends his gratitude to the CAIP Center at Rutgers University for
39233 providing him with computing resources to work on Free Software
39234 since the late 1980s.
39236 * John Gilmore for a donation to the FSF earmarked improving GNU
39239 * Judy Goldberg for c++ contributions.
39241 * Torbjorn Granlund for various fixes and the c-torture testsuite,
39242 multiply- and divide-by-constant optimization, improved long long
39243 support, improved leaf function register allocation, and his
39244 direction via the steering committee.
39246 * Anthony Green for his `-Os' contributions and Java front end work.
39248 * Stu Grossman for gdb hacking, allowing GCJ developers to debug
39251 * Michael K. Gschwind contributed the port to the PDP-11.
39253 * Ron Guilmette implemented the `protoize' and `unprotoize' tools,
39254 the support for Dwarf symbolic debugging information, and much of
39255 the support for System V Release 4. He has also worked heavily on
39256 the Intel 386 and 860 support.
39258 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload
39261 * Bruno Haible for improvements in the runtime overhead for EH, new
39262 warnings and assorted bug fixes.
39264 * Andrew Haley for his amazing Java compiler and library efforts.
39266 * Chris Hanson assisted in making GCC work on HP-UX for the 9000
39269 * Michael Hayes for various thankless work he's done trying to get
39270 the c30/c40 ports functional. Lots of loop and unroll
39271 improvements and fixes.
39273 * Dara Hazeghi for wading through myriads of target-specific bug
39276 * Kate Hedstrom for staking the G77 folks with an initial testsuite.
39278 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64
39279 work, loop opts, and generally fixing lots of old problems we've
39280 ignored for years, flow rewrite and lots of further stuff,
39281 including reviewing tons of patches.
39283 * Aldy Hernandez for working on the PowerPC port, SIMD support, and
39286 * Nobuyuki Hikichi of Software Research Associates, Tokyo,
39287 contributed the support for the Sony NEWS machine.
39289 * Kazu Hirata for caring and feeding the Renesas H8/300 port and
39292 * Katherine Holcomb for work on GNU Fortran.
39294 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots
39295 of testing and bug fixing, particularly of GCC configury code.
39297 * Steve Holmgren for MachTen patches.
39299 * Jan Hubicka for his x86 port improvements.
39301 * Falk Hueffner for working on C and optimization bug reports.
39303 * Bernardo Innocenti for his m68k work, including merging of
39304 ColdFire improvements and uClinux support.
39306 * Christian Iseli for various bug fixes.
39308 * Kamil Iskra for general m68k hacking.
39310 * Lee Iverson for random fixes and MIPS testing.
39312 * Andreas Jaeger for testing and benchmarking of GCC and various bug
39315 * Jakub Jelinek for his SPARC work and sibling call optimizations as
39316 well as lots of bug fixes and test cases, and for improving the
39319 * Janis Johnson for ia64 testing and fixes, her quality improvement
39320 sidetracks, and web page maintenance.
39322 * Kean Johnston for SCO OpenServer support and various fixes.
39324 * Tim Josling for the sample language treelang based originally on
39325 Richard Kenner's "toy" language.
39327 * Nicolai Josuttis for additional libstdc++ documentation.
39329 * Klaus Kaempf for his ongoing work to make alpha-vms a viable
39332 * Steven G. Kargl for work on GNU Fortran.
39334 * David Kashtan of SRI adapted GCC to VMS.
39336 * Ryszard Kabatek for many, many libstdc++ bug fixes and
39337 optimizations of strings, especially member functions, and for
39340 * Geoffrey Keating for his ongoing work to make the PPC work for
39341 GNU/Linux and his automatic regression tester.
39343 * Brendan Kehoe for his ongoing work with G++ and for a lot of early
39344 work in just about every part of libstdc++.
39346 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the
39349 * Richard Kenner of the New York University Ultracomputer Research
39350 Laboratory wrote the machine descriptions for the AMD 29000, the
39351 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the
39352 support for instruction attributes. He also made changes to
39353 better support RISC processors including changes to common
39354 subexpression elimination, strength reduction, function calling
39355 sequence handling, and condition code support, in addition to
39356 generalizing the code for frame pointer elimination and delay slot
39357 scheduling. Richard Kenner was also the head maintainer of GCC
39360 * Mumit Khan for various contributions to the Cygwin and Mingw32
39361 ports and maintaining binary releases for Microsoft Windows hosts,
39362 and for massive libstdc++ porting work to Cygwin/Mingw32.
39364 * Robin Kirkham for cpu32 support.
39366 * Mark Klein for PA improvements.
39368 * Thomas Koenig for various bug fixes.
39370 * Bruce Korb for the new and improved fixincludes code.
39372 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3
39375 * Charles LaBrec contributed the support for the Integrated Solutions
39378 * Asher Langton and Mike Kumbera for contributing Cray pointer
39379 support to GNU Fortran, and for other GNU Fortran improvements.
39381 * Jeff Law for his direction via the steering committee,
39382 coordinating the entire egcs project and GCC 2.95, rolling out
39383 snapshots and releases, handling merges from GCC2, reviewing tons
39384 of patches that might have fallen through the cracks else, and
39385 random but extensive hacking.
39387 * Marc Lehmann for his direction via the steering committee and
39388 helping with analysis and improvements of x86 performance.
39390 * Victor Leikehman for work on GNU Fortran.
39392 * Ted Lemon wrote parts of the RTL reader and printer.
39394 * Kriang Lerdsuwanakij for C++ improvements including template as
39395 template parameter support, and many C++ fixes.
39397 * Warren Levy for tremendous work on libgcj (Java Runtime Library)
39398 and random work on the Java front end.
39400 * Alain Lichnewsky ported GCC to the MIPS CPU.
39402 * Oskar Liljeblad for hacking on AWT and his many Java bug reports
39405 * Robert Lipe for OpenServer support, new testsuites, testing, etc.
39407 * Chen Liqin for various S+core related fixes/improvement, and for
39408 maintaining the S+core port.
39410 * Weiwen Liu for testing and various bug fixes.
39412 * Manuel Lo'pez-Iba'n~ez for improving `-Wconversion' and many other
39413 diagnostics fixes and improvements.
39415 * Dave Love for his ongoing work with the Fortran front end and
39418 * Martin von Lo"wis for internal consistency checking infrastructure,
39419 various C++ improvements including namespace support, and tons of
39420 assistance with libstdc++/compiler merges.
39422 * H.J. Lu for his previous contributions to the steering committee,
39423 many x86 bug reports, prototype patches, and keeping the GNU/Linux
39426 * Greg McGary for random fixes and (someday) bounded pointers.
39428 * Andrew MacLeod for his ongoing work in building a real EH system,
39429 various code generation improvements, work on the global
39432 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC
39433 hacking improvements to compile-time performance, overall
39434 knowledge and direction in the area of instruction scheduling, and
39435 design and implementation of the automaton based instruction
39438 * Bob Manson for his behind the scenes work on dejagnu.
39440 * Philip Martin for lots of libstdc++ string and vector iterator
39441 fixes and improvements, and string clean up and testsuites.
39443 * All of the Mauve project contributors, for Java test code.
39445 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements.
39447 * Adam Megacz for his work on the Microsoft Windows port of GCJ.
39449 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS,
39450 powerpc, haifa, ECOFF debug support, and other assorted hacking.
39452 * Jason Merrill for his direction via the steering committee and
39453 leading the G++ effort.
39455 * Martin Michlmayr for testing GCC on several architectures using the
39456 entire Debian archive.
39458 * David Miller for his direction via the steering committee, lots of
39459 SPARC work, improvements in jump.c and interfacing with the Linux
39462 * Gary Miller ported GCC to Charles River Data Systems machines.
39464 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning
39465 the entire libstdc++ testsuite namespace-compatible.
39467 * Mark Mitchell for his direction via the steering committee,
39468 mountains of C++ work, load/store hoisting out of loops, alias
39469 analysis improvements, ISO C `restrict' support, and serving as
39470 release manager for GCC 3.x.
39472 * Alan Modra for various GNU/Linux bits and testing.
39474 * Toon Moene for his direction via the steering committee, Fortran
39475 maintenance, and his ongoing work to make us make Fortran run fast.
39477 * Jason Molenda for major help in the care and feeding of all the
39478 services on the gcc.gnu.org (formerly egcs.cygnus.com)
39479 machine--mail, web services, ftp services, etc etc. Doing all
39480 this work on scrap paper and the backs of envelopes would have
39483 * Catherine Moore for fixing various ugly problems we have sent her
39484 way, including the haifa bug which was killing the Alpha & PowerPC
39487 * Mike Moreton for his various Java patches.
39489 * David Mosberger-Tang for various Alpha improvements, and for the
39490 initial IA-64 port.
39492 * Stephen Moshier contributed the floating point emulator that
39493 assists in cross-compilation and permits support for floating
39494 point numbers wider than 64 bits and for ISO C99 support.
39496 * Bill Moyer for his behind the scenes work on various issues.
39498 * Philippe De Muyter for his work on the m68k port.
39500 * Joseph S. Myers for his work on the PDP-11 port, format checking
39501 and ISO C99 support, and continuous emphasis on (and contributions
39504 * Nathan Myers for his work on libstdc++-v3: architecture and
39505 authorship through the first three snapshots, including
39506 implementation of locale infrastructure, string, shadow C headers,
39507 and the initial project documentation (DESIGN, CHECKLIST, and so
39508 forth). Later, more work on MT-safe string and shadow headers.
39510 * Felix Natter for documentation on porting libstdc++.
39512 * Nathanael Nerode for cleaning up the configuration/build process.
39514 * NeXT, Inc. donated the front end that supports the Objective-C
39517 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to
39518 the search engine setup, various documentation fixes and other
39521 * Geoff Noer for his work on getting cygwin native builds working.
39523 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance
39524 tracking web pages, GIMPLE tuples, and assorted fixes.
39526 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64,
39527 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and
39528 related infrastructure improvements.
39530 * Alexandre Oliva for various build infrastructure improvements,
39531 scripts and amazing testing work, including keeping libtool issues
39534 * Stefan Olsson for work on mt_alloc.
39536 * Melissa O'Neill for various NeXT fixes.
39538 * Rainer Orth for random MIPS work, including improvements to GCC's
39539 o32 ABI support, improvements to dejagnu's MIPS support, Java
39540 configuration clean-ups and porting work, etc.
39542 * Hartmut Penner for work on the s390 port.
39544 * Paul Petersen wrote the machine description for the Alliant FX/8.
39546 * Alexandre Petit-Bianco for implementing much of the Java compiler
39547 and continued Java maintainership.
39549 * Matthias Pfaller for major improvements to the NS32k port.
39551 * Gerald Pfeifer for his direction via the steering committee,
39552 pointing out lots of problems we need to solve, maintenance of the
39553 web pages, and taking care of documentation maintenance in general.
39555 * Andrew Pinski for processing bug reports by the dozen.
39557 * Ovidiu Predescu for his work on the Objective-C front end and
39560 * Jerry Quinn for major performance improvements in C++ formatted
39563 * Ken Raeburn for various improvements to checker, MIPS ports and
39564 various cleanups in the compiler.
39566 * Rolf W. Rasmussen for hacking on AWT.
39568 * David Reese of Sun Microsystems contributed to the Solaris on
39571 * Volker Reichelt for keeping up with the problem reports.
39573 * Joern Rennecke for maintaining the sh port, loop, regmove & reload
39576 * Loren J. Rittle for improvements to libstdc++-v3 including the
39577 FreeBSD port, threading fixes, thread-related configury changes,
39578 critical threading documentation, and solutions to really tricky
39579 I/O problems, as well as keeping GCC properly working on FreeBSD
39580 and continuous testing.
39582 * Craig Rodrigues for processing tons of bug reports.
39584 * Ola Ro"nnerup for work on mt_alloc.
39586 * Gavin Romig-Koch for lots of behind the scenes MIPS work.
39588 * David Ronis inspired and encouraged Craig to rewrite the G77
39589 documentation in texinfo format by contributing a first pass at a
39590 translation of the old `g77-0.5.16/f/DOC' file.
39592 * Ken Rose for fixes to GCC's delay slot filling code.
39594 * Paul Rubin wrote most of the preprocessor.
39596 * Pe'tur Runo'lfsson for major performance improvements in C++
39597 formatted I/O and large file support in C++ filebuf.
39599 * Chip Salzenberg for libstdc++ patches and improvements to locales,
39600 traits, Makefiles, libio, libtool hackery, and "long long" support.
39602 * Juha Sarlin for improvements to the H8 code generator.
39604 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series
39607 * Roger Sayle for improvements to constant folding and GCC's RTL
39608 optimizers as well as for fixing numerous bugs.
39610 * Bradley Schatz for his work on the GCJ FAQ.
39612 * Peter Schauer wrote the code to allow debugging to work on the
39615 * William Schelter did most of the work on the Intel 80386 support.
39617 * Tobias Schlu"ter for work on GNU Fortran.
39619 * Bernd Schmidt for various code generation improvements and major
39620 work in the reload pass as well a serving as release manager for
39623 * Peter Schmid for constant testing of libstdc++--especially
39624 application testing, going above and beyond what was requested for
39625 the release criteria--and libstdc++ header file tweaks.
39627 * Jason Schroeder for jcf-dump patches.
39629 * Andreas Schwab for his work on the m68k port.
39631 * Lars Segerlund for work on GNU Fortran.
39633 * Joel Sherrill for his direction via the steering committee, RTEMS
39634 contributions and RTEMS testing.
39636 * Nathan Sidwell for many C++ fixes/improvements.
39638 * Jeffrey Siegal for helping RMS with the original design of GCC,
39639 some code which handles the parse tree and RTL data structures,
39640 constant folding and help with the original VAX & m68k ports.
39642 * Kenny Simpson for prompting libstdc++ fixes due to defect reports
39643 from the LWG (thereby keeping GCC in line with updates from the
39646 * Franz Sirl for his ongoing work with making the PPC port stable
39649 * Andrey Slepuhin for assorted AIX hacking.
39651 * Trevor Smigiel for contributing the SPU port.
39653 * Christopher Smith did the port for Convex machines.
39655 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports.
39657 * Randy Smith finished the Sun FPA support.
39659 * Scott Snyder for queue, iterator, istream, and string fixes and
39660 libstdc++ testsuite entries. Also for providing the patch to G77
39661 to add rudimentary support for `INTEGER*1', `INTEGER*2', and
39664 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique.
39666 * Richard Stallman, for writing the original GCC and launching the
39669 * Jan Stein of the Chalmers Computer Society provided support for
39670 Genix, as well as part of the 32000 machine description.
39672 * Nigel Stephens for various mips16 related fixes/improvements.
39674 * Jonathan Stone wrote the machine description for the Pyramid
39677 * Graham Stott for various infrastructure improvements.
39679 * John Stracke for his Java HTTP protocol fixes.
39681 * Mike Stump for his Elxsi port, G++ contributions over the years
39682 and more recently his vxworks contributions
39684 * Jeff Sturm for Java porting help, bug fixes, and encouragement.
39686 * Shigeya Suzuki for this fixes for the bsdi platforms.
39688 * Ian Lance Taylor for his mips16 work, general configury hacking,
39691 * Holger Teutsch provided the support for the Clipper CPU.
39693 * Gary Thomas for his ongoing work to make the PPC work for
39696 * Philipp Thomas for random bug fixes throughout the compiler
39698 * Jason Thorpe for thread support in libstdc++ on NetBSD.
39700 * Kresten Krab Thorup wrote the run time support for the Objective-C
39701 language and the fantastic Java bytecode interpreter.
39703 * Michael Tiemann for random bug fixes, the first instruction
39704 scheduler, initial C++ support, function integration, NS32k, SPARC
39705 and M88k machine description work, delay slot scheduling.
39707 * Andreas Tobler for his work porting libgcj to Darwin.
39709 * Teemu Torma for thread safe exception handling support.
39711 * Leonard Tower wrote parts of the parser, RTL generator, and RTL
39712 definitions, and of the VAX machine description.
39714 * Daniel Towner and Hariharan Sandanagobalane contributed and
39715 maintain the picoChip port.
39717 * Tom Tromey for internationalization support and for his many Java
39718 contributions and libgcj maintainership.
39720 * Lassi Tuura for improvements to config.guess to determine HP
39723 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes.
39725 * Andy Vaught for the design and initial implementation of the GNU
39728 * Brent Verner for work with the libstdc++ cshadow files and their
39729 associated configure steps.
39731 * Todd Vierling for contributions for NetBSD ports.
39733 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML
39736 * Dean Wakerley for converting the install documentation from HTML
39737 to texinfo in time for GCC 3.0.
39739 * Krister Walfridsson for random bug fixes.
39741 * Feng Wang for contributions to GNU Fortran.
39743 * Stephen M. Webb for time and effort on making libstdc++ shadow
39744 files work with the tricky Solaris 8+ headers, and for pushing the
39745 build-time header tree.
39747 * John Wehle for various improvements for the x86 code generator,
39748 related infrastructure improvements to help x86 code generation,
39749 value range propagation and other work, WE32k port.
39751 * Ulrich Weigand for work on the s390 port.
39753 * Zack Weinberg for major work on cpplib and various other bug fixes.
39755 * Matt Welsh for help with Linux Threads support in GCJ.
39757 * Urban Widmark for help fixing java.io.
39759 * Mark Wielaard for new Java library code and his work integrating
39762 * Dale Wiles helped port GCC to the Tahoe.
39764 * Bob Wilson from Tensilica, Inc. for the Xtensa port.
39766 * Jim Wilson for his direction via the steering committee, tackling
39767 hard problems in various places that nobody else wanted to work
39768 on, strength reduction and other loop optimizations.
39770 * Paul Woegerer and Tal Agmon for the CRX port.
39772 * Carlo Wood for various fixes.
39774 * Tom Wood for work on the m88k port.
39776 * Canqun Yang for work on GNU Fortran.
39778 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine
39779 description for the Tron architecture (specifically, the Gmicro).
39781 * Kevin Zachmann helped port GCC to the Tahoe.
39783 * Ayal Zaks for Swing Modulo Scheduling (SMS).
39785 * Xiaoqiang Zhang for work on GNU Fortran.
39787 * Gilles Zunino for help porting Java to Irix.
39790 The following people are recognized for their contributions to GNAT,
39791 the Ada front end of GCC:
39794 * Romain Berrendonner
39844 * Hristian Kirtchev
39887 The following people are recognized for their contributions of new
39888 features, bug reports, testing and integration of classpath/libgcj for
39890 * Lillian Angel for `JTree' implementation and lots Free Swing
39891 additions and bug fixes.
39893 * Wolfgang Baer for `GapContent' bug fixes.
39895 * Anthony Balkissoon for `JList', Free Swing 1.5 updates and mouse
39896 event fixes, lots of Free Swing work including `JTable' editing.
39898 * Stuart Ballard for RMI constant fixes.
39900 * Goffredo Baroncelli for `HTTPURLConnection' fixes.
39902 * Gary Benson for `MessageFormat' fixes.
39904 * Daniel Bonniot for `Serialization' fixes.
39906 * Chris Burdess for lots of gnu.xml and http protocol fixes, `StAX'
39907 and `DOM xml:id' support.
39909 * Ka-Hing Cheung for `TreePath' and `TreeSelection' fixes.
39911 * Archie Cobbs for build fixes, VM interface updates,
39912 `URLClassLoader' updates.
39914 * Kelley Cook for build fixes.
39916 * Martin Cordova for Suggestions for better `SocketTimeoutException'.
39918 * David Daney for `BitSet' bug fixes, `HttpURLConnection' rewrite
39921 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo
39922 2D support. Lots of imageio framework additions, lots of AWT and
39923 Free Swing bug fixes.
39925 * Jeroen Frijters for `ClassLoader' and nio cleanups, serialization
39926 fixes, better `Proxy' support, bug fixes and IKVM integration.
39928 * Santiago Gala for `AccessControlContext' fixes.
39930 * Nicolas Geoffray for `VMClassLoader' and `AccessController'
39933 * David Gilbert for `basic' and `metal' icon and plaf support and
39934 lots of documenting, Lots of Free Swing and metal theme additions.
39935 `MetalIconFactory' implementation.
39937 * Anthony Green for `MIDI' framework, `ALSA' and `DSSI' providers.
39939 * Andrew Haley for `Serialization' and `URLClassLoader' fixes, gcj
39942 * Kim Ho for `JFileChooser' implementation.
39944 * Andrew John Hughes for `Locale' and net fixes, URI RFC2986
39945 updates, `Serialization' fixes, `Properties' XML support and
39946 generic branch work, VMIntegration guide update.
39948 * Bastiaan Huisman for `TimeZone' bug fixing.
39950 * Andreas Jaeger for mprec updates.
39952 * Paul Jenner for better `-Werror' support.
39954 * Ito Kazumitsu for `NetworkInterface' implementation and updates.
39956 * Roman Kennke for `BoxLayout', `GrayFilter' and `SplitPane', plus
39957 bug fixes all over. Lots of Free Swing work including styled text.
39959 * Simon Kitching for `String' cleanups and optimization suggestions.
39961 * Michael Koch for configuration fixes, `Locale' updates, bug and
39964 * Guilhem Lavaux for configuration, thread and channel fixes and
39965 Kaffe integration. JCL native `Pointer' updates. Logger bug fixes.
39967 * David Lichteblau for JCL support library global/local reference
39970 * Aaron Luchko for JDWP updates and documentation fixes.
39972 * Ziga Mahkovec for `Graphics2D' upgraded to Cairo 0.5 and new regex
39975 * Sven de Marothy for BMP imageio support, CSS and `TextLayout'
39976 fixes. `GtkImage' rewrite, 2D, awt, free swing and date/time fixes
39977 and implementing the Qt4 peers.
39979 * Casey Marshall for crypto algorithm fixes, `FileChannel' lock,
39980 `SystemLogger' and `FileHandler' rotate implementations, NIO
39981 `FileChannel.map' support, security and policy updates.
39983 * Bryce McKinlay for RMI work.
39985 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus
39986 testing and documenting.
39988 * Kalle Olavi Niemitalo for build fixes.
39990 * Rainer Orth for build fixes.
39992 * Andrew Overholt for `File' locking fixes.
39994 * Ingo Proetel for `Image', `Logger' and `URLClassLoader' updates.
39996 * Olga Rodimina for `MenuSelectionManager' implementation.
39998 * Jan Roehrich for `BasicTreeUI' and `JTree' fixes.
40000 * Julian Scheid for documentation updates and gjdoc support.
40002 * Christian Schlichtherle for zip fixes and cleanups.
40004 * Robert Schuster for documentation updates and beans fixes,
40005 `TreeNode' enumerations and `ActionCommand' and various fixes, XML
40006 and URL, AWT and Free Swing bug fixes.
40008 * Keith Seitz for lots of JDWP work.
40010 * Christian Thalinger for 64-bit cleanups, Configuration and VM
40011 interface fixes and `CACAO' integration, `fdlibm' updates.
40013 * Gael Thomas for `VMClassLoader' boot packages support suggestions.
40015 * Andreas Tobler for Darwin and Solaris testing and fixing, `Qt4'
40016 support for Darwin/OS X, `Graphics2D' support, `gtk+' updates.
40018 * Dalibor Topic for better `DEBUG' support, build cleanups and Kaffe
40019 integration. `Qt4' build infrastructure, `SHA1PRNG' and
40020 `GdkPixbugDecoder' updates.
40022 * Tom Tromey for Eclipse integration, generics work, lots of bug
40023 fixes and gcj integration including coordinating The Big Merge.
40025 * Mark Wielaard for bug fixes, packaging and release management,
40026 `Clipboard' implementation, system call interrupts and network
40027 timeouts and `GdkPixpufDecoder' fixes.
40030 In addition to the above, all of which also contributed time and
40031 energy in testing GCC, we would like to thank the following for their
40032 contributions to testing:
40034 * Michael Abd-El-Malek
40044 * David Billinghurst
40048 * Stephane Bortzmeyer
40058 * Bradford Castalia
40080 * Charles-Antoine Gauthier
40102 * Kevin B. Hendricks
40106 * Christian Joensson
40114 * Anand Krishnaswamy
40116 * A. O. V. Le Blanc
40180 * Pedro A. M. Vazquez
40190 And finally we'd like to thank everyone who uses the compiler, provides
40191 feedback and generally reminds us why we're doing this work in the first
40195 File: gcc.info, Node: Option Index, Next: Keyword Index, Prev: Contributors, Up: Top
40200 GCC's command line options are indexed here without any initial `-' or
40201 `--'. Where an option has both positive and negative forms (such as
40202 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are
40203 indexed under the most appropriate form; it may sometimes be useful to
40204 look up both forms.
40209 * ###: Overall Options. (line 204)
40210 * -fdump-statistics: Debugging Options. (line 623)
40211 * A: Preprocessor Options.
40213 * all_load: Darwin Options. (line 112)
40214 * allowable_client: Darwin Options. (line 199)
40215 * ansi <1>: Non-bugs. (line 107)
40216 * ansi <2>: Other Builtins. (line 22)
40217 * ansi <3>: Preprocessor Options.
40219 * ansi <4>: C Dialect Options. (line 11)
40220 * ansi: Standards. (line 16)
40221 * arch_errors_fatal: Darwin Options. (line 116)
40222 * aux-info: C Dialect Options. (line 140)
40223 * b: Target Options. (line 13)
40224 * B: Directory Options. (line 41)
40225 * bcopy-builtin: PDP-11 Options. (line 32)
40226 * Bdynamic: VxWorks Options. (line 22)
40227 * bind_at_load: Darwin Options. (line 120)
40228 * Bstatic: VxWorks Options. (line 22)
40229 * bundle: Darwin Options. (line 125)
40230 * bundle_loader: Darwin Options. (line 129)
40231 * c: Link Options. (line 20)
40232 * C: Preprocessor Options.
40234 * c: Overall Options. (line 159)
40235 * client_name: Darwin Options. (line 199)
40236 * combine: Overall Options. (line 215)
40237 * compatibility_version: Darwin Options. (line 199)
40238 * coverage: Debugging Options. (line 272)
40239 * current_version: Darwin Options. (line 199)
40240 * D: Preprocessor Options.
40242 * d: Debugging Options. (line 336)
40243 * dA: Debugging Options. (line 539)
40244 * dD <1>: Preprocessor Options.
40246 * dD: Debugging Options. (line 543)
40247 * dead_strip: Darwin Options. (line 199)
40248 * dependency-file: Darwin Options. (line 199)
40249 * dH: Debugging Options. (line 547)
40250 * dI: Preprocessor Options.
40252 * dM: Preprocessor Options.
40254 * dm: Debugging Options. (line 550)
40255 * dN: Preprocessor Options.
40257 * dP: Debugging Options. (line 559)
40258 * dp: Debugging Options. (line 554)
40259 * dU: Preprocessor Options.
40261 * dumpmachine: Debugging Options. (line 952)
40262 * dumpspecs: Debugging Options. (line 960)
40263 * dumpversion: Debugging Options. (line 956)
40264 * dv: Debugging Options. (line 563)
40265 * dx: Debugging Options. (line 568)
40266 * dy: Debugging Options. (line 572)
40267 * dylib_file: Darwin Options. (line 199)
40268 * dylinker_install_name: Darwin Options. (line 199)
40269 * dynamic: Darwin Options. (line 199)
40270 * dynamiclib: Darwin Options. (line 133)
40271 * E <1>: Link Options. (line 20)
40272 * E: Overall Options. (line 180)
40273 * EB <1>: MIPS Options. (line 7)
40274 * EB: ARC Options. (line 12)
40275 * EL <1>: MIPS Options. (line 10)
40276 * EL: ARC Options. (line 9)
40277 * exported_symbols_list: Darwin Options. (line 199)
40278 * F: Darwin Options. (line 32)
40279 * fabi-version: C++ Dialect Options.
40281 * falign-functions: Optimize Options. (line 1184)
40282 * falign-jumps: Optimize Options. (line 1234)
40283 * falign-labels: Optimize Options. (line 1202)
40284 * falign-loops: Optimize Options. (line 1220)
40285 * fargument-alias: Code Gen Options. (line 413)
40286 * fargument-noalias: Code Gen Options. (line 413)
40287 * fargument-noalias-anything: Code Gen Options. (line 413)
40288 * fargument-noalias-global: Code Gen Options. (line 413)
40289 * fassociative-math: Optimize Options. (line 1419)
40290 * fasynchronous-unwind-tables: Code Gen Options. (line 64)
40291 * fauto-inc-dec: Optimize Options. (line 455)
40292 * fbounds-check: Code Gen Options. (line 15)
40293 * fbranch-probabilities: Optimize Options. (line 1545)
40294 * fbranch-target-load-optimize: Optimize Options. (line 1653)
40295 * fbranch-target-load-optimize2: Optimize Options. (line 1659)
40296 * fbtr-bb-exclusive: Optimize Options. (line 1663)
40297 * fcall-saved: Code Gen Options. (line 262)
40298 * fcall-used: Code Gen Options. (line 248)
40299 * fcaller-saves: Optimize Options. (line 676)
40300 * fcheck-data-deps: Optimize Options. (line 897)
40301 * fcheck-new: C++ Dialect Options.
40303 * fcommon: Variable Attributes.
40305 * fcond-mismatch: C Dialect Options. (line 258)
40306 * fconserve-space: C++ Dialect Options.
40308 * fconserve-stack: Optimize Options. (line 689)
40309 * fconstant-string-class: Objective-C and Objective-C++ Dialect Options.
40311 * fcprop-registers: Optimize Options. (line 1292)
40312 * fcrossjumping: Optimize Options. (line 448)
40313 * fcse-follow-jumps: Optimize Options. (line 376)
40314 * fcse-skip-blocks: Optimize Options. (line 385)
40315 * fcx-fortran-rules: Optimize Options. (line 1531)
40316 * fcx-limited-range: Optimize Options. (line 1519)
40317 * fdata-sections: Optimize Options. (line 1634)
40318 * fdbg-cnt: Debugging Options. (line 325)
40319 * fdbg-cnt-list: Debugging Options. (line 322)
40320 * fdce: Optimize Options. (line 461)
40321 * fdebug-prefix-map: Debugging Options. (line 219)
40322 * fdelayed-branch: Optimize Options. (line 557)
40323 * fdelete-null-pointer-checks: Optimize Options. (line 484)
40324 * fdiagnostics-show-location: Language Independent Options.
40326 * fdiagnostics-show-option: Language Independent Options.
40328 * fdirectives-only: Preprocessor Options.
40330 * fdollars-in-identifiers <1>: Interoperation. (line 146)
40331 * fdollars-in-identifiers: Preprocessor Options.
40333 * fdse: Optimize Options. (line 465)
40334 * fdump-class-hierarchy: Debugging Options. (line 597)
40335 * fdump-ipa: Debugging Options. (line 605)
40336 * fdump-noaddr: Debugging Options. (line 575)
40337 * fdump-rtl-alignments: Debugging Options. (line 351)
40338 * fdump-rtl-all: Debugging Options. (line 536)
40339 * fdump-rtl-asmcons: Debugging Options. (line 354)
40340 * fdump-rtl-auto_inc_dec: Debugging Options. (line 358)
40341 * fdump-rtl-barriers: Debugging Options. (line 362)
40342 * fdump-rtl-bbpart: Debugging Options. (line 365)
40343 * fdump-rtl-bbro: Debugging Options. (line 368)
40344 * fdump-rtl-btl2: Debugging Options. (line 372)
40345 * fdump-rtl-bypass: Debugging Options. (line 376)
40346 * fdump-rtl-ce1: Debugging Options. (line 387)
40347 * fdump-rtl-ce2: Debugging Options. (line 387)
40348 * fdump-rtl-ce3: Debugging Options. (line 387)
40349 * fdump-rtl-combine: Debugging Options. (line 379)
40350 * fdump-rtl-compgotos: Debugging Options. (line 382)
40351 * fdump-rtl-cprop_hardreg: Debugging Options. (line 391)
40352 * fdump-rtl-csa: Debugging Options. (line 394)
40353 * fdump-rtl-cse1: Debugging Options. (line 398)
40354 * fdump-rtl-cse2: Debugging Options. (line 398)
40355 * fdump-rtl-dbr: Debugging Options. (line 405)
40356 * fdump-rtl-dce: Debugging Options. (line 402)
40357 * fdump-rtl-dce1: Debugging Options. (line 409)
40358 * fdump-rtl-dce2: Debugging Options. (line 409)
40359 * fdump-rtl-dfinish: Debugging Options. (line 533)
40360 * fdump-rtl-dfinit: Debugging Options. (line 533)
40361 * fdump-rtl-eh: Debugging Options. (line 413)
40362 * fdump-rtl-eh_ranges: Debugging Options. (line 416)
40363 * fdump-rtl-expand: Debugging Options. (line 419)
40364 * fdump-rtl-fwprop1: Debugging Options. (line 423)
40365 * fdump-rtl-fwprop2: Debugging Options. (line 423)
40366 * fdump-rtl-gcse1: Debugging Options. (line 428)
40367 * fdump-rtl-gcse2: Debugging Options. (line 428)
40368 * fdump-rtl-init-regs: Debugging Options. (line 432)
40369 * fdump-rtl-initvals: Debugging Options. (line 435)
40370 * fdump-rtl-into_cfglayout: Debugging Options. (line 438)
40371 * fdump-rtl-ira: Debugging Options. (line 441)
40372 * fdump-rtl-jump: Debugging Options. (line 444)
40373 * fdump-rtl-loop2: Debugging Options. (line 447)
40374 * fdump-rtl-mach: Debugging Options. (line 451)
40375 * fdump-rtl-mode_sw: Debugging Options. (line 455)
40376 * fdump-rtl-outof_cfglayout: Debugging Options. (line 461)
40377 * fdump-rtl-peephole2: Debugging Options. (line 464)
40378 * fdump-rtl-postreload: Debugging Options. (line 467)
40379 * fdump-rtl-pro_and_epilogue: Debugging Options. (line 470)
40380 * fdump-rtl-regclass: Debugging Options. (line 533)
40381 * fdump-rtl-regmove: Debugging Options. (line 473)
40382 * fdump-rtl-rnreg: Debugging Options. (line 458)
40383 * fdump-rtl-sched1: Debugging Options. (line 477)
40384 * fdump-rtl-sched2: Debugging Options. (line 477)
40385 * fdump-rtl-see: Debugging Options. (line 481)
40386 * fdump-rtl-seqabstr: Debugging Options. (line 484)
40387 * fdump-rtl-shorten: Debugging Options. (line 487)
40388 * fdump-rtl-sibling: Debugging Options. (line 490)
40389 * fdump-rtl-sms: Debugging Options. (line 503)
40390 * fdump-rtl-split1: Debugging Options. (line 497)
40391 * fdump-rtl-split2: Debugging Options. (line 497)
40392 * fdump-rtl-split3: Debugging Options. (line 497)
40393 * fdump-rtl-split4: Debugging Options. (line 497)
40394 * fdump-rtl-split5: Debugging Options. (line 497)
40395 * fdump-rtl-stack: Debugging Options. (line 507)
40396 * fdump-rtl-subreg1: Debugging Options. (line 513)
40397 * fdump-rtl-subreg2: Debugging Options. (line 513)
40398 * fdump-rtl-subregs_of_mode_finish: Debugging Options. (line 533)
40399 * fdump-rtl-subregs_of_mode_init: Debugging Options. (line 533)
40400 * fdump-rtl-unshare: Debugging Options. (line 517)
40401 * fdump-rtl-vartrack: Debugging Options. (line 520)
40402 * fdump-rtl-vregs: Debugging Options. (line 523)
40403 * fdump-rtl-web: Debugging Options. (line 526)
40404 * fdump-translation-unit: Debugging Options. (line 588)
40405 * fdump-tree: Debugging Options. (line 634)
40406 * fdump-tree-alias: Debugging Options. (line 719)
40407 * fdump-tree-all: Debugging Options. (line 804)
40408 * fdump-tree-ccp: Debugging Options. (line 723)
40409 * fdump-tree-cfg: Debugging Options. (line 699)
40410 * fdump-tree-ch: Debugging Options. (line 711)
40411 * fdump-tree-copyprop: Debugging Options. (line 739)
40412 * fdump-tree-copyrename: Debugging Options. (line 785)
40413 * fdump-tree-dce: Debugging Options. (line 747)
40414 * fdump-tree-dom: Debugging Options. (line 765)
40415 * fdump-tree-dse: Debugging Options. (line 770)
40416 * fdump-tree-forwprop: Debugging Options. (line 780)
40417 * fdump-tree-fre: Debugging Options. (line 735)
40418 * fdump-tree-gimple: Debugging Options. (line 694)
40419 * fdump-tree-mudflap: Debugging Options. (line 751)
40420 * fdump-tree-nrv: Debugging Options. (line 790)
40421 * fdump-tree-phiopt: Debugging Options. (line 775)
40422 * fdump-tree-pre: Debugging Options. (line 731)
40423 * fdump-tree-sink: Debugging Options. (line 761)
40424 * fdump-tree-sra: Debugging Options. (line 756)
40425 * fdump-tree-ssa: Debugging Options. (line 715)
40426 * fdump-tree-store_copyprop: Debugging Options. (line 743)
40427 * fdump-tree-storeccp: Debugging Options. (line 727)
40428 * fdump-tree-vcg: Debugging Options. (line 703)
40429 * fdump-tree-vect: Debugging Options. (line 795)
40430 * fdump-tree-vrp: Debugging Options. (line 800)
40431 * fdump-unnumbered: Debugging Options. (line 581)
40432 * fdwarf2-cfi-asm: Debugging Options. (line 223)
40433 * fdyn-ipa: Optimize Options. (line 1344)
40434 * fearly-inlining: Optimize Options. (line 220)
40435 * feliminate-dwarf2-dups: Debugging Options. (line 136)
40436 * feliminate-unused-debug-symbols: Debugging Options. (line 52)
40437 * feliminate-unused-debug-types: Debugging Options. (line 964)
40438 * fexceptions: Code Gen Options. (line 34)
40439 * fexec-charset: Preprocessor Options.
40441 * fexpensive-optimizations: Optimize Options. (line 497)
40442 * fextended-identifiers: Preprocessor Options.
40444 * ffast-math: Optimize Options. (line 1370)
40445 * ffinite-math-only: Optimize Options. (line 1443)
40446 * ffix-and-continue: Darwin Options. (line 106)
40447 * ffixed: Code Gen Options. (line 236)
40448 * ffloat-store <1>: Disappointments. (line 77)
40449 * ffloat-store: Optimize Options. (line 1356)
40450 * ffor-scope: C++ Dialect Options.
40452 * fforward-propagate: Optimize Options. (line 149)
40453 * ffreestanding <1>: Function Attributes.
40455 * ffreestanding <2>: Warning Options. (line 194)
40456 * ffreestanding <3>: C Dialect Options. (line 211)
40457 * ffreestanding: Standards. (line 84)
40458 * ffriend-injection: C++ Dialect Options.
40460 * ffunction-sections: Optimize Options. (line 1634)
40461 * fgcse: Optimize Options. (line 399)
40462 * fgcse-after-reload: Optimize Options. (line 435)
40463 * fgcse-las: Optimize Options. (line 428)
40464 * fgcse-lm: Optimize Options. (line 410)
40465 * fgcse-sm: Optimize Options. (line 419)
40466 * fgnu-runtime: Objective-C and Objective-C++ Dialect Options.
40468 * fgnu89-inline: C Dialect Options. (line 120)
40469 * fhosted: C Dialect Options. (line 204)
40470 * fif-conversion: Optimize Options. (line 469)
40471 * fif-conversion2: Optimize Options. (line 478)
40472 * filelist: Darwin Options. (line 199)
40473 * findirect-data: Darwin Options. (line 106)
40474 * findirect-inlining: Optimize Options. (line 193)
40475 * finhibit-size-directive: Code Gen Options. (line 158)
40476 * finline-functions: Optimize Options. (line 201)
40477 * finline-functions-called-once: Optimize Options. (line 212)
40478 * finline-limit: Optimize Options. (line 230)
40479 * finline-small-functions: Optimize Options. (line 185)
40480 * finput-charset: Preprocessor Options.
40482 * finstrument-functions <1>: Function Attributes.
40484 * finstrument-functions: Code Gen Options. (line 292)
40485 * finstrument-functions-exclude-file-list: Code Gen Options. (line 329)
40486 * finstrument-functions-exclude-function-list: Code Gen Options.
40488 * fipa-cp: Optimize Options. (line 742)
40489 * fipa-cp-clone: Optimize Options. (line 750)
40490 * fipa-matrix-reorg: Optimize Options. (line 760)
40491 * fipa-pta: Optimize Options. (line 738)
40492 * fipa-pure-const: Optimize Options. (line 715)
40493 * fipa-reference: Optimize Options. (line 719)
40494 * fipa-struct-reorg: Optimize Options. (line 723)
40495 * fira-coalesce: Optimize Options. (line 536)
40496 * fira-verbose: Optimize Options. (line 552)
40497 * fivopts: Optimize Options. (line 933)
40498 * fkeep-inline-functions <1>: Inline. (line 51)
40499 * fkeep-inline-functions: Optimize Options. (line 256)
40500 * fkeep-static-consts: Optimize Options. (line 263)
40501 * flat_namespace: Darwin Options. (line 199)
40502 * flax-vector-conversions: C Dialect Options. (line 263)
40503 * fleading-underscore: Code Gen Options. (line 430)
40504 * fmem-report: Debugging Options. (line 247)
40505 * fmerge-all-constants: Optimize Options. (line 282)
40506 * fmerge-constants: Optimize Options. (line 272)
40507 * fmerge-debug-strings: Debugging Options. (line 211)
40508 * fmessage-length: Language Independent Options.
40510 * fmodulo-sched: Optimize Options. (line 293)
40511 * fmodulo-sched-allow-regmoves: Optimize Options. (line 298)
40512 * fmove-loop-invariants: Optimize Options. (line 1624)
40513 * fms-extensions <1>: Unnamed Fields. (line 37)
40514 * fms-extensions <2>: C++ Dialect Options.
40516 * fms-extensions: C Dialect Options. (line 229)
40517 * fmudflap: Optimize Options. (line 338)
40518 * fmudflapir: Optimize Options. (line 338)
40519 * fmudflapth: Optimize Options. (line 338)
40520 * fnext-runtime: Objective-C and Objective-C++ Dialect Options.
40522 * fno-access-control: C++ Dialect Options.
40524 * fno-asm: C Dialect Options. (line 156)
40525 * fno-branch-count-reg: Optimize Options. (line 305)
40526 * fno-builtin <1>: Other Builtins. (line 14)
40527 * fno-builtin <2>: Function Attributes.
40529 * fno-builtin <3>: Warning Options. (line 194)
40530 * fno-builtin: C Dialect Options. (line 170)
40531 * fno-common <1>: Variable Attributes.
40533 * fno-common: Code Gen Options. (line 135)
40534 * fno-default-inline <1>: Inline. (line 71)
40535 * fno-default-inline <2>: Optimize Options. (line 134)
40536 * fno-default-inline: C++ Dialect Options.
40538 * fno-defer-pop: Optimize Options. (line 141)
40539 * fno-dwarf2-cfi-asm: Debugging Options. (line 223)
40540 * fno-elide-constructors: C++ Dialect Options.
40542 * fno-enforce-eh-specs: C++ Dialect Options.
40544 * fno-for-scope: C++ Dialect Options.
40546 * fno-function-cse: Optimize Options. (line 315)
40547 * fno-gnu-keywords: C++ Dialect Options.
40549 * fno-guess-branch-probability: Optimize Options. (line 1056)
40550 * fno-ident: Code Gen Options. (line 155)
40551 * fno-implement-inlines <1>: C++ Interface. (line 75)
40552 * fno-implement-inlines: C++ Dialect Options.
40554 * fno-implicit-inline-templates: C++ Dialect Options.
40556 * fno-implicit-templates <1>: Template Instantiation.
40558 * fno-implicit-templates: C++ Dialect Options.
40560 * fno-inline: Optimize Options. (line 179)
40561 * fno-ira-share-save-slots: Optimize Options. (line 540)
40562 * fno-ira-share-spill-slots: Optimize Options. (line 546)
40563 * fno-jump-tables: Code Gen Options. (line 228)
40564 * fno-math-errno: Optimize Options. (line 1384)
40565 * fno-merge-debug-strings: Debugging Options. (line 211)
40566 * fno-nil-receivers: Objective-C and Objective-C++ Dialect Options.
40568 * fno-nonansi-builtins: C++ Dialect Options.
40570 * fno-operator-names: C++ Dialect Options.
40572 * fno-optional-diags: C++ Dialect Options.
40574 * fno-peephole: Optimize Options. (line 1047)
40575 * fno-peephole2: Optimize Options. (line 1047)
40576 * fno-rtti: C++ Dialect Options.
40578 * fno-sched-interblock: Optimize Options. (line 583)
40579 * fno-sched-spec: Optimize Options. (line 588)
40580 * fno-show-column: Preprocessor Options.
40582 * fno-signed-bitfields: C Dialect Options. (line 296)
40583 * fno-signed-zeros: Optimize Options. (line 1455)
40584 * fno-stack-limit: Code Gen Options. (line 396)
40585 * fno-threadsafe-statics: C++ Dialect Options.
40587 * fno-toplevel-reorder: Optimize Options. (line 1254)
40588 * fno-trapping-math: Optimize Options. (line 1465)
40589 * fno-unsigned-bitfields: C Dialect Options. (line 296)
40590 * fno-use-cxa-get-exception-ptr: C++ Dialect Options.
40592 * fno-weak: C++ Dialect Options.
40594 * fno-working-directory: Preprocessor Options.
40596 * fno-zero-initialized-in-bss: Optimize Options. (line 326)
40597 * fnon-call-exceptions: Code Gen Options. (line 48)
40598 * fobjc-call-cxx-cdtors: Objective-C and Objective-C++ Dialect Options.
40600 * fobjc-direct-dispatch: Objective-C and Objective-C++ Dialect Options.
40602 * fobjc-exceptions: Objective-C and Objective-C++ Dialect Options.
40604 * fobjc-gc: Objective-C and Objective-C++ Dialect Options.
40606 * fomit-frame-pointer: Optimize Options. (line 158)
40607 * fopenmp: C Dialect Options. (line 221)
40608 * foptimize-register-move: Optimize Options. (line 504)
40609 * foptimize-sibling-calls: Optimize Options. (line 174)
40610 * force_cpusubtype_ALL: Darwin Options. (line 138)
40611 * force_flat_namespace: Darwin Options. (line 199)
40612 * fpack-struct: Code Gen Options. (line 279)
40613 * fpcc-struct-return <1>: Incompatibilities. (line 170)
40614 * fpcc-struct-return: Code Gen Options. (line 70)
40615 * fpch-deps: Preprocessor Options.
40617 * fpch-preprocess: Preprocessor Options.
40619 * fpeel-loops: Optimize Options. (line 1616)
40620 * fpermissive: C++ Dialect Options.
40622 * fPIC: Code Gen Options. (line 205)
40623 * fpic: Code Gen Options. (line 184)
40624 * fPIE: Code Gen Options. (line 218)
40625 * fpie: Code Gen Options. (line 218)
40626 * fpost-ipa-mem-report: Debugging Options. (line 253)
40627 * fpre-ipa-mem-report: Debugging Options. (line 251)
40628 * fpredictive-commoning: Optimize Options. (line 1029)
40629 * fprefetch-loop-arrays: Optimize Options. (line 1036)
40630 * fpreprocessed: Preprocessor Options.
40632 * fprofile-arcs <1>: Other Builtins. (line 242)
40633 * fprofile-arcs: Debugging Options. (line 257)
40634 * fprofile-correction: Optimize Options. (line 1299)
40635 * fprofile-dir: Optimize Options. (line 1306)
40636 * fprofile-generate: Optimize Options. (line 1316)
40637 * fprofile-use: Optimize Options. (line 1329)
40638 * fprofile-values: Optimize Options. (line 1564)
40639 * frandom-string: Debugging Options. (line 833)
40640 * freciprocal-math: Optimize Options. (line 1434)
40641 * frecord-gcc-switches: Code Gen Options. (line 174)
40642 * freg-struct-return: Code Gen Options. (line 88)
40643 * fregmove: Optimize Options. (line 504)
40644 * frename-registers: Optimize Options. (line 1583)
40645 * freorder-blocks: Optimize Options. (line 1073)
40646 * freorder-blocks-and-partition: Optimize Options. (line 1079)
40647 * freorder-functions: Optimize Options. (line 1090)
40648 * freplace-objc-classes: Objective-C and Objective-C++ Dialect Options.
40650 * frepo <1>: Template Instantiation.
40652 * frepo: C++ Dialect Options.
40654 * frerun-cse-after-loop: Optimize Options. (line 393)
40655 * freschedule-modulo-scheduled-loops: Optimize Options. (line 652)
40656 * frounding-math: Optimize Options. (line 1480)
40657 * fsched-spec-load: Optimize Options. (line 593)
40658 * fsched-spec-load-dangerous: Optimize Options. (line 598)
40659 * fsched-stalled-insns: Optimize Options. (line 604)
40660 * fsched-stalled-insns-dep: Optimize Options. (line 614)
40661 * fsched-verbose: Debugging Options. (line 843)
40662 * fsched2-use-superblocks: Optimize Options. (line 624)
40663 * fsched2-use-traces: Optimize Options. (line 635)
40664 * fschedule-insns: Optimize Options. (line 564)
40665 * fschedule-insns2: Optimize Options. (line 574)
40666 * fsection-anchors: Optimize Options. (line 1679)
40667 * fsee: Optimize Options. (line 647)
40668 * fsel-sched-pipelining: Optimize Options. (line 666)
40669 * fsel-sched-pipelining-outer-loops: Optimize Options. (line 671)
40670 * fselective-scheduling: Optimize Options. (line 658)
40671 * fselective-scheduling2: Optimize Options. (line 662)
40672 * fshort-double: Code Gen Options. (line 117)
40673 * fshort-enums <1>: Non-bugs. (line 42)
40674 * fshort-enums <2>: Type Attributes. (line 113)
40675 * fshort-enums <3>: Structures unions enumerations and bit-fields implementation.
40677 * fshort-enums: Code Gen Options. (line 106)
40678 * fshort-wchar: Code Gen Options. (line 125)
40679 * fsignaling-nans: Optimize Options. (line 1500)
40680 * fsigned-bitfields <1>: Non-bugs. (line 57)
40681 * fsigned-bitfields: C Dialect Options. (line 296)
40682 * fsigned-char <1>: Characters implementation.
40684 * fsigned-char: C Dialect Options. (line 286)
40685 * fsingle-precision-constant: Optimize Options. (line 1515)
40686 * fsplit-ivs-in-unroller: Optimize Options. (line 1010)
40687 * fsplit-wide-types: Optimize Options. (line 368)
40688 * fstack-check: Code Gen Options. (line 357)
40689 * fstack-limit-register: Code Gen Options. (line 396)
40690 * fstack-limit-symbol: Code Gen Options. (line 396)
40691 * fstack-protector: Optimize Options. (line 1667)
40692 * fstack-protector-all: Optimize Options. (line 1676)
40693 * fstats: C++ Dialect Options.
40695 * fstrict-aliasing: Optimize Options. (line 1103)
40696 * fstrict-overflow: Optimize Options. (line 1149)
40697 * fsyntax-only: Warning Options. (line 14)
40698 * ftabstop: Preprocessor Options.
40700 * ftemplate-depth: C++ Dialect Options.
40702 * ftest-coverage: Debugging Options. (line 313)
40703 * fthread-jumps: Optimize Options. (line 359)
40704 * ftime-report: Debugging Options. (line 243)
40705 * ftls-model: Code Gen Options. (line 441)
40706 * ftracer: Optimize Options. (line 993)
40707 * ftrapv: Code Gen Options. (line 22)
40708 * ftree-builtin-call-dce: Optimize Options. (line 788)
40709 * ftree-ccp: Optimize Options. (line 774)
40710 * ftree-ch: Optimize Options. (line 808)
40711 * ftree-copy-prop: Optimize Options. (line 710)
40712 * ftree-copyrename: Optimize Options. (line 953)
40713 * ftree-dce: Optimize Options. (line 784)
40714 * ftree-dominator-opts: Optimize Options. (line 794)
40715 * ftree-dse: Optimize Options. (line 801)
40716 * ftree-fre: Optimize Options. (line 703)
40717 * ftree-loop-im: Optimize Options. (line 918)
40718 * ftree-loop-ivcanon: Optimize Options. (line 927)
40719 * ftree-loop-linear: Optimize Options. (line 819)
40720 * ftree-loop-optimize: Optimize Options. (line 815)
40721 * ftree-parallelize-loops: Optimize Options. (line 938)
40722 * ftree-pre: Optimize Options. (line 699)
40723 * ftree-reassoc: Optimize Options. (line 695)
40724 * ftree-sink: Optimize Options. (line 770)
40725 * ftree-sra: Optimize Options. (line 947)
40726 * ftree-ter: Optimize Options. (line 960)
40727 * ftree-vect-loop-version: Optimize Options. (line 972)
40728 * ftree-vectorize: Optimize Options. (line 968)
40729 * ftree-vectorizer-verbose: Debugging Options. (line 808)
40730 * ftree-vrp: Optimize Options. (line 984)
40731 * funit-at-a-time: Optimize Options. (line 1247)
40732 * funroll-all-loops: Optimize Options. (line 1004)
40733 * funroll-loops: Optimize Options. (line 998)
40734 * funsafe-loop-optimizations: Optimize Options. (line 440)
40735 * funsafe-math-optimizations: Optimize Options. (line 1402)
40736 * funsigned-bitfields <1>: Non-bugs. (line 57)
40737 * funsigned-bitfields <2>: Structures unions enumerations and bit-fields implementation.
40739 * funsigned-bitfields: C Dialect Options. (line 296)
40740 * funsigned-char <1>: Characters implementation.
40742 * funsigned-char: C Dialect Options. (line 268)
40743 * funswitch-loops: Optimize Options. (line 1628)
40744 * funwind-tables: Code Gen Options. (line 57)
40745 * fuse-cxa-atexit: C++ Dialect Options.
40747 * fvar-tracking: Debugging Options. (line 888)
40748 * fvariable-expansion-in-unroller: Optimize Options. (line 1024)
40749 * fvect-cost-model: Optimize Options. (line 981)
40750 * fverbose-asm: Code Gen Options. (line 165)
40751 * fvisibility: Code Gen Options. (line 449)
40752 * fvisibility-inlines-hidden: C++ Dialect Options.
40754 * fvisibility-ms-compat: C++ Dialect Options.
40756 * fvpt: Optimize Options. (line 1574)
40757 * fweb: Optimize Options. (line 1266)
40758 * fwhole-program: Optimize Options. (line 1277)
40759 * fwide-exec-charset: Preprocessor Options.
40761 * fworking-directory: Preprocessor Options.
40763 * fwrapv: Code Gen Options. (line 26)
40764 * fzero-link: Objective-C and Objective-C++ Dialect Options.
40766 * G <1>: System V Options. (line 10)
40767 * G <2>: RS/6000 and PowerPC Options.
40769 * G <3>: MIPS Options. (line 314)
40770 * G: M32R/D Options. (line 57)
40771 * g: Debugging Options. (line 10)
40772 * gcoff: Debugging Options. (line 70)
40773 * gdwarf-2: Debugging Options. (line 88)
40774 * gdwarf-4: Debugging Options. (line 95)
40775 * gen-decls: Objective-C and Objective-C++ Dialect Options.
40777 * gfull: Darwin Options. (line 71)
40778 * ggdb: Debugging Options. (line 38)
40779 * gnu-ld: HPPA Options. (line 111)
40780 * gstabs: Debugging Options. (line 44)
40781 * gstabs+: Debugging Options. (line 64)
40782 * gused: Darwin Options. (line 66)
40783 * gvms: Debugging Options. (line 103)
40784 * gxcoff: Debugging Options. (line 75)
40785 * gxcoff+: Debugging Options. (line 80)
40786 * H: Preprocessor Options.
40788 * headerpad_max_install_names: Darwin Options. (line 199)
40789 * help <1>: Preprocessor Options.
40791 * help: Overall Options. (line 231)
40792 * hp-ld: HPPA Options. (line 123)
40793 * I <1>: Directory Options. (line 10)
40794 * I: Preprocessor Options.
40796 * I- <1>: Directory Options. (line 107)
40797 * I-: Preprocessor Options.
40799 * idirafter: Preprocessor Options.
40801 * iframework: Darwin Options. (line 59)
40802 * imacros: Preprocessor Options.
40804 * image_base: Darwin Options. (line 199)
40805 * imultilib: Preprocessor Options.
40807 * include: Preprocessor Options.
40809 * init: Darwin Options. (line 199)
40810 * install_name: Darwin Options. (line 199)
40811 * iprefix: Preprocessor Options.
40813 * iquote <1>: Directory Options. (line 31)
40814 * iquote: Preprocessor Options.
40816 * isysroot: Preprocessor Options.
40818 * isystem: Preprocessor Options.
40820 * iwithprefix: Preprocessor Options.
40822 * iwithprefixbefore: Preprocessor Options.
40824 * keep_private_externs: Darwin Options. (line 199)
40825 * L: Directory Options. (line 37)
40826 * l: Link Options. (line 26)
40827 * lobjc: Link Options. (line 53)
40828 * M: Preprocessor Options.
40830 * m1: SH Options. (line 9)
40831 * m10: PDP-11 Options. (line 29)
40832 * m128bit-long-double: i386 and x86-64 Options.
40834 * m16-bit: CRIS Options. (line 64)
40835 * m2: SH Options. (line 12)
40836 * m210: MCore Options. (line 43)
40837 * m3: SH Options. (line 18)
40838 * m31: S/390 and zSeries Options.
40840 * m32 <1>: SPARC Options. (line 191)
40841 * m32 <2>: RS/6000 and PowerPC Options.
40843 * m32: i386 and x86-64 Options.
40845 * m32-bit: CRIS Options. (line 64)
40846 * m32r: M32R/D Options. (line 15)
40847 * m32r2: M32R/D Options. (line 9)
40848 * m32rx: M32R/D Options. (line 12)
40849 * m340: MCore Options. (line 43)
40850 * m3dnow: i386 and x86-64 Options.
40852 * m3e: SH Options. (line 21)
40853 * m4: SH Options. (line 35)
40854 * m4-nofpu: SH Options. (line 24)
40855 * m4-single: SH Options. (line 31)
40856 * m4-single-only: SH Options. (line 27)
40857 * m40: PDP-11 Options. (line 23)
40858 * m45: PDP-11 Options. (line 26)
40859 * m4a: SH Options. (line 50)
40860 * m4a-nofpu: SH Options. (line 38)
40861 * m4a-single: SH Options. (line 46)
40862 * m4a-single-only: SH Options. (line 42)
40863 * m4al: SH Options. (line 53)
40864 * m4byte-functions: MCore Options. (line 27)
40865 * m5200: M680x0 Options. (line 143)
40866 * m5206e: M680x0 Options. (line 152)
40867 * m528x: M680x0 Options. (line 156)
40868 * m5307: M680x0 Options. (line 160)
40869 * m5407: M680x0 Options. (line 164)
40870 * m64 <1>: SPARC Options. (line 191)
40871 * m64 <2>: S/390 and zSeries Options.
40873 * m64 <3>: RS/6000 and PowerPC Options.
40875 * m64: i386 and x86-64 Options.
40877 * m68000: M680x0 Options. (line 91)
40878 * m68010: M680x0 Options. (line 99)
40879 * m68020: M680x0 Options. (line 105)
40880 * m68020-40: M680x0 Options. (line 174)
40881 * m68020-60: M680x0 Options. (line 183)
40882 * m68030: M680x0 Options. (line 110)
40883 * m68040: M680x0 Options. (line 115)
40884 * m68060: M680x0 Options. (line 124)
40885 * m6811: M68hc1x Options. (line 13)
40886 * m6812: M68hc1x Options. (line 18)
40887 * m68881: M680x0 Options. (line 193)
40888 * m68hc11: M68hc1x Options. (line 13)
40889 * m68hc12: M68hc1x Options. (line 18)
40890 * m68hcs12: M68hc1x Options. (line 23)
40891 * m68S12: M68hc1x Options. (line 23)
40892 * m8-bit: CRIS Options. (line 64)
40893 * m96bit-long-double: i386 and x86-64 Options.
40895 * mabi <1>: RS/6000 and PowerPC Options.
40897 * mabi: ARM Options. (line 10)
40898 * mabi-mmixware: MMIX Options. (line 20)
40899 * mabi=32: MIPS Options. (line 129)
40900 * mabi=64: MIPS Options. (line 129)
40901 * mabi=eabi: MIPS Options. (line 129)
40902 * mabi=gnu: MMIX Options. (line 20)
40903 * mabi=ibmlongdouble: RS/6000 and PowerPC Options.
40905 * mabi=ieeelongdouble: RS/6000 and PowerPC Options.
40907 * mabi=n32: MIPS Options. (line 129)
40908 * mabi=no-spe: RS/6000 and PowerPC Options.
40910 * mabi=o64: MIPS Options. (line 129)
40911 * mabi=spe: RS/6000 and PowerPC Options.
40913 * mabicalls: MIPS Options. (line 153)
40914 * mabort-on-noreturn: ARM Options. (line 149)
40915 * mabshi: PDP-11 Options. (line 55)
40916 * mac0: PDP-11 Options. (line 16)
40917 * macc-4: FRV Options. (line 113)
40918 * macc-8: FRV Options. (line 116)
40919 * maccumulate-outgoing-args: i386 and x86-64 Options.
40921 * madjust-unroll: SH Options. (line 196)
40922 * mads: RS/6000 and PowerPC Options.
40924 * maix-struct-return: RS/6000 and PowerPC Options.
40926 * maix32: RS/6000 and PowerPC Options.
40928 * maix64: RS/6000 and PowerPC Options.
40930 * malign-300: H8/300 Options. (line 31)
40931 * malign-double: i386 and x86-64 Options.
40933 * malign-int: M680x0 Options. (line 263)
40934 * malign-labels: FRV Options. (line 104)
40935 * malign-loops: M32R/D Options. (line 73)
40936 * malign-natural: RS/6000 and PowerPC Options.
40938 * malign-power: RS/6000 and PowerPC Options.
40940 * malloc-cc: FRV Options. (line 25)
40941 * malpha-as: DEC Alpha Options. (line 159)
40942 * maltivec: RS/6000 and PowerPC Options.
40944 * mam33: MN10300 Options. (line 17)
40945 * mandroid: ARM Options. (line 264)
40946 * mapcs: ARM Options. (line 22)
40947 * mapcs-frame: ARM Options. (line 14)
40948 * mapp-regs <1>: V850 Options. (line 57)
40949 * mapp-regs: SPARC Options. (line 10)
40950 * march <1>: S/390 and zSeries Options.
40952 * march <2>: MIPS Options. (line 14)
40953 * march <3>: M680x0 Options. (line 12)
40954 * march <4>: i386 and x86-64 Options.
40956 * march <5>: HPPA Options. (line 9)
40957 * march <6>: CRIS Options. (line 10)
40958 * march: ARM Options. (line 112)
40959 * masm=DIALECT: i386 and x86-64 Options.
40961 * mauto-incdec: M68hc1x Options. (line 26)
40962 * mauto-pic: IA-64 Options. (line 50)
40963 * mavoid-indexed-addresses: RS/6000 and PowerPC Options.
40965 * mb: SH Options. (line 58)
40966 * mbackchain: S/390 and zSeries Options.
40968 * mbase-addresses: MMIX Options. (line 54)
40969 * mbcopy: PDP-11 Options. (line 36)
40970 * mbig: RS/6000 and PowerPC Options.
40972 * mbig-endian <1>: RS/6000 and PowerPC Options.
40974 * mbig-endian <2>: MCore Options. (line 39)
40975 * mbig-endian <3>: IA-64 Options. (line 9)
40976 * mbig-endian: ARM Options. (line 72)
40977 * mbig-switch <1>: V850 Options. (line 52)
40978 * mbig-switch: HPPA Options. (line 23)
40979 * mbigtable: SH Options. (line 74)
40980 * mbit-align: RS/6000 and PowerPC Options.
40982 * mbitfield: M680x0 Options. (line 231)
40983 * mbitops: SH Options. (line 78)
40984 * mbranch-cheap: PDP-11 Options. (line 65)
40985 * mbranch-cost: MIPS Options. (line 610)
40986 * mbranch-cost=NUMBER: M32R/D Options. (line 82)
40987 * mbranch-expensive: PDP-11 Options. (line 61)
40988 * mbranch-hints: SPU Options. (line 27)
40989 * mbranch-likely: MIPS Options. (line 617)
40990 * mbranch-predict: MMIX Options. (line 49)
40991 * mbss-plt: RS/6000 and PowerPC Options.
40993 * mbuild-constants: DEC Alpha Options. (line 142)
40994 * mbwx: DEC Alpha Options. (line 171)
40995 * mc68000: M680x0 Options. (line 91)
40996 * mc68020: M680x0 Options. (line 105)
40997 * mcall-gnu: RS/6000 and PowerPC Options.
40999 * mcall-linux: RS/6000 and PowerPC Options.
41001 * mcall-netbsd: RS/6000 and PowerPC Options.
41003 * mcall-prologues: AVR Options. (line 43)
41004 * mcall-solaris: RS/6000 and PowerPC Options.
41006 * mcall-sysv: RS/6000 and PowerPC Options.
41008 * mcall-sysv-eabi: RS/6000 and PowerPC Options.
41010 * mcall-sysv-noeabi: RS/6000 and PowerPC Options.
41012 * mcallee-super-interworking: ARM Options. (line 238)
41013 * mcaller-super-interworking: ARM Options. (line 244)
41014 * mcallgraph-data: MCore Options. (line 31)
41015 * mcc-init: CRIS Options. (line 41)
41016 * mcfv4e: M680x0 Options. (line 168)
41017 * mcheck-zero-division: MIPS Options. (line 425)
41018 * mcirrus-fix-invalid-insns: ARM Options. (line 189)
41019 * mcix: DEC Alpha Options. (line 171)
41020 * mcld: i386 and x86-64 Options.
41022 * mcmodel=embmedany: SPARC Options. (line 213)
41023 * mcmodel=kernel: i386 and x86-64 Options.
41025 * mcmodel=large: i386 and x86-64 Options.
41027 * mcmodel=medany: SPARC Options. (line 207)
41028 * mcmodel=medium: i386 and x86-64 Options.
41030 * mcmodel=medlow: SPARC Options. (line 196)
41031 * mcmodel=medmid: SPARC Options. (line 201)
41032 * mcmodel=small: i386 and x86-64 Options.
41034 * mcmpb: RS/6000 and PowerPC Options.
41036 * mcode-readable: MIPS Options. (line 385)
41037 * mcond-exec: FRV Options. (line 152)
41038 * mcond-move: FRV Options. (line 128)
41039 * mconsole: i386 and x86-64 Windows Options.
41041 * mconst-align: CRIS Options. (line 55)
41042 * mconst16: Xtensa Options. (line 10)
41043 * mconstant-gp: IA-64 Options. (line 46)
41044 * mcorea: Blackfin Options. (line 149)
41045 * mcoreb: Blackfin Options. (line 155)
41046 * mcpu <1>: SPARC Options. (line 96)
41047 * mcpu <2>: RS/6000 and PowerPC Options.
41049 * mcpu <3>: picoChip Options. (line 9)
41050 * mcpu <4>: M680x0 Options. (line 28)
41051 * mcpu <5>: i386 and x86-64 Options.
41053 * mcpu <6>: FRV Options. (line 212)
41054 * mcpu <7>: DEC Alpha Options. (line 223)
41055 * mcpu <8>: CRIS Options. (line 10)
41056 * mcpu <9>: ARM Options. (line 84)
41057 * mcpu: ARC Options. (line 23)
41058 * mcpu32: M680x0 Options. (line 134)
41059 * mcpu= <1>: M32C Options. (line 7)
41060 * mcpu=: Blackfin Options. (line 7)
41061 * mcsync-anomaly: Blackfin Options. (line 55)
41062 * mcx16: i386 and x86-64 Options.
41064 * mcygwin: i386 and x86-64 Windows Options.
41066 * MD: Preprocessor Options.
41068 * mdalign: SH Options. (line 64)
41069 * mdata: ARC Options. (line 30)
41070 * mdata-align: CRIS Options. (line 55)
41071 * mdebug <1>: S/390 and zSeries Options.
41073 * mdebug: M32R/D Options. (line 69)
41074 * mdec-asm: PDP-11 Options. (line 78)
41075 * mdisable-callt: V850 Options. (line 80)
41076 * mdisable-fpregs: HPPA Options. (line 33)
41077 * mdisable-indexing: HPPA Options. (line 40)
41078 * mdiv <1>: MCore Options. (line 15)
41079 * mdiv: M680x0 Options. (line 205)
41080 * mdiv=STRATEGY: SH Options. (line 141)
41081 * mdivide-breaks: MIPS Options. (line 431)
41082 * mdivide-traps: MIPS Options. (line 431)
41083 * mdivsi3_libfunc=NAME: SH Options. (line 182)
41084 * mdll: i386 and x86-64 Windows Options.
41086 * mdlmzb: RS/6000 and PowerPC Options.
41088 * mdmx: MIPS Options. (line 278)
41089 * mdouble: FRV Options. (line 38)
41090 * mdouble-float <1>: RS/6000 and PowerPC Options.
41092 * mdouble-float: MIPS Options. (line 236)
41093 * mdsp: MIPS Options. (line 255)
41094 * mdspr2: MIPS Options. (line 261)
41095 * mdual-nops: SPU Options. (line 55)
41096 * mdwarf2-asm: IA-64 Options. (line 79)
41097 * mdword: FRV Options. (line 32)
41098 * mdynamic-no-pic: RS/6000 and PowerPC Options.
41100 * meabi: RS/6000 and PowerPC Options.
41102 * mearly-stop-bits: IA-64 Options. (line 85)
41103 * meb: Score Options. (line 9)
41104 * mel: Score Options. (line 12)
41105 * melf <1>: MMIX Options. (line 44)
41106 * melf: CRIS Options. (line 87)
41107 * memb: RS/6000 and PowerPC Options.
41109 * membedded-data: MIPS Options. (line 372)
41110 * memregs=: M32C Options. (line 21)
41111 * mep: V850 Options. (line 16)
41112 * mepsilon: MMIX Options. (line 15)
41113 * merror-reloc: SPU Options. (line 10)
41114 * mesa: S/390 and zSeries Options.
41116 * metrax100: CRIS Options. (line 26)
41117 * metrax4: CRIS Options. (line 26)
41118 * mexplicit-relocs <1>: MIPS Options. (line 416)
41119 * mexplicit-relocs: DEC Alpha Options. (line 184)
41120 * mextern-sdata: MIPS Options. (line 334)
41121 * MF: Preprocessor Options.
41123 * mfast-fp: Blackfin Options. (line 128)
41124 * mfast-indirect-calls: HPPA Options. (line 52)
41125 * mfaster-structs: SPARC Options. (line 71)
41126 * mfdpic: FRV Options. (line 56)
41127 * mfix: DEC Alpha Options. (line 171)
41128 * mfix-and-continue: Darwin Options. (line 106)
41129 * mfix-cortex-m3-ldrd: ARC Options. (line 36)
41130 * mfix-r10000: MIPS Options. (line 502)
41131 * mfix-r4000: MIPS Options. (line 481)
41132 * mfix-r4400: MIPS Options. (line 495)
41133 * mfix-sb1: MIPS Options. (line 534)
41134 * mfix-vr4120: MIPS Options. (line 513)
41135 * mfix-vr4130: MIPS Options. (line 527)
41136 * mfixed-cc: FRV Options. (line 28)
41137 * mfixed-range <1>: SPU Options. (line 47)
41138 * mfixed-range <2>: SH Options. (line 189)
41139 * mfixed-range <3>: IA-64 Options. (line 90)
41140 * mfixed-range: HPPA Options. (line 59)
41141 * mflip-mips16: MIPS Options. (line 109)
41142 * mfloat-abi: ARM Options. (line 41)
41143 * mfloat-gprs: RS/6000 and PowerPC Options.
41145 * mfloat-ieee: DEC Alpha Options. (line 179)
41146 * mfloat-vax: DEC Alpha Options. (line 179)
41147 * mfloat32: PDP-11 Options. (line 52)
41148 * mfloat64: PDP-11 Options. (line 48)
41149 * mflush-func: MIPS Options. (line 601)
41150 * mflush-func=NAME: M32R/D Options. (line 94)
41151 * mflush-trap=NUMBER: M32R/D Options. (line 87)
41152 * mfmovd: SH Options. (line 81)
41153 * mfp: ARM Options. (line 124)
41154 * mfp-exceptions: MIPS Options. (line 628)
41155 * mfp-reg: DEC Alpha Options. (line 25)
41156 * mfp-rounding-mode: DEC Alpha Options. (line 85)
41157 * mfp-trap-mode: DEC Alpha Options. (line 63)
41158 * mfp32: MIPS Options. (line 219)
41159 * mfp64: MIPS Options. (line 222)
41160 * mfpe: ARM Options. (line 124)
41161 * mfpr-32: FRV Options. (line 13)
41162 * mfpr-64: FRV Options. (line 16)
41163 * mfprnd: RS/6000 and PowerPC Options.
41165 * mfpu <1>: SPARC Options. (line 20)
41166 * mfpu <2>: RS/6000 and PowerPC Options.
41168 * mfpu <3>: PDP-11 Options. (line 9)
41169 * mfpu: ARM Options. (line 124)
41170 * mfull-toc: RS/6000 and PowerPC Options.
41172 * mfused-madd <1>: Xtensa Options. (line 19)
41173 * mfused-madd <2>: S/390 and zSeries Options.
41175 * mfused-madd <3>: RS/6000 and PowerPC Options.
41177 * mfused-madd <4>: MIPS Options. (line 466)
41178 * mfused-madd: i386 and x86-64 Options.
41180 * mg: VAX Options. (line 17)
41181 * MG: Preprocessor Options.
41183 * mgas <1>: HPPA Options. (line 75)
41184 * mgas: DEC Alpha Options. (line 159)
41185 * mgen-cell-microcode: RS/6000 and PowerPC Options.
41187 * mgettrcost=NUMBER: SH Options. (line 211)
41188 * mglibc: GNU/Linux Options. (line 9)
41189 * mgnu: VAX Options. (line 13)
41190 * mgnu-as: IA-64 Options. (line 18)
41191 * mgnu-ld: IA-64 Options. (line 23)
41192 * mgotplt: CRIS Options. (line 81)
41193 * mgp32: MIPS Options. (line 213)
41194 * mgp64: MIPS Options. (line 216)
41195 * mgpopt: MIPS Options. (line 357)
41196 * mgpr-32: FRV Options. (line 7)
41197 * mgpr-64: FRV Options. (line 10)
41198 * mgprel-ro: FRV Options. (line 79)
41199 * mh: H8/300 Options. (line 14)
41200 * mhard-dfp <1>: S/390 and zSeries Options.
41202 * mhard-dfp: RS/6000 and PowerPC Options.
41204 * mhard-float <1>: SPARC Options. (line 20)
41205 * mhard-float <2>: S/390 and zSeries Options.
41207 * mhard-float <3>: RS/6000 and PowerPC Options.
41209 * mhard-float <4>: MIPS Options. (line 225)
41210 * mhard-float <5>: M680x0 Options. (line 193)
41211 * mhard-float <6>: FRV Options. (line 19)
41212 * mhard-float: ARM Options. (line 62)
41213 * mhard-quad-float: SPARC Options. (line 41)
41214 * mhardlit: MCore Options. (line 10)
41215 * mhint-max-distance: SPU Options. (line 67)
41216 * mhint-max-nops: SPU Options. (line 61)
41217 * mhitachi: SH Options. (line 84)
41218 * micplb: Blackfin Options. (line 168)
41219 * mid-shared-library: Blackfin Options. (line 76)
41220 * mieee <1>: SH Options. (line 99)
41221 * mieee: DEC Alpha Options. (line 39)
41222 * mieee-conformant: DEC Alpha Options. (line 134)
41223 * mieee-fp: i386 and x86-64 Options.
41225 * mieee-with-inexact: DEC Alpha Options. (line 52)
41226 * milp32: IA-64 Options. (line 114)
41227 * mimpure-text: SPARC Options. (line 81)
41228 * mincoming-stack-boundary: i386 and x86-64 Options.
41230 * mindexed-addressing: SH Options. (line 201)
41231 * minit-stack: AVR Options. (line 35)
41232 * minline-all-stringops: i386 and x86-64 Options.
41234 * minline-compares: i386 and x86-64 Options.
41236 * minline-float-divide-max-throughput: IA-64 Options. (line 58)
41237 * minline-float-divide-min-latency: IA-64 Options. (line 54)
41238 * minline-ic_invalidate: SH Options. (line 106)
41239 * minline-int-divide-max-throughput: IA-64 Options. (line 66)
41240 * minline-int-divide-min-latency: IA-64 Options. (line 62)
41241 * minline-plt <1>: FRV Options. (line 64)
41242 * minline-plt: Blackfin Options. (line 133)
41243 * minline-sqrt-max-throughput: IA-64 Options. (line 74)
41244 * minline-sqrt-min-latency: IA-64 Options. (line 70)
41245 * minline-stringops-dynamically: i386 and x86-64 Options.
41247 * minmax: M68hc1x Options. (line 31)
41248 * minsert-sched-nops: RS/6000 and PowerPC Options.
41250 * mint16: PDP-11 Options. (line 40)
41251 * mint32 <1>: PDP-11 Options. (line 44)
41252 * mint32: H8/300 Options. (line 28)
41253 * mint8: AVR Options. (line 55)
41254 * minterlink-mips16: MIPS Options. (line 116)
41255 * minvalid-symbols: SH Options. (line 234)
41256 * mips1: MIPS Options. (line 76)
41257 * mips16: MIPS Options. (line 101)
41258 * mips2: MIPS Options. (line 79)
41259 * mips3: MIPS Options. (line 82)
41260 * mips32: MIPS Options. (line 88)
41261 * mips32r2: MIPS Options. (line 91)
41262 * mips3d: MIPS Options. (line 284)
41263 * mips4: MIPS Options. (line 85)
41264 * mips64: MIPS Options. (line 94)
41265 * mips64r2: MIPS Options. (line 97)
41266 * misel: RS/6000 and PowerPC Options.
41268 * misize: SH Options. (line 118)
41269 * missue-rate=NUMBER: M32R/D Options. (line 79)
41270 * mjump-in-delay: HPPA Options. (line 28)
41271 * mkernel: Darwin Options. (line 84)
41272 * mknuthdiv: MMIX Options. (line 33)
41273 * ml: SH Options. (line 61)
41274 * mlarge-data: DEC Alpha Options. (line 195)
41275 * mlarge-data-threshold=NUMBER: i386 and x86-64 Options.
41277 * mlarge-mem: SPU Options. (line 35)
41278 * mlarge-text: DEC Alpha Options. (line 213)
41279 * mleaf-id-shared-library: Blackfin Options. (line 87)
41280 * mlibfuncs: MMIX Options. (line 10)
41281 * mlibrary-pic: FRV Options. (line 110)
41282 * mlinked-fp: FRV Options. (line 94)
41283 * mlinker-opt: HPPA Options. (line 85)
41284 * mlinux: CRIS Options. (line 91)
41285 * mlittle: RS/6000 and PowerPC Options.
41287 * mlittle-endian <1>: SPARC Options. (line 185)
41288 * mlittle-endian <2>: RS/6000 and PowerPC Options.
41290 * mlittle-endian <3>: MCore Options. (line 39)
41291 * mlittle-endian <4>: IA-64 Options. (line 13)
41292 * mlittle-endian: ARM Options. (line 68)
41293 * mllsc: MIPS Options. (line 241)
41294 * mlocal-sdata: MIPS Options. (line 322)
41295 * mlong-calls <1>: V850 Options. (line 10)
41296 * mlong-calls <2>: MIPS Options. (line 452)
41297 * mlong-calls <3>: M68hc1x Options. (line 35)
41298 * mlong-calls <4>: FRV Options. (line 99)
41299 * mlong-calls <5>: Blackfin Options. (line 116)
41300 * mlong-calls: ARM Options. (line 154)
41301 * mlong-double-128: S/390 and zSeries Options.
41303 * mlong-double-64: S/390 and zSeries Options.
41305 * mlong-load-store: HPPA Options. (line 66)
41306 * mlong32: MIPS Options. (line 297)
41307 * mlong64: MIPS Options. (line 292)
41308 * mlongcall: RS/6000 and PowerPC Options.
41310 * mlongcalls: Xtensa Options. (line 67)
41311 * mlow-64k: Blackfin Options. (line 65)
41312 * mlp64: IA-64 Options. (line 114)
41313 * MM: Preprocessor Options.
41315 * mmac <1>: Score Options. (line 21)
41316 * mmac: CRX Options. (line 9)
41317 * mmad: MIPS Options. (line 461)
41318 * mmangle-cpu: ARC Options. (line 15)
41319 * mmax: DEC Alpha Options. (line 171)
41320 * mmax-stack-frame: CRIS Options. (line 22)
41321 * mmcu: AVR Options. (line 9)
41322 * MMD: Preprocessor Options.
41324 * mmedia: FRV Options. (line 44)
41325 * mmemcpy: MIPS Options. (line 446)
41326 * mmemory-latency: DEC Alpha Options. (line 276)
41327 * mmfcrf: RS/6000 and PowerPC Options.
41329 * mmfpgpr: RS/6000 and PowerPC Options.
41331 * mminimal-toc: RS/6000 and PowerPC Options.
41333 * mmmx: i386 and x86-64 Options.
41335 * mmodel=large: M32R/D Options. (line 33)
41336 * mmodel=medium: M32R/D Options. (line 27)
41337 * mmodel=small: M32R/D Options. (line 18)
41338 * mmt: MIPS Options. (line 289)
41339 * mmul-bug-workaround: CRIS Options. (line 31)
41340 * mmuladd: FRV Options. (line 50)
41341 * mmulhw: RS/6000 and PowerPC Options.
41343 * mmult-bug: MN10300 Options. (line 9)
41344 * mmulti-cond-exec: FRV Options. (line 176)
41345 * mmulticore: Blackfin Options. (line 137)
41346 * mmultiple: RS/6000 and PowerPC Options.
41348 * mmvcle: S/390 and zSeries Options.
41350 * mmvme: RS/6000 and PowerPC Options.
41352 * mn: H8/300 Options. (line 20)
41353 * mnested-cond-exec: FRV Options. (line 189)
41354 * mnew-mnemonics: RS/6000 and PowerPC Options.
41356 * mnhwloop: Score Options. (line 15)
41357 * mno-3dnow: i386 and x86-64 Options.
41359 * mno-4byte-functions: MCore Options. (line 27)
41360 * mno-abicalls: MIPS Options. (line 153)
41361 * mno-abshi: PDP-11 Options. (line 58)
41362 * mno-ac0: PDP-11 Options. (line 20)
41363 * mno-align-double: i386 and x86-64 Options.
41365 * mno-align-int: M680x0 Options. (line 263)
41366 * mno-align-loops: M32R/D Options. (line 76)
41367 * mno-align-stringops: i386 and x86-64 Options.
41369 * mno-altivec: RS/6000 and PowerPC Options.
41371 * mno-am33: MN10300 Options. (line 20)
41372 * mno-app-regs <1>: V850 Options. (line 61)
41373 * mno-app-regs: SPARC Options. (line 10)
41374 * mno-avoid-indexed-addresses: RS/6000 and PowerPC Options.
41376 * mno-backchain: S/390 and zSeries Options.
41378 * mno-base-addresses: MMIX Options. (line 54)
41379 * mno-bit-align: RS/6000 and PowerPC Options.
41381 * mno-bitfield: M680x0 Options. (line 227)
41382 * mno-branch-likely: MIPS Options. (line 617)
41383 * mno-branch-predict: MMIX Options. (line 49)
41384 * mno-bwx: DEC Alpha Options. (line 171)
41385 * mno-callgraph-data: MCore Options. (line 31)
41386 * mno-check-zero-division: MIPS Options. (line 425)
41387 * mno-cirrus-fix-invalid-insns: ARM Options. (line 189)
41388 * mno-cix: DEC Alpha Options. (line 171)
41389 * mno-cmpb: RS/6000 and PowerPC Options.
41391 * mno-cond-exec: FRV Options. (line 158)
41392 * mno-cond-move: FRV Options. (line 134)
41393 * mno-const-align: CRIS Options. (line 55)
41394 * mno-const16: Xtensa Options. (line 10)
41395 * mno-crt0: MN10300 Options. (line 31)
41396 * mno-csync-anomaly: Blackfin Options. (line 61)
41397 * mno-cygwin: i386 and x86-64 Windows Options.
41399 * mno-data-align: CRIS Options. (line 55)
41400 * mno-debug: S/390 and zSeries Options.
41402 * mno-div <1>: MCore Options. (line 15)
41403 * mno-div: M680x0 Options. (line 205)
41404 * mno-dlmzb: RS/6000 and PowerPC Options.
41406 * mno-double: FRV Options. (line 41)
41407 * mno-dsp: MIPS Options. (line 255)
41408 * mno-dspr2: MIPS Options. (line 261)
41409 * mno-dwarf2-asm: IA-64 Options. (line 79)
41410 * mno-dword: FRV Options. (line 35)
41411 * mno-eabi: RS/6000 and PowerPC Options.
41413 * mno-early-stop-bits: IA-64 Options. (line 85)
41414 * mno-eflags: FRV Options. (line 125)
41415 * mno-embedded-data: MIPS Options. (line 372)
41416 * mno-ep: V850 Options. (line 16)
41417 * mno-epsilon: MMIX Options. (line 15)
41418 * mno-explicit-relocs <1>: MIPS Options. (line 416)
41419 * mno-explicit-relocs: DEC Alpha Options. (line 184)
41420 * mno-extern-sdata: MIPS Options. (line 334)
41421 * mno-fancy-math-387: i386 and x86-64 Options.
41423 * mno-faster-structs: SPARC Options. (line 71)
41424 * mno-fix: DEC Alpha Options. (line 171)
41425 * mno-fix-r10000: MIPS Options. (line 502)
41426 * mno-fix-r4000: MIPS Options. (line 481)
41427 * mno-fix-r4400: MIPS Options. (line 495)
41428 * mno-float32: PDP-11 Options. (line 48)
41429 * mno-float64: PDP-11 Options. (line 52)
41430 * mno-flush-func: M32R/D Options. (line 99)
41431 * mno-flush-trap: M32R/D Options. (line 91)
41432 * mno-fp-in-toc: RS/6000 and PowerPC Options.
41434 * mno-fp-regs: DEC Alpha Options. (line 25)
41435 * mno-fp-ret-in-387: i386 and x86-64 Options.
41437 * mno-fprnd: RS/6000 and PowerPC Options.
41439 * mno-fpu: SPARC Options. (line 25)
41440 * mno-fused-madd <1>: Xtensa Options. (line 19)
41441 * mno-fused-madd <2>: S/390 and zSeries Options.
41443 * mno-fused-madd <3>: RS/6000 and PowerPC Options.
41445 * mno-fused-madd: MIPS Options. (line 466)
41446 * mno-gnu-as: IA-64 Options. (line 18)
41447 * mno-gnu-ld: IA-64 Options. (line 23)
41448 * mno-gotplt: CRIS Options. (line 81)
41449 * mno-gpopt: MIPS Options. (line 357)
41450 * mno-hard-dfp <1>: S/390 and zSeries Options.
41452 * mno-hard-dfp: RS/6000 and PowerPC Options.
41454 * mno-hardlit: MCore Options. (line 10)
41455 * mno-id-shared-library: Blackfin Options. (line 83)
41456 * mno-ieee-fp: i386 and x86-64 Options.
41458 * mno-int16: PDP-11 Options. (line 44)
41459 * mno-int32: PDP-11 Options. (line 40)
41460 * mno-interlink-mips16: MIPS Options. (line 116)
41461 * mno-interrupts: AVR Options. (line 39)
41462 * mno-isel: RS/6000 and PowerPC Options.
41464 * mno-knuthdiv: MMIX Options. (line 33)
41465 * mno-leaf-id-shared-library: Blackfin Options. (line 93)
41466 * mno-libfuncs: MMIX Options. (line 10)
41467 * mno-llsc: MIPS Options. (line 241)
41468 * mno-local-sdata: MIPS Options. (line 322)
41469 * mno-long-calls <1>: V850 Options. (line 10)
41470 * mno-long-calls <2>: MIPS Options. (line 452)
41471 * mno-long-calls <3>: M68hc1x Options. (line 35)
41472 * mno-long-calls <4>: HPPA Options. (line 136)
41473 * mno-long-calls <5>: Blackfin Options. (line 116)
41474 * mno-long-calls: ARM Options. (line 154)
41475 * mno-longcall: RS/6000 and PowerPC Options.
41477 * mno-longcalls: Xtensa Options. (line 67)
41478 * mno-low-64k: Blackfin Options. (line 69)
41479 * mno-lsim: FR30 Options. (line 14)
41480 * mno-mad: MIPS Options. (line 461)
41481 * mno-max: DEC Alpha Options. (line 171)
41482 * mno-mdmx: MIPS Options. (line 278)
41483 * mno-media: FRV Options. (line 47)
41484 * mno-memcpy: MIPS Options. (line 446)
41485 * mno-mfcrf: RS/6000 and PowerPC Options.
41487 * mno-mfpgpr: RS/6000 and PowerPC Options.
41489 * mno-mips16: MIPS Options. (line 101)
41490 * mno-mips3d: MIPS Options. (line 284)
41491 * mno-mmx: i386 and x86-64 Options.
41493 * mno-mt: MIPS Options. (line 289)
41494 * mno-mul-bug-workaround: CRIS Options. (line 31)
41495 * mno-muladd: FRV Options. (line 53)
41496 * mno-mulhw: RS/6000 and PowerPC Options.
41498 * mno-mult-bug: MN10300 Options. (line 13)
41499 * mno-multi-cond-exec: FRV Options. (line 183)
41500 * mno-multiple: RS/6000 and PowerPC Options.
41502 * mno-mvcle: S/390 and zSeries Options.
41504 * mno-nested-cond-exec: FRV Options. (line 195)
41505 * mno-optimize-membar: FRV Options. (line 205)
41506 * mno-pack: FRV Options. (line 122)
41507 * mno-packed-stack: S/390 and zSeries Options.
41509 * mno-paired: RS/6000 and PowerPC Options.
41511 * mno-paired-single: MIPS Options. (line 272)
41512 * mno-pic: IA-64 Options. (line 26)
41513 * mno-plt: MIPS Options. (line 180)
41514 * mno-popcntb: RS/6000 and PowerPC Options.
41516 * mno-power: RS/6000 and PowerPC Options.
41518 * mno-power2: RS/6000 and PowerPC Options.
41520 * mno-powerpc: RS/6000 and PowerPC Options.
41522 * mno-powerpc-gfxopt: RS/6000 and PowerPC Options.
41524 * mno-powerpc-gpopt: RS/6000 and PowerPC Options.
41526 * mno-powerpc64: RS/6000 and PowerPC Options.
41528 * mno-prolog-function: V850 Options. (line 23)
41529 * mno-prologue-epilogue: CRIS Options. (line 71)
41530 * mno-prototype: RS/6000 and PowerPC Options.
41532 * mno-push-args: i386 and x86-64 Options.
41534 * mno-register-names: IA-64 Options. (line 37)
41535 * mno-regnames: RS/6000 and PowerPC Options.
41537 * mno-relax-immediate: MCore Options. (line 19)
41538 * mno-relocatable: RS/6000 and PowerPC Options.
41540 * mno-relocatable-lib: RS/6000 and PowerPC Options.
41542 * mno-rtd: M680x0 Options. (line 258)
41543 * mno-scc: FRV Options. (line 146)
41544 * mno-sched-ar-data-spec: IA-64 Options. (line 128)
41545 * mno-sched-ar-in-data-spec: IA-64 Options. (line 149)
41546 * mno-sched-br-data-spec: IA-64 Options. (line 121)
41547 * mno-sched-br-in-data-spec: IA-64 Options. (line 142)
41548 * mno-sched-control-ldc: IA-64 Options. (line 168)
41549 * mno-sched-control-spec: IA-64 Options. (line 135)
41550 * mno-sched-count-spec-in-critical-path: IA-64 Options. (line 194)
41551 * mno-sched-in-control-spec: IA-64 Options. (line 156)
41552 * mno-sched-ldc: IA-64 Options. (line 162)
41553 * mno-sched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
41554 * mno-sched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
41555 * mno-sched-prolog: ARM Options. (line 32)
41556 * mno-sched-spec-verbose: IA-64 Options. (line 176)
41557 * mno-sdata <1>: RS/6000 and PowerPC Options.
41559 * mno-sdata: IA-64 Options. (line 42)
41560 * mno-sep-data: Blackfin Options. (line 111)
41561 * mno-serialize-volatile: Xtensa Options. (line 35)
41562 * mno-short: M680x0 Options. (line 222)
41563 * mno-side-effects: CRIS Options. (line 46)
41564 * mno-single-exit: MMIX Options. (line 66)
41565 * mno-slow-bytes: MCore Options. (line 35)
41566 * mno-small-exec: S/390 and zSeries Options.
41568 * mno-smartmips: MIPS Options. (line 268)
41569 * mno-soft-float: DEC Alpha Options. (line 10)
41570 * mno-space-regs: HPPA Options. (line 45)
41571 * mno-spe: RS/6000 and PowerPC Options.
41573 * mno-specld-anomaly: Blackfin Options. (line 51)
41574 * mno-split: PDP-11 Options. (line 71)
41575 * mno-split-addresses: MIPS Options. (line 410)
41576 * mno-sse: i386 and x86-64 Options.
41578 * mno-stack-align: CRIS Options. (line 55)
41579 * mno-stack-bias: SPARC Options. (line 222)
41580 * mno-strict-align <1>: RS/6000 and PowerPC Options.
41582 * mno-strict-align: M680x0 Options. (line 283)
41583 * mno-string: RS/6000 and PowerPC Options.
41585 * mno-sum-in-toc: RS/6000 and PowerPC Options.
41587 * mno-swdiv: RS/6000 and PowerPC Options.
41589 * mno-sym32: MIPS Options. (line 307)
41590 * mno-tablejump: AVR Options. (line 47)
41591 * mno-target-align: Xtensa Options. (line 54)
41592 * mno-text-section-literals: Xtensa Options. (line 42)
41593 * mno-toc: RS/6000 and PowerPC Options.
41595 * mno-toplevel-symbols: MMIX Options. (line 40)
41596 * mno-tpf-trace: S/390 and zSeries Options.
41598 * mno-unaligned-doubles: SPARC Options. (line 59)
41599 * mno-uninit-const-in-rodata: MIPS Options. (line 380)
41600 * mno-update: RS/6000 and PowerPC Options.
41602 * mno-v8plus: SPARC Options. (line 170)
41603 * mno-vis: SPARC Options. (line 177)
41604 * mno-vliw-branch: FRV Options. (line 170)
41605 * mno-volatile-asm-stop: IA-64 Options. (line 32)
41606 * mno-vrsave: RS/6000 and PowerPC Options.
41608 * mno-wide-bitfields: MCore Options. (line 23)
41609 * mno-xgot <1>: MIPS Options. (line 190)
41610 * mno-xgot: M680x0 Options. (line 315)
41611 * mno-xl-compat: RS/6000 and PowerPC Options.
41613 * mno-zero-extend: MMIX Options. (line 27)
41614 * mnobitfield: M680x0 Options. (line 227)
41615 * mnomacsave: SH Options. (line 95)
41616 * mnominmax: M68hc1x Options. (line 31)
41617 * mnop-fun-dllimport: i386 and x86-64 Windows Options.
41619 * mold-mnemonics: RS/6000 and PowerPC Options.
41621 * momit-leaf-frame-pointer <1>: i386 and x86-64 Options.
41623 * momit-leaf-frame-pointer: Blackfin Options. (line 39)
41624 * mone-byte-bool: Darwin Options. (line 92)
41625 * moptimize-membar: FRV Options. (line 201)
41626 * MP: Preprocessor Options.
41628 * mpa-risc-1-0: HPPA Options. (line 19)
41629 * mpa-risc-1-1: HPPA Options. (line 19)
41630 * mpa-risc-2-0: HPPA Options. (line 19)
41631 * mpack: FRV Options. (line 119)
41632 * mpacked-stack: S/390 and zSeries Options.
41634 * mpadstruct: SH Options. (line 121)
41635 * mpaired: RS/6000 and PowerPC Options.
41637 * mpaired-single: MIPS Options. (line 272)
41638 * mpc32: i386 and x86-64 Options.
41640 * mpc64: i386 and x86-64 Options.
41642 * mpc80: i386 and x86-64 Options.
41644 * mpcrel: M680x0 Options. (line 275)
41645 * mpdebug: CRIS Options. (line 35)
41646 * mpe: RS/6000 and PowerPC Options.
41648 * mpic-register: ARM Options. (line 185)
41649 * mplt: MIPS Options. (line 180)
41650 * mpoke-function-name: ARM Options. (line 199)
41651 * mpopcntb: RS/6000 and PowerPC Options.
41653 * mportable-runtime: HPPA Options. (line 71)
41654 * mpower: RS/6000 and PowerPC Options.
41656 * mpower2: RS/6000 and PowerPC Options.
41658 * mpowerpc: RS/6000 and PowerPC Options.
41660 * mpowerpc-gfxopt: RS/6000 and PowerPC Options.
41662 * mpowerpc-gpopt: RS/6000 and PowerPC Options.
41664 * mpowerpc64: RS/6000 and PowerPC Options.
41666 * mprefergot: SH Options. (line 128)
41667 * mpreferred-stack-boundary: i386 and x86-64 Options.
41669 * mprioritize-restricted-insns: RS/6000 and PowerPC Options.
41671 * mprolog-function: V850 Options. (line 23)
41672 * mprologue-epilogue: CRIS Options. (line 71)
41673 * mprototype: RS/6000 and PowerPC Options.
41675 * mpt-fixed: SH Options. (line 215)
41676 * mpush-args <1>: i386 and x86-64 Options.
41678 * mpush-args: CRX Options. (line 13)
41679 * MQ: Preprocessor Options.
41681 * mr10k-cache-barrier: MIPS Options. (line 539)
41682 * mrecip: i386 and x86-64 Options.
41684 * mregister-names: IA-64 Options. (line 37)
41685 * mregnames: RS/6000 and PowerPC Options.
41687 * mregparm: i386 and x86-64 Options.
41689 * mrelax <1>: SH Options. (line 70)
41690 * mrelax <2>: MN10300 Options. (line 34)
41691 * mrelax: H8/300 Options. (line 9)
41692 * mrelax-immediate: MCore Options. (line 19)
41693 * mrelocatable: RS/6000 and PowerPC Options.
41695 * mrelocatable-lib: RS/6000 and PowerPC Options.
41697 * mreturn-pointer-on-d0: MN10300 Options. (line 24)
41698 * mrodata: ARC Options. (line 30)
41699 * mrtd <1>: Function Attributes.
41701 * mrtd <2>: M680x0 Options. (line 236)
41702 * mrtd: i386 and x86-64 Options.
41704 * mrtp: VxWorks Options. (line 11)
41705 * ms: H8/300 Options. (line 17)
41706 * ms2600: H8/300 Options. (line 24)
41707 * msafe-dma: SPU Options. (line 17)
41708 * msafe-hints: SPU Options. (line 72)
41709 * msahf: i386 and x86-64 Options.
41711 * mscc: FRV Options. (line 140)
41712 * msched-ar-data-spec: IA-64 Options. (line 128)
41713 * msched-ar-in-data-spec: IA-64 Options. (line 149)
41714 * msched-br-data-spec: IA-64 Options. (line 121)
41715 * msched-br-in-data-spec: IA-64 Options. (line 142)
41716 * msched-control-ldc: IA-64 Options. (line 168)
41717 * msched-control-spec: IA-64 Options. (line 135)
41718 * msched-costly-dep: RS/6000 and PowerPC Options.
41720 * msched-count-spec-in-critical-path: IA-64 Options. (line 194)
41721 * msched-in-control-spec: IA-64 Options. (line 156)
41722 * msched-ldc: IA-64 Options. (line 162)
41723 * msched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
41724 * msched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
41725 * msched-spec-verbose: IA-64 Options. (line 176)
41726 * mschedule: HPPA Options. (line 78)
41727 * mscore5: Score Options. (line 25)
41728 * mscore5u: Score Options. (line 28)
41729 * mscore7: Score Options. (line 31)
41730 * mscore7d: Score Options. (line 34)
41731 * msda: V850 Options. (line 40)
41732 * msdata <1>: RS/6000 and PowerPC Options.
41734 * msdata: IA-64 Options. (line 42)
41735 * msdata=data: RS/6000 and PowerPC Options.
41737 * msdata=default: RS/6000 and PowerPC Options.
41739 * msdata=eabi: RS/6000 and PowerPC Options.
41741 * msdata=none <1>: RS/6000 and PowerPC Options.
41743 * msdata=none: M32R/D Options. (line 40)
41744 * msdata=sdata: M32R/D Options. (line 49)
41745 * msdata=sysv: RS/6000 and PowerPC Options.
41747 * msdata=use: M32R/D Options. (line 53)
41748 * msdram: Blackfin Options. (line 162)
41749 * msecure-plt: RS/6000 and PowerPC Options.
41751 * msep-data: Blackfin Options. (line 105)
41752 * mserialize-volatile: Xtensa Options. (line 35)
41753 * mshared-library-id: Blackfin Options. (line 98)
41754 * mshort <1>: M68hc1x Options. (line 40)
41755 * mshort: M680x0 Options. (line 216)
41756 * msim <1>: Xstormy16 Options. (line 9)
41757 * msim <2>: RS/6000 and PowerPC Options.
41759 * msim <3>: M32C Options. (line 13)
41760 * msim: Blackfin Options. (line 32)
41761 * msimple-fpu: RS/6000 and PowerPC Options.
41763 * msingle-exit: MMIX Options. (line 66)
41764 * msingle-float <1>: RS/6000 and PowerPC Options.
41766 * msingle-float: MIPS Options. (line 232)
41767 * msingle-pic-base: ARM Options. (line 179)
41768 * msio: HPPA Options. (line 105)
41769 * msize: AVR Options. (line 32)
41770 * mslow-bytes: MCore Options. (line 35)
41771 * msmall-data: DEC Alpha Options. (line 195)
41772 * msmall-exec: S/390 and zSeries Options.
41774 * msmall-mem: SPU Options. (line 35)
41775 * msmall-model: FR30 Options. (line 9)
41776 * msmall-text: DEC Alpha Options. (line 213)
41777 * msmartmips: MIPS Options. (line 268)
41778 * msoft-float <1>: SPARC Options. (line 25)
41779 * msoft-float <2>: S/390 and zSeries Options.
41781 * msoft-float <3>: RS/6000 and PowerPC Options.
41783 * msoft-float <4>: PDP-11 Options. (line 13)
41784 * msoft-float <5>: MIPS Options. (line 228)
41785 * msoft-float <6>: M680x0 Options. (line 199)
41786 * msoft-float <7>: i386 and x86-64 Options.
41788 * msoft-float <8>: HPPA Options. (line 91)
41789 * msoft-float <9>: FRV Options. (line 22)
41790 * msoft-float <10>: DEC Alpha Options. (line 10)
41791 * msoft-float: ARM Options. (line 65)
41792 * msoft-quad-float: SPARC Options. (line 45)
41793 * msoft-reg-count: M68hc1x Options. (line 43)
41794 * mspace <1>: V850 Options. (line 30)
41795 * mspace: SH Options. (line 125)
41796 * mspe: RS/6000 and PowerPC Options.
41798 * mspecld-anomaly: Blackfin Options. (line 46)
41799 * msplit: PDP-11 Options. (line 68)
41800 * msplit-addresses: MIPS Options. (line 410)
41801 * msse: i386 and x86-64 Options.
41803 * msse2avx: i386 and x86-64 Options.
41805 * msseregparm: i386 and x86-64 Options.
41807 * mstack-align: CRIS Options. (line 55)
41808 * mstack-bias: SPARC Options. (line 222)
41809 * mstack-check-l1: Blackfin Options. (line 72)
41810 * mstack-guard: S/390 and zSeries Options.
41812 * mstack-increment: MCore Options. (line 50)
41813 * mstack-size: S/390 and zSeries Options.
41815 * mstackrealign: i386 and x86-64 Options.
41817 * mstdmain: SPU Options. (line 40)
41818 * mstrict-align <1>: RS/6000 and PowerPC Options.
41820 * mstrict-align: M680x0 Options. (line 283)
41821 * mstring: RS/6000 and PowerPC Options.
41823 * mstringop-strategy=ALG: i386 and x86-64 Options.
41825 * mstructure-size-boundary: ARM Options. (line 134)
41826 * msvr4-struct-return: RS/6000 and PowerPC Options.
41828 * mswdiv: RS/6000 and PowerPC Options.
41830 * msym32: MIPS Options. (line 307)
41831 * mt: IA-64 Options. (line 106)
41832 * MT: Preprocessor Options.
41834 * mtarget-align: Xtensa Options. (line 54)
41835 * mtda: V850 Options. (line 34)
41836 * mtext: ARC Options. (line 30)
41837 * mtext-section-literals: Xtensa Options. (line 42)
41838 * mthread: i386 and x86-64 Windows Options.
41840 * mthreads: i386 and x86-64 Options.
41842 * mthumb: ARM Options. (line 220)
41843 * mthumb-interwork: ARM Options. (line 25)
41844 * mtiny-stack: AVR Options. (line 52)
41845 * mtls-direct-seg-refs: i386 and x86-64 Options.
41847 * mtls-size: IA-64 Options. (line 97)
41848 * mtoc: RS/6000 and PowerPC Options.
41850 * mtomcat-stats: FRV Options. (line 209)
41851 * mtoplevel-symbols: MMIX Options. (line 40)
41852 * mtp: ARM Options. (line 250)
41853 * mtpcs-frame: ARM Options. (line 226)
41854 * mtpcs-leaf-frame: ARM Options. (line 232)
41855 * mtpf-trace: S/390 and zSeries Options.
41857 * mtrap-precision: DEC Alpha Options. (line 109)
41858 * mtune <1>: SPARC Options. (line 158)
41859 * mtune <2>: S/390 and zSeries Options.
41861 * mtune <3>: RS/6000 and PowerPC Options.
41863 * mtune <4>: MIPS Options. (line 61)
41864 * mtune <5>: M680x0 Options. (line 66)
41865 * mtune <6>: IA-64 Options. (line 101)
41866 * mtune <7>: i386 and x86-64 Options.
41868 * mtune <8>: DEC Alpha Options. (line 267)
41869 * mtune <9>: CRIS Options. (line 16)
41870 * mtune: ARM Options. (line 102)
41871 * muclibc: GNU/Linux Options. (line 13)
41872 * muls: Score Options. (line 18)
41873 * multcost=NUMBER: SH Options. (line 138)
41874 * multi_module: Darwin Options. (line 199)
41875 * multilib-library-pic: FRV Options. (line 89)
41876 * multiply_defined: Darwin Options. (line 199)
41877 * multiply_defined_unused: Darwin Options. (line 199)
41878 * munaligned-doubles: SPARC Options. (line 59)
41879 * muninit-const-in-rodata: MIPS Options. (line 380)
41880 * munix: VAX Options. (line 9)
41881 * munix-asm: PDP-11 Options. (line 74)
41882 * munsafe-dma: SPU Options. (line 17)
41883 * mupdate: RS/6000 and PowerPC Options.
41885 * musermode: SH Options. (line 133)
41886 * mv850: V850 Options. (line 49)
41887 * mv850e: V850 Options. (line 69)
41888 * mv850e1: V850 Options. (line 64)
41889 * mv8plus: SPARC Options. (line 170)
41890 * mveclibabi: i386 and x86-64 Options.
41892 * mvis: SPARC Options. (line 177)
41893 * mvliw-branch: FRV Options. (line 164)
41894 * mvms-return-codes: DEC Alpha/VMS Options.
41896 * mvolatile-asm-stop: IA-64 Options. (line 32)
41897 * mvr4130-align: MIPS Options. (line 638)
41898 * mvrsave: RS/6000 and PowerPC Options.
41900 * mvxworks: RS/6000 and PowerPC Options.
41902 * mwarn-cell-microcode: RS/6000 and PowerPC Options.
41904 * mwarn-dynamicstack: S/390 and zSeries Options.
41906 * mwarn-framesize: S/390 and zSeries Options.
41908 * mwarn-reloc: SPU Options. (line 10)
41909 * mwide-bitfields: MCore Options. (line 23)
41910 * mwin32: i386 and x86-64 Windows Options.
41912 * mwindows: i386 and x86-64 Windows Options.
41914 * mword-relocations: ARM Options. (line 258)
41915 * mwords-little-endian: ARM Options. (line 76)
41916 * mxgot <1>: MIPS Options. (line 190)
41917 * mxgot: M680x0 Options. (line 315)
41918 * mxilinx-fpu: RS/6000 and PowerPC Options.
41920 * mxl-compat: RS/6000 and PowerPC Options.
41922 * myellowknife: RS/6000 and PowerPC Options.
41924 * mzarch: S/390 and zSeries Options.
41926 * mzda: V850 Options. (line 45)
41927 * mzero-extend: MMIX Options. (line 27)
41928 * no-canonical-prefixes: Overall Options. (line 348)
41929 * no-integrated-cpp: C Dialect Options. (line 240)
41930 * no-lsim: MCore Options. (line 46)
41931 * no-red-zone: i386 and x86-64 Options.
41933 * no_dead_strip_inits_and_terms: Darwin Options. (line 199)
41934 * noall_load: Darwin Options. (line 199)
41935 * nocpp: MIPS Options. (line 476)
41936 * nodefaultlibs: Link Options. (line 62)
41937 * nofixprebinding: Darwin Options. (line 199)
41938 * nolibdld: HPPA Options. (line 188)
41939 * nomultidefs: Darwin Options. (line 199)
41940 * non-static: VxWorks Options. (line 16)
41941 * noprebind: Darwin Options. (line 199)
41942 * noseglinkedit: Darwin Options. (line 199)
41943 * nostartfiles: Link Options. (line 57)
41944 * nostdinc: Preprocessor Options.
41946 * nostdinc++ <1>: Preprocessor Options.
41948 * nostdinc++: C++ Dialect Options.
41950 * nostdlib: Link Options. (line 71)
41951 * o: Preprocessor Options.
41953 * O: Optimize Options. (line 29)
41954 * o: Overall Options. (line 187)
41955 * O0: Optimize Options. (line 106)
41956 * O1: Optimize Options. (line 29)
41957 * O2: Optimize Options. (line 67)
41958 * O3: Optimize Options. (line 100)
41959 * Os: Optimize Options. (line 110)
41960 * P: Preprocessor Options.
41962 * p: Debugging Options. (line 227)
41963 * pagezero_size: Darwin Options. (line 199)
41964 * param: Optimize Options. (line 1703)
41965 * pass-exit-codes: Overall Options. (line 145)
41966 * pedantic <1>: Warnings and Errors.
41968 * pedantic <2>: Alternate Keywords. (line 29)
41969 * pedantic <3>: C Extensions. (line 6)
41970 * pedantic <4>: Preprocessor Options.
41972 * pedantic <5>: Warning Options. (line 53)
41973 * pedantic: Standards. (line 16)
41974 * pedantic-errors <1>: Warnings and Errors.
41976 * pedantic-errors <2>: Non-bugs. (line 216)
41977 * pedantic-errors <3>: Preprocessor Options.
41979 * pedantic-errors <4>: Warning Options. (line 95)
41980 * pedantic-errors: Standards. (line 16)
41981 * pg: Debugging Options. (line 233)
41982 * pie: Link Options. (line 92)
41983 * pipe: Overall Options. (line 209)
41984 * prebind: Darwin Options. (line 199)
41985 * prebind_all_twolevel_modules: Darwin Options. (line 199)
41986 * preprocessor: Preprocessor Options.
41988 * print-file-name: Debugging Options. (line 898)
41989 * print-libgcc-file-name: Debugging Options. (line 919)
41990 * print-multi-directory: Debugging Options. (line 904)
41991 * print-multi-lib: Debugging Options. (line 909)
41992 * print-objc-runtime-info: Objective-C and Objective-C++ Dialect Options.
41994 * print-prog-name: Debugging Options. (line 916)
41995 * print-search-dirs: Debugging Options. (line 927)
41996 * print-sysroot: Debugging Options. (line 940)
41997 * print-sysroot-headers-suffix: Debugging Options. (line 947)
41998 * private_bundle: Darwin Options. (line 199)
41999 * pthread <1>: SPARC Options. (line 242)
42000 * pthread <2>: RS/6000 and PowerPC Options.
42002 * pthread: IA-64 Options. (line 106)
42003 * pthreads: SPARC Options. (line 236)
42004 * Q: Debugging Options. (line 239)
42005 * Qn: System V Options. (line 18)
42006 * Qy: System V Options. (line 14)
42007 * rdynamic: Link Options. (line 98)
42008 * read_only_relocs: Darwin Options. (line 199)
42009 * remap: Preprocessor Options.
42011 * s: Link Options. (line 105)
42012 * S <1>: Link Options. (line 20)
42013 * S: Overall Options. (line 170)
42014 * save-temps: Debugging Options. (line 860)
42015 * sectalign: Darwin Options. (line 199)
42016 * sectcreate: Darwin Options. (line 199)
42017 * sectobjectsymbols: Darwin Options. (line 199)
42018 * sectorder: Darwin Options. (line 199)
42019 * seg1addr: Darwin Options. (line 199)
42020 * seg_addr_table: Darwin Options. (line 199)
42021 * seg_addr_table_filename: Darwin Options. (line 199)
42022 * segaddr: Darwin Options. (line 199)
42023 * seglinkedit: Darwin Options. (line 199)
42024 * segprot: Darwin Options. (line 199)
42025 * segs_read_only_addr: Darwin Options. (line 199)
42026 * segs_read_write_addr: Darwin Options. (line 199)
42027 * shared: Link Options. (line 114)
42028 * shared-libgcc: Link Options. (line 122)
42029 * sim: CRIS Options. (line 95)
42030 * sim2: CRIS Options. (line 101)
42031 * single_module: Darwin Options. (line 199)
42032 * specs: Directory Options. (line 84)
42033 * static <1>: HPPA Options. (line 192)
42034 * static <2>: Darwin Options. (line 199)
42035 * static: Link Options. (line 109)
42036 * static-libgcc: Link Options. (line 122)
42037 * std <1>: Non-bugs. (line 107)
42038 * std <2>: Other Builtins. (line 22)
42039 * std <3>: C Dialect Options. (line 47)
42040 * std: Standards. (line 16)
42041 * std=: Preprocessor Options.
42043 * sub_library: Darwin Options. (line 199)
42044 * sub_umbrella: Darwin Options. (line 199)
42045 * symbolic: Link Options. (line 157)
42046 * sysroot: Directory Options. (line 92)
42047 * T: Link Options. (line 163)
42048 * target-help <1>: Preprocessor Options.
42050 * target-help: Overall Options. (line 240)
42051 * threads <1>: SPARC Options. (line 230)
42052 * threads: HPPA Options. (line 205)
42053 * time: Debugging Options. (line 874)
42054 * tls: FRV Options. (line 75)
42055 * TLS: FRV Options. (line 72)
42056 * traditional <1>: Incompatibilities. (line 6)
42057 * traditional: C Dialect Options. (line 252)
42058 * traditional-cpp <1>: Preprocessor Options.
42060 * traditional-cpp: C Dialect Options. (line 252)
42061 * trigraphs <1>: Preprocessor Options.
42063 * trigraphs: C Dialect Options. (line 236)
42064 * twolevel_namespace: Darwin Options. (line 199)
42065 * u: Link Options. (line 196)
42066 * U: Preprocessor Options.
42068 * umbrella: Darwin Options. (line 199)
42069 * undef: Preprocessor Options.
42071 * undefined: Darwin Options. (line 199)
42072 * unexported_symbols_list: Darwin Options. (line 199)
42073 * V: Target Options. (line 25)
42074 * v <1>: Preprocessor Options.
42076 * v: Overall Options. (line 198)
42077 * version <1>: Preprocessor Options.
42079 * version: Overall Options. (line 352)
42080 * W: Incompatibilities. (line 64)
42081 * w: Preprocessor Options.
42083 * W: Warning Options. (line 146)
42084 * w: Warning Options. (line 18)
42085 * Wa: Assembler Options. (line 9)
42086 * Wabi: C++ Dialect Options.
42088 * Waddress: Warning Options. (line 953)
42089 * Waggregate-return: Warning Options. (line 971)
42090 * Wall <1>: Standard Libraries. (line 6)
42091 * Wall <2>: Preprocessor Options.
42093 * Wall: Warning Options. (line 99)
42094 * Warray-bounds: Warning Options. (line 691)
42095 * Wassign-intercept: Objective-C and Objective-C++ Dialect Options.
42097 * Wattributes: Warning Options. (line 976)
42098 * Wbad-function-cast: Warning Options. (line 869)
42099 * Wbuiltin-macro-redefined: Warning Options. (line 982)
42100 * Wcast-align: Warning Options. (line 889)
42101 * Wcast-qual: Warning Options. (line 884)
42102 * Wchar-subscripts: Warning Options. (line 184)
42103 * Wclobbered: Warning Options. (line 909)
42104 * Wcomment <1>: Preprocessor Options.
42106 * Wcomment: Warning Options. (line 189)
42107 * Wcomments: Preprocessor Options.
42109 * Wconversion: Warning Options. (line 913)
42110 * Wcoverage-mismatch: Language Independent Options.
42112 * Wctor-dtor-privacy: C++ Dialect Options.
42114 * Wdeclaration-after-statement: Warning Options. (line 812)
42115 * Wdeprecated: Warning Options. (line 1119)
42116 * Wdeprecated-declarations: Warning Options. (line 1123)
42117 * Wdisabled-optimization: Warning Options. (line 1272)
42118 * Wdiv-by-zero: Warning Options. (line 696)
42119 * weak_reference_mismatches: Darwin Options. (line 199)
42120 * Weffc++: C++ Dialect Options.
42122 * Wempty-body: Warning Options. (line 932)
42123 * Wendif-labels <1>: Preprocessor Options.
42125 * Wendif-labels: Warning Options. (line 822)
42126 * Wenum-compare: Warning Options. (line 936)
42127 * Werror <1>: Preprocessor Options.
42129 * Werror: Warning Options. (line 21)
42130 * Werror=: Warning Options. (line 24)
42131 * Wextra: Warning Options. (line 146)
42132 * Wfatal-errors: Warning Options. (line 38)
42133 * Wfloat-equal: Warning Options. (line 712)
42134 * Wformat <1>: Function Attributes.
42136 * Wformat: Warning Options. (line 194)
42137 * Wformat-contains-nul: Warning Options. (line 233)
42138 * Wformat-extra-args: Warning Options. (line 237)
42139 * Wformat-nonliteral <1>: Function Attributes.
42141 * Wformat-nonliteral: Warning Options. (line 255)
42142 * Wformat-security: Warning Options. (line 260)
42143 * Wformat-y2k: Warning Options. (line 229)
42144 * Wformat-zero-length: Warning Options. (line 251)
42145 * Wformat=2: Warning Options. (line 271)
42146 * Wframe-larger-than: Warning Options. (line 834)
42147 * whatsloaded: Darwin Options. (line 199)
42148 * whyload: Darwin Options. (line 199)
42149 * Wignored-qualifiers: Warning Options. (line 310)
42150 * Wimplicit: Warning Options. (line 306)
42151 * Wimplicit-function-declaration: Warning Options. (line 300)
42152 * Wimplicit-int: Warning Options. (line 296)
42153 * Winit-self: Warning Options. (line 283)
42154 * Winline <1>: Inline. (line 63)
42155 * Winline: Warning Options. (line 1211)
42156 * Wint-to-pointer-cast: Warning Options. (line 1238)
42157 * Winvalid-offsetof: Warning Options. (line 1224)
42158 * Winvalid-pch: Warning Options. (line 1246)
42159 * Wl: Link Options. (line 188)
42160 * Wlarger-than-LEN: Warning Options. (line 831)
42161 * Wlarger-than=LEN: Warning Options. (line 831)
42162 * Wlogical-op: Warning Options. (line 966)
42163 * Wlong-long: Warning Options. (line 1250)
42164 * Wmain: Warning Options. (line 321)
42165 * Wmissing-braces: Warning Options. (line 328)
42166 * Wmissing-declarations: Warning Options. (line 1017)
42167 * Wmissing-field-initializers: Warning Options. (line 1025)
42168 * Wmissing-format-attribute: Warning Options. (line 1051)
42169 * Wmissing-include-dirs: Warning Options. (line 338)
42170 * Wmissing-noreturn: Warning Options. (line 1043)
42171 * Wmissing-parameter-type: Warning Options. (line 1003)
42172 * Wmissing-prototypes: Warning Options. (line 1011)
42173 * Wmultichar: Warning Options. (line 1070)
42174 * Wnested-externs: Warning Options. (line 1186)
42175 * Wno-abi: C++ Dialect Options.
42177 * Wno-address: Warning Options. (line 953)
42178 * Wno-aggregate-return: Warning Options. (line 971)
42179 * Wno-all: Warning Options. (line 99)
42180 * Wno-array-bounds: Warning Options. (line 691)
42181 * Wno-assign-intercept: Objective-C and Objective-C++ Dialect Options.
42183 * Wno-attributes: Warning Options. (line 976)
42184 * Wno-bad-function-cast: Warning Options. (line 869)
42185 * Wno-builtin-macro-redefined: Warning Options. (line 982)
42186 * Wno-cast-align: Warning Options. (line 889)
42187 * Wno-cast-qual: Warning Options. (line 884)
42188 * Wno-char-subscripts: Warning Options. (line 184)
42189 * Wno-clobbered: Warning Options. (line 909)
42190 * Wno-comment: Warning Options. (line 189)
42191 * Wno-conversion: Warning Options. (line 913)
42192 * Wno-ctor-dtor-privacy: C++ Dialect Options.
42194 * Wno-declaration-after-statement: Warning Options. (line 812)
42195 * Wno-deprecated: Warning Options. (line 1119)
42196 * Wno-deprecated-declarations: Warning Options. (line 1123)
42197 * Wno-disabled-optimization: Warning Options. (line 1272)
42198 * Wno-div-by-zero: Warning Options. (line 696)
42199 * Wno-effc++: C++ Dialect Options.
42201 * Wno-empty-body: Warning Options. (line 932)
42202 * Wno-endif-labels: Warning Options. (line 822)
42203 * Wno-enum-compare: Warning Options. (line 936)
42204 * Wno-error: Warning Options. (line 21)
42205 * Wno-error=: Warning Options. (line 24)
42206 * Wno-extra: Warning Options. (line 146)
42207 * Wno-fatal-errors: Warning Options. (line 38)
42208 * Wno-float-equal: Warning Options. (line 712)
42209 * Wno-format: Warning Options. (line 194)
42210 * Wno-format-contains-nul: Warning Options. (line 233)
42211 * Wno-format-extra-args: Warning Options. (line 237)
42212 * Wno-format-nonliteral: Warning Options. (line 255)
42213 * Wno-format-security: Warning Options. (line 260)
42214 * Wno-format-y2k: Warning Options. (line 229)
42215 * Wno-format-zero-length: Warning Options. (line 251)
42216 * Wno-format=2: Warning Options. (line 271)
42217 * Wno-ignored-qualifiers: Warning Options. (line 310)
42218 * Wno-implicit: Warning Options. (line 306)
42219 * Wno-implicit-function-declaration: Warning Options. (line 300)
42220 * Wno-implicit-int: Warning Options. (line 296)
42221 * Wno-init-self: Warning Options. (line 283)
42222 * Wno-inline: Warning Options. (line 1211)
42223 * Wno-int-to-pointer-cast: Warning Options. (line 1238)
42224 * Wno-invalid-offsetof: Warning Options. (line 1224)
42225 * Wno-invalid-pch: Warning Options. (line 1246)
42226 * Wno-logical-op: Warning Options. (line 966)
42227 * Wno-long-long: Warning Options. (line 1250)
42228 * Wno-main: Warning Options. (line 321)
42229 * Wno-missing-braces: Warning Options. (line 328)
42230 * Wno-missing-declarations: Warning Options. (line 1017)
42231 * Wno-missing-field-initializers: Warning Options. (line 1025)
42232 * Wno-missing-format-attribute: Warning Options. (line 1051)
42233 * Wno-missing-include-dirs: Warning Options. (line 338)
42234 * Wno-missing-noreturn: Warning Options. (line 1043)
42235 * Wno-missing-parameter-type: Warning Options. (line 1003)
42236 * Wno-missing-prototypes: Warning Options. (line 1011)
42237 * Wno-mudflap: Warning Options. (line 1292)
42238 * Wno-multichar: Warning Options. (line 1070)
42239 * Wno-nested-externs: Warning Options. (line 1186)
42240 * Wno-non-template-friend: C++ Dialect Options.
42242 * Wno-non-virtual-dtor: C++ Dialect Options.
42244 * Wno-nonnull: Warning Options. (line 276)
42245 * Wno-old-style-cast: C++ Dialect Options.
42247 * Wno-old-style-declaration: Warning Options. (line 993)
42248 * Wno-old-style-definition: Warning Options. (line 999)
42249 * Wno-overflow: Warning Options. (line 1129)
42250 * Wno-overlength-strings: Warning Options. (line 1296)
42251 * Wno-overloaded-virtual: C++ Dialect Options.
42253 * Wno-override-init: Warning Options. (line 1132)
42254 * Wno-packed: Warning Options. (line 1140)
42255 * Wno-packed-bitfield-compat: Warning Options. (line 1157)
42256 * Wno-padded: Warning Options. (line 1174)
42257 * Wno-parentheses: Warning Options. (line 341)
42258 * Wno-pedantic-ms-format: Warning Options. (line 849)
42259 * Wno-pmf-conversions <1>: Bound member functions.
42261 * Wno-pmf-conversions: C++ Dialect Options.
42263 * Wno-pointer-arith: Warning Options. (line 855)
42264 * Wno-pointer-sign: Warning Options. (line 1281)
42265 * Wno-pointer-to-int-cast: Warning Options. (line 1242)
42266 * Wno-pragmas: Warning Options. (line 594)
42267 * Wno-protocol: Objective-C and Objective-C++ Dialect Options.
42269 * Wno-redundant-decls: Warning Options. (line 1181)
42270 * Wno-reorder: C++ Dialect Options.
42272 * Wno-return-type: Warning Options. (line 431)
42273 * Wno-selector: Objective-C and Objective-C++ Dialect Options.
42275 * Wno-sequence-point: Warning Options. (line 385)
42276 * Wno-shadow: Warning Options. (line 826)
42277 * Wno-sign-compare: Warning Options. (line 940)
42278 * Wno-sign-conversion: Warning Options. (line 947)
42279 * Wno-sign-promo: C++ Dialect Options.
42281 * Wno-stack-protector: Warning Options. (line 1287)
42282 * Wno-strict-aliasing: Warning Options. (line 599)
42283 * Wno-strict-aliasing=n: Warning Options. (line 607)
42284 * Wno-strict-null-sentinel: C++ Dialect Options.
42286 * Wno-strict-overflow: Warning Options. (line 640)
42287 * Wno-strict-prototypes: Warning Options. (line 987)
42288 * Wno-strict-selector-match: Objective-C and Objective-C++ Dialect Options.
42290 * Wno-switch: Warning Options. (line 446)
42291 * Wno-switch-default: Warning Options. (line 454)
42292 * Wno-switch-enum: Warning Options. (line 457)
42293 * Wno-sync-nand: Warning Options. (line 463)
42294 * Wno-system-headers: Warning Options. (line 701)
42295 * Wno-traditional: Warning Options. (line 727)
42296 * Wno-traditional-conversion: Warning Options. (line 804)
42297 * Wno-trigraphs: Warning Options. (line 468)
42298 * Wno-type-limits: Warning Options. (line 862)
42299 * Wno-undeclared-selector: Objective-C and Objective-C++ Dialect Options.
42301 * Wno-undef: Warning Options. (line 819)
42302 * Wno-uninitialized: Warning Options. (line 517)
42303 * Wno-unknown-pragmas: Warning Options. (line 587)
42304 * Wno-unreachable-code: Warning Options. (line 1189)
42305 * Wno-unsafe-loop-optimizations: Warning Options. (line 843)
42306 * Wno-unused: Warning Options. (line 510)
42307 * Wno-unused-function: Warning Options. (line 473)
42308 * Wno-unused-label: Warning Options. (line 478)
42309 * Wno-unused-parameter: Warning Options. (line 485)
42310 * Wno-unused-value: Warning Options. (line 500)
42311 * Wno-unused-variable: Warning Options. (line 492)
42312 * Wno-variadic-macros: Warning Options. (line 1256)
42313 * Wno-vla: Warning Options. (line 1262)
42314 * Wno-volatile-register-var: Warning Options. (line 1266)
42315 * Wno-write-strings: Warning Options. (line 895)
42316 * Wnon-template-friend: C++ Dialect Options.
42318 * Wnon-virtual-dtor: C++ Dialect Options.
42320 * Wnonnull: Warning Options. (line 276)
42321 * Wnormalized=: Warning Options. (line 1076)
42322 * Wold-style-cast: C++ Dialect Options.
42324 * Wold-style-declaration: Warning Options. (line 993)
42325 * Wold-style-definition: Warning Options. (line 999)
42326 * Woverflow: Warning Options. (line 1129)
42327 * Woverlength-strings: Warning Options. (line 1296)
42328 * Woverloaded-virtual: C++ Dialect Options.
42330 * Woverride-init: Warning Options. (line 1132)
42331 * Wp: Preprocessor Options.
42333 * Wpacked: Warning Options. (line 1140)
42334 * Wpacked-bitfield-compat: Warning Options. (line 1157)
42335 * Wpadded: Warning Options. (line 1174)
42336 * Wparentheses: Warning Options. (line 341)
42337 * Wpedantic-ms-format: Warning Options. (line 849)
42338 * Wpmf-conversions: C++ Dialect Options.
42340 * Wpointer-arith <1>: Pointer Arith. (line 13)
42341 * Wpointer-arith: Warning Options. (line 855)
42342 * Wpointer-sign: Warning Options. (line 1281)
42343 * Wpointer-to-int-cast: Warning Options. (line 1242)
42344 * Wpragmas: Warning Options. (line 594)
42345 * Wprotocol: Objective-C and Objective-C++ Dialect Options.
42347 * wrapper: Overall Options. (line 355)
42348 * Wredundant-decls: Warning Options. (line 1181)
42349 * Wreorder: C++ Dialect Options.
42351 * Wreturn-type: Warning Options. (line 431)
42352 * Wselector: Objective-C and Objective-C++ Dialect Options.
42354 * Wsequence-point: Warning Options. (line 385)
42355 * Wshadow: Warning Options. (line 826)
42356 * Wsign-compare: Warning Options. (line 940)
42357 * Wsign-conversion: Warning Options. (line 947)
42358 * Wsign-promo: C++ Dialect Options.
42360 * Wstack-protector: Warning Options. (line 1287)
42361 * Wstrict-aliasing: Warning Options. (line 599)
42362 * Wstrict-aliasing=n: Warning Options. (line 607)
42363 * Wstrict-null-sentinel: C++ Dialect Options.
42365 * Wstrict-overflow: Warning Options. (line 640)
42366 * Wstrict-prototypes: Warning Options. (line 987)
42367 * Wstrict-selector-match: Objective-C and Objective-C++ Dialect Options.
42369 * Wswitch: Warning Options. (line 446)
42370 * Wswitch-default: Warning Options. (line 454)
42371 * Wswitch-enum: Warning Options. (line 457)
42372 * Wsync-nand: Warning Options. (line 463)
42373 * Wsystem-headers <1>: Preprocessor Options.
42375 * Wsystem-headers: Warning Options. (line 701)
42376 * Wtraditional <1>: Preprocessor Options.
42378 * Wtraditional: Warning Options. (line 727)
42379 * Wtraditional-conversion <1>: Protoize Caveats. (line 31)
42380 * Wtraditional-conversion: Warning Options. (line 804)
42381 * Wtrigraphs <1>: Preprocessor Options.
42383 * Wtrigraphs: Warning Options. (line 468)
42384 * Wtype-limits: Warning Options. (line 862)
42385 * Wundeclared-selector: Objective-C and Objective-C++ Dialect Options.
42387 * Wundef <1>: Preprocessor Options.
42389 * Wundef: Warning Options. (line 819)
42390 * Wuninitialized: Warning Options. (line 517)
42391 * Wunknown-pragmas: Warning Options. (line 587)
42392 * Wunreachable-code: Warning Options. (line 1189)
42393 * Wunsafe-loop-optimizations: Warning Options. (line 843)
42394 * Wunused: Warning Options. (line 510)
42395 * Wunused-function: Warning Options. (line 473)
42396 * Wunused-label: Warning Options. (line 478)
42397 * Wunused-macros: Preprocessor Options.
42399 * Wunused-parameter: Warning Options. (line 485)
42400 * Wunused-value: Warning Options. (line 500)
42401 * Wunused-variable: Warning Options. (line 492)
42402 * Wvariadic-macros: Warning Options. (line 1256)
42403 * Wvla: Warning Options. (line 1262)
42404 * Wvolatile-register-var: Warning Options. (line 1266)
42405 * Wwrite-strings: Warning Options. (line 895)
42406 * x <1>: Preprocessor Options.
42408 * x: Overall Options. (line 122)
42409 * Xassembler: Assembler Options. (line 13)
42410 * Xbind-lazy: VxWorks Options. (line 26)
42411 * Xbind-now: VxWorks Options. (line 30)
42412 * Xlinker: Link Options. (line 169)
42413 * Ym: System V Options. (line 26)
42414 * YP: System V Options. (line 22)
42417 File: gcc.info, Node: Keyword Index, Prev: Option Index, Up: Top
42425 * ! in constraint: Multi-Alternative. (line 33)
42426 * # in constraint: Modifiers. (line 57)
42427 * #pragma: Pragmas. (line 6)
42428 * #pragma implementation: C++ Interface. (line 39)
42429 * #pragma implementation, implied: C++ Interface. (line 46)
42430 * #pragma interface: C++ Interface. (line 20)
42431 * #pragma, reason for not using: Function Attributes.
42433 * $: Dollar Signs. (line 6)
42434 * % in constraint: Modifiers. (line 45)
42435 * %include: Spec Files. (line 27)
42436 * %include_noerr: Spec Files. (line 31)
42437 * %rename: Spec Files. (line 35)
42438 * & in constraint: Modifiers. (line 25)
42439 * ': Incompatibilities. (line 116)
42440 * (: Constructing Calls. (line 53)
42441 * * in constraint: Modifiers. (line 62)
42442 * + in constraint: Modifiers. (line 12)
42443 * -lgcc, use with -nodefaultlibs: Link Options. (line 79)
42444 * -lgcc, use with -nostdlib: Link Options. (line 79)
42445 * -nodefaultlibs and unresolved references: Link Options. (line 79)
42446 * -nostdlib and unresolved references: Link Options. (line 79)
42447 * .sdata/.sdata2 references (PowerPC): RS/6000 and PowerPC Options.
42449 * //: C++ Comments. (line 6)
42450 * 0 in constraint: Simple Constraints. (line 117)
42451 * < in constraint: Simple Constraints. (line 48)
42452 * = in constraint: Modifiers. (line 8)
42453 * > in constraint: Simple Constraints. (line 52)
42454 * ? in constraint: Multi-Alternative. (line 27)
42455 * ?: extensions: Conditionals. (line 6)
42456 * ?: side effect: Conditionals. (line 20)
42457 * _ in variables in macros: Typeof. (line 42)
42458 * __builtin___clear_cache: Other Builtins. (line 274)
42459 * __builtin___fprintf_chk: Object Size Checking.
42461 * __builtin___memcpy_chk: Object Size Checking.
42463 * __builtin___memmove_chk: Object Size Checking.
42465 * __builtin___mempcpy_chk: Object Size Checking.
42467 * __builtin___memset_chk: Object Size Checking.
42469 * __builtin___printf_chk: Object Size Checking.
42471 * __builtin___snprintf_chk: Object Size Checking.
42473 * __builtin___sprintf_chk: Object Size Checking.
42475 * __builtin___stpcpy_chk: Object Size Checking.
42477 * __builtin___strcat_chk: Object Size Checking.
42479 * __builtin___strcpy_chk: Object Size Checking.
42481 * __builtin___strncat_chk: Object Size Checking.
42483 * __builtin___strncpy_chk: Object Size Checking.
42485 * __builtin___vfprintf_chk: Object Size Checking.
42487 * __builtin___vprintf_chk: Object Size Checking.
42489 * __builtin___vsnprintf_chk: Object Size Checking.
42491 * __builtin___vsprintf_chk: Object Size Checking.
42493 * __builtin_apply: Constructing Calls. (line 31)
42494 * __builtin_apply_args: Constructing Calls. (line 20)
42495 * __builtin_bswap32: Other Builtins. (line 493)
42496 * __builtin_bswap64: Other Builtins. (line 498)
42497 * __builtin_choose_expr: Other Builtins. (line 156)
42498 * __builtin_clz: Other Builtins. (line 426)
42499 * __builtin_clzl: Other Builtins. (line 444)
42500 * __builtin_clzll: Other Builtins. (line 464)
42501 * __builtin_constant_p: Other Builtins. (line 196)
42502 * __builtin_ctz: Other Builtins. (line 430)
42503 * __builtin_ctzl: Other Builtins. (line 448)
42504 * __builtin_ctzll: Other Builtins. (line 468)
42505 * __builtin_expect: Other Builtins. (line 242)
42506 * __builtin_ffs: Other Builtins. (line 422)
42507 * __builtin_ffsl: Other Builtins. (line 440)
42508 * __builtin_ffsll: Other Builtins. (line 460)
42509 * __builtin_fpclassify: Other Builtins. (line 6)
42510 * __builtin_frame_address: Return Address. (line 34)
42511 * __builtin_huge_val: Other Builtins. (line 325)
42512 * __builtin_huge_valf: Other Builtins. (line 330)
42513 * __builtin_huge_vall: Other Builtins. (line 333)
42514 * __builtin_inf: Other Builtins. (line 348)
42515 * __builtin_infd128: Other Builtins. (line 358)
42516 * __builtin_infd32: Other Builtins. (line 352)
42517 * __builtin_infd64: Other Builtins. (line 355)
42518 * __builtin_inff: Other Builtins. (line 362)
42519 * __builtin_infl: Other Builtins. (line 367)
42520 * __builtin_isfinite: Other Builtins. (line 6)
42521 * __builtin_isgreater: Other Builtins. (line 6)
42522 * __builtin_isgreaterequal: Other Builtins. (line 6)
42523 * __builtin_isinf_sign: Other Builtins. (line 6)
42524 * __builtin_isless: Other Builtins. (line 6)
42525 * __builtin_islessequal: Other Builtins. (line 6)
42526 * __builtin_islessgreater: Other Builtins. (line 6)
42527 * __builtin_isnormal: Other Builtins. (line 6)
42528 * __builtin_isunordered: Other Builtins. (line 6)
42529 * __builtin_nan: Other Builtins. (line 378)
42530 * __builtin_nand128: Other Builtins. (line 400)
42531 * __builtin_nand32: Other Builtins. (line 394)
42532 * __builtin_nand64: Other Builtins. (line 397)
42533 * __builtin_nanf: Other Builtins. (line 404)
42534 * __builtin_nanl: Other Builtins. (line 407)
42535 * __builtin_nans: Other Builtins. (line 411)
42536 * __builtin_nansf: Other Builtins. (line 415)
42537 * __builtin_nansl: Other Builtins. (line 418)
42538 * __builtin_object_size: Object Size Checking.
42540 * __builtin_offsetof: Offsetof. (line 6)
42541 * __builtin_parity: Other Builtins. (line 437)
42542 * __builtin_parityl: Other Builtins. (line 456)
42543 * __builtin_parityll: Other Builtins. (line 476)
42544 * __builtin_popcount: Other Builtins. (line 434)
42545 * __builtin_popcountl: Other Builtins. (line 452)
42546 * __builtin_popcountll: Other Builtins. (line 472)
42547 * __builtin_powi: Other Builtins. (line 6)
42548 * __builtin_powif: Other Builtins. (line 6)
42549 * __builtin_powil: Other Builtins. (line 6)
42550 * __builtin_prefetch: Other Builtins. (line 286)
42551 * __builtin_return: Constructing Calls. (line 48)
42552 * __builtin_return_address: Return Address. (line 11)
42553 * __builtin_trap: Other Builtins. (line 266)
42554 * __builtin_types_compatible_p: Other Builtins. (line 110)
42555 * __complex__ keyword: Complex. (line 6)
42556 * __declspec(dllexport): Function Attributes.
42558 * __declspec(dllimport): Function Attributes.
42560 * __extension__: Alternate Keywords. (line 29)
42561 * __float128 data type: Floating Types. (line 6)
42562 * __float80 data type: Floating Types. (line 6)
42563 * __func__ identifier: Function Names. (line 6)
42564 * __FUNCTION__ identifier: Function Names. (line 6)
42565 * __imag__ keyword: Complex. (line 27)
42566 * __PRETTY_FUNCTION__ identifier: Function Names. (line 6)
42567 * __real__ keyword: Complex. (line 27)
42568 * __STDC_HOSTED__: Standards. (line 13)
42569 * __sync_add_and_fetch: Atomic Builtins. (line 61)
42570 * __sync_and_and_fetch: Atomic Builtins. (line 61)
42571 * __sync_bool_compare_and_swap: Atomic Builtins. (line 73)
42572 * __sync_fetch_and_add: Atomic Builtins. (line 45)
42573 * __sync_fetch_and_and: Atomic Builtins. (line 45)
42574 * __sync_fetch_and_nand: Atomic Builtins. (line 45)
42575 * __sync_fetch_and_or: Atomic Builtins. (line 45)
42576 * __sync_fetch_and_sub: Atomic Builtins. (line 45)
42577 * __sync_fetch_and_xor: Atomic Builtins. (line 45)
42578 * __sync_lock_release: Atomic Builtins. (line 103)
42579 * __sync_lock_test_and_set: Atomic Builtins. (line 85)
42580 * __sync_nand_and_fetch: Atomic Builtins. (line 61)
42581 * __sync_or_and_fetch: Atomic Builtins. (line 61)
42582 * __sync_sub_and_fetch: Atomic Builtins. (line 61)
42583 * __sync_synchronize: Atomic Builtins. (line 82)
42584 * __sync_val_compare_and_swap: Atomic Builtins. (line 73)
42585 * __sync_xor_and_fetch: Atomic Builtins. (line 61)
42586 * __thread: Thread-Local. (line 6)
42587 * _Accum data type: Fixed-Point. (line 6)
42588 * _Complex keyword: Complex. (line 6)
42589 * _Decimal128 data type: Decimal Float. (line 6)
42590 * _Decimal32 data type: Decimal Float. (line 6)
42591 * _Decimal64 data type: Decimal Float. (line 6)
42592 * _exit: Other Builtins. (line 6)
42593 * _Exit: Other Builtins. (line 6)
42594 * _Fract data type: Fixed-Point. (line 6)
42595 * _Sat data type: Fixed-Point. (line 6)
42596 * ABI: Compatibility. (line 6)
42597 * abort: Other Builtins. (line 6)
42598 * abs: Other Builtins. (line 6)
42599 * accessing volatiles: Volatiles. (line 6)
42600 * acos: Other Builtins. (line 6)
42601 * acosf: Other Builtins. (line 6)
42602 * acosh: Other Builtins. (line 6)
42603 * acoshf: Other Builtins. (line 6)
42604 * acoshl: Other Builtins. (line 6)
42605 * acosl: Other Builtins. (line 6)
42606 * Ada: G++ and GCC. (line 6)
42607 * additional floating types: Floating Types. (line 6)
42608 * address constraints: Simple Constraints. (line 144)
42609 * address of a label: Labels as Values. (line 6)
42610 * address_operand: Simple Constraints. (line 148)
42611 * alias attribute: Function Attributes.
42613 * aliasing of parameters: Code Gen Options. (line 409)
42614 * aligned attribute <1>: Type Attributes. (line 31)
42615 * aligned attribute <2>: Variable Attributes.
42617 * aligned attribute: Function Attributes.
42619 * alignment: Alignment. (line 6)
42620 * alloc_size attribute: Function Attributes.
42622 * alloca: Other Builtins. (line 6)
42623 * alloca vs variable-length arrays: Variable Length. (line 27)
42624 * Allow nesting in an interrupt handler on the Blackfin processor.: Function Attributes.
42626 * alternate keywords: Alternate Keywords. (line 6)
42627 * always_inline function attribute: Function Attributes.
42629 * AMD x86-64 Options: i386 and x86-64 Options.
42631 * AMD1: Standards. (line 13)
42632 * ANSI C: Standards. (line 13)
42633 * ANSI C standard: Standards. (line 13)
42634 * ANSI C89: Standards. (line 13)
42635 * ANSI support: C Dialect Options. (line 10)
42636 * ANSI X3.159-1989: Standards. (line 13)
42637 * apostrophes: Incompatibilities. (line 116)
42638 * application binary interface: Compatibility. (line 6)
42639 * ARC Options: ARC Options. (line 6)
42640 * ARM [Annotated C++ Reference Manual]: Backwards Compatibility.
42642 * ARM options: ARM Options. (line 6)
42643 * arrays of length zero: Zero Length. (line 6)
42644 * arrays of variable length: Variable Length. (line 6)
42645 * arrays, non-lvalue: Subscripting. (line 6)
42646 * artificial function attribute: Function Attributes.
42648 * asin: Other Builtins. (line 6)
42649 * asinf: Other Builtins. (line 6)
42650 * asinh: Other Builtins. (line 6)
42651 * asinhf: Other Builtins. (line 6)
42652 * asinhl: Other Builtins. (line 6)
42653 * asinl: Other Builtins. (line 6)
42654 * asm constraints: Constraints. (line 6)
42655 * asm expressions: Extended Asm. (line 6)
42656 * assembler instructions: Extended Asm. (line 6)
42657 * assembler names for identifiers: Asm Labels. (line 6)
42658 * assembly code, invalid: Bug Criteria. (line 12)
42659 * atan: Other Builtins. (line 6)
42660 * atan2: Other Builtins. (line 6)
42661 * atan2f: Other Builtins. (line 6)
42662 * atan2l: Other Builtins. (line 6)
42663 * atanf: Other Builtins. (line 6)
42664 * atanh: Other Builtins. (line 6)
42665 * atanhf: Other Builtins. (line 6)
42666 * atanhl: Other Builtins. (line 6)
42667 * atanl: Other Builtins. (line 6)
42668 * attribute of types: Type Attributes. (line 6)
42669 * attribute of variables: Variable Attributes.
42671 * attribute syntax: Attribute Syntax. (line 6)
42672 * autoincrement/decrement addressing: Simple Constraints. (line 30)
42673 * automatic inline for C++ member fns: Inline. (line 71)
42674 * AVR Options: AVR Options. (line 6)
42675 * Backwards Compatibility: Backwards Compatibility.
42677 * base class members: Name lookup. (line 6)
42678 * bcmp: Other Builtins. (line 6)
42679 * below100 attribute: Variable Attributes.
42681 * binary compatibility: Compatibility. (line 6)
42682 * Binary constants using the 0b prefix: Binary constants. (line 6)
42683 * Blackfin Options: Blackfin Options. (line 6)
42684 * bound pointer to member function: Bound member functions.
42686 * bounds checking: Optimize Options. (line 338)
42687 * bug criteria: Bug Criteria. (line 6)
42688 * bugs: Bugs. (line 6)
42689 * bugs, known: Trouble. (line 6)
42690 * built-in functions <1>: Other Builtins. (line 6)
42691 * built-in functions: C Dialect Options. (line 170)
42692 * bzero: Other Builtins. (line 6)
42693 * C compilation options: Invoking GCC. (line 17)
42694 * C intermediate output, nonexistent: G++ and GCC. (line 35)
42695 * C language extensions: C Extensions. (line 6)
42696 * C language, traditional: C Dialect Options. (line 250)
42697 * C standard: Standards. (line 13)
42698 * C standards: Standards. (line 13)
42699 * c++: Invoking G++. (line 14)
42700 * C++: G++ and GCC. (line 30)
42701 * C++ comments: C++ Comments. (line 6)
42702 * C++ compilation options: Invoking GCC. (line 23)
42703 * C++ interface and implementation headers: C++ Interface. (line 6)
42704 * C++ language extensions: C++ Extensions. (line 6)
42705 * C++ member fns, automatically inline: Inline. (line 71)
42706 * C++ misunderstandings: C++ Misunderstandings.
42708 * C++ options, command line: C++ Dialect Options.
42710 * C++ pragmas, effect on inlining: C++ Interface. (line 66)
42711 * C++ source file suffixes: Invoking G++. (line 6)
42712 * C++ static data, declaring and defining: Static Definitions.
42714 * C89: Standards. (line 13)
42715 * C90: Standards. (line 13)
42716 * C94: Standards. (line 13)
42717 * C95: Standards. (line 13)
42718 * C99: Standards. (line 13)
42719 * C9X: Standards. (line 13)
42720 * C_INCLUDE_PATH: Environment Variables.
42722 * cabs: Other Builtins. (line 6)
42723 * cabsf: Other Builtins. (line 6)
42724 * cabsl: Other Builtins. (line 6)
42725 * cacos: Other Builtins. (line 6)
42726 * cacosf: Other Builtins. (line 6)
42727 * cacosh: Other Builtins. (line 6)
42728 * cacoshf: Other Builtins. (line 6)
42729 * cacoshl: Other Builtins. (line 6)
42730 * cacosl: Other Builtins. (line 6)
42731 * calling functions through the function vector on H8/300, M16C, M32C and SH2A processors: Function Attributes.
42733 * calloc: Other Builtins. (line 6)
42734 * carg: Other Builtins. (line 6)
42735 * cargf: Other Builtins. (line 6)
42736 * cargl: Other Builtins. (line 6)
42737 * case labels in initializers: Designated Inits. (line 6)
42738 * case ranges: Case Ranges. (line 6)
42739 * casin: Other Builtins. (line 6)
42740 * casinf: Other Builtins. (line 6)
42741 * casinh: Other Builtins. (line 6)
42742 * casinhf: Other Builtins. (line 6)
42743 * casinhl: Other Builtins. (line 6)
42744 * casinl: Other Builtins. (line 6)
42745 * cast to a union: Cast to Union. (line 6)
42746 * catan: Other Builtins. (line 6)
42747 * catanf: Other Builtins. (line 6)
42748 * catanh: Other Builtins. (line 6)
42749 * catanhf: Other Builtins. (line 6)
42750 * catanhl: Other Builtins. (line 6)
42751 * catanl: Other Builtins. (line 6)
42752 * cbrt: Other Builtins. (line 6)
42753 * cbrtf: Other Builtins. (line 6)
42754 * cbrtl: Other Builtins. (line 6)
42755 * ccos: Other Builtins. (line 6)
42756 * ccosf: Other Builtins. (line 6)
42757 * ccosh: Other Builtins. (line 6)
42758 * ccoshf: Other Builtins. (line 6)
42759 * ccoshl: Other Builtins. (line 6)
42760 * ccosl: Other Builtins. (line 6)
42761 * ceil: Other Builtins. (line 6)
42762 * ceilf: Other Builtins. (line 6)
42763 * ceill: Other Builtins. (line 6)
42764 * cexp: Other Builtins. (line 6)
42765 * cexpf: Other Builtins. (line 6)
42766 * cexpl: Other Builtins. (line 6)
42767 * character set, execution: Preprocessor Options.
42769 * character set, input: Preprocessor Options.
42771 * character set, input normalization: Warning Options. (line 1076)
42772 * character set, wide execution: Preprocessor Options.
42774 * cimag: Other Builtins. (line 6)
42775 * cimagf: Other Builtins. (line 6)
42776 * cimagl: Other Builtins. (line 6)
42777 * cleanup attribute: Variable Attributes.
42779 * clog: Other Builtins. (line 6)
42780 * clogf: Other Builtins. (line 6)
42781 * clogl: Other Builtins. (line 6)
42782 * COBOL: G++ and GCC. (line 23)
42783 * code generation conventions: Code Gen Options. (line 6)
42784 * code, mixed with declarations: Mixed Declarations. (line 6)
42785 * cold function attribute: Function Attributes.
42787 * command options: Invoking GCC. (line 6)
42788 * comments, C++ style: C++ Comments. (line 6)
42789 * common attribute: Variable Attributes.
42791 * comparison of signed and unsigned values, warning: Warning Options.
42793 * compiler bugs, reporting: Bug Reporting. (line 6)
42794 * compiler compared to C++ preprocessor: G++ and GCC. (line 35)
42795 * compiler options, C++: C++ Dialect Options.
42797 * compiler options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
42799 * compiler version, specifying: Target Options. (line 6)
42800 * COMPILER_PATH: Environment Variables.
42802 * complex conjugation: Complex. (line 34)
42803 * complex numbers: Complex. (line 6)
42804 * compound literals: Compound Literals. (line 6)
42805 * computed gotos: Labels as Values. (line 6)
42806 * conditional expressions, extensions: Conditionals. (line 6)
42807 * conflicting types: Disappointments. (line 21)
42808 * conj: Other Builtins. (line 6)
42809 * conjf: Other Builtins. (line 6)
42810 * conjl: Other Builtins. (line 6)
42811 * const applied to function: Function Attributes.
42813 * const function attribute: Function Attributes.
42815 * constants in constraints: Simple Constraints. (line 60)
42816 * constraint modifier characters: Modifiers. (line 6)
42817 * constraint, matching: Simple Constraints. (line 129)
42818 * constraints, asm: Constraints. (line 6)
42819 * constraints, machine specific: Machine Constraints.
42821 * constructing calls: Constructing Calls. (line 6)
42822 * constructor expressions: Compound Literals. (line 6)
42823 * constructor function attribute: Function Attributes.
42825 * contributors: Contributors. (line 6)
42826 * copysign: Other Builtins. (line 6)
42827 * copysignf: Other Builtins. (line 6)
42828 * copysignl: Other Builtins. (line 6)
42829 * core dump: Bug Criteria. (line 9)
42830 * cos: Other Builtins. (line 6)
42831 * cosf: Other Builtins. (line 6)
42832 * cosh: Other Builtins. (line 6)
42833 * coshf: Other Builtins. (line 6)
42834 * coshl: Other Builtins. (line 6)
42835 * cosl: Other Builtins. (line 6)
42836 * CPATH: Environment Variables.
42838 * CPLUS_INCLUDE_PATH: Environment Variables.
42840 * cpow: Other Builtins. (line 6)
42841 * cpowf: Other Builtins. (line 6)
42842 * cpowl: Other Builtins. (line 6)
42843 * cproj: Other Builtins. (line 6)
42844 * cprojf: Other Builtins. (line 6)
42845 * cprojl: Other Builtins. (line 6)
42846 * creal: Other Builtins. (line 6)
42847 * crealf: Other Builtins. (line 6)
42848 * creall: Other Builtins. (line 6)
42849 * CRIS Options: CRIS Options. (line 6)
42850 * cross compiling: Target Options. (line 6)
42851 * CRX Options: CRX Options. (line 6)
42852 * csin: Other Builtins. (line 6)
42853 * csinf: Other Builtins. (line 6)
42854 * csinh: Other Builtins. (line 6)
42855 * csinhf: Other Builtins. (line 6)
42856 * csinhl: Other Builtins. (line 6)
42857 * csinl: Other Builtins. (line 6)
42858 * csqrt: Other Builtins. (line 6)
42859 * csqrtf: Other Builtins. (line 6)
42860 * csqrtl: Other Builtins. (line 6)
42861 * ctan: Other Builtins. (line 6)
42862 * ctanf: Other Builtins. (line 6)
42863 * ctanh: Other Builtins. (line 6)
42864 * ctanhf: Other Builtins. (line 6)
42865 * ctanhl: Other Builtins. (line 6)
42866 * ctanl: Other Builtins. (line 6)
42867 * Darwin options: Darwin Options. (line 6)
42868 * dcgettext: Other Builtins. (line 6)
42869 * DD integer suffix: Decimal Float. (line 6)
42870 * dd integer suffix: Decimal Float. (line 6)
42871 * deallocating variable length arrays: Variable Length. (line 23)
42872 * debugging information options: Debugging Options. (line 6)
42873 * decimal floating types: Decimal Float. (line 6)
42874 * declaration scope: Incompatibilities. (line 80)
42875 * declarations inside expressions: Statement Exprs. (line 6)
42876 * declarations, mixed with code: Mixed Declarations. (line 6)
42877 * declaring attributes of functions: Function Attributes.
42879 * declaring static data in C++: Static Definitions. (line 6)
42880 * defining static data in C++: Static Definitions. (line 6)
42881 * dependencies for make as output: Environment Variables.
42883 * dependencies, make: Preprocessor Options.
42885 * DEPENDENCIES_OUTPUT: Environment Variables.
42887 * dependent name lookup: Name lookup. (line 6)
42888 * deprecated attribute: Variable Attributes.
42890 * deprecated attribute.: Function Attributes.
42892 * designated initializers: Designated Inits. (line 6)
42893 * designator lists: Designated Inits. (line 94)
42894 * designators: Designated Inits. (line 61)
42895 * destructor function attribute: Function Attributes.
42897 * DF integer suffix: Decimal Float. (line 6)
42898 * df integer suffix: Decimal Float. (line 6)
42899 * dgettext: Other Builtins. (line 6)
42900 * diagnostic messages: Language Independent Options.
42902 * dialect options: C Dialect Options. (line 6)
42903 * digits in constraint: Simple Constraints. (line 117)
42904 * directory options: Directory Options. (line 6)
42905 * DL integer suffix: Decimal Float. (line 6)
42906 * dl integer suffix: Decimal Float. (line 6)
42907 * dollar signs in identifier names: Dollar Signs. (line 6)
42908 * double-word arithmetic: Long Long. (line 6)
42909 * downward funargs: Nested Functions. (line 6)
42910 * drem: Other Builtins. (line 6)
42911 * dremf: Other Builtins. (line 6)
42912 * dreml: Other Builtins. (line 6)
42913 * E in constraint: Simple Constraints. (line 79)
42914 * earlyclobber operand: Modifiers. (line 25)
42915 * eight bit data on the H8/300, H8/300H, and H8S: Function Attributes.
42917 * empty structures: Empty Structures. (line 6)
42918 * environment variables: Environment Variables.
42920 * erf: Other Builtins. (line 6)
42921 * erfc: Other Builtins. (line 6)
42922 * erfcf: Other Builtins. (line 6)
42923 * erfcl: Other Builtins. (line 6)
42924 * erff: Other Builtins. (line 6)
42925 * erfl: Other Builtins. (line 6)
42926 * error function attribute: Function Attributes.
42928 * error messages: Warnings and Errors.
42930 * escaped newlines: Escaped Newlines. (line 6)
42931 * exception handler functions on the Blackfin processor: Function Attributes.
42933 * exclamation point: Multi-Alternative. (line 33)
42934 * exit: Other Builtins. (line 6)
42935 * exp: Other Builtins. (line 6)
42936 * exp10: Other Builtins. (line 6)
42937 * exp10f: Other Builtins. (line 6)
42938 * exp10l: Other Builtins. (line 6)
42939 * exp2: Other Builtins. (line 6)
42940 * exp2f: Other Builtins. (line 6)
42941 * exp2l: Other Builtins. (line 6)
42942 * expf: Other Builtins. (line 6)
42943 * expl: Other Builtins. (line 6)
42944 * explicit register variables: Explicit Reg Vars. (line 6)
42945 * expm1: Other Builtins. (line 6)
42946 * expm1f: Other Builtins. (line 6)
42947 * expm1l: Other Builtins. (line 6)
42948 * expressions containing statements: Statement Exprs. (line 6)
42949 * expressions, constructor: Compound Literals. (line 6)
42950 * extended asm: Extended Asm. (line 6)
42951 * extensible constraints: Simple Constraints. (line 153)
42952 * extensions, ?:: Conditionals. (line 6)
42953 * extensions, C language: C Extensions. (line 6)
42954 * extensions, C++ language: C++ Extensions. (line 6)
42955 * external declaration scope: Incompatibilities. (line 80)
42956 * externally_visible attribute.: Function Attributes.
42958 * F in constraint: Simple Constraints. (line 84)
42959 * fabs: Other Builtins. (line 6)
42960 * fabsf: Other Builtins. (line 6)
42961 * fabsl: Other Builtins. (line 6)
42962 * fatal signal: Bug Criteria. (line 9)
42963 * fdim: Other Builtins. (line 6)
42964 * fdimf: Other Builtins. (line 6)
42965 * fdiml: Other Builtins. (line 6)
42966 * FDL, GNU Free Documentation License: GNU Free Documentation License.
42968 * ffs: Other Builtins. (line 6)
42969 * file name suffix: Overall Options. (line 14)
42970 * file names: Link Options. (line 10)
42971 * fixed-point types: Fixed-Point. (line 6)
42972 * flatten function attribute: Function Attributes.
42974 * flexible array members: Zero Length. (line 6)
42975 * float as function value type: Incompatibilities. (line 141)
42976 * floating point precision <1>: Disappointments. (line 68)
42977 * floating point precision: Optimize Options. (line 1360)
42978 * floor: Other Builtins. (line 6)
42979 * floorf: Other Builtins. (line 6)
42980 * floorl: Other Builtins. (line 6)
42981 * fma: Other Builtins. (line 6)
42982 * fmaf: Other Builtins. (line 6)
42983 * fmal: Other Builtins. (line 6)
42984 * fmax: Other Builtins. (line 6)
42985 * fmaxf: Other Builtins. (line 6)
42986 * fmaxl: Other Builtins. (line 6)
42987 * fmin: Other Builtins. (line 6)
42988 * fminf: Other Builtins. (line 6)
42989 * fminl: Other Builtins. (line 6)
42990 * fmod: Other Builtins. (line 6)
42991 * fmodf: Other Builtins. (line 6)
42992 * fmodl: Other Builtins. (line 6)
42993 * force_align_arg_pointer attribute: Function Attributes.
42995 * format function attribute: Function Attributes.
42997 * format_arg function attribute: Function Attributes.
42999 * Fortran: G++ and GCC. (line 6)
43000 * forwarding calls: Constructing Calls. (line 6)
43001 * fprintf: Other Builtins. (line 6)
43002 * fprintf_unlocked: Other Builtins. (line 6)
43003 * fputs: Other Builtins. (line 6)
43004 * fputs_unlocked: Other Builtins. (line 6)
43005 * FR30 Options: FR30 Options. (line 6)
43006 * freestanding environment: Standards. (line 13)
43007 * freestanding implementation: Standards. (line 13)
43008 * frexp: Other Builtins. (line 6)
43009 * frexpf: Other Builtins. (line 6)
43010 * frexpl: Other Builtins. (line 6)
43011 * FRV Options: FRV Options. (line 6)
43012 * fscanf: Other Builtins. (line 6)
43013 * fscanf, and constant strings: Incompatibilities. (line 17)
43014 * function addressability on the M32R/D: Function Attributes.
43016 * function attributes: Function Attributes.
43018 * function pointers, arithmetic: Pointer Arith. (line 6)
43019 * function prototype declarations: Function Prototypes.
43021 * function without a prologue/epilogue code: Function Attributes.
43023 * function, size of pointer to: Pointer Arith. (line 6)
43024 * functions called via pointer on the RS/6000 and PowerPC: Function Attributes.
43026 * functions in arbitrary sections: Function Attributes.
43028 * functions that are passed arguments in registers on the 386: Function Attributes.
43030 * functions that behave like malloc: Function Attributes.
43032 * functions that do not pop the argument stack on the 386: Function Attributes.
43034 * functions that do pop the argument stack on the 386: Function Attributes.
43036 * functions that have different compilation options on the 386: Function Attributes.
43038 * functions that have different optimization options: Function Attributes.
43040 * functions that have no side effects: Function Attributes.
43042 * functions that never return: Function Attributes.
43044 * functions that pop the argument stack on the 386: Function Attributes.
43046 * functions that return more than once: Function Attributes.
43048 * functions which do not handle memory bank switching on 68HC11/68HC12: Function Attributes.
43050 * functions which handle memory bank switching: Function Attributes.
43052 * functions with non-null pointer arguments: Function Attributes.
43054 * functions with printf, scanf, strftime or strfmon style arguments: Function Attributes.
43056 * g in constraint: Simple Constraints. (line 110)
43057 * G in constraint: Simple Constraints. (line 88)
43058 * g++: Invoking G++. (line 14)
43059 * G++: G++ and GCC. (line 30)
43060 * gamma: Other Builtins. (line 6)
43061 * gamma_r: Other Builtins. (line 6)
43062 * gammaf: Other Builtins. (line 6)
43063 * gammaf_r: Other Builtins. (line 6)
43064 * gammal: Other Builtins. (line 6)
43065 * gammal_r: Other Builtins. (line 6)
43066 * GCC: G++ and GCC. (line 6)
43067 * GCC command options: Invoking GCC. (line 6)
43068 * GCC_EXEC_PREFIX: Environment Variables.
43070 * gcc_struct: Type Attributes. (line 309)
43071 * gcc_struct attribute: Variable Attributes.
43073 * gcov: Debugging Options. (line 271)
43074 * gettext: Other Builtins. (line 6)
43075 * global offset table: Code Gen Options. (line 184)
43076 * global register after longjmp: Global Reg Vars. (line 66)
43077 * global register variables: Global Reg Vars. (line 6)
43078 * GNAT: G++ and GCC. (line 30)
43079 * GNU C Compiler: G++ and GCC. (line 6)
43080 * GNU Compiler Collection: G++ and GCC. (line 6)
43081 * gnu_inline function attribute: Function Attributes.
43083 * goto with computed label: Labels as Values. (line 6)
43084 * gprof: Debugging Options. (line 232)
43085 * grouping options: Invoking GCC. (line 26)
43086 * H in constraint: Simple Constraints. (line 88)
43087 * hardware models and configurations, specifying: Submodel Options.
43089 * hex floats: Hex Floats. (line 6)
43090 * HK fixed-suffix: Fixed-Point. (line 6)
43091 * hk fixed-suffix: Fixed-Point. (line 6)
43092 * hosted environment <1>: C Dialect Options. (line 204)
43093 * hosted environment: Standards. (line 13)
43094 * hosted implementation: Standards. (line 13)
43095 * hot function attribute: Function Attributes.
43097 * HPPA Options: HPPA Options. (line 6)
43098 * HR fixed-suffix: Fixed-Point. (line 6)
43099 * hr fixed-suffix: Fixed-Point. (line 6)
43100 * hypot: Other Builtins. (line 6)
43101 * hypotf: Other Builtins. (line 6)
43102 * hypotl: Other Builtins. (line 6)
43103 * I in constraint: Simple Constraints. (line 71)
43104 * i in constraint: Simple Constraints. (line 60)
43105 * i386 and x86-64 Windows Options: i386 and x86-64 Windows Options.
43107 * i386 Options: i386 and x86-64 Options.
43109 * IA-64 Options: IA-64 Options. (line 6)
43110 * IBM RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
43112 * identifier names, dollar signs in: Dollar Signs. (line 6)
43113 * identifiers, names in assembler code: Asm Labels. (line 6)
43114 * ilogb: Other Builtins. (line 6)
43115 * ilogbf: Other Builtins. (line 6)
43116 * ilogbl: Other Builtins. (line 6)
43117 * imaxabs: Other Builtins. (line 6)
43118 * implementation-defined behavior, C language: C Implementation.
43120 * implied #pragma implementation: C++ Interface. (line 46)
43121 * incompatibilities of GCC: Incompatibilities. (line 6)
43122 * increment operators: Bug Criteria. (line 17)
43123 * index: Other Builtins. (line 6)
43124 * indirect calls on ARM: Function Attributes.
43126 * indirect calls on MIPS: Function Attributes.
43128 * init_priority attribute: C++ Attributes. (line 9)
43129 * initializations in expressions: Compound Literals. (line 6)
43130 * initializers with labeled elements: Designated Inits. (line 6)
43131 * initializers, non-constant: Initializers. (line 6)
43132 * inline automatic for C++ member fns: Inline. (line 71)
43133 * inline functions: Inline. (line 6)
43134 * inline functions, omission of: Inline. (line 51)
43135 * inlining and C++ pragmas: C++ Interface. (line 66)
43136 * installation trouble: Trouble. (line 6)
43137 * integrating function code: Inline. (line 6)
43138 * Intel 386 Options: i386 and x86-64 Options.
43140 * interface and implementation headers, C++: C++ Interface. (line 6)
43141 * intermediate C version, nonexistent: G++ and GCC. (line 35)
43142 * interrupt handler functions: Function Attributes.
43144 * interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors: Function Attributes.
43146 * interrupt service routines on ARM: Function Attributes.
43148 * interrupt thread functions on fido: Function Attributes.
43150 * introduction: Top. (line 6)
43151 * invalid assembly code: Bug Criteria. (line 12)
43152 * invalid input: Bug Criteria. (line 42)
43153 * invoking g++: Invoking G++. (line 22)
43154 * isalnum: Other Builtins. (line 6)
43155 * isalpha: Other Builtins. (line 6)
43156 * isascii: Other Builtins. (line 6)
43157 * isblank: Other Builtins. (line 6)
43158 * iscntrl: Other Builtins. (line 6)
43159 * isdigit: Other Builtins. (line 6)
43160 * isgraph: Other Builtins. (line 6)
43161 * islower: Other Builtins. (line 6)
43162 * ISO 9899: Standards. (line 13)
43163 * ISO C: Standards. (line 13)
43164 * ISO C standard: Standards. (line 13)
43165 * ISO C90: Standards. (line 13)
43166 * ISO C94: Standards. (line 13)
43167 * ISO C95: Standards. (line 13)
43168 * ISO C99: Standards. (line 13)
43169 * ISO C9X: Standards. (line 13)
43170 * ISO support: C Dialect Options. (line 10)
43171 * ISO/IEC 9899: Standards. (line 13)
43172 * isprint: Other Builtins. (line 6)
43173 * ispunct: Other Builtins. (line 6)
43174 * isspace: Other Builtins. (line 6)
43175 * isupper: Other Builtins. (line 6)
43176 * iswalnum: Other Builtins. (line 6)
43177 * iswalpha: Other Builtins. (line 6)
43178 * iswblank: Other Builtins. (line 6)
43179 * iswcntrl: Other Builtins. (line 6)
43180 * iswdigit: Other Builtins. (line 6)
43181 * iswgraph: Other Builtins. (line 6)
43182 * iswlower: Other Builtins. (line 6)
43183 * iswprint: Other Builtins. (line 6)
43184 * iswpunct: Other Builtins. (line 6)
43185 * iswspace: Other Builtins. (line 6)
43186 * iswupper: Other Builtins. (line 6)
43187 * iswxdigit: Other Builtins. (line 6)
43188 * isxdigit: Other Builtins. (line 6)
43189 * j0: Other Builtins. (line 6)
43190 * j0f: Other Builtins. (line 6)
43191 * j0l: Other Builtins. (line 6)
43192 * j1: Other Builtins. (line 6)
43193 * j1f: Other Builtins. (line 6)
43194 * j1l: Other Builtins. (line 6)
43195 * Java: G++ and GCC. (line 6)
43196 * java_interface attribute: C++ Attributes. (line 29)
43197 * jn: Other Builtins. (line 6)
43198 * jnf: Other Builtins. (line 6)
43199 * jnl: Other Builtins. (line 6)
43200 * K fixed-suffix: Fixed-Point. (line 6)
43201 * k fixed-suffix: Fixed-Point. (line 6)
43202 * keywords, alternate: Alternate Keywords. (line 6)
43203 * known causes of trouble: Trouble. (line 6)
43204 * l1_data variable attribute: Variable Attributes.
43206 * l1_data_A variable attribute: Variable Attributes.
43208 * l1_data_B variable attribute: Variable Attributes.
43210 * l1_text function attribute: Function Attributes.
43212 * labeled elements in initializers: Designated Inits. (line 6)
43213 * labels as values: Labels as Values. (line 6)
43214 * labs: Other Builtins. (line 6)
43215 * LANG: Environment Variables.
43217 * language dialect options: C Dialect Options. (line 6)
43218 * LC_ALL: Environment Variables.
43220 * LC_CTYPE: Environment Variables.
43222 * LC_MESSAGES: Environment Variables.
43224 * ldexp: Other Builtins. (line 6)
43225 * ldexpf: Other Builtins. (line 6)
43226 * ldexpl: Other Builtins. (line 6)
43227 * length-zero arrays: Zero Length. (line 6)
43228 * lgamma: Other Builtins. (line 6)
43229 * lgamma_r: Other Builtins. (line 6)
43230 * lgammaf: Other Builtins. (line 6)
43231 * lgammaf_r: Other Builtins. (line 6)
43232 * lgammal: Other Builtins. (line 6)
43233 * lgammal_r: Other Builtins. (line 6)
43234 * Libraries: Link Options. (line 24)
43235 * LIBRARY_PATH: Environment Variables.
43237 * link options: Link Options. (line 6)
43238 * linker script: Link Options. (line 163)
43239 * LK fixed-suffix: Fixed-Point. (line 6)
43240 * lk fixed-suffix: Fixed-Point. (line 6)
43241 * LL integer suffix: Long Long. (line 6)
43242 * llabs: Other Builtins. (line 6)
43243 * LLK fixed-suffix: Fixed-Point. (line 6)
43244 * llk fixed-suffix: Fixed-Point. (line 6)
43245 * LLR fixed-suffix: Fixed-Point. (line 6)
43246 * llr fixed-suffix: Fixed-Point. (line 6)
43247 * llrint: Other Builtins. (line 6)
43248 * llrintf: Other Builtins. (line 6)
43249 * llrintl: Other Builtins. (line 6)
43250 * llround: Other Builtins. (line 6)
43251 * llroundf: Other Builtins. (line 6)
43252 * llroundl: Other Builtins. (line 6)
43253 * load address instruction: Simple Constraints. (line 144)
43254 * local labels: Local Labels. (line 6)
43255 * local variables in macros: Typeof. (line 42)
43256 * local variables, specifying registers: Local Reg Vars. (line 6)
43257 * locale: Environment Variables.
43259 * locale definition: Environment Variables.
43261 * log: Other Builtins. (line 6)
43262 * log10: Other Builtins. (line 6)
43263 * log10f: Other Builtins. (line 6)
43264 * log10l: Other Builtins. (line 6)
43265 * log1p: Other Builtins. (line 6)
43266 * log1pf: Other Builtins. (line 6)
43267 * log1pl: Other Builtins. (line 6)
43268 * log2: Other Builtins. (line 6)
43269 * log2f: Other Builtins. (line 6)
43270 * log2l: Other Builtins. (line 6)
43271 * logb: Other Builtins. (line 6)
43272 * logbf: Other Builtins. (line 6)
43273 * logbl: Other Builtins. (line 6)
43274 * logf: Other Builtins. (line 6)
43275 * logl: Other Builtins. (line 6)
43276 * long long data types: Long Long. (line 6)
43277 * longjmp: Global Reg Vars. (line 66)
43278 * longjmp incompatibilities: Incompatibilities. (line 39)
43279 * longjmp warnings: Warning Options. (line 570)
43280 * LR fixed-suffix: Fixed-Point. (line 6)
43281 * lr fixed-suffix: Fixed-Point. (line 6)
43282 * lrint: Other Builtins. (line 6)
43283 * lrintf: Other Builtins. (line 6)
43284 * lrintl: Other Builtins. (line 6)
43285 * lround: Other Builtins. (line 6)
43286 * lroundf: Other Builtins. (line 6)
43287 * lroundl: Other Builtins. (line 6)
43288 * m in constraint: Simple Constraints. (line 17)
43289 * M32C options: M32C Options. (line 6)
43290 * M32R/D options: M32R/D Options. (line 6)
43291 * M680x0 options: M680x0 Options. (line 6)
43292 * M68hc1x options: M68hc1x Options. (line 6)
43293 * machine dependent options: Submodel Options. (line 6)
43294 * machine specific constraints: Machine Constraints.
43296 * macro with variable arguments: Variadic Macros. (line 6)
43297 * macros containing asm: Extended Asm. (line 241)
43298 * macros, inline alternative: Inline. (line 6)
43299 * macros, local labels: Local Labels. (line 6)
43300 * macros, local variables in: Typeof. (line 42)
43301 * macros, statements in expressions: Statement Exprs. (line 6)
43302 * macros, types of arguments: Typeof. (line 6)
43303 * make: Preprocessor Options.
43305 * malloc: Other Builtins. (line 6)
43306 * malloc attribute: Function Attributes.
43308 * matching constraint: Simple Constraints. (line 129)
43309 * MCore options: MCore Options. (line 6)
43310 * member fns, automatically inline: Inline. (line 71)
43311 * memchr: Other Builtins. (line 6)
43312 * memcmp: Other Builtins. (line 6)
43313 * memcpy: Other Builtins. (line 6)
43314 * memory references in constraints: Simple Constraints. (line 17)
43315 * mempcpy: Other Builtins. (line 6)
43316 * memset: Other Builtins. (line 6)
43317 * Mercury: G++ and GCC. (line 23)
43318 * message formatting: Language Independent Options.
43320 * messages, warning: Warning Options. (line 6)
43321 * messages, warning and error: Warnings and Errors.
43323 * middle-operands, omitted: Conditionals. (line 6)
43324 * MIPS options: MIPS Options. (line 6)
43325 * mips16 attribute: Function Attributes.
43327 * misunderstandings in C++: C++ Misunderstandings.
43329 * mixed declarations and code: Mixed Declarations. (line 6)
43330 * mktemp, and constant strings: Incompatibilities. (line 13)
43331 * MMIX Options: MMIX Options. (line 6)
43332 * MN10300 options: MN10300 Options. (line 6)
43333 * mode attribute: Variable Attributes.
43335 * modf: Other Builtins. (line 6)
43336 * modff: Other Builtins. (line 6)
43337 * modfl: Other Builtins. (line 6)
43338 * modifiers in constraints: Modifiers. (line 6)
43339 * ms_abi attribute: Function Attributes.
43341 * ms_struct: Type Attributes. (line 309)
43342 * ms_struct attribute: Variable Attributes.
43344 * mudflap: Optimize Options. (line 338)
43345 * multiple alternative constraints: Multi-Alternative. (line 6)
43346 * multiprecision arithmetic: Long Long. (line 6)
43347 * n in constraint: Simple Constraints. (line 65)
43348 * names used in assembler code: Asm Labels. (line 6)
43349 * naming convention, implementation headers: C++ Interface. (line 46)
43350 * nearbyint: Other Builtins. (line 6)
43351 * nearbyintf: Other Builtins. (line 6)
43352 * nearbyintl: Other Builtins. (line 6)
43353 * nested functions: Nested Functions. (line 6)
43354 * newlines (escaped): Escaped Newlines. (line 6)
43355 * nextafter: Other Builtins. (line 6)
43356 * nextafterf: Other Builtins. (line 6)
43357 * nextafterl: Other Builtins. (line 6)
43358 * nexttoward: Other Builtins. (line 6)
43359 * nexttowardf: Other Builtins. (line 6)
43360 * nexttowardl: Other Builtins. (line 6)
43361 * NFC: Warning Options. (line 1076)
43362 * NFKC: Warning Options. (line 1076)
43363 * NMI handler functions on the Blackfin processor: Function Attributes.
43365 * no_instrument_function function attribute: Function Attributes.
43367 * nocommon attribute: Variable Attributes.
43369 * noinline function attribute: Function Attributes.
43371 * nomips16 attribute: Function Attributes.
43373 * non-constant initializers: Initializers. (line 6)
43374 * non-static inline function: Inline. (line 85)
43375 * nonnull function attribute: Function Attributes.
43377 * noreturn function attribute: Function Attributes.
43379 * nothrow function attribute: Function Attributes.
43381 * o in constraint: Simple Constraints. (line 23)
43382 * OBJC_INCLUDE_PATH: Environment Variables.
43384 * Objective-C <1>: Standards. (line 153)
43385 * Objective-C: G++ and GCC. (line 6)
43386 * Objective-C and Objective-C++ options, command line: Objective-C and Objective-C++ Dialect Options.
43388 * Objective-C++ <1>: Standards. (line 153)
43389 * Objective-C++: G++ and GCC. (line 6)
43390 * offsettable address: Simple Constraints. (line 23)
43391 * old-style function definitions: Function Prototypes.
43393 * omitted middle-operands: Conditionals. (line 6)
43394 * open coding: Inline. (line 6)
43395 * openmp parallel: C Dialect Options. (line 221)
43396 * operand constraints, asm: Constraints. (line 6)
43397 * optimize function attribute: Function Attributes.
43399 * optimize options: Optimize Options. (line 6)
43400 * options to control diagnostics formatting: Language Independent Options.
43402 * options to control warnings: Warning Options. (line 6)
43403 * options, C++: C++ Dialect Options.
43405 * options, code generation: Code Gen Options. (line 6)
43406 * options, debugging: Debugging Options. (line 6)
43407 * options, dialect: C Dialect Options. (line 6)
43408 * options, directory search: Directory Options. (line 6)
43409 * options, GCC command: Invoking GCC. (line 6)
43410 * options, grouping: Invoking GCC. (line 26)
43411 * options, linking: Link Options. (line 6)
43412 * options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
43414 * options, optimization: Optimize Options. (line 6)
43415 * options, order: Invoking GCC. (line 30)
43416 * options, preprocessor: Preprocessor Options.
43418 * order of evaluation, side effects: Non-bugs. (line 196)
43419 * order of options: Invoking GCC. (line 30)
43420 * other register constraints: Simple Constraints. (line 153)
43421 * output file option: Overall Options. (line 186)
43422 * overloaded virtual fn, warning: C++ Dialect Options.
43424 * p in constraint: Simple Constraints. (line 144)
43425 * packed attribute: Variable Attributes.
43427 * parameter forward declaration: Variable Length. (line 60)
43428 * parameters, aliased: Code Gen Options. (line 409)
43429 * Pascal: G++ and GCC. (line 23)
43430 * PDP-11 Options: PDP-11 Options. (line 6)
43431 * PIC: Code Gen Options. (line 184)
43432 * picoChip options: picoChip Options. (line 6)
43433 * pmf: Bound member functions.
43435 * pointer arguments: Function Attributes.
43437 * pointer to member function: Bound member functions.
43439 * portions of temporary objects, pointers to: Temporaries. (line 6)
43440 * pow: Other Builtins. (line 6)
43441 * pow10: Other Builtins. (line 6)
43442 * pow10f: Other Builtins. (line 6)
43443 * pow10l: Other Builtins. (line 6)
43444 * PowerPC options: PowerPC Options. (line 6)
43445 * powf: Other Builtins. (line 6)
43446 * powl: Other Builtins. (line 6)
43447 * pragma GCC optimize: Function Specific Option Pragmas.
43449 * pragma GCC pop_options: Function Specific Option Pragmas.
43451 * pragma GCC push_options: Function Specific Option Pragmas.
43453 * pragma GCC reset_options: Function Specific Option Pragmas.
43455 * pragma GCC target: Function Specific Option Pragmas.
43457 * pragma, align: Solaris Pragmas. (line 11)
43458 * pragma, diagnostic: Diagnostic Pragmas. (line 14)
43459 * pragma, extern_prefix: Symbol-Renaming Pragmas.
43461 * pragma, fini: Solaris Pragmas. (line 19)
43462 * pragma, init: Solaris Pragmas. (line 24)
43463 * pragma, long_calls: ARM Pragmas. (line 11)
43464 * pragma, long_calls_off: ARM Pragmas. (line 17)
43465 * pragma, longcall: RS/6000 and PowerPC Pragmas.
43467 * pragma, mark: Darwin Pragmas. (line 11)
43468 * pragma, memregs: M32C Pragmas. (line 7)
43469 * pragma, no_long_calls: ARM Pragmas. (line 14)
43470 * pragma, options align: Darwin Pragmas. (line 14)
43471 * pragma, pop_macro: Push/Pop Macro Pragmas.
43473 * pragma, push_macro: Push/Pop Macro Pragmas.
43475 * pragma, reason for not using: Function Attributes.
43477 * pragma, redefine_extname: Symbol-Renaming Pragmas.
43479 * pragma, segment: Darwin Pragmas. (line 21)
43480 * pragma, unused: Darwin Pragmas. (line 24)
43481 * pragma, visibility: Visibility Pragmas. (line 8)
43482 * pragma, weak: Weak Pragmas. (line 10)
43483 * pragmas: Pragmas. (line 6)
43484 * pragmas in C++, effect on inlining: C++ Interface. (line 66)
43485 * pragmas, interface and implementation: C++ Interface. (line 6)
43486 * pragmas, warning of unknown: Warning Options. (line 587)
43487 * precompiled headers: Precompiled Headers.
43489 * preprocessing numbers: Incompatibilities. (line 173)
43490 * preprocessing tokens: Incompatibilities. (line 173)
43491 * preprocessor options: Preprocessor Options.
43493 * printf: Other Builtins. (line 6)
43494 * printf_unlocked: Other Builtins. (line 6)
43495 * prof: Debugging Options. (line 226)
43496 * progmem variable attribute: Variable Attributes.
43498 * promotion of formal parameters: Function Prototypes.
43500 * pure function attribute: Function Attributes.
43502 * push address instruction: Simple Constraints. (line 144)
43503 * putchar: Other Builtins. (line 6)
43504 * puts: Other Builtins. (line 6)
43505 * Q floating point suffix: Floating Types. (line 6)
43506 * q floating point suffix: Floating Types. (line 6)
43507 * qsort, and global register variables: Global Reg Vars. (line 42)
43508 * question mark: Multi-Alternative. (line 27)
43509 * R fixed-suffix: Fixed-Point. (line 6)
43510 * r fixed-suffix: Fixed-Point. (line 6)
43511 * r in constraint: Simple Constraints. (line 56)
43512 * ranges in case statements: Case Ranges. (line 6)
43513 * read-only strings: Incompatibilities. (line 9)
43514 * register variable after longjmp: Global Reg Vars. (line 66)
43515 * registers: Extended Asm. (line 6)
43516 * registers for local variables: Local Reg Vars. (line 6)
43517 * registers in constraints: Simple Constraints. (line 56)
43518 * registers, global allocation: Explicit Reg Vars. (line 6)
43519 * registers, global variables in: Global Reg Vars. (line 6)
43520 * regparm attribute: Function Attributes.
43522 * relocation truncated to fit (ColdFire): M680x0 Options. (line 325)
43523 * relocation truncated to fit (MIPS): MIPS Options. (line 198)
43524 * remainder: Other Builtins. (line 6)
43525 * remainderf: Other Builtins. (line 6)
43526 * remainderl: Other Builtins. (line 6)
43527 * remquo: Other Builtins. (line 6)
43528 * remquof: Other Builtins. (line 6)
43529 * remquol: Other Builtins. (line 6)
43530 * reordering, warning: C++ Dialect Options.
43532 * reporting bugs: Bugs. (line 6)
43533 * resbank attribute: Function Attributes.
43535 * rest argument (in macro): Variadic Macros. (line 6)
43536 * restricted pointers: Restricted Pointers.
43538 * restricted references: Restricted Pointers.
43540 * restricted this pointer: Restricted Pointers.
43542 * returns_twice attribute: Function Attributes.
43544 * rindex: Other Builtins. (line 6)
43545 * rint: Other Builtins. (line 6)
43546 * rintf: Other Builtins. (line 6)
43547 * rintl: Other Builtins. (line 6)
43548 * round: Other Builtins. (line 6)
43549 * roundf: Other Builtins. (line 6)
43550 * roundl: Other Builtins. (line 6)
43551 * RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
43553 * RTTI: Vague Linkage. (line 43)
43554 * run-time options: Code Gen Options. (line 6)
43555 * s in constraint: Simple Constraints. (line 92)
43556 * S/390 and zSeries Options: S/390 and zSeries Options.
43558 * save all registers on the Blackfin, H8/300, H8/300H, and H8S: Function Attributes.
43560 * scalb: Other Builtins. (line 6)
43561 * scalbf: Other Builtins. (line 6)
43562 * scalbl: Other Builtins. (line 6)
43563 * scalbln: Other Builtins. (line 6)
43564 * scalblnf: Other Builtins. (line 6)
43565 * scalbn: Other Builtins. (line 6)
43566 * scalbnf: Other Builtins. (line 6)
43567 * scanf, and constant strings: Incompatibilities. (line 17)
43568 * scanfnl: Other Builtins. (line 6)
43569 * scope of a variable length array: Variable Length. (line 23)
43570 * scope of declaration: Disappointments. (line 21)
43571 * scope of external declarations: Incompatibilities. (line 80)
43572 * Score Options: Score Options. (line 6)
43573 * search path: Directory Options. (line 6)
43574 * section function attribute: Function Attributes.
43576 * section variable attribute: Variable Attributes.
43578 * sentinel function attribute: Function Attributes.
43580 * setjmp: Global Reg Vars. (line 66)
43581 * setjmp incompatibilities: Incompatibilities. (line 39)
43582 * shared strings: Incompatibilities. (line 9)
43583 * shared variable attribute: Variable Attributes.
43585 * side effect in ?:: Conditionals. (line 20)
43586 * side effects, macro argument: Statement Exprs. (line 35)
43587 * side effects, order of evaluation: Non-bugs. (line 196)
43588 * signal handler functions on the AVR processors: Function Attributes.
43590 * signbit: Other Builtins. (line 6)
43591 * signbitd128: Other Builtins. (line 6)
43592 * signbitd32: Other Builtins. (line 6)
43593 * signbitd64: Other Builtins. (line 6)
43594 * signbitf: Other Builtins. (line 6)
43595 * signbitl: Other Builtins. (line 6)
43596 * signed and unsigned values, comparison warning: Warning Options.
43598 * significand: Other Builtins. (line 6)
43599 * significandf: Other Builtins. (line 6)
43600 * significandl: Other Builtins. (line 6)
43601 * simple constraints: Simple Constraints. (line 6)
43602 * sin: Other Builtins. (line 6)
43603 * sincos: Other Builtins. (line 6)
43604 * sincosf: Other Builtins. (line 6)
43605 * sincosl: Other Builtins. (line 6)
43606 * sinf: Other Builtins. (line 6)
43607 * sinh: Other Builtins. (line 6)
43608 * sinhf: Other Builtins. (line 6)
43609 * sinhl: Other Builtins. (line 6)
43610 * sinl: Other Builtins. (line 6)
43611 * sizeof: Typeof. (line 6)
43612 * smaller data references: M32R/D Options. (line 57)
43613 * smaller data references (PowerPC): RS/6000 and PowerPC Options.
43615 * snprintf: Other Builtins. (line 6)
43616 * SPARC options: SPARC Options. (line 6)
43617 * Spec Files: Spec Files. (line 6)
43618 * specified registers: Explicit Reg Vars. (line 6)
43619 * specifying compiler version and target machine: Target Options.
43621 * specifying hardware config: Submodel Options. (line 6)
43622 * specifying machine version: Target Options. (line 6)
43623 * specifying registers for local variables: Local Reg Vars. (line 6)
43624 * speed of compilation: Precompiled Headers.
43626 * sprintf: Other Builtins. (line 6)
43627 * SPU options: SPU Options. (line 6)
43628 * sqrt: Other Builtins. (line 6)
43629 * sqrtf: Other Builtins. (line 6)
43630 * sqrtl: Other Builtins. (line 6)
43631 * sscanf: Other Builtins. (line 6)
43632 * sscanf, and constant strings: Incompatibilities. (line 17)
43633 * sseregparm attribute: Function Attributes.
43635 * statements inside expressions: Statement Exprs. (line 6)
43636 * static data in C++, declaring and defining: Static Definitions.
43638 * stpcpy: Other Builtins. (line 6)
43639 * stpncpy: Other Builtins. (line 6)
43640 * strcasecmp: Other Builtins. (line 6)
43641 * strcat: Other Builtins. (line 6)
43642 * strchr: Other Builtins. (line 6)
43643 * strcmp: Other Builtins. (line 6)
43644 * strcpy: Other Builtins. (line 6)
43645 * strcspn: Other Builtins. (line 6)
43646 * strdup: Other Builtins. (line 6)
43647 * strfmon: Other Builtins. (line 6)
43648 * strftime: Other Builtins. (line 6)
43649 * string constants: Incompatibilities. (line 9)
43650 * strlen: Other Builtins. (line 6)
43651 * strncasecmp: Other Builtins. (line 6)
43652 * strncat: Other Builtins. (line 6)
43653 * strncmp: Other Builtins. (line 6)
43654 * strncpy: Other Builtins. (line 6)
43655 * strndup: Other Builtins. (line 6)
43656 * strpbrk: Other Builtins. (line 6)
43657 * strrchr: Other Builtins. (line 6)
43658 * strspn: Other Builtins. (line 6)
43659 * strstr: Other Builtins. (line 6)
43660 * struct: Unnamed Fields. (line 6)
43661 * structures: Incompatibilities. (line 146)
43662 * structures, constructor expression: Compound Literals. (line 6)
43663 * submodel options: Submodel Options. (line 6)
43664 * subscripting: Subscripting. (line 6)
43665 * subscripting and function values: Subscripting. (line 6)
43666 * suffixes for C++ source: Invoking G++. (line 6)
43667 * SUNPRO_DEPENDENCIES: Environment Variables.
43669 * suppressing warnings: Warning Options. (line 6)
43670 * surprises in C++: C++ Misunderstandings.
43672 * syntax checking: Warning Options. (line 13)
43673 * syscall_linkage attribute: Function Attributes.
43675 * system headers, warnings from: Warning Options. (line 701)
43676 * sysv_abi attribute: Function Attributes.
43678 * tan: Other Builtins. (line 6)
43679 * tanf: Other Builtins. (line 6)
43680 * tanh: Other Builtins. (line 6)
43681 * tanhf: Other Builtins. (line 6)
43682 * tanhl: Other Builtins. (line 6)
43683 * tanl: Other Builtins. (line 6)
43684 * target function attribute: Function Attributes.
43686 * target machine, specifying: Target Options. (line 6)
43687 * target options: Target Options. (line 6)
43688 * target("abm") attribute: Function Attributes.
43690 * target("aes") attribute: Function Attributes.
43692 * target("align-stringops") attribute: Function Attributes.
43694 * target("arch=ARCH") attribute: Function Attributes.
43696 * target("cld") attribute: Function Attributes.
43698 * target("fancy-math-387") attribute: Function Attributes.
43700 * target("fpmath=FPMATH") attribute: Function Attributes.
43702 * target("fused-madd") attribute: Function Attributes.
43704 * target("ieee-fp") attribute: Function Attributes.
43706 * target("inline-all-stringops") attribute: Function Attributes.
43708 * target("inline-stringops-dynamically") attribute: Function Attributes.
43710 * target("mmx") attribute: Function Attributes.
43712 * target("pclmul") attribute: Function Attributes.
43714 * target("popcnt") attribute: Function Attributes.
43716 * target("recip") attribute: Function Attributes.
43718 * target("sse") attribute: Function Attributes.
43720 * target("sse2") attribute: Function Attributes.
43722 * target("sse3") attribute: Function Attributes.
43724 * target("sse4") attribute: Function Attributes.
43726 * target("sse4.1") attribute: Function Attributes.
43728 * target("sse4.2") attribute: Function Attributes.
43730 * target("sse4a") attribute: Function Attributes.
43732 * target("sse5") attribute: Function Attributes.
43734 * target("ssse3") attribute: Function Attributes.
43736 * target("tune=TUNE") attribute: Function Attributes.
43738 * TC1: Standards. (line 13)
43739 * TC2: Standards. (line 13)
43740 * TC3: Standards. (line 13)
43741 * Technical Corrigenda: Standards. (line 13)
43742 * Technical Corrigendum 1: Standards. (line 13)
43743 * Technical Corrigendum 2: Standards. (line 13)
43744 * Technical Corrigendum 3: Standards. (line 13)
43745 * template instantiation: Template Instantiation.
43747 * temporaries, lifetime of: Temporaries. (line 6)
43748 * tgamma: Other Builtins. (line 6)
43749 * tgammaf: Other Builtins. (line 6)
43750 * tgammal: Other Builtins. (line 6)
43751 * Thread-Local Storage: Thread-Local. (line 6)
43752 * thunks: Nested Functions. (line 6)
43753 * tiny data section on the H8/300H and H8S: Function Attributes.
43755 * TLS: Thread-Local. (line 6)
43756 * tls_model attribute: Variable Attributes.
43758 * TMPDIR: Environment Variables.
43760 * toascii: Other Builtins. (line 6)
43761 * tolower: Other Builtins. (line 6)
43762 * toupper: Other Builtins. (line 6)
43763 * towlower: Other Builtins. (line 6)
43764 * towupper: Other Builtins. (line 6)
43765 * traditional C language: C Dialect Options. (line 250)
43766 * trunc: Other Builtins. (line 6)
43767 * truncf: Other Builtins. (line 6)
43768 * truncl: Other Builtins. (line 6)
43769 * two-stage name lookup: Name lookup. (line 6)
43770 * type alignment: Alignment. (line 6)
43771 * type attributes: Type Attributes. (line 6)
43772 * type_info: Vague Linkage. (line 43)
43773 * typedef names as function parameters: Incompatibilities. (line 97)
43774 * typeof: Typeof. (line 6)
43775 * UHK fixed-suffix: Fixed-Point. (line 6)
43776 * uhk fixed-suffix: Fixed-Point. (line 6)
43777 * UHR fixed-suffix: Fixed-Point. (line 6)
43778 * uhr fixed-suffix: Fixed-Point. (line 6)
43779 * UK fixed-suffix: Fixed-Point. (line 6)
43780 * uk fixed-suffix: Fixed-Point. (line 6)
43781 * ULK fixed-suffix: Fixed-Point. (line 6)
43782 * ulk fixed-suffix: Fixed-Point. (line 6)
43783 * ULL integer suffix: Long Long. (line 6)
43784 * ULLK fixed-suffix: Fixed-Point. (line 6)
43785 * ullk fixed-suffix: Fixed-Point. (line 6)
43786 * ULLR fixed-suffix: Fixed-Point. (line 6)
43787 * ullr fixed-suffix: Fixed-Point. (line 6)
43788 * ULR fixed-suffix: Fixed-Point. (line 6)
43789 * ulr fixed-suffix: Fixed-Point. (line 6)
43790 * undefined behavior: Bug Criteria. (line 17)
43791 * undefined function value: Bug Criteria. (line 17)
43792 * underscores in variables in macros: Typeof. (line 42)
43793 * union: Unnamed Fields. (line 6)
43794 * union, casting to a: Cast to Union. (line 6)
43795 * unions: Incompatibilities. (line 146)
43796 * unknown pragmas, warning: Warning Options. (line 587)
43797 * unresolved references and -nodefaultlibs: Link Options. (line 79)
43798 * unresolved references and -nostdlib: Link Options. (line 79)
43799 * unused attribute.: Function Attributes.
43801 * UR fixed-suffix: Fixed-Point. (line 6)
43802 * ur fixed-suffix: Fixed-Point. (line 6)
43803 * used attribute.: Function Attributes.
43805 * User stack pointer in interrupts on the Blackfin: Function Attributes.
43807 * V in constraint: Simple Constraints. (line 43)
43808 * V850 Options: V850 Options. (line 6)
43809 * vague linkage: Vague Linkage. (line 6)
43810 * value after longjmp: Global Reg Vars. (line 66)
43811 * variable addressability on the IA-64: Function Attributes.
43813 * variable addressability on the M32R/D: Variable Attributes.
43815 * variable alignment: Alignment. (line 6)
43816 * variable attributes: Variable Attributes.
43818 * variable number of arguments: Variadic Macros. (line 6)
43819 * variable-length array scope: Variable Length. (line 23)
43820 * variable-length arrays: Variable Length. (line 6)
43821 * variables in specified registers: Explicit Reg Vars. (line 6)
43822 * variables, local, in macros: Typeof. (line 42)
43823 * variadic macros: Variadic Macros. (line 6)
43824 * VAX options: VAX Options. (line 6)
43825 * version_id attribute: Function Attributes.
43827 * vfprintf: Other Builtins. (line 6)
43828 * vfscanf: Other Builtins. (line 6)
43829 * visibility attribute: Function Attributes.
43831 * VLAs: Variable Length. (line 6)
43832 * void pointers, arithmetic: Pointer Arith. (line 6)
43833 * void, size of pointer to: Pointer Arith. (line 6)
43834 * volatile access: Volatiles. (line 6)
43835 * volatile applied to function: Function Attributes.
43837 * volatile read: Volatiles. (line 6)
43838 * volatile write: Volatiles. (line 6)
43839 * vprintf: Other Builtins. (line 6)
43840 * vscanf: Other Builtins. (line 6)
43841 * vsnprintf: Other Builtins. (line 6)
43842 * vsprintf: Other Builtins. (line 6)
43843 * vsscanf: Other Builtins. (line 6)
43844 * vtable: Vague Linkage. (line 28)
43845 * VxWorks Options: VxWorks Options. (line 6)
43846 * W floating point suffix: Floating Types. (line 6)
43847 * w floating point suffix: Floating Types. (line 6)
43848 * warn_unused_result attribute: Function Attributes.
43850 * warning for comparison of signed and unsigned values: Warning Options.
43852 * warning for overloaded virtual fn: C++ Dialect Options.
43854 * warning for reordering of member initializers: C++ Dialect Options.
43856 * warning for unknown pragmas: Warning Options. (line 587)
43857 * warning function attribute: Function Attributes.
43859 * warning messages: Warning Options. (line 6)
43860 * warnings from system headers: Warning Options. (line 701)
43861 * warnings vs errors: Warnings and Errors.
43863 * weak attribute: Function Attributes.
43865 * weakref attribute: Function Attributes.
43867 * whitespace: Incompatibilities. (line 112)
43868 * X in constraint: Simple Constraints. (line 114)
43869 * X3.159-1989: Standards. (line 13)
43870 * x86-64 options: x86-64 Options. (line 6)
43871 * x86-64 Options: i386 and x86-64 Options.
43873 * Xstormy16 Options: Xstormy16 Options. (line 6)
43874 * Xtensa Options: Xtensa Options. (line 6)
43875 * y0: Other Builtins. (line 6)
43876 * y0f: Other Builtins. (line 6)
43877 * y0l: Other Builtins. (line 6)
43878 * y1: Other Builtins. (line 6)
43879 * y1f: Other Builtins. (line 6)
43880 * y1l: Other Builtins. (line 6)
43881 * yn: Other Builtins. (line 6)
43882 * ynf: Other Builtins. (line 6)
43883 * ynl: Other Builtins. (line 6)
43884 * zero-length arrays: Zero Length. (line 6)
43885 * zero-size structures: Empty Structures. (line 6)
43886 * zSeries options: zSeries Options. (line 6)
43892 Node: G++ and GCC
\7f3776
43893 Node: Standards
\7f5841
43894 Node: Invoking GCC
\7f14816
43895 Node: Option Summary
\7f18645
43896 Node: Overall Options
\7f51393
43897 Node: Invoking G++
\7f65900
43898 Node: C Dialect Options
\7f67423
43899 Node: C++ Dialect Options
\7f81314
43900 Node: Objective-C and Objective-C++ Dialect Options
\7f102304
43901 Node: Language Independent Options
\7f114085
43902 Node: Warning Options
\7f116855
43903 Node: Debugging Options
\7f175202
43904 Node: Optimize Options
\7f214824
43905 Ref: Type-punning
\7f261626
43906 Node: Preprocessor Options
\7f320496
43907 Ref: Wtrigraphs
\7f324581
43908 Ref: dashMF
\7f329329
43909 Ref: fdollars-in-identifiers
\7f339848
43910 Node: Assembler Options
\7f348409
43911 Node: Link Options
\7f349114
43912 Ref: Link Options-Footnote-1
\7f358584
43913 Node: Directory Options
\7f358918
43914 Node: Spec Files
\7f364980
43915 Node: Target Options
\7f385319
43916 Node: Submodel Options
\7f386837
43917 Node: ARC Options
\7f388536
43918 Node: ARM Options
\7f390023
43919 Node: AVR Options
\7f403599
43920 Node: Blackfin Options
\7f405817
43921 Node: CRIS Options
\7f413709
43922 Node: CRX Options
\7f417450
43923 Node: Darwin Options
\7f417875
43924 Node: DEC Alpha Options
\7f425368
43925 Node: DEC Alpha/VMS Options
\7f437284
43926 Node: FR30 Options
\7f437670
43927 Node: FRV Options
\7f438245
43928 Node: GNU/Linux Options
\7f444962
43929 Node: H8/300 Options
\7f445420
43930 Node: HPPA Options
\7f446487
43931 Node: i386 and x86-64 Options
\7f455987
43932 Node: IA-64 Options
\7f484161
43933 Node: M32C Options
\7f491486
43934 Node: M32R/D Options
\7f492777
43935 Node: M680x0 Options
\7f496364
43936 Node: M68hc1x Options
\7f510184
43937 Node: MCore Options
\7f511752
43938 Node: MIPS Options
\7f513260
43939 Node: MMIX Options
\7f539295
43940 Node: MN10300 Options
\7f541777
43941 Node: PDP-11 Options
\7f543199
43942 Node: picoChip Options
\7f545039
43943 Node: PowerPC Options
\7f547238
43944 Node: RS/6000 and PowerPC Options
\7f547474
43945 Node: S/390 and zSeries Options
\7f578221
43946 Node: Score Options
\7f586169
43947 Node: SH Options
\7f586997
43948 Node: SPARC Options
\7f597275
43949 Node: SPU Options
\7f608248
43950 Node: System V Options
\7f611536
43951 Node: V850 Options
\7f612359
43952 Node: VAX Options
\7f615499
43953 Node: VxWorks Options
\7f616047
43954 Node: x86-64 Options
\7f617202
43955 Node: i386 and x86-64 Windows Options
\7f617420
43956 Node: Xstormy16 Options
\7f619739
43957 Node: Xtensa Options
\7f620028
43958 Node: zSeries Options
\7f624175
43959 Node: Code Gen Options
\7f624371
43960 Node: Environment Variables
\7f648950
43961 Node: Precompiled Headers
\7f656846
43962 Node: Running Protoize
\7f663072
43963 Node: C Implementation
\7f669409
43964 Node: Translation implementation
\7f671072
43965 Node: Environment implementation
\7f671646
43966 Node: Identifiers implementation
\7f672196
43967 Node: Characters implementation
\7f673250
43968 Node: Integers implementation
\7f676056
43969 Node: Floating point implementation
\7f677881
43970 Node: Arrays and pointers implementation
\7f680810
43971 Ref: Arrays and pointers implementation-Footnote-1
\7f682245
43972 Node: Hints implementation
\7f682369
43973 Node: Structures unions enumerations and bit-fields implementation
\7f683835
43974 Node: Qualifiers implementation
\7f685821
43975 Node: Declarators implementation
\7f687593
43976 Node: Statements implementation
\7f687935
43977 Node: Preprocessing directives implementation
\7f688262
43978 Node: Library functions implementation
\7f690367
43979 Node: Architecture implementation
\7f691007
43980 Node: Locale-specific behavior implementation
\7f691710
43981 Node: C Extensions
\7f692015
43982 Node: Statement Exprs
\7f696626
43983 Node: Local Labels
\7f701139
43984 Node: Labels as Values
\7f704118
43985 Ref: Labels as Values-Footnote-1
\7f706491
43986 Node: Nested Functions
\7f706674
43987 Node: Constructing Calls
\7f710568
43988 Node: Typeof
\7f715291
43989 Node: Conditionals
\7f718457
43990 Node: Long Long
\7f719348
43991 Node: Complex
\7f720849
43992 Node: Floating Types
\7f723419
43993 Node: Decimal Float
\7f724538
43994 Node: Hex Floats
\7f726527
43995 Node: Fixed-Point
\7f727568
43996 Node: Zero Length
\7f730853
43997 Node: Empty Structures
\7f734131
43998 Node: Variable Length
\7f734547
43999 Node: Variadic Macros
\7f737314
44000 Node: Escaped Newlines
\7f739696
44001 Node: Subscripting
\7f740535
44002 Node: Pointer Arith
\7f741258
44003 Node: Initializers
\7f741826
44004 Node: Compound Literals
\7f742322
44005 Node: Designated Inits
\7f744497
44006 Node: Case Ranges
\7f748152
44007 Node: Cast to Union
\7f748835
44008 Node: Mixed Declarations
\7f749931
44009 Node: Function Attributes
\7f750437
44010 Node: Attribute Syntax
\7f813052
44011 Node: Function Prototypes
\7f823322
44012 Node: C++ Comments
\7f825103
44013 Node: Dollar Signs
\7f825622
44014 Node: Character Escapes
\7f826087
44015 Node: Alignment
\7f826381
44016 Node: Variable Attributes
\7f827755
44017 Ref: i386 Variable Attributes
\7f842345
44018 Node: Type Attributes
\7f848330
44019 Ref: i386 Type Attributes
\7f861951
44020 Ref: PowerPC Type Attributes
\7f862791
44021 Ref: SPU Type Attributes
\7f863653
44022 Node: Inline
\7f863944
44023 Node: Extended Asm
\7f868891
44024 Ref: Example of asm with clobbered asm reg
\7f874977
44025 Node: Constraints
\7f889196
44026 Node: Simple Constraints
\7f890046
44027 Node: Multi-Alternative
\7f896717
44028 Node: Modifiers
\7f898434
44029 Node: Machine Constraints
\7f901328
44030 Node: Asm Labels
\7f933541
44031 Node: Explicit Reg Vars
\7f935217
44032 Node: Global Reg Vars
\7f936825
44033 Node: Local Reg Vars
\7f941375
44034 Node: Alternate Keywords
\7f943816
44035 Node: Incomplete Enums
\7f945244
44036 Node: Function Names
\7f946001
44037 Node: Return Address
\7f948163
44038 Node: Vector Extensions
\7f950960
44039 Node: Offsetof
\7f954462
44040 Node: Atomic Builtins
\7f955276
44041 Node: Object Size Checking
\7f960654
44042 Node: Other Builtins
\7f966082
44043 Node: Target Builtins
\7f990890
44044 Node: Alpha Built-in Functions
\7f991784
44045 Node: ARM iWMMXt Built-in Functions
\7f994783
44046 Node: ARM NEON Intrinsics
\7f1001502
44047 Node: Blackfin Built-in Functions
\7f1209340
44048 Node: FR-V Built-in Functions
\7f1209954
44049 Node: Argument Types
\7f1210813
44050 Node: Directly-mapped Integer Functions
\7f1212569
44051 Node: Directly-mapped Media Functions
\7f1213651
44052 Node: Raw read/write Functions
\7f1220683
44053 Node: Other Built-in Functions
\7f1221595
44054 Node: X86 Built-in Functions
\7f1222784
44055 Node: MIPS DSP Built-in Functions
\7f1267175
44056 Node: MIPS Paired-Single Support
\7f1279622
44057 Node: MIPS Loongson Built-in Functions
\7f1281123
44058 Node: Paired-Single Arithmetic
\7f1287641
44059 Node: Paired-Single Built-in Functions
\7f1288587
44060 Node: MIPS-3D Built-in Functions
\7f1291257
44061 Node: picoChip Built-in Functions
\7f1296632
44062 Node: Other MIPS Built-in Functions
\7f1297994
44063 Node: PowerPC AltiVec Built-in Functions
\7f1298518
44064 Node: SPARC VIS Built-in Functions
\7f1399942
44065 Node: SPU Built-in Functions
\7f1401634
44066 Node: Target Format Checks
\7f1403416
44067 Node: Solaris Format Checks
\7f1403823
44068 Node: Pragmas
\7f1404220
44069 Node: ARM Pragmas
\7f1404914
44070 Node: M32C Pragmas
\7f1405517
44071 Node: RS/6000 and PowerPC Pragmas
\7f1406093
44072 Node: Darwin Pragmas
\7f1406835
44073 Node: Solaris Pragmas
\7f1407902
44074 Node: Symbol-Renaming Pragmas
\7f1409063
44075 Node: Structure-Packing Pragmas
\7f1411685
44076 Node: Weak Pragmas
\7f1413337
44077 Node: Diagnostic Pragmas
\7f1414139
44078 Node: Visibility Pragmas
\7f1416773
44079 Node: Push/Pop Macro Pragmas
\7f1417525
44080 Node: Function Specific Option Pragmas
\7f1418498
44081 Node: Unnamed Fields
\7f1420713
44082 Node: Thread-Local
\7f1422223
44083 Node: C99 Thread-Local Edits
\7f1424332
44084 Node: C++98 Thread-Local Edits
\7f1426344
44085 Node: Binary constants
\7f1429789
44086 Node: C++ Extensions
\7f1430460
44087 Node: Volatiles
\7f1432102
44088 Node: Restricted Pointers
\7f1434778
44089 Node: Vague Linkage
\7f1436372
44090 Node: C++ Interface
\7f1440028
44091 Ref: C++ Interface-Footnote-1
\7f1444325
44092 Node: Template Instantiation
\7f1444462
44093 Node: Bound member functions
\7f1451474
44094 Node: C++ Attributes
\7f1453017
44095 Node: Namespace Association
\7f1454675
44096 Node: Type Traits
\7f1456089
44097 Node: Java Exceptions
\7f1461636
44098 Node: Deprecated Features
\7f1463033
44099 Node: Backwards Compatibility
\7f1465998
44100 Node: Objective-C
\7f1467356
44101 Node: Executing code before main
\7f1467937
44102 Node: What you can and what you cannot do in +load
\7f1470543
44103 Node: Type encoding
\7f1472710
44104 Node: Garbage Collection
\7f1476097
44105 Node: Constant string objects
\7f1478721
44106 Node: compatibility_alias
\7f1481229
44107 Node: Compatibility
\7f1482107
44108 Node: Gcov
\7f1488674
44109 Node: Gcov Intro
\7f1489205
44110 Node: Invoking Gcov
\7f1491921
44111 Node: Gcov and Optimization
\7f1504002
44112 Node: Gcov Data Files
\7f1506655
44113 Node: Cross-profiling
\7f1507793
44114 Node: Trouble
\7f1509619
44115 Node: Actual Bugs
\7f1511175
44116 Node: Cross-Compiler Problems
\7f1511915
44117 Node: Interoperation
\7f1512329
44118 Node: Incompatibilities
\7f1519466
44119 Node: Fixed Headers
\7f1527616
44120 Node: Standard Libraries
\7f1529279
44121 Node: Disappointments
\7f1530651
44122 Node: C++ Misunderstandings
\7f1535009
44123 Node: Static Definitions
\7f1535828
44124 Node: Name lookup
\7f1536881
44125 Ref: Name lookup-Footnote-1
\7f1541659
44126 Node: Temporaries
\7f1541846
44127 Node: Copy Assignment
\7f1543822
44128 Node: Protoize Caveats
\7f1545629
44129 Node: Non-bugs
\7f1549602
44130 Node: Warnings and Errors
\7f1560106
44131 Node: Bugs
\7f1561870
44132 Node: Bug Criteria
\7f1562434
44133 Node: Bug Reporting
\7f1564644
44134 Node: Service
\7f1564865
44135 Node: Contributing
\7f1565684
44136 Node: Funding
\7f1566424
44137 Node: GNU Project
\7f1568913
44138 Node: Copying
\7f1569559
44139 Node: GNU Free Documentation License
\7f1607087
44140 Node: Contributors
\7f1629493
44141 Node: Option Index
\7f1665820
44142 Node: Keyword Index
\7f1825284