1 This is doc/gcc.info, produced by makeinfo version 4.11 from
2 ../../../../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 -pipe -pass-exit-codes
402 -x LANGUAGE -v -### --help[=CLASS[,...]] --target-help
403 --version -wrapper@FILE -fplugin=FILE -fplugin-arg-NAME=ARG
406 *Note Options Controlling C Dialect: C Dialect Options.
407 -ansi -std=STANDARD -fgnu89-inline
409 -fno-asm -fno-builtin -fno-builtin-FUNCTION
410 -fhosted -ffreestanding -fopenmp -fms-extensions
411 -trigraphs -no-integrated-cpp -traditional -traditional-cpp
412 -fallow-single-precision -fcond-mismatch -flax-vector-conversions
413 -fsigned-bitfields -fsigned-char
414 -funsigned-bitfields -funsigned-char
416 _C++ Language Options_
417 *Note Options Controlling C++ Dialect: C++ Dialect Options.
418 -fabi-version=N -fno-access-control -fcheck-new
419 -fconserve-space -ffriend-injection
420 -fno-elide-constructors
421 -fno-enforce-eh-specs
422 -ffor-scope -fno-for-scope -fno-gnu-keywords
423 -fno-implicit-templates
424 -fno-implicit-inline-templates
425 -fno-implement-inlines -fms-extensions
426 -fno-nonansi-builtins -fno-operator-names
427 -fno-optional-diags -fpermissive
428 -frepo -fno-rtti -fstats -ftemplate-depth-N
429 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++
430 -fno-default-inline -fvisibility-inlines-hidden
431 -fvisibility-ms-compat
432 -Wabi -Wctor-dtor-privacy
433 -Wnon-virtual-dtor -Wreorder
434 -Weffc++ -Wstrict-null-sentinel
435 -Wno-non-template-friend -Wold-style-cast
436 -Woverloaded-virtual -Wno-pmf-conversions
439 _Objective-C and Objective-C++ Language Options_
440 *Note Options Controlling Objective-C and Objective-C++ Dialects:
441 Objective-C and Objective-C++ Dialect Options.
442 -fconstant-string-class=CLASS-NAME
443 -fgnu-runtime -fnext-runtime
445 -fobjc-call-cxx-cdtors
446 -fobjc-direct-dispatch
449 -freplace-objc-classes
453 -Wno-protocol -Wselector
454 -Wstrict-selector-match
455 -Wundeclared-selector
457 _Language Independent Options_
458 *Note Options to Control Diagnostic Messages Formatting: Language
461 -fdiagnostics-show-location=[once|every-line]
462 -fdiagnostics-show-option
465 *Note Options to Request or Suppress Warnings: Warning Options.
466 -fsyntax-only -pedantic -pedantic-errors
467 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds
468 -Wno-attributes -Wno-builtin-macro-redefined
469 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual
470 -Wchar-subscripts -Wclobbered -Wcomment
471 -Wconversion -Wcoverage-mismatch -Wno-deprecated
472 -Wno-deprecated-declarations -Wdisabled-optimization
473 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels
475 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2
476 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral
477 -Wformat-security -Wformat-y2k
478 -Wframe-larger-than=LEN -Wignored-qualifiers
479 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int
481 -Wno-int-to-pointer-cast -Wno-invalid-offsetof
482 -Winvalid-pch -Wlarger-than=LEN -Wunsafe-loop-optimizations
483 -Wlogical-op -Wlong-long
484 -Wmain -Wmissing-braces -Wmissing-field-initializers
485 -Wmissing-format-attribute -Wmissing-include-dirs
486 -Wmissing-noreturn -Wno-mudflap
487 -Wno-multichar -Wnonnull -Wno-overflow
488 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded
489 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format
490 -Wpointer-arith -Wno-pointer-to-int-cast
492 -Wreturn-type -Wsequence-point -Wshadow
493 -Wsign-compare -Wsign-conversion -Wstack-protector
494 -Wstrict-aliasing -Wstrict-aliasing=n
495 -Wstrict-overflow -Wstrict-overflow=N
496 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand
497 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized
498 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code
499 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
500 -Wunused-value -Wunused-variable
501 -Wvariadic-macros -Wvla
502 -Wvolatile-register-var -Wwrite-strings
504 _C and Objective-C-only Warning Options_
505 -Wbad-function-cast -Wmissing-declarations
506 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs
507 -Wold-style-declaration -Wold-style-definition
508 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion
509 -Wdeclaration-after-statement -Wpointer-sign
512 *Note Options for Debugging Your Program or GCC: Debugging Options.
513 -dLETTERS -dumpspecs -dumpmachine -dumpversion
514 -fdbg-cnt-list -fdbg-cnt=COUNTER-VALUE-LIST
515 -fdump-noaddr -fdump-unnumbered
516 -fdump-translation-unit[-N]
517 -fdump-class-hierarchy[-N]
518 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline
521 -fdump-tree-original[-N]
522 -fdump-tree-optimized[-N]
523 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias
525 -fdump-tree-ssa[-N] -fdump-tree-pre[-N]
526 -fdump-tree-ccp[-N] -fdump-tree-dce[-N]
527 -fdump-tree-gimple[-raw] -fdump-tree-mudflap[-N]
530 -fdump-tree-phiopt[-N]
531 -fdump-tree-forwprop[-N]
532 -fdump-tree-copyrename[-N]
533 -fdump-tree-nrv -fdump-tree-vect
538 -ftree-vectorizer-verbose=N
539 -fdump-tree-storeccp[-N]
540 -feliminate-dwarf2-dups -feliminate-unused-debug-types
541 -feliminate-unused-debug-symbols -femit-class-debug-always
542 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs
543 -frandom-seed=STRING -fsched-verbose=N
544 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose
545 -ftest-coverage -ftime-report -fvar-tracking
546 -g -gLEVEL -gcoff -gdwarf-2
547 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+
548 -fno-merge-debug-strings -fno-dwarf2-cfi-asm
549 -fdebug-prefix-map=OLD=NEW
550 -femit-struct-debug-baseonly -femit-struct-debug-reduced
551 -femit-struct-debug-detailed[=SPEC-LIST]
552 -p -pg -print-file-name=LIBRARY -print-libgcc-file-name
553 -print-multi-directory -print-multi-lib
554 -print-prog-name=PROGRAM -print-search-dirs -Q
555 -print-sysroot -print-sysroot-headers-suffix
558 _Optimization Options_
559 *Note Options that Control Optimization: Optimize Options.
560 -falign-functions[=N] -falign-jumps[=N]
561 -falign-labels[=N] -falign-loops[=N] -fassociative-math
562 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize
563 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves
564 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping
565 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range
566 -fdata-sections -fdce -fdce
567 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse
568 -fearly-inlining -fexpensive-optimizations -ffast-math
569 -ffinite-math-only -ffloat-store -fforward-propagate
570 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm
571 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining
572 -finline-functions -finline-functions-called-once -finline-limit=N
573 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta
574 -fipa-pure-const -fipa-reference -fipa-struct-reorg
575 -fipa-type-escape -fira-algorithm=ALGORITHM
576 -fira-region=REGION -fira-coalesce -fno-ira-share-save-slots
577 -fno-ira-share-spill-slots -fira-verbose=N
578 -fivopts -fkeep-inline-functions -fkeep-static-consts
579 -floop-block -floop-interchange -floop-strip-mine
580 -fmerge-all-constants -fmerge-constants -fmodulo-sched
581 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap
582 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline
583 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability
584 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2
585 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros
586 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss
587 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls
588 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays
589 -fprofile-correction -fprofile-dir=PATH -fprofile-generate
590 -fprofile-generate=PATH
591 -fprofile-use -fprofile-use=PATH -fprofile-values
592 -freciprocal-math -fregmove -frename-registers -freorder-blocks
593 -freorder-blocks-and-partition -freorder-functions
594 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops
595 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks
596 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous
597 -fsched-stalled-insns-dep[=N] -fsched-stalled-insns[=N]
598 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee
599 -fselective-scheduling -fselective-scheduling2
600 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops
601 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller
602 -fsplit-wide-types -fstack-protector -fstack-protector-all
603 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer
604 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop
605 -ftree-copyrename -ftree-dce
606 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im
607 -ftree-loop-distribution
608 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize
609 -ftree-parallelize-loops=N -ftree-pre -ftree-reassoc
610 -ftree-sink -ftree-sra -ftree-switch-conversion
611 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp
612 -funit-at-a-time -funroll-all-loops -funroll-loops
613 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops
614 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb
617 -O -O0 -O1 -O2 -O3 -Os
619 _Preprocessor Options_
620 *Note Options Controlling the Preprocessor: Preprocessor Options.
626 -include FILE -imacros FILE
627 -iprefix FILE -iwithprefix DIR
628 -iwithprefixbefore DIR -isystem DIR
629 -imultilib DIR -isysroot DIR
630 -M -MM -MF -MG -MP -MQ -MT -nostdinc
631 -P -fworking-directory -remap
632 -trigraphs -undef -UMACRO -Wp,OPTION
633 -Xpreprocessor OPTION
636 *Note Passing Options to the Assembler: Assembler Options.
637 -Wa,OPTION -Xassembler OPTION
640 *Note Options for Linking: Link Options.
641 OBJECT-FILE-NAME -lLIBRARY
642 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic
643 -s -static -static-libgcc -shared -shared-libgcc -symbolic
644 -T SCRIPT -Wl,OPTION -Xlinker OPTION
648 *Note Options for Directory Search: Directory Options.
649 -BPREFIX -IDIR -iquoteDIR -LDIR
650 -specs=FILE -I- --sysroot=DIR
653 *Note Target Options::.
654 -V VERSION -b MACHINE
656 _Machine Dependent Options_
657 *Note Hardware Models and Configurations: Submodel Options.
661 -mmangle-cpu -mcpu=CPU -mtext=TEXT-SECTION
662 -mdata=DATA-SECTION -mrodata=READONLY-DATA-SECTION
665 -mapcs-frame -mno-apcs-frame
667 -mapcs-stack-check -mno-apcs-stack-check
668 -mapcs-float -mno-apcs-float
669 -mapcs-reentrant -mno-apcs-reentrant
670 -msched-prolog -mno-sched-prolog
671 -mlittle-endian -mbig-endian -mwords-little-endian
672 -mfloat-abi=NAME -msoft-float -mhard-float -mfpe
673 -mthumb-interwork -mno-thumb-interwork
674 -mcpu=NAME -march=NAME -mfpu=NAME
675 -mstructure-size-boundary=N
677 -mlong-calls -mno-long-calls
678 -msingle-pic-base -mno-single-pic-base
681 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns
684 -mtpcs-frame -mtpcs-leaf-frame
685 -mcaller-super-interworking -mcallee-super-interworking
692 -mmcu=MCU -msize -minit-stack=N -mno-interrupts
693 -mcall-prologues -mno-tablejump -mtiny-stack -mint8
696 -mcpu=CPU[-SIREVISION]
697 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer
698 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly
699 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library
700 -mno-id-shared-library -mshared-library-id=N
701 -mleaf-id-shared-library -mno-leaf-id-shared-library
702 -msep-data -mno-sep-data -mlong-calls -mno-long-calls
703 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram
707 -mcpu=CPU -march=CPU -mtune=CPU
708 -mmax-stack-frame=N -melinux-stacksize=N
709 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects
710 -mstack-align -mdata-align -mconst-align
711 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt
712 -melf -maout -melinux -mlinux -sim -sim2
713 -mmul-bug-workaround -mno-mul-bug-workaround
719 -all_load -allowable_client -arch -arch_errors_fatal
720 -arch_only -bind_at_load -bundle -bundle_loader
721 -client_name -compatibility_version -current_version
723 -dependency-file -dylib_file -dylinker_install_name
724 -dynamic -dynamiclib -exported_symbols_list
725 -filelist -flat_namespace -force_cpusubtype_ALL
726 -force_flat_namespace -headerpad_max_install_names
728 -image_base -init -install_name -keep_private_externs
729 -multi_module -multiply_defined -multiply_defined_unused
730 -noall_load -no_dead_strip_inits_and_terms
731 -nofixprebinding -nomultidefs -noprebind -noseglinkedit
732 -pagezero_size -prebind -prebind_all_twolevel_modules
733 -private_bundle -read_only_relocs -sectalign
734 -sectobjectsymbols -whyload -seg1addr
735 -sectcreate -sectobjectsymbols -sectorder
736 -segaddr -segs_read_only_addr -segs_read_write_addr
737 -seg_addr_table -seg_addr_table_filename -seglinkedit
738 -segprot -segs_read_only_addr -segs_read_write_addr
739 -single_module -static -sub_library -sub_umbrella
740 -twolevel_namespace -umbrella -undefined
741 -unexported_symbols_list -weak_reference_mismatches
742 -whatsloaded -F -gused -gfull -mmacosx-version-min=VERSION
743 -mkernel -mone-byte-bool
746 -mno-fp-regs -msoft-float -malpha-as -mgas
747 -mieee -mieee-with-inexact -mieee-conformant
748 -mfp-trap-mode=MODE -mfp-rounding-mode=MODE
749 -mtrap-precision=MODE -mbuild-constants
750 -mcpu=CPU-TYPE -mtune=CPU-TYPE
751 -mbwx -mmax -mfix -mcix
752 -mfloat-vax -mfloat-ieee
753 -mexplicit-relocs -msmall-data -mlarge-data
754 -msmall-text -mlarge-text
755 -mmemory-latency=TIME
757 _DEC Alpha/VMS Options_
761 -msmall-model -mno-lsim
764 -mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64
765 -mhard-float -msoft-float
766 -malloc-cc -mfixed-cc -mdword -mno-dword
768 -mmedia -mno-media -mmuladd -mno-muladd
769 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic
770 -mlinked-fp -mlong-calls -malign-labels
771 -mlibrary-pic -macc-4 -macc-8
772 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move
773 -moptimize-membar -mno-optimize-membar
774 -mscc -mno-scc -mcond-exec -mno-cond-exec
775 -mvliw-branch -mno-vliw-branch
776 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec
777 -mno-nested-cond-exec -mtomcat-stats
785 -mrelax -mh -ms -mn -mint32 -malign-300
788 -march=ARCHITECTURE-TYPE
789 -mbig-switch -mdisable-fpregs -mdisable-indexing
790 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld
791 -mfixed-range=REGISTER-RANGE
792 -mjump-in-delay -mlinker-opt -mlong-calls
793 -mlong-load-store -mno-big-switch -mno-disable-fpregs
794 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
795 -mno-jump-in-delay -mno-long-load-store
796 -mno-portable-runtime -mno-soft-float
797 -mno-space-regs -msoft-float -mpa-risc-1-0
798 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
799 -mschedule=CPU-TYPE -mspace-regs -msio -mwsio
800 -munix=UNIX-STD -nolibdld -static -threads
802 _i386 and x86-64 Options_
803 -mtune=CPU-TYPE -march=CPU-TYPE
805 -masm=DIALECT -mno-fancy-math-387
806 -mno-fp-ret-in-387 -msoft-float
807 -mno-wide-multiply -mrtd -malign-double
808 -mpreferred-stack-boundary=NUM
809 -mincoming-stack-boundary=NUM
810 -mcld -mcx16 -msahf -mrecip
811 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx
813 -msse4a -m3dnow -mpopcnt -mabm -msse5
814 -mthreads -mno-align-stringops -minline-all-stringops
815 -minline-stringops-dynamically -mstringop-strategy=ALG
816 -mpush-args -maccumulate-outgoing-args -m128bit-long-double
817 -m96bit-long-double -mregparm=NUM -msseregparm
818 -mveclibabi=TYPE -mpc32 -mpc64 -mpc80 -mstackrealign
819 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs
821 -m32 -m64 -mlarge-data-threshold=NUM
822 -mfused-madd -mno-fused-madd -msse2avx
825 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic
826 -mvolatile-asm-stop -mregister-names -mno-sdata
827 -mconstant-gp -mauto-pic -minline-float-divide-min-latency
828 -minline-float-divide-max-throughput
829 -minline-int-divide-min-latency
830 -minline-int-divide-max-throughput
831 -minline-sqrt-min-latency -minline-sqrt-max-throughput
832 -mno-dwarf2-asm -mearly-stop-bits
833 -mfixed-range=REGISTER-RANGE -mtls-size=TLS-SIZE
834 -mtune=CPU-TYPE -mt -pthread -milp32 -mlp64
835 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec
836 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec
837 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose
838 -mno-sched-prefer-non-data-spec-insns
839 -mno-sched-prefer-non-control-spec-insns
840 -mno-sched-count-spec-in-critical-path
845 -malign-loops -mno-align-loops
848 -mmodel=CODE-SIZE-MODEL-TYPE
850 -mno-flush-func -mflush-func=NAME
851 -mno-flush-trap -mflush-trap=NUMBER
855 -mcpu=CPU -msim -memregs=NUMBER
858 -march=ARCH -mcpu=CPU -mtune=TUNE
859 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
860 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407
861 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020
862 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort
863 -mno-short -mhard-float -m68881 -msoft-float -mpcrel
864 -malign-int -mstrict-align -msep-data -mno-sep-data
865 -mshared-library-id=n -mid-shared-library -mno-id-shared-library
869 -m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12
870 -mauto-incdec -minmax -mlong-calls -mshort
871 -msoft-reg-count=COUNT
874 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates
875 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
876 -m4byte-functions -mno-4byte-functions -mcallgraph-data
877 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
878 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
881 -EL -EB -march=ARCH -mtune=ARCH
882 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2
884 -mips16 -mno-mips16 -mflip-mips16
885 -minterlink-mips16 -mno-interlink-mips16
886 -mabi=ABI -mabicalls -mno-abicalls
887 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot
888 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float
889 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2
891 -msmartmips -mno-smartmips
892 -mpaired-single -mno-paired-single -mdmx -mno-mdmx
893 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc
894 -mlong64 -mlong32 -msym32 -mno-sym32
895 -GNUM -mlocal-sdata -mno-local-sdata
896 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt
897 -membedded-data -mno-embedded-data
898 -muninit-const-in-rodata -mno-uninit-const-in-rodata
899 -mcode-readable=SETTING
900 -msplit-addresses -mno-split-addresses
901 -mexplicit-relocs -mno-explicit-relocs
902 -mcheck-zero-division -mno-check-zero-division
903 -mdivide-traps -mdivide-breaks
904 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls
905 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp
906 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400
907 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120
908 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1
909 -mflush-func=FUNC -mno-flush-func
910 -mbranch-cost=NUM -mbranch-likely -mno-branch-likely
911 -mfp-exceptions -mno-fp-exceptions
912 -mvr4130-align -mno-vr4130-align
915 -mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu
916 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols
917 -melf -mbranch-predict -mno-branch-predict -mbase-addresses
918 -mno-base-addresses -msingle-exit -mno-single-exit
921 -mmult-bug -mno-mult-bug
924 -mreturn-pointer-on-d0
928 -mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10
929 -mbcopy -mbcopy-builtin -mint32 -mno-int16
930 -mint16 -mno-int32 -mfloat32 -mno-float64
931 -mfloat64 -mno-float32 -mabshi -mno-abshi
932 -mbranch-expensive -mbranch-cheap
933 -msplit -mno-split -munix-asm -mdec-asm
936 -mae=AE_TYPE -mvliw-lookahead=N
937 -msymbol-as-address -mno-inefficient-warnings
939 _PowerPC Options_ See RS/6000 and PowerPC Options.
941 _RS/6000 and PowerPC Options_
944 -mpower -mno-power -mpower2 -mno-power2
945 -mpowerpc -mpowerpc64 -mno-powerpc
946 -maltivec -mno-altivec
947 -mpowerpc-gpopt -mno-powerpc-gpopt
948 -mpowerpc-gfxopt -mno-powerpc-gfxopt
949 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd
950 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp
951 -mnew-mnemonics -mold-mnemonics
952 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc
953 -m64 -m32 -mxl-compat -mno-xl-compat -mpe
954 -malign-power -malign-natural
955 -msoft-float -mhard-float -mmultiple -mno-multiple
956 -msingle-float -mdouble-float -msimple-fpu
957 -mstring -mno-string -mupdate -mno-update
958 -mavoid-indexed-addresses -mno-avoid-indexed-addresses
959 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
960 -mstrict-align -mno-strict-align -mrelocatable
961 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
962 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
963 -mdynamic-no-pic -maltivec -mswdiv
964 -mprioritize-restricted-insns=PRIORITY
965 -msched-costly-dep=DEPENDENCE_TYPE
966 -minsert-sched-nops=SCHEME
967 -mcall-sysv -mcall-netbsd
968 -maix-struct-return -msvr4-struct-return
969 -mabi=ABI-TYPE -msecure-plt -mbss-plt
975 -mgen-cell-microcode -mwarn-cell-microcode
979 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double
980 -mprototype -mno-prototype
981 -msim -mmvme -mads -myellowknife -memb -msdata
982 -msdata=OPT -mvxworks -G NUM -pthread
984 _S/390 and zSeries Options_
985 -mtune=CPU-TYPE -march=CPU-TYPE
986 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp
987 -mlong-double-64 -mlong-double-128
988 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack
989 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle
990 -m64 -m31 -mdebug -mno-debug -mesa -mzarch
991 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd
992 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard
999 -mscore5 -mscore5u -mscore7 -mscore7d
1002 -m1 -m2 -m2e -m3 -m3e
1003 -m4-nofpu -m4-single-only -m4-single -m4
1004 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al
1005 -m5-64media -m5-64media-nofpu
1006 -m5-32media -m5-32media-nofpu
1007 -m5-compact -m5-compact-nofpu
1008 -mb -ml -mdalign -mrelax
1009 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave
1010 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace
1011 -mprefergot -musermode -multcost=NUMBER -mdiv=STRATEGY
1012 -mdivsi3_libfunc=NAME -mfixed-range=REGISTER-RANGE
1013 -madjust-unroll -mindexed-addressing -mgettrcost=NUMBER -mpt-fixed
1020 -m32 -m64 -mapp-regs -mno-app-regs
1021 -mfaster-structs -mno-faster-structs
1022 -mfpu -mno-fpu -mhard-float -msoft-float
1023 -mhard-quad-float -msoft-quad-float
1024 -mimpure-text -mno-impure-text -mlittle-endian
1025 -mstack-bias -mno-stack-bias
1026 -munaligned-doubles -mno-unaligned-doubles
1027 -mv8plus -mno-v8plus -mvis -mno-vis
1028 -threads -pthreads -pthread
1031 -mwarn-reloc -merror-reloc
1032 -msafe-dma -munsafe-dma
1034 -msmall-mem -mlarge-mem -mstdmain
1035 -mfixed-range=REGISTER-RANGE
1038 -Qy -Qn -YP,PATHS -Ym,DIR
1041 -mlong-calls -mno-long-calls -mep -mno-ep
1042 -mprolog-function -mno-prolog-function -mspace
1043 -mtda=N -msda=N -mzda=N
1044 -mapp-regs -mno-app-regs
1045 -mdisable-callt -mno-disable-callt
1054 -mrtp -non-static -Bstatic -Bdynamic
1055 -Xbind-lazy -Xbind-now
1057 _x86-64 Options_ See i386 and x86-64 Options.
1059 _i386 and x86-64 Windows Options_
1060 -mconsole -mcygwin -mno-cygwin -mdll
1061 -mnop-fun-dllimport -mthread -mwin32 -mwindows
1067 -mconst16 -mno-const16
1068 -mfused-madd -mno-fused-madd
1069 -mserialize-volatile -mno-serialize-volatile
1070 -mtext-section-literals -mno-text-section-literals
1071 -mtarget-align -mno-target-align
1072 -mlongcalls -mno-longcalls
1074 _zSeries Options_ See S/390 and zSeries Options.
1076 _Code Generation Options_
1077 *Note Options for Code Generation Conventions: Code Gen Options.
1078 -fcall-saved-REG -fcall-used-REG
1079 -ffixed-REG -fexceptions
1080 -fnon-call-exceptions -funwind-tables
1081 -fasynchronous-unwind-tables
1082 -finhibit-size-directive -finstrument-functions
1083 -finstrument-functions-exclude-function-list=SYM,SYM,...
1084 -finstrument-functions-exclude-file-list=FILE,FILE,...
1085 -fno-common -fno-ident
1086 -fpcc-struct-return -fpic -fPIC -fpie -fPIE
1088 -frecord-gcc-switches
1089 -freg-struct-return -fshort-enums
1090 -fshort-double -fshort-wchar
1091 -fverbose-asm -fpack-struct[=N] -fstack-check
1092 -fstack-limit-register=REG -fstack-limit-symbol=SYM
1093 -fno-stack-limit -fargument-alias -fargument-noalias
1094 -fargument-noalias-global -fargument-noalias-anything
1095 -fleading-underscore -ftls-model=MODEL
1096 -ftrapv -fwrapv -fbounds-check
1102 * Overall Options:: Controlling the kind of output:
1103 an executable, object files, assembler files,
1104 or preprocessed source.
1105 * C Dialect Options:: Controlling the variant of C language compiled.
1106 * C++ Dialect Options:: Variations on C++.
1107 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1109 * Language Independent Options:: Controlling how diagnostics should be
1111 * Warning Options:: How picky should the compiler be?
1112 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1113 * Optimize Options:: How much optimization?
1114 * Preprocessor Options:: Controlling header files and macro definitions.
1115 Also, getting dependency information for Make.
1116 * Assembler Options:: Passing options to the assembler.
1117 * Link Options:: Specifying libraries and so on.
1118 * Directory Options:: Where to find header files and libraries.
1119 Where to find the compiler executable files.
1120 * Spec Files:: How to pass switches to sub-processes.
1121 * Target Options:: Running a cross-compiler, or an old version of GCC.
1124 File: gcc.info, Node: Overall Options, Next: Invoking G++, Prev: Option Summary, Up: Invoking GCC
1126 3.2 Options Controlling the Kind of Output
1127 ==========================================
1129 Compilation can involve up to four stages: preprocessing, compilation
1130 proper, assembly and linking, always in that order. GCC is capable of
1131 preprocessing and compiling several files either into several assembler
1132 input files, or into one assembler input file; then each assembler
1133 input file produces an object file, and linking combines all the object
1134 files (those newly compiled, and those specified as input) into an
1137 For any given input file, the file name suffix determines what kind of
1138 compilation is done:
1141 C source code which must be preprocessed.
1144 C source code which should not be preprocessed.
1147 C++ source code which should not be preprocessed.
1150 Objective-C source code. Note that you must link with the
1151 `libobjc' library to make an Objective-C program work.
1154 Objective-C source code which should not be preprocessed.
1158 Objective-C++ source code. Note that you must link with the
1159 `libobjc' library to make an Objective-C++ program work. Note
1160 that `.M' refers to a literal capital M.
1163 Objective-C++ source code which should not be preprocessed.
1166 C, C++, Objective-C or Objective-C++ header file to be turned into
1167 a precompiled header.
1176 C++ source code which must be preprocessed. Note that in `.cxx',
1177 the last two letters must both be literally `x'. Likewise, `.C'
1178 refers to a literal capital C.
1182 Objective-C++ source code which must be preprocessed.
1185 Objective-C++ source code which should not be preprocessed.
1195 C++ header file to be turned into a precompiled header.
1200 Fixed form Fortran source code which should not be preprocessed.
1207 Fixed form Fortran source code which must be preprocessed (with
1208 the traditional preprocessor).
1214 Free form Fortran source code which should not be preprocessed.
1220 Free form Fortran source code which must be preprocessed (with the
1221 traditional preprocessor).
1224 Ada source code file which contains a library unit declaration (a
1225 declaration of a package, subprogram, or generic, or a generic
1226 instantiation), or a library unit renaming declaration (a package,
1227 generic, or subprogram renaming declaration). Such files are also
1231 Ada source code file containing a library unit body (a subprogram
1232 or package body). Such files are also called "bodies".
1239 Assembler code which must be preprocessed.
1242 An object file to be fed straight into linking. Any file name
1243 with no recognized suffix is treated this way.
1245 You can specify the input language explicitly with the `-x' option:
1248 Specify explicitly the LANGUAGE for the following input files
1249 (rather than letting the compiler choose a default based on the
1250 file name suffix). This option applies to all following input
1251 files until the next `-x' option. Possible values for LANGUAGE
1253 c c-header c-cpp-output
1254 c++ c++-header c++-cpp-output
1255 objective-c objective-c-header objective-c-cpp-output
1256 objective-c++ objective-c++-header objective-c++-cpp-output
1257 assembler assembler-with-cpp
1259 f77 f77-cpp-input f95 f95-cpp-input
1263 Turn off any specification of a language, so that subsequent files
1264 are handled according to their file name suffixes (as they are if
1265 `-x' has not been used at all).
1268 Normally the `gcc' program will exit with the code of 1 if any
1269 phase of the compiler returns a non-success return code. If you
1270 specify `-pass-exit-codes', the `gcc' program will instead return
1271 with numerically highest error produced by any phase that returned
1272 an error indication. The C, C++, and Fortran frontends return 4,
1273 if an internal compiler error is encountered.
1275 If you only want some of the stages of compilation, you can use `-x'
1276 (or filename suffixes) to tell `gcc' where to start, and one of the
1277 options `-c', `-S', or `-E' to say where `gcc' is to stop. Note that
1278 some combinations (for example, `-x cpp-output -E') instruct `gcc' to
1282 Compile or assemble the source files, but do not link. The linking
1283 stage simply is not done. The ultimate output is in the form of an
1284 object file for each source file.
1286 By default, the object file name for a source file is made by
1287 replacing the suffix `.c', `.i', `.s', etc., with `.o'.
1289 Unrecognized input files, not requiring compilation or assembly,
1293 Stop after the stage of compilation proper; do not assemble. The
1294 output is in the form of an assembler code file for each
1295 non-assembler input file specified.
1297 By default, the assembler file name for a source file is made by
1298 replacing the suffix `.c', `.i', etc., with `.s'.
1300 Input files that don't require compilation are ignored.
1303 Stop after the preprocessing stage; do not run the compiler
1304 proper. The output is in the form of preprocessed source code,
1305 which is sent to the standard output.
1307 Input files which don't require preprocessing are ignored.
1310 Place output in file FILE. This applies regardless to whatever
1311 sort of output is being produced, whether it be an executable file,
1312 an object file, an assembler file or preprocessed C code.
1314 If `-o' is not specified, the default is to put an executable file
1315 in `a.out', the object file for `SOURCE.SUFFIX' in `SOURCE.o', its
1316 assembler file in `SOURCE.s', a precompiled header file in
1317 `SOURCE.SUFFIX.gch', and all preprocessed C source on standard
1321 Print (on standard error output) the commands executed to run the
1322 stages of compilation. Also print the version number of the
1323 compiler driver program and of the preprocessor and the compiler
1327 Like `-v' except the commands are not executed and all command
1328 arguments are quoted. This is useful for shell scripts to capture
1329 the driver-generated command lines.
1332 Use pipes rather than temporary files for communication between the
1333 various stages of compilation. This fails to work on some systems
1334 where the assembler is unable to read from a pipe; but the GNU
1335 assembler has no trouble.
1338 If you are compiling multiple source files, this option tells the
1339 driver to pass all the source files to the compiler at once (for
1340 those languages for which the compiler can handle this). This
1341 will allow intermodule analysis (IMA) to be performed by the
1342 compiler. Currently the only language for which this is supported
1343 is C. If you pass source files for multiple languages to the
1344 driver, using this option, the driver will invoke the compiler(s)
1345 that support IMA once each, passing each compiler all the source
1346 files appropriate for it. For those languages that do not support
1347 IMA this option will be ignored, and the compiler will be invoked
1348 once for each source file in that language. If you use this
1349 option in conjunction with `-save-temps', the compiler will
1350 generate multiple pre-processed files (one for each source file),
1351 but only one (combined) `.o' or `.s' file.
1354 Print (on the standard output) a description of the command line
1355 options understood by `gcc'. If the `-v' option is also specified
1356 then `--help' will also be passed on to the various processes
1357 invoked by `gcc', so that they can display the command line options
1358 they accept. If the `-Wextra' option has also been specified
1359 (prior to the `--help' option), then command line options which
1360 have no documentation associated with them will also be displayed.
1363 Print (on the standard output) a description of target-specific
1364 command line options for each tool. For some targets extra
1365 target-specific information may also be printed.
1367 `--help={CLASS|[^]QUALIFIER}[,...]'
1368 Print (on the standard output) a description of the command line
1369 options understood by the compiler that fit into all specified
1370 classes and qualifiers. These are the supported classes:
1373 This will display all of the optimization options supported
1377 This will display all of the options controlling warning
1378 messages produced by the compiler.
1381 This will display target-specific options. Unlike the
1382 `--target-help' option however, target-specific options of the
1383 linker and assembler will not be displayed. This is because
1384 those tools do not currently support the extended `--help='
1388 This will display the values recognized by the `--param'
1392 This will display the options supported for LANGUAGE, where
1393 LANGUAGE is the name of one of the languages supported in this
1397 This will display the options that are common to all
1400 These are the supported qualifiers:
1403 Display only those options which are undocumented.
1406 Display options which take an argument that appears after an
1407 equal sign in the same continuous piece of text, such as:
1411 Display options which take an argument that appears as a
1412 separate word following the original option, such as: `-o
1415 Thus for example to display all the undocumented target-specific
1416 switches supported by the compiler the following can be used:
1418 --help=target,undocumented
1420 The sense of a qualifier can be inverted by prefixing it with the
1421 `^' character, so for example to display all binary warning
1422 options (i.e., ones that are either on or off and that do not take
1423 an argument), which have a description the following can be used:
1425 --help=warnings,^joined,^undocumented
1427 The argument to `--help=' should not consist solely of inverted
1430 Combining several classes is possible, although this usually
1431 restricts the output by so much that there is nothing to display.
1432 One case where it does work however is when one of the classes is
1433 TARGET. So for example to display all the target-specific
1434 optimization options the following can be used:
1436 --help=target,optimizers
1438 The `--help=' option can be repeated on the command line. Each
1439 successive use will display its requested class of options,
1440 skipping those that have already been displayed.
1442 If the `-Q' option appears on the command line before the
1443 `--help=' option, then the descriptive text displayed by `--help='
1444 is changed. Instead of describing the displayed options, an
1445 indication is given as to whether the option is enabled, disabled
1446 or set to a specific value (assuming that the compiler knows this
1447 at the point where the `--help=' option is used).
1449 Here is a truncated example from the ARM port of `gcc':
1451 % gcc -Q -mabi=2 --help=target -c
1452 The following options are target specific:
1454 -mabort-on-noreturn [disabled]
1457 The output is sensitive to the effects of previous command line
1458 options, so for example it is possible to find out which
1459 optimizations are enabled at `-O2' by using:
1461 -Q -O2 --help=optimizers
1463 Alternatively you can discover which binary optimizations are
1464 enabled by `-O3' by using:
1466 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1467 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1468 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1471 Display the version number and copyrights of the invoked GCC.
1474 Invoke all subcommands under a wrapper program. It takes a single
1475 comma separated list as an argument, which will be used to invoke
1478 gcc -c t.c -wrapper gdb,--args
1480 This will invoke all subprograms of gcc under "gdb -args", thus
1481 cc1 invocation will be "gdb -args cc1 ...".
1484 Load the plugin code in file NAME.so, assumed to be a shared
1485 object to be dlopen'd by the compiler. The base name of the
1486 shared object file is used to identify the plugin for the purposes
1487 of argument parsing (See `-fplugin-arg-NAME-KEY=VALUE' below).
1488 Each plugin should define the callback functions specified in the
1491 `-fplugin-arg-NAME-KEY=VALUE'
1492 Define an argument called KEY with a value of VALUE for the plugin
1496 Read command-line options from FILE. The options read are
1497 inserted in place of the original @FILE option. If FILE does not
1498 exist, or cannot be read, then the option will be treated
1499 literally, and not removed.
1501 Options in FILE are separated by whitespace. A whitespace
1502 character may be included in an option by surrounding the entire
1503 option in either single or double quotes. Any character
1504 (including a backslash) may be included by prefixing the character
1505 to be included with a backslash. The FILE may itself contain
1506 additional @FILE options; any such options will be processed
1510 File: gcc.info, Node: Invoking G++, Next: C Dialect Options, Prev: Overall Options, Up: Invoking GCC
1512 3.3 Compiling C++ Programs
1513 ==========================
1515 C++ source files conventionally use one of the suffixes `.C', `.cc',
1516 `.cpp', `.CPP', `.c++', `.cp', or `.cxx'; C++ header files often use
1517 `.hh', `.hpp', `.H', or (for shared template code) `.tcc'; and
1518 preprocessed C++ files use the suffix `.ii'. GCC recognizes files with
1519 these names and compiles them as C++ programs even if you call the
1520 compiler the same way as for compiling C programs (usually with the
1523 However, the use of `gcc' does not add the C++ library. `g++' is a
1524 program that calls GCC and treats `.c', `.h' and `.i' files as C++
1525 source files instead of C source files unless `-x' is used, and
1526 automatically specifies linking against the C++ library. This program
1527 is also useful when precompiling a C header file with a `.h' extension
1528 for use in C++ compilations. On many systems, `g++' is also installed
1529 with the name `c++'.
1531 When you compile C++ programs, you may specify many of the same
1532 command-line options that you use for compiling programs in any
1533 language; or command-line options meaningful for C and related
1534 languages; or options that are meaningful only for C++ programs. *Note
1535 Options Controlling C Dialect: C Dialect Options, for explanations of
1536 options for languages related to C. *Note Options Controlling C++
1537 Dialect: C++ Dialect Options, for explanations of options that are
1538 meaningful only for C++ programs.
1541 File: gcc.info, Node: C Dialect Options, Next: C++ Dialect Options, Prev: Invoking G++, Up: Invoking GCC
1543 3.4 Options Controlling C Dialect
1544 =================================
1546 The following options control the dialect of C (or languages derived
1547 from C, such as C++, Objective-C and Objective-C++) that the compiler
1551 In C mode, this is equivalent to `-std=c89'. In C++ mode, it is
1552 equivalent to `-std=c++98'.
1554 This turns off certain features of GCC that are incompatible with
1555 ISO C90 (when compiling C code), or of standard C++ (when
1556 compiling C++ code), such as the `asm' and `typeof' keywords, and
1557 predefined macros such as `unix' and `vax' that identify the type
1558 of system you are using. It also enables the undesirable and
1559 rarely used ISO trigraph feature. For the C compiler, it disables
1560 recognition of C++ style `//' comments as well as the `inline'
1563 The alternate keywords `__asm__', `__extension__', `__inline__'
1564 and `__typeof__' continue to work despite `-ansi'. You would not
1565 want to use them in an ISO C program, of course, but it is useful
1566 to put them in header files that might be included in compilations
1567 done with `-ansi'. Alternate predefined macros such as `__unix__'
1568 and `__vax__' are also available, with or without `-ansi'.
1570 The `-ansi' option does not cause non-ISO programs to be rejected
1571 gratuitously. For that, `-pedantic' is required in addition to
1572 `-ansi'. *Note Warning Options::.
1574 The macro `__STRICT_ANSI__' is predefined when the `-ansi' option
1575 is used. Some header files may notice this macro and refrain from
1576 declaring certain functions or defining certain macros that the
1577 ISO standard doesn't call for; this is to avoid interfering with
1578 any programs that might use these names for other things.
1580 Functions that would normally be built in but do not have semantics
1581 defined by ISO C (such as `alloca' and `ffs') are not built-in
1582 functions when `-ansi' is used. *Note Other built-in functions
1583 provided by GCC: Other Builtins, for details of the functions
1587 Determine the language standard. *Note Language Standards
1588 Supported by GCC: Standards, for details of these standard
1589 versions. This option is currently only supported when compiling
1592 The compiler can accept several base standards, such as `c89' or
1593 `c++98', and GNU dialects of those standards, such as `gnu89' or
1594 `gnu++98'. By specifying a base standard, the compiler will
1595 accept all programs following that standard and those using GNU
1596 extensions that do not contradict it. For example, `-std=c89'
1597 turns off certain features of GCC that are incompatible with ISO
1598 C90, such as the `asm' and `typeof' keywords, but not other GNU
1599 extensions that do not have a meaning in ISO C90, such as omitting
1600 the middle term of a `?:' expression. On the other hand, by
1601 specifying a GNU dialect of a standard, all features the compiler
1602 support are enabled, even when those features change the meaning
1603 of the base standard and some strict-conforming programs may be
1604 rejected. The particular standard is used by `-pedantic' to
1605 identify which features are GNU extensions given that version of
1606 the standard. For example `-std=gnu89 -pedantic' would warn about
1607 C++ style `//' comments, while `-std=gnu99 -pedantic' would not.
1609 A value for this option must be provided; possible values are
1613 Support all ISO C90 programs (certain GNU extensions that
1614 conflict with ISO C90 are disabled). Same as `-ansi' for C
1618 ISO C90 as modified in amendment 1.
1624 ISO C99. Note that this standard is not yet fully supported;
1625 see `http://gcc.gnu.org/gcc-4.4/c99status.html' for more
1626 information. The names `c9x' and `iso9899:199x' are
1630 GNU dialect of ISO C90 (including some C99 features). This is
1631 the default for C code.
1635 GNU dialect of ISO C99. When ISO C99 is fully implemented in
1636 GCC, this will become the default. The name `gnu9x' is
1640 The 1998 ISO C++ standard plus amendments. Same as `-ansi' for
1644 GNU dialect of `-std=c++98'. This is the default for C++
1648 The working draft of the upcoming ISO C++0x standard. This
1649 option enables experimental features that are likely to be
1650 included in C++0x. The working draft is constantly changing,
1651 and any feature that is enabled by this flag may be removed
1652 from future versions of GCC if it is not part of the C++0x
1656 GNU dialect of `-std=c++0x'. This option enables experimental
1657 features that may be removed in future versions of GCC.
1660 The option `-fgnu89-inline' tells GCC to use the traditional GNU
1661 semantics for `inline' functions when in C99 mode. *Note An
1662 Inline Function is As Fast As a Macro: Inline. This option is
1663 accepted and ignored by GCC versions 4.1.3 up to but not including
1664 4.3. In GCC versions 4.3 and later it changes the behavior of GCC
1665 in C99 mode. Using this option is roughly equivalent to adding the
1666 `gnu_inline' function attribute to all inline functions (*note
1667 Function Attributes::).
1669 The option `-fno-gnu89-inline' explicitly tells GCC to use the C99
1670 semantics for `inline' when in C99 or gnu99 mode (i.e., it
1671 specifies the default behavior). This option was first supported
1672 in GCC 4.3. This option is not supported in C89 or gnu89 mode.
1674 The preprocessor macros `__GNUC_GNU_INLINE__' and
1675 `__GNUC_STDC_INLINE__' may be used to check which semantics are in
1676 effect for `inline' functions. *Note Common Predefined Macros:
1677 (cpp)Common Predefined Macros.
1679 `-aux-info FILENAME'
1680 Output to the given filename prototyped declarations for all
1681 functions declared and/or defined in a translation unit, including
1682 those in header files. This option is silently ignored in any
1683 language other than C.
1685 Besides declarations, the file indicates, in comments, the origin
1686 of each declaration (source file and line), whether the
1687 declaration was implicit, prototyped or unprototyped (`I', `N' for
1688 new or `O' for old, respectively, in the first character after the
1689 line number and the colon), and whether it came from a declaration
1690 or a definition (`C' or `F', respectively, in the following
1691 character). In the case of function definitions, a K&R-style list
1692 of arguments followed by their declarations is also provided,
1693 inside comments, after the declaration.
1696 Do not recognize `asm', `inline' or `typeof' as a keyword, so that
1697 code can use these words as identifiers. You can use the keywords
1698 `__asm__', `__inline__' and `__typeof__' instead. `-ansi' implies
1701 In C++, this switch only affects the `typeof' keyword, since `asm'
1702 and `inline' are standard keywords. You may want to use the
1703 `-fno-gnu-keywords' flag instead, which has the same effect. In
1704 C99 mode (`-std=c99' or `-std=gnu99'), this switch only affects
1705 the `asm' and `typeof' keywords, since `inline' is a standard
1709 `-fno-builtin-FUNCTION'
1710 Don't recognize built-in functions that do not begin with
1711 `__builtin_' as prefix. *Note Other built-in functions provided
1712 by GCC: Other Builtins, for details of the functions affected,
1713 including those which are not built-in functions when `-ansi' or
1714 `-std' options for strict ISO C conformance are used because they
1715 do not have an ISO standard meaning.
1717 GCC normally generates special code to handle certain built-in
1718 functions more efficiently; for instance, calls to `alloca' may
1719 become single instructions that adjust the stack directly, and
1720 calls to `memcpy' may become inline copy loops. The resulting
1721 code is often both smaller and faster, but since the function
1722 calls no longer appear as such, you cannot set a breakpoint on
1723 those calls, nor can you change the behavior of the functions by
1724 linking with a different library. In addition, when a function is
1725 recognized as a built-in function, GCC may use information about
1726 that function to warn about problems with calls to that function,
1727 or to generate more efficient code, even if the resulting code
1728 still contains calls to that function. For example, warnings are
1729 given with `-Wformat' for bad calls to `printf', when `printf' is
1730 built in, and `strlen' is known not to modify global memory.
1732 With the `-fno-builtin-FUNCTION' option only the built-in function
1733 FUNCTION is disabled. FUNCTION must not begin with `__builtin_'.
1734 If a function is named that is not built-in in this version of
1735 GCC, this option is ignored. There is no corresponding
1736 `-fbuiltin-FUNCTION' option; if you wish to enable built-in
1737 functions selectively when using `-fno-builtin' or
1738 `-ffreestanding', you may define macros such as:
1740 #define abs(n) __builtin_abs ((n))
1741 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1744 Assert that compilation takes place in a hosted environment. This
1745 implies `-fbuiltin'. A hosted environment is one in which the
1746 entire standard library is available, and in which `main' has a
1747 return type of `int'. Examples are nearly everything except a
1748 kernel. This is equivalent to `-fno-freestanding'.
1751 Assert that compilation takes place in a freestanding environment.
1752 This implies `-fno-builtin'. A freestanding environment is one in
1753 which the standard library may not exist, and program startup may
1754 not necessarily be at `main'. The most obvious example is an OS
1755 kernel. This is equivalent to `-fno-hosted'.
1757 *Note Language Standards Supported by GCC: Standards, for details
1758 of freestanding and hosted environments.
1761 Enable handling of OpenMP directives `#pragma omp' in C/C++ and
1762 `!$omp' in Fortran. When `-fopenmp' is specified, the compiler
1763 generates parallel code according to the OpenMP Application
1764 Program Interface v2.5 `http://www.openmp.org/'. This option
1765 implies `-pthread', and thus is only supported on targets that
1766 have support for `-pthread'.
1769 Accept some non-standard constructs used in Microsoft header files.
1771 Some cases of unnamed fields in structures and unions are only
1772 accepted with this option. *Note Unnamed struct/union fields
1773 within structs/unions: Unnamed Fields, for details.
1776 Support ISO C trigraphs. The `-ansi' option (and `-std' options
1777 for strict ISO C conformance) implies `-trigraphs'.
1779 `-no-integrated-cpp'
1780 Performs a compilation in two passes: preprocessing and compiling.
1781 This option allows a user supplied "cc1", "cc1plus", or "cc1obj"
1782 via the `-B' option. The user supplied compilation step can then
1783 add in an additional preprocessing step after normal preprocessing
1784 but before compiling. The default is to use the integrated cpp
1787 The semantics of this option will change if "cc1", "cc1plus", and
1788 "cc1obj" are merged.
1792 Formerly, these options caused GCC to attempt to emulate a
1793 pre-standard C compiler. They are now only supported with the
1794 `-E' switch. The preprocessor continues to support a pre-standard
1795 mode. See the GNU CPP manual for details.
1798 Allow conditional expressions with mismatched types in the second
1799 and third arguments. The value of such an expression is void.
1800 This option is not supported for C++.
1802 `-flax-vector-conversions'
1803 Allow implicit conversions between vectors with differing numbers
1804 of elements and/or incompatible element types. This option should
1805 not be used for new code.
1808 Let the type `char' be unsigned, like `unsigned char'.
1810 Each kind of machine has a default for what `char' should be. It
1811 is either like `unsigned char' by default or like `signed char' by
1814 Ideally, a portable program should always use `signed char' or
1815 `unsigned char' when it depends on the signedness of an object.
1816 But many programs have been written to use plain `char' and expect
1817 it to be signed, or expect it to be unsigned, depending on the
1818 machines they were written for. This option, and its inverse, let
1819 you make such a program work with the opposite default.
1821 The type `char' is always a distinct type from each of `signed
1822 char' or `unsigned char', even though its behavior is always just
1823 like one of those two.
1826 Let the type `char' be signed, like `signed char'.
1828 Note that this is equivalent to `-fno-unsigned-char', which is the
1829 negative form of `-funsigned-char'. Likewise, the option
1830 `-fno-signed-char' is equivalent to `-funsigned-char'.
1832 `-fsigned-bitfields'
1833 `-funsigned-bitfields'
1834 `-fno-signed-bitfields'
1835 `-fno-unsigned-bitfields'
1836 These options control whether a bit-field is signed or unsigned,
1837 when the declaration does not use either `signed' or `unsigned'.
1838 By default, such a bit-field is signed, because this is
1839 consistent: the basic integer types such as `int' are signed types.
1842 File: gcc.info, Node: C++ Dialect Options, Next: Objective-C and Objective-C++ Dialect Options, Prev: C Dialect Options, Up: Invoking GCC
1844 3.5 Options Controlling C++ Dialect
1845 ===================================
1847 This section describes the command-line options that are only meaningful
1848 for C++ programs; but you can also use most of the GNU compiler options
1849 regardless of what language your program is in. For example, you might
1850 compile a file `firstClass.C' like this:
1852 g++ -g -frepo -O -c firstClass.C
1854 In this example, only `-frepo' is an option meant only for C++
1855 programs; you can use the other options with any language supported by
1858 Here is a list of options that are _only_ for compiling C++ programs:
1861 Use version N of the C++ ABI. Version 2 is the version of the C++
1862 ABI that first appeared in G++ 3.4. Version 1 is the version of
1863 the C++ ABI that first appeared in G++ 3.2. Version 0 will always
1864 be the version that conforms most closely to the C++ ABI
1865 specification. Therefore, the ABI obtained using version 0 will
1866 change as ABI bugs are fixed.
1868 The default is version 2.
1870 `-fno-access-control'
1871 Turn off all access checking. This switch is mainly useful for
1872 working around bugs in the access control code.
1875 Check that the pointer returned by `operator new' is non-null
1876 before attempting to modify the storage allocated. This check is
1877 normally unnecessary because the C++ standard specifies that
1878 `operator new' will only return `0' if it is declared `throw()',
1879 in which case the compiler will always check the return value even
1880 without this option. In all other cases, when `operator new' has
1881 a non-empty exception specification, memory exhaustion is
1882 signalled by throwing `std::bad_alloc'. See also `new (nothrow)'.
1885 Put uninitialized or runtime-initialized global variables into the
1886 common segment, as C does. This saves space in the executable at
1887 the cost of not diagnosing duplicate definitions. If you compile
1888 with this flag and your program mysteriously crashes after
1889 `main()' has completed, you may have an object that is being
1890 destroyed twice because two definitions were merged.
1892 This option is no longer useful on most targets, now that support
1893 has been added for putting variables into BSS without making them
1896 `-ffriend-injection'
1897 Inject friend functions into the enclosing namespace, so that they
1898 are visible outside the scope of the class in which they are
1899 declared. Friend functions were documented to work this way in
1900 the old Annotated C++ Reference Manual, and versions of G++ before
1901 4.1 always worked that way. However, in ISO C++ a friend function
1902 which is not declared in an enclosing scope can only be found
1903 using argument dependent lookup. This option causes friends to be
1904 injected as they were in earlier releases.
1906 This option is for compatibility, and may be removed in a future
1909 `-fno-elide-constructors'
1910 The C++ standard allows an implementation to omit creating a
1911 temporary which is only used to initialize another object of the
1912 same type. Specifying this option disables that optimization, and
1913 forces G++ to call the copy constructor in all cases.
1915 `-fno-enforce-eh-specs'
1916 Don't generate code to check for violation of exception
1917 specifications at runtime. This option violates the C++ standard,
1918 but may be useful for reducing code size in production builds,
1919 much like defining `NDEBUG'. This does not give user code
1920 permission to throw exceptions in violation of the exception
1921 specifications; the compiler will still optimize based on the
1922 specifications, so throwing an unexpected exception will result in
1927 If `-ffor-scope' is specified, the scope of variables declared in
1928 a for-init-statement is limited to the `for' loop itself, as
1929 specified by the C++ standard. If `-fno-for-scope' is specified,
1930 the scope of variables declared in a for-init-statement extends to
1931 the end of the enclosing scope, as was the case in old versions of
1932 G++, and other (traditional) implementations of C++.
1934 The default if neither flag is given to follow the standard, but
1935 to allow and give a warning for old-style code that would
1936 otherwise be invalid, or have different behavior.
1939 Do not recognize `typeof' as a keyword, so that code can use this
1940 word as an identifier. You can use the keyword `__typeof__'
1941 instead. `-ansi' implies `-fno-gnu-keywords'.
1943 `-fno-implicit-templates'
1944 Never emit code for non-inline templates which are instantiated
1945 implicitly (i.e. by use); only emit code for explicit
1946 instantiations. *Note Template Instantiation::, for more
1949 `-fno-implicit-inline-templates'
1950 Don't emit code for implicit instantiations of inline templates,
1951 either. The default is to handle inlines differently so that
1952 compiles with and without optimization will need the same set of
1953 explicit instantiations.
1955 `-fno-implement-inlines'
1956 To save space, do not emit out-of-line copies of inline functions
1957 controlled by `#pragma implementation'. This will cause linker
1958 errors if these functions are not inlined everywhere they are
1962 Disable pedantic warnings about constructs used in MFC, such as
1963 implicit int and getting a pointer to member function via
1964 non-standard syntax.
1966 `-fno-nonansi-builtins'
1967 Disable built-in declarations of functions that are not mandated by
1968 ANSI/ISO C. These include `ffs', `alloca', `_exit', `index',
1969 `bzero', `conjf', and other related functions.
1971 `-fno-operator-names'
1972 Do not treat the operator name keywords `and', `bitand', `bitor',
1973 `compl', `not', `or' and `xor' as synonyms as keywords.
1975 `-fno-optional-diags'
1976 Disable diagnostics that the standard says a compiler does not
1977 need to issue. Currently, the only such diagnostic issued by G++
1978 is the one for a name having multiple meanings within a class.
1981 Downgrade some diagnostics about nonconformant code from errors to
1982 warnings. Thus, using `-fpermissive' will allow some
1983 nonconforming code to compile.
1986 Enable automatic template instantiation at link time. This option
1987 also implies `-fno-implicit-templates'. *Note Template
1988 Instantiation::, for more information.
1991 Disable generation of information about every class with virtual
1992 functions for use by the C++ runtime type identification features
1993 (`dynamic_cast' and `typeid'). If you don't use those parts of
1994 the language, you can save some space by using this flag. Note
1995 that exception handling uses the same information, but it will
1996 generate it as needed. The `dynamic_cast' operator can still be
1997 used for casts that do not require runtime type information, i.e.
1998 casts to `void *' or to unambiguous base classes.
2001 Emit statistics about front-end processing at the end of the
2002 compilation. This information is generally only useful to the G++
2005 `-ftemplate-depth-N'
2006 Set the maximum instantiation depth for template classes to N. A
2007 limit on the template instantiation depth is needed to detect
2008 endless recursions during template class instantiation. ANSI/ISO
2009 C++ conforming programs must not rely on a maximum depth greater
2012 `-fno-threadsafe-statics'
2013 Do not emit the extra code to use the routines specified in the C++
2014 ABI for thread-safe initialization of local statics. You can use
2015 this option to reduce code size slightly in code that doesn't need
2019 Register destructors for objects with static storage duration with
2020 the `__cxa_atexit' function rather than the `atexit' function.
2021 This option is required for fully standards-compliant handling of
2022 static destructors, but will only work if your C library supports
2025 `-fno-use-cxa-get-exception-ptr'
2026 Don't use the `__cxa_get_exception_ptr' runtime routine. This
2027 will cause `std::uncaught_exception' to be incorrect, but is
2028 necessary if the runtime routine is not available.
2030 `-fvisibility-inlines-hidden'
2031 This switch declares that the user does not attempt to compare
2032 pointers to inline methods where the addresses of the two functions
2033 were taken in different shared objects.
2035 The effect of this is that GCC may, effectively, mark inline
2036 methods with `__attribute__ ((visibility ("hidden")))' so that
2037 they do not appear in the export table of a DSO and do not require
2038 a PLT indirection when used within the DSO. Enabling this option
2039 can have a dramatic effect on load and link times of a DSO as it
2040 massively reduces the size of the dynamic export table when the
2041 library makes heavy use of templates.
2043 The behavior of this switch is not quite the same as marking the
2044 methods as hidden directly, because it does not affect static
2045 variables local to the function or cause the compiler to deduce
2046 that the function is defined in only one shared object.
2048 You may mark a method as having a visibility explicitly to negate
2049 the effect of the switch for that method. For example, if you do
2050 want to compare pointers to a particular inline method, you might
2051 mark it as having default visibility. Marking the enclosing class
2052 with explicit visibility will have no effect.
2054 Explicitly instantiated inline methods are unaffected by this
2055 option as their linkage might otherwise cross a shared library
2056 boundary. *Note Template Instantiation::.
2058 `-fvisibility-ms-compat'
2059 This flag attempts to use visibility settings to make GCC's C++
2060 linkage model compatible with that of Microsoft Visual Studio.
2062 The flag makes these changes to GCC's linkage model:
2064 1. It sets the default visibility to `hidden', like
2065 `-fvisibility=hidden'.
2067 2. Types, but not their members, are not hidden by default.
2069 3. The One Definition Rule is relaxed for types without explicit
2070 visibility specifications which are defined in more than one
2071 different shared object: those declarations are permitted if
2072 they would have been permitted when this option was not used.
2074 In new code it is better to use `-fvisibility=hidden' and export
2075 those classes which are intended to be externally visible.
2076 Unfortunately it is possible for code to rely, perhaps
2077 accidentally, on the Visual Studio behavior.
2079 Among the consequences of these changes are that static data
2080 members of the same type with the same name but defined in
2081 different shared objects will be different, so changing one will
2082 not change the other; and that pointers to function members
2083 defined in different shared objects may not compare equal. When
2084 this flag is given, it is a violation of the ODR to define types
2085 with the same name differently.
2088 Do not use weak symbol support, even if it is provided by the
2089 linker. By default, G++ will use weak symbols if they are
2090 available. This option exists only for testing, and should not be
2091 used by end-users; it will result in inferior code and has no
2092 benefits. This option may be removed in a future release of G++.
2095 Do not search for header files in the standard directories
2096 specific to C++, but do still search the other standard
2097 directories. (This option is used when building the C++ library.)
2099 In addition, these optimization, warning, and code generation options
2100 have meanings only for C++ programs:
2102 `-fno-default-inline'
2103 Do not assume `inline' for functions defined inside a class scope.
2104 *Note Options That Control Optimization: Optimize Options. Note
2105 that these functions will have linkage like inline functions; they
2106 just won't be inlined by default.
2108 `-Wabi (C, Objective-C, C++ and Objective-C++ only)'
2109 Warn when G++ generates code that is probably not compatible with
2110 the vendor-neutral C++ ABI. Although an effort has been made to
2111 warn about all such cases, there are probably some cases that are
2112 not warned about, even though G++ is generating incompatible code.
2113 There may also be cases where warnings are emitted even though the
2114 code that is generated will be compatible.
2116 You should rewrite your code to avoid these warnings if you are
2117 concerned about the fact that code generated by G++ may not be
2118 binary compatible with code generated by other compilers.
2120 The known incompatibilities at this point include:
2122 * Incorrect handling of tail-padding for bit-fields. G++ may
2123 attempt to pack data into the same byte as a base class. For
2126 struct A { virtual void f(); int f1 : 1; };
2127 struct B : public A { int f2 : 1; };
2129 In this case, G++ will place `B::f2' into the same byte
2130 as`A::f1'; other compilers will not. You can avoid this
2131 problem by explicitly padding `A' so that its size is a
2132 multiple of the byte size on your platform; that will cause
2133 G++ and other compilers to layout `B' identically.
2135 * Incorrect handling of tail-padding for virtual bases. G++
2136 does not use tail padding when laying out virtual bases. For
2139 struct A { virtual void f(); char c1; };
2140 struct B { B(); char c2; };
2141 struct C : public A, public virtual B {};
2143 In this case, G++ will not place `B' into the tail-padding for
2144 `A'; other compilers will. You can avoid this problem by
2145 explicitly padding `A' so that its size is a multiple of its
2146 alignment (ignoring virtual base classes); that will cause
2147 G++ and other compilers to layout `C' identically.
2149 * Incorrect handling of bit-fields with declared widths greater
2150 than that of their underlying types, when the bit-fields
2151 appear in a union. For example:
2153 union U { int i : 4096; };
2155 Assuming that an `int' does not have 4096 bits, G++ will make
2156 the union too small by the number of bits in an `int'.
2158 * Empty classes can be placed at incorrect offsets. For
2168 struct C : public B, public A {};
2170 G++ will place the `A' base class of `C' at a nonzero offset;
2171 it should be placed at offset zero. G++ mistakenly believes
2172 that the `A' data member of `B' is already at offset zero.
2174 * Names of template functions whose types involve `typename' or
2175 template template parameters can be mangled incorrectly.
2177 template <typename Q>
2178 void f(typename Q::X) {}
2180 template <template <typename> class Q>
2181 void f(typename Q<int>::X) {}
2183 Instantiations of these templates may be mangled incorrectly.
2186 It also warns psABI related changes. The known psABI changes at
2189 * For SYSV/x86-64, when passing union with long double, it is
2190 changed to pass in memory as specified in psABI. For example:
2197 `union U' will always be passed in memory.
2200 `-Wctor-dtor-privacy (C++ and Objective-C++ only)'
2201 Warn when a class seems unusable because all the constructors or
2202 destructors in that class are private, and it has neither friends
2203 nor public static member functions.
2205 `-Wnon-virtual-dtor (C++ and Objective-C++ only)'
2206 Warn when a class has virtual functions and accessible non-virtual
2207 destructor, in which case it would be possible but unsafe to delete
2208 an instance of a derived class through a pointer to the base class.
2209 This warning is also enabled if -Weffc++ is specified.
2211 `-Wreorder (C++ and Objective-C++ only)'
2212 Warn when the order of member initializers given in the code does
2213 not match the order in which they must be executed. For instance:
2218 A(): j (0), i (1) { }
2221 The compiler will rearrange the member initializers for `i' and
2222 `j' to match the declaration order of the members, emitting a
2223 warning to that effect. This warning is enabled by `-Wall'.
2225 The following `-W...' options are not affected by `-Wall'.
2227 `-Weffc++ (C++ and Objective-C++ only)'
2228 Warn about violations of the following style guidelines from Scott
2229 Meyers' `Effective C++' book:
2231 * Item 11: Define a copy constructor and an assignment
2232 operator for classes with dynamically allocated memory.
2234 * Item 12: Prefer initialization to assignment in constructors.
2236 * Item 14: Make destructors virtual in base classes.
2238 * Item 15: Have `operator=' return a reference to `*this'.
2240 * Item 23: Don't try to return a reference when you must
2244 Also warn about violations of the following style guidelines from
2245 Scott Meyers' `More Effective C++' book:
2247 * Item 6: Distinguish between prefix and postfix forms of
2248 increment and decrement operators.
2250 * Item 7: Never overload `&&', `||', or `,'.
2253 When selecting this option, be aware that the standard library
2254 headers do not obey all of these guidelines; use `grep -v' to
2255 filter out those warnings.
2257 `-Wstrict-null-sentinel (C++ and Objective-C++ only)'
2258 Warn also about the use of an uncasted `NULL' as sentinel. When
2259 compiling only with GCC this is a valid sentinel, as `NULL' is
2260 defined to `__null'. Although it is a null pointer constant not a
2261 null pointer, it is guaranteed to be of the same size as a
2262 pointer. But this use is not portable across different compilers.
2264 `-Wno-non-template-friend (C++ and Objective-C++ only)'
2265 Disable warnings when non-templatized friend functions are declared
2266 within a template. Since the advent of explicit template
2267 specification support in G++, if the name of the friend is an
2268 unqualified-id (i.e., `friend foo(int)'), the C++ language
2269 specification demands that the friend declare or define an
2270 ordinary, nontemplate function. (Section 14.5.3). Before G++
2271 implemented explicit specification, unqualified-ids could be
2272 interpreted as a particular specialization of a templatized
2273 function. Because this non-conforming behavior is no longer the
2274 default behavior for G++, `-Wnon-template-friend' allows the
2275 compiler to check existing code for potential trouble spots and is
2276 on by default. This new compiler behavior can be turned off with
2277 `-Wno-non-template-friend' which keeps the conformant compiler code
2278 but disables the helpful warning.
2280 `-Wold-style-cast (C++ and Objective-C++ only)'
2281 Warn if an old-style (C-style) cast to a non-void type is used
2282 within a C++ program. The new-style casts (`dynamic_cast',
2283 `static_cast', `reinterpret_cast', and `const_cast') are less
2284 vulnerable to unintended effects and much easier to search for.
2286 `-Woverloaded-virtual (C++ and Objective-C++ only)'
2287 Warn when a function declaration hides virtual functions from a
2288 base class. For example, in:
2294 struct B: public A {
2298 the `A' class version of `f' is hidden in `B', and code like:
2303 will fail to compile.
2305 `-Wno-pmf-conversions (C++ and Objective-C++ only)'
2306 Disable the diagnostic for converting a bound pointer to member
2307 function to a plain pointer.
2309 `-Wsign-promo (C++ and Objective-C++ only)'
2310 Warn when overload resolution chooses a promotion from unsigned or
2311 enumerated type to a signed type, over a conversion to an unsigned
2312 type of the same size. Previous versions of G++ would try to
2313 preserve unsignedness, but the standard mandates the current
2318 A& operator = (int);
2327 In this example, G++ will synthesize a default `A& operator =
2328 (const A&);', while cfront will use the user-defined `operator ='.
2331 File: gcc.info, Node: Objective-C and Objective-C++ Dialect Options, Next: Language Independent Options, Prev: C++ Dialect Options, Up: Invoking GCC
2333 3.6 Options Controlling Objective-C and Objective-C++ Dialects
2334 ==============================================================
2336 (NOTE: This manual does not describe the Objective-C and Objective-C++
2337 languages themselves. See *Note Language Standards Supported by GCC:
2338 Standards, for references.)
2340 This section describes the command-line options that are only
2341 meaningful for Objective-C and Objective-C++ programs, but you can also
2342 use most of the language-independent GNU compiler options. For
2343 example, you might compile a file `some_class.m' like this:
2345 gcc -g -fgnu-runtime -O -c some_class.m
2347 In this example, `-fgnu-runtime' is an option meant only for
2348 Objective-C and Objective-C++ programs; you can use the other options
2349 with any language supported by GCC.
2351 Note that since Objective-C is an extension of the C language,
2352 Objective-C compilations may also use options specific to the C
2353 front-end (e.g., `-Wtraditional'). Similarly, Objective-C++
2354 compilations may use C++-specific options (e.g., `-Wabi').
2356 Here is a list of options that are _only_ for compiling Objective-C
2357 and Objective-C++ programs:
2359 `-fconstant-string-class=CLASS-NAME'
2360 Use CLASS-NAME as the name of the class to instantiate for each
2361 literal string specified with the syntax `@"..."'. The default
2362 class name is `NXConstantString' if the GNU runtime is being used,
2363 and `NSConstantString' if the NeXT runtime is being used (see
2364 below). The `-fconstant-cfstrings' option, if also present, will
2365 override the `-fconstant-string-class' setting and cause `@"..."'
2366 literals to be laid out as constant CoreFoundation strings.
2369 Generate object code compatible with the standard GNU Objective-C
2370 runtime. This is the default for most types of systems.
2373 Generate output compatible with the NeXT runtime. This is the
2374 default for NeXT-based systems, including Darwin and Mac OS X.
2375 The macro `__NEXT_RUNTIME__' is predefined if (and only if) this
2378 `-fno-nil-receivers'
2379 Assume that all Objective-C message dispatches (e.g., `[receiver
2380 message:arg]') in this translation unit ensure that the receiver
2381 is not `nil'. This allows for more efficient entry points in the
2382 runtime to be used. Currently, this option is only available in
2383 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2385 `-fobjc-call-cxx-cdtors'
2386 For each Objective-C class, check if any of its instance variables
2387 is a C++ object with a non-trivial default constructor. If so,
2388 synthesize a special `- (id) .cxx_construct' instance method that
2389 will run non-trivial default constructors on any such instance
2390 variables, in order, and then return `self'. Similarly, check if
2391 any instance variable is a C++ object with a non-trivial
2392 destructor, and if so, synthesize a special `- (void)
2393 .cxx_destruct' method that will run all such default destructors,
2396 The `- (id) .cxx_construct' and/or `- (void) .cxx_destruct' methods
2397 thusly generated will only operate on instance variables declared
2398 in the current Objective-C class, and not those inherited from
2399 superclasses. It is the responsibility of the Objective-C runtime
2400 to invoke all such methods in an object's inheritance hierarchy.
2401 The `- (id) .cxx_construct' methods will be invoked by the runtime
2402 immediately after a new object instance is allocated; the `-
2403 (void) .cxx_destruct' methods will be invoked immediately before
2404 the runtime deallocates an object instance.
2406 As of this writing, only the NeXT runtime on Mac OS X 10.4 and
2407 later has support for invoking the `- (id) .cxx_construct' and `-
2408 (void) .cxx_destruct' methods.
2410 `-fobjc-direct-dispatch'
2411 Allow fast jumps to the message dispatcher. On Darwin this is
2412 accomplished via the comm page.
2415 Enable syntactic support for structured exception handling in
2416 Objective-C, similar to what is offered by C++ and Java. This
2417 option is unavailable in conjunction with the NeXT runtime on Mac
2418 OS X 10.2 and earlier.
2425 @catch (AnObjCClass *exc) {
2432 @catch (AnotherClass *exc) {
2435 @catch (id allOthers) {
2444 The `@throw' statement may appear anywhere in an Objective-C or
2445 Objective-C++ program; when used inside of a `@catch' block, the
2446 `@throw' may appear without an argument (as shown above), in which
2447 case the object caught by the `@catch' will be rethrown.
2449 Note that only (pointers to) Objective-C objects may be thrown and
2450 caught using this scheme. When an object is thrown, it will be
2451 caught by the nearest `@catch' clause capable of handling objects
2452 of that type, analogously to how `catch' blocks work in C++ and
2453 Java. A `@catch(id ...)' clause (as shown above) may also be
2454 provided to catch any and all Objective-C exceptions not caught by
2455 previous `@catch' clauses (if any).
2457 The `@finally' clause, if present, will be executed upon exit from
2458 the immediately preceding `@try ... @catch' section. This will
2459 happen regardless of whether any exceptions are thrown, caught or
2460 rethrown inside the `@try ... @catch' section, analogously to the
2461 behavior of the `finally' clause in Java.
2463 There are several caveats to using the new exception mechanism:
2465 * Although currently designed to be binary compatible with
2466 `NS_HANDLER'-style idioms provided by the `NSException'
2467 class, the new exceptions can only be used on Mac OS X 10.3
2468 (Panther) and later systems, due to additional functionality
2469 needed in the (NeXT) Objective-C runtime.
2471 * As mentioned above, the new exceptions do not support handling
2472 types other than Objective-C objects. Furthermore, when
2473 used from Objective-C++, the Objective-C exception model does
2474 not interoperate with C++ exceptions at this time. This
2475 means you cannot `@throw' an exception from Objective-C and
2476 `catch' it in C++, or vice versa (i.e., `throw ... @catch').
2478 The `-fobjc-exceptions' switch also enables the use of
2479 synchronization blocks for thread-safe execution:
2481 @synchronized (ObjCClass *guard) {
2485 Upon entering the `@synchronized' block, a thread of execution
2486 shall first check whether a lock has been placed on the
2487 corresponding `guard' object by another thread. If it has, the
2488 current thread shall wait until the other thread relinquishes its
2489 lock. Once `guard' becomes available, the current thread will
2490 place its own lock on it, execute the code contained in the
2491 `@synchronized' block, and finally relinquish the lock (thereby
2492 making `guard' available to other threads).
2494 Unlike Java, Objective-C does not allow for entire methods to be
2495 marked `@synchronized'. Note that throwing exceptions out of
2496 `@synchronized' blocks is allowed, and will cause the guarding
2497 object to be unlocked properly.
2500 Enable garbage collection (GC) in Objective-C and Objective-C++
2503 `-freplace-objc-classes'
2504 Emit a special marker instructing `ld(1)' not to statically link in
2505 the resulting object file, and allow `dyld(1)' to load it in at
2506 run time instead. This is used in conjunction with the
2507 Fix-and-Continue debugging mode, where the object file in question
2508 may be recompiled and dynamically reloaded in the course of
2509 program execution, without the need to restart the program itself.
2510 Currently, Fix-and-Continue functionality is only available in
2511 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2514 When compiling for the NeXT runtime, the compiler ordinarily
2515 replaces calls to `objc_getClass("...")' (when the name of the
2516 class is known at compile time) with static class references that
2517 get initialized at load time, which improves run-time performance.
2518 Specifying the `-fzero-link' flag suppresses this behavior and
2519 causes calls to `objc_getClass("...")' to be retained. This is
2520 useful in Zero-Link debugging mode, since it allows for individual
2521 class implementations to be modified during program execution.
2524 Dump interface declarations for all classes seen in the source
2525 file to a file named `SOURCENAME.decl'.
2527 `-Wassign-intercept (Objective-C and Objective-C++ only)'
2528 Warn whenever an Objective-C assignment is being intercepted by the
2531 `-Wno-protocol (Objective-C and Objective-C++ only)'
2532 If a class is declared to implement a protocol, a warning is
2533 issued for every method in the protocol that is not implemented by
2534 the class. The default behavior is to issue a warning for every
2535 method not explicitly implemented in the class, even if a method
2536 implementation is inherited from the superclass. If you use the
2537 `-Wno-protocol' option, then methods inherited from the superclass
2538 are considered to be implemented, and no warning is issued for
2541 `-Wselector (Objective-C and Objective-C++ only)'
2542 Warn if multiple methods of different types for the same selector
2543 are found during compilation. The check is performed on the list
2544 of methods in the final stage of compilation. Additionally, a
2545 check is performed for each selector appearing in a
2546 `@selector(...)' expression, and a corresponding method for that
2547 selector has been found during compilation. Because these checks
2548 scan the method table only at the end of compilation, these
2549 warnings are not produced if the final stage of compilation is not
2550 reached, for example because an error is found during compilation,
2551 or because the `-fsyntax-only' option is being used.
2553 `-Wstrict-selector-match (Objective-C and Objective-C++ only)'
2554 Warn if multiple methods with differing argument and/or return
2555 types are found for a given selector when attempting to send a
2556 message using this selector to a receiver of type `id' or `Class'.
2557 When this flag is off (which is the default behavior), the
2558 compiler will omit such warnings if any differences found are
2559 confined to types which share the same size and alignment.
2561 `-Wundeclared-selector (Objective-C and Objective-C++ only)'
2562 Warn if a `@selector(...)' expression referring to an undeclared
2563 selector is found. A selector is considered undeclared if no
2564 method with that name has been declared before the
2565 `@selector(...)' expression, either explicitly in an `@interface'
2566 or `@protocol' declaration, or implicitly in an `@implementation'
2567 section. This option always performs its checks as soon as a
2568 `@selector(...)' expression is found, while `-Wselector' only
2569 performs its checks in the final stage of compilation. This also
2570 enforces the coding style convention that methods and selectors
2571 must be declared before being used.
2573 `-print-objc-runtime-info'
2574 Generate C header describing the largest structure that is passed
2579 File: gcc.info, Node: Language Independent Options, Next: Warning Options, Prev: Objective-C and Objective-C++ Dialect Options, Up: Invoking GCC
2581 3.7 Options to Control Diagnostic Messages Formatting
2582 =====================================================
2584 Traditionally, diagnostic messages have been formatted irrespective of
2585 the output device's aspect (e.g. its width, ...). The options described
2586 below can be used to control the diagnostic messages formatting
2587 algorithm, e.g. how many characters per line, how often source location
2588 information should be reported. Right now, only the C++ front end can
2589 honor these options. However it is expected, in the near future, that
2590 the remaining front ends would be able to digest them correctly.
2592 `-fmessage-length=N'
2593 Try to format error messages so that they fit on lines of about N
2594 characters. The default is 72 characters for `g++' and 0 for the
2595 rest of the front ends supported by GCC. If N is zero, then no
2596 line-wrapping will be done; each error message will appear on a
2599 `-fdiagnostics-show-location=once'
2600 Only meaningful in line-wrapping mode. Instructs the diagnostic
2601 messages reporter to emit _once_ source location information; that
2602 is, in case the message is too long to fit on a single physical
2603 line and has to be wrapped, the source location won't be emitted
2604 (as prefix) again, over and over, in subsequent continuation
2605 lines. This is the default behavior.
2607 `-fdiagnostics-show-location=every-line'
2608 Only meaningful in line-wrapping mode. Instructs the diagnostic
2609 messages reporter to emit the same source location information (as
2610 prefix) for physical lines that result from the process of breaking
2611 a message which is too long to fit on a single line.
2613 `-fdiagnostics-show-option'
2614 This option instructs the diagnostic machinery to add text to each
2615 diagnostic emitted, which indicates which command line option
2616 directly controls that diagnostic, when such an option is known to
2617 the diagnostic machinery.
2619 `-Wcoverage-mismatch'
2620 Warn if feedback profiles do not match when using the
2621 `-fprofile-use' option. If a source file was changed between
2622 `-fprofile-gen' and `-fprofile-use', the files with the profile
2623 feedback can fail to match the source file and GCC can not use the
2624 profile feedback information. By default, GCC emits an error
2625 message in this case. The option `-Wcoverage-mismatch' emits a
2626 warning instead of an error. GCC does not use appropriate
2627 feedback profiles, so using this option can result in poorly
2628 optimized code. This option is useful only in the case of very
2629 minor changes such as bug fixes to an existing code-base.
2633 File: gcc.info, Node: Warning Options, Next: Debugging Options, Prev: Language Independent Options, Up: Invoking GCC
2635 3.8 Options to Request or Suppress Warnings
2636 ===========================================
2638 Warnings are diagnostic messages that report constructions which are
2639 not inherently erroneous but which are risky or suggest there may have
2642 The following language-independent options do not enable specific
2643 warnings but control the kinds of diagnostics produced by GCC.
2646 Check the code for syntax errors, but don't do anything beyond
2650 Inhibit all warning messages.
2653 Make all warnings into errors.
2656 Make the specified warning into an error. The specifier for a
2657 warning is appended, for example `-Werror=switch' turns the
2658 warnings controlled by `-Wswitch' into errors. This switch takes a
2659 negative form, to be used to negate `-Werror' for specific
2660 warnings, for example `-Wno-error=switch' makes `-Wswitch'
2661 warnings not be errors, even when `-Werror' is in effect. You can
2662 use the `-fdiagnostics-show-option' option to have each
2663 controllable warning amended with the option which controls it, to
2664 determine what to use with this option.
2666 Note that specifying `-Werror='FOO automatically implies `-W'FOO.
2667 However, `-Wno-error='FOO does not imply anything.
2670 This option causes the compiler to abort compilation on the first
2671 error occurred rather than trying to keep going and printing
2672 further error messages.
2675 You can request many specific warnings with options beginning `-W',
2676 for example `-Wimplicit' to request warnings on implicit declarations.
2677 Each of these specific warning options also has a negative form
2678 beginning `-Wno-' to turn off warnings; for example, `-Wno-implicit'.
2679 This manual lists only one of the two forms, whichever is not the
2680 default. For further, language-specific options also refer to *note
2681 C++ Dialect Options:: and *note Objective-C and Objective-C++ Dialect
2685 Issue all the warnings demanded by strict ISO C and ISO C++;
2686 reject all programs that use forbidden extensions, and some other
2687 programs that do not follow ISO C and ISO C++. For ISO C, follows
2688 the version of the ISO C standard specified by any `-std' option
2691 Valid ISO C and ISO C++ programs should compile properly with or
2692 without this option (though a rare few will require `-ansi' or a
2693 `-std' option specifying the required version of ISO C). However,
2694 without this option, certain GNU extensions and traditional C and
2695 C++ features are supported as well. With this option, they are
2698 `-pedantic' does not cause warning messages for use of the
2699 alternate keywords whose names begin and end with `__'. Pedantic
2700 warnings are also disabled in the expression that follows
2701 `__extension__'. However, only system header files should use
2702 these escape routes; application programs should avoid them.
2703 *Note Alternate Keywords::.
2705 Some users try to use `-pedantic' to check programs for strict ISO
2706 C conformance. They soon find that it does not do quite what they
2707 want: it finds some non-ISO practices, but not all--only those for
2708 which ISO C _requires_ a diagnostic, and some others for which
2709 diagnostics have been added.
2711 A feature to report any failure to conform to ISO C might be
2712 useful in some instances, but would require considerable
2713 additional work and would be quite different from `-pedantic'. We
2714 don't have plans to support such a feature in the near future.
2716 Where the standard specified with `-std' represents a GNU extended
2717 dialect of C, such as `gnu89' or `gnu99', there is a corresponding
2718 "base standard", the version of ISO C on which the GNU extended
2719 dialect is based. Warnings from `-pedantic' are given where they
2720 are required by the base standard. (It would not make sense for
2721 such warnings to be given only for features not in the specified
2722 GNU C dialect, since by definition the GNU dialects of C include
2723 all features the compiler supports with the given option, and
2724 there would be nothing to warn about.)
2727 Like `-pedantic', except that errors are produced rather than
2731 This enables all the warnings about constructions that some users
2732 consider questionable, and that are easy to avoid (or modify to
2733 prevent the warning), even in conjunction with macros. This also
2734 enables some language-specific warnings described in *note C++
2735 Dialect Options:: and *note Objective-C and Objective-C++ Dialect
2738 `-Wall' turns on the following warning flags:
2741 -Warray-bounds (only with `-O2')
2745 -Wimplicit-function-declaration
2748 -Wmain (only for C/ObjC and unless `-ffreestanding')
2756 -Wsign-compare (only in C++)
2767 -Wvolatile-register-var
2769 Note that some warning flags are not implied by `-Wall'. Some of
2770 them warn about constructions that users generally do not consider
2771 questionable, but which occasionally you might wish to check for;
2772 others warn about constructions that are necessary or hard to
2773 avoid in some cases, and there is no simple way to modify the code
2774 to suppress the warning. Some of them are enabled by `-Wextra' but
2775 many of them must be enabled individually.
2778 This enables some extra warning flags that are not enabled by
2779 `-Wall'. (This option used to be called `-W'. The older name is
2780 still supported, but the newer name is more descriptive.)
2784 -Wignored-qualifiers
2785 -Wmissing-field-initializers
2786 -Wmissing-parameter-type (C only)
2787 -Wold-style-declaration (C only)
2792 -Wunused-parameter (only with `-Wunused' or `-Wall')
2794 The option `-Wextra' also prints warning messages for the
2797 * A pointer is compared against integer zero with `<', `<=',
2800 * (C++ only) An enumerator and a non-enumerator both appear in a
2801 conditional expression.
2803 * (C++ only) Ambiguous virtual bases.
2805 * (C++ only) Subscripting an array which has been declared
2808 * (C++ only) Taking the address of a variable which has been
2809 declared `register'.
2811 * (C++ only) A base class is not initialized in a derived
2812 class' copy constructor.
2816 Warn if an array subscript has type `char'. This is a common cause
2817 of error, as programmers often forget that this type is signed on
2818 some machines. This warning is enabled by `-Wall'.
2821 Warn whenever a comment-start sequence `/*' appears in a `/*'
2822 comment, or whenever a Backslash-Newline appears in a `//' comment.
2823 This warning is enabled by `-Wall'.
2826 Check calls to `printf' and `scanf', etc., to make sure that the
2827 arguments supplied have types appropriate to the format string
2828 specified, and that the conversions specified in the format string
2829 make sense. This includes standard functions, and others
2830 specified by format attributes (*note Function Attributes::), in
2831 the `printf', `scanf', `strftime' and `strfmon' (an X/Open
2832 extension, not in the C standard) families (or other
2833 target-specific families). Which functions are checked without
2834 format attributes having been specified depends on the standard
2835 version selected, and such checks of functions without the
2836 attribute specified are disabled by `-ffreestanding' or
2839 The formats are checked against the format features supported by
2840 GNU libc version 2.2. These include all ISO C90 and C99 features,
2841 as well as features from the Single Unix Specification and some
2842 BSD and GNU extensions. Other library implementations may not
2843 support all these features; GCC does not support warning about
2844 features that go beyond a particular library's limitations.
2845 However, if `-pedantic' is used with `-Wformat', warnings will be
2846 given about format features not in the selected standard version
2847 (but not for `strfmon' formats, since those are not in any version
2848 of the C standard). *Note Options Controlling C Dialect: C
2851 Since `-Wformat' also checks for null format arguments for several
2852 functions, `-Wformat' also implies `-Wnonnull'.
2854 `-Wformat' is included in `-Wall'. For more control over some
2855 aspects of format checking, the options `-Wformat-y2k',
2856 `-Wno-format-extra-args', `-Wno-format-zero-length',
2857 `-Wformat-nonliteral', `-Wformat-security', and `-Wformat=2' are
2858 available, but are not included in `-Wall'.
2861 If `-Wformat' is specified, also warn about `strftime' formats
2862 which may yield only a two-digit year.
2864 `-Wno-format-contains-nul'
2865 If `-Wformat' is specified, do not warn about format strings that
2868 `-Wno-format-extra-args'
2869 If `-Wformat' is specified, do not warn about excess arguments to a
2870 `printf' or `scanf' format function. The C standard specifies
2871 that such arguments are ignored.
2873 Where the unused arguments lie between used arguments that are
2874 specified with `$' operand number specifications, normally
2875 warnings are still given, since the implementation could not know
2876 what type to pass to `va_arg' to skip the unused arguments.
2877 However, in the case of `scanf' formats, this option will suppress
2878 the warning if the unused arguments are all pointers, since the
2879 Single Unix Specification says that such unused arguments are
2882 `-Wno-format-zero-length (C and Objective-C only)'
2883 If `-Wformat' is specified, do not warn about zero-length formats.
2884 The C standard specifies that zero-length formats are allowed.
2886 `-Wformat-nonliteral'
2887 If `-Wformat' is specified, also warn if the format string is not a
2888 string literal and so cannot be checked, unless the format function
2889 takes its format arguments as a `va_list'.
2892 If `-Wformat' is specified, also warn about uses of format
2893 functions that represent possible security problems. At present,
2894 this warns about calls to `printf' and `scanf' functions where the
2895 format string is not a string literal and there are no format
2896 arguments, as in `printf (foo);'. This may be a security hole if
2897 the format string came from untrusted input and contains `%n'.
2898 (This is currently a subset of what `-Wformat-nonliteral' warns
2899 about, but in future warnings may be added to `-Wformat-security'
2900 that are not included in `-Wformat-nonliteral'.)
2903 Enable `-Wformat' plus format checks not included in `-Wformat'.
2904 Currently equivalent to `-Wformat -Wformat-nonliteral
2905 -Wformat-security -Wformat-y2k'.
2907 `-Wnonnull (C and Objective-C only)'
2908 Warn about passing a null pointer for arguments marked as
2909 requiring a non-null value by the `nonnull' function attribute.
2911 `-Wnonnull' is included in `-Wall' and `-Wformat'. It can be
2912 disabled with the `-Wno-nonnull' option.
2914 `-Winit-self (C, C++, Objective-C and Objective-C++ only)'
2915 Warn about uninitialized variables which are initialized with
2916 themselves. Note this option can only be used with the
2917 `-Wuninitialized' option.
2919 For example, GCC will warn about `i' being uninitialized in the
2920 following snippet only when `-Winit-self' has been specified:
2927 `-Wimplicit-int (C and Objective-C only)'
2928 Warn when a declaration does not specify a type. This warning is
2931 `-Wimplicit-function-declaration (C and Objective-C only)'
2932 Give a warning whenever a function is used before being declared.
2933 In C99 mode (`-std=c99' or `-std=gnu99'), this warning is enabled
2934 by default and it is made into an error by `-pedantic-errors'.
2935 This warning is also enabled by `-Wall'.
2938 Same as `-Wimplicit-int' and `-Wimplicit-function-declaration'.
2939 This warning is enabled by `-Wall'.
2941 `-Wignored-qualifiers (C and C++ only)'
2942 Warn if the return type of a function has a type qualifier such as
2943 `const'. For ISO C such a type qualifier has no effect, since the
2944 value returned by a function is not an lvalue. For C++, the
2945 warning is only emitted for scalar types or `void'. ISO C
2946 prohibits qualified `void' return types on function definitions,
2947 so such return types always receive a warning even without this
2950 This warning is also enabled by `-Wextra'.
2953 Warn if the type of `main' is suspicious. `main' should be a
2954 function with external linkage, returning int, taking either zero
2955 arguments, two, or three arguments of appropriate types. This
2956 warning is enabled by default in C++ and is enabled by either
2957 `-Wall' or `-pedantic'.
2960 Warn if an aggregate or union initializer is not fully bracketed.
2961 In the following example, the initializer for `a' is not fully
2962 bracketed, but that for `b' is fully bracketed.
2964 int a[2][2] = { 0, 1, 2, 3 };
2965 int b[2][2] = { { 0, 1 }, { 2, 3 } };
2967 This warning is enabled by `-Wall'.
2969 `-Wmissing-include-dirs (C, C++, Objective-C and Objective-C++ only)'
2970 Warn if a user-supplied include directory does not exist.
2973 Warn if parentheses are omitted in certain contexts, such as when
2974 there is an assignment in a context where a truth value is
2975 expected, or when operators are nested whose precedence people
2976 often get confused about.
2978 Also warn if a comparison like `x<=y<=z' appears; this is
2979 equivalent to `(x<=y ? 1 : 0) <= z', which is a different
2980 interpretation from that of ordinary mathematical notation.
2982 Also warn about constructions where there may be confusion to which
2983 `if' statement an `else' branch belongs. Here is an example of
2994 In C/C++, every `else' branch belongs to the innermost possible
2995 `if' statement, which in this example is `if (b)'. This is often
2996 not what the programmer expected, as illustrated in the above
2997 example by indentation the programmer chose. When there is the
2998 potential for this confusion, GCC will issue a warning when this
2999 flag is specified. To eliminate the warning, add explicit braces
3000 around the innermost `if' statement so there is no way the `else'
3001 could belong to the enclosing `if'. The resulting code would look
3014 This warning is enabled by `-Wall'.
3017 Warn about code that may have undefined semantics because of
3018 violations of sequence point rules in the C and C++ standards.
3020 The C and C++ standards defines the order in which expressions in
3021 a C/C++ program are evaluated in terms of "sequence points", which
3022 represent a partial ordering between the execution of parts of the
3023 program: those executed before the sequence point, and those
3024 executed after it. These occur after the evaluation of a full
3025 expression (one which is not part of a larger expression), after
3026 the evaluation of the first operand of a `&&', `||', `? :' or `,'
3027 (comma) operator, before a function is called (but after the
3028 evaluation of its arguments and the expression denoting the called
3029 function), and in certain other places. Other than as expressed
3030 by the sequence point rules, the order of evaluation of
3031 subexpressions of an expression is not specified. All these rules
3032 describe only a partial order rather than a total order, since,
3033 for example, if two functions are called within one expression
3034 with no sequence point between them, the order in which the
3035 functions are called is not specified. However, the standards
3036 committee have ruled that function calls do not overlap.
3038 It is not specified when between sequence points modifications to
3039 the values of objects take effect. Programs whose behavior
3040 depends on this have undefined behavior; the C and C++ standards
3041 specify that "Between the previous and next sequence point an
3042 object shall have its stored value modified at most once by the
3043 evaluation of an expression. Furthermore, the prior value shall
3044 be read only to determine the value to be stored.". If a program
3045 breaks these rules, the results on any particular implementation
3046 are entirely unpredictable.
3048 Examples of code with undefined behavior are `a = a++;', `a[n] =
3049 b[n++]' and `a[i++] = i;'. Some more complicated cases are not
3050 diagnosed by this option, and it may give an occasional false
3051 positive result, but in general it has been found fairly effective
3052 at detecting this sort of problem in programs.
3054 The standard is worded confusingly, therefore there is some debate
3055 over the precise meaning of the sequence point rules in subtle
3056 cases. Links to discussions of the problem, including proposed
3057 formal definitions, may be found on the GCC readings page, at
3058 `http://gcc.gnu.org/readings.html'.
3060 This warning is enabled by `-Wall' for C and C++.
3063 Warn whenever a function is defined with a return-type that
3064 defaults to `int'. Also warn about any `return' statement with no
3065 return-value in a function whose return-type is not `void'
3066 (falling off the end of the function body is considered returning
3067 without a value), and about a `return' statement with a expression
3068 in a function whose return-type is `void'.
3070 For C++, a function without return type always produces a
3071 diagnostic message, even when `-Wno-return-type' is specified.
3072 The only exceptions are `main' and functions defined in system
3075 This warning is enabled by `-Wall'.
3078 Warn whenever a `switch' statement has an index of enumerated type
3079 and lacks a `case' for one or more of the named codes of that
3080 enumeration. (The presence of a `default' label prevents this
3081 warning.) `case' labels outside the enumeration range also
3082 provoke warnings when this option is used. This warning is
3086 Warn whenever a `switch' statement does not have a `default' case.
3089 Warn whenever a `switch' statement has an index of enumerated type
3090 and lacks a `case' for one or more of the named codes of that
3091 enumeration. `case' labels outside the enumeration range also
3092 provoke warnings when this option is used.
3094 `-Wsync-nand (C and C++ only)'
3095 Warn when `__sync_fetch_and_nand' and `__sync_nand_and_fetch'
3096 built-in functions are used. These functions changed semantics in
3100 Warn if any trigraphs are encountered that might change the
3101 meaning of the program (trigraphs within comments are not warned
3102 about). This warning is enabled by `-Wall'.
3105 Warn whenever a static function is declared but not defined or a
3106 non-inline static function is unused. This warning is enabled by
3110 Warn whenever a label is declared but not used. This warning is
3113 To suppress this warning use the `unused' attribute (*note
3114 Variable Attributes::).
3116 `-Wunused-parameter'
3117 Warn whenever a function parameter is unused aside from its
3120 To suppress this warning use the `unused' attribute (*note
3121 Variable Attributes::).
3124 Warn whenever a local variable or non-constant static variable is
3125 unused aside from its declaration. This warning is enabled by
3128 To suppress this warning use the `unused' attribute (*note
3129 Variable Attributes::).
3132 Warn whenever a statement computes a result that is explicitly not
3133 used. To suppress this warning cast the unused expression to
3134 `void'. This includes an expression-statement or the left-hand
3135 side of a comma expression that contains no side effects. For
3136 example, an expression such as `x[i,j]' will cause a warning, while
3137 `x[(void)i,j]' will not.
3139 This warning is enabled by `-Wall'.
3142 All the above `-Wunused' options combined.
3144 In order to get a warning about an unused function parameter, you
3145 must either specify `-Wextra -Wunused' (note that `-Wall' implies
3146 `-Wunused'), or separately specify `-Wunused-parameter'.
3149 Warn if an automatic variable is used without first being
3150 initialized or if a variable may be clobbered by a `setjmp' call.
3151 In C++, warn if a non-static reference or non-static `const' member
3152 appears in a class without constructors.
3154 If you want to warn about code which uses the uninitialized value
3155 of the variable in its own initializer, use the `-Winit-self'
3158 These warnings occur for individual uninitialized or clobbered
3159 elements of structure, union or array variables as well as for
3160 variables which are uninitialized or clobbered as a whole. They do
3161 not occur for variables or elements declared `volatile'. Because
3162 these warnings depend on optimization, the exact variables or
3163 elements for which there are warnings will depend on the precise
3164 optimization options and version of GCC used.
3166 Note that there may be no warning about a variable that is used
3167 only to compute a value that itself is never used, because such
3168 computations may be deleted by data flow analysis before the
3169 warnings are printed.
3171 These warnings are made optional because GCC is not smart enough
3172 to see all the reasons why the code might be correct despite
3173 appearing to have an error. Here is one example of how this can
3189 If the value of `y' is always 1, 2 or 3, then `x' is always
3190 initialized, but GCC doesn't know this. Here is another common
3195 if (change_y) save_y = y, y = new_y;
3197 if (change_y) y = save_y;
3200 This has no bug because `save_y' is used only if it is set.
3202 This option also warns when a non-volatile automatic variable
3203 might be changed by a call to `longjmp'. These warnings as well
3204 are possible only in optimizing compilation.
3206 The compiler sees only the calls to `setjmp'. It cannot know
3207 where `longjmp' will be called; in fact, a signal handler could
3208 call it at any point in the code. As a result, you may get a
3209 warning even when there is in fact no problem because `longjmp'
3210 cannot in fact be called at the place which would cause a problem.
3212 Some spurious warnings can be avoided if you declare all the
3213 functions you use that never return as `noreturn'. *Note Function
3216 This warning is enabled by `-Wall' or `-Wextra'.
3219 Warn when a #pragma directive is encountered which is not
3220 understood by GCC. If this command line option is used, warnings
3221 will even be issued for unknown pragmas in system header files.
3222 This is not the case if the warnings were only enabled by the
3223 `-Wall' command line option.
3226 Do not warn about misuses of pragmas, such as incorrect parameters,
3227 invalid syntax, or conflicts between pragmas. See also
3228 `-Wunknown-pragmas'.
3231 This option is only active when `-fstrict-aliasing' is active. It
3232 warns about code which might break the strict aliasing rules that
3233 the compiler is using for optimization. The warning does not
3234 catch all cases, but does attempt to catch the more common
3235 pitfalls. It is included in `-Wall'. It is equivalent to
3236 `-Wstrict-aliasing=3'
3238 `-Wstrict-aliasing=n'
3239 This option is only active when `-fstrict-aliasing' is active. It
3240 warns about code which might break the strict aliasing rules that
3241 the compiler is using for optimization. Higher levels correspond
3242 to higher accuracy (fewer false positives). Higher levels also
3243 correspond to more effort, similar to the way -O works.
3244 `-Wstrict-aliasing' is equivalent to `-Wstrict-aliasing=n', with
3247 Level 1: Most aggressive, quick, least accurate. Possibly useful
3248 when higher levels do not warn but -fstrict-aliasing still breaks
3249 the code, as it has very few false negatives. However, it has
3250 many false positives. Warns for all pointer conversions between
3251 possibly incompatible types, even if never dereferenced. Runs in
3254 Level 2: Aggressive, quick, not too precise. May still have many
3255 false positives (not as many as level 1 though), and few false
3256 negatives (but possibly more than level 1). Unlike level 1, it
3257 only warns when an address is taken. Warns about incomplete
3258 types. Runs in the frontend only.
3260 Level 3 (default for `-Wstrict-aliasing'): Should have very few
3261 false positives and few false negatives. Slightly slower than
3262 levels 1 or 2 when optimization is enabled. Takes care of the
3263 common punn+dereference pattern in the frontend:
3264 `*(int*)&some_float'. If optimization is enabled, it also runs in
3265 the backend, where it deals with multiple statement cases using
3266 flow-sensitive points-to information. Only warns when the
3267 converted pointer is dereferenced. Does not warn about incomplete
3271 `-Wstrict-overflow=N'
3272 This option is only active when `-fstrict-overflow' is active. It
3273 warns about cases where the compiler optimizes based on the
3274 assumption that signed overflow does not occur. Note that it does
3275 not warn about all cases where the code might overflow: it only
3276 warns about cases where the compiler implements some optimization.
3277 Thus this warning depends on the optimization level.
3279 An optimization which assumes that signed overflow does not occur
3280 is perfectly safe if the values of the variables involved are such
3281 that overflow never does, in fact, occur. Therefore this warning
3282 can easily give a false positive: a warning about code which is not
3283 actually a problem. To help focus on important issues, several
3284 warning levels are defined. No warnings are issued for the use of
3285 undefined signed overflow when estimating how many iterations a
3286 loop will require, in particular when determining whether a loop
3287 will be executed at all.
3289 `-Wstrict-overflow=1'
3290 Warn about cases which are both questionable and easy to
3291 avoid. For example: `x + 1 > x'; with `-fstrict-overflow',
3292 the compiler will simplify this to `1'. This level of
3293 `-Wstrict-overflow' is enabled by `-Wall'; higher levels are
3294 not, and must be explicitly requested.
3296 `-Wstrict-overflow=2'
3297 Also warn about other cases where a comparison is simplified
3298 to a constant. For example: `abs (x) >= 0'. This can only be
3299 simplified when `-fstrict-overflow' is in effect, because
3300 `abs (INT_MIN)' overflows to `INT_MIN', which is less than
3301 zero. `-Wstrict-overflow' (with no level) is the same as
3302 `-Wstrict-overflow=2'.
3304 `-Wstrict-overflow=3'
3305 Also warn about other cases where a comparison is simplified.
3306 For example: `x + 1 > 1' will be simplified to `x > 0'.
3308 `-Wstrict-overflow=4'
3309 Also warn about other simplifications not covered by the
3310 above cases. For example: `(x * 10) / 5' will be simplified
3313 `-Wstrict-overflow=5'
3314 Also warn about cases where the compiler reduces the
3315 magnitude of a constant involved in a comparison. For
3316 example: `x + 2 > y' will be simplified to `x + 1 >= y'.
3317 This is reported only at the highest warning level because
3318 this simplification applies to many comparisons, so this
3319 warning level will give a very large number of false
3323 This option is only active when `-ftree-vrp' is active (default
3324 for -O2 and above). It warns about subscripts to arrays that are
3325 always out of bounds. This warning is enabled by `-Wall'.
3328 Do not warn about compile-time integer division by zero. Floating
3329 point division by zero is not warned about, as it can be a
3330 legitimate way of obtaining infinities and NaNs.
3333 Print warning messages for constructs found in system header files.
3334 Warnings from system headers are normally suppressed, on the
3335 assumption that they usually do not indicate real problems and
3336 would only make the compiler output harder to read. Using this
3337 command line option tells GCC to emit warnings from system headers
3338 as if they occurred in user code. However, note that using
3339 `-Wall' in conjunction with this option will _not_ warn about
3340 unknown pragmas in system headers--for that, `-Wunknown-pragmas'
3344 Warn if floating point values are used in equality comparisons.
3346 The idea behind this is that sometimes it is convenient (for the
3347 programmer) to consider floating-point values as approximations to
3348 infinitely precise real numbers. If you are doing this, then you
3349 need to compute (by analyzing the code, or in some other way) the
3350 maximum or likely maximum error that the computation introduces,
3351 and allow for it when performing comparisons (and when producing
3352 output, but that's a different problem). In particular, instead
3353 of testing for equality, you would check to see whether the two
3354 values have ranges that overlap; and this is done with the
3355 relational operators, so equality comparisons are probably
3358 `-Wtraditional (C and Objective-C only)'
3359 Warn about certain constructs that behave differently in
3360 traditional and ISO C. Also warn about ISO C constructs that have
3361 no traditional C equivalent, and/or problematic constructs which
3364 * Macro parameters that appear within string literals in the
3365 macro body. In traditional C macro replacement takes place
3366 within string literals, but does not in ISO C.
3368 * In traditional C, some preprocessor directives did not exist.
3369 Traditional preprocessors would only consider a line to be a
3370 directive if the `#' appeared in column 1 on the line.
3371 Therefore `-Wtraditional' warns about directives that
3372 traditional C understands but would ignore because the `#'
3373 does not appear as the first character on the line. It also
3374 suggests you hide directives like `#pragma' not understood by
3375 traditional C by indenting them. Some traditional
3376 implementations would not recognize `#elif', so it suggests
3377 avoiding it altogether.
3379 * A function-like macro that appears without arguments.
3381 * The unary plus operator.
3383 * The `U' integer constant suffix, or the `F' or `L' floating
3384 point constant suffixes. (Traditional C does support the `L'
3385 suffix on integer constants.) Note, these suffixes appear in
3386 macros defined in the system headers of most modern systems,
3387 e.g. the `_MIN'/`_MAX' macros in `<limits.h>'. Use of these
3388 macros in user code might normally lead to spurious warnings,
3389 however GCC's integrated preprocessor has enough context to
3390 avoid warning in these cases.
3392 * A function declared external in one block and then used after
3393 the end of the block.
3395 * A `switch' statement has an operand of type `long'.
3397 * A non-`static' function declaration follows a `static' one.
3398 This construct is not accepted by some traditional C
3401 * The ISO type of an integer constant has a different width or
3402 signedness from its traditional type. This warning is only
3403 issued if the base of the constant is ten. I.e. hexadecimal
3404 or octal values, which typically represent bit patterns, are
3407 * Usage of ISO string concatenation is detected.
3409 * Initialization of automatic aggregates.
3411 * Identifier conflicts with labels. Traditional C lacks a
3412 separate namespace for labels.
3414 * Initialization of unions. If the initializer is zero, the
3415 warning is omitted. This is done under the assumption that
3416 the zero initializer in user code appears conditioned on e.g.
3417 `__STDC__' to avoid missing initializer warnings and relies
3418 on default initialization to zero in the traditional C case.
3420 * Conversions by prototypes between fixed/floating point values
3421 and vice versa. The absence of these prototypes when
3422 compiling with traditional C would cause serious problems.
3423 This is a subset of the possible conversion warnings, for the
3424 full set use `-Wtraditional-conversion'.
3426 * Use of ISO C style function definitions. This warning
3427 intentionally is _not_ issued for prototype declarations or
3428 variadic functions because these ISO C features will appear
3429 in your code when using libiberty's traditional C
3430 compatibility macros, `PARAMS' and `VPARAMS'. This warning
3431 is also bypassed for nested functions because that feature is
3432 already a GCC extension and thus not relevant to traditional
3435 `-Wtraditional-conversion (C and Objective-C only)'
3436 Warn if a prototype causes a type conversion that is different
3437 from what would happen to the same argument in the absence of a
3438 prototype. This includes conversions of fixed point to floating
3439 and vice versa, and conversions changing the width or signedness
3440 of a fixed point argument except when the same as the default
3443 `-Wdeclaration-after-statement (C and Objective-C only)'
3444 Warn when a declaration is found after a statement in a block.
3445 This construct, known from C++, was introduced with ISO C99 and is
3446 by default allowed in GCC. It is not supported by ISO C90 and was
3447 not supported by GCC versions before GCC 3.0. *Note Mixed
3451 Warn if an undefined identifier is evaluated in an `#if' directive.
3454 Do not warn whenever an `#else' or an `#endif' are followed by
3458 Warn whenever a local variable shadows another local variable,
3459 parameter or global variable or whenever a built-in function is
3463 Warn whenever an object of larger than LEN bytes is defined.
3465 `-Wframe-larger-than=LEN'
3466 Warn if the size of a function frame is larger than LEN bytes.
3467 The computation done to determine the stack frame size is
3468 approximate and not conservative. The actual requirements may be
3469 somewhat greater than LEN even if you do not get a warning. In
3470 addition, any space allocated via `alloca', variable-length
3471 arrays, or related constructs is not included by the compiler when
3472 determining whether or not to issue a warning.
3474 `-Wunsafe-loop-optimizations'
3475 Warn if the loop cannot be optimized because the compiler could not
3476 assume anything on the bounds of the loop indices. With
3477 `-funsafe-loop-optimizations' warn if the compiler made such
3480 `-Wno-pedantic-ms-format (MinGW targets only)'
3481 Disables the warnings about non-ISO `printf' / `scanf' format
3482 width specifiers `I32', `I64', and `I' used on Windows targets
3483 depending on the MS runtime, when you are using the options
3484 `-Wformat' and `-pedantic' without gnu-extensions.
3487 Warn about anything that depends on the "size of" a function type
3488 or of `void'. GNU C assigns these types a size of 1, for
3489 convenience in calculations with `void *' pointers and pointers to
3490 functions. In C++, warn also when an arithmetic operation involves
3491 `NULL'. This warning is also enabled by `-pedantic'.
3494 Warn if a comparison is always true or always false due to the
3495 limited range of the data type, but do not warn for constant
3496 expressions. For example, warn if an unsigned variable is
3497 compared against zero with `<' or `>='. This warning is also
3498 enabled by `-Wextra'.
3500 `-Wbad-function-cast (C and Objective-C only)'
3501 Warn whenever a function call is cast to a non-matching type. For
3502 example, warn if `int malloc()' is cast to `anything *'.
3504 `-Wc++-compat (C and Objective-C only)'
3505 Warn about ISO C constructs that are outside of the common subset
3506 of ISO C and ISO C++, e.g. request for implicit conversion from
3507 `void *' to a pointer to non-`void' type.
3509 `-Wc++0x-compat (C++ and Objective-C++ only)'
3510 Warn about C++ constructs whose meaning differs between ISO C++
3511 1998 and ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will
3512 become keywords in ISO C++ 200x. This warning is enabled by
3516 Warn whenever a pointer is cast so as to remove a type qualifier
3517 from the target type. For example, warn if a `const char *' is
3518 cast to an ordinary `char *'.
3521 Warn whenever a pointer is cast such that the required alignment
3522 of the target is increased. For example, warn if a `char *' is
3523 cast to an `int *' on machines where integers can only be accessed
3524 at two- or four-byte boundaries.
3527 When compiling C, give string constants the type `const
3528 char[LENGTH]' so that copying the address of one into a
3529 non-`const' `char *' pointer will get a warning. These warnings
3530 will help you find at compile time code that can try to write into
3531 a string constant, but only if you have been very careful about
3532 using `const' in declarations and prototypes. Otherwise, it will
3533 just be a nuisance. This is why we did not make `-Wall' request
3536 When compiling C++, warn about the deprecated conversion from
3537 string literals to `char *'. This warning is enabled by default
3541 Warn for variables that might be changed by `longjmp' or `vfork'.
3542 This warning is also enabled by `-Wextra'.
3545 Warn for implicit conversions that may alter a value. This includes
3546 conversions between real and integer, like `abs (x)' when `x' is
3547 `double'; conversions between signed and unsigned, like `unsigned
3548 ui = -1'; and conversions to smaller types, like `sqrtf (M_PI)'.
3549 Do not warn for explicit casts like `abs ((int) x)' and `ui =
3550 (unsigned) -1', or if the value is not changed by the conversion
3551 like in `abs (2.0)'. Warnings about conversions between signed
3552 and unsigned integers can be disabled by using
3553 `-Wno-sign-conversion'.
3555 For C++, also warn for conversions between `NULL' and non-pointer
3556 types; confusing overload resolution for user-defined conversions;
3557 and conversions that will never use a type conversion operator:
3558 conversions to `void', the same type, a base class or a reference
3559 to them. Warnings about conversions between signed and unsigned
3560 integers are disabled by default in C++ unless `-Wsign-conversion'
3561 is explicitly enabled.
3564 Warn if an empty body occurs in an `if', `else' or `do while'
3565 statement. This warning is also enabled by `-Wextra'.
3567 `-Wenum-compare (C++ and Objective-C++ only)'
3568 Warn about a comparison between values of different enum types.
3569 This warning is enabled by default.
3572 Warn when a comparison between signed and unsigned values could
3573 produce an incorrect result when the signed value is converted to
3574 unsigned. This warning is also enabled by `-Wextra'; to get the
3575 other warnings of `-Wextra' without this warning, use `-Wextra
3579 Warn for implicit conversions that may change the sign of an
3580 integer value, like assigning a signed integer expression to an
3581 unsigned integer variable. An explicit cast silences the warning.
3582 In C, this option is enabled also by `-Wconversion'.
3585 Warn about suspicious uses of memory addresses. These include using
3586 the address of a function in a conditional expression, such as
3587 `void func(void); if (func)', and comparisons against the memory
3588 address of a string literal, such as `if (x == "abc")'. Such uses
3589 typically indicate a programmer error: the address of a function
3590 always evaluates to true, so their use in a conditional usually
3591 indicate that the programmer forgot the parentheses in a function
3592 call; and comparisons against string literals result in unspecified
3593 behavior and are not portable in C, so they usually indicate that
3594 the programmer intended to use `strcmp'. This warning is enabled
3598 Warn about suspicious uses of logical operators in expressions.
3599 This includes using logical operators in contexts where a bit-wise
3600 operator is likely to be expected.
3602 `-Waggregate-return'
3603 Warn if any functions that return structures or unions are defined
3604 or called. (In languages where you can return an array, this also
3608 Do not warn if an unexpected `__attribute__' is used, such as
3609 unrecognized attributes, function attributes applied to variables,
3610 etc. This will not stop errors for incorrect use of supported
3613 `-Wno-builtin-macro-redefined'
3614 Do not warn if certain built-in macros are redefined. This
3615 suppresses warnings for redefinition of `__TIMESTAMP__',
3616 `__TIME__', `__DATE__', `__FILE__', and `__BASE_FILE__'.
3618 `-Wstrict-prototypes (C and Objective-C only)'
3619 Warn if a function is declared or defined without specifying the
3620 argument types. (An old-style function definition is permitted
3621 without a warning if preceded by a declaration which specifies the
3624 `-Wold-style-declaration (C and Objective-C only)'
3625 Warn for obsolescent usages, according to the C Standard, in a
3626 declaration. For example, warn if storage-class specifiers like
3627 `static' are not the first things in a declaration. This warning
3628 is also enabled by `-Wextra'.
3630 `-Wold-style-definition (C and Objective-C only)'
3631 Warn if an old-style function definition is used. A warning is
3632 given even if there is a previous prototype.
3634 `-Wmissing-parameter-type (C and Objective-C only)'
3635 A function parameter is declared without a type specifier in
3636 K&R-style functions:
3640 This warning is also enabled by `-Wextra'.
3642 `-Wmissing-prototypes (C and Objective-C only)'
3643 Warn if a global function is defined without a previous prototype
3644 declaration. This warning is issued even if the definition itself
3645 provides a prototype. The aim is to detect global functions that
3646 fail to be declared in header files.
3648 `-Wmissing-declarations'
3649 Warn if a global function is defined without a previous
3650 declaration. Do so even if the definition itself provides a
3651 prototype. Use this option to detect global functions that are
3652 not declared in header files. In C++, no warnings are issued for
3653 function templates, or for inline functions, or for functions in
3654 anonymous namespaces.
3656 `-Wmissing-field-initializers'
3657 Warn if a structure's initializer has some fields missing. For
3658 example, the following code would cause such a warning, because
3659 `x.h' is implicitly zero:
3661 struct s { int f, g, h; };
3662 struct s x = { 3, 4 };
3664 This option does not warn about designated initializers, so the
3665 following modification would not trigger a warning:
3667 struct s { int f, g, h; };
3668 struct s x = { .f = 3, .g = 4 };
3670 This warning is included in `-Wextra'. To get other `-Wextra'
3671 warnings without this one, use `-Wextra
3672 -Wno-missing-field-initializers'.
3674 `-Wmissing-noreturn'
3675 Warn about functions which might be candidates for attribute
3676 `noreturn'. Note these are only possible candidates, not absolute
3677 ones. Care should be taken to manually verify functions actually
3678 do not ever return before adding the `noreturn' attribute,
3679 otherwise subtle code generation bugs could be introduced. You
3680 will not get a warning for `main' in hosted C environments.
3682 `-Wmissing-format-attribute'
3683 Warn about function pointers which might be candidates for `format'
3684 attributes. Note these are only possible candidates, not absolute
3685 ones. GCC will guess that function pointers with `format'
3686 attributes that are used in assignment, initialization, parameter
3687 passing or return statements should have a corresponding `format'
3688 attribute in the resulting type. I.e. the left-hand side of the
3689 assignment or initialization, the type of the parameter variable,
3690 or the return type of the containing function respectively should
3691 also have a `format' attribute to avoid the warning.
3693 GCC will also warn about function definitions which might be
3694 candidates for `format' attributes. Again, these are only
3695 possible candidates. GCC will guess that `format' attributes
3696 might be appropriate for any function that calls a function like
3697 `vprintf' or `vscanf', but this might not always be the case, and
3698 some functions for which `format' attributes are appropriate may
3702 Do not warn if a multicharacter constant (`'FOOF'') is used.
3703 Usually they indicate a typo in the user's code, as they have
3704 implementation-defined values, and should not be used in portable
3707 `-Wnormalized=<none|id|nfc|nfkc>'
3708 In ISO C and ISO C++, two identifiers are different if they are
3709 different sequences of characters. However, sometimes when
3710 characters outside the basic ASCII character set are used, you can
3711 have two different character sequences that look the same. To
3712 avoid confusion, the ISO 10646 standard sets out some
3713 "normalization rules" which when applied ensure that two sequences
3714 that look the same are turned into the same sequence. GCC can
3715 warn you if you are using identifiers which have not been
3716 normalized; this option controls that warning.
3718 There are four levels of warning that GCC supports. The default is
3719 `-Wnormalized=nfc', which warns about any identifier which is not
3720 in the ISO 10646 "C" normalized form, "NFC". NFC is the
3721 recommended form for most uses.
3723 Unfortunately, there are some characters which ISO C and ISO C++
3724 allow in identifiers that when turned into NFC aren't allowable as
3725 identifiers. That is, there's no way to use these symbols in
3726 portable ISO C or C++ and have all your identifiers in NFC.
3727 `-Wnormalized=id' suppresses the warning for these characters. It
3728 is hoped that future versions of the standards involved will
3729 correct this, which is why this option is not the default.
3731 You can switch the warning off for all characters by writing
3732 `-Wnormalized=none'. You would only want to do this if you were
3733 using some other normalization scheme (like "D"), because
3734 otherwise you can easily create bugs that are literally impossible
3737 Some characters in ISO 10646 have distinct meanings but look
3738 identical in some fonts or display methodologies, especially once
3739 formatting has been applied. For instance `\u207F', "SUPERSCRIPT
3740 LATIN SMALL LETTER N", will display just like a regular `n' which
3741 has been placed in a superscript. ISO 10646 defines the "NFKC"
3742 normalization scheme to convert all these into a standard form as
3743 well, and GCC will warn if your code is not in NFKC if you use
3744 `-Wnormalized=nfkc'. This warning is comparable to warning about
3745 every identifier that contains the letter O because it might be
3746 confused with the digit 0, and so is not the default, but may be
3747 useful as a local coding convention if the programming environment
3748 is unable to be fixed to display these characters distinctly.
3751 Do not warn about usage of deprecated features. *Note Deprecated
3754 `-Wno-deprecated-declarations'
3755 Do not warn about uses of functions (*note Function Attributes::),
3756 variables (*note Variable Attributes::), and types (*note Type
3757 Attributes::) marked as deprecated by using the `deprecated'
3761 Do not warn about compile-time overflow in constant expressions.
3763 `-Woverride-init (C and Objective-C only)'
3764 Warn if an initialized field without side effects is overridden
3765 when using designated initializers (*note Designated Initializers:
3768 This warning is included in `-Wextra'. To get other `-Wextra'
3769 warnings without this one, use `-Wextra -Wno-override-init'.
3772 Warn if a structure is given the packed attribute, but the packed
3773 attribute has no effect on the layout or size of the structure.
3774 Such structures may be mis-aligned for little benefit. For
3775 instance, in this code, the variable `f.x' in `struct bar' will be
3776 misaligned even though `struct bar' does not itself have the
3782 } __attribute__((packed));
3788 `-Wpacked-bitfield-compat'
3789 The 4.1, 4.2 and 4.3 series of GCC ignore the `packed' attribute
3790 on bit-fields of type `char'. This has been fixed in GCC 4.4 but
3791 the change can lead to differences in the structure layout. GCC
3792 informs you when the offset of such a field has changed in GCC 4.4.
3793 For example there is no longer a 4-bit padding between field `a'
3794 and `b' in this structure:
3800 } __attribute__ ((packed));
3802 This warning is enabled by default. Use
3803 `-Wno-packed-bitfield-compat' to disable this warning.
3806 Warn if padding is included in a structure, either to align an
3807 element of the structure or to align the whole structure.
3808 Sometimes when this happens it is possible to rearrange the fields
3809 of the structure to reduce the padding and so make the structure
3813 Warn if anything is declared more than once in the same scope,
3814 even in cases where multiple declaration is valid and changes
3817 `-Wnested-externs (C and Objective-C only)'
3818 Warn if an `extern' declaration is encountered within a function.
3820 `-Wunreachable-code'
3821 Warn if the compiler detects that code will never be executed.
3823 This option is intended to warn when the compiler detects that at
3824 least a whole line of source code will never be executed, because
3825 some condition is never satisfied or because it is after a
3826 procedure that never returns.
3828 It is possible for this option to produce a warning even though
3829 there are circumstances under which part of the affected line can
3830 be executed, so care should be taken when removing
3831 apparently-unreachable code.
3833 For instance, when a function is inlined, a warning may mean that
3834 the line is unreachable in only one inlined copy of the function.
3836 This option is not made part of `-Wall' because in a debugging
3837 version of a program there is often substantial code which checks
3838 correct functioning of the program and is, hopefully, unreachable
3839 because the program does work. Another common use of unreachable
3840 code is to provide behavior which is selectable at compile-time.
3843 Warn if a function can not be inlined and it was declared as
3844 inline. Even with this option, the compiler will not warn about
3845 failures to inline functions declared in system headers.
3847 The compiler uses a variety of heuristics to determine whether or
3848 not to inline a function. For example, the compiler takes into
3849 account the size of the function being inlined and the amount of
3850 inlining that has already been done in the current function.
3851 Therefore, seemingly insignificant changes in the source program
3852 can cause the warnings produced by `-Winline' to appear or
3855 `-Wno-invalid-offsetof (C++ and Objective-C++ only)'
3856 Suppress warnings from applying the `offsetof' macro to a non-POD
3857 type. According to the 1998 ISO C++ standard, applying `offsetof'
3858 to a non-POD type is undefined. In existing C++ implementations,
3859 however, `offsetof' typically gives meaningful results even when
3860 applied to certain kinds of non-POD types. (Such as a simple
3861 `struct' that fails to be a POD type only by virtue of having a
3862 constructor.) This flag is for users who are aware that they are
3863 writing nonportable code and who have deliberately chosen to
3864 ignore the warning about it.
3866 The restrictions on `offsetof' may be relaxed in a future version
3867 of the C++ standard.
3869 `-Wno-int-to-pointer-cast (C and Objective-C only)'
3870 Suppress warnings from casts to pointer type of an integer of a
3873 `-Wno-pointer-to-int-cast (C and Objective-C only)'
3874 Suppress warnings from casts from a pointer to an integer type of a
3878 Warn if a precompiled header (*note Precompiled Headers::) is
3879 found in the search path but can't be used.
3882 Warn if `long long' type is used. This is default. To inhibit
3883 the warning messages, use `-Wno-long-long'. Flags `-Wlong-long'
3884 and `-Wno-long-long' are taken into account only when `-pedantic'
3888 Warn if variadic macros are used in pedantic ISO C90 mode, or the
3889 GNU alternate syntax when in pedantic ISO C99 mode. This is
3890 default. To inhibit the warning messages, use
3891 `-Wno-variadic-macros'.
3894 Warn if variable length array is used in the code. `-Wno-vla'
3895 will prevent the `-pedantic' warning of the variable length array.
3897 `-Wvolatile-register-var'
3898 Warn if a register variable is declared volatile. The volatile
3899 modifier does not inhibit all optimizations that may eliminate
3900 reads and/or writes to register variables. This warning is
3903 `-Wdisabled-optimization'
3904 Warn if a requested optimization pass is disabled. This warning
3905 does not generally indicate that there is anything wrong with your
3906 code; it merely indicates that GCC's optimizers were unable to
3907 handle the code effectively. Often, the problem is that your code
3908 is too big or too complex; GCC will refuse to optimize programs
3909 when the optimization itself is likely to take inordinate amounts
3912 `-Wpointer-sign (C and Objective-C only)'
3913 Warn for pointer argument passing or assignment with different
3914 signedness. This option is only supported for C and Objective-C.
3915 It is implied by `-Wall' and by `-pedantic', which can be disabled
3916 with `-Wno-pointer-sign'.
3919 This option is only active when `-fstack-protector' is active. It
3920 warns about functions that will not be protected against stack
3924 Suppress warnings about constructs that cannot be instrumented by
3927 `-Woverlength-strings'
3928 Warn about string constants which are longer than the "minimum
3929 maximum" length specified in the C standard. Modern compilers
3930 generally allow string constants which are much longer than the
3931 standard's minimum limit, but very portable programs should avoid
3932 using longer strings.
3934 The limit applies _after_ string constant concatenation, and does
3935 not count the trailing NUL. In C89, the limit was 509 characters;
3936 in C99, it was raised to 4095. C++98 does not specify a normative
3937 minimum maximum, so we do not diagnose overlength strings in C++.
3939 This option is implied by `-pedantic', and can be disabled with
3940 `-Wno-overlength-strings'.
3943 File: gcc.info, Node: Debugging Options, Next: Optimize Options, Prev: Warning Options, Up: Invoking GCC
3945 3.9 Options for Debugging Your Program or GCC
3946 =============================================
3948 GCC has various special options that are used for debugging either your
3952 Produce debugging information in the operating system's native
3953 format (stabs, COFF, XCOFF, or DWARF 2). GDB can work with this
3954 debugging information.
3956 On most systems that use stabs format, `-g' enables use of extra
3957 debugging information that only GDB can use; this extra information
3958 makes debugging work better in GDB but will probably make other
3959 debuggers crash or refuse to read the program. If you want to
3960 control for certain whether to generate the extra information, use
3961 `-gstabs+', `-gstabs', `-gxcoff+', `-gxcoff', or `-gvms' (see
3964 GCC allows you to use `-g' with `-O'. The shortcuts taken by
3965 optimized code may occasionally produce surprising results: some
3966 variables you declared may not exist at all; flow of control may
3967 briefly move where you did not expect it; some statements may not
3968 be executed because they compute constant results or their values
3969 were already at hand; some statements may execute in different
3970 places because they were moved out of loops.
3972 Nevertheless it proves possible to debug optimized output. This
3973 makes it reasonable to use the optimizer for programs that might
3976 The following options are useful when GCC is generated with the
3977 capability for more than one debugging format.
3980 Produce debugging information for use by GDB. This means to use
3981 the most expressive format available (DWARF 2, stabs, or the
3982 native format if neither of those are supported), including GDB
3983 extensions if at all possible.
3986 Produce debugging information in stabs format (if that is
3987 supported), without GDB extensions. This is the format used by
3988 DBX on most BSD systems. On MIPS, Alpha and System V Release 4
3989 systems this option produces stabs debugging output which is not
3990 understood by DBX or SDB. On System V Release 4 systems this
3991 option requires the GNU assembler.
3993 `-feliminate-unused-debug-symbols'
3994 Produce debugging information in stabs format (if that is
3995 supported), for only symbols that are actually used.
3997 `-femit-class-debug-always'
3998 Instead of emitting debugging information for a C++ class in only
3999 one object file, emit it in all object files using the class.
4000 This option should be used only with debuggers that are unable to
4001 handle the way GCC normally emits debugging information for
4002 classes because using this option will increase the size of
4003 debugging information by as much as a factor of two.
4006 Produce debugging information in stabs format (if that is
4007 supported), using GNU extensions understood only by the GNU
4008 debugger (GDB). The use of these extensions is likely to make
4009 other debuggers crash or refuse to read the program.
4012 Produce debugging information in COFF format (if that is
4013 supported). This is the format used by SDB on most System V
4014 systems prior to System V Release 4.
4017 Produce debugging information in XCOFF format (if that is
4018 supported). This is the format used by the DBX debugger on IBM
4022 Produce debugging information in XCOFF format (if that is
4023 supported), using GNU extensions understood only by the GNU
4024 debugger (GDB). The use of these extensions is likely to make
4025 other debuggers crash or refuse to read the program, and may cause
4026 assemblers other than the GNU assembler (GAS) to fail with an
4030 Produce debugging information in DWARF version 2 format (if that is
4031 supported). This is the format used by DBX on IRIX 6. With this
4032 option, GCC uses features of DWARF version 3 when they are useful;
4033 version 3 is upward compatible with version 2, but may still cause
4034 problems for older debuggers.
4037 Produce debugging information in DWARF version 4 format (if that is
4038 supported). With this option, GCC uses features of DWARF version 4
4039 when they are useful, including the placement of most type
4040 information in separate comdat sections. The DWARF version 4
4041 format is still a draft specification, and this option is
4042 currently experimental.
4045 Produce debugging information in VMS debug format (if that is
4046 supported). This is the format used by DEBUG on VMS systems.
4054 Request debugging information and also use LEVEL to specify how
4055 much information. The default level is 2.
4057 Level 0 produces no debug information at all. Thus, `-g0' negates
4060 Level 1 produces minimal information, enough for making backtraces
4061 in parts of the program that you don't plan to debug. This
4062 includes descriptions of functions and external variables, but no
4063 information about local variables and no line numbers.
4065 Level 3 includes extra information, such as all the macro
4066 definitions present in the program. Some debuggers support macro
4067 expansion when you use `-g3'.
4069 `-gdwarf-2' does not accept a concatenated debug level, because
4070 GCC used to support an option `-gdwarf' that meant to generate
4071 debug information in version 1 of the DWARF format (which is very
4072 different from version 2), and it would have been too confusing.
4073 That debug format is long obsolete, but the option cannot be
4074 changed now. Instead use an additional `-gLEVEL' option to change
4075 the debug level for DWARF2.
4077 `-feliminate-dwarf2-dups'
4078 Compress DWARF2 debugging information by eliminating duplicated
4079 information about each symbol. This option only makes sense when
4080 generating DWARF2 debugging information with `-gdwarf-2'.
4082 `-femit-struct-debug-baseonly'
4083 Emit debug information for struct-like types only when the base
4084 name of the compilation source file matches the base name of file
4085 in which the struct was defined.
4087 This option substantially reduces the size of debugging
4088 information, but at significant potential loss in type information
4089 to the debugger. See `-femit-struct-debug-reduced' for a less
4090 aggressive option. See `-femit-struct-debug-detailed' for more
4093 This option works only with DWARF 2.
4095 `-femit-struct-debug-reduced'
4096 Emit debug information for struct-like types only when the base
4097 name of the compilation source file matches the base name of file
4098 in which the type was defined, unless the struct is a template or
4099 defined in a system header.
4101 This option significantly reduces the size of debugging
4102 information, with some potential loss in type information to the
4103 debugger. See `-femit-struct-debug-baseonly' for a more
4104 aggressive option. See `-femit-struct-debug-detailed' for more
4107 This option works only with DWARF 2.
4109 `-femit-struct-debug-detailed[=SPEC-LIST]'
4110 Specify the struct-like types for which the compiler will generate
4111 debug information. The intent is to reduce duplicate struct debug
4112 information between different object files within the same program.
4114 This option is a detailed version of `-femit-struct-debug-reduced'
4115 and `-femit-struct-debug-baseonly', which will serve for most
4118 A specification has the syntax
4119 [`dir:'|`ind:'][`ord:'|`gen:'](`any'|`sys'|`base'|`none')
4121 The optional first word limits the specification to structs that
4122 are used directly (`dir:') or used indirectly (`ind:'). A struct
4123 type is used directly when it is the type of a variable, member.
4124 Indirect uses arise through pointers to structs. That is, when
4125 use of an incomplete struct would be legal, the use is indirect.
4126 An example is `struct one direct; struct two * indirect;'.
4128 The optional second word limits the specification to ordinary
4129 structs (`ord:') or generic structs (`gen:'). Generic structs are
4130 a bit complicated to explain. For C++, these are non-explicit
4131 specializations of template classes, or non-template classes
4132 within the above. Other programming languages have generics, but
4133 `-femit-struct-debug-detailed' does not yet implement them.
4135 The third word specifies the source files for those structs for
4136 which the compiler will emit debug information. The values `none'
4137 and `any' have the normal meaning. The value `base' means that
4138 the base of name of the file in which the type declaration appears
4139 must match the base of the name of the main compilation file. In
4140 practice, this means that types declared in `foo.c' and `foo.h'
4141 will have debug information, but types declared in other header
4142 will not. The value `sys' means those types satisfying `base' or
4143 declared in system or compiler headers.
4145 You may need to experiment to determine the best settings for your
4148 The default is `-femit-struct-debug-detailed=all'.
4150 This option works only with DWARF 2.
4152 `-fno-merge-debug-strings'
4153 Direct the linker to not merge together strings in the debugging
4154 information which are identical in different object files.
4155 Merging is not supported by all assemblers or linkers. Merging
4156 decreases the size of the debug information in the output file at
4157 the cost of increasing link processing time. Merging is enabled
4160 `-fdebug-prefix-map=OLD=NEW'
4161 When compiling files in directory `OLD', record debugging
4162 information describing them as in `NEW' instead.
4164 `-fno-dwarf2-cfi-asm'
4165 Emit DWARF 2 unwind info as compiler generated `.eh_frame' section
4166 instead of using GAS `.cfi_*' directives.
4169 Generate extra code to write profile information suitable for the
4170 analysis program `prof'. You must use this option when compiling
4171 the source files you want data about, and you must also use it when
4175 Generate extra code to write profile information suitable for the
4176 analysis program `gprof'. You must use this option when compiling
4177 the source files you want data about, and you must also use it when
4181 Makes the compiler print out each function name as it is compiled,
4182 and print some statistics about each pass when it finishes.
4185 Makes the compiler print some statistics about the time consumed
4186 by each pass when it finishes.
4189 Makes the compiler print some statistics about permanent memory
4190 allocation when it finishes.
4192 `-fpre-ipa-mem-report'
4194 `-fpost-ipa-mem-report'
4195 Makes the compiler print some statistics about permanent memory
4196 allocation before or after interprocedural optimization.
4199 Add code so that program flow "arcs" are instrumented. During
4200 execution the program records how many times each branch and call
4201 is executed and how many times it is taken or returns. When the
4202 compiled program exits it saves this data to a file called
4203 `AUXNAME.gcda' for each source file. The data may be used for
4204 profile-directed optimizations (`-fbranch-probabilities'), or for
4205 test coverage analysis (`-ftest-coverage'). Each object file's
4206 AUXNAME is generated from the name of the output file, if
4207 explicitly specified and it is not the final executable, otherwise
4208 it is the basename of the source file. In both cases any suffix
4209 is removed (e.g. `foo.gcda' for input file `dir/foo.c', or
4210 `dir/foo.gcda' for output file specified as `-o dir/foo.o').
4211 *Note Cross-profiling::.
4214 This option is used to compile and link code instrumented for
4215 coverage analysis. The option is a synonym for `-fprofile-arcs'
4216 `-ftest-coverage' (when compiling) and `-lgcov' (when linking).
4217 See the documentation for those options for more details.
4219 * Compile the source files with `-fprofile-arcs' plus
4220 optimization and code generation options. For test coverage
4221 analysis, use the additional `-ftest-coverage' option. You
4222 do not need to profile every source file in a program.
4224 * Link your object files with `-lgcov' or `-fprofile-arcs' (the
4225 latter implies the former).
4227 * Run the program on a representative workload to generate the
4228 arc profile information. This may be repeated any number of
4229 times. You can run concurrent instances of your program, and
4230 provided that the file system supports locking, the data
4231 files will be correctly updated. Also `fork' calls are
4232 detected and correctly handled (double counting will not
4235 * For profile-directed optimizations, compile the source files
4236 again with the same optimization and code generation options
4237 plus `-fbranch-probabilities' (*note Options that Control
4238 Optimization: Optimize Options.).
4240 * For test coverage analysis, use `gcov' to produce human
4241 readable information from the `.gcno' and `.gcda' files.
4242 Refer to the `gcov' documentation for further information.
4245 With `-fprofile-arcs', for each function of your program GCC
4246 creates a program flow graph, then finds a spanning tree for the
4247 graph. Only arcs that are not on the spanning tree have to be
4248 instrumented: the compiler adds code to count the number of times
4249 that these arcs are executed. When an arc is the only exit or
4250 only entrance to a block, the instrumentation code can be added to
4251 the block; otherwise, a new basic block must be created to hold
4252 the instrumentation code.
4255 Produce a notes file that the `gcov' code-coverage utility (*note
4256 `gcov'--a Test Coverage Program: Gcov.) can use to show program
4257 coverage. Each source file's note file is called `AUXNAME.gcno'.
4258 Refer to the `-fprofile-arcs' option above for a description of
4259 AUXNAME and instructions on how to generate test coverage data.
4260 Coverage data will match the source files more closely, if you do
4264 Print the name and the counter upperbound for all debug counters.
4266 `-fdbg-cnt=COUNTER-VALUE-LIST'
4267 Set the internal debug counter upperbound. COUNTER-VALUE-LIST is a
4268 comma-separated list of NAME:VALUE pairs which sets the upperbound
4269 of each debug counter NAME to VALUE. All debug counters have the
4270 initial upperbound of UINT_MAX, thus dbg_cnt() returns true always
4271 unless the upperbound is set by this option. e.g. With
4272 -fdbg-cnt=dce:10,tail_call:0 dbg_cnt(dce) will return true only
4273 for first 10 invocations and dbg_cnt(tail_call) will return false
4278 Says to make debugging dumps during compilation at times specified
4279 by LETTERS. This is used for debugging the RTL-based passes of the
4280 compiler. The file names for most of the dumps are made by
4281 appending a pass number and a word to the DUMPNAME, and the files
4282 are created in the directory of the output file. DUMPNAME is
4283 generated from the name of the output file, if explicitly specified
4284 and it is not an executable, otherwise it is the basename of the
4285 source file. These switches may have different effects when `-E'
4286 is used for preprocessing.
4288 Debug dumps can be enabled with a `-fdump-rtl' switch or some `-d'
4289 option LETTERS. Here are the possible letters for use in PASS and
4290 LETTERS, and their meanings:
4292 `-fdump-rtl-alignments'
4293 Dump after branch alignments have been computed.
4295 `-fdump-rtl-asmcons'
4296 Dump after fixing rtl statements that have unsatisfied in/out
4299 `-fdump-rtl-auto_inc_dec'
4300 Dump after auto-inc-dec discovery. This pass is only run on
4301 architectures that have auto inc or auto dec instructions.
4303 `-fdump-rtl-barriers'
4304 Dump after cleaning up the barrier instructions.
4307 Dump after partitioning hot and cold basic blocks.
4310 Dump after block reordering.
4314 `-fdump-rtl-btl1' and `-fdump-rtl-btl2' enable dumping after
4315 the two branch target load optimization passes.
4318 Dump after jump bypassing and control flow optimizations.
4320 `-fdump-rtl-combine'
4321 Dump after the RTL instruction combination pass.
4323 `-fdump-rtl-compgotos'
4324 Dump after duplicating the computed gotos.
4329 `-fdump-rtl-ce1', `-fdump-rtl-ce2', and `-fdump-rtl-ce3'
4330 enable dumping after the three if conversion passes.
4332 `-fdump-rtl-cprop_hardreg'
4333 Dump after hard register copy propagation.
4336 Dump after combining stack adjustments.
4340 `-fdump-rtl-cse1' and `-fdump-rtl-cse2' enable dumping after
4341 the two common sub-expression elimination passes.
4344 Dump after the standalone dead code elimination passes.
4347 Dump after delayed branch scheduling.
4351 `-fdump-rtl-dce1' and `-fdump-rtl-dce2' enable dumping after
4352 the two dead store elimination passes.
4355 Dump after finalization of EH handling code.
4357 `-fdump-rtl-eh_ranges'
4358 Dump after conversion of EH handling range regions.
4361 Dump after RTL generation.
4363 `-fdump-rtl-fwprop1'
4364 `-fdump-rtl-fwprop2'
4365 `-fdump-rtl-fwprop1' and `-fdump-rtl-fwprop2' enable dumping
4366 after the two forward propagation passes.
4370 `-fdump-rtl-gcse1' and `-fdump-rtl-gcse2' enable dumping
4371 after global common subexpression elimination.
4373 `-fdump-rtl-init-regs'
4374 Dump after the initialization of the registers.
4376 `-fdump-rtl-initvals'
4377 Dump after the computation of the initial value sets.
4379 `-fdump-rtl-into_cfglayout'
4380 Dump after converting to cfglayout mode.
4383 Dump after iterated register allocation.
4386 Dump after the second jump optimization.
4389 `-fdump-rtl-loop2' enables dumping after the rtl loop
4390 optimization passes.
4393 Dump after performing the machine dependent reorganization
4394 pass, if that pass exists.
4396 `-fdump-rtl-mode_sw'
4397 Dump after removing redundant mode switches.
4400 Dump after register renumbering.
4402 `-fdump-rtl-outof_cfglayout'
4403 Dump after converting from cfglayout mode.
4405 `-fdump-rtl-peephole2'
4406 Dump after the peephole pass.
4408 `-fdump-rtl-postreload'
4409 Dump after post-reload optimizations.
4411 `-fdump-rtl-pro_and_epilogue'
4412 Dump after generating the function pro and epilogues.
4414 `-fdump-rtl-regmove'
4415 Dump after the register move pass.
4419 `-fdump-rtl-sched1' and `-fdump-rtl-sched2' enable dumping
4420 after the basic block scheduling passes.
4423 Dump after sign extension elimination.
4425 `-fdump-rtl-seqabstr'
4426 Dump after common sequence discovery.
4428 `-fdump-rtl-shorten'
4429 Dump after shortening branches.
4431 `-fdump-rtl-sibling'
4432 Dump after sibling call optimizations.
4439 `-fdump-rtl-split1', `-fdump-rtl-split2',
4440 `-fdump-rtl-split3', `-fdump-rtl-split4' and
4441 `-fdump-rtl-split5' enable dumping after five rounds of
4442 instruction splitting.
4445 Dump after modulo scheduling. This pass is only run on some
4449 Dump after conversion from GCC's "flat register file"
4450 registers to the x87's stack-like registers. This pass is
4451 only run on x86 variants.
4453 `-fdump-rtl-subreg1'
4454 `-fdump-rtl-subreg2'
4455 `-fdump-rtl-subreg1' and `-fdump-rtl-subreg2' enable dumping
4456 after the two subreg expansion passes.
4458 `-fdump-rtl-unshare'
4459 Dump after all rtl has been unshared.
4461 `-fdump-rtl-vartrack'
4462 Dump after variable tracking.
4465 Dump after converting virtual registers to hard registers.
4468 Dump after live range splitting.
4470 `-fdump-rtl-regclass'
4471 `-fdump-rtl-subregs_of_mode_init'
4472 `-fdump-rtl-subregs_of_mode_finish'
4474 `-fdump-rtl-dfinish'
4475 These dumps are defined but always produce empty files.
4478 Produce all the dumps listed above.
4481 Annotate the assembler output with miscellaneous debugging
4485 Dump all macro definitions, at the end of preprocessing, in
4486 addition to normal output.
4489 Produce a core dump whenever an error occurs.
4492 Print statistics on memory usage, at the end of the run, to
4496 Annotate the assembler output with a comment indicating which
4497 pattern and alternative was used. The length of each
4498 instruction is also printed.
4501 Dump the RTL in the assembler output as a comment before each
4502 instruction. Also turns on `-dp' annotation.
4505 For each of the other indicated dump files
4506 (`-fdump-rtl-PASS'), dump a representation of the control
4507 flow graph suitable for viewing with VCG to `FILE.PASS.vcg'.
4510 Just generate RTL for a function instead of compiling it.
4511 Usually used with `-fdump-rtl-expand'.
4514 Dump debugging information during parsing, to standard error.
4517 When doing debugging dumps, suppress address output. This makes
4518 it more feasible to use diff on debugging dumps for compiler
4519 invocations with different compiler binaries and/or different text
4520 / bss / data / heap / stack / dso start locations.
4523 When doing debugging dumps, suppress instruction numbers and
4524 address output. This makes it more feasible to use diff on
4525 debugging dumps for compiler invocations with different options,
4526 in particular with and without `-g'.
4528 `-fdump-translation-unit (C++ only)'
4529 `-fdump-translation-unit-OPTIONS (C++ only)'
4530 Dump a representation of the tree structure for the entire
4531 translation unit to a file. The file name is made by appending
4532 `.tu' to the source file name, and the file is created in the same
4533 directory as the output file. If the `-OPTIONS' form is used,
4534 OPTIONS controls the details of the dump as described for the
4535 `-fdump-tree' options.
4537 `-fdump-class-hierarchy (C++ only)'
4538 `-fdump-class-hierarchy-OPTIONS (C++ only)'
4539 Dump a representation of each class's hierarchy and virtual
4540 function table layout to a file. The file name is made by
4541 appending `.class' to the source file name, and the file is
4542 created in the same directory as the output file. If the
4543 `-OPTIONS' form is used, OPTIONS controls the details of the dump
4544 as described for the `-fdump-tree' options.
4547 Control the dumping at various stages of inter-procedural analysis
4548 language tree to a file. The file name is generated by appending a
4549 switch specific suffix to the source file name, and the file is
4550 created in the same directory as the output file. The following
4554 Enables all inter-procedural analysis dumps.
4557 Dumps information about call-graph optimization, unused
4558 function removal, and inlining decisions.
4561 Dump after function inlining.
4564 `-fdump-statistics-OPTION'
4565 Enable and control dumping of pass statistics in a separate file.
4566 The file name is generated by appending a suffix ending in
4567 `.statistics' to the source file name, and the file is created in
4568 the same directory as the output file. If the `-OPTION' form is
4569 used, `-stats' will cause counters to be summed over the whole
4570 compilation unit while `-details' will dump every event as the
4571 passes generate them. The default with no option is to sum
4572 counters for each function compiled.
4574 `-fdump-tree-SWITCH'
4575 `-fdump-tree-SWITCH-OPTIONS'
4576 Control the dumping at various stages of processing the
4577 intermediate language tree to a file. The file name is generated
4578 by appending a switch specific suffix to the source file name, and
4579 the file is created in the same directory as the output file. If
4580 the `-OPTIONS' form is used, OPTIONS is a list of `-' separated
4581 options that control the details of the dump. Not all options are
4582 applicable to all dumps, those which are not meaningful will be
4583 ignored. The following options are available
4586 Print the address of each node. Usually this is not
4587 meaningful as it changes according to the environment and
4588 source file. Its primary use is for tying up a dump file
4589 with a debug environment.
4592 Inhibit dumping of members of a scope or body of a function
4593 merely because that scope has been reached. Only dump such
4594 items when they are directly reachable by some other path.
4595 When dumping pretty-printed trees, this option inhibits
4596 dumping the bodies of control structures.
4599 Print a raw representation of the tree. By default, trees are
4600 pretty-printed into a C-like representation.
4603 Enable more detailed dumps (not honored by every dump option).
4606 Enable dumping various statistics about the pass (not honored
4607 by every dump option).
4610 Enable showing basic block boundaries (disabled in raw dumps).
4613 Enable showing virtual operands for every statement.
4616 Enable showing line numbers for statements.
4619 Enable showing the unique ID (`DECL_UID') for each variable.
4622 Enable showing the tree dump for each statement.
4625 Turn on all options, except `raw', `slim', `verbose' and
4628 The following tree dumps are possible:
4630 Dump before any tree based optimization, to `FILE.original'.
4633 Dump after all tree based optimization, to `FILE.optimized'.
4636 Dump each function before and after the gimplification pass
4637 to a file. The file name is made by appending `.gimple' to
4638 the source file name.
4641 Dump the control flow graph of each function to a file. The
4642 file name is made by appending `.cfg' to the source file name.
4645 Dump the control flow graph of each function to a file in VCG
4646 format. The file name is made by appending `.vcg' to the
4647 source file name. Note that if the file contains more than
4648 one function, the generated file cannot be used directly by
4649 VCG. You will need to cut and paste each function's graph
4650 into its own separate file first.
4653 Dump each function after copying loop headers. The file name
4654 is made by appending `.ch' to the source file name.
4657 Dump SSA related information to a file. The file name is
4658 made by appending `.ssa' to the source file name.
4661 Dump aliasing information for each function. The file name
4662 is made by appending `.alias' to the source file name.
4665 Dump each function after CCP. The file name is made by
4666 appending `.ccp' to the source file name.
4669 Dump each function after STORE-CCP. The file name is made by
4670 appending `.storeccp' to the source file name.
4673 Dump trees after partial redundancy elimination. The file
4674 name is made by appending `.pre' to the source file name.
4677 Dump trees after full redundancy elimination. The file name
4678 is made by appending `.fre' to the source file name.
4681 Dump trees after copy propagation. The file name is made by
4682 appending `.copyprop' to the source file name.
4685 Dump trees after store copy-propagation. The file name is
4686 made by appending `.store_copyprop' to the source file name.
4689 Dump each function after dead code elimination. The file
4690 name is made by appending `.dce' to the source file name.
4693 Dump each function after adding mudflap instrumentation. The
4694 file name is made by appending `.mudflap' to the source file
4698 Dump each function after performing scalar replacement of
4699 aggregates. The file name is made by appending `.sra' to the
4703 Dump each function after performing code sinking. The file
4704 name is made by appending `.sink' to the source file name.
4707 Dump each function after applying dominator tree
4708 optimizations. The file name is made by appending `.dom' to
4709 the source file name.
4712 Dump each function after applying dead store elimination.
4713 The file name is made by appending `.dse' to the source file
4717 Dump each function after optimizing PHI nodes into
4718 straightline code. The file name is made by appending
4719 `.phiopt' to the source file name.
4722 Dump each function after forward propagating single use
4723 variables. The file name is made by appending `.forwprop' to
4724 the source file name.
4727 Dump each function after applying the copy rename
4728 optimization. The file name is made by appending
4729 `.copyrename' to the source file name.
4732 Dump each function after applying the named return value
4733 optimization on generic trees. The file name is made by
4734 appending `.nrv' to the source file name.
4737 Dump each function after applying vectorization of loops.
4738 The file name is made by appending `.vect' to the source file
4742 Dump each function after Value Range Propagation (VRP). The
4743 file name is made by appending `.vrp' to the source file name.
4746 Enable all the available tree dumps with the flags provided
4749 `-ftree-vectorizer-verbose=N'
4750 This option controls the amount of debugging output the vectorizer
4751 prints. This information is written to standard error, unless
4752 `-fdump-tree-all' or `-fdump-tree-vect' is specified, in which
4753 case it is output to the usual dump listing file, `.vect'. For
4754 N=0 no diagnostic information is reported. If N=1 the vectorizer
4755 reports each loop that got vectorized, and the total number of
4756 loops that got vectorized. If N=2 the vectorizer also reports
4757 non-vectorized loops that passed the first analysis phase
4758 (vect_analyze_loop_form) - i.e. countable, inner-most, single-bb,
4759 single-entry/exit loops. This is the same verbosity level that
4760 `-fdump-tree-vect-stats' uses. Higher verbosity levels mean
4761 either more information dumped for each reported loop, or same
4762 amount of information reported for more loops: If N=3, alignment
4763 related information is added to the reports. If N=4,
4764 data-references related information (e.g. memory dependences,
4765 memory access-patterns) is added to the reports. If N=5, the
4766 vectorizer reports also non-vectorized inner-most loops that did
4767 not pass the first analysis phase (i.e., may not be countable, or
4768 may have complicated control-flow). If N=6, the vectorizer
4769 reports also non-vectorized nested loops. For N=7, all the
4770 information the vectorizer generates during its analysis and
4771 transformation is reported. This is the same verbosity level that
4772 `-fdump-tree-vect-details' uses.
4774 `-frandom-seed=STRING'
4775 This option provides a seed that GCC uses when it would otherwise
4776 use random numbers. It is used to generate certain symbol names
4777 that have to be different in every compiled file. It is also used
4778 to place unique stamps in coverage data files and the object files
4779 that produce them. You can use the `-frandom-seed' option to
4780 produce reproducibly identical object files.
4782 The STRING should be different for every file you compile.
4785 On targets that use instruction scheduling, this option controls
4786 the amount of debugging output the scheduler prints. This
4787 information is written to standard error, unless
4788 `-fdump-rtl-sched1' or `-fdump-rtl-sched2' is specified, in which
4789 case it is output to the usual dump listing file, `.sched' or
4790 `.sched2' respectively. However for N greater than nine, the
4791 output is always printed to standard error.
4793 For N greater than zero, `-fsched-verbose' outputs the same
4794 information as `-fdump-rtl-sched1' and `-fdump-rtl-sched2'. For N
4795 greater than one, it also output basic block probabilities,
4796 detailed ready list information and unit/insn info. For N greater
4797 than two, it includes RTL at abort point, control-flow and regions
4798 info. And for N over four, `-fsched-verbose' also includes
4802 Store the usual "temporary" intermediate files permanently; place
4803 them in the current directory and name them based on the source
4804 file. Thus, compiling `foo.c' with `-c -save-temps' would produce
4805 files `foo.i' and `foo.s', as well as `foo.o'. This creates a
4806 preprocessed `foo.i' output file even though the compiler now
4807 normally uses an integrated preprocessor.
4809 When used in combination with the `-x' command line option,
4810 `-save-temps' is sensible enough to avoid over writing an input
4811 source file with the same extension as an intermediate file. The
4812 corresponding intermediate file may be obtained by renaming the
4813 source file before using `-save-temps'.
4816 Report the CPU time taken by each subprocess in the compilation
4817 sequence. For C source files, this is the compiler proper and
4818 assembler (plus the linker if linking is done). The output looks
4824 The first number on each line is the "user time", that is time
4825 spent executing the program itself. The second number is "system
4826 time", time spent executing operating system routines on behalf of
4827 the program. Both numbers are in seconds.
4830 Run variable tracking pass. It computes where variables are
4831 stored at each position in code. Better debugging information is
4832 then generated (if the debugging information format supports this
4835 It is enabled by default when compiling with optimization (`-Os',
4836 `-O', `-O2', ...), debugging information (`-g') and the debug info
4839 `-print-file-name=LIBRARY'
4840 Print the full absolute name of the library file LIBRARY that
4841 would be used when linking--and don't do anything else. With this
4842 option, GCC does not compile or link anything; it just prints the
4845 `-print-multi-directory'
4846 Print the directory name corresponding to the multilib selected by
4847 any other switches present in the command line. This directory is
4848 supposed to exist in `GCC_EXEC_PREFIX'.
4851 Print the mapping from multilib directory names to compiler
4852 switches that enable them. The directory name is separated from
4853 the switches by `;', and each switch starts with an `@' instead of
4854 the `-', without spaces between multiple switches. This is
4855 supposed to ease shell-processing.
4857 `-print-prog-name=PROGRAM'
4858 Like `-print-file-name', but searches for a program such as `cpp'.
4860 `-print-libgcc-file-name'
4861 Same as `-print-file-name=libgcc.a'.
4863 This is useful when you use `-nostdlib' or `-nodefaultlibs' but
4864 you do want to link with `libgcc.a'. You can do
4866 gcc -nostdlib FILES... `gcc -print-libgcc-file-name`
4868 `-print-search-dirs'
4869 Print the name of the configured installation directory and a list
4870 of program and library directories `gcc' will search--and don't do
4873 This is useful when `gcc' prints the error message `installation
4874 problem, cannot exec cpp0: No such file or directory'. To resolve
4875 this you either need to put `cpp0' and the other compiler
4876 components where `gcc' expects to find them, or you can set the
4877 environment variable `GCC_EXEC_PREFIX' to the directory where you
4878 installed them. Don't forget the trailing `/'. *Note Environment
4882 Print the target sysroot directory that will be used during
4883 compilation. This is the target sysroot specified either at
4884 configure time or using the `--sysroot' option, possibly with an
4885 extra suffix that depends on compilation options. If no target
4886 sysroot is specified, the option prints nothing.
4888 `-print-sysroot-headers-suffix'
4889 Print the suffix added to the target sysroot when searching for
4890 headers, or give an error if the compiler is not configured with
4891 such a suffix--and don't do anything else.
4894 Print the compiler's target machine (for example,
4895 `i686-pc-linux-gnu')--and don't do anything else.
4898 Print the compiler version (for example, `3.0')--and don't do
4902 Print the compiler's built-in specs--and don't do anything else.
4903 (This is used when GCC itself is being built.) *Note Spec Files::.
4905 `-feliminate-unused-debug-types'
4906 Normally, when producing DWARF2 output, GCC will emit debugging
4907 information for all types declared in a compilation unit,
4908 regardless of whether or not they are actually used in that
4909 compilation unit. Sometimes this is useful, such as if, in the
4910 debugger, you want to cast a value to a type that is not actually
4911 used in your program (but is declared). More often, however, this
4912 results in a significant amount of wasted space. With this
4913 option, GCC will avoid producing debug symbol output for types
4914 that are nowhere used in the source file being compiled.
4917 File: gcc.info, Node: Optimize Options, Next: Preprocessor Options, Prev: Debugging Options, Up: Invoking GCC
4919 3.10 Options That Control Optimization
4920 ======================================
4922 These options control various sorts of optimizations.
4924 Without any optimization option, the compiler's goal is to reduce the
4925 cost of compilation and to make debugging produce the expected results.
4926 Statements are independent: if you stop the program with a breakpoint
4927 between statements, you can then assign a new value to any variable or
4928 change the program counter to any other statement in the function and
4929 get exactly the results you would expect from the source code.
4931 Turning on optimization flags makes the compiler attempt to improve
4932 the performance and/or code size at the expense of compilation time and
4933 possibly the ability to debug the program.
4935 The compiler performs optimization based on the knowledge it has of the
4936 program. Compiling multiple files at once to a single output file mode
4937 allows the compiler to use information gained from all of the files
4938 when compiling each of them.
4940 Not all optimizations are controlled directly by a flag. Only
4941 optimizations that have a flag are listed.
4945 Optimize. Optimizing compilation takes somewhat more time, and a
4946 lot more memory for a large function.
4948 With `-O', the compiler tries to reduce code size and execution
4949 time, without performing any optimizations that take a great deal
4950 of compilation time.
4952 `-O' turns on the following optimization flags:
4959 -fguess-branch-probability
4962 -finline-small-functions
4967 -ftree-builtin-call-dce
4972 -ftree-dominator-opts
4979 `-O' also turns on `-fomit-frame-pointer' on machines where doing
4980 so does not interfere with debugging.
4983 Optimize even more. GCC performs nearly all supported
4984 optimizations that do not involve a space-speed tradeoff. As
4985 compared to `-O', this option increases both compilation time and
4986 the performance of the generated code.
4988 `-O2' turns on all optimization flags specified by `-O'. It also
4989 turns on the following optimization flags:
4991 -falign-functions -falign-jumps
4992 -falign-loops -falign-labels
4995 -fcse-follow-jumps -fcse-skip-blocks
4996 -fdelete-null-pointer-checks
4997 -fexpensive-optimizations
5000 -foptimize-sibling-calls
5003 -freorder-blocks -freorder-functions
5004 -frerun-cse-after-loop
5005 -fsched-interblock -fsched-spec
5006 -fschedule-insns -fschedule-insns2
5007 -fstrict-aliasing -fstrict-overflow
5008 -ftree-switch-conversion
5012 Please note the warning under `-fgcse' about invoking `-O2' on
5013 programs that use computed gotos.
5016 Optimize yet more. `-O3' turns on all optimizations specified by
5017 `-O2' and also turns on the `-finline-functions',
5018 `-funswitch-loops', `-fpredictive-commoning',
5019 `-fgcse-after-reload' and `-ftree-vectorize' options.
5022 Reduce compilation time and make debugging produce the expected
5023 results. This is the default.
5026 Optimize for size. `-Os' enables all `-O2' optimizations that do
5027 not typically increase code size. It also performs further
5028 optimizations designed to reduce code size.
5030 `-Os' disables the following optimization flags:
5031 -falign-functions -falign-jumps -falign-loops
5032 -falign-labels -freorder-blocks -freorder-blocks-and-partition
5033 -fprefetch-loop-arrays -ftree-vect-loop-version
5035 If you use multiple `-O' options, with or without level numbers,
5036 the last such option is the one that is effective.
5038 Options of the form `-fFLAG' specify machine-independent flags. Most
5039 flags have both positive and negative forms; the negative form of
5040 `-ffoo' would be `-fno-foo'. In the table below, only one of the forms
5041 is listed--the one you typically will use. You can figure out the
5042 other form by either removing `no-' or adding it.
5044 The following options control specific optimizations. They are either
5045 activated by `-O' options or are related to ones that are. You can use
5046 the following flags in the rare cases when "fine-tuning" of
5047 optimizations to be performed is desired.
5049 `-fno-default-inline'
5050 Do not make member functions inline by default merely because they
5051 are defined inside the class scope (C++ only). Otherwise, when
5052 you specify `-O', member functions defined inside class scope are
5053 compiled inline by default; i.e., you don't need to add `inline'
5054 in front of the member function name.
5057 Always pop the arguments to each function call as soon as that
5058 function returns. For machines which must pop arguments after a
5059 function call, the compiler normally lets arguments accumulate on
5060 the stack for several function calls and pops them all at once.
5062 Disabled at levels `-O', `-O2', `-O3', `-Os'.
5064 `-fforward-propagate'
5065 Perform a forward propagation pass on RTL. The pass tries to
5066 combine two instructions and checks if the result can be
5067 simplified. If loop unrolling is active, two passes are performed
5068 and the second is scheduled after loop unrolling.
5070 This option is enabled by default at optimization levels `-O2',
5073 `-fomit-frame-pointer'
5074 Don't keep the frame pointer in a register for functions that
5075 don't need one. This avoids the instructions to save, set up and
5076 restore frame pointers; it also makes an extra register available
5077 in many functions. *It also makes debugging impossible on some
5080 On some machines, such as the VAX, this flag has no effect, because
5081 the standard calling sequence automatically handles the frame
5082 pointer and nothing is saved by pretending it doesn't exist. The
5083 machine-description macro `FRAME_POINTER_REQUIRED' controls
5084 whether a target machine supports this flag. *Note Register
5085 Usage: (gccint)Registers.
5087 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5089 `-foptimize-sibling-calls'
5090 Optimize sibling and tail recursive calls.
5092 Enabled at levels `-O2', `-O3', `-Os'.
5095 Don't pay attention to the `inline' keyword. Normally this option
5096 is used to keep the compiler from expanding any functions inline.
5097 Note that if you are not optimizing, no functions can be expanded
5100 `-finline-small-functions'
5101 Integrate functions into their callers when their body is smaller
5102 than expected function call code (so overall size of program gets
5103 smaller). The compiler heuristically decides which functions are
5104 simple enough to be worth integrating in this way.
5106 Enabled at level `-O2'.
5108 `-findirect-inlining'
5109 Inline also indirect calls that are discovered to be known at
5110 compile time thanks to previous inlining. This option has any
5111 effect only when inlining itself is turned on by the
5112 `-finline-functions' or `-finline-small-functions' options.
5114 Enabled at level `-O2'.
5116 `-finline-functions'
5117 Integrate all simple functions into their callers. The compiler
5118 heuristically decides which functions are simple enough to be worth
5119 integrating in this way.
5121 If all calls to a given function are integrated, and the function
5122 is declared `static', then the function is normally not output as
5123 assembler code in its own right.
5125 Enabled at level `-O3'.
5127 `-finline-functions-called-once'
5128 Consider all `static' functions called once for inlining into their
5129 caller even if they are not marked `inline'. If a call to a given
5130 function is integrated, then the function is not output as
5131 assembler code in its own right.
5133 Enabled at levels `-O1', `-O2', `-O3' and `-Os'.
5136 Inline functions marked by `always_inline' and functions whose
5137 body seems smaller than the function call overhead early before
5138 doing `-fprofile-generate' instrumentation and real inlining pass.
5139 Doing so makes profiling significantly cheaper and usually
5140 inlining faster on programs having large chains of nested wrapper
5146 By default, GCC limits the size of functions that can be inlined.
5147 This flag allows coarse control of this limit. N is the size of
5148 functions that can be inlined in number of pseudo instructions.
5150 Inlining is actually controlled by a number of parameters, which
5151 may be specified individually by using `--param NAME=VALUE'. The
5152 `-finline-limit=N' option sets some of these parameters as follows:
5154 `max-inline-insns-single'
5157 `max-inline-insns-auto'
5160 See below for a documentation of the individual parameters
5161 controlling inlining and for the defaults of these parameters.
5163 _Note:_ there may be no value to `-finline-limit' that results in
5166 _Note:_ pseudo instruction represents, in this particular context,
5167 an abstract measurement of function's size. In no way does it
5168 represent a count of assembly instructions and as such its exact
5169 meaning might change from one release to an another.
5171 `-fkeep-inline-functions'
5172 In C, emit `static' functions that are declared `inline' into the
5173 object file, even if the function has been inlined into all of its
5174 callers. This switch does not affect functions using the `extern
5175 inline' extension in GNU C89. In C++, emit any and all inline
5176 functions into the object file.
5178 `-fkeep-static-consts'
5179 Emit variables declared `static const' when optimization isn't
5180 turned on, even if the variables aren't referenced.
5182 GCC enables this option by default. If you want to force the
5183 compiler to check if the variable was referenced, regardless of
5184 whether or not optimization is turned on, use the
5185 `-fno-keep-static-consts' option.
5188 Attempt to merge identical constants (string constants and
5189 floating point constants) across compilation units.
5191 This option is the default for optimized compilation if the
5192 assembler and linker support it. Use `-fno-merge-constants' to
5193 inhibit this behavior.
5195 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5197 `-fmerge-all-constants'
5198 Attempt to merge identical constants and identical variables.
5200 This option implies `-fmerge-constants'. In addition to
5201 `-fmerge-constants' this considers e.g. even constant initialized
5202 arrays or initialized constant variables with integral or floating
5203 point types. Languages like C or C++ require each variable,
5204 including multiple instances of the same variable in recursive
5205 calls, to have distinct locations, so using this option will
5206 result in non-conforming behavior.
5209 Perform swing modulo scheduling immediately before the first
5210 scheduling pass. This pass looks at innermost loops and reorders
5211 their instructions by overlapping different iterations.
5213 `-fmodulo-sched-allow-regmoves'
5214 Perform more aggressive SMS based modulo scheduling with register
5215 moves allowed. By setting this flag certain anti-dependences
5216 edges will be deleted which will trigger the generation of
5217 reg-moves based on the life-range analysis. This option is
5218 effective only with `-fmodulo-sched' enabled.
5220 `-fno-branch-count-reg'
5221 Do not use "decrement and branch" instructions on a count register,
5222 but instead generate a sequence of instructions that decrement a
5223 register, compare it against zero, then branch based upon the
5224 result. This option is only meaningful on architectures that
5225 support such instructions, which include x86, PowerPC, IA-64 and
5228 The default is `-fbranch-count-reg'.
5231 Do not put function addresses in registers; make each instruction
5232 that calls a constant function contain the function's address
5235 This option results in less efficient code, but some strange hacks
5236 that alter the assembler output may be confused by the
5237 optimizations performed when this option is not used.
5239 The default is `-ffunction-cse'
5241 `-fno-zero-initialized-in-bss'
5242 If the target supports a BSS section, GCC by default puts
5243 variables that are initialized to zero into BSS. This can save
5244 space in the resulting code.
5246 This option turns off this behavior because some programs
5247 explicitly rely on variables going to the data section. E.g., so
5248 that the resulting executable can find the beginning of that
5249 section and/or make assumptions based on that.
5251 The default is `-fzero-initialized-in-bss'.
5253 `-fmudflap -fmudflapth -fmudflapir'
5254 For front-ends that support it (C and C++), instrument all risky
5255 pointer/array dereferencing operations, some standard library
5256 string/heap functions, and some other associated constructs with
5257 range/validity tests. Modules so instrumented should be immune to
5258 buffer overflows, invalid heap use, and some other classes of C/C++
5259 programming errors. The instrumentation relies on a separate
5260 runtime library (`libmudflap'), which will be linked into a
5261 program if `-fmudflap' is given at link time. Run-time behavior
5262 of the instrumented program is controlled by the `MUDFLAP_OPTIONS'
5263 environment variable. See `env MUDFLAP_OPTIONS=-help a.out' for
5266 Use `-fmudflapth' instead of `-fmudflap' to compile and to link if
5267 your program is multi-threaded. Use `-fmudflapir', in addition to
5268 `-fmudflap' or `-fmudflapth', if instrumentation should ignore
5269 pointer reads. This produces less instrumentation (and therefore
5270 faster execution) and still provides some protection against
5271 outright memory corrupting writes, but allows erroneously read
5272 data to propagate within a program.
5275 Perform optimizations where we check to see if a jump branches to a
5276 location where another comparison subsumed by the first is found.
5277 If so, the first branch is redirected to either the destination of
5278 the second branch or a point immediately following it, depending
5279 on whether the condition is known to be true or false.
5281 Enabled at levels `-O2', `-O3', `-Os'.
5283 `-fsplit-wide-types'
5284 When using a type that occupies multiple registers, such as `long
5285 long' on a 32-bit system, split the registers apart and allocate
5286 them independently. This normally generates better code for those
5287 types, but may make debugging more difficult.
5289 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5291 `-fcse-follow-jumps'
5292 In common subexpression elimination (CSE), scan through jump
5293 instructions when the target of the jump is not reached by any
5294 other path. For example, when CSE encounters an `if' statement
5295 with an `else' clause, CSE will follow the jump when the condition
5298 Enabled at levels `-O2', `-O3', `-Os'.
5301 This is similar to `-fcse-follow-jumps', but causes CSE to follow
5302 jumps which conditionally skip over blocks. When CSE encounters a
5303 simple `if' statement with no else clause, `-fcse-skip-blocks'
5304 causes CSE to follow the jump around the body of the `if'.
5306 Enabled at levels `-O2', `-O3', `-Os'.
5308 `-frerun-cse-after-loop'
5309 Re-run common subexpression elimination after loop optimizations
5312 Enabled at levels `-O2', `-O3', `-Os'.
5315 Perform a global common subexpression elimination pass. This pass
5316 also performs global constant and copy propagation.
5318 _Note:_ When compiling a program using computed gotos, a GCC
5319 extension, you may get better runtime performance if you disable
5320 the global common subexpression elimination pass by adding
5321 `-fno-gcse' to the command line.
5323 Enabled at levels `-O2', `-O3', `-Os'.
5326 When `-fgcse-lm' is enabled, global common subexpression
5327 elimination will attempt to move loads which are only killed by
5328 stores into themselves. This allows a loop containing a
5329 load/store sequence to be changed to a load outside the loop, and
5330 a copy/store within the loop.
5332 Enabled by default when gcse is enabled.
5335 When `-fgcse-sm' is enabled, a store motion pass is run after
5336 global common subexpression elimination. This pass will attempt
5337 to move stores out of loops. When used in conjunction with
5338 `-fgcse-lm', loops containing a load/store sequence can be changed
5339 to a load before the loop and a store after the loop.
5341 Not enabled at any optimization level.
5344 When `-fgcse-las' is enabled, the global common subexpression
5345 elimination pass eliminates redundant loads that come after stores
5346 to the same memory location (both partial and full redundancies).
5348 Not enabled at any optimization level.
5350 `-fgcse-after-reload'
5351 When `-fgcse-after-reload' is enabled, a redundant load elimination
5352 pass is performed after reload. The purpose of this pass is to
5353 cleanup redundant spilling.
5355 `-funsafe-loop-optimizations'
5356 If given, the loop optimizer will assume that loop indices do not
5357 overflow, and that the loops with nontrivial exit condition are not
5358 infinite. This enables a wider range of loop optimizations even if
5359 the loop optimizer itself cannot prove that these assumptions are
5360 valid. Using `-Wunsafe-loop-optimizations', the compiler will
5361 warn you if it finds this kind of loop.
5364 Perform cross-jumping transformation. This transformation unifies
5365 equivalent code and save code size. The resulting code may or may
5366 not perform better than without cross-jumping.
5368 Enabled at levels `-O2', `-O3', `-Os'.
5371 Combine increments or decrements of addresses with memory accesses.
5372 This pass is always skipped on architectures that do not have
5373 instructions to support this. Enabled by default at `-O' and
5374 higher on architectures that support this.
5377 Perform dead code elimination (DCE) on RTL. Enabled by default at
5381 Perform dead store elimination (DSE) on RTL. Enabled by default
5385 Attempt to transform conditional jumps into branch-less
5386 equivalents. This include use of conditional moves, min, max, set
5387 flags and abs instructions, and some tricks doable by standard
5388 arithmetics. The use of conditional execution on chips where it
5389 is available is controlled by `if-conversion2'.
5391 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5394 Use conditional execution (where available) to transform
5395 conditional jumps into branch-less equivalents.
5397 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5399 `-fdelete-null-pointer-checks'
5400 Use global dataflow analysis to identify and eliminate useless
5401 checks for null pointers. The compiler assumes that dereferencing
5402 a null pointer would have halted the program. If a pointer is
5403 checked after it has already been dereferenced, it cannot be null.
5405 In some environments, this assumption is not true, and programs can
5406 safely dereference null pointers. Use
5407 `-fno-delete-null-pointer-checks' to disable this optimization for
5408 programs which depend on that behavior.
5410 Enabled at levels `-O2', `-O3', `-Os'.
5412 `-fexpensive-optimizations'
5413 Perform a number of minor optimizations that are relatively
5416 Enabled at levels `-O2', `-O3', `-Os'.
5418 `-foptimize-register-move'
5420 Attempt to reassign register numbers in move instructions and as
5421 operands of other simple instructions in order to maximize the
5422 amount of register tying. This is especially helpful on machines
5423 with two-operand instructions.
5425 Note `-fregmove' and `-foptimize-register-move' are the same
5428 Enabled at levels `-O2', `-O3', `-Os'.
5430 `-fira-algorithm=ALGORITHM'
5431 Use specified coloring algorithm for the integrated register
5432 allocator. The ALGORITHM argument should be `priority' or `CB'.
5433 The first algorithm specifies Chow's priority coloring, the second
5434 one specifies Chaitin-Briggs coloring. The second algorithm can
5435 be unimplemented for some architectures. If it is implemented, it
5436 is the default because Chaitin-Briggs coloring as a rule generates
5439 `-fira-region=REGION'
5440 Use specified regions for the integrated register allocator. The
5441 REGION argument should be one of `all', `mixed', or `one'. The
5442 first value means using all loops as register allocation regions,
5443 the second value which is the default means using all loops except
5444 for loops with small register pressure as the regions, and third
5445 one means using all function as a single region. The first value
5446 can give best result for machines with small size and irregular
5447 register set, the third one results in faster and generates decent
5448 code and the smallest size code, and the default value usually
5449 give the best results in most cases and for most architectures.
5452 Do optimistic register coalescing. This option might be
5453 profitable for architectures with big regular register files.
5455 `-fno-ira-share-save-slots'
5456 Switch off sharing stack slots used for saving call used hard
5457 registers living through a call. Each hard register will get a
5458 separate stack slot and as a result function stack frame will be
5461 `-fno-ira-share-spill-slots'
5462 Switch off sharing stack slots allocated for pseudo-registers.
5463 Each pseudo-register which did not get a hard register will get a
5464 separate stack slot and as a result function stack frame will be
5468 Set up how verbose dump file for the integrated register allocator
5469 will be. Default value is 5. If the value is greater or equal to
5470 10, the dump file will be stderr as if the value were N minus 10.
5473 If supported for the target machine, attempt to reorder
5474 instructions to exploit instruction slots available after delayed
5475 branch instructions.
5477 Enabled at levels `-O', `-O2', `-O3', `-Os'.
5480 If supported for the target machine, attempt to reorder
5481 instructions to eliminate execution stalls due to required data
5482 being unavailable. This helps machines that have slow floating
5483 point or memory load instructions by allowing other instructions
5484 to be issued until the result of the load or floating point
5485 instruction is required.
5487 Enabled at levels `-O2', `-O3', `-Os'.
5490 Similar to `-fschedule-insns', but requests an additional pass of
5491 instruction scheduling after register allocation has been done.
5492 This is especially useful on machines with a relatively small
5493 number of registers and where memory load instructions take more
5496 Enabled at levels `-O2', `-O3', `-Os'.
5498 `-fno-sched-interblock'
5499 Don't schedule instructions across basic blocks. This is normally
5500 enabled by default when scheduling before register allocation, i.e.
5501 with `-fschedule-insns' or at `-O2' or higher.
5504 Don't allow speculative motion of non-load instructions. This is
5505 normally enabled by default when scheduling before register
5506 allocation, i.e. with `-fschedule-insns' or at `-O2' or higher.
5509 Allow speculative motion of some load instructions. This only
5510 makes sense when scheduling before register allocation, i.e. with
5511 `-fschedule-insns' or at `-O2' or higher.
5513 `-fsched-spec-load-dangerous'
5514 Allow speculative motion of more 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-stalled-insns'
5519 `-fsched-stalled-insns=N'
5520 Define how many insns (if any) can be moved prematurely from the
5521 queue of stalled insns into the ready list, during the second
5522 scheduling pass. `-fno-sched-stalled-insns' means that no insns
5523 will be moved prematurely, `-fsched-stalled-insns=0' means there
5524 is no limit on how many queued insns can be moved prematurely.
5525 `-fsched-stalled-insns' without a value is equivalent to
5526 `-fsched-stalled-insns=1'.
5528 `-fsched-stalled-insns-dep'
5529 `-fsched-stalled-insns-dep=N'
5530 Define how many insn groups (cycles) will be examined for a
5531 dependency on a stalled insn that is candidate for premature
5532 removal from the queue of stalled insns. This has an effect only
5533 during the second scheduling pass, and only if
5534 `-fsched-stalled-insns' is used. `-fno-sched-stalled-insns-dep'
5535 is equivalent to `-fsched-stalled-insns-dep=0'.
5536 `-fsched-stalled-insns-dep' without a value is equivalent to
5537 `-fsched-stalled-insns-dep=1'.
5539 `-fsched2-use-superblocks'
5540 When scheduling after register allocation, do use superblock
5541 scheduling algorithm. Superblock scheduling allows motion across
5542 basic block boundaries resulting on faster schedules. This option
5543 is experimental, as not all machine descriptions used by GCC model
5544 the CPU closely enough to avoid unreliable results from the
5547 This only makes sense when scheduling after register allocation,
5548 i.e. with `-fschedule-insns2' or at `-O2' or higher.
5550 `-fsched2-use-traces'
5551 Use `-fsched2-use-superblocks' algorithm when scheduling after
5552 register allocation and additionally perform code duplication in
5553 order to increase the size of superblocks using tracer pass. See
5554 `-ftracer' for details on trace formation.
5556 This mode should produce faster but significantly longer programs.
5557 Also without `-fbranch-probabilities' the traces constructed may
5558 not match the reality and hurt the performance. This only makes
5559 sense when scheduling after register allocation, i.e. with
5560 `-fschedule-insns2' or at `-O2' or higher.
5563 Eliminate redundant sign extension instructions and move the
5564 non-redundant ones to optimal placement using lazy code motion
5567 `-freschedule-modulo-scheduled-loops'
5568 The modulo scheduling comes before the traditional scheduling, if
5569 a loop was modulo scheduled we may want to prevent the later
5570 scheduling passes from changing its schedule, we use this option
5573 `-fselective-scheduling'
5574 Schedule instructions using selective scheduling algorithm.
5575 Selective scheduling runs instead of the first scheduler pass.
5577 `-fselective-scheduling2'
5578 Schedule instructions using selective scheduling algorithm.
5579 Selective scheduling runs instead of the second scheduler pass.
5581 `-fsel-sched-pipelining'
5582 Enable software pipelining of innermost loops during selective
5583 scheduling. This option has no effect until one of
5584 `-fselective-scheduling' or `-fselective-scheduling2' is turned on.
5586 `-fsel-sched-pipelining-outer-loops'
5587 When pipelining loops during selective scheduling, also pipeline
5588 outer loops. This option has no effect until
5589 `-fsel-sched-pipelining' is turned on.
5592 Enable values to be allocated in registers that will be clobbered
5593 by function calls, by emitting extra instructions to save and
5594 restore the registers around such calls. Such allocation is done
5595 only when it seems to result in better code than would otherwise
5598 This option is always enabled by default on certain machines,
5599 usually those which have no call-preserved registers to use
5602 Enabled at levels `-O2', `-O3', `-Os'.
5605 Attempt to minimize stack usage. The compiler will attempt to use
5606 less stack space, even if that makes the program slower. This
5607 option implies setting the `large-stack-frame' parameter to 100
5608 and the `large-stack-frame-growth' parameter to 400.
5611 Perform reassociation on trees. This flag is enabled by default
5615 Perform partial redundancy elimination (PRE) on trees. This flag
5616 is enabled by default at `-O2' and `-O3'.
5619 Perform full redundancy elimination (FRE) on trees. The difference
5620 between FRE and PRE is that FRE only considers expressions that
5621 are computed on all paths leading to the redundant computation.
5622 This analysis is faster than PRE, though it exposes fewer
5623 redundancies. This flag is enabled by default at `-O' and higher.
5626 Perform copy propagation on trees. This pass eliminates
5627 unnecessary copy operations. This flag is enabled by default at
5631 Discover which functions are pure or constant. Enabled by default
5635 Discover which static variables do not escape cannot escape the
5636 compilation unit. Enabled by default at `-O' and higher.
5638 `-fipa-struct-reorg'
5639 Perform structure reorganization optimization, that change C-like
5640 structures layout in order to better utilize spatial locality.
5641 This transformation is affective for programs containing arrays of
5642 structures. Available in two compilation modes: profile-based
5643 (enabled with `-fprofile-generate') or static (which uses built-in
5644 heuristics). Require `-fipa-type-escape' to provide the safety of
5645 this transformation. It works only in whole program mode, so it
5646 requires `-fwhole-program' and `-combine' to be enabled.
5647 Structures considered `cold' by this transformation are not
5648 affected (see `--param struct-reorg-cold-struct-ratio=VALUE').
5650 With this flag, the program debug info reflects a new structure
5654 Perform interprocedural pointer analysis. This option is
5655 experimental and does not affect generated code.
5658 Perform interprocedural constant propagation. This optimization
5659 analyzes the program to determine when values passed to functions
5660 are constants and then optimizes accordingly. This optimization
5661 can substantially increase performance if the application has
5662 constants passed to functions. This flag is enabled by default at
5663 `-O2', `-Os' and `-O3'.
5666 Perform function cloning to make interprocedural constant
5667 propagation stronger. When enabled, interprocedural constant
5668 propagation will perform function cloning when externally visible
5669 function can be called with constant arguments. Because this
5670 optimization can create multiple copies of functions, it may
5671 significantly increase code size (see `--param
5672 ipcp-unit-growth=VALUE'). This flag is enabled by default at
5675 `-fipa-matrix-reorg'
5676 Perform matrix flattening and transposing. Matrix flattening
5677 tries to replace a m-dimensional matrix with its equivalent
5678 n-dimensional matrix, where n < m. This reduces the level of
5679 indirection needed for accessing the elements of the matrix. The
5680 second optimization is matrix transposing that attempts to change
5681 the order of the matrix's dimensions in order to improve cache
5682 locality. Both optimizations need the `-fwhole-program' flag.
5683 Transposing is enabled only if profiling information is available.
5686 Perform forward store motion on trees. This flag is enabled by
5687 default at `-O' and higher.
5690 Perform sparse conditional constant propagation (CCP) on trees.
5691 This pass only operates on local scalar variables and is enabled
5692 by default at `-O' and higher.
5694 `-ftree-switch-conversion'
5695 Perform conversion of simple initializations in a switch to
5696 initializations from a scalar array. This flag is enabled by
5697 default at `-O2' and higher.
5700 Perform dead code elimination (DCE) on trees. This flag is
5701 enabled by default at `-O' and higher.
5703 `-ftree-builtin-call-dce'
5704 Perform conditional dead code elimination (DCE) for calls to
5705 builtin functions that may set `errno' but are otherwise
5706 side-effect free. This flag is enabled by default at `-O2' and
5707 higher if `-Os' is not also specified.
5709 `-ftree-dominator-opts'
5710 Perform a variety of simple scalar cleanups (constant/copy
5711 propagation, redundancy elimination, range propagation and
5712 expression simplification) based on a dominator tree traversal.
5713 This also performs jump threading (to reduce jumps to jumps). This
5714 flag is enabled by default at `-O' and higher.
5717 Perform dead store elimination (DSE) on trees. A dead store is a
5718 store into a memory location which will later be overwritten by
5719 another store without any intervening loads. In this case the
5720 earlier store can be deleted. This flag is enabled by default at
5724 Perform loop header copying on trees. This is beneficial since it
5725 increases effectiveness of code motion optimizations. It also
5726 saves one jump. This flag is enabled by default at `-O' and
5727 higher. It is not enabled for `-Os', since it usually increases
5730 `-ftree-loop-optimize'
5731 Perform loop optimizations on trees. This flag is enabled by
5732 default at `-O' and higher.
5734 `-ftree-loop-linear'
5735 Perform linear loop transformations on tree. This flag can
5736 improve cache performance and allow further loop optimizations to
5739 `-floop-interchange'
5740 Perform loop interchange transformations on loops. Interchanging
5741 two nested loops switches the inner and outer loops. For example,
5745 A(J, I) = A(J, I) * C
5748 loop interchange will transform the loop as if the user had
5752 A(J, I) = A(J, I) * C
5755 which can be beneficial when `N' is larger than the caches,
5756 because in Fortran, the elements of an array are stored in memory
5757 contiguously by column, and the original loop iterates over rows,
5758 potentially creating at each access a cache miss. This
5759 optimization applies to all the languages supported by GCC and is
5760 not limited to Fortran. To use this code transformation, GCC has
5761 to be configured with `--with-ppl' and `--with-cloog' to enable the
5762 Graphite loop transformation infrastructure.
5765 Perform loop strip mining transformations on loops. Strip mining
5766 splits a loop into two nested loops. The outer loop has strides
5767 equal to the strip size and the inner loop has strides of the
5768 original loop within a strip. For example, given a loop like:
5772 loop strip mining will transform the loop as if the user had
5775 DO I = II, min (II + 3, N)
5779 This optimization applies to all the languages supported by GCC
5780 and is not limited to Fortran. To use this code transformation,
5781 GCC has to be configured with `--with-ppl' and `--with-cloog' to
5782 enable the Graphite loop transformation infrastructure.
5785 Perform loop blocking transformations on loops. Blocking strip
5786 mines each loop in the loop nest such that the memory accesses of
5787 the element loops fit inside caches. For example, given a loop
5791 A(J, I) = B(I) + C(J)
5794 loop blocking will transform the loop as if the user had written:
5797 DO I = II, min (II + 63, N)
5798 DO J = JJ, min (JJ + 63, M)
5799 A(J, I) = B(I) + C(J)
5804 which can be beneficial when `M' is larger than the caches,
5805 because the innermost loop will iterate over a smaller amount of
5806 data that can be kept in the caches. This optimization applies to
5807 all the languages supported by GCC and is not limited to Fortran.
5808 To use this code transformation, GCC has to be configured with
5809 `--with-ppl' and `--with-cloog' to enable the Graphite loop
5810 transformation infrastructure.
5813 Compare the results of several data dependence analyzers. This
5814 option is used for debugging the data dependence analyzers.
5816 `-ftree-loop-distribution'
5817 Perform loop distribution. This flag can improve cache
5818 performance on big loop bodies and allow further loop
5819 optimizations, like parallelization or vectorization, to take
5820 place. For example, the loop
5834 Perform loop invariant motion on trees. This pass moves only
5835 invariants that would be hard to handle at RTL level (function
5836 calls, operations that expand to nontrivial sequences of insns).
5837 With `-funswitch-loops' it also moves operands of conditions that
5838 are invariant out of the loop, so that we can use just trivial
5839 invariantness analysis in loop unswitching. The pass also includes
5842 `-ftree-loop-ivcanon'
5843 Create a canonical counter for number of iterations in the loop
5844 for that determining number of iterations requires complicated
5845 analysis. Later optimizations then may determine the number
5846 easily. Useful especially in connection with unrolling.
5849 Perform induction variable optimizations (strength reduction,
5850 induction variable merging and induction variable elimination) on
5853 `-ftree-parallelize-loops=n'
5854 Parallelize loops, i.e., split their iteration space to run in n
5855 threads. This is only possible for loops whose iterations are
5856 independent and can be arbitrarily reordered. The optimization is
5857 only profitable on multiprocessor machines, for loops that are
5858 CPU-intensive, rather than constrained e.g. by memory bandwidth.
5859 This option implies `-pthread', and thus is only supported on
5860 targets that have support for `-pthread'.
5863 Perform scalar replacement of aggregates. This pass replaces
5864 structure references with scalars to prevent committing structures
5865 to memory too early. This flag is enabled by default at `-O' and
5869 Perform copy renaming on trees. This pass attempts to rename
5870 compiler temporaries to other variables at copy locations, usually
5871 resulting in variable names which more closely resemble the
5872 original variables. This flag is enabled by default at `-O' and
5876 Perform temporary expression replacement during the SSA->normal
5877 phase. Single use/single def temporaries are replaced at their
5878 use location with their defining expression. This results in
5879 non-GIMPLE code, but gives the expanders much more complex trees
5880 to work on resulting in better RTL generation. This is enabled by
5881 default at `-O' and higher.
5884 Perform loop vectorization on trees. This flag is enabled by
5887 `-ftree-vect-loop-version'
5888 Perform loop versioning when doing loop vectorization on trees.
5889 When a loop appears to be vectorizable except that data alignment
5890 or data dependence cannot be determined at compile time then
5891 vectorized and non-vectorized versions of the loop are generated
5892 along with runtime checks for alignment or dependence to control
5893 which version is executed. This option is enabled by default
5894 except at level `-Os' where it is disabled.
5897 Enable cost model for vectorization.
5900 Perform Value Range Propagation on trees. This is similar to the
5901 constant propagation pass, but instead of values, ranges of values
5902 are propagated. This allows the optimizers to remove unnecessary
5903 range checks like array bound checks and null pointer checks.
5904 This is enabled by default at `-O2' and higher. Null pointer check
5905 elimination is only done if `-fdelete-null-pointer-checks' is
5909 Perform tail duplication to enlarge superblock size. This
5910 transformation simplifies the control flow of the function
5911 allowing other optimizations to do better job.
5914 Unroll loops whose number of iterations can be determined at
5915 compile time or upon entry to the loop. `-funroll-loops' implies
5916 `-frerun-cse-after-loop'. This option makes code larger, and may
5917 or may not make it run faster.
5919 `-funroll-all-loops'
5920 Unroll all loops, even if their number of iterations is uncertain
5921 when the loop is entered. This usually makes programs run more
5922 slowly. `-funroll-all-loops' implies the same options as
5925 `-fsplit-ivs-in-unroller'
5926 Enables expressing of values of induction variables in later
5927 iterations of the unrolled loop using the value in the first
5928 iteration. This breaks long dependency chains, thus improving
5929 efficiency of the scheduling passes.
5931 Combination of `-fweb' and CSE is often sufficient to obtain the
5932 same effect. However in cases the loop body is more complicated
5933 than a single basic block, this is not reliable. It also does not
5934 work at all on some of the architectures due to restrictions in
5937 This optimization is enabled by default.
5939 `-fvariable-expansion-in-unroller'
5940 With this option, the compiler will create multiple copies of some
5941 local variables when unrolling a loop which can result in superior
5944 `-fpredictive-commoning'
5945 Perform predictive commoning optimization, i.e., reusing
5946 computations (especially memory loads and stores) performed in
5947 previous iterations of loops.
5949 This option is enabled at level `-O3'.
5951 `-fprefetch-loop-arrays'
5952 If supported by the target machine, generate instructions to
5953 prefetch memory to improve the performance of loops that access
5956 This option may generate better or worse code; results are highly
5957 dependent on the structure of loops within the source code.
5959 Disabled at level `-Os'.
5963 Disable any machine-specific peephole optimizations. The
5964 difference between `-fno-peephole' and `-fno-peephole2' is in how
5965 they are implemented in the compiler; some targets use one, some
5966 use the other, a few use both.
5968 `-fpeephole' is enabled by default. `-fpeephole2' enabled at
5969 levels `-O2', `-O3', `-Os'.
5971 `-fno-guess-branch-probability'
5972 Do not guess branch probabilities using heuristics.
5974 GCC will use heuristics to guess branch probabilities if they are
5975 not provided by profiling feedback (`-fprofile-arcs'). These
5976 heuristics are based on the control flow graph. If some branch
5977 probabilities are specified by `__builtin_expect', then the
5978 heuristics will be used to guess branch probabilities for the rest
5979 of the control flow graph, taking the `__builtin_expect' info into
5980 account. The interactions between the heuristics and
5981 `__builtin_expect' can be complex, and in some cases, it may be
5982 useful to disable the heuristics so that the effects of
5983 `__builtin_expect' are easier to understand.
5985 The default is `-fguess-branch-probability' at levels `-O', `-O2',
5989 Reorder basic blocks in the compiled function in order to reduce
5990 number of taken branches and improve code locality.
5992 Enabled at levels `-O2', `-O3'.
5994 `-freorder-blocks-and-partition'
5995 In addition to reordering basic blocks in the compiled function,
5996 in order to reduce number of taken branches, partitions hot and
5997 cold basic blocks into separate sections of the assembly and .o
5998 files, to improve paging and cache locality performance.
6000 This optimization is automatically turned off in the presence of
6001 exception handling, for linkonce sections, for functions with a
6002 user-defined section attribute and on any architecture that does
6003 not support named sections.
6005 `-freorder-functions'
6006 Reorder functions in the object file in order to improve code
6007 locality. This is implemented by using special subsections
6008 `.text.hot' for most frequently executed functions and
6009 `.text.unlikely' for unlikely executed functions. Reordering is
6010 done by the linker so object file format must support named
6011 sections and linker must place them in a reasonable way.
6013 Also profile feedback must be available in to make this option
6014 effective. See `-fprofile-arcs' for details.
6016 Enabled at levels `-O2', `-O3', `-Os'.
6019 Allows the compiler to assume the strictest aliasing rules
6020 applicable to the language being compiled. For C (and C++), this
6021 activates optimizations based on the type of expressions. In
6022 particular, an object of one type is assumed never to reside at
6023 the same address as an object of a different type, unless the
6024 types are almost the same. For example, an `unsigned int' can
6025 alias an `int', but not a `void*' or a `double'. A character type
6026 may alias any other type.
6028 Pay special attention to code like this:
6039 The practice of reading from a different union member than the one
6040 most recently written to (called "type-punning") is common. Even
6041 with `-fstrict-aliasing', type-punning is allowed, provided the
6042 memory is accessed through the union type. So, the code above
6043 will work as expected. *Note Structures unions enumerations and
6044 bit-fields implementation::. However, this code might not:
6053 Similarly, access by taking the address, casting the resulting
6054 pointer and dereferencing the result has undefined behavior, even
6055 if the cast uses a union type, e.g.:
6058 return ((union a_union *) &d)->i;
6061 The `-fstrict-aliasing' option is enabled at levels `-O2', `-O3',
6065 Allow the compiler to assume strict signed overflow rules,
6066 depending on the language being compiled. For C (and C++) this
6067 means that overflow when doing arithmetic with signed numbers is
6068 undefined, which means that the compiler may assume that it will
6069 not happen. This permits various optimizations. For example, the
6070 compiler will assume that an expression like `i + 10 > i' will
6071 always be true for signed `i'. This assumption is only valid if
6072 signed overflow is undefined, as the expression is false if `i +
6073 10' overflows when using twos complement arithmetic. When this
6074 option is in effect any attempt to determine whether an operation
6075 on signed numbers will overflow must be written carefully to not
6076 actually involve overflow.
6078 This option also allows the compiler to assume strict pointer
6079 semantics: given a pointer to an object, if adding an offset to
6080 that pointer does not produce a pointer to the same object, the
6081 addition is undefined. This permits the compiler to conclude that
6082 `p + u > p' is always true for a pointer `p' and unsigned integer
6083 `u'. This assumption is only valid because pointer wraparound is
6084 undefined, as the expression is false if `p + u' overflows using
6085 twos complement arithmetic.
6087 See also the `-fwrapv' option. Using `-fwrapv' means that integer
6088 signed overflow is fully defined: it wraps. When `-fwrapv' is
6089 used, there is no difference between `-fstrict-overflow' and
6090 `-fno-strict-overflow' for integers. With `-fwrapv' certain types
6091 of overflow are permitted. For example, if the compiler gets an
6092 overflow when doing arithmetic on constants, the overflowed value
6093 can still be used with `-fwrapv', but not otherwise.
6095 The `-fstrict-overflow' option is enabled at levels `-O2', `-O3',
6099 `-falign-functions=N'
6100 Align the start of functions to the next power-of-two greater than
6101 N, skipping up to N bytes. For instance, `-falign-functions=32'
6102 aligns functions to the next 32-byte boundary, but
6103 `-falign-functions=24' would align to the next 32-byte boundary
6104 only if this can be done by skipping 23 bytes or less.
6106 `-fno-align-functions' and `-falign-functions=1' are equivalent
6107 and mean that functions will not be aligned.
6109 Some assemblers only support this flag when N is a power of two;
6110 in that case, it is rounded up.
6112 If N is not specified or is zero, use a machine-dependent default.
6114 Enabled at levels `-O2', `-O3'.
6118 Align all branch targets to a power-of-two boundary, skipping up to
6119 N bytes like `-falign-functions'. This option can easily make
6120 code slower, because it must insert dummy operations for when the
6121 branch target is reached in the usual flow of the code.
6123 `-fno-align-labels' and `-falign-labels=1' are equivalent and mean
6124 that labels will not be aligned.
6126 If `-falign-loops' or `-falign-jumps' are applicable and are
6127 greater than this value, then their values are used instead.
6129 If N is not specified or is zero, use a machine-dependent default
6130 which is very likely to be `1', meaning no alignment.
6132 Enabled at levels `-O2', `-O3'.
6136 Align loops to a power-of-two boundary, skipping up to N bytes
6137 like `-falign-functions'. The hope is that the loop will be
6138 executed many times, which will make up for any execution of the
6141 `-fno-align-loops' and `-falign-loops=1' are equivalent and mean
6142 that loops will not be aligned.
6144 If N is not specified or is zero, use a machine-dependent default.
6146 Enabled at levels `-O2', `-O3'.
6150 Align branch targets to a power-of-two boundary, for branch targets
6151 where the targets can only be reached by jumping, skipping up to N
6152 bytes like `-falign-functions'. In this case, no dummy operations
6155 `-fno-align-jumps' and `-falign-jumps=1' are equivalent and mean
6156 that loops will not be aligned.
6158 If N is not specified or is zero, use a machine-dependent default.
6160 Enabled at levels `-O2', `-O3'.
6163 This option is left for compatibility reasons. `-funit-at-a-time'
6164 has no effect, while `-fno-unit-at-a-time' implies
6165 `-fno-toplevel-reorder' and `-fno-section-anchors'.
6169 `-fno-toplevel-reorder'
6170 Do not reorder top-level functions, variables, and `asm'
6171 statements. Output them in the same order that they appear in the
6172 input file. When this option is used, unreferenced static
6173 variables will not be removed. This option is intended to support
6174 existing code which relies on a particular ordering. For new
6175 code, it is better to use attributes.
6177 Enabled at level `-O0'. When disabled explicitly, it also imply
6178 `-fno-section-anchors' that is otherwise enabled at `-O0' on some
6182 Constructs webs as commonly used for register allocation purposes
6183 and assign each web individual pseudo register. This allows the
6184 register allocation pass to operate on pseudos directly, but also
6185 strengthens several other optimization passes, such as CSE, loop
6186 optimizer and trivial dead code remover. It can, however, make
6187 debugging impossible, since variables will no longer stay in a
6190 Enabled by default with `-funroll-loops'.
6193 Assume that the current compilation unit represents whole program
6194 being compiled. All public functions and variables with the
6195 exception of `main' and those merged by attribute
6196 `externally_visible' become static functions and in a affect gets
6197 more aggressively optimized by interprocedural optimizers. While
6198 this option is equivalent to proper use of `static' keyword for
6199 programs consisting of single file, in combination with option
6200 `--combine' this flag can be used to compile most of smaller scale
6201 C programs since the functions and variables become local for the
6202 whole combined compilation unit, not for the single source file
6205 This option is not supported for Fortran programs.
6208 After register allocation and post-register allocation instruction
6209 splitting, we perform a copy-propagation pass to try to reduce
6210 scheduling dependencies and occasionally eliminate the copy.
6212 Enabled at levels `-O', `-O2', `-O3', `-Os'.
6214 `-fprofile-correction'
6215 Profiles collected using an instrumented binary for multi-threaded
6216 programs may be inconsistent due to missed counter updates. When
6217 this option is specified, GCC will use heuristics to correct or
6218 smooth out such inconsistencies. By default, GCC will emit an
6219 error message when an inconsistent profile is detected.
6221 `-fprofile-dir=PATH'
6222 Set the directory to search the profile data files in to PATH.
6223 This option affects only the profile data generated by
6224 `-fprofile-generate', `-ftest-coverage', `-fprofile-arcs' and used
6225 by `-fprofile-use' and `-fbranch-probabilities' and its related
6226 options. By default, GCC will use the current directory as PATH
6227 thus the profile data file will appear in the same directory as
6230 `-fprofile-generate'
6231 `-fprofile-generate=PATH'
6232 Enable options usually used for instrumenting application to
6233 produce profile useful for later recompilation with profile
6234 feedback based optimization. You must use `-fprofile-generate'
6235 both when compiling and when linking your program.
6237 The following options are enabled: `-fprofile-arcs',
6238 `-fprofile-values', `-fvpt'.
6240 If PATH is specified, GCC will look at the PATH to find the
6241 profile feedback data files. See `-fprofile-dir'.
6244 `-fprofile-use=PATH'
6245 Enable profile feedback directed optimizations, and optimizations
6246 generally profitable only with profile feedback available.
6248 The following options are enabled: `-fbranch-probabilities',
6249 `-fvpt', `-funroll-loops', `-fpeel-loops', `-ftracer'
6251 By default, GCC emits an error message if the feedback profiles do
6252 not match the source code. This error can be turned into a
6253 warning by using `-Wcoverage-mismatch'. Note this may result in
6254 poorly optimized code.
6256 If PATH is specified, GCC will look at the PATH to find the
6257 profile feedback data files. See `-fprofile-dir'.
6259 The following options control compiler behavior regarding floating
6260 point arithmetic. These options trade off between speed and
6261 correctness. All must be specifically enabled.
6264 Do not store floating point variables in registers, and inhibit
6265 other options that might change whether a floating point value is
6266 taken from a register or memory.
6268 This option prevents undesirable excess precision on machines such
6269 as the 68000 where the floating registers (of the 68881) keep more
6270 precision than a `double' is supposed to have. Similarly for the
6271 x86 architecture. For most programs, the excess precision does
6272 only good, but a few programs rely on the precise definition of
6273 IEEE floating point. Use `-ffloat-store' for such programs, after
6274 modifying them to store all pertinent intermediate computations
6278 Sets `-fno-math-errno', `-funsafe-math-optimizations',
6279 `-ffinite-math-only', `-fno-rounding-math', `-fno-signaling-nans'
6280 and `-fcx-limited-range'.
6282 This option causes the preprocessor macro `__FAST_MATH__' to be
6285 This option is not turned on by any `-O' option since it can
6286 result in incorrect output for programs which depend on an exact
6287 implementation of IEEE or ISO rules/specifications for math
6288 functions. It may, however, yield faster code for programs that do
6289 not require the guarantees of these specifications.
6292 Do not set ERRNO after calling math functions that are executed
6293 with a single instruction, e.g., sqrt. A program that relies on
6294 IEEE exceptions for math error handling may want to use this flag
6295 for speed while maintaining IEEE arithmetic compatibility.
6297 This option is not turned on by any `-O' option since it can
6298 result in incorrect output for programs which depend on an exact
6299 implementation of IEEE or ISO rules/specifications for math
6300 functions. It may, however, yield faster code for programs that do
6301 not require the guarantees of these specifications.
6303 The default is `-fmath-errno'.
6305 On Darwin systems, the math library never sets `errno'. There is
6306 therefore no reason for the compiler to consider the possibility
6307 that it might, and `-fno-math-errno' is the default.
6309 `-funsafe-math-optimizations'
6310 Allow optimizations for floating-point arithmetic that (a) assume
6311 that arguments and results are valid and (b) may violate IEEE or
6312 ANSI standards. When used at link-time, it may include libraries
6313 or startup files that change the default FPU control word or other
6314 similar optimizations.
6316 This option is not turned on by any `-O' option since it can
6317 result in incorrect output for programs which depend on an exact
6318 implementation of IEEE or ISO rules/specifications for math
6319 functions. It may, however, yield faster code for programs that do
6320 not require the guarantees of these specifications. Enables
6321 `-fno-signed-zeros', `-fno-trapping-math', `-fassociative-math'
6322 and `-freciprocal-math'.
6324 The default is `-fno-unsafe-math-optimizations'.
6326 `-fassociative-math'
6327 Allow re-association of operands in series of floating-point
6328 operations. This violates the ISO C and C++ language standard by
6329 possibly changing computation result. NOTE: re-ordering may
6330 change the sign of zero as well as ignore NaNs and inhibit or
6331 create underflow or overflow (and thus cannot be used on a code
6332 which relies on rounding behavior like `(x + 2**52) - 2**52)'.
6333 May also reorder floating-point comparisons and thus may not be
6334 used when ordered comparisons are required. This option requires
6335 that both `-fno-signed-zeros' and `-fno-trapping-math' be in
6336 effect. Moreover, it doesn't make much sense with
6339 The default is `-fno-associative-math'.
6342 Allow the reciprocal of a value to be used instead of dividing by
6343 the value if this enables optimizations. For example `x / y' can
6344 be replaced with `x * (1/y)' which is useful if `(1/y)' is subject
6345 to common subexpression elimination. Note that this loses
6346 precision and increases the number of flops operating on the value.
6348 The default is `-fno-reciprocal-math'.
6350 `-ffinite-math-only'
6351 Allow optimizations for floating-point arithmetic that assume that
6352 arguments and results are not NaNs or +-Infs.
6354 This option is not turned on by any `-O' option since it can
6355 result in incorrect output for programs which depend on an exact
6356 implementation of IEEE or ISO rules/specifications for math
6357 functions. It may, however, yield faster code for programs that do
6358 not require the guarantees of these specifications.
6360 The default is `-fno-finite-math-only'.
6363 Allow optimizations for floating point arithmetic that ignore the
6364 signedness of zero. IEEE arithmetic specifies the behavior of
6365 distinct +0.0 and -0.0 values, which then prohibits simplification
6366 of expressions such as x+0.0 or 0.0*x (even with
6367 `-ffinite-math-only'). This option implies that the sign of a
6368 zero result isn't significant.
6370 The default is `-fsigned-zeros'.
6372 `-fno-trapping-math'
6373 Compile code assuming that floating-point operations cannot
6374 generate user-visible traps. These traps include division by
6375 zero, overflow, underflow, inexact result and invalid operation.
6376 This option requires that `-fno-signaling-nans' be in effect.
6377 Setting this option may allow faster code if one relies on
6378 "non-stop" IEEE arithmetic, for example.
6380 This option should never be turned on by any `-O' option since it
6381 can result in incorrect output for programs which depend on an
6382 exact implementation of IEEE or ISO rules/specifications for math
6385 The default is `-ftrapping-math'.
6388 Disable transformations and optimizations that assume default
6389 floating point rounding behavior. This is round-to-zero for all
6390 floating point to integer conversions, and round-to-nearest for
6391 all other arithmetic truncations. This option should be specified
6392 for programs that change the FP rounding mode dynamically, or that
6393 may be executed with a non-default rounding mode. This option
6394 disables constant folding of floating point expressions at
6395 compile-time (which may be affected by rounding mode) and
6396 arithmetic transformations that are unsafe in the presence of
6397 sign-dependent rounding modes.
6399 The default is `-fno-rounding-math'.
6401 This option is experimental and does not currently guarantee to
6402 disable all GCC optimizations that are affected by rounding mode.
6403 Future versions of GCC may provide finer control of this setting
6404 using C99's `FENV_ACCESS' pragma. This command line option will
6405 be used to specify the default state for `FENV_ACCESS'.
6407 `-frtl-abstract-sequences'
6408 It is a size optimization method. This option is to find identical
6409 sequences of code, which can be turned into pseudo-procedures and
6410 then replace all occurrences with calls to the newly created
6411 subroutine. It is kind of an opposite of `-finline-functions'.
6412 This optimization runs at RTL level.
6415 Compile code assuming that IEEE signaling NaNs may generate
6416 user-visible traps during floating-point operations. Setting this
6417 option disables optimizations that may change the number of
6418 exceptions visible with signaling NaNs. This option implies
6421 This option causes the preprocessor macro `__SUPPORT_SNAN__' to be
6424 The default is `-fno-signaling-nans'.
6426 This option is experimental and does not currently guarantee to
6427 disable all GCC optimizations that affect signaling NaN behavior.
6429 `-fsingle-precision-constant'
6430 Treat floating point constant as single precision constant instead
6431 of implicitly converting it to double precision constant.
6433 `-fcx-limited-range'
6434 When enabled, this option states that a range reduction step is not
6435 needed when performing complex division. Also, there is no
6436 checking whether the result of a complex multiplication or
6437 division is `NaN + I*NaN', with an attempt to rescue the situation
6438 in that case. The default is `-fno-cx-limited-range', but is
6439 enabled by `-ffast-math'.
6441 This option controls the default setting of the ISO C99
6442 `CX_LIMITED_RANGE' pragma. Nevertheless, the option applies to
6445 `-fcx-fortran-rules'
6446 Complex multiplication and division follow Fortran rules. Range
6447 reduction is done as part of complex division, but there is no
6448 checking whether the result of a complex multiplication or
6449 division is `NaN + I*NaN', with an attempt to rescue the situation
6452 The default is `-fno-cx-fortran-rules'.
6455 The following options control optimizations that may improve
6456 performance, but are not enabled by any `-O' options. This section
6457 includes experimental options that may produce broken code.
6459 `-fbranch-probabilities'
6460 After running a program compiled with `-fprofile-arcs' (*note
6461 Options for Debugging Your Program or `gcc': Debugging Options.),
6462 you can compile it a second time using `-fbranch-probabilities',
6463 to improve optimizations based on the number of times each branch
6464 was taken. When the program compiled with `-fprofile-arcs' exits
6465 it saves arc execution counts to a file called `SOURCENAME.gcda'
6466 for each source file. The information in this data file is very
6467 dependent on the structure of the generated code, so you must use
6468 the same source code and the same optimization options for both
6471 With `-fbranch-probabilities', GCC puts a `REG_BR_PROB' note on
6472 each `JUMP_INSN' and `CALL_INSN'. These can be used to improve
6473 optimization. Currently, they are only used in one place: in
6474 `reorg.c', instead of guessing which path a branch is mostly to
6475 take, the `REG_BR_PROB' values are used to exactly determine which
6476 path is taken more often.
6479 If combined with `-fprofile-arcs', it adds code so that some data
6480 about values of expressions in the program is gathered.
6482 With `-fbranch-probabilities', it reads back the data gathered
6483 from profiling values of expressions and adds `REG_VALUE_PROFILE'
6484 notes to instructions for their later usage in optimizations.
6486 Enabled with `-fprofile-generate' and `-fprofile-use'.
6489 If combined with `-fprofile-arcs', it instructs the compiler to add
6490 a code to gather information about values of expressions.
6492 With `-fbranch-probabilities', it reads back the data gathered and
6493 actually performs the optimizations based on them. Currently the
6494 optimizations include specialization of division operation using
6495 the knowledge about the value of the denominator.
6497 `-frename-registers'
6498 Attempt to avoid false dependencies in scheduled code by making use
6499 of registers left over after register allocation. This
6500 optimization will most benefit processors with lots of registers.
6501 Depending on the debug information format adopted by the target,
6502 however, it can make debugging impossible, since variables will no
6503 longer stay in a "home register".
6505 Enabled by default with `-funroll-loops'.
6508 Perform tail duplication to enlarge superblock size. This
6509 transformation simplifies the control flow of the function
6510 allowing other optimizations to do better job.
6512 Enabled with `-fprofile-use'.
6515 Unroll loops whose number of iterations can be determined at
6516 compile time or upon entry to the loop. `-funroll-loops' implies
6517 `-frerun-cse-after-loop', `-fweb' and `-frename-registers'. It
6518 also turns on complete loop peeling (i.e. complete removal of
6519 loops with small constant number of iterations). This option
6520 makes code larger, and may or may not make it run faster.
6522 Enabled with `-fprofile-use'.
6524 `-funroll-all-loops'
6525 Unroll all loops, even if their number of iterations is uncertain
6526 when the loop is entered. This usually makes programs run more
6527 slowly. `-funroll-all-loops' implies the same options as
6531 Peels the loops for that there is enough information that they do
6532 not roll much (from profile feedback). It also turns on complete
6533 loop peeling (i.e. complete removal of loops with small constant
6534 number of iterations).
6536 Enabled with `-fprofile-use'.
6538 `-fmove-loop-invariants'
6539 Enables the loop invariant motion pass in the RTL loop optimizer.
6540 Enabled at level `-O1'
6543 Move branches with loop invariant conditions out of the loop, with
6544 duplicates of the loop on both branches (modified according to
6545 result of the condition).
6547 `-ffunction-sections'
6549 Place each function or data item into its own section in the output
6550 file if the target supports arbitrary sections. The name of the
6551 function or the name of the data item determines the section's name
6554 Use these options on systems where the linker can perform
6555 optimizations to improve locality of reference in the instruction
6556 space. Most systems using the ELF object format and SPARC
6557 processors running Solaris 2 have linkers with such optimizations.
6558 AIX may have these optimizations in the future.
6560 Only use these options when there are significant benefits from
6561 doing so. When you specify these options, the assembler and
6562 linker will create larger object and executable files and will
6563 also be slower. You will not be able to use `gprof' on all
6564 systems if you specify this option and you may have problems with
6565 debugging if you specify both this option and `-g'.
6567 `-fbranch-target-load-optimize'
6568 Perform branch target register load optimization before prologue /
6569 epilogue threading. The use of target registers can typically be
6570 exposed only during reload, thus hoisting loads out of loops and
6571 doing inter-block scheduling needs a separate optimization pass.
6573 `-fbranch-target-load-optimize2'
6574 Perform branch target register load optimization after prologue /
6577 `-fbtr-bb-exclusive'
6578 When performing branch target register load optimization, don't
6579 reuse branch target registers in within any basic block.
6582 Emit extra code to check for buffer overflows, such as stack
6583 smashing attacks. This is done by adding a guard variable to
6584 functions with vulnerable objects. This includes functions that
6585 call alloca, and functions with buffers larger than 8 bytes. The
6586 guards are initialized when a function is entered and then checked
6587 when the function exits. If a guard check fails, an error message
6588 is printed and the program exits.
6590 `-fstack-protector-all'
6591 Like `-fstack-protector' except that all functions are protected.
6594 Try to reduce the number of symbolic address calculations by using
6595 shared "anchor" symbols to address nearby objects. This
6596 transformation can help to reduce the number of GOT entries and
6597 GOT accesses on some targets.
6599 For example, the implementation of the following function `foo':
6602 int foo (void) { return a + b + c; }
6604 would usually calculate the addresses of all three variables, but
6605 if you compile it with `-fsection-anchors', it will access the
6606 variables from a common anchor point instead. The effect is
6607 similar to the following pseudocode (which isn't valid C):
6611 register int *xr = &x;
6612 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6615 Not all targets support this option.
6617 `--param NAME=VALUE'
6618 In some places, GCC uses various constants to control the amount of
6619 optimization that is done. For example, GCC will not inline
6620 functions that contain more that a certain number of instructions.
6621 You can control some of these constants on the command-line using
6622 the `--param' option.
6624 The names of specific parameters, and the meaning of the values,
6625 are tied to the internals of the compiler, and are subject to
6626 change without notice in future releases.
6628 In each case, the VALUE is an integer. The allowable choices for
6629 NAME are given in the following table:
6631 `sra-max-structure-size'
6632 The maximum structure size, in bytes, at which the scalar
6633 replacement of aggregates (SRA) optimization will perform
6634 block copies. The default value, 0, implies that GCC will
6635 select the most appropriate size itself.
6637 `sra-field-structure-ratio'
6638 The threshold ratio (as a percentage) between instantiated
6639 fields and the complete structure size. We say that if the
6640 ratio of the number of bytes in instantiated fields to the
6641 number of bytes in the complete structure exceeds this
6642 parameter, then block copies are not used. The default is 75.
6644 `struct-reorg-cold-struct-ratio'
6645 The threshold ratio (as a percentage) between a structure
6646 frequency and the frequency of the hottest structure in the
6647 program. This parameter is used by struct-reorg optimization
6648 enabled by `-fipa-struct-reorg'. We say that if the ratio of
6649 a structure frequency, calculated by profiling, to the
6650 hottest structure frequency in the program is less than this
6651 parameter, then structure reorganization is not applied to
6652 this structure. The default is 10.
6654 `predictable-branch-cost-outcome'
6655 When branch is predicted to be taken with probability lower
6656 than this threshold (in percent), then it is considered well
6657 predictable. The default is 10.
6659 `max-crossjump-edges'
6660 The maximum number of incoming edges to consider for
6661 crossjumping. The algorithm used by `-fcrossjumping' is
6662 O(N^2) in the number of edges incoming to each block.
6663 Increasing values mean more aggressive optimization, making
6664 the compile time increase with probably small improvement in
6667 `min-crossjump-insns'
6668 The minimum number of instructions which must be matched at
6669 the end of two blocks before crossjumping will be performed
6670 on them. This value is ignored in the case where all
6671 instructions in the block being crossjumped from are matched.
6672 The default value is 5.
6674 `max-grow-copy-bb-insns'
6675 The maximum code size expansion factor when copying basic
6676 blocks instead of jumping. The expansion is relative to a
6677 jump instruction. The default value is 8.
6679 `max-goto-duplication-insns'
6680 The maximum number of instructions to duplicate to a block
6681 that jumps to a computed goto. To avoid O(N^2) behavior in a
6682 number of passes, GCC factors computed gotos early in the
6683 compilation process, and unfactors them as late as possible.
6684 Only computed jumps at the end of a basic blocks with no more
6685 than max-goto-duplication-insns are unfactored. The default
6688 `max-delay-slot-insn-search'
6689 The maximum number of instructions to consider when looking
6690 for an instruction to fill a delay slot. If more than this
6691 arbitrary number of instructions is searched, the time
6692 savings from filling the delay slot will be minimal so stop
6693 searching. Increasing values mean more aggressive
6694 optimization, making the compile time increase with probably
6695 small improvement in executable run time.
6697 `max-delay-slot-live-search'
6698 When trying to fill delay slots, the maximum number of
6699 instructions to consider when searching for a block with
6700 valid live register information. Increasing this arbitrarily
6701 chosen value means more aggressive optimization, increasing
6702 the compile time. This parameter should be removed when the
6703 delay slot code is rewritten to maintain the control-flow
6707 The approximate maximum amount of memory that will be
6708 allocated in order to perform the global common subexpression
6709 elimination optimization. If more memory than specified is
6710 required, the optimization will not be done.
6713 The maximum number of passes of GCSE to run. The default is
6716 `max-pending-list-length'
6717 The maximum number of pending dependencies scheduling will
6718 allow before flushing the current state and starting over.
6719 Large functions with few branches or calls can create
6720 excessively large lists which needlessly consume memory and
6723 `max-inline-insns-single'
6724 Several parameters control the tree inliner used in gcc.
6725 This number sets the maximum number of instructions (counted
6726 in GCC's internal representation) in a single function that
6727 the tree inliner will consider for inlining. This only
6728 affects functions declared inline and methods implemented in
6729 a class declaration (C++). The default value is 450.
6731 `max-inline-insns-auto'
6732 When you use `-finline-functions' (included in `-O3'), a lot
6733 of functions that would otherwise not be considered for
6734 inlining by the compiler will be investigated. To those
6735 functions, a different (more restrictive) limit compared to
6736 functions declared inline can be applied. The default value
6739 `large-function-insns'
6740 The limit specifying really large functions. For functions
6741 larger than this limit after inlining, inlining is
6742 constrained by `--param large-function-growth'. This
6743 parameter is useful primarily to avoid extreme compilation
6744 time caused by non-linear algorithms used by the backend.
6745 The default value is 2700.
6747 `large-function-growth'
6748 Specifies maximal growth of large function caused by inlining
6749 in percents. The default value is 100 which limits large
6750 function growth to 2.0 times the original size.
6753 The limit specifying large translation unit. Growth caused
6754 by inlining of units larger than this limit is limited by
6755 `--param inline-unit-growth'. For small units this might be
6756 too tight (consider unit consisting of function A that is
6757 inline and B that just calls A three time. If B is small
6758 relative to A, the growth of unit is 300\% and yet such
6759 inlining is very sane. For very large units consisting of
6760 small inlineable functions however the overall unit growth
6761 limit is needed to avoid exponential explosion of code size.
6762 Thus for smaller units, the size is increased to `--param
6763 large-unit-insns' before applying `--param
6764 inline-unit-growth'. The default is 10000
6766 `inline-unit-growth'
6767 Specifies maximal overall growth of the compilation unit
6768 caused by inlining. The default value is 30 which limits
6769 unit growth to 1.3 times the original size.
6772 Specifies maximal overall growth of the compilation unit
6773 caused by interprocedural constant propagation. The default
6774 value is 10 which limits unit growth to 1.1 times the
6778 The limit specifying large stack frames. While inlining the
6779 algorithm is trying to not grow past this limit too much.
6780 Default value is 256 bytes.
6782 `large-stack-frame-growth'
6783 Specifies maximal growth of large stack frames caused by
6784 inlining in percents. The default value is 1000 which limits
6785 large stack frame growth to 11 times the original size.
6787 `max-inline-insns-recursive'
6788 `max-inline-insns-recursive-auto'
6789 Specifies maximum number of instructions out-of-line copy of
6790 self recursive inline function can grow into by performing
6793 For functions declared inline `--param
6794 max-inline-insns-recursive' is taken into account. For
6795 function not declared inline, recursive inlining happens only
6796 when `-finline-functions' (included in `-O3') is enabled and
6797 `--param max-inline-insns-recursive-auto' is used. The
6798 default value is 450.
6800 `max-inline-recursive-depth'
6801 `max-inline-recursive-depth-auto'
6802 Specifies maximum recursion depth used by the recursive
6805 For functions declared inline `--param
6806 max-inline-recursive-depth' is taken into account. For
6807 function not declared inline, recursive inlining happens only
6808 when `-finline-functions' (included in `-O3') is enabled and
6809 `--param max-inline-recursive-depth-auto' is used. The
6812 `min-inline-recursive-probability'
6813 Recursive inlining is profitable only for function having
6814 deep recursion in average and can hurt for function having
6815 little recursion depth by increasing the prologue size or
6816 complexity of function body to other optimizers.
6818 When profile feedback is available (see `-fprofile-generate')
6819 the actual recursion depth can be guessed from probability
6820 that function will recurse via given call expression. This
6821 parameter limits inlining only to call expression whose
6822 probability exceeds given threshold (in percents). The
6823 default value is 10.
6826 Specify cost of call instruction relative to simple
6827 arithmetics operations (having cost of 1). Increasing this
6828 cost disqualifies inlining of non-leaf functions and at the
6829 same time increases size of leaf function that is believed to
6830 reduce function size by being inlined. In effect it
6831 increases amount of inlining for code having large
6832 abstraction penalty (many functions that just pass the
6833 arguments to other functions) and decrease inlining for code
6834 with low abstraction penalty. The default value is 12.
6836 `min-vect-loop-bound'
6837 The minimum number of iterations under which a loop will not
6838 get vectorized when `-ftree-vectorize' is used. The number
6839 of iterations after vectorization needs to be greater than
6840 the value specified by this option to allow vectorization.
6841 The default value is 0.
6843 `max-unrolled-insns'
6844 The maximum number of instructions that a loop should have if
6845 that loop is unrolled, and if the loop is unrolled, it
6846 determines how many times the loop code is unrolled.
6848 `max-average-unrolled-insns'
6849 The maximum number of instructions biased by probabilities of
6850 their execution that a loop should have if that loop is
6851 unrolled, and if the loop is unrolled, it determines how many
6852 times the loop code is unrolled.
6855 The maximum number of unrollings of a single loop.
6858 The maximum number of instructions that a loop should have if
6859 that loop is peeled, and if the loop is peeled, it determines
6860 how many times the loop code is peeled.
6863 The maximum number of peelings of a single loop.
6865 `max-completely-peeled-insns'
6866 The maximum number of insns of a completely peeled loop.
6868 `max-completely-peel-times'
6869 The maximum number of iterations of a loop to be suitable for
6872 `max-unswitch-insns'
6873 The maximum number of insns of an unswitched loop.
6875 `max-unswitch-level'
6876 The maximum number of branches unswitched in a single loop.
6879 The minimum cost of an expensive expression in the loop
6882 `iv-consider-all-candidates-bound'
6883 Bound on number of candidates for induction variables below
6884 that all candidates are considered for each use in induction
6885 variable optimizations. Only the most relevant candidates
6886 are considered if there are more candidates, to avoid
6887 quadratic time complexity.
6889 `iv-max-considered-uses'
6890 The induction variable optimizations give up on loops that
6891 contain more induction variable uses.
6893 `iv-always-prune-cand-set-bound'
6894 If number of candidates in the set is smaller than this value,
6895 we always try to remove unnecessary ivs from the set during
6896 its optimization when a new iv is added to the set.
6898 `scev-max-expr-size'
6899 Bound on size of expressions used in the scalar evolutions
6900 analyzer. Large expressions slow the analyzer.
6903 The maximum number of variables in an Omega constraint system.
6904 The default value is 128.
6907 The maximum number of inequalities in an Omega constraint
6908 system. The default value is 256.
6911 The maximum number of equalities in an Omega constraint
6912 system. The default value is 128.
6914 `omega-max-wild-cards'
6915 The maximum number of wildcard variables that the Omega
6916 solver will be able to insert. The default value is 18.
6918 `omega-hash-table-size'
6919 The size of the hash table in the Omega solver. The default
6923 The maximal number of keys used by the Omega solver. The
6924 default value is 500.
6926 `omega-eliminate-redundant-constraints'
6927 When set to 1, use expensive methods to eliminate all
6928 redundant constraints. The default value is 0.
6930 `vect-max-version-for-alignment-checks'
6931 The maximum number of runtime checks that can be performed
6932 when doing loop versioning for alignment in the vectorizer.
6933 See option ftree-vect-loop-version for more information.
6935 `vect-max-version-for-alias-checks'
6936 The maximum number of runtime checks that can be performed
6937 when doing loop versioning for alias in the vectorizer. See
6938 option ftree-vect-loop-version for more information.
6940 `max-iterations-to-track'
6941 The maximum number of iterations of a loop the brute force
6942 algorithm for analysis of # of iterations of the loop tries
6945 `hot-bb-count-fraction'
6946 Select fraction of the maximal count of repetitions of basic
6947 block in program given basic block needs to have to be
6950 `hot-bb-frequency-fraction'
6951 Select fraction of the maximal frequency of executions of
6952 basic block in function given basic block needs to have to be
6955 `max-predicted-iterations'
6956 The maximum number of loop iterations we predict statically.
6957 This is useful in cases where function contain single loop
6958 with known bound and other loop with unknown. We predict the
6959 known number of iterations correctly, while the unknown
6960 number of iterations average to roughly 10. This means that
6961 the loop without bounds would appear artificially cold
6962 relative to the other one.
6965 Select fraction of the maximal frequency of executions of
6966 basic block in function given basic block will get aligned.
6968 `align-loop-iterations'
6969 A loop expected to iterate at lest the selected number of
6970 iterations will get aligned.
6972 `tracer-dynamic-coverage'
6973 `tracer-dynamic-coverage-feedback'
6974 This value is used to limit superblock formation once the
6975 given percentage of executed instructions is covered. This
6976 limits unnecessary code size expansion.
6978 The `tracer-dynamic-coverage-feedback' is used only when
6979 profile feedback is available. The real profiles (as opposed
6980 to statically estimated ones) are much less balanced allowing
6981 the threshold to be larger value.
6983 `tracer-max-code-growth'
6984 Stop tail duplication once code growth has reached given
6985 percentage. This is rather hokey argument, as most of the
6986 duplicates will be eliminated later in cross jumping, so it
6987 may be set to much higher values than is the desired code
6990 `tracer-min-branch-ratio'
6991 Stop reverse growth when the reverse probability of best edge
6992 is less than this threshold (in percent).
6994 `tracer-min-branch-ratio'
6995 `tracer-min-branch-ratio-feedback'
6996 Stop forward growth if the best edge do have probability
6997 lower than this threshold.
6999 Similarly to `tracer-dynamic-coverage' two values are
7000 present, one for compilation for profile feedback and one for
7001 compilation without. The value for compilation with profile
7002 feedback needs to be more conservative (higher) in order to
7003 make tracer effective.
7005 `max-cse-path-length'
7006 Maximum number of basic blocks on path that cse considers.
7010 The maximum instructions CSE process before flushing. The
7014 Maximum number of virtual operands per function allowed to
7015 represent aliases before triggering the alias partitioning
7016 heuristic. Alias partitioning reduces compile times and
7017 memory consumption needed for aliasing at the expense of
7018 precision loss in alias information. The default value for
7019 this parameter is 100 for -O1, 500 for -O2 and 1000 for -O3.
7021 Notice that if a function contains more memory statements
7022 than the value of this parameter, it is not really possible
7023 to achieve this reduction. In this case, the compiler will
7024 use the number of memory statements as the value for
7028 Average number of virtual operands per statement allowed to
7029 represent aliases before triggering the alias partitioning
7030 heuristic. This works in conjunction with
7031 `max-aliased-vops'. If a function contains more than
7032 `max-aliased-vops' virtual operators, then memory symbols
7033 will be grouped into memory partitions until either the total
7034 number of virtual operators is below `max-aliased-vops' or
7035 the average number of virtual operators per memory statement
7036 is below `avg-aliased-vops'. The default value for this
7037 parameter is 1 for -O1 and -O2, and 3 for -O3.
7040 GCC uses a garbage collector to manage its own memory
7041 allocation. This parameter specifies the minimum percentage
7042 by which the garbage collector's heap should be allowed to
7043 expand between collections. Tuning this may improve
7044 compilation speed; it has no effect on code generation.
7046 The default is 30% + 70% * (RAM/1GB) with an upper bound of
7047 100% when RAM >= 1GB. If `getrlimit' is available, the
7048 notion of "RAM" is the smallest of actual RAM and
7049 `RLIMIT_DATA' or `RLIMIT_AS'. If GCC is not able to
7050 calculate RAM on a particular platform, the lower bound of
7051 30% is used. Setting this parameter and `ggc-min-heapsize'
7052 to zero causes a full collection to occur at every
7053 opportunity. This is extremely slow, but can be useful for
7057 Minimum size of the garbage collector's heap before it begins
7058 bothering to collect garbage. The first collection occurs
7059 after the heap expands by `ggc-min-expand'% beyond
7060 `ggc-min-heapsize'. Again, tuning this may improve
7061 compilation speed, and has no effect on code generation.
7063 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit
7064 which tries to ensure that RLIMIT_DATA or RLIMIT_AS are not
7065 exceeded, but with a lower bound of 4096 (four megabytes) and
7066 an upper bound of 131072 (128 megabytes). If GCC is not able
7067 to calculate RAM on a particular platform, the lower bound is
7068 used. Setting this parameter very large effectively disables
7069 garbage collection. Setting this parameter and
7070 `ggc-min-expand' to zero causes a full collection to occur at
7073 `max-reload-search-insns'
7074 The maximum number of instruction reload should look backward
7075 for equivalent register. Increasing values mean more
7076 aggressive optimization, making the compile time increase
7077 with probably slightly better performance. The default value
7080 `max-cselib-memory-locations'
7081 The maximum number of memory locations cselib should take
7082 into account. Increasing values mean more aggressive
7083 optimization, making the compile time increase with probably
7084 slightly better performance. The default value is 500.
7086 `reorder-blocks-duplicate'
7087 `reorder-blocks-duplicate-feedback'
7088 Used by basic block reordering pass to decide whether to use
7089 unconditional branch or duplicate the code on its
7090 destination. Code is duplicated when its estimated size is
7091 smaller than this value multiplied by the estimated size of
7092 unconditional jump in the hot spots of the program.
7094 The `reorder-block-duplicate-feedback' is used only when
7095 profile feedback is available and may be set to higher values
7096 than `reorder-block-duplicate' since information about the
7097 hot spots is more accurate.
7099 `max-sched-ready-insns'
7100 The maximum number of instructions ready to be issued the
7101 scheduler should consider at any given time during the first
7102 scheduling pass. Increasing values mean more thorough
7103 searches, making the compilation time increase with probably
7104 little benefit. The default value is 100.
7106 `max-sched-region-blocks'
7107 The maximum number of blocks in a region to be considered for
7108 interblock scheduling. The default value is 10.
7110 `max-pipeline-region-blocks'
7111 The maximum number of blocks in a region to be considered for
7112 pipelining in the selective scheduler. The default value is
7115 `max-sched-region-insns'
7116 The maximum number of insns in a region to be considered for
7117 interblock scheduling. The default value is 100.
7119 `max-pipeline-region-insns'
7120 The maximum number of insns in a region to be considered for
7121 pipelining in the selective scheduler. The default value is
7125 The minimum probability (in percents) of reaching a source
7126 block for interblock speculative scheduling. The default
7129 `max-sched-extend-regions-iters'
7130 The maximum number of iterations through CFG to extend
7131 regions. 0 - disable region extension, N - do at most N
7132 iterations. The default value is 0.
7134 `max-sched-insn-conflict-delay'
7135 The maximum conflict delay for an insn to be considered for
7136 speculative motion. The default value is 3.
7138 `sched-spec-prob-cutoff'
7139 The minimal probability of speculation success (in percents),
7140 so that speculative insn will be scheduled. The default
7143 `sched-mem-true-dep-cost'
7144 Minimal distance (in CPU cycles) between store and load
7145 targeting same memory locations. The default value is 1.
7147 `selsched-max-lookahead'
7148 The maximum size of the lookahead window of selective
7149 scheduling. It is a depth of search for available
7150 instructions. The default value is 50.
7152 `selsched-max-sched-times'
7153 The maximum number of times that an instruction will be
7154 scheduled during selective scheduling. This is the limit on
7155 the number of iterations through which the instruction may be
7156 pipelined. The default value is 2.
7158 `selsched-max-insns-to-rename'
7159 The maximum number of best instructions in the ready list
7160 that are considered for renaming in the selective scheduler.
7161 The default value is 2.
7163 `max-last-value-rtl'
7164 The maximum size measured as number of RTLs that can be
7165 recorded in an expression in combiner for a pseudo register
7166 as last known value of that register. The default is 10000.
7168 `integer-share-limit'
7169 Small integer constants can use a shared data structure,
7170 reducing the compiler's memory usage and increasing its
7171 speed. This sets the maximum value of a shared integer
7172 constant. The default value is 256.
7174 `min-virtual-mappings'
7175 Specifies the minimum number of virtual mappings in the
7176 incremental SSA updater that should be registered to trigger
7177 the virtual mappings heuristic defined by
7178 virtual-mappings-ratio. The default value is 100.
7180 `virtual-mappings-ratio'
7181 If the number of virtual mappings is virtual-mappings-ratio
7182 bigger than the number of virtual symbols to be updated, then
7183 the incremental SSA updater switches to a full update for
7184 those symbols. The default ratio is 3.
7187 The minimum size of buffers (i.e. arrays) that will receive
7188 stack smashing protection when `-fstack-protection' is used.
7190 `max-jump-thread-duplication-stmts'
7191 Maximum number of statements allowed in a block that needs to
7192 be duplicated when threading jumps.
7194 `max-fields-for-field-sensitive'
7195 Maximum number of fields in a structure we will treat in a
7196 field sensitive manner during pointer analysis. The default
7197 is zero for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7200 Estimate on average number of instructions that are executed
7201 before prefetch finishes. The distance we prefetch ahead is
7202 proportional to this constant. Increasing this number may
7203 also lead to less streams being prefetched (see
7204 `simultaneous-prefetches').
7206 `simultaneous-prefetches'
7207 Maximum number of prefetches that can run at the same time.
7209 `l1-cache-line-size'
7210 The size of cache line in L1 cache, in bytes.
7213 The size of L1 cache, in kilobytes.
7216 The size of L2 cache, in kilobytes.
7218 `use-canonical-types'
7219 Whether the compiler should use the "canonical" type system.
7220 By default, this should always be 1, which uses a more
7221 efficient internal mechanism for comparing types in C++ and
7222 Objective-C++. However, if bugs in the canonical type system
7223 are causing compilation failures, set this value to 0 to
7224 disable canonical types.
7226 `switch-conversion-max-branch-ratio'
7227 Switch initialization conversion will refuse to create arrays
7228 that are bigger than `switch-conversion-max-branch-ratio'
7229 times the number of branches in the switch.
7231 `max-partial-antic-length'
7232 Maximum length of the partial antic set computed during the
7233 tree partial redundancy elimination optimization
7234 (`-ftree-pre') when optimizing at `-O3' and above. For some
7235 sorts of source code the enhanced partial redundancy
7236 elimination optimization can run away, consuming all of the
7237 memory available on the host machine. This parameter sets a
7238 limit on the length of the sets that are computed, which
7239 prevents the runaway behavior. Setting a value of 0 for this
7240 parameter will allow an unlimited set length.
7242 `sccvn-max-scc-size'
7243 Maximum size of a strongly connected component (SCC) during
7244 SCCVN processing. If this limit is hit, SCCVN processing for
7245 the whole function will not be done and optimizations
7246 depending on it will be disabled. The default maximum SCC
7250 IRA uses a regional register allocation by default. If a
7251 function contains loops more than number given by the
7252 parameter, only at most given number of the most frequently
7253 executed loops will form regions for the regional register
7254 allocation. The default value of the parameter is 100.
7256 `ira-max-conflict-table-size'
7257 Although IRA uses a sophisticated algorithm of compression
7258 conflict table, the table can be still big for huge
7259 functions. If the conflict table for a function could be
7260 more than size in MB given by the parameter, the conflict
7261 table is not built and faster, simpler, and lower quality
7262 register allocation algorithm will be used. The algorithm do
7263 not use pseudo-register conflicts. The default value of the
7266 `loop-invariant-max-bbs-in-loop'
7267 Loop invariant motion can be very expensive, both in compile
7268 time and in amount of needed compile time memory, with very
7269 large loops. Loops with more basic blocks than this
7270 parameter won't have loop invariant motion optimization
7271 performed on them. The default value of the parameter is
7272 1000 for -O1 and 10000 for -O2 and above.
7275 This is a switch to turn on live range shrinking optimization.
7278 This is used as a control knob to enable different
7279 transformations in the live range shrinking phase. Values of
7280 1, 2, and 4 are used to enable upward motion, downward
7281 motion, and tree reshaping transformations respectively. The
7282 values can be bitwise ORed.
7284 `reg-pressure-min-bb-factor'
7285 A performance tuning knob to control register pressure. When
7286 the size (in the number of gimple statements) of a basic
7287 block in a loop is larger than the threshold specified by
7288 this parameter multiplied by the number of available
7289 registers, live range shrinking optimization is enabled.
7291 `reg-pressure-min-tree'
7292 The minimal size (number of leaves) of a tree to be reshaped
7293 in the Live Range Shrinking optimization.
7297 File: gcc.info, Node: Preprocessor Options, Next: Assembler Options, Prev: Optimize Options, Up: Invoking GCC
7299 3.11 Options Controlling the Preprocessor
7300 =========================================
7302 These options control the C preprocessor, which is run on each C source
7303 file before actual compilation.
7305 If you use the `-E' option, nothing is done except preprocessing.
7306 Some of these options make sense only together with `-E' because they
7307 cause the preprocessor output to be unsuitable for actual compilation.
7309 You can use `-Wp,OPTION' to bypass the compiler driver and pass
7310 OPTION directly through to the preprocessor. If OPTION contains
7311 commas, it is split into multiple options at the commas. However,
7312 many options are modified, translated or interpreted by the
7313 compiler driver before being passed to the preprocessor, and `-Wp'
7314 forcibly bypasses this phase. The preprocessor's direct interface
7315 is undocumented and subject to change, so whenever possible you
7316 should avoid using `-Wp' and let the driver handle the options
7319 `-Xpreprocessor OPTION'
7320 Pass OPTION as an option to the preprocessor. You can use this to
7321 supply system-specific preprocessor options which GCC does not
7322 know how to recognize.
7324 If you want to pass an option that takes an argument, you must use
7325 `-Xpreprocessor' twice, once for the option and once for the
7329 Predefine NAME as a macro, with definition `1'.
7331 `-D NAME=DEFINITION'
7332 The contents of DEFINITION are tokenized and processed as if they
7333 appeared during translation phase three in a `#define' directive.
7334 In particular, the definition will be truncated by embedded
7337 If you are invoking the preprocessor from a shell or shell-like
7338 program you may need to use the shell's quoting syntax to protect
7339 characters such as spaces that have a meaning in the shell syntax.
7341 If you wish to define a function-like macro on the command line,
7342 write its argument list with surrounding parentheses before the
7343 equals sign (if any). Parentheses are meaningful to most shells,
7344 so you will need to quote the option. With `sh' and `csh',
7345 `-D'NAME(ARGS...)=DEFINITION'' works.
7347 `-D' and `-U' options are processed in the order they are given on
7348 the command line. All `-imacros FILE' and `-include FILE' options
7349 are processed after all `-D' and `-U' options.
7352 Cancel any previous definition of NAME, either built in or
7353 provided with a `-D' option.
7356 Do not predefine any system-specific or GCC-specific macros. The
7357 standard predefined macros remain defined.
7360 Add the directory DIR to the list of directories to be searched
7361 for header files. Directories named by `-I' are searched before
7362 the standard system include directories. If the directory DIR is
7363 a standard system include directory, the option is ignored to
7364 ensure that the default search order for system directories and
7365 the special treatment of system headers are not defeated . If DIR
7366 begins with `=', then the `=' will be replaced by the sysroot
7367 prefix; see `--sysroot' and `-isysroot'.
7370 Write output to FILE. This is the same as specifying FILE as the
7371 second non-option argument to `cpp'. `gcc' has a different
7372 interpretation of a second non-option argument, so you must use
7373 `-o' to specify the output file.
7376 Turns on all optional warnings which are desirable for normal code.
7377 At present this is `-Wcomment', `-Wtrigraphs', `-Wmultichar' and a
7378 warning about integer promotion causing a change of sign in `#if'
7379 expressions. Note that many of the preprocessor's warnings are on
7380 by default and have no options to control them.
7384 Warn whenever a comment-start sequence `/*' appears in a `/*'
7385 comment, or whenever a backslash-newline appears in a `//' comment.
7386 (Both forms have the same effect.)
7389 Most trigraphs in comments cannot affect the meaning of the
7390 program. However, a trigraph that would form an escaped newline
7391 (`??/' at the end of a line) can, by changing where the comment
7392 begins or ends. Therefore, only trigraphs that would form escaped
7393 newlines produce warnings inside a comment.
7395 This option is implied by `-Wall'. If `-Wall' is not given, this
7396 option is still enabled unless trigraphs are enabled. To get
7397 trigraph conversion without warnings, but get the other `-Wall'
7398 warnings, use `-trigraphs -Wall -Wno-trigraphs'.
7401 Warn about certain constructs that behave differently in
7402 traditional and ISO C. Also warn about ISO C constructs that have
7403 no traditional C equivalent, and problematic constructs which
7407 Warn whenever an identifier which is not a macro is encountered in
7408 an `#if' directive, outside of `defined'. Such identifiers are
7412 Warn about macros defined in the main file that are unused. A
7413 macro is "used" if it is expanded or tested for existence at least
7414 once. The preprocessor will also warn if the macro has not been
7415 used at the time it is redefined or undefined.
7417 Built-in macros, macros defined on the command line, and macros
7418 defined in include files are not warned about.
7420 _Note:_ If a macro is actually used, but only used in skipped
7421 conditional blocks, then CPP will report it as unused. To avoid
7422 the warning in such a case, you might improve the scope of the
7423 macro's definition by, for example, moving it into the first
7424 skipped block. Alternatively, you could provide a dummy use with
7427 #if defined the_macro_causing_the_warning
7431 Warn whenever an `#else' or an `#endif' are followed by text.
7432 This usually happens in code of the form
7440 The second and third `FOO' should be in comments, but often are not
7441 in older programs. This warning is on by default.
7444 Make all warnings into hard errors. Source code which triggers
7445 warnings will be rejected.
7448 Issue warnings for code in system headers. These are normally
7449 unhelpful in finding bugs in your own code, therefore suppressed.
7450 If you are responsible for the system library, you may want to see
7454 Suppress all warnings, including those which GNU CPP issues by
7458 Issue all the mandatory diagnostics listed in the C standard.
7459 Some of them are left out by default, since they trigger
7460 frequently on harmless code.
7463 Issue all the mandatory diagnostics, and make all mandatory
7464 diagnostics into errors. This includes mandatory diagnostics that
7465 GCC issues without `-pedantic' but treats as warnings.
7468 Instead of outputting the result of preprocessing, output a rule
7469 suitable for `make' describing the dependencies of the main source
7470 file. The preprocessor outputs one `make' rule containing the
7471 object file name for that source file, a colon, and the names of
7472 all the included files, including those coming from `-include' or
7473 `-imacros' command line options.
7475 Unless specified explicitly (with `-MT' or `-MQ'), the object file
7476 name consists of the name of the source file with any suffix
7477 replaced with object file suffix and with any leading directory
7478 parts removed. If there are many included files then the rule is
7479 split into several lines using `\'-newline. The rule has no
7482 This option does not suppress the preprocessor's debug output,
7483 such as `-dM'. To avoid mixing such debug output with the
7484 dependency rules you should explicitly specify the dependency
7485 output file with `-MF', or use an environment variable like
7486 `DEPENDENCIES_OUTPUT' (*note Environment Variables::). Debug
7487 output will still be sent to the regular output stream as normal.
7489 Passing `-M' to the driver implies `-E', and suppresses warnings
7490 with an implicit `-w'.
7493 Like `-M' but do not mention header files that are found in system
7494 header directories, nor header files that are included, directly
7495 or indirectly, from such a header.
7497 This implies that the choice of angle brackets or double quotes in
7498 an `#include' directive does not in itself determine whether that
7499 header will appear in `-MM' dependency output. This is a slight
7500 change in semantics from GCC versions 3.0 and earlier.
7503 When used with `-M' or `-MM', specifies a file to write the
7504 dependencies to. If no `-MF' switch is given the preprocessor
7505 sends the rules to the same place it would have sent preprocessed
7508 When used with the driver options `-MD' or `-MMD', `-MF' overrides
7509 the default dependency output file.
7512 In conjunction with an option such as `-M' requesting dependency
7513 generation, `-MG' assumes missing header files are generated files
7514 and adds them to the dependency list without raising an error.
7515 The dependency filename is taken directly from the `#include'
7516 directive without prepending any path. `-MG' also suppresses
7517 preprocessed output, as a missing header file renders this useless.
7519 This feature is used in automatic updating of makefiles.
7522 This option instructs CPP to add a phony target for each dependency
7523 other than the main file, causing each to depend on nothing. These
7524 dummy rules work around errors `make' gives if you remove header
7525 files without updating the `Makefile' to match.
7527 This is typical output:
7529 test.o: test.c test.h
7534 Change the target of the rule emitted by dependency generation. By
7535 default CPP takes the name of the main input file, deletes any
7536 directory components and any file suffix such as `.c', and appends
7537 the platform's usual object suffix. The result is the target.
7539 An `-MT' option will set the target to be exactly the string you
7540 specify. If you want multiple targets, you can specify them as a
7541 single argument to `-MT', or use multiple `-MT' options.
7543 For example, `-MT '$(objpfx)foo.o'' might give
7545 $(objpfx)foo.o: foo.c
7548 Same as `-MT', but it quotes any characters which are special to
7549 Make. `-MQ '$(objpfx)foo.o'' gives
7551 $$(objpfx)foo.o: foo.c
7553 The default target is automatically quoted, as if it were given
7557 `-MD' is equivalent to `-M -MF FILE', except that `-E' is not
7558 implied. The driver determines FILE based on whether an `-o'
7559 option is given. If it is, the driver uses its argument but with
7560 a suffix of `.d', otherwise it takes the name of the input file,
7561 removes any directory components and suffix, and applies a `.d'
7564 If `-MD' is used in conjunction with `-E', any `-o' switch is
7565 understood to specify the dependency output file (*note -MF:
7566 dashMF.), but if used without `-E', each `-o' is understood to
7567 specify a target object file.
7569 Since `-E' is not implied, `-MD' can be used to generate a
7570 dependency output file as a side-effect of the compilation process.
7573 Like `-MD' except mention only user header files, not system
7577 When using precompiled headers (*note Precompiled Headers::), this
7578 flag will cause the dependency-output flags to also list the files
7579 from the precompiled header's dependencies. If not specified only
7580 the precompiled header would be listed and not the files that were
7581 used to create it because those files are not consulted when a
7582 precompiled header is used.
7585 This option allows use of a precompiled header (*note Precompiled
7586 Headers::) together with `-E'. It inserts a special `#pragma',
7587 `#pragma GCC pch_preprocess "<filename>"' in the output to mark
7588 the place where the precompiled header was found, and its
7589 filename. When `-fpreprocessed' is in use, GCC recognizes this
7590 `#pragma' and loads the PCH.
7592 This option is off by default, because the resulting preprocessed
7593 output is only really suitable as input to GCC. It is switched on
7596 You should not write this `#pragma' in your own code, but it is
7597 safe to edit the filename if the PCH file is available in a
7598 different location. The filename may be absolute or it may be
7599 relative to GCC's current directory.
7604 `-x assembler-with-cpp'
7605 Specify the source language: C, C++, Objective-C, or assembly.
7606 This has nothing to do with standards conformance or extensions;
7607 it merely selects which base syntax to expect. If you give none
7608 of these options, cpp will deduce the language from the extension
7609 of the source file: `.c', `.cc', `.m', or `.S'. Some other common
7610 extensions for C++ and assembly are also recognized. If cpp does
7611 not recognize the extension, it will treat the file as C; this is
7612 the most generic mode.
7614 _Note:_ Previous versions of cpp accepted a `-lang' option which
7615 selected both the language and the standards conformance level.
7616 This option has been removed, because it conflicts with the `-l'
7621 Specify the standard to which the code should conform. Currently
7622 CPP knows about C and C++ standards; others may be added in the
7625 STANDARD may be one of:
7628 The ISO C standard from 1990. `c89' is the customary
7629 shorthand for this version of the standard.
7631 The `-ansi' option is equivalent to `-std=c89'.
7634 The 1990 C standard, as amended in 1994.
7640 The revised ISO C standard, published in December 1999.
7641 Before publication, this was known as C9X.
7644 The 1990 C standard plus GNU extensions. This is the default.
7648 The 1999 C standard plus GNU extensions.
7651 The 1998 ISO C++ standard plus amendments.
7654 The same as `-std=c++98' plus GNU extensions. This is the
7655 default for C++ code.
7658 Split the include path. Any directories specified with `-I'
7659 options before `-I-' are searched only for headers requested with
7660 `#include "FILE"'; they are not searched for `#include <FILE>'.
7661 If additional directories are specified with `-I' options after
7662 the `-I-', those directories are searched for all `#include'
7665 In addition, `-I-' inhibits the use of the directory of the current
7666 file directory as the first search directory for `#include "FILE"'.
7667 This option has been deprecated.
7670 Do not search the standard system directories for header files.
7671 Only the directories you have specified with `-I' options (and the
7672 directory of the current file, if appropriate) are searched.
7675 Do not search for header files in the C++-specific standard
7676 directories, but do still search the other standard directories.
7677 (This option is used when building the C++ library.)
7680 Process FILE as if `#include "file"' appeared as the first line of
7681 the primary source file. However, the first directory searched
7682 for FILE is the preprocessor's working directory _instead of_ the
7683 directory containing the main source file. If not found there, it
7684 is searched for in the remainder of the `#include "..."' search
7687 If multiple `-include' options are given, the files are included
7688 in the order they appear on the command line.
7691 Exactly like `-include', except that any output produced by
7692 scanning FILE is thrown away. Macros it defines remain defined.
7693 This allows you to acquire all the macros from a header without
7694 also processing its declarations.
7696 All files specified by `-imacros' are processed before all files
7697 specified by `-include'.
7700 Search DIR for header files, but do it _after_ all directories
7701 specified with `-I' and the standard system directories have been
7702 exhausted. DIR is treated as a system include directory. If DIR
7703 begins with `=', then the `=' will be replaced by the sysroot
7704 prefix; see `--sysroot' and `-isysroot'.
7707 Specify PREFIX as the prefix for subsequent `-iwithprefix'
7708 options. If the prefix represents a directory, you should include
7712 `-iwithprefixbefore DIR'
7713 Append DIR to the prefix specified previously with `-iprefix', and
7714 add the resulting directory to the include search path.
7715 `-iwithprefixbefore' puts it in the same place `-I' would;
7716 `-iwithprefix' puts it where `-idirafter' would.
7719 This option is like the `--sysroot' option, but applies only to
7720 header files. See the `--sysroot' option for more information.
7723 Use DIR as a subdirectory of the directory containing
7724 target-specific C++ headers.
7727 Search DIR for header files, after all directories specified by
7728 `-I' but before the standard system directories. Mark it as a
7729 system directory, so that it gets the same special treatment as is
7730 applied to the standard system directories. If DIR begins with
7731 `=', then the `=' will be replaced by the sysroot prefix; see
7732 `--sysroot' and `-isysroot'.
7735 Search DIR only for header files requested with `#include "FILE"';
7736 they are not searched for `#include <FILE>', before all
7737 directories specified by `-I' and before the standard system
7738 directories. If DIR begins with `=', then the `=' will be replaced
7739 by the sysroot prefix; see `--sysroot' and `-isysroot'.
7742 When preprocessing, handle directives, but do not expand macros.
7744 The option's behavior depends on the `-E' and `-fpreprocessed'
7747 With `-E', preprocessing is limited to the handling of directives
7748 such as `#define', `#ifdef', and `#error'. Other preprocessor
7749 operations, such as macro expansion and trigraph conversion are
7750 not performed. In addition, the `-dD' option is implicitly
7753 With `-fpreprocessed', predefinition of command line and most
7754 builtin macros is disabled. Macros such as `__LINE__', which are
7755 contextually dependent, are handled normally. This enables
7756 compilation of files previously preprocessed with `-E
7759 With both `-E' and `-fpreprocessed', the rules for
7760 `-fpreprocessed' take precedence. This enables full preprocessing
7761 of files previously preprocessed with `-E -fdirectives-only'.
7763 `-fdollars-in-identifiers'
7764 Accept `$' in identifiers.
7766 `-fextended-identifiers'
7767 Accept universal character names in identifiers. This option is
7768 experimental; in a future version of GCC, it will be enabled by
7769 default for C99 and C++.
7772 Indicate to the preprocessor that the input file has already been
7773 preprocessed. This suppresses things like macro expansion,
7774 trigraph conversion, escaped newline splicing, and processing of
7775 most directives. The preprocessor still recognizes and removes
7776 comments, so that you can pass a file preprocessed with `-C' to
7777 the compiler without problems. In this mode the integrated
7778 preprocessor is little more than a tokenizer for the front ends.
7780 `-fpreprocessed' is implicit if the input file has one of the
7781 extensions `.i', `.ii' or `.mi'. These are the extensions that
7782 GCC uses for preprocessed files created by `-save-temps'.
7785 Set the distance between tab stops. This helps the preprocessor
7786 report correct column numbers in warnings or errors, even if tabs
7787 appear on the line. If the value is less than 1 or greater than
7788 100, the option is ignored. The default is 8.
7790 `-fexec-charset=CHARSET'
7791 Set the execution character set, used for string and character
7792 constants. The default is UTF-8. CHARSET can be any encoding
7793 supported by the system's `iconv' library routine.
7795 `-fwide-exec-charset=CHARSET'
7796 Set the wide execution character set, used for wide string and
7797 character constants. The default is UTF-32 or UTF-16, whichever
7798 corresponds to the width of `wchar_t'. As with `-fexec-charset',
7799 CHARSET can be any encoding supported by the system's `iconv'
7800 library routine; however, you will have problems with encodings
7801 that do not fit exactly in `wchar_t'.
7803 `-finput-charset=CHARSET'
7804 Set the input character set, used for translation from the
7805 character set of the input file to the source character set used
7806 by GCC. If the locale does not specify, or GCC cannot get this
7807 information from the locale, the default is UTF-8. This can be
7808 overridden by either the locale or this command line option.
7809 Currently the command line option takes precedence if there's a
7810 conflict. CHARSET can be any encoding supported by the system's
7811 `iconv' library routine.
7813 `-fworking-directory'
7814 Enable generation of linemarkers in the preprocessor output that
7815 will let the compiler know the current working directory at the
7816 time of preprocessing. When this option is enabled, the
7817 preprocessor will emit, after the initial linemarker, a second
7818 linemarker with the current working directory followed by two
7819 slashes. GCC will use this directory, when it's present in the
7820 preprocessed input, as the directory emitted as the current
7821 working directory in some debugging information formats. This
7822 option is implicitly enabled if debugging information is enabled,
7823 but this can be inhibited with the negated form
7824 `-fno-working-directory'. If the `-P' flag is present in the
7825 command line, this option has no effect, since no `#line'
7826 directives are emitted whatsoever.
7829 Do not print column numbers in diagnostics. This may be necessary
7830 if diagnostics are being scanned by a program that does not
7831 understand the column numbers, such as `dejagnu'.
7833 `-A PREDICATE=ANSWER'
7834 Make an assertion with the predicate PREDICATE and answer ANSWER.
7835 This form is preferred to the older form `-A PREDICATE(ANSWER)',
7836 which is still supported, because it does not use shell special
7839 `-A -PREDICATE=ANSWER'
7840 Cancel an assertion with the predicate PREDICATE and answer ANSWER.
7843 CHARS is a sequence of one or more of the following characters,
7844 and must not be preceded by a space. Other characters are
7845 interpreted by the compiler proper, or reserved for future
7846 versions of GCC, and so are silently ignored. If you specify
7847 characters whose behavior conflicts, the result is undefined.
7850 Instead of the normal output, generate a list of `#define'
7851 directives for all the macros defined during the execution of
7852 the preprocessor, including predefined macros. This gives
7853 you a way of finding out what is predefined in your version
7854 of the preprocessor. Assuming you have no file `foo.h', the
7857 touch foo.h; cpp -dM foo.h
7859 will show all the predefined macros.
7861 If you use `-dM' without the `-E' option, `-dM' is
7862 interpreted as a synonym for `-fdump-rtl-mach'. *Note
7863 Debugging Options: (gcc)Debugging Options.
7866 Like `M' except in two respects: it does _not_ include the
7867 predefined macros, and it outputs _both_ the `#define'
7868 directives and the result of preprocessing. Both kinds of
7869 output go to the standard output file.
7872 Like `D', but emit only the macro names, not their expansions.
7875 Output `#include' directives in addition to the result of
7879 Like `D' except that only macros that are expanded, or whose
7880 definedness is tested in preprocessor directives, are output;
7881 the output is delayed until the use or test of the macro; and
7882 `#undef' directives are also output for macros tested but
7883 undefined at the time.
7886 Inhibit generation of linemarkers in the output from the
7887 preprocessor. This might be useful when running the preprocessor
7888 on something that is not C code, and will be sent to a program
7889 which might be confused by the linemarkers.
7892 Do not discard comments. All comments are passed through to the
7893 output file, except for comments in processed directives, which
7894 are deleted along with the directive.
7896 You should be prepared for side effects when using `-C'; it causes
7897 the preprocessor to treat comments as tokens in their own right.
7898 For example, comments appearing at the start of what would be a
7899 directive line have the effect of turning that line into an
7900 ordinary source line, since the first token on the line is no
7904 Do not discard comments, including during macro expansion. This is
7905 like `-C', except that comments contained within macros are also
7906 passed through to the output file where the macro is expanded.
7908 In addition to the side-effects of the `-C' option, the `-CC'
7909 option causes all C++-style comments inside a macro to be
7910 converted to C-style comments. This is to prevent later use of
7911 that macro from inadvertently commenting out the remainder of the
7914 The `-CC' option is generally used to support lint comments.
7917 Try to imitate the behavior of old-fashioned C preprocessors, as
7918 opposed to ISO C preprocessors.
7921 Process trigraph sequences. These are three-character sequences,
7922 all starting with `??', that are defined by ISO C to stand for
7923 single characters. For example, `??/' stands for `\', so `'??/n''
7924 is a character constant for a newline. By default, GCC ignores
7925 trigraphs, but in standard-conforming modes it converts them. See
7926 the `-std' and `-ansi' options.
7928 The nine trigraphs and their replacements are
7930 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??-
7931 Replacement: [ ] { } # \ ^ | ~
7934 Enable special code to work around file systems which only permit
7935 very short file names, such as MS-DOS.
7939 Print text describing all the command line options instead of
7940 preprocessing anything.
7943 Verbose mode. Print out GNU CPP's version number at the beginning
7944 of execution, and report the final form of the include path.
7947 Print the name of each header file used, in addition to other
7948 normal activities. Each name is indented to show how deep in the
7949 `#include' stack it is. Precompiled header files are also
7950 printed, even if they are found to be invalid; an invalid
7951 precompiled header file is printed with `...x' and a valid one
7956 Print out GNU CPP's version number. With one dash, proceed to
7957 preprocess as normal. With two dashes, exit immediately.
7960 File: gcc.info, Node: Assembler Options, Next: Link Options, Prev: Preprocessor Options, Up: Invoking GCC
7962 3.12 Passing Options to the Assembler
7963 =====================================
7965 You can pass options to the assembler.
7968 Pass OPTION as an option to the assembler. If OPTION contains
7969 commas, it is split into multiple options at the commas.
7971 `-Xassembler OPTION'
7972 Pass OPTION as an option to the assembler. You can use this to
7973 supply system-specific assembler options which GCC does not know
7976 If you want to pass an option that takes an argument, you must use
7977 `-Xassembler' twice, once for the option and once for the argument.
7981 File: gcc.info, Node: Link Options, Next: Directory Options, Prev: Assembler Options, Up: Invoking GCC
7983 3.13 Options for Linking
7984 ========================
7986 These options come into play when the compiler links object files into
7987 an executable output file. They are meaningless if the compiler is not
7991 A file name that does not end in a special recognized suffix is
7992 considered to name an object file or library. (Object files are
7993 distinguished from libraries by the linker according to the file
7994 contents.) If linking is done, these object files are used as
7995 input to the linker.
8000 If any of these options is used, then the linker is not run, and
8001 object file names should not be used as arguments. *Note Overall
8006 Search the library named LIBRARY when linking. (The second
8007 alternative with the library as a separate argument is only for
8008 POSIX compliance and is not recommended.)
8010 It makes a difference where in the command you write this option;
8011 the linker searches and processes libraries and object files in
8012 the order they are specified. Thus, `foo.o -lz bar.o' searches
8013 library `z' after file `foo.o' but before `bar.o'. If `bar.o'
8014 refers to functions in `z', those functions may not be loaded.
8016 The linker searches a standard list of directories for the library,
8017 which is actually a file named `libLIBRARY.a'. The linker then
8018 uses this file as if it had been specified precisely by name.
8020 The directories searched include several standard system
8021 directories plus any that you specify with `-L'.
8023 Normally the files found this way are library files--archive files
8024 whose members are object files. The linker handles an archive
8025 file by scanning through it for members which define symbols that
8026 have so far been referenced but not defined. But if the file that
8027 is found is an ordinary object file, it is linked in the usual
8028 fashion. The only difference between using an `-l' option and
8029 specifying a file name is that `-l' surrounds LIBRARY with `lib'
8030 and `.a' and searches several directories.
8033 You need this special case of the `-l' option in order to link an
8034 Objective-C or Objective-C++ program.
8037 Do not use the standard system startup files when linking. The
8038 standard system libraries are used normally, unless `-nostdlib' or
8039 `-nodefaultlibs' is used.
8042 Do not use the standard system libraries when linking. Only the
8043 libraries you specify will be passed to the linker. The standard
8044 startup files are used normally, unless `-nostartfiles' is used.
8045 The compiler may generate calls to `memcmp', `memset', `memcpy'
8046 and `memmove'. These entries are usually resolved by entries in
8047 libc. These entry points should be supplied through some other
8048 mechanism when this option is specified.
8051 Do not use the standard system startup files or libraries when
8052 linking. No startup files and only the libraries you specify will
8053 be passed to the linker. The compiler may generate calls to
8054 `memcmp', `memset', `memcpy' and `memmove'. These entries are
8055 usually resolved by entries in libc. These entry points should be
8056 supplied through some other mechanism when this option is
8059 One of the standard libraries bypassed by `-nostdlib' and
8060 `-nodefaultlibs' is `libgcc.a', a library of internal subroutines
8061 that GCC uses to overcome shortcomings of particular machines, or
8062 special needs for some languages. (*Note Interfacing to GCC
8063 Output: (gccint)Interface, for more discussion of `libgcc.a'.) In
8064 most cases, you need `libgcc.a' even when you want to avoid other
8065 standard libraries. In other words, when you specify `-nostdlib'
8066 or `-nodefaultlibs' you should usually specify `-lgcc' as well.
8067 This ensures that you have no unresolved references to internal GCC
8068 library subroutines. (For example, `__main', used to ensure C++
8069 constructors will be called; *note `collect2': (gccint)Collect2.)
8072 Produce a position independent executable on targets which support
8073 it. For predictable results, you must also specify the same set
8074 of options that were used to generate code (`-fpie', `-fPIE', or
8075 model suboptions) when you specify this option.
8078 Pass the flag `-export-dynamic' to the ELF linker, on targets that
8079 support it. This instructs the linker to add all symbols, not only
8080 used ones, to the dynamic symbol table. This option is needed for
8081 some uses of `dlopen' or to allow obtaining backtraces from within
8085 Remove all symbol table and relocation information from the
8089 On systems that support dynamic linking, this prevents linking
8090 with the shared libraries. On other systems, this option has no
8094 Produce a shared object which can then be linked with other
8095 objects to form an executable. Not all systems support this
8096 option. For predictable results, you must also specify the same
8097 set of options that were used to generate code (`-fpic', `-fPIC',
8098 or model suboptions) when you specify this option.(1)
8102 On systems that provide `libgcc' as a shared library, these options
8103 force the use of either the shared or static version respectively.
8104 If no shared version of `libgcc' was built when the compiler was
8105 configured, these options have no effect.
8107 There are several situations in which an application should use the
8108 shared `libgcc' instead of the static version. The most common of
8109 these is when the application wishes to throw and catch exceptions
8110 across different shared libraries. In that case, each of the
8111 libraries as well as the application itself should use the shared
8114 Therefore, the G++ and GCJ drivers automatically add
8115 `-shared-libgcc' whenever you build a shared library or a main
8116 executable, because C++ and Java programs typically use
8117 exceptions, so this is the right thing to do.
8119 If, instead, you use the GCC driver to create shared libraries,
8120 you may find that they will not always be linked with the shared
8121 `libgcc'. If GCC finds, at its configuration time, that you have
8122 a non-GNU linker or a GNU linker that does not support option
8123 `--eh-frame-hdr', it will link the shared version of `libgcc' into
8124 shared libraries by default. Otherwise, it will take advantage of
8125 the linker and optimize away the linking with the shared version
8126 of `libgcc', linking with the static version of libgcc by default.
8127 This allows exceptions to propagate through such shared libraries,
8128 without incurring relocation costs at library load time.
8130 However, if a library or main executable is supposed to throw or
8131 catch exceptions, you must link it using the G++ or GCJ driver, as
8132 appropriate for the languages used in the program, or using the
8133 option `-shared-libgcc', such that it is linked with the shared
8137 Bind references to global symbols when building a shared object.
8138 Warn about any unresolved references (unless overridden by the
8139 link editor option `-Xlinker -z -Xlinker defs'). Only a few
8140 systems support this option.
8143 Use SCRIPT as the linker script. This option is supported by most
8144 systems using the GNU linker. On some targets, such as bare-board
8145 targets without an operating system, the `-T' option may be
8146 required when linking to avoid references to undefined symbols.
8149 Pass OPTION as an option to the linker. You can use this to
8150 supply system-specific linker options which GCC does not know how
8153 If you want to pass an option that takes a separate argument, you
8154 must use `-Xlinker' twice, once for the option and once for the
8155 argument. For example, to pass `-assert definitions', you must
8156 write `-Xlinker -assert -Xlinker definitions'. It does not work
8157 to write `-Xlinker "-assert definitions"', because this passes the
8158 entire string as a single argument, which is not what the linker
8161 When using the GNU linker, it is usually more convenient to pass
8162 arguments to linker options using the `OPTION=VALUE' syntax than
8163 as separate arguments. For example, you can specify `-Xlinker
8164 -Map=output.map' rather than `-Xlinker -Map -Xlinker output.map'.
8165 Other linkers may not support this syntax for command-line options.
8168 Pass OPTION as an option to the linker. If OPTION contains
8169 commas, it is split into multiple options at the commas. You can
8170 use this syntax to pass an argument to the option. For example,
8171 `-Wl,-Map,output.map' passes `-Map output.map' to the linker.
8172 When using the GNU linker, you can also get the same effect with
8173 `-Wl,-Map=output.map'.
8176 Pretend the symbol SYMBOL is undefined, to force linking of
8177 library modules to define it. You can use `-u' multiple times with
8178 different symbols to force loading of additional library modules.
8180 ---------- Footnotes ----------
8182 (1) On some systems, `gcc -shared' needs to build supplementary stub
8183 code for constructors to work. On multi-libbed systems, `gcc -shared'
8184 must select the correct support libraries to link against. Failing to
8185 supply the correct flags may lead to subtle defects. Supplying them in
8186 cases where they are not necessary is innocuous.
8189 File: gcc.info, Node: Directory Options, Next: Spec Files, Prev: Link Options, Up: Invoking GCC
8191 3.14 Options for Directory Search
8192 =================================
8194 These options specify directories to search for header files, for
8195 libraries and for parts of the compiler:
8198 Add the directory DIR to the head of the list of directories to be
8199 searched for header files. This can be used to override a system
8200 header file, substituting your own version, since these
8201 directories are searched before the system header file
8202 directories. However, you should not use this option to add
8203 directories that contain vendor-supplied system header files (use
8204 `-isystem' for that). If you use more than one `-I' option, the
8205 directories are scanned in left-to-right order; the standard
8206 system directories come after.
8208 If a standard system include directory, or a directory specified
8209 with `-isystem', is also specified with `-I', the `-I' option will
8210 be ignored. The directory will still be searched but as a system
8211 directory at its normal position in the system include chain.
8212 This is to ensure that GCC's procedure to fix buggy system headers
8213 and the ordering for the include_next directive are not
8214 inadvertently changed. If you really need to change the search
8215 order for system directories, use the `-nostdinc' and/or
8219 Add the directory DIR to the head of the list of directories to be
8220 searched for header files only for the case of `#include "FILE"';
8221 they are not searched for `#include <FILE>', otherwise just like
8225 Add directory DIR to the list of directories to be searched for
8229 This option specifies where to find the executables, libraries,
8230 include files, and data files of the compiler itself.
8232 The compiler driver program runs one or more of the subprograms
8233 `cpp', `cc1', `as' and `ld'. It tries PREFIX as a prefix for each
8234 program it tries to run, both with and without `MACHINE/VERSION/'
8235 (*note Target Options::).
8237 For each subprogram to be run, the compiler driver first tries the
8238 `-B' prefix, if any. If that name is not found, or if `-B' was
8239 not specified, the driver tries two standard prefixes, which are
8240 `/usr/lib/gcc/' and `/usr/local/lib/gcc/'. If neither of those
8241 results in a file name that is found, the unmodified program name
8242 is searched for using the directories specified in your `PATH'
8243 environment variable.
8245 The compiler will check to see if the path provided by the `-B'
8246 refers to a directory, and if necessary it will add a directory
8247 separator character at the end of the path.
8249 `-B' prefixes that effectively specify directory names also apply
8250 to libraries in the linker, because the compiler translates these
8251 options into `-L' options for the linker. They also apply to
8252 includes files in the preprocessor, because the compiler
8253 translates these options into `-isystem' options for the
8254 preprocessor. In this case, the compiler appends `include' to the
8257 The run-time support file `libgcc.a' can also be searched for using
8258 the `-B' prefix, if needed. If it is not found there, the two
8259 standard prefixes above are tried, and that is all. The file is
8260 left out of the link if it is not found by those means.
8262 Another way to specify a prefix much like the `-B' prefix is to use
8263 the environment variable `GCC_EXEC_PREFIX'. *Note Environment
8266 As a special kludge, if the path provided by `-B' is
8267 `[dir/]stageN/', where N is a number in the range 0 to 9, then it
8268 will be replaced by `[dir/]include'. This is to help with
8269 boot-strapping the compiler.
8272 Process FILE after the compiler reads in the standard `specs'
8273 file, in order to override the defaults that the `gcc' driver
8274 program uses when determining what switches to pass to `cc1',
8275 `cc1plus', `as', `ld', etc. More than one `-specs=FILE' can be
8276 specified on the command line, and they are processed in order,
8280 Use DIR as the logical root directory for headers and libraries.
8281 For example, if the compiler would normally search for headers in
8282 `/usr/include' and libraries in `/usr/lib', it will instead search
8283 `DIR/usr/include' and `DIR/usr/lib'.
8285 If you use both this option and the `-isysroot' option, then the
8286 `--sysroot' option will apply to libraries, but the `-isysroot'
8287 option will apply to header files.
8289 The GNU linker (beginning with version 2.16) has the necessary
8290 support for this option. If your linker does not support this
8291 option, the header file aspect of `--sysroot' will still work, but
8292 the library aspect will not.
8295 This option has been deprecated. Please use `-iquote' instead for
8296 `-I' directories before the `-I-' and remove the `-I-'. Any
8297 directories you specify with `-I' options before the `-I-' option
8298 are searched only for the case of `#include "FILE"'; they are not
8299 searched for `#include <FILE>'.
8301 If additional directories are specified with `-I' options after
8302 the `-I-', these directories are searched for all `#include'
8303 directives. (Ordinarily _all_ `-I' directories are used this way.)
8305 In addition, the `-I-' option inhibits the use of the current
8306 directory (where the current input file came from) as the first
8307 search directory for `#include "FILE"'. There is no way to
8308 override this effect of `-I-'. With `-I.' you can specify
8309 searching the directory which was current when the compiler was
8310 invoked. That is not exactly the same as what the preprocessor
8311 does by default, but it is often satisfactory.
8313 `-I-' does not inhibit the use of the standard system directories
8314 for header files. Thus, `-I-' and `-nostdinc' are independent.
8317 File: gcc.info, Node: Spec Files, Next: Target Options, Prev: Directory Options, Up: Invoking GCC
8319 3.15 Specifying subprocesses and the switches to pass to them
8320 =============================================================
8322 `gcc' is a driver program. It performs its job by invoking a sequence
8323 of other programs to do the work of compiling, assembling and linking.
8324 GCC interprets its command-line parameters and uses these to deduce
8325 which programs it should invoke, and which command-line options it
8326 ought to place on their command lines. This behavior is controlled by
8327 "spec strings". In most cases there is one spec string for each
8328 program that GCC can invoke, but a few programs have multiple spec
8329 strings to control their behavior. The spec strings built into GCC can
8330 be overridden by using the `-specs=' command-line switch to specify a
8333 "Spec files" are plaintext files that are used to construct spec
8334 strings. They consist of a sequence of directives separated by blank
8335 lines. The type of directive is determined by the first non-whitespace
8336 character on the line and it can be one of the following:
8339 Issues a COMMAND to the spec file processor. The commands that can
8343 Search for FILE and insert its text at the current point in
8346 `%include_noerr <FILE>'
8347 Just like `%include', but do not generate an error message if
8348 the include file cannot be found.
8350 `%rename OLD_NAME NEW_NAME'
8351 Rename the spec string OLD_NAME to NEW_NAME.
8355 This tells the compiler to create, override or delete the named
8356 spec string. All lines after this directive up to the next
8357 directive or blank line are considered to be the text for the spec
8358 string. If this results in an empty string then the spec will be
8359 deleted. (Or, if the spec did not exist, then nothing will
8360 happened.) Otherwise, if the spec does not currently exist a new
8361 spec will be created. If the spec does exist then its contents
8362 will be overridden by the text of this directive, unless the first
8363 character of that text is the `+' character, in which case the
8364 text will be appended to the spec.
8367 Creates a new `[SUFFIX] spec' pair. All lines after this directive
8368 and up to the next directive or blank line are considered to make
8369 up the spec string for the indicated suffix. When the compiler
8370 encounters an input file with the named suffix, it will processes
8371 the spec string in order to work out how to compile that file.
8377 This says that any input file whose name ends in `.ZZ' should be
8378 passed to the program `z-compile', which should be invoked with the
8379 command-line switch `-input' and with the result of performing the
8380 `%i' substitution. (See below.)
8382 As an alternative to providing a spec string, the text that
8383 follows a suffix directive can be one of the following:
8386 This says that the suffix is an alias for a known LANGUAGE.
8387 This is similar to using the `-x' command-line switch to GCC
8388 to specify a language explicitly. For example:
8393 Says that .ZZ files are, in fact, C++ source files.
8396 This causes an error messages saying:
8398 NAME compiler not installed on this system.
8400 GCC already has an extensive list of suffixes built into it. This
8401 directive will add an entry to the end of the list of suffixes, but
8402 since the list is searched from the end backwards, it is
8403 effectively possible to override earlier entries using this
8407 GCC has the following spec strings built into it. Spec files can
8408 override these strings or create their own. Note that individual
8409 targets can also add their own spec strings to this list.
8411 asm Options to pass to the assembler
8412 asm_final Options to pass to the assembler post-processor
8413 cpp Options to pass to the C preprocessor
8414 cc1 Options to pass to the C compiler
8415 cc1plus Options to pass to the C++ compiler
8416 endfile Object files to include at the end of the link
8417 link Options to pass to the linker
8418 lib Libraries to include on the command line to the linker
8419 libgcc Decides which GCC support library to pass to the linker
8420 linker Sets the name of the linker
8421 predefines Defines to be passed to the C preprocessor
8422 signed_char Defines to pass to CPP to say whether `char' is signed
8424 startfile Object files to include at the start of the link
8426 Here is a small example of a spec file:
8431 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8433 This example renames the spec called `lib' to `old_lib' and then
8434 overrides the previous definition of `lib' with a new one. The new
8435 definition adds in some extra command-line options before including the
8436 text of the old definition.
8438 "Spec strings" are a list of command-line options to be passed to their
8439 corresponding program. In addition, the spec strings can contain
8440 `%'-prefixed sequences to substitute variable text or to conditionally
8441 insert text into the command line. Using these constructs it is
8442 possible to generate quite complex command lines.
8444 Here is a table of all defined `%'-sequences for spec strings. Note
8445 that spaces are not generated automatically around the results of
8446 expanding these sequences. Therefore you can concatenate them together
8447 or combine them with constant text in a single argument.
8450 Substitute one `%' into the program name or argument.
8453 Substitute the name of the input file being processed.
8456 Substitute the basename of the input file being processed. This
8457 is the substring up to (and not including) the last period and not
8458 including the directory.
8461 This is the same as `%b', but include the file suffix (text after
8465 Marks the argument containing or following the `%d' as a temporary
8466 file name, so that that file will be deleted if GCC exits
8467 successfully. Unlike `%g', this contributes no text to the
8471 Substitute a file name that has suffix SUFFIX and is chosen once
8472 per compilation, and mark the argument in the same way as `%d'.
8473 To reduce exposure to denial-of-service attacks, the file name is
8474 now chosen in a way that is hard to predict even when previously
8475 chosen file names are known. For example, `%g.s ... %g.o ... %g.s'
8476 might turn into `ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s'. SUFFIX
8477 matches the regexp `[.A-Za-z]*' or the special string `%O', which
8478 is treated exactly as if `%O' had been preprocessed. Previously,
8479 `%g' was simply substituted with a file name chosen once per
8480 compilation, without regard to any appended suffix (which was
8481 therefore treated just like ordinary text), making such attacks
8482 more likely to succeed.
8485 Like `%g', but generates a new temporary file name even if
8486 `%uSUFFIX' was already seen.
8489 Substitutes the last file name generated with `%uSUFFIX',
8490 generating a new one if there is no such last file name. In the
8491 absence of any `%uSUFFIX', this is just like `%gSUFFIX', except
8492 they don't share the same suffix _space_, so `%g.s ... %U.s ...
8493 %g.s ... %U.s' would involve the generation of two distinct file
8494 names, one for each `%g.s' and another for each `%U.s'.
8495 Previously, `%U' was simply substituted with a file name chosen
8496 for the previous `%u', without regard to any appended suffix.
8499 Substitutes the name of the `HOST_BIT_BUCKET', if any, and if it is
8500 writable, and if save-temps is off; otherwise, substitute the name
8501 of a temporary file, just like `%u'. This temporary file is not
8502 meant for communication between processes, but rather as a junk
8507 Like `%g', except if `-pipe' is in effect. In that case `%|'
8508 substitutes a single dash and `%m' substitutes nothing at all.
8509 These are the two most common ways to instruct a program that it
8510 should read from standard input or write to standard output. If
8511 you need something more elaborate you can use an `%{pipe:`X'}'
8512 construct: see for example `f/lang-specs.h'.
8515 Substitutes .SUFFIX for the suffixes of a matched switch's args
8516 when it is subsequently output with `%*'. SUFFIX is terminated by
8517 the next space or %.
8520 Marks the argument containing or following the `%w' as the
8521 designated output file of this compilation. This puts the argument
8522 into the sequence of arguments that `%o' will substitute later.
8525 Substitutes the names of all the output files, with spaces
8526 automatically placed around them. You should write spaces around
8527 the `%o' as well or the results are undefined. `%o' is for use in
8528 the specs for running the linker. Input files whose names have no
8529 recognized suffix are not compiled at all, but they are included
8530 among the output files, so they will be linked.
8533 Substitutes the suffix for object files. Note that this is
8534 handled specially when it immediately follows `%g, %u, or %U',
8535 because of the need for those to form complete file names. The
8536 handling is such that `%O' is treated exactly as if it had already
8537 been substituted, except that `%g, %u, and %U' do not currently
8538 support additional SUFFIX characters following `%O' as they would
8539 following, for example, `.o'.
8542 Substitutes the standard macro predefinitions for the current
8543 target machine. Use this when running `cpp'.
8546 Like `%p', but puts `__' before and after the name of each
8547 predefined macro, except for macros that start with `__' or with
8548 `_L', where L is an uppercase letter. This is for ISO C.
8551 Substitute any of `-iprefix' (made from `GCC_EXEC_PREFIX'),
8552 `-isysroot' (made from `TARGET_SYSTEM_ROOT'), `-isystem' (made
8553 from `COMPILER_PATH' and `-B' options) and `-imultilib' as
8557 Current argument is the name of a library or startup file of some
8558 sort. Search for that file in a standard list of directories and
8559 substitute the full name found.
8562 Print STR as an error message. STR is terminated by a newline.
8563 Use this when inconsistent options are detected.
8566 Substitute the contents of spec string NAME at this point.
8569 Like `%(...)' but put `__' around `-D' arguments.
8572 Accumulate an option for `%X'.
8575 Output the accumulated linker options specified by `-Wl' or a `%x'
8579 Output the accumulated assembler options specified by `-Wa'.
8582 Output the accumulated preprocessor options specified by `-Wp'.
8585 Process the `asm' spec. This is used to compute the switches to
8586 be passed to the assembler.
8589 Process the `asm_final' spec. This is a spec string for passing
8590 switches to an assembler post-processor, if such a program is
8594 Process the `link' spec. This is the spec for computing the
8595 command line passed to the linker. Typically it will make use of
8596 the `%L %G %S %D and %E' sequences.
8599 Dump out a `-L' option for each directory that GCC believes might
8600 contain startup files. If the target supports multilibs then the
8601 current multilib directory will be prepended to each of these
8605 Process the `lib' spec. This is a spec string for deciding which
8606 libraries should be included on the command line to the linker.
8609 Process the `libgcc' spec. This is a spec string for deciding
8610 which GCC support library should be included on the command line
8614 Process the `startfile' spec. This is a spec for deciding which
8615 object files should be the first ones passed to the linker.
8616 Typically this might be a file named `crt0.o'.
8619 Process the `endfile' spec. This is a spec string that specifies
8620 the last object files that will be passed to the linker.
8623 Process the `cpp' spec. This is used to construct the arguments
8624 to be passed to the C preprocessor.
8627 Process the `cc1' spec. This is used to construct the options to
8628 be passed to the actual C compiler (`cc1').
8631 Process the `cc1plus' spec. This is used to construct the options
8632 to be passed to the actual C++ compiler (`cc1plus').
8635 Substitute the variable part of a matched option. See below.
8636 Note that each comma in the substituted string is replaced by a
8640 Remove all occurrences of `-S' from the command line. Note--this
8641 command is position dependent. `%' commands in the spec string
8642 before this one will see `-S', `%' commands in the spec string
8643 after this one will not.
8646 Call the named function FUNCTION, passing it ARGS. ARGS is first
8647 processed as a nested spec string, then split into an argument
8648 vector in the usual fashion. The function returns a string which
8649 is processed as if it had appeared literally as part of the
8652 The following built-in spec functions are provided:
8655 The `getenv' spec function takes two arguments: an environment
8656 variable name and a string. If the environment variable is
8657 not defined, a fatal error is issued. Otherwise, the return
8658 value is the value of the environment variable concatenated
8659 with the string. For example, if `TOPDIR' is defined as
8660 `/path/to/top', then:
8662 %:getenv(TOPDIR /include)
8664 expands to `/path/to/top/include'.
8667 The `if-exists' spec function takes one argument, an absolute
8668 pathname to a file. If the file exists, `if-exists' returns
8669 the pathname. Here is a small example of its usage:
8672 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8675 The `if-exists-else' spec function is similar to the
8676 `if-exists' spec function, except that it takes two
8677 arguments. The first argument is an absolute pathname to a
8678 file. If the file exists, `if-exists-else' returns the
8679 pathname. If it does not exist, it returns the second
8680 argument. This way, `if-exists-else' can be used to select
8681 one file or another, based on the existence of the first.
8682 Here is a small example of its usage:
8685 crt0%O%s %:if-exists(crti%O%s) \
8686 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8689 The `replace-outfile' spec function takes two arguments. It
8690 looks for the first argument in the outfiles array and
8691 replaces it with the second argument. Here is a small
8692 example of its usage:
8694 %{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)}
8696 ``print-asm-header''
8697 The `print-asm-header' function takes no arguments and simply
8698 prints a banner like:
8703 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8705 It is used to separate compiler options from assembler options
8706 in the `--target-help' output.
8709 Substitutes the `-S' switch, if that switch was given to GCC. If
8710 that switch was not specified, this substitutes nothing. Note that
8711 the leading dash is omitted when specifying this option, and it is
8712 automatically inserted if the substitution is performed. Thus the
8713 spec string `%{foo}' would match the command-line option `-foo'
8714 and would output the command line option `-foo'.
8717 Like %{`S'} but mark last argument supplied within as a file to be
8721 Substitutes all the switches specified to GCC whose names start
8722 with `-S', but which also take an argument. This is used for
8723 switches like `-o', `-D', `-I', etc. GCC considers `-o foo' as
8724 being one switch whose names starts with `o'. %{o*} would
8725 substitute this text, including the space. Thus two arguments
8729 Like %{`S'*}, but preserve order of `S' and `T' options (the order
8730 of `S' and `T' in the spec is not significant). There can be any
8731 number of ampersand-separated variables; for each the wild card is
8732 optional. Useful for CPP as `%{D*&U*&A*}'.
8735 Substitutes `X', if the `-S' switch was given to GCC.
8738 Substitutes `X', if the `-S' switch was _not_ given to GCC.
8741 Substitutes `X' if one or more switches whose names start with
8742 `-S' are specified to GCC. Normally `X' is substituted only once,
8743 no matter how many such switches appeared. However, if `%*'
8744 appears somewhere in `X', then `X' will be substituted once for
8745 each matching switch, with the `%*' replaced by the part of that
8746 switch that matched the `*'.
8749 Substitutes `X', if processing a file with suffix `S'.
8752 Substitutes `X', if _not_ processing a file with suffix `S'.
8755 Substitutes `X', if processing a file for language `S'.
8758 Substitutes `X', if not processing a file for language `S'.
8761 Substitutes `X' if either `-S' or `-P' was given to GCC. This may
8762 be combined with `!', `.', `,', and `*' sequences as well,
8763 although they have a stronger binding than the `|'. If `%*'
8764 appears in `X', all of the alternatives must be starred, and only
8765 the first matching alternative is substituted.
8767 For example, a spec string like this:
8769 %{.c:-foo} %{!.c:-bar} %{.c|d:-baz} %{!.c|d:-boggle}
8771 will output the following command-line options from the following
8772 input command-line options:
8776 -d fred.c -foo -baz -boggle
8777 -d jim.d -bar -baz -boggle
8780 If `S' was given to GCC, substitutes `X'; else if `T' was given to
8781 GCC, substitutes `Y'; else substitutes `D'. There can be as many
8782 clauses as you need. This may be combined with `.', `,', `!',
8783 `|', and `*' as needed.
8786 The conditional text `X' in a %{`S':`X'} or similar construct may
8787 contain other nested `%' constructs or spaces, or even newlines. They
8788 are processed as usual, as described above. Trailing white space in
8789 `X' is ignored. White space may also appear anywhere on the left side
8790 of the colon in these constructs, except between `.' or `*' and the
8793 The `-O', `-f', `-m', and `-W' switches are handled specifically in
8794 these constructs. If another value of `-O' or the negated form of a
8795 `-f', `-m', or `-W' switch is found later in the command line, the
8796 earlier switch value is ignored, except with {`S'*} where `S' is just
8797 one letter, which passes all matching options.
8799 The character `|' at the beginning of the predicate text is used to
8800 indicate that a command should be piped to the following command, but
8801 only if `-pipe' is specified.
8803 It is built into GCC which switches take arguments and which do not.
8804 (You might think it would be useful to generalize this to allow each
8805 compiler's spec to say which switches take arguments. But this cannot
8806 be done in a consistent fashion. GCC cannot even decide which input
8807 files have been specified without knowing which switches take arguments,
8808 and it must know which input files to compile in order to tell which
8811 GCC also knows implicitly that arguments starting in `-l' are to be
8812 treated as compiler output files, and passed to the linker in their
8813 proper position among the other output files.
8816 File: gcc.info, Node: Target Options, Next: Submodel Options, Prev: Spec Files, Up: Invoking GCC
8818 3.16 Specifying Target Machine and Compiler Version
8819 ===================================================
8821 The usual way to run GCC is to run the executable called `gcc', or
8822 `<machine>-gcc' when cross-compiling, or `<machine>-gcc-<version>' to
8823 run a version other than the one that was installed last. Sometimes
8824 this is inconvenient, so GCC provides options that will switch to
8825 another cross-compiler or version.
8828 The argument MACHINE specifies the target machine for compilation.
8830 The value to use for MACHINE is the same as was specified as the
8831 machine type when configuring GCC as a cross-compiler. For
8832 example, if a cross-compiler was configured with `configure
8833 arm-elf', meaning to compile for an arm processor with elf
8834 binaries, then you would specify `-b arm-elf' to run that cross
8835 compiler. Because there are other options beginning with `-b', the
8836 configuration must contain a hyphen, or `-b' alone should be one
8837 argument followed by the configuration in the next argument.
8840 The argument VERSION specifies which version of GCC to run. This
8841 is useful when multiple versions are installed. For example,
8842 VERSION might be `4.0', meaning to run GCC version 4.0.
8844 The `-V' and `-b' options work by running the
8845 `<machine>-gcc-<version>' executable, so there's no real reason to use
8846 them if you can just run that directly.
8849 File: gcc.info, Node: Submodel Options, Next: Code Gen Options, Prev: Target Options, Up: Invoking GCC
8851 3.17 Hardware Models and Configurations
8852 =======================================
8854 Earlier we discussed the standard option `-b' which chooses among
8855 different installed compilers for completely different target machines,
8856 such as VAX vs. 68000 vs. 80386.
8858 In addition, each of these target machine types can have its own
8859 special options, starting with `-m', to choose among various hardware
8860 models or configurations--for example, 68010 vs 68020, floating
8861 coprocessor or none. A single installed version of the compiler can
8862 compile for any model or configuration, according to the options
8865 Some configurations of the compiler also support additional special
8866 options, usually for compatibility with other compilers on the same
8874 * Blackfin Options::
8878 * DEC Alpha Options::
8879 * DEC Alpha/VMS Options::
8882 * GNU/Linux Options::
8885 * i386 and x86-64 Options::
8886 * i386 and x86-64 Windows Options::
8897 * picoChip Options::
8899 * RS/6000 and PowerPC Options::
8900 * S/390 and zSeries Options::
8905 * System V Options::
8910 * Xstormy16 Options::
8915 File: gcc.info, Node: ARC Options, Next: ARM Options, Up: Submodel Options
8920 These options are defined for ARC implementations:
8923 Compile code for little endian mode. This is the default.
8926 Compile code for big endian mode.
8929 Prepend the name of the cpu to all public symbol names. In
8930 multiple-processor systems, there are many ARC variants with
8931 different instruction and register set characteristics. This flag
8932 prevents code compiled for one cpu to be linked with code compiled
8933 for another. No facility exists for handling variants that are
8934 "almost identical". This is an all or nothing option.
8937 Compile code for ARC variant CPU. Which variants are supported
8938 depend on the configuration. All variants support `-mcpu=base',
8939 this is the default.
8941 `-mtext=TEXT-SECTION'
8942 `-mdata=DATA-SECTION'
8943 `-mrodata=READONLY-DATA-SECTION'
8944 Put functions, data, and readonly data in TEXT-SECTION,
8945 DATA-SECTION, and READONLY-DATA-SECTION respectively by default.
8946 This can be overridden with the `section' attribute. *Note
8947 Variable Attributes::.
8949 `-mfix-cortex-m3-ldrd'
8950 Some Cortex-M3 cores can cause data corruption when `ldrd'
8951 instructions with overlapping destination and base registers are
8952 used. This option avoids generating these instructions. This
8953 option is enabled by default when `-mcpu=cortex-m3' is specified.
8957 File: gcc.info, Node: ARM Options, Next: AVR Options, Prev: ARC Options, Up: Submodel Options
8962 These `-m' options are defined for Advanced RISC Machines (ARM)
8966 Generate code for the specified ABI. Permissible values are:
8967 `apcs-gnu', `atpcs', `aapcs', `aapcs-linux' and `iwmmxt'.
8970 Generate a stack frame that is compliant with the ARM Procedure
8971 Call Standard for all functions, even if this is not strictly
8972 necessary for correct execution of the code. Specifying
8973 `-fomit-frame-pointer' with this option will cause the stack
8974 frames not to be generated for leaf functions. The default is
8978 This is a synonym for `-mapcs-frame'.
8981 Generate code which supports calling between the ARM and Thumb
8982 instruction sets. Without this option the two instruction sets
8983 cannot be reliably used inside one program. The default is
8984 `-mno-thumb-interwork', since slightly larger code is generated
8985 when `-mthumb-interwork' is specified.
8988 Prevent the reordering of instructions in the function prolog, or
8989 the merging of those instruction with the instructions in the
8990 function's body. This means that all functions will start with a
8991 recognizable set of instructions (or in fact one of a choice from
8992 a small set of different function prologues), and this information
8993 can be used to locate the start if functions inside an executable
8994 piece of code. The default is `-msched-prolog'.
8997 Specifies which floating-point ABI to use. Permissible values
8998 are: `soft', `softfp' and `hard'.
9000 Specifying `soft' causes GCC to generate output containing library
9001 calls for floating-point operations. `softfp' allows the
9002 generation of code using hardware floating-point instructions, but
9003 still uses the soft-float calling conventions. `hard' allows
9004 generation of floating-point instructions and uses FPU-specific
9005 calling conventions.
9007 Using `-mfloat-abi=hard' with VFP coprocessors is not supported.
9008 Use `-mfloat-abi=softfp' with the appropriate `-mfpu' option to
9009 allow the compiler to generate code that makes use of the hardware
9010 floating-point capabilities for these CPUs.
9012 The default depends on the specific target configuration. Note
9013 that the hard-float and soft-float ABIs are not link-compatible;
9014 you must compile your entire program with the same ABI, and link
9015 with a compatible set of libraries.
9018 Equivalent to `-mfloat-abi=hard'.
9021 Equivalent to `-mfloat-abi=soft'.
9024 Generate code for a processor running in little-endian mode. This
9025 is the default for all standard configurations.
9028 Generate code for a processor running in big-endian mode; the
9029 default is to compile code for a little-endian processor.
9031 `-mwords-little-endian'
9032 This option only applies when generating code for big-endian
9033 processors. Generate code for a little-endian word order but a
9034 big-endian byte order. That is, a byte order of the form
9035 `32107654'. Note: this option should only be used if you require
9036 compatibility with code for big-endian ARM processors generated by
9037 versions of the compiler prior to 2.8.
9040 This specifies the name of the target ARM processor. GCC uses
9041 this name to determine what kind of instructions it can emit when
9042 generating assembly code. Permissible names are: `arm2', `arm250',
9043 `arm3', `arm6', `arm60', `arm600', `arm610', `arm620', `arm7',
9044 `arm7m', `arm7d', `arm7dm', `arm7di', `arm7dmi', `arm70', `arm700',
9045 `arm700i', `arm710', `arm710c', `arm7100', `arm720', `arm7500',
9046 `arm7500fe', `arm7tdmi', `arm7tdmi-s', `arm710t', `arm720t',
9047 `arm740t', `strongarm', `strongarm110', `strongarm1100',
9048 `strongarm1110', `arm8', `arm810', `arm9', `arm9e', `arm920',
9049 `arm920t', `arm922t', `arm946e-s', `arm966e-s', `arm968e-s',
9050 `arm926ej-s', `arm940t', `arm9tdmi', `arm10tdmi', `arm1020t',
9051 `arm1026ej-s', `arm10e', `arm1020e', `arm1022e', `arm1136j-s',
9052 `arm1136jf-s', `mpcore', `mpcorenovfp', `arm1156t2-s',
9053 `arm1176jz-s', `arm1176jzf-s', `cortex-a8', `cortex-a9',
9054 `cortex-r4', `cortex-r4f', `cortex-m3', `cortex-m1', `xscale',
9055 `iwmmxt', `iwmmxt2', `ep9312'.
9058 This option is very similar to the `-mcpu=' option, except that
9059 instead of specifying the actual target processor type, and hence
9060 restricting which instructions can be used, it specifies that GCC
9061 should tune the performance of the code as if the target were of
9062 the type specified in this option, but still choosing the
9063 instructions that it will generate based on the cpu specified by a
9064 `-mcpu=' option. For some ARM implementations better performance
9065 can be obtained by using this option.
9068 This specifies the name of the target ARM architecture. GCC uses
9069 this name to determine what kind of instructions it can emit when
9070 generating assembly code. This option can be used in conjunction
9071 with or instead of the `-mcpu=' option. Permissible names are:
9072 `armv2', `armv2a', `armv3', `armv3m', `armv4', `armv4t', `armv5',
9073 `armv5t', `armv5e', `armv5te', `armv6', `armv6j', `armv6t2',
9074 `armv6z', `armv6zk', `armv6-m', `armv7', `armv7-a', `armv7-r',
9075 `armv7-m', `iwmmxt', `iwmmxt2', `ep9312'.
9080 This specifies what floating point hardware (or hardware
9081 emulation) is available on the target. Permissible names are:
9082 `fpa', `fpe2', `fpe3', `maverick', `vfp', `vfpv3', `vfpv3-d16' and
9083 `neon'. `-mfp' and `-mfpe' are synonyms for `-mfpu'=`fpe'NUMBER,
9084 for compatibility with older versions of GCC.
9086 If `-msoft-float' is specified this specifies the format of
9087 floating point values.
9089 `-mstructure-size-boundary=N'
9090 The size of all structures and unions will be rounded up to a
9091 multiple of the number of bits set by this option. Permissible
9092 values are 8, 32 and 64. The default value varies for different
9093 toolchains. For the COFF targeted toolchain the default value is
9094 8. A value of 64 is only allowed if the underlying ABI supports
9097 Specifying the larger number can produce faster, more efficient
9098 code, but can also increase the size of the program. Different
9099 values are potentially incompatible. Code compiled with one value
9100 cannot necessarily expect to work with code or libraries compiled
9101 with another value, if they exchange information using structures
9104 `-mabort-on-noreturn'
9105 Generate a call to the function `abort' at the end of a `noreturn'
9106 function. It will be executed if the function tries to return.
9110 Tells the compiler to perform function calls by first loading the
9111 address of the function into a register and then performing a
9112 subroutine call on this register. This switch is needed if the
9113 target function will lie outside of the 64 megabyte addressing
9114 range of the offset based version of subroutine call instruction.
9116 Even if this switch is enabled, not all function calls will be
9117 turned into long calls. The heuristic is that static functions,
9118 functions which have the `short-call' attribute, functions that
9119 are inside the scope of a `#pragma no_long_calls' directive and
9120 functions whose definitions have already been compiled within the
9121 current compilation unit, will not be turned into long calls. The
9122 exception to this rule is that weak function definitions,
9123 functions with the `long-call' attribute or the `section'
9124 attribute, and functions that are within the scope of a `#pragma
9125 long_calls' directive, will always be turned into long calls.
9127 This feature is not enabled by default. Specifying
9128 `-mno-long-calls' will restore the default behavior, as will
9129 placing the function calls within the scope of a `#pragma
9130 long_calls_off' directive. Note these switches have no effect on
9131 how the compiler generates code to handle function calls via
9135 Treat the register used for PIC addressing as read-only, rather
9136 than loading it in the prologue for each function. The run-time
9137 system is responsible for initializing this register with an
9138 appropriate value before execution begins.
9140 `-mpic-register=REG'
9141 Specify the register to be used for PIC addressing. The default
9142 is R10 unless stack-checking is enabled, when R9 is used.
9144 `-mcirrus-fix-invalid-insns'
9145 Insert NOPs into the instruction stream to in order to work around
9146 problems with invalid Maverick instruction combinations. This
9147 option is only valid if the `-mcpu=ep9312' option has been used to
9148 enable generation of instructions for the Cirrus Maverick floating
9149 point co-processor. This option is not enabled by default, since
9150 the problem is only present in older Maverick implementations.
9151 The default can be re-enabled by use of the
9152 `-mno-cirrus-fix-invalid-insns' switch.
9154 `-mpoke-function-name'
9155 Write the name of each function into the text section, directly
9156 preceding the function prologue. The generated code is similar to
9160 .ascii "arm_poke_function_name", 0
9163 .word 0xff000000 + (t1 - t0)
9164 arm_poke_function_name
9166 stmfd sp!, {fp, ip, lr, pc}
9169 When performing a stack backtrace, code can inspect the value of
9170 `pc' stored at `fp + 0'. If the trace function then looks at
9171 location `pc - 12' and the top 8 bits are set, then we know that
9172 there is a function name embedded immediately preceding this
9173 location and has length `((pc[-3]) & 0xff000000)'.
9176 Generate code for the Thumb instruction set. The default is to
9177 use the 32-bit ARM instruction set. This option automatically
9178 enables either 16-bit Thumb-1 or mixed 16/32-bit Thumb-2
9179 instructions based on the `-mcpu=NAME' and `-march=NAME' options.
9182 Generate a stack frame that is compliant with the Thumb Procedure
9183 Call Standard for all non-leaf functions. (A leaf function is one
9184 that does not call any other functions.) The default is
9188 Generate a stack frame that is compliant with the Thumb Procedure
9189 Call Standard for all leaf functions. (A leaf function is one
9190 that does not call any other functions.) The default is
9191 `-mno-apcs-leaf-frame'.
9193 `-mcallee-super-interworking'
9194 Gives all externally visible functions in the file being compiled
9195 an ARM instruction set header which switches to Thumb mode before
9196 executing the rest of the function. This allows these functions
9197 to be called from non-interworking code.
9199 `-mcaller-super-interworking'
9200 Allows calls via function pointers (including virtual functions) to
9201 execute correctly regardless of whether the target code has been
9202 compiled for interworking or not. There is a small overhead in
9203 the cost of executing a function pointer if this option is enabled.
9206 Specify the access model for the thread local storage pointer.
9207 The valid models are `soft', which generates calls to
9208 `__aeabi_read_tp', `cp15', which fetches the thread pointer from
9209 `cp15' directly (supported in the arm6k architecture), and `auto',
9210 which uses the best available method for the selected processor.
9211 The default setting is `auto'.
9213 `-mword-relocations'
9214 Only generate absolute relocations on word sized values (i.e.
9215 R_ARM_ABS32). This is enabled by default on targets (uClinux,
9216 SymbianOS) where the runtime loader imposes this restriction, and
9217 when `-fpic' or `-fPIC' is specified.
9220 Enable Android specific compilier options.
9222 If this option is used, a preprocessor macro `__ANDROID__' is
9223 defined and has the value 1 during compilation. The option also
9224 implies C/C++ options `-fno-exceptions' `-fpic' `-mthumb-interwork'
9225 `-fno-short-enums' and C++ option `-fno-rtti'. These implied
9226 options can be overridden. For example RTTI in C++ code can still
9227 be enabled with -frtti even when -mandroid is also used.
9229 Linking options depend on whether a static executable, a dynamic
9230 executable or a shared library is built. When `-static' is given,
9231 `-mandroid' implies linking flag `-Bstatic', start file
9232 `crtbegin_static.o' and end file `crtend_android.o'.
9234 When `-shared' is given, `-mandroid' implies the linking flag
9235 `-Bsymbolic' and no start and end files.
9237 When none of `-static' and `-shared' is given, `-mandroid' implies
9238 linking flags `-Bdynamic -dynamic-linker /system/bin/linker',
9239 start file `crtbegin_dynamic.o' and end file `crtend_android.o'.
9240 The dynamic linker used can be overriden by another
9241 `-dynamic-linker' in command line.
9243 The linking option `-ldl' is also added if `-static' is not given.
9245 If more than one of `-dynamic', `-static' and `-shared' are given,
9246 behaviour of `-mandroid' is undefined.
9250 File: gcc.info, Node: AVR Options, Next: Blackfin Options, Prev: ARM Options, Up: Submodel Options
9255 These options are defined for AVR implementations:
9258 Specify ATMEL AVR instruction set or MCU type.
9260 Instruction set avr1 is for the minimal AVR core, not supported by
9261 the C compiler, only for assembler programs (MCU types: at90s1200,
9262 attiny10, attiny11, attiny12, attiny15, attiny28).
9264 Instruction set avr2 (default) is for the classic AVR core with up
9265 to 8K program memory space (MCU types: at90s2313, at90s2323,
9266 attiny22, at90s2333, at90s2343, at90s4414, at90s4433, at90s4434,
9267 at90s8515, at90c8534, at90s8535).
9269 Instruction set avr3 is for the classic AVR core with up to 128K
9270 program memory space (MCU types: atmega103, atmega603, at43usb320,
9273 Instruction set avr4 is for the enhanced AVR core with up to 8K
9274 program memory space (MCU types: atmega8, atmega83, atmega85).
9276 Instruction set avr5 is for the enhanced AVR core with up to 128K
9277 program memory space (MCU types: atmega16, atmega161, atmega163,
9278 atmega32, atmega323, atmega64, atmega128, at43usb355, at94k).
9281 Output instruction sizes to the asm file.
9284 Specify the initial stack address, which may be a symbol or
9285 numeric value, `__stack' is the default.
9288 Generated code is not compatible with hardware interrupts. Code
9289 size will be smaller.
9292 Functions prologues/epilogues expanded as call to appropriate
9293 subroutines. Code size will be smaller.
9296 Do not generate tablejump insns which sometimes increase code size.
9297 The option is now deprecated in favor of the equivalent
9301 Change only the low 8 bits of the stack pointer.
9304 Assume int to be 8 bit integer. This affects the sizes of all
9305 types: A char will be 1 byte, an int will be 1 byte, an long will
9306 be 2 bytes and long long will be 4 bytes. Please note that this
9307 option does not comply to the C standards, but it will provide you
9308 with smaller code size.
9311 File: gcc.info, Node: Blackfin Options, Next: CRIS Options, Prev: AVR Options, Up: Submodel Options
9313 3.17.4 Blackfin Options
9314 -----------------------
9316 `-mcpu=CPU[-SIREVISION]'
9317 Specifies the name of the target Blackfin processor. Currently,
9318 CPU can be one of `bf512', `bf514', `bf516', `bf518', `bf522',
9319 `bf523', `bf524', `bf525', `bf526', `bf527', `bf531', `bf532',
9320 `bf533', `bf534', `bf536', `bf537', `bf538', `bf539', `bf542',
9321 `bf544', `bf547', `bf548', `bf549', `bf561'. The optional
9322 SIREVISION specifies the silicon revision of the target Blackfin
9323 processor. Any workarounds available for the targeted silicon
9324 revision will be enabled. If SIREVISION is `none', no workarounds
9325 are enabled. If SIREVISION is `any', all workarounds for the
9326 targeted processor will be enabled. The `__SILICON_REVISION__'
9327 macro is defined to two hexadecimal digits representing the major
9328 and minor numbers in the silicon revision. If SIREVISION is
9329 `none', the `__SILICON_REVISION__' is not defined. If SIREVISION
9330 is `any', the `__SILICON_REVISION__' is defined to be `0xffff'.
9331 If this optional SIREVISION is not used, GCC assumes the latest
9332 known silicon revision of the targeted Blackfin processor.
9334 Support for `bf561' is incomplete. For `bf561', Only the
9335 processor macro is defined. Without this option, `bf532' is used
9336 as the processor by default. The corresponding predefined
9337 processor macros for CPU is to be defined. And for `bfin-elf'
9338 toolchain, this causes the hardware BSP provided by libgloss to be
9339 linked in if `-msim' is not given.
9342 Specifies that the program will be run on the simulator. This
9343 causes the simulator BSP provided by libgloss to be linked in.
9344 This option has effect only for `bfin-elf' toolchain. Certain
9345 other options, such as `-mid-shared-library' and `-mfdpic', imply
9348 `-momit-leaf-frame-pointer'
9349 Don't keep the frame pointer in a register for leaf functions.
9350 This avoids the instructions to save, set up and restore frame
9351 pointers and makes an extra register available in leaf functions.
9352 The option `-fomit-frame-pointer' removes the frame pointer for
9353 all functions which might make debugging harder.
9356 When enabled, the compiler will ensure that the generated code
9357 does not contain speculative loads after jump instructions. If
9358 this option is used, `__WORKAROUND_SPECULATIVE_LOADS' is defined.
9360 `-mno-specld-anomaly'
9361 Don't generate extra code to prevent speculative loads from
9365 When enabled, the compiler will ensure that the generated code
9366 does not contain CSYNC or SSYNC instructions too soon after
9367 conditional branches. If this option is used,
9368 `__WORKAROUND_SPECULATIVE_SYNCS' is defined.
9370 `-mno-csync-anomaly'
9371 Don't generate extra code to prevent CSYNC or SSYNC instructions
9372 from occurring too soon after a conditional branch.
9375 When enabled, the compiler is free to take advantage of the
9376 knowledge that the entire program fits into the low 64k of memory.
9379 Assume that the program is arbitrarily large. This is the default.
9382 Do stack checking using information placed into L1 scratchpad
9383 memory by the uClinux kernel.
9385 `-mid-shared-library'
9386 Generate code that supports shared libraries via the library ID
9387 method. This allows for execute in place and shared libraries in
9388 an environment without virtual memory management. This option
9389 implies `-fPIC'. With a `bfin-elf' target, this option implies
9392 `-mno-id-shared-library'
9393 Generate code that doesn't assume ID based shared libraries are
9394 being used. This is the default.
9396 `-mleaf-id-shared-library'
9397 Generate code that supports shared libraries via the library ID
9398 method, but assumes that this library or executable won't link
9399 against any other ID shared libraries. That allows the compiler
9400 to use faster code for jumps and calls.
9402 `-mno-leaf-id-shared-library'
9403 Do not assume that the code being compiled won't link against any
9404 ID shared libraries. Slower code will be generated for jump and
9407 `-mshared-library-id=n'
9408 Specified the identification number of the ID based shared library
9409 being compiled. Specifying a value of 0 will generate more
9410 compact code, specifying other values will force the allocation of
9411 that number to the current library but is no more space or time
9412 efficient than omitting this option.
9415 Generate code that allows the data segment to be located in a
9416 different area of memory from the text segment. This allows for
9417 execute in place in an environment without virtual memory
9418 management by eliminating relocations against the text section.
9421 Generate code that assumes that the data segment follows the text
9422 segment. This is the default.
9426 Tells the compiler to perform function calls by first loading the
9427 address of the function into a register and then performing a
9428 subroutine call on this register. This switch is needed if the
9429 target function will lie outside of the 24 bit addressing range of
9430 the offset based version of subroutine call instruction.
9432 This feature is not enabled by default. Specifying
9433 `-mno-long-calls' will restore the default behavior. Note these
9434 switches have no effect on how the compiler generates code to
9435 handle function calls via function pointers.
9438 Link with the fast floating-point library. This library relaxes
9439 some of the IEEE floating-point standard's rules for checking
9440 inputs against Not-a-Number (NAN), in the interest of performance.
9443 Enable inlining of PLT entries in function calls to functions that
9444 are not known to bind locally. It has no effect without `-mfdpic'.
9447 Build standalone application for multicore Blackfin processor.
9448 Proper start files and link scripts will be used to support
9449 multicore. This option defines `__BFIN_MULTICORE'. It can only be
9450 used with `-mcpu=bf561[-SIREVISION]'. It can be used with
9451 `-mcorea' or `-mcoreb'. If it's used without `-mcorea' or
9452 `-mcoreb', single application/dual core programming model is used.
9453 In this model, the main function of Core B should be named as
9454 coreb_main. If it's used with `-mcorea' or `-mcoreb', one
9455 application per core programming model is used. If this option is
9456 not used, single core application programming model is used.
9459 Build standalone application for Core A of BF561 when using one
9460 application per core programming model. Proper start files and
9461 link scripts will be used to support Core A. This option defines
9462 `__BFIN_COREA'. It must be used with `-mmulticore'.
9465 Build standalone application for Core B of BF561 when using one
9466 application per core programming model. Proper start files and
9467 link scripts will be used to support Core B. This option defines
9468 `__BFIN_COREB'. When this option is used, coreb_main should be
9469 used instead of main. It must be used with `-mmulticore'.
9472 Build standalone application for SDRAM. Proper start files and
9473 link scripts will be used to put the application into SDRAM.
9474 Loader should initialize SDRAM before loading the application into
9475 SDRAM. This option defines `__BFIN_SDRAM'.
9478 Assume that ICPLBs are enabled at runtime. This has an effect on
9479 certain anomaly workarounds. For Linux targets, the default is to
9480 assume ICPLBs are enabled; for standalone applications the default
9484 File: gcc.info, Node: CRIS Options, Next: CRX Options, Prev: Blackfin Options, Up: Submodel Options
9489 These options are defined specifically for the CRIS ports.
9491 `-march=ARCHITECTURE-TYPE'
9492 `-mcpu=ARCHITECTURE-TYPE'
9493 Generate code for the specified architecture. The choices for
9494 ARCHITECTURE-TYPE are `v3', `v8' and `v10' for respectively
9495 ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is `v0' except for
9496 cris-axis-linux-gnu, where the default is `v10'.
9498 `-mtune=ARCHITECTURE-TYPE'
9499 Tune to ARCHITECTURE-TYPE everything applicable about the generated
9500 code, except for the ABI and the set of available instructions.
9501 The choices for ARCHITECTURE-TYPE are the same as for
9502 `-march=ARCHITECTURE-TYPE'.
9504 `-mmax-stack-frame=N'
9505 Warn when the stack frame of a function exceeds N bytes.
9509 The options `-metrax4' and `-metrax100' are synonyms for
9510 `-march=v3' and `-march=v8' respectively.
9512 `-mmul-bug-workaround'
9513 `-mno-mul-bug-workaround'
9514 Work around a bug in the `muls' and `mulu' instructions for CPU
9515 models where it applies. This option is active by default.
9518 Enable CRIS-specific verbose debug-related information in the
9519 assembly code. This option also has the effect to turn off the
9520 `#NO_APP' formatted-code indicator to the assembler at the
9521 beginning of the assembly file.
9524 Do not use condition-code results from previous instruction;
9525 always emit compare and test instructions before use of condition
9529 Do not emit instructions with side-effects in addressing modes
9530 other than post-increment.
9538 These options (no-options) arranges (eliminate arrangements) for
9539 the stack-frame, individual data and constants to be aligned for
9540 the maximum single data access size for the chosen CPU model. The
9541 default is to arrange for 32-bit alignment. ABI details such as
9542 structure layout are not affected by these options.
9547 Similar to the stack- data- and const-align options above, these
9548 options arrange for stack-frame, writable data and constants to
9549 all be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit
9552 `-mno-prologue-epilogue'
9553 `-mprologue-epilogue'
9554 With `-mno-prologue-epilogue', the normal function prologue and
9555 epilogue that sets up the stack-frame are omitted and no return
9556 instructions or return sequences are generated in the code. Use
9557 this option only together with visual inspection of the compiled
9558 code: no warnings or errors are generated when call-saved
9559 registers must be saved, or storage for local variable needs to be
9564 With `-fpic' and `-fPIC', don't generate (do generate) instruction
9565 sequences that load addresses for functions from the PLT part of
9566 the GOT rather than (traditional on other architectures) calls to
9567 the PLT. The default is `-mgotplt'.
9570 Legacy no-op option only recognized with the cris-axis-elf and
9571 cris-axis-linux-gnu targets.
9574 Legacy no-op option only recognized with the cris-axis-linux-gnu
9578 This option, recognized for the cris-axis-elf arranges to link
9579 with input-output functions from a simulator library. Code,
9580 initialized data and zero-initialized data are allocated
9584 Like `-sim', but pass linker options to locate initialized data at
9585 0x40000000 and zero-initialized data at 0x80000000.
9588 File: gcc.info, Node: CRX Options, Next: Darwin Options, Prev: CRIS Options, Up: Submodel Options
9593 These options are defined specifically for the CRX ports.
9596 Enable the use of multiply-accumulate instructions. Disabled by
9600 Push instructions will be used to pass outgoing arguments when
9601 functions are called. Enabled by default.
9604 File: gcc.info, Node: Darwin Options, Next: DEC Alpha Options, Prev: CRX Options, Up: Submodel Options
9606 3.17.7 Darwin Options
9607 ---------------------
9609 These options are defined for all architectures running the Darwin
9612 FSF GCC on Darwin does not create "fat" object files; it will create
9613 an object file for the single architecture that it was built to target.
9614 Apple's GCC on Darwin does create "fat" files if multiple `-arch'
9615 options are used; it does so by running the compiler or linker multiple
9616 times and joining the results together with `lipo'.
9618 The subtype of the file created (like `ppc7400' or `ppc970' or `i686')
9619 is determined by the flags that specify the ISA that GCC is targetting,
9620 like `-mcpu' or `-march'. The `-force_cpusubtype_ALL' option can be
9621 used to override this.
9623 The Darwin tools vary in their behavior when presented with an ISA
9624 mismatch. The assembler, `as', will only permit instructions to be
9625 used that are valid for the subtype of the file it is generating, so
9626 you cannot put 64-bit instructions in an `ppc750' object file. The
9627 linker for shared libraries, `/usr/bin/libtool', will fail and print an
9628 error if asked to create a shared library with a less restrictive
9629 subtype than its input files (for instance, trying to put a `ppc970'
9630 object file in a `ppc7400' library). The linker for executables, `ld',
9631 will quietly give the executable the most restrictive subtype of any of
9635 Add the framework directory DIR to the head of the list of
9636 directories to be searched for header files. These directories are
9637 interleaved with those specified by `-I' options and are scanned
9638 in a left-to-right order.
9640 A framework directory is a directory with frameworks in it. A
9641 framework is a directory with a `"Headers"' and/or
9642 `"PrivateHeaders"' directory contained directly in it that ends in
9643 `".framework"'. The name of a framework is the name of this
9644 directory excluding the `".framework"'. Headers associated with
9645 the framework are found in one of those two directories, with
9646 `"Headers"' being searched first. A subframework is a framework
9647 directory that is in a framework's `"Frameworks"' directory.
9648 Includes of subframework headers can only appear in a header of a
9649 framework that contains the subframework, or in a sibling
9650 subframework header. Two subframeworks are siblings if they occur
9651 in the same framework. A subframework should not have the same
9652 name as a framework, a warning will be issued if this is violated.
9653 Currently a subframework cannot have subframeworks, in the future,
9654 the mechanism may be extended to support this. The standard
9655 frameworks can be found in `"/System/Library/Frameworks"' and
9656 `"/Library/Frameworks"'. An example include looks like `#include
9657 <Framework/header.h>', where `Framework' denotes the name of the
9658 framework and header.h is found in the `"PrivateHeaders"' or
9659 `"Headers"' directory.
9662 Like `-F' except the directory is a treated as a system directory.
9663 The main difference between this `-iframework' and `-F' is that
9664 with `-iframework' the compiler does not warn about constructs
9665 contained within header files found via DIR. This option is valid
9666 only for the C family of languages.
9669 Emit debugging information for symbols that are used. For STABS
9670 debugging format, this enables `-feliminate-unused-debug-symbols'.
9671 This is by default ON.
9674 Emit debugging information for all symbols and types.
9676 `-mmacosx-version-min=VERSION'
9677 The earliest version of MacOS X that this executable will run on
9678 is VERSION. Typical values of VERSION include `10.1', `10.2', and
9681 If the compiler was built to use the system's headers by default,
9682 then the default for this option is the system version on which the
9683 compiler is running, otherwise the default is to make choices which
9684 are compatible with as many systems and code bases as possible.
9687 Enable kernel development mode. The `-mkernel' option sets
9688 `-static', `-fno-common', `-fno-cxa-atexit', `-fno-exceptions',
9689 `-fno-non-call-exceptions', `-fapple-kext', `-fno-weak' and
9690 `-fno-rtti' where applicable. This mode also sets `-mno-altivec',
9691 `-msoft-float', `-fno-builtin' and `-mlong-branch' for PowerPC
9695 Override the defaults for `bool' so that `sizeof(bool)==1'. By
9696 default `sizeof(bool)' is `4' when compiling for Darwin/PowerPC
9697 and `1' when compiling for Darwin/x86, so this option has no
9700 *Warning:* The `-mone-byte-bool' switch causes GCC to generate
9701 code that is not binary compatible with code generated without
9702 that switch. Using this switch may require recompiling all other
9703 modules in a program, including system libraries. Use this switch
9704 to conform to a non-default data model.
9706 `-mfix-and-continue'
9707 `-ffix-and-continue'
9709 Generate code suitable for fast turn around development. Needed to
9710 enable gdb to dynamically load `.o' files into already running
9711 programs. `-findirect-data' and `-ffix-and-continue' are provided
9712 for backwards compatibility.
9715 Loads all members of static archive libraries. See man ld(1) for
9718 `-arch_errors_fatal'
9719 Cause the errors having to do with files that have the wrong
9720 architecture to be fatal.
9723 Causes the output file to be marked such that the dynamic linker
9724 will bind all undefined references when the file is loaded or
9728 Produce a Mach-o bundle format file. See man ld(1) for more
9731 `-bundle_loader EXECUTABLE'
9732 This option specifies the EXECUTABLE that will be loading the build
9733 output file being linked. See man ld(1) for more information.
9736 When passed this option, GCC will produce a dynamic library
9737 instead of an executable when linking, using the Darwin `libtool'
9740 `-force_cpusubtype_ALL'
9741 This causes GCC's output file to have the ALL subtype, instead of
9742 one controlled by the `-mcpu' or `-march' option.
9744 `-allowable_client CLIENT_NAME'
9746 `-compatibility_version'
9751 `-dylinker_install_name'
9753 `-exported_symbols_list'
9756 `-force_flat_namespace'
9757 `-headerpad_max_install_names'
9761 `-keep_private_externs'
9764 `-multiply_defined_unused'
9766 `-no_dead_strip_inits_and_terms'
9773 `-prebind_all_twolevel_modules'
9777 `-sectobjectsymbols'
9781 `-sectobjectsymbols'
9784 `-segs_read_only_addr'
9785 `-segs_read_write_addr'
9787 `-seg_addr_table_filename'
9790 `-segs_read_only_addr'
9791 `-segs_read_write_addr'
9796 `-twolevel_namespace'
9799 `-unexported_symbols_list'
9800 `-weak_reference_mismatches'
9802 These options are passed to the Darwin linker. The Darwin linker
9803 man page describes them in detail.
9806 File: gcc.info, Node: DEC Alpha Options, Next: DEC Alpha/VMS Options, Prev: Darwin Options, Up: Submodel Options
9808 3.17.8 DEC Alpha Options
9809 ------------------------
9811 These `-m' options are defined for the DEC Alpha implementations:
9815 Use (do not use) the hardware floating-point instructions for
9816 floating-point operations. When `-msoft-float' is specified,
9817 functions in `libgcc.a' will be used to perform floating-point
9818 operations. Unless they are replaced by routines that emulate the
9819 floating-point operations, or compiled in such a way as to call
9820 such emulations routines, these routines will issue floating-point
9821 operations. If you are compiling for an Alpha without
9822 floating-point operations, you must ensure that the library is
9823 built so as not to call them.
9825 Note that Alpha implementations without floating-point operations
9826 are required to have floating-point registers.
9830 Generate code that uses (does not use) the floating-point register
9831 set. `-mno-fp-regs' implies `-msoft-float'. If the floating-point
9832 register set is not used, floating point operands are passed in
9833 integer registers as if they were integers and floating-point
9834 results are passed in `$0' instead of `$f0'. This is a
9835 non-standard calling sequence, so any function with a
9836 floating-point argument or return value called by code compiled
9837 with `-mno-fp-regs' must also be compiled with that option.
9839 A typical use of this option is building a kernel that does not
9840 use, and hence need not save and restore, any floating-point
9844 The Alpha architecture implements floating-point hardware
9845 optimized for maximum performance. It is mostly compliant with
9846 the IEEE floating point standard. However, for full compliance,
9847 software assistance is required. This option generates code fully
9848 IEEE compliant code _except_ that the INEXACT-FLAG is not
9849 maintained (see below). If this option is turned on, the
9850 preprocessor macro `_IEEE_FP' is defined during compilation. The
9851 resulting code is less efficient but is able to correctly support
9852 denormalized numbers and exceptional IEEE values such as
9853 not-a-number and plus/minus infinity. Other Alpha compilers call
9854 this option `-ieee_with_no_inexact'.
9856 `-mieee-with-inexact'
9857 This is like `-mieee' except the generated code also maintains the
9858 IEEE INEXACT-FLAG. Turning on this option causes the generated
9859 code to implement fully-compliant IEEE math. In addition to
9860 `_IEEE_FP', `_IEEE_FP_EXACT' is defined as a preprocessor macro.
9861 On some Alpha implementations the resulting code may execute
9862 significantly slower than the code generated by default. Since
9863 there is very little code that depends on the INEXACT-FLAG, you
9864 should normally not specify this option. Other Alpha compilers
9865 call this option `-ieee_with_inexact'.
9867 `-mfp-trap-mode=TRAP-MODE'
9868 This option controls what floating-point related traps are enabled.
9869 Other Alpha compilers call this option `-fptm TRAP-MODE'. The
9870 trap mode can be set to one of four values:
9873 This is the default (normal) setting. The only traps that
9874 are enabled are the ones that cannot be disabled in software
9875 (e.g., division by zero trap).
9878 In addition to the traps enabled by `n', underflow traps are
9882 Like `u', but the instructions are marked to be safe for
9883 software completion (see Alpha architecture manual for
9887 Like `su', but inexact traps are enabled as well.
9889 `-mfp-rounding-mode=ROUNDING-MODE'
9890 Selects the IEEE rounding mode. Other Alpha compilers call this
9891 option `-fprm ROUNDING-MODE'. The ROUNDING-MODE can be one of:
9894 Normal IEEE rounding mode. Floating point numbers are
9895 rounded towards the nearest machine number or towards the
9896 even machine number in case of a tie.
9899 Round towards minus infinity.
9902 Chopped rounding mode. Floating point numbers are rounded
9906 Dynamic rounding mode. A field in the floating point control
9907 register (FPCR, see Alpha architecture reference manual)
9908 controls the rounding mode in effect. The C library
9909 initializes this register for rounding towards plus infinity.
9910 Thus, unless your program modifies the FPCR, `d' corresponds
9911 to round towards plus infinity.
9913 `-mtrap-precision=TRAP-PRECISION'
9914 In the Alpha architecture, floating point traps are imprecise.
9915 This means without software assistance it is impossible to recover
9916 from a floating trap and program execution normally needs to be
9917 terminated. GCC can generate code that can assist operating
9918 system trap handlers in determining the exact location that caused
9919 a floating point trap. Depending on the requirements of an
9920 application, different levels of precisions can be selected:
9923 Program precision. This option is the default and means a
9924 trap handler can only identify which program caused a
9925 floating point exception.
9928 Function precision. The trap handler can determine the
9929 function that caused a floating point exception.
9932 Instruction precision. The trap handler can determine the
9933 exact instruction that caused a floating point exception.
9935 Other Alpha compilers provide the equivalent options called
9936 `-scope_safe' and `-resumption_safe'.
9939 This option marks the generated code as IEEE conformant. You must
9940 not use this option unless you also specify `-mtrap-precision=i'
9941 and either `-mfp-trap-mode=su' or `-mfp-trap-mode=sui'. Its only
9942 effect is to emit the line `.eflag 48' in the function prologue of
9943 the generated assembly file. Under DEC Unix, this has the effect
9944 that IEEE-conformant math library routines will be linked in.
9947 Normally GCC examines a 32- or 64-bit integer constant to see if
9948 it can construct it from smaller constants in two or three
9949 instructions. If it cannot, it will output the constant as a
9950 literal and generate code to load it from the data segment at
9953 Use this option to require GCC to construct _all_ integer constants
9954 using code, even if it takes more instructions (the maximum is
9957 You would typically use this option to build a shared library
9958 dynamic loader. Itself a shared library, it must relocate itself
9959 in memory before it can find the variables and constants in its
9964 Select whether to generate code to be assembled by the
9965 vendor-supplied assembler (`-malpha-as') or by the GNU assembler
9976 Indicate whether GCC should generate code to use the optional BWX,
9977 CIX, FIX and MAX instruction sets. The default is to use the
9978 instruction sets supported by the CPU type specified via `-mcpu='
9979 option or that of the CPU on which GCC was built if none was
9984 Generate code that uses (does not use) VAX F and G floating point
9985 arithmetic instead of IEEE single and double precision.
9988 `-mno-explicit-relocs'
9989 Older Alpha assemblers provided no way to generate symbol
9990 relocations except via assembler macros. Use of these macros does
9991 not allow optimal instruction scheduling. GNU binutils as of
9992 version 2.12 supports a new syntax that allows the compiler to
9993 explicitly mark which relocations should apply to which
9994 instructions. This option is mostly useful for debugging, as GCC
9995 detects the capabilities of the assembler when it is built and
9996 sets the default accordingly.
10000 When `-mexplicit-relocs' is in effect, static data is accessed via
10001 "gp-relative" relocations. When `-msmall-data' is used, objects 8
10002 bytes long or smaller are placed in a "small data area" (the
10003 `.sdata' and `.sbss' sections) and are accessed via 16-bit
10004 relocations off of the `$gp' register. This limits the size of
10005 the small data area to 64KB, but allows the variables to be
10006 directly accessed via a single instruction.
10008 The default is `-mlarge-data'. With this option the data area is
10009 limited to just below 2GB. Programs that require more than 2GB of
10010 data must use `malloc' or `mmap' to allocate the data in the heap
10011 instead of in the program's data segment.
10013 When generating code for shared libraries, `-fpic' implies
10014 `-msmall-data' and `-fPIC' implies `-mlarge-data'.
10018 When `-msmall-text' is used, the compiler assumes that the code of
10019 the entire program (or shared library) fits in 4MB, and is thus
10020 reachable with a branch instruction. When `-msmall-data' is used,
10021 the compiler can assume that all local symbols share the same
10022 `$gp' value, and thus reduce the number of instructions required
10023 for a function call from 4 to 1.
10025 The default is `-mlarge-text'.
10028 Set the instruction set and instruction scheduling parameters for
10029 machine type CPU_TYPE. You can specify either the `EV' style name
10030 or the corresponding chip number. GCC supports scheduling
10031 parameters for the EV4, EV5 and EV6 family of processors and will
10032 choose the default values for the instruction set from the
10033 processor you specify. If you do not specify a processor type,
10034 GCC will default to the processor on which the compiler was built.
10036 Supported values for CPU_TYPE are
10041 Schedules as an EV4 and has no instruction set extensions.
10045 Schedules as an EV5 and has no instruction set extensions.
10049 Schedules as an EV5 and supports the BWX extension.
10054 Schedules as an EV5 and supports the BWX and MAX extensions.
10058 Schedules as an EV6 and supports the BWX, FIX, and MAX
10063 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX
10066 Native Linux/GNU toolchains also support the value `native', which
10067 selects the best architecture option for the host processor.
10068 `-mcpu=native' has no effect if GCC does not recognize the
10072 Set only the instruction scheduling parameters for machine type
10073 CPU_TYPE. The instruction set is not changed.
10075 Native Linux/GNU toolchains also support the value `native', which
10076 selects the best architecture option for the host processor.
10077 `-mtune=native' has no effect if GCC does not recognize the
10080 `-mmemory-latency=TIME'
10081 Sets the latency the scheduler should assume for typical memory
10082 references as seen by the application. This number is highly
10083 dependent on the memory access patterns used by the application
10084 and the size of the external cache on the machine.
10086 Valid options for TIME are
10089 A decimal number representing clock cycles.
10095 The compiler contains estimates of the number of clock cycles
10096 for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10097 (also called Dcache, Scache, and Bcache), as well as to main
10098 memory. Note that L3 is only valid for EV5.
10102 File: gcc.info, Node: DEC Alpha/VMS Options, Next: FR30 Options, Prev: DEC Alpha Options, Up: Submodel Options
10104 3.17.9 DEC Alpha/VMS Options
10105 ----------------------------
10107 These `-m' options are defined for the DEC Alpha/VMS implementations:
10109 `-mvms-return-codes'
10110 Return VMS condition codes from main. The default is to return
10111 POSIX style condition (e.g. error) codes.
10114 File: gcc.info, Node: FR30 Options, Next: FRV Options, Prev: DEC Alpha/VMS Options, Up: Submodel Options
10116 3.17.10 FR30 Options
10117 --------------------
10119 These options are defined specifically for the FR30 port.
10122 Use the small address space model. This can produce smaller code,
10123 but it does assume that all symbolic values and addresses will fit
10124 into a 20-bit range.
10127 Assume that run-time support has been provided and so there is no
10128 need to include the simulator library (`libsim.a') on the linker
10133 File: gcc.info, Node: FRV Options, Next: GNU/Linux Options, Prev: FR30 Options, Up: Submodel Options
10135 3.17.11 FRV Options
10136 -------------------
10139 Only use the first 32 general purpose registers.
10142 Use all 64 general purpose registers.
10145 Use only the first 32 floating point registers.
10148 Use all 64 floating point registers
10151 Use hardware instructions for floating point operations.
10154 Use library routines for floating point operations.
10157 Dynamically allocate condition code registers.
10160 Do not try to dynamically allocate condition code registers, only
10161 use `icc0' and `fcc0'.
10164 Change ABI to use double word insns.
10167 Do not use double word instructions.
10170 Use floating point double instructions.
10173 Do not use floating point double instructions.
10176 Use media instructions.
10179 Do not use media instructions.
10182 Use multiply and add/subtract instructions.
10185 Do not use multiply and add/subtract instructions.
10188 Select the FDPIC ABI, that uses function descriptors to represent
10189 pointers to functions. Without any PIC/PIE-related options, it
10190 implies `-fPIE'. With `-fpic' or `-fpie', it assumes GOT entries
10191 and small data are within a 12-bit range from the GOT base
10192 address; with `-fPIC' or `-fPIE', GOT offsets are computed with 32
10193 bits. With a `bfin-elf' target, this option implies `-msim'.
10196 Enable inlining of PLT entries in function calls to functions that
10197 are not known to bind locally. It has no effect without `-mfdpic'.
10198 It's enabled by default if optimizing for speed and compiling for
10199 shared libraries (i.e., `-fPIC' or `-fpic'), or when an
10200 optimization option such as `-O3' or above is present in the
10204 Assume a large TLS segment when generating thread-local code.
10207 Do not assume a large TLS segment when generating thread-local
10211 Enable the use of `GPREL' relocations in the FDPIC ABI for data
10212 that is known to be in read-only sections. It's enabled by
10213 default, except for `-fpic' or `-fpie': even though it may help
10214 make the global offset table smaller, it trades 1 instruction for
10215 4. With `-fPIC' or `-fPIE', it trades 3 instructions for 4, one
10216 of which may be shared by multiple symbols, and it avoids the need
10217 for a GOT entry for the referenced symbol, so it's more likely to
10218 be a win. If it is not, `-mno-gprel-ro' can be used to disable it.
10220 `-multilib-library-pic'
10221 Link with the (library, not FD) pic libraries. It's implied by
10222 `-mlibrary-pic', as well as by `-fPIC' and `-fpic' without
10223 `-mfdpic'. You should never have to use it explicitly.
10226 Follow the EABI requirement of always creating a frame pointer
10227 whenever a stack frame is allocated. This option is enabled by
10228 default and can be disabled with `-mno-linked-fp'.
10231 Use indirect addressing to call functions outside the current
10232 compilation unit. This allows the functions to be placed anywhere
10233 within the 32-bit address space.
10236 Try to align labels to an 8-byte boundary by inserting nops into
10237 the previous packet. This option only has an effect when VLIW
10238 packing is enabled. It doesn't create new packets; it merely adds
10239 nops to existing ones.
10242 Generate position-independent EABI code.
10245 Use only the first four media accumulator registers.
10248 Use all eight media accumulator registers.
10251 Pack VLIW instructions.
10254 Do not pack VLIW instructions.
10257 Do not mark ABI switches in e_flags.
10260 Enable the use of conditional-move instructions (default).
10262 This switch is mainly for debugging the compiler and will likely
10263 be removed in a future version.
10266 Disable the use of conditional-move instructions.
10268 This switch is mainly for debugging the compiler and will likely
10269 be removed in a future version.
10272 Enable the use of conditional set instructions (default).
10274 This switch is mainly for debugging the compiler and will likely
10275 be removed in a future version.
10278 Disable the use of conditional set instructions.
10280 This switch is mainly for debugging the compiler and will likely
10281 be removed in a future version.
10284 Enable the use of conditional execution (default).
10286 This switch is mainly for debugging the compiler and will likely
10287 be removed in a future version.
10290 Disable the use of conditional execution.
10292 This switch is mainly for debugging the compiler and will likely
10293 be removed in a future version.
10296 Run a pass to pack branches into VLIW instructions (default).
10298 This switch is mainly for debugging the compiler and will likely
10299 be removed in a future version.
10302 Do not run a pass to pack branches into VLIW instructions.
10304 This switch is mainly for debugging the compiler and will likely
10305 be removed in a future version.
10307 `-mmulti-cond-exec'
10308 Enable optimization of `&&' and `||' in conditional execution
10311 This switch is mainly for debugging the compiler and will likely
10312 be removed in a future version.
10314 `-mno-multi-cond-exec'
10315 Disable optimization of `&&' and `||' in conditional execution.
10317 This switch is mainly for debugging the compiler and will likely
10318 be removed in a future version.
10320 `-mnested-cond-exec'
10321 Enable nested conditional execution optimizations (default).
10323 This switch is mainly for debugging the compiler and will likely
10324 be removed in a future version.
10326 `-mno-nested-cond-exec'
10327 Disable nested conditional execution optimizations.
10329 This switch is mainly for debugging the compiler and will likely
10330 be removed in a future version.
10332 `-moptimize-membar'
10333 This switch removes redundant `membar' instructions from the
10334 compiler generated code. It is enabled by default.
10336 `-mno-optimize-membar'
10337 This switch disables the automatic removal of redundant `membar'
10338 instructions from the generated code.
10341 Cause gas to print out tomcat statistics.
10344 Select the processor type for which to generate code. Possible
10345 values are `frv', `fr550', `tomcat', `fr500', `fr450', `fr405',
10346 `fr400', `fr300' and `simple'.
10350 File: gcc.info, Node: GNU/Linux Options, Next: H8/300 Options, Prev: FRV Options, Up: Submodel Options
10352 3.17.12 GNU/Linux Options
10353 -------------------------
10355 These `-m' options are defined for GNU/Linux targets:
10358 Use the GNU C library instead of uClibc. This is the default
10359 except on `*-*-linux-*uclibc*' targets.
10362 Use uClibc instead of the GNU C library. This is the default on
10363 `*-*-linux-*uclibc*' targets.
10366 File: gcc.info, Node: H8/300 Options, Next: HPPA Options, Prev: GNU/Linux Options, Up: Submodel Options
10368 3.17.13 H8/300 Options
10369 ----------------------
10371 These `-m' options are defined for the H8/300 implementations:
10374 Shorten some address references at link time, when possible; uses
10375 the linker option `-relax'. *Note `ld' and the H8/300:
10376 (ld)H8/300, for a fuller description.
10379 Generate code for the H8/300H.
10382 Generate code for the H8S.
10385 Generate code for the H8S and H8/300H in the normal mode. This
10386 switch must be used either with `-mh' or `-ms'.
10389 Generate code for the H8S/2600. This switch must be used with
10393 Make `int' data 32 bits by default.
10396 On the H8/300H and H8S, use the same alignment rules as for the
10397 H8/300. The default for the H8/300H and H8S is to align longs and
10398 floats on 4 byte boundaries. `-malign-300' causes them to be
10399 aligned on 2 byte boundaries. This option has no effect on the
10403 File: gcc.info, Node: HPPA Options, Next: i386 and x86-64 Options, Prev: H8/300 Options, Up: Submodel Options
10405 3.17.14 HPPA Options
10406 --------------------
10408 These `-m' options are defined for the HPPA family of computers:
10410 `-march=ARCHITECTURE-TYPE'
10411 Generate code for the specified architecture. The choices for
10412 ARCHITECTURE-TYPE are `1.0' for PA 1.0, `1.1' for PA 1.1, and
10413 `2.0' for PA 2.0 processors. Refer to `/usr/lib/sched.models' on
10414 an HP-UX system to determine the proper architecture option for
10415 your machine. Code compiled for lower numbered architectures will
10416 run on higher numbered architectures, but not the other way around.
10421 Synonyms for `-march=1.0', `-march=1.1', and `-march=2.0'
10425 Generate code suitable for big switch tables. Use this option
10426 only if the assembler/linker complain about out of range branches
10427 within a switch table.
10430 Fill delay slots of function calls with unconditional jump
10431 instructions by modifying the return pointer for the function call
10432 to be the target of the conditional jump.
10435 Prevent floating point registers from being used in any manner.
10436 This is necessary for compiling kernels which perform lazy context
10437 switching of floating point registers. If you use this option and
10438 attempt to perform floating point operations, the compiler will
10441 `-mdisable-indexing'
10442 Prevent the compiler from using indexing address modes. This
10443 avoids some rather obscure problems when compiling MIG generated
10447 Generate code that assumes the target has no space registers.
10448 This allows GCC to generate faster indirect calls and use unscaled
10449 index address modes.
10451 Such code is suitable for level 0 PA systems and kernels.
10453 `-mfast-indirect-calls'
10454 Generate code that assumes calls never cross space boundaries.
10455 This allows GCC to emit code which performs faster indirect calls.
10457 This option will not work in the presence of shared libraries or
10460 `-mfixed-range=REGISTER-RANGE'
10461 Generate code treating the given register range as fixed registers.
10462 A fixed register is one that the register allocator can not use.
10463 This is useful when compiling kernel code. A register range is
10464 specified as two registers separated by a dash. Multiple register
10465 ranges can be specified separated by a comma.
10467 `-mlong-load-store'
10468 Generate 3-instruction load and store sequences as sometimes
10469 required by the HP-UX 10 linker. This is equivalent to the `+k'
10470 option to the HP compilers.
10472 `-mportable-runtime'
10473 Use the portable calling conventions proposed by HP for ELF
10477 Enable the use of assembler directives only GAS understands.
10479 `-mschedule=CPU-TYPE'
10480 Schedule code according to the constraints for the machine type
10481 CPU-TYPE. The choices for CPU-TYPE are `700' `7100', `7100LC',
10482 `7200', `7300' and `8000'. Refer to `/usr/lib/sched.models' on an
10483 HP-UX system to determine the proper scheduling option for your
10484 machine. The default scheduling is `8000'.
10487 Enable the optimization pass in the HP-UX linker. Note this makes
10488 symbolic debugging impossible. It also triggers a bug in the
10489 HP-UX 8 and HP-UX 9 linkers in which they give bogus error
10490 messages when linking some programs.
10493 Generate output containing library calls for floating point.
10494 *Warning:* the requisite libraries are not available for all HPPA
10495 targets. Normally the facilities of the machine's usual C
10496 compiler are used, but this cannot be done directly in
10497 cross-compilation. You must make your own arrangements to provide
10498 suitable library functions for cross-compilation.
10500 `-msoft-float' changes the calling convention in the output file;
10501 therefore, it is only useful if you compile _all_ of a program with
10502 this option. In particular, you need to compile `libgcc.a', the
10503 library that comes with GCC, with `-msoft-float' in order for this
10507 Generate the predefine, `_SIO', for server IO. The default is
10508 `-mwsio'. This generates the predefines, `__hp9000s700',
10509 `__hp9000s700__' and `_WSIO', for workstation IO. These options
10510 are available under HP-UX and HI-UX.
10513 Use GNU ld specific options. This passes `-shared' to ld when
10514 building a shared library. It is the default when GCC is
10515 configured, explicitly or implicitly, with the GNU linker. This
10516 option does not have any affect on which ld is called, it only
10517 changes what parameters are passed to that ld. The ld that is
10518 called is determined by the `--with-ld' configure option, GCC's
10519 program search path, and finally by the user's `PATH'. The linker
10520 used by GCC can be printed using `which `gcc
10521 -print-prog-name=ld`'. This option is only available on the 64
10522 bit HP-UX GCC, i.e. configured with `hppa*64*-*-hpux*'.
10525 Use HP ld specific options. This passes `-b' to ld when building
10526 a shared library and passes `+Accept TypeMismatch' to ld on all
10527 links. It is the default when GCC is configured, explicitly or
10528 implicitly, with the HP linker. This option does not have any
10529 affect on which ld is called, it only changes what parameters are
10530 passed to that ld. The ld that is called is determined by the
10531 `--with-ld' configure option, GCC's program search path, and
10532 finally by the user's `PATH'. The linker used by GCC can be
10533 printed using `which `gcc -print-prog-name=ld`'. This option is
10534 only available on the 64 bit HP-UX GCC, i.e. configured with
10535 `hppa*64*-*-hpux*'.
10538 Generate code that uses long call sequences. This ensures that a
10539 call is always able to reach linker generated stubs. The default
10540 is to generate long calls only when the distance from the call
10541 site to the beginning of the function or translation unit, as the
10542 case may be, exceeds a predefined limit set by the branch type
10543 being used. The limits for normal calls are 7,600,000 and 240,000
10544 bytes, respectively for the PA 2.0 and PA 1.X architectures.
10545 Sibcalls are always limited at 240,000 bytes.
10547 Distances are measured from the beginning of functions when using
10548 the `-ffunction-sections' option, or when using the `-mgas' and
10549 `-mno-portable-runtime' options together under HP-UX with the SOM
10552 It is normally not desirable to use this option as it will degrade
10553 performance. However, it may be useful in large applications,
10554 particularly when partial linking is used to build the application.
10556 The types of long calls used depends on the capabilities of the
10557 assembler and linker, and the type of code being generated. The
10558 impact on systems that support long absolute calls, and long pic
10559 symbol-difference or pc-relative calls should be relatively small.
10560 However, an indirect call is used on 32-bit ELF systems in pic code
10561 and it is quite long.
10564 Generate compiler predefines and select a startfile for the
10565 specified UNIX standard. The choices for UNIX-STD are `93', `95'
10566 and `98'. `93' is supported on all HP-UX versions. `95' is
10567 available on HP-UX 10.10 and later. `98' is available on HP-UX
10568 11.11 and later. The default values are `93' for HP-UX 10.00,
10569 `95' for HP-UX 10.10 though to 11.00, and `98' for HP-UX 11.11 and
10572 `-munix=93' provides the same predefines as GCC 3.3 and 3.4.
10573 `-munix=95' provides additional predefines for `XOPEN_UNIX' and
10574 `_XOPEN_SOURCE_EXTENDED', and the startfile `unix95.o'.
10575 `-munix=98' provides additional predefines for `_XOPEN_UNIX',
10576 `_XOPEN_SOURCE_EXTENDED', `_INCLUDE__STDC_A1_SOURCE' and
10577 `_INCLUDE_XOPEN_SOURCE_500', and the startfile `unix98.o'.
10579 It is _important_ to note that this option changes the interfaces
10580 for various library routines. It also affects the operational
10581 behavior of the C library. Thus, _extreme_ care is needed in
10584 Library code that is intended to operate with more than one UNIX
10585 standard must test, set and restore the variable
10586 __XPG4_EXTENDED_MASK as appropriate. Most GNU software doesn't
10587 provide this capability.
10590 Suppress the generation of link options to search libdld.sl when
10591 the `-static' option is specified on HP-UX 10 and later.
10594 The HP-UX implementation of setlocale in libc has a dependency on
10595 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10596 when the `-static' option is specified, special link options are
10597 needed to resolve this dependency.
10599 On HP-UX 10 and later, the GCC driver adds the necessary options to
10600 link with libdld.sl when the `-static' option is specified. This
10601 causes the resulting binary to be dynamic. On the 64-bit port,
10602 the linkers generate dynamic binaries by default in any case. The
10603 `-nolibdld' option can be used to prevent the GCC driver from
10604 adding these link options.
10607 Add support for multithreading with the "dce thread" library under
10608 HP-UX. This option sets flags for both the preprocessor and
10612 File: gcc.info, Node: i386 and x86-64 Options, Next: i386 and x86-64 Windows Options, Prev: HPPA Options, Up: Submodel Options
10614 3.17.15 Intel 386 and AMD x86-64 Options
10615 ----------------------------------------
10617 These `-m' options are defined for the i386 and x86-64 family of
10621 Tune to CPU-TYPE everything applicable about the generated code,
10622 except for the ABI and the set of available instructions. The
10623 choices for CPU-TYPE are:
10625 Produce code optimized for the most common IA32/AMD64/EM64T
10626 processors. If you know the CPU on which your code will run,
10627 then you should use the corresponding `-mtune' option instead
10628 of `-mtune=generic'. But, if you do not know exactly what
10629 CPU users of your application will have, then you should use
10632 As new processors are deployed in the marketplace, the
10633 behavior of this option will change. Therefore, if you
10634 upgrade to a newer version of GCC, the code generated option
10635 will change to reflect the processors that were most common
10636 when that version of GCC was released.
10638 There is no `-march=generic' option because `-march'
10639 indicates the instruction set the compiler can use, and there
10640 is no generic instruction set applicable to all processors.
10641 In contrast, `-mtune' indicates the processor (or, in this
10642 case, collection of processors) for which the code is
10646 This selects the CPU to tune for at compilation time by
10647 determining the processor type of the compiling machine.
10648 Using `-mtune=native' will produce code optimized for the
10649 local machine under the constraints of the selected
10650 instruction set. Using `-march=native' will enable all
10651 instruction subsets supported by the local machine (hence the
10652 result might not run on different machines).
10655 Original Intel's i386 CPU.
10658 Intel's i486 CPU. (No scheduling is implemented for this
10662 Intel Pentium CPU with no MMX support.
10665 Intel PentiumMMX CPU based on Pentium core with MMX
10666 instruction set support.
10669 Intel PentiumPro CPU.
10672 Same as `generic', but when used as `march' option, PentiumPro
10673 instruction set will be used, so the code will run on all
10677 Intel Pentium2 CPU based on PentiumPro core with MMX
10678 instruction set support.
10680 _pentium3, pentium3m_
10681 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE
10682 instruction set support.
10685 Low power version of Intel Pentium3 CPU with MMX, SSE and
10686 SSE2 instruction set support. Used by Centrino notebooks.
10688 _pentium4, pentium4m_
10689 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set
10693 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2
10694 and SSE3 instruction set support.
10697 Improved version of Intel Pentium4 CPU with 64-bit
10698 extensions, MMX, SSE, SSE2 and SSE3 instruction set support.
10701 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3
10702 and SSSE3 instruction set support.
10705 AMD K6 CPU with MMX instruction set support.
10708 Improved versions of AMD K6 CPU with MMX and 3dNOW!
10709 instruction set support.
10711 _athlon, athlon-tbird_
10712 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE
10713 prefetch instructions support.
10715 _athlon-4, athlon-xp, athlon-mp_
10716 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and
10717 full SSE instruction set support.
10719 _k8, opteron, athlon64, athlon-fx_
10720 AMD K8 core based CPUs with x86-64 instruction set support.
10721 (This supersets MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and
10722 64-bit instruction set extensions.)
10724 _k8-sse3, opteron-sse3, athlon64-sse3_
10725 Improved versions of k8, opteron and athlon64 with SSE3
10726 instruction set support.
10728 _amdfam10, barcelona_
10729 AMD Family 10h core based CPUs with x86-64 instruction set
10730 support. (This supersets MMX, SSE, SSE2, SSE3, SSE4A,
10731 3dNOW!, enhanced 3dNOW!, ABM and 64-bit instruction set
10735 IDT Winchip C6 CPU, dealt in same way as i486 with additional
10736 MMX instruction set support.
10739 IDT Winchip2 CPU, dealt in same way as i486 with additional
10740 MMX and 3dNOW! instruction set support.
10743 Via C3 CPU with MMX and 3dNOW! instruction set support. (No
10744 scheduling is implemented for this chip.)
10747 Via C3-2 CPU with MMX and SSE instruction set support. (No
10748 scheduling is implemented for this chip.)
10751 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10753 While picking a specific CPU-TYPE will schedule things
10754 appropriately for that particular chip, the compiler will not
10755 generate any code that does not run on the i386 without the
10756 `-march=CPU-TYPE' option being used.
10759 Generate instructions for the machine type CPU-TYPE. The choices
10760 for CPU-TYPE are the same as for `-mtune'. Moreover, specifying
10761 `-march=CPU-TYPE' implies `-mtune=CPU-TYPE'.
10764 A deprecated synonym for `-mtune'.
10767 Generate floating point arithmetics for selected unit UNIT. The
10768 choices for UNIT are:
10771 Use the standard 387 floating point coprocessor present
10772 majority of chips and emulated otherwise. Code compiled with
10773 this option will run almost everywhere. The temporary
10774 results are computed in 80bit precision instead of precision
10775 specified by the type resulting in slightly different results
10776 compared to most of other chips. See `-ffloat-store' for
10777 more detailed description.
10779 This is the default choice for i386 compiler.
10782 Use scalar floating point instructions present in the SSE
10783 instruction set. This instruction set is supported by
10784 Pentium3 and newer chips, in the AMD line by Athlon-4,
10785 Athlon-xp and Athlon-mp chips. The earlier version of SSE
10786 instruction set supports only single precision arithmetics,
10787 thus the double and extended precision arithmetics is still
10788 done using 387. Later version, present only in Pentium4 and
10789 the future AMD x86-64 chips supports double precision
10792 For the i386 compiler, you need to use `-march=CPU-TYPE',
10793 `-msse' or `-msse2' switches to enable SSE extensions and
10794 make this option effective. For the x86-64 compiler, these
10795 extensions are enabled by default.
10797 The resulting code should be considerably faster in the
10798 majority of cases and avoid the numerical instability
10799 problems of 387 code, but may break some existing code that
10800 expects temporaries to be 80bit.
10802 This is the default choice for the x86-64 compiler.
10807 Attempt to utilize both instruction sets at once. This
10808 effectively double the amount of available registers and on
10809 chips with separate execution units for 387 and SSE the
10810 execution resources too. Use this option with care, as it is
10811 still experimental, because the GCC register allocator does
10812 not model separate functional units well resulting in
10813 instable performance.
10816 Output asm instructions using selected DIALECT. Supported choices
10817 are `intel' or `att' (the default one). Darwin does not support
10822 Control whether or not the compiler uses IEEE floating point
10823 comparisons. These handle correctly the case where the result of a
10824 comparison is unordered.
10827 Generate output containing library calls for floating point.
10828 *Warning:* the requisite libraries are not part of GCC. Normally
10829 the facilities of the machine's usual C compiler are used, but
10830 this can't be done directly in cross-compilation. You must make
10831 your own arrangements to provide suitable library functions for
10834 On machines where a function returns floating point results in the
10835 80387 register stack, some floating point opcodes may be emitted
10836 even if `-msoft-float' is used.
10838 `-mno-fp-ret-in-387'
10839 Do not use the FPU registers for return values of functions.
10841 The usual calling convention has functions return values of types
10842 `float' and `double' in an FPU register, even if there is no FPU.
10843 The idea is that the operating system should emulate an FPU.
10845 The option `-mno-fp-ret-in-387' causes such values to be returned
10846 in ordinary CPU registers instead.
10848 `-mno-fancy-math-387'
10849 Some 387 emulators do not support the `sin', `cos' and `sqrt'
10850 instructions for the 387. Specify this option to avoid generating
10851 those instructions. This option is the default on FreeBSD,
10852 OpenBSD and NetBSD. This option is overridden when `-march'
10853 indicates that the target cpu will always have an FPU and so the
10854 instruction will not need emulation. As of revision 2.6.1, these
10855 instructions are not generated unless you also use the
10856 `-funsafe-math-optimizations' switch.
10859 `-mno-align-double'
10860 Control whether GCC aligns `double', `long double', and `long
10861 long' variables on a two word boundary or a one word boundary.
10862 Aligning `double' variables on a two word boundary will produce
10863 code that runs somewhat faster on a `Pentium' at the expense of
10866 On x86-64, `-malign-double' is enabled by default.
10868 *Warning:* if you use the `-malign-double' switch, structures
10869 containing the above types will be aligned differently than the
10870 published application binary interface specifications for the 386
10871 and will not be binary compatible with structures in code compiled
10872 without that switch.
10874 `-m96bit-long-double'
10875 `-m128bit-long-double'
10876 These switches control the size of `long double' type. The i386
10877 application binary interface specifies the size to be 96 bits, so
10878 `-m96bit-long-double' is the default in 32 bit mode.
10880 Modern architectures (Pentium and newer) would prefer `long double'
10881 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10882 conforming to the ABI, this would not be possible. So specifying a
10883 `-m128bit-long-double' will align `long double' to a 16 byte
10884 boundary by padding the `long double' with an additional 32 bit
10887 In the x86-64 compiler, `-m128bit-long-double' is the default
10888 choice as its ABI specifies that `long double' is to be aligned on
10891 Notice that neither of these options enable any extra precision
10892 over the x87 standard of 80 bits for a `long double'.
10894 *Warning:* if you override the default value for your target ABI,
10895 the structures and arrays containing `long double' variables will
10896 change their size as well as function calling convention for
10897 function taking `long double' will be modified. Hence they will
10898 not be binary compatible with arrays or structures in code
10899 compiled without that switch.
10901 `-mlarge-data-threshold=NUMBER'
10902 When `-mcmodel=medium' is specified, the data greater than
10903 THRESHOLD are placed in large data section. This value must be the
10904 same across all object linked into the binary and defaults to
10908 Use a different function-calling convention, in which functions
10909 that take a fixed number of arguments return with the `ret' NUM
10910 instruction, which pops their arguments while returning. This
10911 saves one instruction in the caller since there is no need to pop
10912 the arguments there.
10914 You can specify that an individual function is called with this
10915 calling sequence with the function attribute `stdcall'. You can
10916 also override the `-mrtd' option by using the function attribute
10917 `cdecl'. *Note Function Attributes::.
10919 *Warning:* this calling convention is incompatible with the one
10920 normally used on Unix, so you cannot use it if you need to call
10921 libraries compiled with the Unix compiler.
10923 Also, you must provide function prototypes for all functions that
10924 take variable numbers of arguments (including `printf'); otherwise
10925 incorrect code will be generated for calls to those functions.
10927 In addition, seriously incorrect code will result if you call a
10928 function with too many arguments. (Normally, extra arguments are
10929 harmlessly ignored.)
10932 Control how many registers are used to pass integer arguments. By
10933 default, no registers are used to pass arguments, and at most 3
10934 registers can be used. You can control this behavior for a
10935 specific function by using the function attribute `regparm'.
10936 *Note Function Attributes::.
10938 *Warning:* if you use this switch, and NUM is nonzero, then you
10939 must build all modules with the same value, including any
10940 libraries. This includes the system libraries and startup modules.
10943 Use SSE register passing conventions for float and double arguments
10944 and return values. You can control this behavior for a specific
10945 function by using the function attribute `sseregparm'. *Note
10946 Function Attributes::.
10948 *Warning:* if you use this switch then you must build all modules
10949 with the same value, including any libraries. This includes the
10950 system libraries and startup modules.
10955 Set 80387 floating-point precision to 32, 64 or 80 bits. When
10956 `-mpc32' is specified, the significands of results of
10957 floating-point operations are rounded to 24 bits (single
10958 precision); `-mpc64' rounds the significands of results of
10959 floating-point operations to 53 bits (double precision) and
10960 `-mpc80' rounds the significands of results of floating-point
10961 operations to 64 bits (extended double precision), which is the
10962 default. When this option is used, floating-point operations in
10963 higher precisions are not available to the programmer without
10964 setting the FPU control word explicitly.
10966 Setting the rounding of floating-point operations to less than the
10967 default 80 bits can speed some programs by 2% or more. Note that
10968 some mathematical libraries assume that extended precision (80
10969 bit) floating-point operations are enabled by default; routines in
10970 such libraries could suffer significant loss of accuracy,
10971 typically through so-called "catastrophic cancellation", when this
10972 option is used to set the precision to less than extended
10976 Realign the stack at entry. On the Intel x86, the `-mstackrealign'
10977 option will generate an alternate prologue and epilogue that
10978 realigns the runtime stack if necessary. This supports mixing
10979 legacy codes that keep a 4-byte aligned stack with modern codes
10980 that keep a 16-byte stack for SSE compatibility. See also the
10981 attribute `force_align_arg_pointer', applicable to individual
10984 `-mpreferred-stack-boundary=NUM'
10985 Attempt to keep the stack boundary aligned to a 2 raised to NUM
10986 byte boundary. If `-mpreferred-stack-boundary' is not specified,
10987 the default is 4 (16 bytes or 128 bits).
10989 `-mincoming-stack-boundary=NUM'
10990 Assume the incoming stack is aligned to a 2 raised to NUM byte
10991 boundary. If `-mincoming-stack-boundary' is not specified, the
10992 one specified by `-mpreferred-stack-boundary' will be used.
10994 On Pentium and PentiumPro, `double' and `long double' values
10995 should be aligned to an 8 byte boundary (see `-malign-double') or
10996 suffer significant run time performance penalties. On Pentium
10997 III, the Streaming SIMD Extension (SSE) data type `__m128' may not
10998 work properly if it is not 16 byte aligned.
11000 To ensure proper alignment of this values on the stack, the stack
11001 boundary must be as aligned as that required by any value stored
11002 on the stack. Further, every function must be generated such that
11003 it keeps the stack aligned. Thus calling a function compiled with
11004 a higher preferred stack boundary from a function compiled with a
11005 lower preferred stack boundary will most likely misalign the
11006 stack. It is recommended that libraries that use callbacks always
11007 use the default setting.
11009 This extra alignment does consume extra stack space, and generally
11010 increases code size. Code that is sensitive to stack space usage,
11011 such as embedded systems and operating system kernels, may want to
11012 reduce the preferred alignment to `-mpreferred-stack-boundary=2'.
11046 These switches enable or disable the use of instructions in the
11047 MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A,
11048 SSE5, ABM or 3DNow! extended instruction sets. These extensions
11049 are also available as built-in functions: see *note X86 Built-in
11050 Functions::, for details of the functions enabled and disabled by
11053 To have SSE/SSE2 instructions generated automatically from
11054 floating-point code (as opposed to 387 instructions), see
11057 GCC depresses SSEx instructions when `-mavx' is used. Instead, it
11058 generates new AVX instructions or AVX equivalence for all SSEx
11059 instructions when needed.
11061 These options will enable GCC to use these extended instructions in
11062 generated code, even without `-mfpmath=sse'. Applications which
11063 perform runtime CPU detection must compile separate files for each
11064 supported architecture, using the appropriate flags. In
11065 particular, the file containing the CPU detection code should be
11066 compiled without these options.
11069 This option instructs GCC to emit a `cld' instruction in the
11070 prologue of functions that use string instructions. String
11071 instructions depend on the DF flag to select between autoincrement
11072 or autodecrement mode. While the ABI specifies the DF flag to be
11073 cleared on function entry, some operating systems violate this
11074 specification by not clearing the DF flag in their exception
11075 dispatchers. The exception handler can be invoked with the DF flag
11076 set which leads to wrong direction mode, when string instructions
11077 are used. This option can be enabled by default on 32-bit x86
11078 targets by configuring GCC with the `--enable-cld' configure
11079 option. Generation of `cld' instructions can be suppressed with
11080 the `-mno-cld' compiler option in this case.
11083 This option will enable GCC to use CMPXCHG16B instruction in
11084 generated code. CMPXCHG16B allows for atomic operations on
11085 128-bit double quadword (or oword) data types. This is useful for
11086 high resolution counters that could be updated by multiple
11087 processors (or cores). This instruction is generated as part of
11088 atomic built-in functions: see *note Atomic Builtins:: for details.
11091 This option will enable GCC to use SAHF instruction in generated
11092 64-bit code. Early Intel CPUs with Intel 64 lacked LAHF and SAHF
11093 instructions supported by AMD64 until introduction of Pentium 4 G1
11094 step in December 2005. LAHF and SAHF are load and store
11095 instructions, respectively, for certain status flags. In 64-bit
11096 mode, SAHF instruction is used to optimize `fmod', `drem' or
11097 `remainder' built-in functions: see *note Other Builtins:: for
11101 This option will enable GCC to use RCPSS and RSQRTSS instructions
11102 (and their vectorized variants RCPPS and RSQRTPS) with an
11103 additional Newton-Raphson step to increase precision instead of
11104 DIVSS and SQRTSS (and their vectorized variants) for single
11105 precision floating point arguments. These instructions are
11106 generated only when `-funsafe-math-optimizations' is enabled
11107 together with `-finite-math-only' and `-fno-trapping-math'. Note
11108 that while the throughput of the sequence is higher than the
11109 throughput of the non-reciprocal instruction, the precision of the
11110 sequence can be decreased by up to 2 ulp (i.e. the inverse of 1.0
11111 equals 0.99999994).
11114 Specifies the ABI type to use for vectorizing intrinsics using an
11115 external library. Supported types are `svml' for the Intel short
11116 vector math library and `acml' for the AMD math core library style
11117 of interfacing. GCC will currently emit calls to `vmldExp2',
11118 `vmldLn2', `vmldLog102', `vmldLog102', `vmldPow2', `vmldTanh2',
11119 `vmldTan2', `vmldAtan2', `vmldAtanh2', `vmldCbrt2', `vmldSinh2',
11120 `vmldSin2', `vmldAsinh2', `vmldAsin2', `vmldCosh2', `vmldCos2',
11121 `vmldAcosh2', `vmldAcos2', `vmlsExp4', `vmlsLn4', `vmlsLog104',
11122 `vmlsLog104', `vmlsPow4', `vmlsTanh4', `vmlsTan4', `vmlsAtan4',
11123 `vmlsAtanh4', `vmlsCbrt4', `vmlsSinh4', `vmlsSin4', `vmlsAsinh4',
11124 `vmlsAsin4', `vmlsCosh4', `vmlsCos4', `vmlsAcosh4' and `vmlsAcos4'
11125 for corresponding function type when `-mveclibabi=svml' is used
11126 and `__vrd2_sin', `__vrd2_cos', `__vrd2_exp', `__vrd2_log',
11127 `__vrd2_log2', `__vrd2_log10', `__vrs4_sinf', `__vrs4_cosf',
11128 `__vrs4_expf', `__vrs4_logf', `__vrs4_log2f', `__vrs4_log10f' and
11129 `__vrs4_powf' for corresponding function type when
11130 `-mveclibabi=acml' is used. Both `-ftree-vectorize' and
11131 `-funsafe-math-optimizations' have to be enabled. A SVML or ACML
11132 ABI compatible library will have to be specified at link time.
11136 Use PUSH operations to store outgoing parameters. This method is
11137 shorter and usually equally fast as method using SUB/MOV
11138 operations and is enabled by default. In some cases disabling it
11139 may improve performance because of improved scheduling and reduced
11142 `-maccumulate-outgoing-args'
11143 If enabled, the maximum amount of space required for outgoing
11144 arguments will be computed in the function prologue. This is
11145 faster on most modern CPUs because of reduced dependencies,
11146 improved scheduling and reduced stack usage when preferred stack
11147 boundary is not equal to 2. The drawback is a notable increase in
11148 code size. This switch implies `-mno-push-args'.
11151 Support thread-safe exception handling on `Mingw32'. Code that
11152 relies on thread-safe exception handling must compile and link all
11153 code with the `-mthreads' option. When compiling, `-mthreads'
11154 defines `-D_MT'; when linking, it links in a special thread helper
11155 library `-lmingwthrd' which cleans up per thread exception
11158 `-mno-align-stringops'
11159 Do not align destination of inlined string operations. This
11160 switch reduces code size and improves performance in case the
11161 destination is already aligned, but GCC doesn't know about it.
11163 `-minline-all-stringops'
11164 By default GCC inlines string operations only when destination is
11165 known to be aligned at least to 4 byte boundary. This enables
11166 more inlining, increase code size, but may improve performance of
11167 code that depends on fast memcpy, strlen and memset for short
11170 `-minline-stringops-dynamically'
11171 For string operation of unknown size, inline runtime checks so for
11172 small blocks inline code is used, while for large blocks library
11175 `-mstringop-strategy=ALG'
11176 Overwrite internal decision heuristic about particular algorithm
11177 to inline string operation with. The allowed values are
11178 `rep_byte', `rep_4byte', `rep_8byte' for expanding using i386
11179 `rep' prefix of specified size, `byte_loop', `loop',
11180 `unrolled_loop' for expanding inline loop, `libcall' for always
11181 expanding library call.
11183 `-momit-leaf-frame-pointer'
11184 Don't keep the frame pointer in a register for leaf functions.
11185 This avoids the instructions to save, set up and restore frame
11186 pointers and makes an extra register available in leaf functions.
11187 The option `-fomit-frame-pointer' removes the frame pointer for
11188 all functions which might make debugging harder.
11190 `-mtls-direct-seg-refs'
11191 `-mno-tls-direct-seg-refs'
11192 Controls whether TLS variables may be accessed with offsets from
11193 the TLS segment register (`%gs' for 32-bit, `%fs' for 64-bit), or
11194 whether the thread base pointer must be added. Whether or not this
11195 is legal depends on the operating system, and whether it maps the
11196 segment to cover the entire TLS area.
11198 For systems that use GNU libc, the default is on.
11202 Enable automatic generation of fused floating point multiply-add
11203 instructions if the ISA supports such instructions. The
11204 -mfused-madd option is on by default. The fused multiply-add
11205 instructions have a different rounding behavior compared to
11206 executing a multiply followed by an add.
11210 Specify that the assembler should encode SSE instructions with VEX
11211 prefix. The option `-mavx' turns this on by default.
11213 These `-m' switches are supported in addition to the above on AMD
11214 x86-64 processors in 64-bit environments.
11218 Generate code for a 32-bit or 64-bit environment. The 32-bit
11219 environment sets int, long and pointer to 32 bits and generates
11220 code that runs on any i386 system. The 64-bit environment sets
11221 int to 32 bits and long and pointer to 64 bits and generates code
11222 for AMD's x86-64 architecture. For darwin only the -m64 option
11223 turns off the `-fno-pic' and `-mdynamic-no-pic' options.
11226 Do not use a so called red zone for x86-64 code. The red zone is
11227 mandated by the x86-64 ABI, it is a 128-byte area beyond the
11228 location of the stack pointer that will not be modified by signal
11229 or interrupt handlers and therefore can be used for temporary data
11230 without adjusting the stack pointer. The flag `-mno-red-zone'
11231 disables this red zone.
11234 Generate code for the small code model: the program and its
11235 symbols must be linked in the lower 2 GB of the address space.
11236 Pointers are 64 bits. Programs can be statically or dynamically
11237 linked. This is the default code model.
11240 Generate code for the kernel code model. The kernel runs in the
11241 negative 2 GB of the address space. This model has to be used for
11245 Generate code for the medium model: The program is linked in the
11246 lower 2 GB of the address space. Small symbols are also placed
11247 there. Symbols with sizes larger than `-mlarge-data-threshold'
11248 are put into large data or bss sections and can be located above
11249 2GB. Programs can be statically or dynamically linked.
11252 Generate code for the large model: This model makes no assumptions
11253 about addresses and sizes of sections.
11256 File: gcc.info, Node: IA-64 Options, Next: M32C Options, Prev: i386 and x86-64 Windows Options, Up: Submodel Options
11258 3.17.16 IA-64 Options
11259 ---------------------
11261 These are the `-m' options defined for the Intel IA-64 architecture.
11264 Generate code for a big endian target. This is the default for
11268 Generate code for a little endian target. This is the default for
11269 AIX5 and GNU/Linux.
11273 Generate (or don't) code for the GNU assembler. This is the
11278 Generate (or don't) code for the GNU linker. This is the default.
11281 Generate code that does not use a global pointer register. The
11282 result is not position independent code, and violates the IA-64
11285 `-mvolatile-asm-stop'
11286 `-mno-volatile-asm-stop'
11287 Generate (or don't) a stop bit immediately before and after
11288 volatile asm statements.
11291 `-mno-register-names'
11292 Generate (or don't) `in', `loc', and `out' register names for the
11293 stacked registers. This may make assembler output more readable.
11297 Disable (or enable) optimizations that use the small data section.
11298 This may be useful for working around optimizer bugs.
11301 Generate code that uses a single constant global pointer value.
11302 This is useful when compiling kernel code.
11305 Generate code that is self-relocatable. This implies
11306 `-mconstant-gp'. This is useful when compiling firmware code.
11308 `-minline-float-divide-min-latency'
11309 Generate code for inline divides of floating point values using
11310 the minimum latency algorithm.
11312 `-minline-float-divide-max-throughput'
11313 Generate code for inline divides of floating point values using
11314 the maximum throughput algorithm.
11316 `-minline-int-divide-min-latency'
11317 Generate code for inline divides of integer values using the
11318 minimum latency algorithm.
11320 `-minline-int-divide-max-throughput'
11321 Generate code for inline divides of integer values using the
11322 maximum throughput algorithm.
11324 `-minline-sqrt-min-latency'
11325 Generate code for inline square roots using the minimum latency
11328 `-minline-sqrt-max-throughput'
11329 Generate code for inline square roots using the maximum throughput
11334 Don't (or do) generate assembler code for the DWARF2 line number
11335 debugging info. This may be useful when not using the GNU
11338 `-mearly-stop-bits'
11339 `-mno-early-stop-bits'
11340 Allow stop bits to be placed earlier than immediately preceding the
11341 instruction that triggered the stop bit. This can improve
11342 instruction scheduling, but does not always do so.
11344 `-mfixed-range=REGISTER-RANGE'
11345 Generate code treating the given register range as fixed registers.
11346 A fixed register is one that the register allocator can not use.
11347 This is useful when compiling kernel code. A register range is
11348 specified as two registers separated by a dash. Multiple register
11349 ranges can be specified separated by a comma.
11351 `-mtls-size=TLS-SIZE'
11352 Specify bit size of immediate TLS offsets. Valid values are 14,
11356 Tune the instruction scheduling for a particular CPU, Valid values
11357 are itanium, itanium1, merced, itanium2, and mckinley.
11361 Add support for multithreading using the POSIX threads library.
11362 This option sets flags for both the preprocessor and linker. It
11363 does not affect the thread safety of object code produced by the
11364 compiler or that of libraries supplied with it. These are HP-UX
11369 Generate code for a 32-bit or 64-bit environment. The 32-bit
11370 environment sets int, long and pointer to 32 bits. The 64-bit
11371 environment sets int to 32 bits and long and pointer to 64 bits.
11372 These are HP-UX specific flags.
11374 `-mno-sched-br-data-spec'
11375 `-msched-br-data-spec'
11376 (Dis/En)able data speculative scheduling before reload. This will
11377 result in generation of the ld.a instructions and the
11378 corresponding check instructions (ld.c / chk.a). The default is
11381 `-msched-ar-data-spec'
11382 `-mno-sched-ar-data-spec'
11383 (En/Dis)able data speculative scheduling after reload. This will
11384 result in generation of the ld.a instructions and the
11385 corresponding check instructions (ld.c / chk.a). The default is
11388 `-mno-sched-control-spec'
11389 `-msched-control-spec'
11390 (Dis/En)able control speculative scheduling. This feature is
11391 available only during region scheduling (i.e. before reload).
11392 This will result in generation of the ld.s instructions and the
11393 corresponding check instructions chk.s . The default is 'disable'.
11395 `-msched-br-in-data-spec'
11396 `-mno-sched-br-in-data-spec'
11397 (En/Dis)able speculative scheduling of the instructions that are
11398 dependent on the data speculative loads before reload. This is
11399 effective only with `-msched-br-data-spec' enabled. The default
11402 `-msched-ar-in-data-spec'
11403 `-mno-sched-ar-in-data-spec'
11404 (En/Dis)able speculative scheduling of the instructions that are
11405 dependent on the data speculative loads after reload. This is
11406 effective only with `-msched-ar-data-spec' enabled. The default
11409 `-msched-in-control-spec'
11410 `-mno-sched-in-control-spec'
11411 (En/Dis)able speculative scheduling of the instructions that are
11412 dependent on the control speculative loads. This is effective
11413 only with `-msched-control-spec' enabled. The default is 'enable'.
11417 (En/Dis)able use of simple data speculation checks ld.c . If
11418 disabled, only chk.a instructions will be emitted to check data
11419 speculative loads. The default is 'enable'.
11421 `-mno-sched-control-ldc'
11422 `-msched-control-ldc'
11423 (Dis/En)able use of ld.c instructions to check control speculative
11424 loads. If enabled, in case of control speculative load with no
11425 speculatively scheduled dependent instructions this load will be
11426 emitted as ld.sa and ld.c will be used to check it. The default
11429 `-mno-sched-spec-verbose'
11430 `-msched-spec-verbose'
11431 (Dis/En)able printing of the information about speculative motions.
11433 `-mno-sched-prefer-non-data-spec-insns'
11434 `-msched-prefer-non-data-spec-insns'
11435 If enabled, data speculative instructions will be chosen for
11436 schedule only if there are no other choices at the moment. This
11437 will make the use of the data speculation much more conservative.
11438 The default is 'disable'.
11440 `-mno-sched-prefer-non-control-spec-insns'
11441 `-msched-prefer-non-control-spec-insns'
11442 If enabled, control speculative instructions will be chosen for
11443 schedule only if there are no other choices at the moment. This
11444 will make the use of the control speculation much more
11445 conservative. The default is 'disable'.
11447 `-mno-sched-count-spec-in-critical-path'
11448 `-msched-count-spec-in-critical-path'
11449 If enabled, speculative dependencies will be considered during
11450 computation of the instructions priorities. This will make the
11451 use of the speculation a bit more conservative. The default is
11456 File: gcc.info, Node: M32C Options, Next: M32R/D Options, Prev: IA-64 Options, Up: Submodel Options
11458 3.17.17 M32C Options
11459 --------------------
11462 Select the CPU for which code is generated. NAME may be one of
11463 `r8c' for the R8C/Tiny series, `m16c' for the M16C (up to /60)
11464 series, `m32cm' for the M16C/80 series, or `m32c' for the M32C/80
11468 Specifies that the program will be run on the simulator. This
11469 causes an alternate runtime library to be linked in which
11470 supports, for example, file I/O. You must not use this option
11471 when generating programs that will run on real hardware; you must
11472 provide your own runtime library for whatever I/O functions are
11476 Specifies the number of memory-based pseudo-registers GCC will use
11477 during code generation. These pseudo-registers will be used like
11478 real registers, so there is a tradeoff between GCC's ability to
11479 fit the code into available registers, and the performance penalty
11480 of using memory instead of registers. Note that all modules in a
11481 program must be compiled with the same value for this option.
11482 Because of that, you must not use this option with the default
11483 runtime libraries gcc builds.
11487 File: gcc.info, Node: M32R/D Options, Next: M680x0 Options, Prev: M32C Options, Up: Submodel Options
11489 3.17.18 M32R/D Options
11490 ----------------------
11492 These `-m' options are defined for Renesas M32R/D architectures:
11495 Generate code for the M32R/2.
11498 Generate code for the M32R/X.
11501 Generate code for the M32R. This is the default.
11504 Assume all objects live in the lower 16MB of memory (so that their
11505 addresses can be loaded with the `ld24' instruction), and assume
11506 all subroutines are reachable with the `bl' instruction. This is
11509 The addressability of a particular object can be set with the
11513 Assume objects may be anywhere in the 32-bit address space (the
11514 compiler will generate `seth/add3' instructions to load their
11515 addresses), and assume all subroutines are reachable with the `bl'
11519 Assume objects may be anywhere in the 32-bit address space (the
11520 compiler will generate `seth/add3' instructions to load their
11521 addresses), and assume subroutines may not be reachable with the
11522 `bl' instruction (the compiler will generate the much slower
11523 `seth/add3/jl' instruction sequence).
11526 Disable use of the small data area. Variables will be put into
11527 one of `.data', `bss', or `.rodata' (unless the `section'
11528 attribute has been specified). This is the default.
11530 The small data area consists of sections `.sdata' and `.sbss'.
11531 Objects may be explicitly put in the small data area with the
11532 `section' attribute using one of these sections.
11535 Put small global and static data in the small data area, but do not
11536 generate special code to reference them.
11539 Put small global and static data in the small data area, and
11540 generate special instructions to reference them.
11543 Put global and static objects less than or equal to NUM bytes into
11544 the small data or bss sections instead of the normal data or bss
11545 sections. The default value of NUM is 8. The `-msdata' option
11546 must be set to one of `sdata' or `use' for this option to have any
11549 All modules should be compiled with the same `-G NUM' value.
11550 Compiling with different values of NUM may or may not work; if it
11551 doesn't the linker will give an error message--incorrect code will
11555 Makes the M32R specific code in the compiler display some
11556 statistics that might help in debugging programs.
11559 Align all loops to a 32-byte boundary.
11562 Do not enforce a 32-byte alignment for loops. This is the default.
11564 `-missue-rate=NUMBER'
11565 Issue NUMBER instructions per cycle. NUMBER can only be 1 or 2.
11567 `-mbranch-cost=NUMBER'
11568 NUMBER can only be 1 or 2. If it is 1 then branches will be
11569 preferred over conditional code, if it is 2, then the opposite will
11572 `-mflush-trap=NUMBER'
11573 Specifies the trap number to use to flush the cache. The default
11574 is 12. Valid numbers are between 0 and 15 inclusive.
11577 Specifies that the cache cannot be flushed by using a trap.
11579 `-mflush-func=NAME'
11580 Specifies the name of the operating system function to call to
11581 flush the cache. The default is __flush_cache_, but a function
11582 call will only be used if a trap is not available.
11585 Indicates that there is no OS function for flushing the cache.
11589 File: gcc.info, Node: M680x0 Options, Next: M68hc1x Options, Prev: M32R/D Options, Up: Submodel Options
11591 3.17.19 M680x0 Options
11592 ----------------------
11594 These are the `-m' options defined for M680x0 and ColdFire processors.
11595 The default settings depend on which architecture was selected when the
11596 compiler was configured; the defaults for the most common choices are
11600 Generate code for a specific M680x0 or ColdFire instruction set
11601 architecture. Permissible values of ARCH for M680x0 architectures
11602 are: `68000', `68010', `68020', `68030', `68040', `68060' and
11603 `cpu32'. ColdFire architectures are selected according to
11604 Freescale's ISA classification and the permissible values are:
11605 `isaa', `isaaplus', `isab' and `isac'.
11607 gcc defines a macro `__mcfARCH__' whenever it is generating code
11608 for a ColdFire target. The ARCH in this macro is one of the
11609 `-march' arguments given above.
11611 When used together, `-march' and `-mtune' select code that runs on
11612 a family of similar processors but that is optimized for a
11613 particular microarchitecture.
11616 Generate code for a specific M680x0 or ColdFire processor. The
11617 M680x0 CPUs are: `68000', `68010', `68020', `68030', `68040',
11618 `68060', `68302', `68332' and `cpu32'. The ColdFire CPUs are
11619 given by the table below, which also classifies the CPUs into
11622 *Family* *`-mcpu' arguments*
11624 `5206' `5202' `5204' `5206'
11626 `5208' `5207' `5208'
11627 `5211a' `5210a' `5211a'
11628 `5213' `5211' `5212' `5213'
11629 `5216' `5214' `5216'
11630 `52235' `52230' `52231' `52232' `52233' `52234' `52235'
11631 `5225' `5224' `5225'
11632 `5235' `5232' `5233' `5234' `5235' `523x'
11635 `5271' `5270' `5271'
11637 `5275' `5274' `5275'
11638 `5282' `5280' `5281' `5282' `528x'
11640 `5329' `5327' `5328' `5329' `532x'
11641 `5373' `5372' `5373' `537x'
11643 `5475' `5470' `5471' `5472' `5473' `5474' `5475' `547x'
11644 `5480' `5481' `5482' `5483' `5484' `5485'
11646 `-mcpu=CPU' overrides `-march=ARCH' if ARCH is compatible with
11647 CPU. Other combinations of `-mcpu' and `-march' are rejected.
11649 gcc defines the macro `__mcf_cpu_CPU' when ColdFire target CPU is
11650 selected. It also defines `__mcf_family_FAMILY', where the value
11651 of FAMILY is given by the table above.
11654 Tune the code for a particular microarchitecture, within the
11655 constraints set by `-march' and `-mcpu'. The M680x0
11656 microarchitectures are: `68000', `68010', `68020', `68030',
11657 `68040', `68060' and `cpu32'. The ColdFire microarchitectures
11658 are: `cfv1', `cfv2', `cfv3', `cfv4' and `cfv4e'.
11660 You can also use `-mtune=68020-40' for code that needs to run
11661 relatively well on 68020, 68030 and 68040 targets.
11662 `-mtune=68020-60' is similar but includes 68060 targets as well.
11663 These two options select the same tuning decisions as `-m68020-40'
11664 and `-m68020-60' respectively.
11666 gcc defines the macros `__mcARCH' and `__mcARCH__' when tuning for
11667 680x0 architecture ARCH. It also defines `mcARCH' unless either
11668 `-ansi' or a non-GNU `-std' option is used. If gcc is tuning for
11669 a range of architectures, as selected by `-mtune=68020-40' or
11670 `-mtune=68020-60', it defines the macros for every architecture in
11673 gcc also defines the macro `__mUARCH__' when tuning for ColdFire
11674 microarchitecture UARCH, where UARCH is one of the arguments given
11679 Generate output for a 68000. This is the default when the
11680 compiler is configured for 68000-based systems. It is equivalent
11683 Use this option for microcontrollers with a 68000 or EC000 core,
11684 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11687 Generate output for a 68010. This is the default when the
11688 compiler is configured for 68010-based systems. It is equivalent
11693 Generate output for a 68020. This is the default when the
11694 compiler is configured for 68020-based systems. It is equivalent
11698 Generate output for a 68030. This is the default when the
11699 compiler is configured for 68030-based systems. It is equivalent
11703 Generate output for a 68040. This is the default when the
11704 compiler is configured for 68040-based systems. It is equivalent
11707 This option inhibits the use of 68881/68882 instructions that have
11708 to be emulated by software on the 68040. Use this option if your
11709 68040 does not have code to emulate those instructions.
11712 Generate output for a 68060. This is the default when the
11713 compiler is configured for 68060-based systems. It is equivalent
11716 This option inhibits the use of 68020 and 68881/68882 instructions
11717 that have to be emulated by software on the 68060. Use this
11718 option if your 68060 does not have code to emulate those
11722 Generate output for a CPU32. This is the default when the
11723 compiler is configured for CPU32-based systems. It is equivalent
11726 Use this option for microcontrollers with a CPU32 or CPU32+ core,
11727 including the 68330, 68331, 68332, 68333, 68334, 68336, 68340,
11728 68341, 68349 and 68360.
11731 Generate output for a 520X ColdFire CPU. This is the default when
11732 the compiler is configured for 520X-based systems. It is
11733 equivalent to `-mcpu=5206', and is now deprecated in favor of that
11736 Use this option for microcontroller with a 5200 core, including
11737 the MCF5202, MCF5203, MCF5204 and MCF5206.
11740 Generate output for a 5206e ColdFire CPU. The option is now
11741 deprecated in favor of the equivalent `-mcpu=5206e'.
11744 Generate output for a member of the ColdFire 528X family. The
11745 option is now deprecated in favor of the equivalent `-mcpu=528x'.
11748 Generate output for a ColdFire 5307 CPU. The option is now
11749 deprecated in favor of the equivalent `-mcpu=5307'.
11752 Generate output for a ColdFire 5407 CPU. The option is now
11753 deprecated in favor of the equivalent `-mcpu=5407'.
11756 Generate output for a ColdFire V4e family CPU (e.g. 547x/548x).
11757 This includes use of hardware floating point instructions. The
11758 option is equivalent to `-mcpu=547x', and is now deprecated in
11759 favor of that option.
11762 Generate output for a 68040, without using any of the new
11763 instructions. This results in code which can run relatively
11764 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11765 generated code does use the 68881 instructions that are emulated
11768 The option is equivalent to `-march=68020' `-mtune=68020-40'.
11771 Generate output for a 68060, without using any of the new
11772 instructions. This results in code which can run relatively
11773 efficiently on either a 68020/68881 or a 68030 or a 68040. The
11774 generated code does use the 68881 instructions that are emulated
11777 The option is equivalent to `-march=68020' `-mtune=68020-60'.
11781 Generate floating-point instructions. This is the default for
11782 68020 and above, and for ColdFire devices that have an FPU. It
11783 defines the macro `__HAVE_68881__' on M680x0 targets and
11784 `__mcffpu__' on ColdFire targets.
11787 Do not generate floating-point instructions; use library calls
11788 instead. This is the default for 68000, 68010, and 68832 targets.
11789 It is also the default for ColdFire devices that have no FPU.
11793 Generate (do not generate) ColdFire hardware divide and remainder
11794 instructions. If `-march' is used without `-mcpu', the default is
11795 "on" for ColdFire architectures and "off" for M680x0
11796 architectures. Otherwise, the default is taken from the target CPU
11797 (either the default CPU, or the one specified by `-mcpu'). For
11798 example, the default is "off" for `-mcpu=5206' and "on" for
11801 gcc defines the macro `__mcfhwdiv__' when this option is enabled.
11804 Consider type `int' to be 16 bits wide, like `short int'.
11805 Additionally, parameters passed on the stack are also aligned to a
11806 16-bit boundary even on targets whose API mandates promotion to
11810 Do not consider type `int' to be 16 bits wide. This is the
11815 Do not use the bit-field instructions. The `-m68000', `-mcpu32'
11816 and `-m5200' options imply `-mnobitfield'.
11819 Do use the bit-field instructions. The `-m68020' option implies
11820 `-mbitfield'. This is the default if you use a configuration
11821 designed for a 68020.
11824 Use a different function-calling convention, in which functions
11825 that take a fixed number of arguments return with the `rtd'
11826 instruction, which pops their arguments while returning. This
11827 saves one instruction in the caller since there is no need to pop
11828 the arguments there.
11830 This calling convention is incompatible with the one normally used
11831 on Unix, so you cannot use it if you need to call libraries
11832 compiled with the Unix compiler.
11834 Also, you must provide function prototypes for all functions that
11835 take variable numbers of arguments (including `printf'); otherwise
11836 incorrect code will be generated for calls to those functions.
11838 In addition, seriously incorrect code will result if you call a
11839 function with too many arguments. (Normally, extra arguments are
11840 harmlessly ignored.)
11842 The `rtd' instruction is supported by the 68010, 68020, 68030,
11843 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11846 Do not use the calling conventions selected by `-mrtd'. This is
11851 Control whether GCC aligns `int', `long', `long long', `float',
11852 `double', and `long double' variables on a 32-bit boundary
11853 (`-malign-int') or a 16-bit boundary (`-mno-align-int'). Aligning
11854 variables on 32-bit boundaries produces code that runs somewhat
11855 faster on processors with 32-bit busses at the expense of more
11858 *Warning:* if you use the `-malign-int' switch, GCC will align
11859 structures containing the above types differently than most
11860 published application binary interface specifications for the m68k.
11863 Use the pc-relative addressing mode of the 68000 directly, instead
11864 of using a global offset table. At present, this option implies
11865 `-fpic', allowing at most a 16-bit offset for pc-relative
11866 addressing. `-fPIC' is not presently supported with `-mpcrel',
11867 though this could be supported for 68020 and higher processors.
11869 `-mno-strict-align'
11871 Do not (do) assume that unaligned memory references will be
11872 handled by the system.
11875 Generate code that allows the data segment to be located in a
11876 different area of memory from the text segment. This allows for
11877 execute in place in an environment without virtual memory
11878 management. This option implies `-fPIC'.
11881 Generate code that assumes that the data segment follows the text
11882 segment. This is the default.
11884 `-mid-shared-library'
11885 Generate code that supports shared libraries via the library ID
11886 method. This allows for execute in place and shared libraries in
11887 an environment without virtual memory management. This option
11890 `-mno-id-shared-library'
11891 Generate code that doesn't assume ID based shared libraries are
11892 being used. This is the default.
11894 `-mshared-library-id=n'
11895 Specified the identification number of the ID based shared library
11896 being compiled. Specifying a value of 0 will generate more
11897 compact code, specifying other values will force the allocation of
11898 that number to the current library but is no more space or time
11899 efficient than omitting this option.
11903 When generating position-independent code for ColdFire, generate
11904 code that works if the GOT has more than 8192 entries. This code
11905 is larger and slower than code generated without this option. On
11906 M680x0 processors, this option is not needed; `-fPIC' suffices.
11908 GCC normally uses a single instruction to load values from the GOT.
11909 While this is relatively efficient, it only works if the GOT is
11910 smaller than about 64k. Anything larger causes the linker to
11911 report an error such as:
11913 relocation truncated to fit: R_68K_GOT16O foobar
11915 If this happens, you should recompile your code with `-mxgot'. It
11916 should then work with very large GOTs. However, code generated
11917 with `-mxgot' is less efficient, since it takes 4 instructions to
11918 fetch the value of a global symbol.
11920 Note that some linkers, including newer versions of the GNU linker,
11921 can create multiple GOTs and sort GOT entries. If you have such a
11922 linker, you should only need to use `-mxgot' when compiling a
11923 single object file that accesses more than 8192 GOT entries. Very
11926 These options have no effect unless GCC is generating
11927 position-independent code.
11931 File: gcc.info, Node: M68hc1x Options, Next: MCore Options, Prev: M680x0 Options, Up: Submodel Options
11933 3.17.20 M68hc1x Options
11934 -----------------------
11936 These are the `-m' options defined for the 68hc11 and 68hc12
11937 microcontrollers. The default values for these options depends on
11938 which style of microcontroller was selected when the compiler was
11939 configured; the defaults for the most common choices are given below.
11943 Generate output for a 68HC11. This is the default when the
11944 compiler is configured for 68HC11-based systems.
11948 Generate output for a 68HC12. This is the default when the
11949 compiler is configured for 68HC12-based systems.
11953 Generate output for a 68HCS12.
11956 Enable the use of 68HC12 pre and post auto-increment and
11957 auto-decrement addressing modes.
11961 Enable the use of 68HC12 min and max instructions.
11965 Treat all calls as being far away (near). If calls are assumed to
11966 be far away, the compiler will use the `call' instruction to call
11967 a function and the `rtc' instruction for returning.
11970 Consider type `int' to be 16 bits wide, like `short int'.
11972 `-msoft-reg-count=COUNT'
11973 Specify the number of pseudo-soft registers which are used for the
11974 code generation. The maximum number is 32. Using more pseudo-soft
11975 register may or may not result in better code depending on the
11976 program. The default is 4 for 68HC11 and 2 for 68HC12.
11980 File: gcc.info, Node: MCore Options, Next: MIPS Options, Prev: M68hc1x Options, Up: Submodel Options
11982 3.17.21 MCore Options
11983 ---------------------
11985 These are the `-m' options defined for the Motorola M*Core processors.
11989 Inline constants into the code stream if it can be done in two
11990 instructions or less.
11994 Use the divide instruction. (Enabled by default).
11996 `-mrelax-immediate'
11997 `-mno-relax-immediate'
11998 Allow arbitrary sized immediates in bit operations.
12001 `-mno-wide-bitfields'
12002 Always treat bit-fields as int-sized.
12004 `-m4byte-functions'
12005 `-mno-4byte-functions'
12006 Force all functions to be aligned to a four byte boundary.
12009 `-mno-callgraph-data'
12010 Emit callgraph information.
12014 Prefer word access when reading byte quantities.
12018 Generate code for a little endian target.
12022 Generate code for the 210 processor.
12025 Assume that run-time support has been provided and so omit the
12026 simulator library (`libsim.a)' from the linker command line.
12028 `-mstack-increment=SIZE'
12029 Set the maximum amount for a single stack increment operation.
12030 Large values can increase the speed of programs which contain
12031 functions that need a large amount of stack space, but they can
12032 also trigger a segmentation fault if the stack is extended too
12033 much. The default value is 0x1000.
12037 File: gcc.info, Node: MIPS Options, Next: MMIX Options, Prev: MCore Options, Up: Submodel Options
12039 3.17.22 MIPS Options
12040 --------------------
12043 Generate big-endian code.
12046 Generate little-endian code. This is the default for `mips*el-*-*'
12050 Generate code that will run on ARCH, which can be the name of a
12051 generic MIPS ISA, or the name of a particular processor. The ISA
12052 names are: `mips1', `mips2', `mips3', `mips4', `mips32',
12053 `mips32r2', `mips64' and `mips64r2'. The processor names are:
12054 `4kc', `4km', `4kp', `4ksc', `4kec', `4kem', `4kep', `4ksd',
12055 `5kc', `5kf', `20kc', `24kc', `24kf2_1', `24kf1_1', `24kec',
12056 `24kef2_1', `24kef1_1', `34kc', `34kf2_1', `34kf1_1', `74kc',
12057 `74kf2_1', `74kf1_1', `74kf3_2', `loongson2e', `loongson2f', `m4k',
12058 `octeon', `orion', `r2000', `r3000', `r3900', `r4000', `r4400',
12059 `r4600', `r4650', `r6000', `r8000', `rm7000', `rm9000', `r10000',
12060 `r12000', `r14000', `r16000', `sb1', `sr71000', `vr4100',
12061 `vr4111', `vr4120', `vr4130', `vr4300', `vr5000', `vr5400',
12062 `vr5500' and `xlr'. The special value `from-abi' selects the most
12063 compatible architecture for the selected ABI (that is, `mips1' for
12064 32-bit ABIs and `mips3' for 64-bit ABIs).
12066 Native Linux/GNU toolchains also support the value `native', which
12067 selects the best architecture option for the host processor.
12068 `-march=native' has no effect if GCC does not recognize the
12071 In processor names, a final `000' can be abbreviated as `k' (for
12072 example, `-march=r2k'). Prefixes are optional, and `vr' may be
12075 Names of the form `Nf2_1' refer to processors with FPUs clocked at
12076 half the rate of the core, names of the form `Nf1_1' refer to
12077 processors with FPUs clocked at the same rate as the core, and
12078 names of the form `Nf3_2' refer to processors with FPUs clocked a
12079 ratio of 3:2 with respect to the core. For compatibility reasons,
12080 `Nf' is accepted as a synonym for `Nf2_1' while `Nx' and `Bfx' are
12081 accepted as synonyms for `Nf1_1'.
12083 GCC defines two macros based on the value of this option. The
12084 first is `_MIPS_ARCH', which gives the name of target
12085 architecture, as a string. The second has the form
12086 `_MIPS_ARCH_FOO', where FOO is the capitalized value of
12087 `_MIPS_ARCH'. For example, `-march=r2000' will set `_MIPS_ARCH'
12088 to `"r2000"' and define the macro `_MIPS_ARCH_R2000'.
12090 Note that the `_MIPS_ARCH' macro uses the processor names given
12091 above. In other words, it will have the full prefix and will not
12092 abbreviate `000' as `k'. In the case of `from-abi', the macro
12093 names the resolved architecture (either `"mips1"' or `"mips3"').
12094 It names the default architecture when no `-march' option is given.
12097 Optimize for ARCH. Among other things, this option controls the
12098 way instructions are scheduled, and the perceived cost of
12099 arithmetic operations. The list of ARCH values is the same as for
12102 When this option is not used, GCC will optimize for the processor
12103 specified by `-march'. By using `-march' and `-mtune' together,
12104 it is possible to generate code that will run on a family of
12105 processors, but optimize the code for one particular member of
12108 `-mtune' defines the macros `_MIPS_TUNE' and `_MIPS_TUNE_FOO',
12109 which work in the same way as the `-march' ones described above.
12112 Equivalent to `-march=mips1'.
12115 Equivalent to `-march=mips2'.
12118 Equivalent to `-march=mips3'.
12121 Equivalent to `-march=mips4'.
12124 Equivalent to `-march=mips32'.
12127 Equivalent to `-march=mips32r2'.
12130 Equivalent to `-march=mips64'.
12133 Equivalent to `-march=mips64r2'.
12137 Generate (do not generate) MIPS16 code. If GCC is targetting a
12138 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE.
12140 MIPS16 code generation can also be controlled on a per-function
12141 basis by means of `mips16' and `nomips16' attributes. *Note
12142 Function Attributes::, for more information.
12145 Generate MIPS16 code on alternating functions. This option is
12146 provided for regression testing of mixed MIPS16/non-MIPS16 code
12147 generation, and is not intended for ordinary use in compiling user
12150 `-minterlink-mips16'
12151 `-mno-interlink-mips16'
12152 Require (do not require) that non-MIPS16 code be link-compatible
12155 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12156 it must either use a call or an indirect jump.
12157 `-minterlink-mips16' therefore disables direct jumps unless GCC
12158 knows that the target of the jump is not MIPS16.
12165 Generate code for the given ABI.
12167 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12168 generates 64-bit code when you select a 64-bit architecture, but
12169 you can use `-mgp32' to get 32-bit code instead.
12171 For information about the O64 ABI, see
12172 `http://gcc.gnu.org/projects/mipso64-abi.html'.
12174 GCC supports a variant of the o32 ABI in which floating-point
12175 registers are 64 rather than 32 bits wide. You can select this
12176 combination with `-mabi=32' `-mfp64'. This ABI relies on the
12177 `mthc1' and `mfhc1' instructions and is therefore only supported
12178 for MIPS32R2 processors.
12180 The register assignments for arguments and return values remain the
12181 same, but each scalar value is passed in a single 64-bit register
12182 rather than a pair of 32-bit registers. For example, scalar
12183 floating-point values are returned in `$f0' only, not a
12184 `$f0'/`$f1' pair. The set of call-saved registers also remains
12185 the same, but all 64 bits are saved.
12189 Generate (do not generate) code that is suitable for SVR4-style
12190 dynamic objects. `-mabicalls' is the default for SVR4-based
12195 Generate (do not generate) code that is fully position-independent,
12196 and that can therefore be linked into shared libraries. This
12197 option only affects `-mabicalls'.
12199 All `-mabicalls' code has traditionally been position-independent,
12200 regardless of options like `-fPIC' and `-fpic'. However, as an
12201 extension, the GNU toolchain allows executables to use absolute
12202 accesses for locally-binding symbols. It can also use shorter GP
12203 initialization sequences and generate direct calls to
12204 locally-defined functions. This mode is selected by `-mno-shared'.
12206 `-mno-shared' depends on binutils 2.16 or higher and generates
12207 objects that can only be linked by the GNU linker. However, the
12208 option does not affect the ABI of the final executable; it only
12209 affects the ABI of relocatable objects. Using `-mno-shared' will
12210 generally make executables both smaller and quicker.
12212 `-mshared' is the default.
12216 Assume (do not assume) that the static and dynamic linkers support
12217 PLTs and copy relocations. This option only affects `-mno-shared
12218 -mabicalls'. For the n64 ABI, this option has no effect without
12221 You can make `-mplt' the default by configuring GCC with
12222 `--with-mips-plt'. The default is `-mno-plt' otherwise.
12226 Lift (do not lift) the usual restrictions on the size of the global
12229 GCC normally uses a single instruction to load values from the GOT.
12230 While this is relatively efficient, it will only work if the GOT
12231 is smaller than about 64k. Anything larger will cause the linker
12232 to report an error such as:
12234 relocation truncated to fit: R_MIPS_GOT16 foobar
12236 If this happens, you should recompile your code with `-mxgot'. It
12237 should then work with very large GOTs, although it will also be
12238 less efficient, since it will take three instructions to fetch the
12239 value of a global symbol.
12241 Note that some linkers can create multiple GOTs. If you have such
12242 a linker, you should only need to use `-mxgot' when a single object
12243 file accesses more than 64k's worth of GOT entries. Very few do.
12245 These options have no effect unless GCC is generating position
12249 Assume that general-purpose registers are 32 bits wide.
12252 Assume that general-purpose registers are 64 bits wide.
12255 Assume that floating-point registers are 32 bits wide.
12258 Assume that floating-point registers are 64 bits wide.
12261 Use floating-point coprocessor instructions.
12264 Do not use floating-point coprocessor instructions. Implement
12265 floating-point calculations using library calls instead.
12268 Assume that the floating-point coprocessor only supports
12269 single-precision operations.
12272 Assume that the floating-point coprocessor supports
12273 double-precision operations. This is the default.
12277 Use (do not use) `ll', `sc', and `sync' instructions to implement
12278 atomic memory built-in functions. When neither option is
12279 specified, GCC will use the instructions if the target architecture
12282 `-mllsc' is useful if the runtime environment can emulate the
12283 instructions and `-mno-llsc' can be useful when compiling for
12284 nonstandard ISAs. You can make either option the default by
12285 configuring GCC with `--with-llsc' and `--without-llsc'
12286 respectively. `--with-llsc' is the default for some
12287 configurations; see the installation documentation for details.
12291 Use (do not use) revision 1 of the MIPS DSP ASE. *Note MIPS DSP
12292 Built-in Functions::. This option defines the preprocessor macro
12293 `__mips_dsp'. It also defines `__mips_dsp_rev' to 1.
12297 Use (do not use) revision 2 of the MIPS DSP ASE. *Note MIPS DSP
12298 Built-in Functions::. This option defines the preprocessor macros
12299 `__mips_dsp' and `__mips_dspr2'. It also defines `__mips_dsp_rev'
12304 Use (do not use) the MIPS SmartMIPS ASE.
12307 `-mno-paired-single'
12308 Use (do not use) paired-single floating-point instructions. *Note
12309 MIPS Paired-Single Support::. This option requires hardware
12310 floating-point support to be enabled.
12314 Use (do not use) MIPS Digital Media Extension instructions. This
12315 option can only be used when generating 64-bit code and requires
12316 hardware floating-point support to be enabled.
12320 Use (do not use) the MIPS-3D ASE. *Note MIPS-3D Built-in
12321 Functions::. The option `-mips3d' implies `-mpaired-single'.
12325 Use (do not use) MT Multithreading instructions.
12328 Force `long' types to be 64 bits wide. See `-mlong32' for an
12329 explanation of the default and the way that the pointer size is
12333 Force `long', `int', and pointer types to be 32 bits wide.
12335 The default size of `int's, `long's and pointers depends on the
12336 ABI. All the supported ABIs use 32-bit `int's. The n64 ABI uses
12337 64-bit `long's, as does the 64-bit EABI; the others use 32-bit
12338 `long's. Pointers are the same size as `long's, or the same size
12339 as integer registers, whichever is smaller.
12343 Assume (do not assume) that all symbols have 32-bit values,
12344 regardless of the selected ABI. This option is useful in
12345 combination with `-mabi=64' and `-mno-abicalls' because it allows
12346 GCC to generate shorter and faster references to symbolic
12350 Put definitions of externally-visible data in a small data section
12351 if that data is no bigger than NUM bytes. GCC can then access the
12352 data more efficiently; see `-mgpopt' for details.
12354 The default `-G' option depends on the configuration.
12358 Extend (do not extend) the `-G' behavior to local data too, such
12359 as to static variables in C. `-mlocal-sdata' is the default for
12360 all configurations.
12362 If the linker complains that an application is using too much
12363 small data, you might want to try rebuilding the less
12364 performance-critical parts with `-mno-local-sdata'. You might
12365 also want to build large libraries with `-mno-local-sdata', so
12366 that the libraries leave more room for the main program.
12369 `-mno-extern-sdata'
12370 Assume (do not assume) that externally-defined data will be in a
12371 small data section if that data is within the `-G' limit.
12372 `-mextern-sdata' is the default for all configurations.
12374 If you compile a module MOD with `-mextern-sdata' `-G NUM'
12375 `-mgpopt', and MOD references a variable VAR that is no bigger
12376 than NUM bytes, you must make sure that VAR is placed in a small
12377 data section. If VAR is defined by another module, you must
12378 either compile that module with a high-enough `-G' setting or
12379 attach a `section' attribute to VAR's definition. If VAR is
12380 common, you must link the application with a high-enough `-G'
12383 The easiest way of satisfying these restrictions is to compile and
12384 link every module with the same `-G' option. However, you may
12385 wish to build a library that supports several different small data
12386 limits. You can do this by compiling the library with the highest
12387 supported `-G' setting and additionally using `-mno-extern-sdata'
12388 to stop the library from making assumptions about
12389 externally-defined data.
12393 Use (do not use) GP-relative accesses for symbols that are known
12394 to be in a small data section; see `-G', `-mlocal-sdata' and
12395 `-mextern-sdata'. `-mgpopt' is the default for all configurations.
12397 `-mno-gpopt' is useful for cases where the `$gp' register might
12398 not hold the value of `_gp'. For example, if the code is part of
12399 a library that might be used in a boot monitor, programs that call
12400 boot monitor routines will pass an unknown value in `$gp'. (In
12401 such situations, the boot monitor itself would usually be compiled
12404 `-mno-gpopt' implies `-mno-local-sdata' and `-mno-extern-sdata'.
12407 `-mno-embedded-data'
12408 Allocate variables to the read-only data section first if
12409 possible, then next in the small data section if possible,
12410 otherwise in data. This gives slightly slower code than the
12411 default, but reduces the amount of RAM required when executing,
12412 and thus may be preferred for some embedded systems.
12414 `-muninit-const-in-rodata'
12415 `-mno-uninit-const-in-rodata'
12416 Put uninitialized `const' variables in the read-only data section.
12417 This option is only meaningful in conjunction with
12420 `-mcode-readable=SETTING'
12421 Specify whether GCC may generate code that reads from executable
12422 sections. There are three possible settings:
12424 `-mcode-readable=yes'
12425 Instructions may freely access executable sections. This is
12426 the default setting.
12428 `-mcode-readable=pcrel'
12429 MIPS16 PC-relative load instructions can access executable
12430 sections, but other instructions must not do so. This option
12431 is useful on 4KSc and 4KSd processors when the code TLBs have
12432 the Read Inhibit bit set. It is also useful on processors
12433 that can be configured to have a dual instruction/data SRAM
12434 interface and that, like the M4K, automatically redirect
12435 PC-relative loads to the instruction RAM.
12437 `-mcode-readable=no'
12438 Instructions must not access executable sections. This
12439 option can be useful on targets that are configured to have a
12440 dual instruction/data SRAM interface but that (unlike the
12441 M4K) do not automatically redirect PC-relative loads to the
12444 `-msplit-addresses'
12445 `-mno-split-addresses'
12446 Enable (disable) use of the `%hi()' and `%lo()' assembler
12447 relocation operators. This option has been superseded by
12448 `-mexplicit-relocs' but is retained for backwards compatibility.
12450 `-mexplicit-relocs'
12451 `-mno-explicit-relocs'
12452 Use (do not use) assembler relocation operators when dealing with
12453 symbolic addresses. The alternative, selected by
12454 `-mno-explicit-relocs', is to use assembler macros instead.
12456 `-mexplicit-relocs' is the default if GCC was configured to use an
12457 assembler that supports relocation operators.
12459 `-mcheck-zero-division'
12460 `-mno-check-zero-division'
12461 Trap (do not trap) on integer division by zero.
12463 The default is `-mcheck-zero-division'.
12467 MIPS systems check for division by zero by generating either a
12468 conditional trap or a break instruction. Using traps results in
12469 smaller code, but is only supported on MIPS II and later. Also,
12470 some versions of the Linux kernel have a bug that prevents trap
12471 from generating the proper signal (`SIGFPE'). Use
12472 `-mdivide-traps' to allow conditional traps on architectures that
12473 support them and `-mdivide-breaks' to force the use of breaks.
12475 The default is usually `-mdivide-traps', but this can be
12476 overridden at configure time using `--with-divide=breaks'.
12477 Divide-by-zero checks can be completely disabled using
12478 `-mno-check-zero-division'.
12482 Force (do not force) the use of `memcpy()' for non-trivial block
12483 moves. The default is `-mno-memcpy', which allows GCC to inline
12484 most constant-sized copies.
12488 Disable (do not disable) use of the `jal' instruction. Calling
12489 functions using `jal' is more efficient but requires the caller
12490 and callee to be in the same 256 megabyte segment.
12492 This option has no effect on abicalls code. The default is
12497 Enable (disable) use of the `mad', `madu' and `mul' instructions,
12498 as provided by the R4650 ISA.
12502 Enable (disable) use of the floating point multiply-accumulate
12503 instructions, when they are available. The default is
12506 When multiply-accumulate instructions are used, the intermediate
12507 product is calculated to infinite precision and is not subject to
12508 the FCSR Flush to Zero bit. This may be undesirable in some
12512 Tell the MIPS assembler to not run its preprocessor over user
12513 assembler files (with a `.s' suffix) when assembling them.
12517 Work around certain R4000 CPU errata:
12518 - A double-word or a variable shift may give an incorrect
12519 result if executed immediately after starting an integer
12522 - A double-word or a variable shift may give an incorrect
12523 result if executed while an integer multiplication is in
12526 - An integer division may give an incorrect result if started
12527 in a delay slot of a taken branch or a jump.
12531 Work around certain R4400 CPU errata:
12532 - A double-word or a variable shift may give an incorrect
12533 result if executed immediately after starting an integer
12538 Work around certain R10000 errata:
12539 - `ll'/`sc' sequences may not behave atomically on revisions
12540 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
12542 This option can only be used if the target architecture supports
12543 branch-likely instructions. `-mfix-r10000' is the default when
12544 `-march=r10000' is used; `-mno-fix-r10000' is the default
12549 Work around certain VR4120 errata:
12550 - `dmultu' does not always produce the correct result.
12552 - `div' and `ddiv' do not always produce the correct result if
12553 one of the operands is negative.
12554 The workarounds for the division errata rely on special functions
12555 in `libgcc.a'. At present, these functions are only provided by
12556 the `mips64vr*-elf' configurations.
12558 Other VR4120 errata require a nop to be inserted between certain
12559 pairs of instructions. These errata are handled by the assembler,
12563 Work around the VR4130 `mflo'/`mfhi' errata. The workarounds are
12564 implemented by the assembler rather than by GCC, although GCC will
12565 avoid using `mflo' and `mfhi' if the VR4130 `macc', `macchi',
12566 `dmacc' and `dmacchi' instructions are available instead.
12570 Work around certain SB-1 CPU core errata. (This flag currently
12571 works around the SB-1 revision 2 "F1" and "F2" floating point
12574 `-mr10k-cache-barrier=SETTING'
12575 Specify whether GCC should insert cache barriers to avoid the
12576 side-effects of speculation on R10K processors.
12578 In common with many processors, the R10K tries to predict the
12579 outcome of a conditional branch and speculatively executes
12580 instructions from the "taken" branch. It later aborts these
12581 instructions if the predicted outcome was wrong. However, on the
12582 R10K, even aborted instructions can have side effects.
12584 This problem only affects kernel stores and, depending on the
12585 system, kernel loads. As an example, a speculatively-executed
12586 store may load the target memory into cache and mark the cache
12587 line as dirty, even if the store itself is later aborted. If a
12588 DMA operation writes to the same area of memory before the "dirty"
12589 line is flushed, the cached data will overwrite the DMA-ed data.
12590 See the R10K processor manual for a full description, including
12591 other potential problems.
12593 One workaround is to insert cache barrier instructions before
12594 every memory access that might be speculatively executed and that
12595 might have side effects even if aborted.
12596 `-mr10k-cache-barrier=SETTING' controls GCC's implementation of
12597 this workaround. It assumes that aborted accesses to any byte in
12598 the following regions will not have side effects:
12600 1. the memory occupied by the current function's stack frame;
12602 2. the memory occupied by an incoming stack argument;
12604 3. the memory occupied by an object with a link-time-constant
12607 It is the kernel's responsibility to ensure that speculative
12608 accesses to these regions are indeed safe.
12610 If the input program contains a function declaration such as:
12614 then the implementation of `foo' must allow `j foo' and `jal foo'
12615 to be executed speculatively. GCC honors this restriction for
12616 functions it compiles itself. It expects non-GCC functions (such
12617 as hand-written assembly code) to do the same.
12619 The option has three forms:
12621 `-mr10k-cache-barrier=load-store'
12622 Insert a cache barrier before a load or store that might be
12623 speculatively executed and that might have side effects even
12626 `-mr10k-cache-barrier=store'
12627 Insert a cache barrier before a store that might be
12628 speculatively executed and that might have side effects even
12631 `-mr10k-cache-barrier=none'
12632 Disable the insertion of cache barriers. This is the default
12635 `-mflush-func=FUNC'
12637 Specifies the function to call to flush the I and D caches, or to
12638 not call any such function. If called, the function must take the
12639 same arguments as the common `_flush_func()', that is, the address
12640 of the memory range for which the cache is being flushed, the size
12641 of the memory range, and the number 3 (to flush both caches). The
12642 default depends on the target GCC was configured for, but commonly
12643 is either `_flush_func' or `__cpu_flush'.
12646 Set the cost of branches to roughly NUM "simple" instructions.
12647 This cost is only a heuristic and is not guaranteed to produce
12648 consistent results across releases. A zero cost redundantly
12649 selects the default, which is based on the `-mtune' setting.
12652 `-mno-branch-likely'
12653 Enable or disable use of Branch Likely instructions, regardless of
12654 the default for the selected architecture. By default, Branch
12655 Likely instructions may be generated if they are supported by the
12656 selected architecture. An exception is for the MIPS32 and MIPS64
12657 architectures and processors which implement those architectures;
12658 for those, Branch Likely instructions will not be generated by
12659 default because the MIPS32 and MIPS64 architectures specifically
12660 deprecate their use.
12663 `-mno-fp-exceptions'
12664 Specifies whether FP exceptions are enabled. This affects how we
12665 schedule FP instructions for some processors. The default is that
12666 FP exceptions are enabled.
12668 For instance, on the SB-1, if FP exceptions are disabled, and we
12669 are emitting 64-bit code, then we can use both FP pipes.
12670 Otherwise, we can only use one FP pipe.
12673 `-mno-vr4130-align'
12674 The VR4130 pipeline is two-way superscalar, but can only issue two
12675 instructions together if the first one is 8-byte aligned. When
12676 this option is enabled, GCC will align pairs of instructions that
12677 it thinks should execute in parallel.
12679 This option only has an effect when optimizing for the VR4130. It
12680 normally makes code faster, but at the expense of making it bigger.
12681 It is enabled by default at optimization level `-O3'.
12684 File: gcc.info, Node: MMIX Options, Next: MN10300 Options, Prev: MIPS Options, Up: Submodel Options
12686 3.17.23 MMIX Options
12687 --------------------
12689 These options are defined for the MMIX:
12693 Specify that intrinsic library functions are being compiled,
12694 passing all values in registers, no matter the size.
12698 Generate floating-point comparison instructions that compare with
12699 respect to the `rE' epsilon register.
12703 Generate code that passes function parameters and return values
12704 that (in the called function) are seen as registers `$0' and up,
12705 as opposed to the GNU ABI which uses global registers `$231' and
12710 When reading data from memory in sizes shorter than 64 bits, use
12711 (do not use) zero-extending load instructions by default, rather
12712 than sign-extending ones.
12716 Make the result of a division yielding a remainder have the same
12717 sign as the divisor. With the default, `-mno-knuthdiv', the sign
12718 of the remainder follows the sign of the dividend. Both methods
12719 are arithmetically valid, the latter being almost exclusively used.
12721 `-mtoplevel-symbols'
12722 `-mno-toplevel-symbols'
12723 Prepend (do not prepend) a `:' to all global symbols, so the
12724 assembly code can be used with the `PREFIX' assembly directive.
12727 Generate an executable in the ELF format, rather than the default
12728 `mmo' format used by the `mmix' simulator.
12731 `-mno-branch-predict'
12732 Use (do not use) the probable-branch instructions, when static
12733 branch prediction indicates a probable branch.
12736 `-mno-base-addresses'
12737 Generate (do not generate) code that uses _base addresses_. Using
12738 a base address automatically generates a request (handled by the
12739 assembler and the linker) for a constant to be set up in a global
12740 register. The register is used for one or more base address
12741 requests within the range 0 to 255 from the value held in the
12742 register. The generally leads to short and fast code, but the
12743 number of different data items that can be addressed is limited.
12744 This means that a program that uses lots of static data may
12745 require `-mno-base-addresses'.
12749 Force (do not force) generated code to have a single exit point in
12753 File: gcc.info, Node: MN10300 Options, Next: PDP-11 Options, Prev: MMIX Options, Up: Submodel Options
12755 3.17.24 MN10300 Options
12756 -----------------------
12758 These `-m' options are defined for Matsushita MN10300 architectures:
12761 Generate code to avoid bugs in the multiply instructions for the
12762 MN10300 processors. This is the default.
12765 Do not generate code to avoid bugs in the multiply instructions
12766 for the MN10300 processors.
12769 Generate code which uses features specific to the AM33 processor.
12772 Do not generate code which uses features specific to the AM33
12773 processor. This is the default.
12775 `-mreturn-pointer-on-d0'
12776 When generating a function which returns a pointer, return the
12777 pointer in both `a0' and `d0'. Otherwise, the pointer is returned
12778 only in a0, and attempts to call such functions without a prototype
12779 would result in errors. Note that this option is on by default;
12780 use `-mno-return-pointer-on-d0' to disable it.
12783 Do not link in the C run-time initialization object file.
12786 Indicate to the linker that it should perform a relaxation
12787 optimization pass to shorten branches, calls and absolute memory
12788 addresses. This option only has an effect when used on the
12789 command line for the final link step.
12791 This option makes symbolic debugging impossible.
12794 File: gcc.info, Node: PDP-11 Options, Next: picoChip Options, Prev: MN10300 Options, Up: Submodel Options
12796 3.17.25 PDP-11 Options
12797 ----------------------
12799 These options are defined for the PDP-11:
12802 Use hardware FPP floating point. This is the default. (FIS
12803 floating point on the PDP-11/40 is not supported.)
12806 Do not use hardware floating point.
12809 Return floating-point results in ac0 (fr0 in Unix assembler
12813 Return floating-point results in memory. This is the default.
12816 Generate code for a PDP-11/40.
12819 Generate code for a PDP-11/45. This is the default.
12822 Generate code for a PDP-11/10.
12825 Use inline `movmemhi' patterns for copying memory. This is the
12829 Do not use inline `movmemhi' patterns for copying memory.
12833 Use 16-bit `int'. This is the default.
12841 Use 64-bit `float'. This is the default.
12845 Use 32-bit `float'.
12848 Use `abshi2' pattern. This is the default.
12851 Do not use `abshi2' pattern.
12853 `-mbranch-expensive'
12854 Pretend that branches are expensive. This is for experimenting
12855 with code generation only.
12858 Do not pretend that branches are expensive. This is the default.
12861 Generate code for a system with split I&D.
12864 Generate code for a system without split I&D. This is the default.
12867 Use Unix assembler syntax. This is the default when configured for
12871 Use DEC assembler syntax. This is the default when configured for
12872 any PDP-11 target other than `pdp11-*-bsd'.
12875 File: gcc.info, Node: picoChip Options, Next: PowerPC Options, Prev: PDP-11 Options, Up: Submodel Options
12877 3.17.26 picoChip Options
12878 ------------------------
12880 These `-m' options are defined for picoChip implementations:
12883 Set the instruction set, register set, and instruction scheduling
12884 parameters for array element type AE_TYPE. Supported values for
12885 AE_TYPE are `ANY', `MUL', and `MAC'.
12887 `-mae=ANY' selects a completely generic AE type. Code generated
12888 with this option will run on any of the other AE types. The code
12889 will not be as efficient as it would be if compiled for a specific
12890 AE type, and some types of operation (e.g., multiplication) will
12891 not work properly on all types of AE.
12893 `-mae=MUL' selects a MUL AE type. This is the most useful AE type
12894 for compiled code, and is the default.
12896 `-mae=MAC' selects a DSP-style MAC AE. Code compiled with this
12897 option may suffer from poor performance of byte (char)
12898 manipulation, since the DSP AE does not provide hardware support
12899 for byte load/stores.
12901 `-msymbol-as-address'
12902 Enable the compiler to directly use a symbol name as an address in
12903 a load/store instruction, without first loading it into a
12904 register. Typically, the use of this option will generate larger
12905 programs, which run faster than when the option isn't used.
12906 However, the results vary from program to program, so it is left
12907 as a user option, rather than being permanently enabled.
12909 `-mno-inefficient-warnings'
12910 Disables warnings about the generation of inefficient code. These
12911 warnings can be generated, for example, when compiling code which
12912 performs byte-level memory operations on the MAC AE type. The MAC
12913 AE has no hardware support for byte-level memory operations, so
12914 all byte load/stores must be synthesized from word load/store
12915 operations. This is inefficient and a warning will be generated
12916 indicating to the programmer that they should rewrite the code to
12917 avoid byte operations, or to target an AE type which has the
12918 necessary hardware support. This option enables the warning to be
12923 File: gcc.info, Node: PowerPC Options, Next: RS/6000 and PowerPC Options, Prev: picoChip Options, Up: Submodel Options
12925 3.17.27 PowerPC Options
12926 -----------------------
12928 These are listed under *Note RS/6000 and PowerPC Options::.
12931 File: gcc.info, Node: RS/6000 and PowerPC Options, Next: S/390 and zSeries Options, Prev: PowerPC Options, Up: Submodel Options
12933 3.17.28 IBM RS/6000 and PowerPC Options
12934 ---------------------------------------
12936 These `-m' options are defined for the IBM RS/6000 and PowerPC:
12944 `-mno-powerpc-gpopt'
12946 `-mno-powerpc-gfxopt'
12961 GCC supports two related instruction set architectures for the
12962 RS/6000 and PowerPC. The "POWER" instruction set are those
12963 instructions supported by the `rios' chip set used in the original
12964 RS/6000 systems and the "PowerPC" instruction set is the
12965 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx
12966 microprocessors, and the IBM 4xx, 6xx, and follow-on
12969 Neither architecture is a subset of the other. However there is a
12970 large common subset of instructions supported by both. An MQ
12971 register is included in processors supporting the POWER
12974 You use these options to specify which instructions are available
12975 on the processor you are using. The default value of these
12976 options is determined when configuring GCC. Specifying the
12977 `-mcpu=CPU_TYPE' overrides the specification of these options. We
12978 recommend you use the `-mcpu=CPU_TYPE' option rather than the
12979 options listed above.
12981 The `-mpower' option allows GCC to generate instructions that are
12982 found only in the POWER architecture and to use the MQ register.
12983 Specifying `-mpower2' implies `-power' and also allows GCC to
12984 generate instructions that are present in the POWER2 architecture
12985 but not the original POWER architecture.
12987 The `-mpowerpc' option allows GCC to generate instructions that
12988 are found only in the 32-bit subset of the PowerPC architecture.
12989 Specifying `-mpowerpc-gpopt' implies `-mpowerpc' and also allows
12990 GCC to use the optional PowerPC architecture instructions in the
12991 General Purpose group, including floating-point square root.
12992 Specifying `-mpowerpc-gfxopt' implies `-mpowerpc' and also allows
12993 GCC to use the optional PowerPC architecture instructions in the
12994 Graphics group, including floating-point select.
12996 The `-mmfcrf' option allows GCC to generate the move from
12997 condition register field instruction implemented on the POWER4
12998 processor and other processors that support the PowerPC V2.01
12999 architecture. The `-mpopcntb' option allows GCC to generate the
13000 popcount and double precision FP reciprocal estimate instruction
13001 implemented on the POWER5 processor and other processors that
13002 support the PowerPC V2.02 architecture. The `-mfprnd' option
13003 allows GCC to generate the FP round to integer instructions
13004 implemented on the POWER5+ processor and other processors that
13005 support the PowerPC V2.03 architecture. The `-mcmpb' option
13006 allows GCC to generate the compare bytes instruction implemented
13007 on the POWER6 processor and other processors that support the
13008 PowerPC V2.05 architecture. The `-mmfpgpr' option allows GCC to
13009 generate the FP move to/from general purpose register instructions
13010 implemented on the POWER6X processor and other processors that
13011 support the extended PowerPC V2.05 architecture. The `-mhard-dfp'
13012 option allows GCC to generate the decimal floating point
13013 instructions implemented on some POWER processors.
13015 The `-mpowerpc64' option allows GCC to generate the additional
13016 64-bit instructions that are found in the full PowerPC64
13017 architecture and to treat GPRs as 64-bit, doubleword quantities.
13018 GCC defaults to `-mno-powerpc64'.
13020 If you specify both `-mno-power' and `-mno-powerpc', GCC will use
13021 only the instructions in the common subset of both architectures
13022 plus some special AIX common-mode calls, and will not use the MQ
13023 register. Specifying both `-mpower' and `-mpowerpc' permits GCC
13024 to use any instruction from either architecture and to allow use
13025 of the MQ register; specify this for the Motorola MPC601.
13029 Select which mnemonics to use in the generated assembler code.
13030 With `-mnew-mnemonics', GCC uses the assembler mnemonics defined
13031 for the PowerPC architecture. With `-mold-mnemonics' it uses the
13032 assembler mnemonics defined for the POWER architecture.
13033 Instructions defined in only one architecture have only one
13034 mnemonic; GCC uses that mnemonic irrespective of which of these
13035 options is specified.
13037 GCC defaults to the mnemonics appropriate for the architecture in
13038 use. Specifying `-mcpu=CPU_TYPE' sometimes overrides the value of
13039 these option. Unless you are building a cross-compiler, you
13040 should normally not specify either `-mnew-mnemonics' or
13041 `-mold-mnemonics', but should instead accept the default.
13044 Set architecture type, register usage, choice of mnemonics, and
13045 instruction scheduling parameters for machine type CPU_TYPE.
13046 Supported values for CPU_TYPE are `401', `403', `405', `405fp',
13047 `440', `440fp', `464', `464fp', `505', `601', `602', `603',
13048 `603e', `604', `604e', `620', `630', `740', `7400', `7450', `750',
13049 `801', `821', `823', `860', `970', `8540', `e300c2', `e300c3',
13050 `e500mc', `ec603e', `G3', `G4', `G5', `power', `power2', `power3',
13051 `power4', `power5', `power5+', `power6', `power6x', `power7'
13052 `common', `powerpc', `powerpc64', `rios', `rios1', `rios2', `rsc',
13055 `-mcpu=common' selects a completely generic processor. Code
13056 generated under this option will run on any POWER or PowerPC
13057 processor. GCC will use only the instructions in the common
13058 subset of both architectures, and will not use the MQ register.
13059 GCC assumes a generic processor model for scheduling purposes.
13061 `-mcpu=power', `-mcpu=power2', `-mcpu=powerpc', and
13062 `-mcpu=powerpc64' specify generic POWER, POWER2, pure 32-bit
13063 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13064 types, with an appropriate, generic processor model assumed for
13065 scheduling purposes.
13067 The other options specify a specific processor. Code generated
13068 under those options will run best on that processor, and may not
13069 run at all on others.
13071 The `-mcpu' options automatically enable or disable the following
13074 -maltivec -mfprnd -mhard-float -mmfcrf -mmultiple
13075 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64
13076 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float
13077 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr
13079 The particular options set for any particular CPU will vary between
13080 compiler versions, depending on what setting seems to produce
13081 optimal code for that CPU; it doesn't necessarily reflect the
13082 actual hardware's capabilities. If you wish to set an individual
13083 option to a particular value, you may specify it after the `-mcpu'
13084 option, like `-mcpu=970 -mno-altivec'.
13086 On AIX, the `-maltivec' and `-mpowerpc64' options are not enabled
13087 or disabled by the `-mcpu' option at present because AIX does not
13088 have full support for these options. You may still enable or
13089 disable them individually if you're sure it'll work in your
13093 Set the instruction scheduling parameters for machine type
13094 CPU_TYPE, but do not set the architecture type, register usage, or
13095 choice of mnemonics, as `-mcpu=CPU_TYPE' would. The same values
13096 for CPU_TYPE are used for `-mtune' as for `-mcpu'. If both are
13097 specified, the code generated will use the architecture,
13098 registers, and mnemonics set by `-mcpu', but the scheduling
13099 parameters set by `-mtune'.
13103 Generate code to compute division as reciprocal estimate and
13104 iterative refinement, creating opportunities for increased
13105 throughput. This feature requires: optional PowerPC Graphics
13106 instruction set for single precision and FRE instruction for
13107 double precision, assuming divides cannot generate user-visible
13108 traps, and the domain values not include Infinities, denormals or
13113 Generate code that uses (does not use) AltiVec instructions, and
13114 also enable the use of built-in functions that allow more direct
13115 access to the AltiVec instruction set. You may also need to set
13116 `-mabi=altivec' to adjust the current ABI with AltiVec ABI
13121 Generate VRSAVE instructions when generating AltiVec code.
13123 `-mgen-cell-microcode'
13124 Generate Cell microcode instructions
13126 `-mwarn-cell-microcode'
13127 Warning when a Cell microcode instruction is going to emitted. An
13128 example of a Cell microcode instruction is a variable shift.
13131 Generate code that allows ld and ld.so to build executables and
13132 shared libraries with non-exec .plt and .got sections. This is a
13133 PowerPC 32-bit SYSV ABI option.
13136 Generate code that uses a BSS .plt section that ld.so fills in, and
13137 requires .plt and .got sections that are both writable and
13138 executable. This is a PowerPC 32-bit SYSV ABI option.
13142 This switch enables or disables the generation of ISEL
13146 This switch has been deprecated. Use `-misel' and `-mno-isel'
13151 This switch enables or disables the generation of SPE simd
13156 This switch enables or disables the generation of PAIRED simd
13160 This option has been deprecated. Use `-mspe' and `-mno-spe'
13163 `-mfloat-gprs=YES/SINGLE/DOUBLE/NO'
13165 This switch enables or disables the generation of floating point
13166 operations on the general purpose registers for architectures that
13169 The argument YES or SINGLE enables the use of single-precision
13170 floating point operations.
13172 The argument DOUBLE enables the use of single and double-precision
13173 floating point operations.
13175 The argument NO disables floating point operations on the general
13178 This option is currently only available on the MPC854x.
13182 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13183 targets (including GNU/Linux). The 32-bit environment sets int,
13184 long and pointer to 32 bits and generates code that runs on any
13185 PowerPC variant. The 64-bit environment sets int to 32 bits and
13186 long and pointer to 64 bits, and generates code for PowerPC64, as
13193 Modify generation of the TOC (Table Of Contents), which is created
13194 for every executable file. The `-mfull-toc' option is selected by
13195 default. In that case, GCC will allocate at least one TOC entry
13196 for each unique non-automatic variable reference in your program.
13197 GCC will also place floating-point constants in the TOC. However,
13198 only 16,384 entries are available in the TOC.
13200 If you receive a linker error message that saying you have
13201 overflowed the available TOC space, you can reduce the amount of
13202 TOC space used with the `-mno-fp-in-toc' and `-mno-sum-in-toc'
13203 options. `-mno-fp-in-toc' prevents GCC from putting floating-point
13204 constants in the TOC and `-mno-sum-in-toc' forces GCC to generate
13205 code to calculate the sum of an address and a constant at run-time
13206 instead of putting that sum into the TOC. You may specify one or
13207 both of these options. Each causes GCC to produce very slightly
13208 slower and larger code at the expense of conserving TOC space.
13210 If you still run out of space in the TOC even when you specify
13211 both of these options, specify `-mminimal-toc' instead. This
13212 option causes GCC to make only one TOC entry for every file. When
13213 you specify this option, GCC will produce code that is slower and
13214 larger but which uses extremely little TOC space. You may wish to
13215 use this option only on files that contain less frequently
13220 Enable 64-bit AIX ABI and calling convention: 64-bit pointers,
13221 64-bit `long' type, and the infrastructure needed to support them.
13222 Specifying `-maix64' implies `-mpowerpc64' and `-mpowerpc', while
13223 `-maix32' disables the 64-bit ABI and implies `-mno-powerpc64'.
13224 GCC defaults to `-maix32'.
13228 Produce code that conforms more closely to IBM XL compiler
13229 semantics when using AIX-compatible ABI. Pass floating-point
13230 arguments to prototyped functions beyond the register save area
13231 (RSA) on the stack in addition to argument FPRs. Do not assume
13232 that most significant double in 128-bit long double value is
13233 properly rounded when comparing values and converting to double.
13234 Use XL symbol names for long double support routines.
13236 The AIX calling convention was extended but not initially
13237 documented to handle an obscure K&R C case of calling a function
13238 that takes the address of its arguments with fewer arguments than
13239 declared. IBM XL compilers access floating point arguments which
13240 do not fit in the RSA from the stack when a subroutine is compiled
13241 without optimization. Because always storing floating-point
13242 arguments on the stack is inefficient and rarely needed, this
13243 option is not enabled by default and only is necessary when
13244 calling subroutines compiled by IBM XL compilers without
13248 Support "IBM RS/6000 SP" "Parallel Environment" (PE). Link an
13249 application written to use message passing with special startup
13250 code to enable the application to run. The system must have PE
13251 installed in the standard location (`/usr/lpp/ppe.poe/'), or the
13252 `specs' file must be overridden with the `-specs=' option to
13253 specify the appropriate directory location. The Parallel
13254 Environment does not support threads, so the `-mpe' option and the
13255 `-pthread' option are incompatible.
13259 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13260 `-malign-natural' overrides the ABI-defined alignment of larger
13261 types, such as floating-point doubles, on their natural size-based
13262 boundary. The option `-malign-power' instructs GCC to follow the
13263 ABI-specified alignment rules. GCC defaults to the standard
13264 alignment defined in the ABI.
13266 On 64-bit Darwin, natural alignment is the default, and
13267 `-malign-power' is not supported.
13271 Generate code that does not use (uses) the floating-point register
13272 set. Software floating point emulation is provided if you use the
13273 `-msoft-float' option, and pass the option to GCC when linking.
13277 Generate code for single or double-precision floating point
13278 operations. `-mdouble-float' implies `-msingle-float'.
13281 Do not generate sqrt and div instructions for hardware floating
13285 Specify type of floating point unit. Valid values are SP_LITE
13286 (equivalent to -msingle-float -msimple-fpu), DP_LITE (equivalent
13287 to -mdouble-float -msimple-fpu), SP_FULL (equivalent to
13288 -msingle-float), and DP_FULL (equivalent to -mdouble-float).
13291 Perform optimizations for floating point unit on Xilinx PPC
13296 Generate code that uses (does not use) the load multiple word
13297 instructions and the store multiple word instructions. These
13298 instructions are generated by default on POWER systems, and not
13299 generated on PowerPC systems. Do not use `-mmultiple' on little
13300 endian PowerPC systems, since those instructions do not work when
13301 the processor is in little endian mode. The exceptions are PPC740
13302 and PPC750 which permit the instructions usage in little endian
13307 Generate code that uses (does not use) the load string instructions
13308 and the store string word instructions to save multiple registers
13309 and do small block moves. These instructions are generated by
13310 default on POWER systems, and not generated on PowerPC systems.
13311 Do not use `-mstring' on little endian PowerPC systems, since those
13312 instructions do not work when the processor is in little endian
13313 mode. The exceptions are PPC740 and PPC750 which permit the
13314 instructions usage in little endian mode.
13318 Generate code that uses (does not use) the load or store
13319 instructions that update the base register to the address of the
13320 calculated memory location. These instructions are generated by
13321 default. If you use `-mno-update', there is a small window
13322 between the time that the stack pointer is updated and the address
13323 of the previous frame is stored, which means code that walks the
13324 stack frame across interrupts or signals may get corrupted data.
13326 `-mavoid-indexed-addresses'
13328 `-mno-avoid-indexed-addresses'
13329 Generate code that tries to avoid (not avoid) the use of indexed
13330 load or store instructions. These instructions can incur a
13331 performance penalty on Power6 processors in certain situations,
13332 such as when stepping through large arrays that cross a 16M
13333 boundary. This option is enabled by default when targetting
13334 Power6 and disabled otherwise.
13338 Generate code that uses (does not use) the floating point multiply
13339 and accumulate instructions. These instructions are generated by
13340 default if hardware floating is used.
13344 Generate code that uses (does not use) the half-word multiply and
13345 multiply-accumulate instructions on the IBM 405, 440 and 464
13346 processors. These instructions are generated by default when
13347 targetting those processors.
13351 Generate code that uses (does not use) the string-search `dlmzb'
13352 instruction on the IBM 405, 440 and 464 processors. This
13353 instruction is generated by default when targetting those
13358 On System V.4 and embedded PowerPC systems do not (do) force
13359 structures and unions that contain bit-fields to be aligned to the
13360 base type of the bit-field.
13362 For example, by default a structure containing nothing but 8
13363 `unsigned' bit-fields of length 1 would be aligned to a 4 byte
13364 boundary and have a size of 4 bytes. By using `-mno-bit-align',
13365 the structure would be aligned to a 1 byte boundary and be one
13368 `-mno-strict-align'
13370 On System V.4 and embedded PowerPC systems do not (do) assume that
13371 unaligned memory references will be handled by the system.
13375 On embedded PowerPC systems generate code that allows (does not
13376 allow) the program to be relocated to a different address at
13377 runtime. If you use `-mrelocatable' on any module, all objects
13378 linked together must be compiled with `-mrelocatable' or
13379 `-mrelocatable-lib'.
13381 `-mrelocatable-lib'
13382 `-mno-relocatable-lib'
13383 On embedded PowerPC systems generate code that allows (does not
13384 allow) the program to be relocated to a different address at
13385 runtime. Modules compiled with `-mrelocatable-lib' can be linked
13386 with either modules compiled without `-mrelocatable' and
13387 `-mrelocatable-lib' or with modules compiled with the
13388 `-mrelocatable' options.
13392 On System V.4 and embedded PowerPC systems do not (do) assume that
13393 register 2 contains a pointer to a global area pointing to the
13394 addresses used in the program.
13398 On System V.4 and embedded PowerPC systems compile code for the
13399 processor in little endian mode. The `-mlittle-endian' option is
13400 the same as `-mlittle'.
13404 On System V.4 and embedded PowerPC systems compile code for the
13405 processor in big endian mode. The `-mbig-endian' option is the
13409 On Darwin and Mac OS X systems, compile code so that it is not
13410 relocatable, but that its external references are relocatable. The
13411 resulting code is suitable for applications, but not shared
13414 `-mprioritize-restricted-insns=PRIORITY'
13415 This option controls the priority that is assigned to
13416 dispatch-slot restricted instructions during the second scheduling
13417 pass. The argument PRIORITY takes the value 0/1/2 to assign
13418 NO/HIGHEST/SECOND-HIGHEST priority to dispatch slot restricted
13421 `-msched-costly-dep=DEPENDENCE_TYPE'
13422 This option controls which dependences are considered costly by
13423 the target during instruction scheduling. The argument
13424 DEPENDENCE_TYPE takes one of the following values: NO: no
13425 dependence is costly, ALL: all dependences are costly,
13426 TRUE_STORE_TO_LOAD: a true dependence from store to load is costly,
13427 STORE_TO_LOAD: any dependence from store to load is costly,
13428 NUMBER: any dependence which latency >= NUMBER is costly.
13430 `-minsert-sched-nops=SCHEME'
13431 This option controls which nop insertion scheme will be used during
13432 the second scheduling pass. The argument SCHEME takes one of the
13433 following values: NO: Don't insert nops. PAD: Pad with nops any
13434 dispatch group which has vacant issue slots, according to the
13435 scheduler's grouping. REGROUP_EXACT: Insert nops to force costly
13436 dependent insns into separate groups. Insert exactly as many nops
13437 as needed to force an insn to a new group, according to the
13438 estimated processor grouping. NUMBER: Insert nops to force costly
13439 dependent insns into separate groups. Insert NUMBER nops to force
13440 an insn to a new group.
13443 On System V.4 and embedded PowerPC systems compile code using
13444 calling conventions that adheres to the March 1995 draft of the
13445 System V Application Binary Interface, PowerPC processor
13446 supplement. This is the default unless you configured GCC using
13447 `powerpc-*-eabiaix'.
13450 Specify both `-mcall-sysv' and `-meabi' options.
13452 `-mcall-sysv-noeabi'
13453 Specify both `-mcall-sysv' and `-mno-eabi' options.
13456 On System V.4 and embedded PowerPC systems compile code for the
13457 Solaris operating system.
13460 On System V.4 and embedded PowerPC systems compile code for the
13461 Linux-based GNU system.
13464 On System V.4 and embedded PowerPC systems compile code for the
13465 Hurd-based GNU system.
13468 On System V.4 and embedded PowerPC systems compile code for the
13469 NetBSD operating system.
13471 `-maix-struct-return'
13472 Return all structures in memory (as specified by the AIX ABI).
13474 `-msvr4-struct-return'
13475 Return structures smaller than 8 bytes in registers (as specified
13479 Extend the current ABI with a particular extension, or remove such
13480 extension. Valid values are ALTIVEC, NO-ALTIVEC, SPE, NO-SPE,
13481 IBMLONGDOUBLE, IEEELONGDOUBLE.
13484 Extend the current ABI with SPE ABI extensions. This does not
13485 change the default ABI, instead it adds the SPE ABI extensions to
13489 Disable Booke SPE ABI extensions for the current ABI.
13491 `-mabi=ibmlongdouble'
13492 Change the current ABI to use IBM extended precision long double.
13493 This is a PowerPC 32-bit SYSV ABI option.
13495 `-mabi=ieeelongdouble'
13496 Change the current ABI to use IEEE extended precision long double.
13497 This is a PowerPC 32-bit Linux ABI option.
13501 On System V.4 and embedded PowerPC systems assume that all calls to
13502 variable argument functions are properly prototyped. Otherwise,
13503 the compiler must insert an instruction before every non
13504 prototyped call to set or clear bit 6 of the condition code
13505 register (CR) to indicate whether floating point values were
13506 passed in the floating point registers in case the function takes
13507 a variable arguments. With `-mprototype', only calls to
13508 prototyped variable argument functions will set or clear the bit.
13511 On embedded PowerPC systems, assume that the startup module is
13512 called `sim-crt0.o' and that the standard C libraries are
13513 `libsim.a' and `libc.a'. This is the default for
13514 `powerpc-*-eabisim' configurations.
13517 On embedded PowerPC systems, assume that the startup module is
13518 called `crt0.o' and the standard C libraries are `libmvme.a' and
13522 On embedded PowerPC systems, assume that the startup module is
13523 called `crt0.o' and the standard C libraries are `libads.a' and
13527 On embedded PowerPC systems, assume that the startup module is
13528 called `crt0.o' and the standard C libraries are `libyk.a' and
13532 On System V.4 and embedded PowerPC systems, specify that you are
13533 compiling for a VxWorks system.
13536 On embedded PowerPC systems, set the PPC_EMB bit in the ELF flags
13537 header to indicate that `eabi' extended relocations are used.
13541 On System V.4 and embedded PowerPC systems do (do not) adhere to
13542 the Embedded Applications Binary Interface (eabi) which is a set of
13543 modifications to the System V.4 specifications. Selecting `-meabi'
13544 means that the stack is aligned to an 8 byte boundary, a function
13545 `__eabi' is called to from `main' to set up the eabi environment,
13546 and the `-msdata' option can use both `r2' and `r13' to point to
13547 two separate small data areas. Selecting `-mno-eabi' means that
13548 the stack is aligned to a 16 byte boundary, do not call an
13549 initialization function from `main', and the `-msdata' option will
13550 only use `r13' to point to a single small data area. The `-meabi'
13551 option is on by default if you configured GCC using one of the
13552 `powerpc*-*-eabi*' options.
13555 On System V.4 and embedded PowerPC systems, put small initialized
13556 `const' global and static data in the `.sdata2' section, which is
13557 pointed to by register `r2'. Put small initialized non-`const'
13558 global and static data in the `.sdata' section, which is pointed
13559 to by register `r13'. Put small uninitialized global and static
13560 data in the `.sbss' section, which is adjacent to the `.sdata'
13561 section. The `-msdata=eabi' option is incompatible with the
13562 `-mrelocatable' option. The `-msdata=eabi' option also sets the
13566 On System V.4 and embedded PowerPC systems, put small global and
13567 static data in the `.sdata' section, which is pointed to by
13568 register `r13'. Put small uninitialized global and static data in
13569 the `.sbss' section, which is adjacent to the `.sdata' section.
13570 The `-msdata=sysv' option is incompatible with the `-mrelocatable'
13575 On System V.4 and embedded PowerPC systems, if `-meabi' is used,
13576 compile code the same as `-msdata=eabi', otherwise compile code the
13577 same as `-msdata=sysv'.
13580 On System V.4 and embedded PowerPC systems, put small global data
13581 in the `.sdata' section. Put small uninitialized global data in
13582 the `.sbss' section. Do not use register `r13' to address small
13583 data however. This is the default behavior unless other `-msdata'
13588 On embedded PowerPC systems, put all initialized global and static
13589 data in the `.data' section, and all uninitialized data in the
13593 On embedded PowerPC systems, put global and static items less than
13594 or equal to NUM bytes into the small data or bss sections instead
13595 of the normal data or bss section. By default, NUM is 8. The `-G
13596 NUM' switch is also passed to the linker. All modules should be
13597 compiled with the same `-G NUM' value.
13601 On System V.4 and embedded PowerPC systems do (do not) emit
13602 register names in the assembly language output using symbolic
13607 By default assume that all calls are far away so that a longer more
13608 expensive calling sequence is required. This is required for calls
13609 further than 32 megabytes (33,554,432 bytes) from the current
13610 location. A short call will be generated if the compiler knows
13611 the call cannot be that far away. This setting can be overridden
13612 by the `shortcall' function attribute, or by `#pragma longcall(0)'.
13614 Some linkers are capable of detecting out-of-range calls and
13615 generating glue code on the fly. On these systems, long calls are
13616 unnecessary and generate slower code. As of this writing, the AIX
13617 linker can do this, as can the GNU linker for PowerPC/64. It is
13618 planned to add this feature to the GNU linker for 32-bit PowerPC
13621 On Darwin/PPC systems, `#pragma longcall' will generate "jbsr
13622 callee, L42", plus a "branch island" (glue code). The two target
13623 addresses represent the callee and the "branch island". The
13624 Darwin/PPC linker will prefer the first address and generate a "bl
13625 callee" if the PPC "bl" instruction will reach the callee directly;
13626 otherwise, the linker will generate "bl L42" to call the "branch
13627 island". The "branch island" is appended to the body of the
13628 calling function; it computes the full 32-bit address of the callee
13631 On Mach-O (Darwin) systems, this option directs the compiler emit
13632 to the glue for every direct call, and the Darwin linker decides
13633 whether to use or discard it.
13635 In the future, we may cause GCC to ignore all longcall
13636 specifications when the linker is known to generate glue.
13639 Adds support for multithreading with the "pthreads" library. This
13640 option sets flags for both the preprocessor and linker.
13644 File: gcc.info, Node: S/390 and zSeries Options, Next: Score Options, Prev: RS/6000 and PowerPC Options, Up: Submodel Options
13646 3.17.29 S/390 and zSeries Options
13647 ---------------------------------
13649 These are the `-m' options defined for the S/390 and zSeries
13654 Use (do not use) the hardware floating-point instructions and
13655 registers for floating-point operations. When `-msoft-float' is
13656 specified, functions in `libgcc.a' will be used to perform
13657 floating-point operations. When `-mhard-float' is specified, the
13658 compiler generates IEEE floating-point instructions. This is the
13663 Use (do not use) the hardware decimal-floating-point instructions
13664 for decimal-floating-point operations. When `-mno-hard-dfp' is
13665 specified, functions in `libgcc.a' will be used to perform
13666 decimal-floating-point operations. When `-mhard-dfp' is
13667 specified, the compiler generates decimal-floating-point hardware
13668 instructions. This is the default for `-march=z9-ec' or higher.
13671 `-mlong-double-128'
13672 These switches control the size of `long double' type. A size of
13673 64bit makes the `long double' type equivalent to the `double'
13674 type. This is the default.
13678 Store (do not store) the address of the caller's frame as
13679 backchain pointer into the callee's stack frame. A backchain may
13680 be needed to allow debugging using tools that do not understand
13681 DWARF-2 call frame information. When `-mno-packed-stack' is in
13682 effect, the backchain pointer is stored at the bottom of the stack
13683 frame; when `-mpacked-stack' is in effect, the backchain is placed
13684 into the topmost word of the 96/160 byte register save area.
13686 In general, code compiled with `-mbackchain' is call-compatible
13687 with code compiled with `-mmo-backchain'; however, use of the
13688 backchain for debugging purposes usually requires that the whole
13689 binary is built with `-mbackchain'. Note that the combination of
13690 `-mbackchain', `-mpacked-stack' and `-mhard-float' is not
13691 supported. In order to build a linux kernel use `-msoft-float'.
13693 The default is to not maintain the backchain.
13696 `-mno-packed-stack'
13697 Use (do not use) the packed stack layout. When
13698 `-mno-packed-stack' is specified, the compiler uses the all fields
13699 of the 96/160 byte register save area only for their default
13700 purpose; unused fields still take up stack space. When
13701 `-mpacked-stack' is specified, register save slots are densely
13702 packed at the top of the register save area; unused space is
13703 reused for other purposes, allowing for more efficient use of the
13704 available stack space. However, when `-mbackchain' is also in
13705 effect, the topmost word of the save area is always used to store
13706 the backchain, and the return address register is always saved two
13707 words below the backchain.
13709 As long as the stack frame backchain is not used, code generated
13710 with `-mpacked-stack' is call-compatible with code generated with
13711 `-mno-packed-stack'. Note that some non-FSF releases of GCC 2.95
13712 for S/390 or zSeries generated code that uses the stack frame
13713 backchain at run time, not just for debugging purposes. Such code
13714 is not call-compatible with code compiled with `-mpacked-stack'.
13715 Also, note that the combination of `-mbackchain', `-mpacked-stack'
13716 and `-mhard-float' is not supported. In order to build a linux
13717 kernel use `-msoft-float'.
13719 The default is to not use the packed stack layout.
13723 Generate (or do not generate) code using the `bras' instruction to
13724 do subroutine calls. This only works reliably if the total
13725 executable size does not exceed 64k. The default is to use the
13726 `basr' instruction instead, which does not have this limitation.
13730 When `-m31' is specified, generate code compliant to the GNU/Linux
13731 for S/390 ABI. When `-m64' is specified, generate code compliant
13732 to the GNU/Linux for zSeries ABI. This allows GCC in particular
13733 to generate 64-bit instructions. For the `s390' targets, the
13734 default is `-m31', while the `s390x' targets default to `-m64'.
13738 When `-mzarch' is specified, generate code using the instructions
13739 available on z/Architecture. When `-mesa' is specified, generate
13740 code using the instructions available on ESA/390. Note that
13741 `-mesa' is not possible with `-m64'. When generating code
13742 compliant to the GNU/Linux for S/390 ABI, the default is `-mesa'.
13743 When generating code compliant to the GNU/Linux for zSeries ABI,
13744 the default is `-mzarch'.
13748 Generate (or do not generate) code using the `mvcle' instruction
13749 to perform block moves. When `-mno-mvcle' is specified, use a
13750 `mvc' loop instead. This is the default unless optimizing for
13755 Print (or do not print) additional debug information when
13756 compiling. The default is to not print debug information.
13759 Generate code that will run on CPU-TYPE, which is the name of a
13760 system representing a certain processor type. Possible values for
13761 CPU-TYPE are `g5', `g6', `z900', `z990', `z9-109', `z9-ec' and
13762 `z10'. When generating code using the instructions available on
13763 z/Architecture, the default is `-march=z900'. Otherwise, the
13764 default is `-march=g5'.
13767 Tune to CPU-TYPE everything applicable about the generated code,
13768 except for the ABI and the set of available instructions. The
13769 list of CPU-TYPE values is the same as for `-march'. The default
13770 is the value used for `-march'.
13774 Generate code that adds (does not add) in TPF OS specific branches
13775 to trace routines in the operating system. This option is off by
13776 default, even when compiling for the TPF OS.
13780 Generate code that uses (does not use) the floating point multiply
13781 and accumulate instructions. These instructions are generated by
13782 default if hardware floating point is used.
13784 `-mwarn-framesize=FRAMESIZE'
13785 Emit a warning if the current function exceeds the given frame
13786 size. Because this is a compile time check it doesn't need to be
13787 a real problem when the program runs. It is intended to identify
13788 functions which most probably cause a stack overflow. It is
13789 useful to be used in an environment with limited stack size e.g.
13792 `-mwarn-dynamicstack'
13793 Emit a warning if the function calls alloca or uses dynamically
13794 sized arrays. This is generally a bad idea with a limited stack
13797 `-mstack-guard=STACK-GUARD'
13798 `-mstack-size=STACK-SIZE'
13799 If these options are provided the s390 back end emits additional
13800 instructions in the function prologue which trigger a trap if the
13801 stack size is STACK-GUARD bytes above the STACK-SIZE (remember
13802 that the stack on s390 grows downward). If the STACK-GUARD option
13803 is omitted the smallest power of 2 larger than the frame size of
13804 the compiled function is chosen. These options are intended to be
13805 used to help debugging stack overflow problems. The additionally
13806 emitted code causes only little overhead and hence can also be
13807 used in production like systems without greater performance
13808 degradation. The given values have to be exact powers of 2 and
13809 STACK-SIZE has to be greater than STACK-GUARD without exceeding
13810 64k. In order to be efficient the extra code makes the assumption
13811 that the stack starts at an address aligned to the value given by
13812 STACK-SIZE. The STACK-GUARD option can only be used in
13813 conjunction with STACK-SIZE.
13816 File: gcc.info, Node: Score Options, Next: SH Options, Prev: S/390 and zSeries Options, Up: Submodel Options
13818 3.17.30 Score Options
13819 ---------------------
13821 These options are defined for Score implementations:
13824 Compile code for big endian mode. This is the default.
13827 Compile code for little endian mode.
13830 Disable generate bcnz instruction.
13833 Enable generate unaligned load and store instruction.
13836 Enable the use of multiply-accumulate instructions. Disabled by
13840 Specify the SCORE5 as the target architecture.
13843 Specify the SCORE5U of the target architecture.
13846 Specify the SCORE7 as the target architecture. This is the default.
13849 Specify the SCORE7D as the target architecture.
13852 File: gcc.info, Node: SH Options, Next: SPARC Options, Prev: Score Options, Up: Submodel Options
13857 These `-m' options are defined for the SH implementations:
13860 Generate code for the SH1.
13863 Generate code for the SH2.
13866 Generate code for the SH2e.
13869 Generate code for the SH3.
13872 Generate code for the SH3e.
13875 Generate code for the SH4 without a floating-point unit.
13878 Generate code for the SH4 with a floating-point unit that only
13879 supports single-precision arithmetic.
13882 Generate code for the SH4 assuming the floating-point unit is in
13883 single-precision mode by default.
13886 Generate code for the SH4.
13889 Generate code for the SH4al-dsp, or for a SH4a in such a way that
13890 the floating-point unit is not used.
13893 Generate code for the SH4a, in such a way that no double-precision
13894 floating point operations are used.
13897 Generate code for the SH4a assuming the floating-point unit is in
13898 single-precision mode by default.
13901 Generate code for the SH4a.
13904 Same as `-m4a-nofpu', except that it implicitly passes `-dsp' to
13905 the assembler. GCC doesn't generate any DSP instructions at the
13909 Compile code for the processor in big endian mode.
13912 Compile code for the processor in little endian mode.
13915 Align doubles at 64-bit boundaries. Note that this changes the
13916 calling conventions, and thus some functions from the standard C
13917 library will not work unless you recompile it first with
13921 Shorten some address references at link time, when possible; uses
13922 the linker option `-relax'.
13925 Use 32-bit offsets in `switch' tables. The default is to use
13929 Enable the use of bit manipulation instructions on SH2A.
13932 Enable the use of the instruction `fmovd'.
13935 Comply with the calling conventions defined by Renesas.
13938 Comply with the calling conventions defined by Renesas.
13941 Comply with the calling conventions defined for GCC before the
13942 Renesas conventions were available. This option is the default
13943 for all targets of the SH toolchain except for `sh-symbianelf'.
13946 Mark the `MAC' register as call-clobbered, even if `-mhitachi' is
13950 Increase IEEE-compliance of floating-point code. At the moment,
13951 this is equivalent to `-fno-finite-math-only'. When generating 16
13952 bit SH opcodes, getting IEEE-conforming results for comparisons of
13953 NANs / infinities incurs extra overhead in every floating point
13954 comparison, therefore the default is set to `-ffinite-math-only'.
13956 `-minline-ic_invalidate'
13957 Inline code to invalidate instruction cache entries after setting
13958 up nested function trampolines. This option has no effect if
13959 -musermode is in effect and the selected code generation option
13960 (e.g. -m4) does not allow the use of the icbi instruction. If the
13961 selected code generation option does not allow the use of the icbi
13962 instruction, and -musermode is not in effect, the inlined code will
13963 manipulate the instruction cache address array directly with an
13964 associative write. This not only requires privileged mode, but it
13965 will also fail if the cache line had been mapped via the TLB and
13966 has become unmapped.
13969 Dump instruction size and location in the assembly code.
13972 This option is deprecated. It pads structures to multiple of 4
13973 bytes, which is incompatible with the SH ABI.
13976 Optimize for space instead of speed. Implied by `-Os'.
13979 When generating position-independent code, emit function calls
13980 using the Global Offset Table instead of the Procedure Linkage
13984 Don't generate privileged mode only code; implies
13985 -mno-inline-ic_invalidate if the inlined code would not work in
13986 user mode. This is the default when the target is `sh-*-linux*'.
13989 Set the cost to assume for a multiply insn.
13992 Set the division strategy to use for SHmedia code. STRATEGY must
13993 be one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l,
13994 inv:call, inv:call2, inv:fp . "fp" performs the operation in
13995 floating point. This has a very high latency, but needs only a
13996 few instructions, so it might be a good choice if your code has
13997 enough easily exploitable ILP to allow the compiler to schedule
13998 the floating point instructions together with other instructions.
13999 Division by zero causes a floating point exception. "inv" uses
14000 integer operations to calculate the inverse of the divisor, and
14001 then multiplies the dividend with the inverse. This strategy
14002 allows cse and hoisting of the inverse calculation. Division by
14003 zero calculates an unspecified result, but does not trap.
14004 "inv:minlat" is a variant of "inv" where if no cse / hoisting
14005 opportunities have been found, or if the entire operation has been
14006 hoisted to the same place, the last stages of the inverse
14007 calculation are intertwined with the final multiply to reduce the
14008 overall latency, at the expense of using a few more instructions,
14009 and thus offering fewer scheduling opportunities with other code.
14010 "call" calls a library function that usually implements the
14011 inv:minlat strategy. This gives high code density for
14012 m5-*media-nofpu compilations. "call2" uses a different entry
14013 point of the same library function, where it assumes that a
14014 pointer to a lookup table has already been set up, which exposes
14015 the pointer load to cse / code hoisting optimizations.
14016 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm
14017 for initial code generation, but if the code stays unoptimized,
14018 revert to the "call", "call2", or "fp" strategies, respectively.
14019 Note that the potentially-trapping side effect of division by zero
14020 is carried by a separate instruction, so it is possible that all
14021 the integer instructions are hoisted out, but the marker for the
14022 side effect stays where it is. A recombination to fp operations
14023 or a call is not possible in that case. "inv20u" and "inv20l" are
14024 variants of the "inv:minlat" strategy. In the case that the
14025 inverse calculation was nor separated from the multiply, they speed
14026 up division where the dividend fits into 20 bits (plus sign where
14027 applicable), by inserting a test to skip a number of operations in
14028 this case; this test slows down the case of larger dividends.
14029 inv20u assumes the case of a such a small dividend to be unlikely,
14030 and inv20l assumes it to be likely.
14032 `-mdivsi3_libfunc=NAME'
14033 Set the name of the library function used for 32 bit signed
14034 division to NAME. This only affect the name used in the call and
14035 inv:call division strategies, and the compiler will still expect
14036 the same sets of input/output/clobbered registers as if this
14037 option was not present.
14039 `-mfixed-range=REGISTER-RANGE'
14040 Generate code treating the given register range as fixed registers.
14041 A fixed register is one that the register allocator can not use.
14042 This is useful when compiling kernel code. A register range is
14043 specified as two registers separated by a dash. Multiple register
14044 ranges can be specified separated by a comma.
14047 Throttle unrolling to avoid thrashing target registers. This
14048 option only has an effect if the gcc code base supports the
14049 TARGET_ADJUST_UNROLL_MAX target hook.
14051 `-mindexed-addressing'
14052 Enable the use of the indexed addressing mode for
14053 SHmedia32/SHcompact. This is only safe if the hardware and/or OS
14054 implement 32 bit wrap-around semantics for the indexed addressing
14055 mode. The architecture allows the implementation of processors
14056 with 64 bit MMU, which the OS could use to get 32 bit addressing,
14057 but since no current hardware implementation supports this or any
14058 other way to make the indexed addressing mode safe to use in the
14059 32 bit ABI, the default is -mno-indexed-addressing.
14061 `-mgettrcost=NUMBER'
14062 Set the cost assumed for the gettr instruction to NUMBER. The
14063 default is 2 if `-mpt-fixed' is in effect, 100 otherwise.
14066 Assume pt* instructions won't trap. This will generally generate
14067 better scheduled code, but is unsafe on current hardware. The
14068 current architecture definition says that ptabs and ptrel trap
14069 when the target anded with 3 is 3. This has the unintentional
14070 effect of making it unsafe to schedule ptabs / ptrel before a
14071 branch, or hoist it out of a loop. For example,
14072 __do_global_ctors, a part of libgcc that runs constructors at
14073 program startup, calls functions in a list which is delimited by
14074 -1. With the -mpt-fixed option, the ptabs will be done before
14075 testing against -1. That means that all the constructors will be
14076 run a bit quicker, but when the loop comes to the end of the list,
14077 the program crashes because ptabs loads -1 into a target register.
14078 Since this option is unsafe for any hardware implementing the
14079 current architecture specification, the default is -mno-pt-fixed.
14080 Unless the user specifies a specific cost with `-mgettrcost',
14081 -mno-pt-fixed also implies `-mgettrcost=100'; this deters register
14082 allocation using target registers for storing ordinary integers.
14084 `-minvalid-symbols'
14085 Assume symbols might be invalid. Ordinary function symbols
14086 generated by the compiler will always be valid to load with
14087 movi/shori/ptabs or movi/shori/ptrel, but with assembler and/or
14088 linker tricks it is possible to generate symbols that will cause
14089 ptabs / ptrel to trap. This option is only meaningful when
14090 `-mno-pt-fixed' is in effect. It will then prevent
14091 cross-basic-block cse, hoisting and most scheduling of symbol
14092 loads. The default is `-mno-invalid-symbols'.
14095 File: gcc.info, Node: SPARC Options, Next: SPU Options, Prev: SH Options, Up: Submodel Options
14097 3.17.32 SPARC Options
14098 ---------------------
14100 These `-m' options are supported on the SPARC:
14104 Specify `-mapp-regs' to generate output using the global registers
14105 2 through 4, which the SPARC SVR4 ABI reserves for applications.
14106 This is the default.
14108 To be fully SVR4 ABI compliant at the cost of some performance
14109 loss, specify `-mno-app-regs'. You should compile libraries and
14110 system software with this option.
14114 Generate output containing floating point instructions. This is
14119 Generate output containing library calls for floating point.
14120 *Warning:* the requisite libraries are not available for all SPARC
14121 targets. Normally the facilities of the machine's usual C
14122 compiler are used, but this cannot be done directly in
14123 cross-compilation. You must make your own arrangements to provide
14124 suitable library functions for cross-compilation. The embedded
14125 targets `sparc-*-aout' and `sparclite-*-*' do provide software
14126 floating point support.
14128 `-msoft-float' changes the calling convention in the output file;
14129 therefore, it is only useful if you compile _all_ of a program with
14130 this option. In particular, you need to compile `libgcc.a', the
14131 library that comes with GCC, with `-msoft-float' in order for this
14134 `-mhard-quad-float'
14135 Generate output containing quad-word (long double) floating point
14138 `-msoft-quad-float'
14139 Generate output containing library calls for quad-word (long
14140 double) floating point instructions. The functions called are
14141 those specified in the SPARC ABI. This is the default.
14143 As of this writing, there are no SPARC implementations that have
14144 hardware support for the quad-word floating point instructions.
14145 They all invoke a trap handler for one of these instructions, and
14146 then the trap handler emulates the effect of the instruction.
14147 Because of the trap handler overhead, this is much slower than
14148 calling the ABI library routines. Thus the `-msoft-quad-float'
14149 option is the default.
14151 `-mno-unaligned-doubles'
14152 `-munaligned-doubles'
14153 Assume that doubles have 8 byte alignment. This is the default.
14155 With `-munaligned-doubles', GCC assumes that doubles have 8 byte
14156 alignment only if they are contained in another type, or if they
14157 have an absolute address. Otherwise, it assumes they have 4 byte
14158 alignment. Specifying this option avoids some rare compatibility
14159 problems with code generated by other compilers. It is not the
14160 default because it results in a performance loss, especially for
14161 floating point code.
14163 `-mno-faster-structs'
14165 With `-mfaster-structs', the compiler assumes that structures
14166 should have 8 byte alignment. This enables the use of pairs of
14167 `ldd' and `std' instructions for copies in structure assignment,
14168 in place of twice as many `ld' and `st' pairs. However, the use
14169 of this changed alignment directly violates the SPARC ABI. Thus,
14170 it's intended only for use on targets where the developer
14171 acknowledges that their resulting code will not be directly in
14172 line with the rules of the ABI.
14175 `-mimpure-text', used in addition to `-shared', tells the compiler
14176 to not pass `-z text' to the linker when linking a shared object.
14177 Using this option, you can link position-dependent code into a
14180 `-mimpure-text' suppresses the "relocations remain against
14181 allocatable but non-writable sections" linker error message.
14182 However, the necessary relocations will trigger copy-on-write, and
14183 the shared object is not actually shared across processes.
14184 Instead of using `-mimpure-text', you should compile all source
14185 code with `-fpic' or `-fPIC'.
14187 This option is only available on SunOS and Solaris.
14190 Set the instruction set, register set, and instruction scheduling
14191 parameters for machine type CPU_TYPE. Supported values for
14192 CPU_TYPE are `v7', `cypress', `v8', `supersparc', `sparclite',
14193 `f930', `f934', `hypersparc', `sparclite86x', `sparclet',
14194 `tsc701', `v9', `ultrasparc', `ultrasparc3', `niagara' and
14197 Default instruction scheduling parameters are used for values that
14198 select an architecture and not an implementation. These are `v7',
14199 `v8', `sparclite', `sparclet', `v9'.
14201 Here is a list of each supported architecture and their supported
14205 v8: supersparc, hypersparc
14206 sparclite: f930, f934, sparclite86x
14208 v9: ultrasparc, ultrasparc3, niagara, niagara2
14210 By default (unless configured otherwise), GCC generates code for
14211 the V7 variant of the SPARC architecture. With `-mcpu=cypress',
14212 the compiler additionally optimizes it for the Cypress CY7C602
14213 chip, as used in the SPARCStation/SPARCServer 3xx series. This is
14214 also appropriate for the older SPARCStation 1, 2, IPX etc.
14216 With `-mcpu=v8', GCC generates code for the V8 variant of the SPARC
14217 architecture. The only difference from V7 code is that the
14218 compiler emits the integer multiply and integer divide
14219 instructions which exist in SPARC-V8 but not in SPARC-V7. With
14220 `-mcpu=supersparc', the compiler additionally optimizes it for the
14221 SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000
14224 With `-mcpu=sparclite', GCC generates code for the SPARClite
14225 variant of the SPARC architecture. This adds the integer
14226 multiply, integer divide step and scan (`ffs') instructions which
14227 exist in SPARClite but not in SPARC-V7. With `-mcpu=f930', the
14228 compiler additionally optimizes it for the Fujitsu MB86930 chip,
14229 which is the original SPARClite, with no FPU. With `-mcpu=f934',
14230 the compiler additionally optimizes it for the Fujitsu MB86934
14231 chip, which is the more recent SPARClite with FPU.
14233 With `-mcpu=sparclet', GCC generates code for the SPARClet variant
14234 of the SPARC architecture. This adds the integer multiply,
14235 multiply/accumulate, integer divide step and scan (`ffs')
14236 instructions which exist in SPARClet but not in SPARC-V7. With
14237 `-mcpu=tsc701', the compiler additionally optimizes it for the
14238 TEMIC SPARClet chip.
14240 With `-mcpu=v9', GCC generates code for the V9 variant of the SPARC
14241 architecture. This adds 64-bit integer and floating-point move
14242 instructions, 3 additional floating-point condition code registers
14243 and conditional move instructions. With `-mcpu=ultrasparc', the
14244 compiler additionally optimizes it for the Sun UltraSPARC I/II/IIi
14245 chips. With `-mcpu=ultrasparc3', the compiler additionally
14246 optimizes it for the Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+
14247 chips. With `-mcpu=niagara', the compiler additionally optimizes
14248 it for Sun UltraSPARC T1 chips. With `-mcpu=niagara2', the
14249 compiler additionally optimizes it for Sun UltraSPARC T2 chips.
14252 Set the instruction scheduling parameters for machine type
14253 CPU_TYPE, but do not set the instruction set or register set that
14254 the option `-mcpu=CPU_TYPE' would.
14256 The same values for `-mcpu=CPU_TYPE' can be used for
14257 `-mtune=CPU_TYPE', but the only useful values are those that
14258 select a particular cpu implementation. Those are `cypress',
14259 `supersparc', `hypersparc', `f930', `f934', `sparclite86x',
14260 `tsc701', `ultrasparc', `ultrasparc3', `niagara', and `niagara2'.
14264 With `-mv8plus', GCC generates code for the SPARC-V8+ ABI. The
14265 difference from the V8 ABI is that the global and out registers are
14266 considered 64-bit wide. This is enabled by default on Solaris in
14267 32-bit mode for all SPARC-V9 processors.
14271 With `-mvis', GCC generates code that takes advantage of the
14272 UltraSPARC Visual Instruction Set extensions. The default is
14275 These `-m' options are supported in addition to the above on SPARC-V9
14276 processors in 64-bit environments:
14279 Generate code for a processor running in little-endian mode. It
14280 is only available for a few configurations and most notably not on
14285 Generate code for a 32-bit or 64-bit environment. The 32-bit
14286 environment sets int, long and pointer to 32 bits. The 64-bit
14287 environment sets int to 32 bits and long and pointer to 64 bits.
14290 Generate code for the Medium/Low code model: 64-bit addresses,
14291 programs must be linked in the low 32 bits of memory. Programs
14292 can be statically or dynamically linked.
14295 Generate code for the Medium/Middle code model: 64-bit addresses,
14296 programs must be linked in the low 44 bits of memory, the text and
14297 data segments must be less than 2GB in size and the data segment
14298 must be located within 2GB of the text segment.
14301 Generate code for the Medium/Anywhere code model: 64-bit
14302 addresses, programs may be linked anywhere in memory, the text and
14303 data segments must be less than 2GB in size and the data segment
14304 must be located within 2GB of the text segment.
14306 `-mcmodel=embmedany'
14307 Generate code for the Medium/Anywhere code model for embedded
14308 systems: 64-bit addresses, the text and data segments must be less
14309 than 2GB in size, both starting anywhere in memory (determined at
14310 link time). The global register %g4 points to the base of the
14311 data segment. Programs are statically linked and PIC is not
14316 With `-mstack-bias', GCC assumes that the stack pointer, and frame
14317 pointer if present, are offset by -2047 which must be added back
14318 when making stack frame references. This is the default in 64-bit
14319 mode. Otherwise, assume no such offset is present.
14321 These switches are supported in addition to the above on Solaris:
14324 Add support for multithreading using the Solaris threads library.
14325 This option sets flags for both the preprocessor and linker. This
14326 option does not affect the thread safety of object code produced
14327 by the compiler or that of libraries supplied with it.
14330 Add support for multithreading using the POSIX threads library.
14331 This option sets flags for both the preprocessor and linker. This
14332 option does not affect the thread safety of object code produced
14333 by the compiler or that of libraries supplied with it.
14336 This is a synonym for `-pthreads'.
14339 File: gcc.info, Node: SPU Options, Next: System V Options, Prev: SPARC Options, Up: Submodel Options
14341 3.17.33 SPU Options
14342 -------------------
14344 These `-m' options are supported on the SPU:
14348 The loader for SPU does not handle dynamic relocations. By
14349 default, GCC will give an error when it generates code that
14350 requires a dynamic relocation. `-mno-error-reloc' disables the
14351 error, `-mwarn-reloc' will generate a warning instead.
14355 Instructions which initiate or test completion of DMA must not be
14356 reordered with respect to loads and stores of the memory which is
14357 being accessed. Users typically address this problem using the
14358 volatile keyword, but that can lead to inefficient code in places
14359 where the memory is known to not change. Rather than mark the
14360 memory as volatile we treat the DMA instructions as potentially
14361 effecting all memory. With `-munsafe-dma' users must use the
14362 volatile keyword to protect memory accesses.
14365 By default, GCC will generate a branch hint instruction to avoid
14366 pipeline stalls for always taken or probably taken branches. A
14367 hint will not be generated closer than 8 instructions away from
14368 its branch. There is little reason to disable them, except for
14369 debugging purposes, or to make an object a little bit smaller.
14373 By default, GCC generates code assuming that addresses are never
14374 larger than 18 bits. With `-mlarge-mem' code is generated that
14375 assumes a full 32 bit address.
14378 By default, GCC links against startup code that assumes the
14379 SPU-style main function interface (which has an unconventional
14380 parameter list). With `-mstdmain', GCC will link your program
14381 against startup code that assumes a C99-style interface to `main',
14382 including a local copy of `argv' strings.
14384 `-mfixed-range=REGISTER-RANGE'
14385 Generate code treating the given register range as fixed registers.
14386 A fixed register is one that the register allocator can not use.
14387 This is useful when compiling kernel code. A register range is
14388 specified as two registers separated by a dash. Multiple register
14389 ranges can be specified separated by a comma.
14393 By default, GCC will insert nops to increase dual issue when it
14394 expects it to increase performance. N can be a value from 0 to
14395 10. A smaller N will insert fewer nops. 10 is the default, 0 is
14396 the same as `-mno-dual-nops'. Disabled with `-Os'.
14398 `-mhint-max-nops=N'
14399 Maximum number of nops to insert for a branch hint. A branch hint
14400 must be at least 8 instructions away from the branch it is
14401 effecting. GCC will insert up to N nops to enforce this,
14402 otherwise it will not generate the branch hint.
14404 `-mhint-max-distance=N'
14405 The encoding of the branch hint instruction limits the hint to be
14406 within 256 instructions of the branch it is effecting. By
14407 default, GCC makes sure it is within 125.
14410 Work around a hardware bug which causes the SPU to stall
14411 indefinitely. By default, GCC will insert the `hbrp' instruction
14412 to make sure this stall won't happen.
14416 File: gcc.info, Node: System V Options, Next: V850 Options, Prev: SPU Options, Up: Submodel Options
14418 3.17.34 Options for System V
14419 ----------------------------
14421 These additional options are available on System V Release 4 for
14422 compatibility with other compilers on those systems:
14425 Create a shared object. It is recommended that `-symbolic' or
14426 `-shared' be used instead.
14429 Identify the versions of each tool used by the compiler, in a
14430 `.ident' assembler directive in the output.
14433 Refrain from adding `.ident' directives to the output file (this is
14437 Search the directories DIRS, and no others, for libraries
14438 specified with `-l'.
14441 Look in the directory DIR to find the M4 preprocessor. The
14442 assembler uses this option.
14445 File: gcc.info, Node: V850 Options, Next: VAX Options, Prev: System V Options, Up: Submodel Options
14447 3.17.35 V850 Options
14448 --------------------
14450 These `-m' options are defined for V850 implementations:
14454 Treat all calls as being far away (near). If calls are assumed to
14455 be far away, the compiler will always load the functions address
14456 up into a register, and call indirect through the pointer.
14460 Do not optimize (do optimize) basic blocks that use the same index
14461 pointer 4 or more times to copy pointer into the `ep' register, and
14462 use the shorter `sld' and `sst' instructions. The `-mep' option
14463 is on by default if you optimize.
14465 `-mno-prolog-function'
14466 `-mprolog-function'
14467 Do not use (do use) external functions to save and restore
14468 registers at the prologue and epilogue of a function. The
14469 external functions are slower, but use less code space if more
14470 than one function saves the same number of registers. The
14471 `-mprolog-function' option is on by default if you optimize.
14474 Try to make the code as small as possible. At present, this just
14475 turns on the `-mep' and `-mprolog-function' options.
14478 Put static or global variables whose size is N bytes or less into
14479 the tiny data area that register `ep' points to. The tiny data
14480 area can hold up to 256 bytes in total (128 bytes for byte
14484 Put static or global variables whose size is N bytes or less into
14485 the small data area that register `gp' points to. The small data
14486 area can hold up to 64 kilobytes.
14489 Put static or global variables whose size is N bytes or less into
14490 the first 32 kilobytes of memory.
14493 Specify that the target processor is the V850.
14496 Generate code suitable for big switch tables. Use this option
14497 only if the assembler/linker complain about out of range branches
14498 within a switch table.
14501 This option will cause r2 and r5 to be used in the code generated
14502 by the compiler. This setting is the default.
14505 This option will cause r2 and r5 to be treated as fixed registers.
14508 Specify that the target processor is the V850E1. The preprocessor
14509 constants `__v850e1__' and `__v850e__' will be defined if this
14513 Specify that the target processor is the V850E. The preprocessor
14514 constant `__v850e__' will be defined if this option is used.
14516 If neither `-mv850' nor `-mv850e' nor `-mv850e1' are defined then
14517 a default target processor will be chosen and the relevant
14518 `__v850*__' preprocessor constant will be defined.
14520 The preprocessor constants `__v850' and `__v851__' are always
14521 defined, regardless of which processor variant is the target.
14524 This option will suppress generation of the CALLT instruction for
14525 the v850e and v850e1 flavors of the v850 architecture. The
14526 default is `-mno-disable-callt' which allows the CALLT instruction
14531 File: gcc.info, Node: VAX Options, Next: VxWorks Options, Prev: V850 Options, Up: Submodel Options
14533 3.17.36 VAX Options
14534 -------------------
14536 These `-m' options are defined for the VAX:
14539 Do not output certain jump instructions (`aobleq' and so on) that
14540 the Unix assembler for the VAX cannot handle across long ranges.
14543 Do output those jump instructions, on the assumption that you will
14544 assemble with the GNU assembler.
14547 Output code for g-format floating point numbers instead of
14551 File: gcc.info, Node: VxWorks Options, Next: x86-64 Options, Prev: VAX Options, Up: Submodel Options
14553 3.17.37 VxWorks Options
14554 -----------------------
14556 The options in this section are defined for all VxWorks targets.
14557 Options specific to the target hardware are listed with the other
14558 options for that target.
14561 GCC can generate code for both VxWorks kernels and real time
14562 processes (RTPs). This option switches from the former to the
14563 latter. It also defines the preprocessor macro `__RTP__'.
14566 Link an RTP executable against shared libraries rather than static
14567 libraries. The options `-static' and `-shared' can also be used
14568 for RTPs (*note Link Options::); `-static' is the default.
14572 These options are passed down to the linker. They are defined for
14573 compatibility with Diab.
14576 Enable lazy binding of function calls. This option is equivalent
14577 to `-Wl,-z,now' and is defined for compatibility with Diab.
14580 Disable lazy binding of function calls. This option is the
14581 default and is defined for compatibility with Diab.
14584 File: gcc.info, Node: x86-64 Options, Next: Xstormy16 Options, Prev: VxWorks Options, Up: Submodel Options
14586 3.17.38 x86-64 Options
14587 ----------------------
14589 These are listed under *Note i386 and x86-64 Options::.
14592 File: gcc.info, Node: i386 and x86-64 Windows Options, Next: IA-64 Options, Prev: i386 and x86-64 Options, Up: Submodel Options
14594 3.17.39 i386 and x86-64 Windows Options
14595 ---------------------------------------
14597 These additional options are available for Windows targets:
14600 This option is available for Cygwin and MinGW targets. It
14601 specifies that a console application is to be generated, by
14602 instructing the linker to set the PE header subsystem type
14603 required for console applications. This is the default behaviour
14604 for Cygwin and MinGW targets.
14607 This option is available for Cygwin targets. It specifies that
14608 the Cygwin internal interface is to be used for predefined
14609 preprocessor macros, C runtime libraries and related linker paths
14610 and options. For Cygwin targets this is the default behaviour.
14611 This option is deprecated and will be removed in a future release.
14614 This option is available for Cygwin targets. It specifies that
14615 the MinGW internal interface is to be used instead of Cygwin's, by
14616 setting MinGW-related predefined macros and linker paths and
14617 default library options. This option is deprecated and will be
14618 removed in a future release.
14621 This option is available for Cygwin and MinGW targets. It
14622 specifies that a DLL - a dynamic link library - is to be
14623 generated, enabling the selection of the required runtime startup
14624 object and entry point.
14626 `-mnop-fun-dllimport'
14627 This option is available for Cygwin and MinGW targets. It
14628 specifies that the dllimport attribute should be ignored.
14631 This option is available for MinGW targets. It specifies that
14632 MinGW-specific thread support is to be used.
14635 This option is available for Cygwin and MinGW targets. It
14636 specifies that the typical Windows pre-defined macros are to be
14637 set in the pre-processor, but does not influence the choice of
14638 runtime library/startup code.
14641 This option is available for Cygwin and MinGW targets. It
14642 specifies that a GUI application is to be generated by instructing
14643 the linker to set the PE header subsystem type appropriately.
14645 See also under *note i386 and x86-64 Options:: for standard options.
14648 File: gcc.info, Node: Xstormy16 Options, Next: Xtensa Options, Prev: x86-64 Options, Up: Submodel Options
14650 3.17.40 Xstormy16 Options
14651 -------------------------
14653 These options are defined for Xstormy16:
14656 Choose startup files and linker script suitable for the simulator.
14659 File: gcc.info, Node: Xtensa Options, Next: zSeries Options, Prev: Xstormy16 Options, Up: Submodel Options
14661 3.17.41 Xtensa Options
14662 ----------------------
14664 These options are supported for Xtensa targets:
14668 Enable or disable use of `CONST16' instructions for loading
14669 constant values. The `CONST16' instruction is currently not a
14670 standard option from Tensilica. When enabled, `CONST16'
14671 instructions are always used in place of the standard `L32R'
14672 instructions. The use of `CONST16' is enabled by default only if
14673 the `L32R' instruction is not available.
14677 Enable or disable use of fused multiply/add and multiply/subtract
14678 instructions in the floating-point option. This has no effect if
14679 the floating-point option is not also enabled. Disabling fused
14680 multiply/add and multiply/subtract instructions forces the
14681 compiler to use separate instructions for the multiply and
14682 add/subtract operations. This may be desirable in some cases
14683 where strict IEEE 754-compliant results are required: the fused
14684 multiply add/subtract instructions do not round the intermediate
14685 result, thereby producing results with _more_ bits of precision
14686 than specified by the IEEE standard. Disabling fused multiply
14687 add/subtract instructions also ensures that the program output is
14688 not sensitive to the compiler's ability to combine multiply and
14689 add/subtract operations.
14691 `-mserialize-volatile'
14692 `-mno-serialize-volatile'
14693 When this option is enabled, GCC inserts `MEMW' instructions before
14694 `volatile' memory references to guarantee sequential consistency.
14695 The default is `-mserialize-volatile'. Use
14696 `-mno-serialize-volatile' to omit the `MEMW' instructions.
14698 `-mtext-section-literals'
14699 `-mno-text-section-literals'
14700 Control the treatment of literal pools. The default is
14701 `-mno-text-section-literals', which places literals in a separate
14702 section in the output file. This allows the literal pool to be
14703 placed in a data RAM/ROM, and it also allows the linker to combine
14704 literal pools from separate object files to remove redundant
14705 literals and improve code size. With `-mtext-section-literals',
14706 the literals are interspersed in the text section in order to keep
14707 them as close as possible to their references. This may be
14708 necessary for large assembly files.
14711 `-mno-target-align'
14712 When this option is enabled, GCC instructs the assembler to
14713 automatically align instructions to reduce branch penalties at the
14714 expense of some code density. The assembler attempts to widen
14715 density instructions to align branch targets and the instructions
14716 following call instructions. If there are not enough preceding
14717 safe density instructions to align a target, no widening will be
14718 performed. The default is `-mtarget-align'. These options do not
14719 affect the treatment of auto-aligned instructions like `LOOP',
14720 which the assembler will always align, either by widening density
14721 instructions or by inserting no-op instructions.
14725 When this option is enabled, GCC instructs the assembler to
14726 translate direct calls to indirect calls unless it can determine
14727 that the target of a direct call is in the range allowed by the
14728 call instruction. This translation typically occurs for calls to
14729 functions in other source files. Specifically, the assembler
14730 translates a direct `CALL' instruction into an `L32R' followed by
14731 a `CALLX' instruction. The default is `-mno-longcalls'. This
14732 option should be used in programs where the call target can
14733 potentially be out of range. This option is implemented in the
14734 assembler, not the compiler, so the assembly code generated by GCC
14735 will still show direct call instructions--look at the disassembled
14736 object code to see the actual instructions. Note that the
14737 assembler will use an indirect call for every cross-file call, not
14738 just those that really will be out of range.
14741 File: gcc.info, Node: zSeries Options, Prev: Xtensa Options, Up: Submodel Options
14743 3.17.42 zSeries Options
14744 -----------------------
14746 These are listed under *Note S/390 and zSeries Options::.
14749 File: gcc.info, Node: Code Gen Options, Next: Environment Variables, Prev: Submodel Options, Up: Invoking GCC
14751 3.18 Options for Code Generation Conventions
14752 ============================================
14754 These machine-independent options control the interface conventions
14755 used in code generation.
14757 Most of them have both positive and negative forms; the negative form
14758 of `-ffoo' would be `-fno-foo'. In the table below, only one of the
14759 forms is listed--the one which is not the default. You can figure out
14760 the other form by either removing `no-' or adding it.
14763 For front-ends that support it, generate additional code to check
14764 that indices used to access arrays are within the declared range.
14765 This is currently only supported by the Java and Fortran
14766 front-ends, where this option defaults to true and false
14770 This option generates traps for signed overflow on addition,
14771 subtraction, multiplication operations.
14774 This option instructs the compiler to assume that signed arithmetic
14775 overflow of addition, subtraction and multiplication wraps around
14776 using twos-complement representation. This flag enables some
14777 optimizations and disables others. This option is enabled by
14778 default for the Java front-end, as required by the Java language
14782 Enable exception handling. Generates extra code needed to
14783 propagate exceptions. For some targets, this implies GCC will
14784 generate frame unwind information for all functions, which can
14785 produce significant data size overhead, although it does not
14786 affect execution. If you do not specify this option, GCC will
14787 enable it by default for languages like C++ which normally require
14788 exception handling, and disable it for languages like C that do
14789 not normally require it. However, you may need to enable this
14790 option when compiling C code that needs to interoperate properly
14791 with exception handlers written in C++. You may also wish to
14792 disable this option if you are compiling older C++ programs that
14793 don't use exception handling.
14795 `-fnon-call-exceptions'
14796 Generate code that allows trapping instructions to throw
14797 exceptions. Note that this requires platform-specific runtime
14798 support that does not exist everywhere. Moreover, it only allows
14799 _trapping_ instructions to throw exceptions, i.e. memory
14800 references or floating point instructions. It does not allow
14801 exceptions to be thrown from arbitrary signal handlers such as
14805 Similar to `-fexceptions', except that it will just generate any
14806 needed static data, but will not affect the generated code in any
14807 other way. You will normally not enable this option; instead, a
14808 language processor that needs this handling would enable it on
14811 `-fasynchronous-unwind-tables'
14812 Generate unwind table in dwarf2 format, if supported by target
14813 machine. The table is exact at each instruction boundary, so it
14814 can be used for stack unwinding from asynchronous events (such as
14815 debugger or garbage collector).
14817 `-fpcc-struct-return'
14818 Return "short" `struct' and `union' values in memory like longer
14819 ones, rather than in registers. This convention is less
14820 efficient, but it has the advantage of allowing intercallability
14821 between GCC-compiled files and files compiled with other
14822 compilers, particularly the Portable C Compiler (pcc).
14824 The precise convention for returning structures in memory depends
14825 on the target configuration macros.
14827 Short structures and unions are those whose size and alignment
14828 match that of some integer type.
14830 *Warning:* code compiled with the `-fpcc-struct-return' switch is
14831 not binary compatible with code compiled with the
14832 `-freg-struct-return' switch. Use it to conform to a non-default
14833 application binary interface.
14835 `-freg-struct-return'
14836 Return `struct' and `union' values in registers when possible.
14837 This is more efficient for small structures than
14838 `-fpcc-struct-return'.
14840 If you specify neither `-fpcc-struct-return' nor
14841 `-freg-struct-return', GCC defaults to whichever convention is
14842 standard for the target. If there is no standard convention, GCC
14843 defaults to `-fpcc-struct-return', except on targets where GCC is
14844 the principal compiler. In those cases, we can choose the
14845 standard, and we chose the more efficient register return
14848 *Warning:* code compiled with the `-freg-struct-return' switch is
14849 not binary compatible with code compiled with the
14850 `-fpcc-struct-return' switch. Use it to conform to a non-default
14851 application binary interface.
14854 Allocate to an `enum' type only as many bytes as it needs for the
14855 declared range of possible values. Specifically, the `enum' type
14856 will be equivalent to the smallest integer type which has enough
14859 *Warning:* the `-fshort-enums' switch causes GCC to generate code
14860 that is not binary compatible with code generated without that
14861 switch. Use it to conform to a non-default application binary
14865 Use the same size for `double' as for `float'.
14867 *Warning:* the `-fshort-double' switch causes GCC to generate code
14868 that is not binary compatible with code generated without that
14869 switch. Use it to conform to a non-default application binary
14873 Override the underlying type for `wchar_t' to be `short unsigned
14874 int' instead of the default for the target. This option is useful
14875 for building programs to run under WINE.
14877 *Warning:* the `-fshort-wchar' switch causes GCC to generate code
14878 that is not binary compatible with code generated without that
14879 switch. Use it to conform to a non-default application binary
14883 In C code, controls the placement of uninitialized global
14884 variables. Unix C compilers have traditionally permitted multiple
14885 definitions of such variables in different compilation units by
14886 placing the variables in a common block. This is the behavior
14887 specified by `-fcommon', and is the default for GCC on most
14888 targets. On the other hand, this behavior is not required by ISO
14889 C, and on some targets may carry a speed or code size penalty on
14890 variable references. The `-fno-common' option specifies that the
14891 compiler should place uninitialized global variables in the data
14892 section of the object file, rather than generating them as common
14893 blocks. This has the effect that if the same variable is declared
14894 (without `extern') in two different compilations, you will get a
14895 multiple-definition error when you link them. In this case, you
14896 must compile with `-fcommon' instead. Compiling with
14897 `-fno-common' is useful on targets for which it provides better
14898 performance, or if you wish to verify that the program will work
14899 on other systems which always treat uninitialized variable
14900 declarations this way.
14903 Ignore the `#ident' directive.
14905 `-finhibit-size-directive'
14906 Don't output a `.size' assembler directive, or anything else that
14907 would cause trouble if the function is split in the middle, and the
14908 two halves are placed at locations far apart in memory. This
14909 option is used when compiling `crtstuff.c'; you should not need to
14910 use it for anything else.
14913 Put extra commentary information in the generated assembly code to
14914 make it more readable. This option is generally only of use to
14915 those who actually need to read the generated assembly code
14916 (perhaps while debugging the compiler itself).
14918 `-fno-verbose-asm', the default, causes the extra information to
14919 be omitted and is useful when comparing two assembler files.
14921 `-frecord-gcc-switches'
14922 This switch causes the command line that was used to invoke the
14923 compiler to be recorded into the object file that is being created.
14924 This switch is only implemented on some targets and the exact
14925 format of the recording is target and binary file format
14926 dependent, but it usually takes the form of a section containing
14927 ASCII text. This switch is related to the `-fverbose-asm' switch,
14928 but that switch only records information in the assembler output
14929 file as comments, so it never reaches the object file.
14932 Generate position-independent code (PIC) suitable for use in a
14933 shared library, if supported for the target machine. Such code
14934 accesses all constant addresses through a global offset table
14935 (GOT). The dynamic loader resolves the GOT entries when the
14936 program starts (the dynamic loader is not part of GCC; it is part
14937 of the operating system). If the GOT size for the linked
14938 executable exceeds a machine-specific maximum size, you get an
14939 error message from the linker indicating that `-fpic' does not
14940 work; in that case, recompile with `-fPIC' instead. (These
14941 maximums are 8k on the SPARC and 32k on the m68k and RS/6000. The
14942 386 has no such limit.)
14944 Position-independent code requires special support, and therefore
14945 works only on certain machines. For the 386, GCC supports PIC for
14946 System V but not for the Sun 386i. Code generated for the IBM
14947 RS/6000 is always position-independent.
14949 When this flag is set, the macros `__pic__' and `__PIC__' are
14953 If supported for the target machine, emit position-independent
14954 code, suitable for dynamic linking and avoiding any limit on the
14955 size of the global offset table. This option makes a difference
14956 on the m68k, PowerPC and SPARC.
14958 Position-independent code requires special support, and therefore
14959 works only on certain machines.
14961 When this flag is set, the macros `__pic__' and `__PIC__' are
14966 These options are similar to `-fpic' and `-fPIC', but generated
14967 position independent code can be only linked into executables.
14968 Usually these options are used when `-pie' GCC option will be used
14971 `-fpie' and `-fPIE' both define the macros `__pie__' and
14972 `__PIE__'. The macros have the value 1 for `-fpie' and 2 for
14976 Do not use jump tables for switch statements even where it would be
14977 more efficient than other code generation strategies. This option
14978 is of use in conjunction with `-fpic' or `-fPIC' for building code
14979 which forms part of a dynamic linker and cannot reference the
14980 address of a jump table. On some targets, jump tables do not
14981 require a GOT and this option is not needed.
14984 Treat the register named REG as a fixed register; generated code
14985 should never refer to it (except perhaps as a stack pointer, frame
14986 pointer or in some other fixed role).
14988 REG must be the name of a register. The register names accepted
14989 are machine-specific and are defined in the `REGISTER_NAMES' macro
14990 in the machine description macro file.
14992 This flag does not have a negative form, because it specifies a
14996 Treat the register named REG as an allocable register that is
14997 clobbered by function calls. It may be allocated for temporaries
14998 or variables that do not live across a call. Functions compiled
14999 this way will not save and restore the register REG.
15001 It is an error to used this flag with the frame pointer or stack
15002 pointer. Use of this flag for other registers that have fixed
15003 pervasive roles in the machine's execution model will produce
15004 disastrous results.
15006 This flag does not have a negative form, because it specifies a
15010 Treat the register named REG as an allocable register saved by
15011 functions. It may be allocated even for temporaries or variables
15012 that live across a call. Functions compiled this way will save
15013 and restore the register REG if they use it.
15015 It is an error to used this flag with the frame pointer or stack
15016 pointer. Use of this flag for other registers that have fixed
15017 pervasive roles in the machine's execution model will produce
15018 disastrous results.
15020 A different sort of disaster will result from the use of this flag
15021 for a register in which function values may be returned.
15023 This flag does not have a negative form, because it specifies a
15026 `-fpack-struct[=N]'
15027 Without a value specified, pack all structure members together
15028 without holes. When a value is specified (which must be a small
15029 power of two), pack structure members according to this value,
15030 representing the maximum alignment (that is, objects with default
15031 alignment requirements larger than this will be output potentially
15032 unaligned at the next fitting location.
15034 *Warning:* the `-fpack-struct' switch causes GCC to generate code
15035 that is not binary compatible with code generated without that
15036 switch. Additionally, it makes the code suboptimal. Use it to
15037 conform to a non-default application binary interface.
15039 `-finstrument-functions'
15040 Generate instrumentation calls for entry and exit to functions.
15041 Just after function entry and just before function exit, the
15042 following profiling functions will be called with the address of
15043 the current function and its call site. (On some platforms,
15044 `__builtin_return_address' does not work beyond the current
15045 function, so the call site information may not be available to the
15046 profiling functions otherwise.)
15048 void __cyg_profile_func_enter (void *this_fn,
15050 void __cyg_profile_func_exit (void *this_fn,
15053 The first argument is the address of the start of the current
15054 function, which may be looked up exactly in the symbol table.
15056 This instrumentation is also done for functions expanded inline in
15057 other functions. The profiling calls will indicate where,
15058 conceptually, the inline function is entered and exited. This
15059 means that addressable versions of such functions must be
15060 available. If all your uses of a function are expanded inline,
15061 this may mean an additional expansion of code size. If you use
15062 `extern inline' in your C code, an addressable version of such
15063 functions must be provided. (This is normally the case anyways,
15064 but if you get lucky and the optimizer always expands the
15065 functions inline, you might have gotten away without providing
15068 A function may be given the attribute `no_instrument_function', in
15069 which case this instrumentation will not be done. This can be
15070 used, for example, for the profiling functions listed above,
15071 high-priority interrupt routines, and any functions from which the
15072 profiling functions cannot safely be called (perhaps signal
15073 handlers, if the profiling routines generate output or allocate
15076 `-finstrument-functions-exclude-file-list=FILE,FILE,...'
15077 Set the list of functions that are excluded from instrumentation
15078 (see the description of `-finstrument-functions'). If the file
15079 that contains a function definition matches with one of FILE, then
15080 that function is not instrumented. The match is done on
15081 substrings: if the FILE parameter is a substring of the file name,
15082 it is considered to be a match.
15085 `-finstrument-functions-exclude-file-list=/bits/stl,include/sys'
15086 will exclude any inline function defined in files whose pathnames
15087 contain `/bits/stl' or `include/sys'.
15089 If, for some reason, you want to include letter `','' in one of
15090 SYM, write `'\,''. For example,
15091 `-finstrument-functions-exclude-file-list='\,\,tmp'' (note the
15092 single quote surrounding the option).
15094 `-finstrument-functions-exclude-function-list=SYM,SYM,...'
15095 This is similar to `-finstrument-functions-exclude-file-list', but
15096 this option sets the list of function names to be excluded from
15097 instrumentation. The function name to be matched is its
15098 user-visible name, such as `vector<int> blah(const vector<int>
15099 &)', not the internal mangled name (e.g.,
15100 `_Z4blahRSt6vectorIiSaIiEE'). The match is done on substrings: if
15101 the SYM parameter is a substring of the function name, it is
15102 considered to be a match.
15105 Generate code to verify that you do not go beyond the boundary of
15106 the stack. You should specify this flag if you are running in an
15107 environment with multiple threads, but only rarely need to specify
15108 it in a single-threaded environment since stack overflow is
15109 automatically detected on nearly all systems if there is only one
15112 Note that this switch does not actually cause checking to be done;
15113 the operating system or the language runtime must do that. The
15114 switch causes generation of code to ensure that they see the stack
15117 You can additionally specify a string parameter: `no' means no
15118 checking, `generic' means force the use of old-style checking,
15119 `specific' means use the best checking method and is equivalent to
15120 bare `-fstack-check'.
15122 Old-style checking is a generic mechanism that requires no specific
15123 target support in the compiler but comes with the following
15126 1. Modified allocation strategy for large objects: they will
15127 always be allocated dynamically if their size exceeds a fixed
15130 2. Fixed limit on the size of the static frame of functions:
15131 when it is topped by a particular function, stack checking is
15132 not reliable and a warning is issued by the compiler.
15134 3. Inefficiency: because of both the modified allocation
15135 strategy and the generic implementation, the performances of
15136 the code are hampered.
15138 Note that old-style stack checking is also the fallback method for
15139 `specific' if no target support has been added in the compiler.
15141 `-fstack-limit-register=REG'
15142 `-fstack-limit-symbol=SYM'
15144 Generate code to ensure that the stack does not grow beyond a
15145 certain value, either the value of a register or the address of a
15146 symbol. If the stack would grow beyond the value, a signal is
15147 raised. For most targets, the signal is raised before the stack
15148 overruns the boundary, so it is possible to catch the signal
15149 without taking special precautions.
15151 For instance, if the stack starts at absolute address `0x80000000'
15152 and grows downwards, you can use the flags
15153 `-fstack-limit-symbol=__stack_limit' and
15154 `-Wl,--defsym,__stack_limit=0x7ffe0000' to enforce a stack limit
15155 of 128KB. Note that this may only work with the GNU linker.
15158 `-fargument-noalias'
15159 `-fargument-noalias-global'
15160 `-fargument-noalias-anything'
15161 Specify the possible relationships among parameters and between
15162 parameters and global data.
15164 `-fargument-alias' specifies that arguments (parameters) may alias
15165 each other and may alias global storage.
15166 `-fargument-noalias' specifies that arguments do not alias each
15167 other, but may alias global storage.
15168 `-fargument-noalias-global' specifies that arguments do not alias
15169 each other and do not alias global storage.
15170 `-fargument-noalias-anything' specifies that arguments do not
15171 alias any other storage.
15173 Each language will automatically use whatever option is required by
15174 the language standard. You should not need to use these options
15177 `-fleading-underscore'
15178 This option and its counterpart, `-fno-leading-underscore',
15179 forcibly change the way C symbols are represented in the object
15180 file. One use is to help link with legacy assembly code.
15182 *Warning:* the `-fleading-underscore' switch causes GCC to
15183 generate code that is not binary compatible with code generated
15184 without that switch. Use it to conform to a non-default
15185 application binary interface. Not all targets provide complete
15186 support for this switch.
15188 `-ftls-model=MODEL'
15189 Alter the thread-local storage model to be used (*note
15190 Thread-Local::). The MODEL argument should be one of
15191 `global-dynamic', `local-dynamic', `initial-exec' or `local-exec'.
15193 The default without `-fpic' is `initial-exec'; with `-fpic' the
15194 default is `global-dynamic'.
15196 `-fvisibility=DEFAULT|INTERNAL|HIDDEN|PROTECTED'
15197 Set the default ELF image symbol visibility to the specified
15198 option--all symbols will be marked with this unless overridden
15199 within the code. Using this feature can very substantially
15200 improve linking and load times of shared object libraries, produce
15201 more optimized code, provide near-perfect API export and prevent
15202 symbol clashes. It is *strongly* recommended that you use this in
15203 any shared objects you distribute.
15205 Despite the nomenclature, `default' always means public ie;
15206 available to be linked against from outside the shared object.
15207 `protected' and `internal' are pretty useless in real-world usage
15208 so the only other commonly used option will be `hidden'. The
15209 default if `-fvisibility' isn't specified is `default', i.e., make
15210 every symbol public--this causes the same behavior as previous
15213 A good explanation of the benefits offered by ensuring ELF symbols
15214 have the correct visibility is given by "How To Write Shared
15215 Libraries" by Ulrich Drepper (which can be found at
15216 `http://people.redhat.com/~drepper/')--however a superior solution
15217 made possible by this option to marking things hidden when the
15218 default is public is to make the default hidden and mark things
15219 public. This is the norm with DLL's on Windows and with
15220 `-fvisibility=hidden' and `__attribute__
15221 ((visibility("default")))' instead of `__declspec(dllexport)' you
15222 get almost identical semantics with identical syntax. This is a
15223 great boon to those working with cross-platform projects.
15225 For those adding visibility support to existing code, you may find
15226 `#pragma GCC visibility' of use. This works by you enclosing the
15227 declarations you wish to set visibility for with (for example)
15228 `#pragma GCC visibility push(hidden)' and `#pragma GCC visibility
15229 pop'. Bear in mind that symbol visibility should be viewed *as
15230 part of the API interface contract* and thus all new code should
15231 always specify visibility when it is not the default ie;
15232 declarations only for use within the local DSO should *always* be
15233 marked explicitly as hidden as so to avoid PLT indirection
15234 overheads--making this abundantly clear also aids readability and
15235 self-documentation of the code. Note that due to ISO C++
15236 specification requirements, operator new and operator delete must
15237 always be of default visibility.
15239 Be aware that headers from outside your project, in particular
15240 system headers and headers from any other library you use, may not
15241 be expecting to be compiled with visibility other than the
15242 default. You may need to explicitly say `#pragma GCC visibility
15243 push(default)' before including any such headers.
15245 `extern' declarations are not affected by `-fvisibility', so a lot
15246 of code can be recompiled with `-fvisibility=hidden' with no
15247 modifications. However, this means that calls to `extern'
15248 functions with no explicit visibility will use the PLT, so it is
15249 more effective to use `__attribute ((visibility))' and/or `#pragma
15250 GCC visibility' to tell the compiler which `extern' declarations
15251 should be treated as hidden.
15253 Note that `-fvisibility' does affect C++ vague linkage entities.
15254 This means that, for instance, an exception class that will be
15255 thrown between DSOs must be explicitly marked with default
15256 visibility so that the `type_info' nodes will be unified between
15259 An overview of these techniques, their benefits and how to use them
15260 is at `http://gcc.gnu.org/wiki/Visibility'.
15264 File: gcc.info, Node: Environment Variables, Next: Precompiled Headers, Prev: Code Gen Options, Up: Invoking GCC
15266 3.19 Environment Variables Affecting GCC
15267 ========================================
15269 This section describes several environment variables that affect how GCC
15270 operates. Some of them work by specifying directories or prefixes to
15271 use when searching for various kinds of files. Some are used to
15272 specify other aspects of the compilation environment.
15274 Note that you can also specify places to search using options such as
15275 `-B', `-I' and `-L' (*note Directory Options::). These take precedence
15276 over places specified using environment variables, which in turn take
15277 precedence over those specified by the configuration of GCC. *Note
15278 Controlling the Compilation Driver `gcc': (gccint)Driver.
15284 These environment variables control the way that GCC uses
15285 localization information that allow GCC to work with different
15286 national conventions. GCC inspects the locale categories
15287 `LC_CTYPE' and `LC_MESSAGES' if it has been configured to do so.
15288 These locale categories can be set to any value supported by your
15289 installation. A typical value is `en_GB.UTF-8' for English in the
15290 United Kingdom encoded in UTF-8.
15292 The `LC_CTYPE' environment variable specifies character
15293 classification. GCC uses it to determine the character boundaries
15294 in a string; this is needed for some multibyte encodings that
15295 contain quote and escape characters that would otherwise be
15296 interpreted as a string end or escape.
15298 The `LC_MESSAGES' environment variable specifies the language to
15299 use in diagnostic messages.
15301 If the `LC_ALL' environment variable is set, it overrides the value
15302 of `LC_CTYPE' and `LC_MESSAGES'; otherwise, `LC_CTYPE' and
15303 `LC_MESSAGES' default to the value of the `LANG' environment
15304 variable. If none of these variables are set, GCC defaults to
15305 traditional C English behavior.
15308 If `TMPDIR' is set, it specifies the directory to use for temporary
15309 files. GCC uses temporary files to hold the output of one stage of
15310 compilation which is to be used as input to the next stage: for
15311 example, the output of the preprocessor, which is the input to the
15315 If `GCC_EXEC_PREFIX' is set, it specifies a prefix to use in the
15316 names of the subprograms executed by the compiler. No slash is
15317 added when this prefix is combined with the name of a subprogram,
15318 but you can specify a prefix that ends with a slash if you wish.
15320 If `GCC_EXEC_PREFIX' is not set, GCC will attempt to figure out an
15321 appropriate prefix to use based on the pathname it was invoked
15324 If GCC cannot find the subprogram using the specified prefix, it
15325 tries looking in the usual places for the subprogram.
15327 The default value of `GCC_EXEC_PREFIX' is `PREFIX/lib/gcc/' where
15328 PREFIX is the prefix to the installed compiler. In many cases
15329 PREFIX is the value of `prefix' when you ran the `configure'
15332 Other prefixes specified with `-B' take precedence over this
15335 This prefix is also used for finding files such as `crt0.o' that
15336 are used for linking.
15338 In addition, the prefix is used in an unusual way in finding the
15339 directories to search for header files. For each of the standard
15340 directories whose name normally begins with `/usr/local/lib/gcc'
15341 (more precisely, with the value of `GCC_INCLUDE_DIR'), GCC tries
15342 replacing that beginning with the specified prefix to produce an
15343 alternate directory name. Thus, with `-Bfoo/', GCC will search
15344 `foo/bar' where it would normally search `/usr/local/lib/bar'.
15345 These alternate directories are searched first; the standard
15346 directories come next. If a standard directory begins with the
15347 configured PREFIX then the value of PREFIX is replaced by
15348 `GCC_EXEC_PREFIX' when looking for header files.
15351 The value of `COMPILER_PATH' is a colon-separated list of
15352 directories, much like `PATH'. GCC tries the directories thus
15353 specified when searching for subprograms, if it can't find the
15354 subprograms using `GCC_EXEC_PREFIX'.
15357 The value of `LIBRARY_PATH' is a colon-separated list of
15358 directories, much like `PATH'. When configured as a native
15359 compiler, GCC tries the directories thus specified when searching
15360 for special linker files, if it can't find them using
15361 `GCC_EXEC_PREFIX'. Linking using GCC also uses these directories
15362 when searching for ordinary libraries for the `-l' option (but
15363 directories specified with `-L' come first).
15366 This variable is used to pass locale information to the compiler.
15367 One way in which this information is used is to determine the
15368 character set to be used when character literals, string literals
15369 and comments are parsed in C and C++. When the compiler is
15370 configured to allow multibyte characters, the following values for
15371 `LANG' are recognized:
15374 Recognize JIS characters.
15377 Recognize SJIS characters.
15380 Recognize EUCJP characters.
15382 If `LANG' is not defined, or if it has some other value, then the
15383 compiler will use mblen and mbtowc as defined by the default
15384 locale to recognize and translate multibyte characters.
15386 Some additional environments variables affect the behavior of the
15391 `CPLUS_INCLUDE_PATH'
15392 `OBJC_INCLUDE_PATH'
15393 Each variable's value is a list of directories separated by a
15394 special character, much like `PATH', in which to look for header
15395 files. The special character, `PATH_SEPARATOR', is
15396 target-dependent and determined at GCC build time. For Microsoft
15397 Windows-based targets it is a semicolon, and for almost all other
15398 targets it is a colon.
15400 `CPATH' specifies a list of directories to be searched as if
15401 specified with `-I', but after any paths given with `-I' options
15402 on the command line. This environment variable is used regardless
15403 of which language is being preprocessed.
15405 The remaining environment variables apply only when preprocessing
15406 the particular language indicated. Each specifies a list of
15407 directories to be searched as if specified with `-isystem', but
15408 after any paths given with `-isystem' options on the command line.
15410 In all these variables, an empty element instructs the compiler to
15411 search its current working directory. Empty elements can appear
15412 at the beginning or end of a path. For instance, if the value of
15413 `CPATH' is `:/special/include', that has the same effect as
15414 `-I. -I/special/include'.
15416 `DEPENDENCIES_OUTPUT'
15417 If this variable is set, its value specifies how to output
15418 dependencies for Make based on the non-system header files
15419 processed by the compiler. System header files are ignored in the
15422 The value of `DEPENDENCIES_OUTPUT' can be just a file name, in
15423 which case the Make rules are written to that file, guessing the
15424 target name from the source file name. Or the value can have the
15425 form `FILE TARGET', in which case the rules are written to file
15426 FILE using TARGET as the target name.
15428 In other words, this environment variable is equivalent to
15429 combining the options `-MM' and `-MF' (*note Preprocessor
15430 Options::), with an optional `-MT' switch too.
15432 `SUNPRO_DEPENDENCIES'
15433 This variable is the same as `DEPENDENCIES_OUTPUT' (see above),
15434 except that system header files are not ignored, so it implies
15435 `-M' rather than `-MM'. However, the dependence on the main input
15436 file is omitted. *Note Preprocessor Options::.
15439 File: gcc.info, Node: Precompiled Headers, Next: Running Protoize, Prev: Environment Variables, Up: Invoking GCC
15441 3.20 Using Precompiled Headers
15442 ==============================
15444 Often large projects have many header files that are included in every
15445 source file. The time the compiler takes to process these header files
15446 over and over again can account for nearly all of the time required to
15447 build the project. To make builds faster, GCC allows users to
15448 `precompile' a header file; then, if builds can use the precompiled
15449 header file they will be much faster.
15451 To create a precompiled header file, simply compile it as you would any
15452 other file, if necessary using the `-x' option to make the driver treat
15453 it as a C or C++ header file. You will probably want to use a tool
15454 like `make' to keep the precompiled header up-to-date when the headers
15455 it contains change.
15457 A precompiled header file will be searched for when `#include' is seen
15458 in the compilation. As it searches for the included file (*note Search
15459 Path: (cpp)Search Path.) the compiler looks for a precompiled header in
15460 each directory just before it looks for the include file in that
15461 directory. The name searched for is the name specified in the
15462 `#include' with `.gch' appended. If the precompiled header file can't
15463 be used, it is ignored.
15465 For instance, if you have `#include "all.h"', and you have `all.h.gch'
15466 in the same directory as `all.h', then the precompiled header file will
15467 be used if possible, and the original header will be used otherwise.
15469 Alternatively, you might decide to put the precompiled header file in a
15470 directory and use `-I' to ensure that directory is searched before (or
15471 instead of) the directory containing the original header. Then, if you
15472 want to check that the precompiled header file is always used, you can
15473 put a file of the same name as the original header in this directory
15474 containing an `#error' command.
15476 This also works with `-include'. So yet another way to use
15477 precompiled headers, good for projects not designed with precompiled
15478 header files in mind, is to simply take most of the header files used by
15479 a project, include them from another header file, precompile that header
15480 file, and `-include' the precompiled header. If the header files have
15481 guards against multiple inclusion, they will be skipped because they've
15482 already been included (in the precompiled header).
15484 If you need to precompile the same header file for different
15485 languages, targets, or compiler options, you can instead make a
15486 _directory_ named like `all.h.gch', and put each precompiled header in
15487 the directory, perhaps using `-o'. It doesn't matter what you call the
15488 files in the directory, every precompiled header in the directory will
15489 be considered. The first precompiled header encountered in the
15490 directory that is valid for this compilation will be used; they're
15491 searched in no particular order.
15493 There are many other possibilities, limited only by your imagination,
15494 good sense, and the constraints of your build system.
15496 A precompiled header file can be used only when these conditions apply:
15498 * Only one precompiled header can be used in a particular
15501 * A precompiled header can't be used once the first C token is seen.
15502 You can have preprocessor directives before a precompiled header;
15503 you can even include a precompiled header from inside another
15504 header, so long as there are no C tokens before the `#include'.
15506 * The precompiled header file must be produced for the same language
15507 as the current compilation. You can't use a C precompiled header
15508 for a C++ compilation.
15510 * The precompiled header file must have been produced by the same
15511 compiler binary as the current compilation is using.
15513 * Any macros defined before the precompiled header is included must
15514 either be defined in the same way as when the precompiled header
15515 was generated, or must not affect the precompiled header, which
15516 usually means that they don't appear in the precompiled header at
15519 The `-D' option is one way to define a macro before a precompiled
15520 header is included; using a `#define' can also do it. There are
15521 also some options that define macros implicitly, like `-O' and
15522 `-Wdeprecated'; the same rule applies to macros defined this way.
15524 * If debugging information is output when using the precompiled
15525 header, using `-g' or similar, the same kind of debugging
15526 information must have been output when building the precompiled
15527 header. However, a precompiled header built using `-g' can be
15528 used in a compilation when no debugging information is being
15531 * The same `-m' options must generally be used when building and
15532 using the precompiled header. *Note Submodel Options::, for any
15533 cases where this rule is relaxed.
15535 * Each of the following options must be the same when building and
15536 using the precompiled header:
15540 * Some other command-line options starting with `-f', `-p', or `-O'
15541 must be defined in the same way as when the precompiled header was
15542 generated. At present, it's not clear which options are safe to
15543 change and which are not; the safest choice is to use exactly the
15544 same options when generating and using the precompiled header.
15545 The following are known to be safe:
15547 -fmessage-length= -fpreprocessed -fsched-interblock
15548 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
15549 -fsched-verbose=<number> -fschedule-insns -fvisibility=
15553 For all of these except the last, the compiler will automatically
15554 ignore the precompiled header if the conditions aren't met. If you
15555 find an option combination that doesn't work and doesn't cause the
15556 precompiled header to be ignored, please consider filing a bug report,
15559 If you do use differing options when generating and using the
15560 precompiled header, the actual behavior will be a mixture of the
15561 behavior for the options. For instance, if you use `-g' to generate
15562 the precompiled header but not when using it, you may or may not get
15563 debugging information for routines in the precompiled header.
15566 File: gcc.info, Node: Running Protoize, Prev: Precompiled Headers, Up: Invoking GCC
15568 3.21 Running Protoize
15569 =====================
15571 The program `protoize' is an optional part of GCC. You can use it to
15572 add prototypes to a program, thus converting the program to ISO C in
15573 one respect. The companion program `unprotoize' does the reverse: it
15574 removes argument types from any prototypes that are found.
15576 When you run these programs, you must specify a set of source files as
15577 command line arguments. The conversion programs start out by compiling
15578 these files to see what functions they define. The information gathered
15579 about a file FOO is saved in a file named `FOO.X'.
15581 After scanning comes actual conversion. The specified files are all
15582 eligible to be converted; any files they include (whether sources or
15583 just headers) are eligible as well.
15585 But not all the eligible files are converted. By default, `protoize'
15586 and `unprotoize' convert only source and header files in the current
15587 directory. You can specify additional directories whose files should
15588 be converted with the `-d DIRECTORY' option. You can also specify
15589 particular files to exclude with the `-x FILE' option. A file is
15590 converted if it is eligible, its directory name matches one of the
15591 specified directory names, and its name within the directory has not
15594 Basic conversion with `protoize' consists of rewriting most function
15595 definitions and function declarations to specify the types of the
15596 arguments. The only ones not rewritten are those for varargs functions.
15598 `protoize' optionally inserts prototype declarations at the beginning
15599 of the source file, to make them available for any calls that precede
15600 the function's definition. Or it can insert prototype declarations
15601 with block scope in the blocks where undeclared functions are called.
15603 Basic conversion with `unprotoize' consists of rewriting most function
15604 declarations to remove any argument types, and rewriting function
15605 definitions to the old-style pre-ISO form.
15607 Both conversion programs print a warning for any function declaration
15608 or definition that they can't convert. You can suppress these warnings
15611 The output from `protoize' or `unprotoize' replaces the original
15612 source file. The original file is renamed to a name ending with
15613 `.save' (for DOS, the saved filename ends in `.sav' without the
15614 original `.c' suffix). If the `.save' (`.sav' for DOS) file already
15615 exists, then the source file is simply discarded.
15617 `protoize' and `unprotoize' both depend on GCC itself to scan the
15618 program and collect information about the functions it uses. So
15619 neither of these programs will work until GCC is installed.
15621 Here is a table of the options you can use with `protoize' and
15622 `unprotoize'. Each option works with both programs unless otherwise
15626 Look for the file `SYSCALLS.c.X' in DIRECTORY, instead of the
15627 usual directory (normally `/usr/local/lib'). This file contains
15628 prototype information about standard system functions. This option
15629 applies only to `protoize'.
15631 `-c COMPILATION-OPTIONS'
15632 Use COMPILATION-OPTIONS as the options when running `gcc' to
15633 produce the `.X' files. The special option `-aux-info' is always
15634 passed in addition, to tell `gcc' to write a `.X' file.
15636 Note that the compilation options must be given as a single
15637 argument to `protoize' or `unprotoize'. If you want to specify
15638 several `gcc' options, you must quote the entire set of
15639 compilation options to make them a single word in the shell.
15641 There are certain `gcc' arguments that you cannot use, because they
15642 would produce the wrong kind of output. These include `-g', `-O',
15643 `-c', `-S', and `-o' If you include these in the
15644 COMPILATION-OPTIONS, they are ignored.
15647 Rename files to end in `.C' (`.cc' for DOS-based file systems)
15648 instead of `.c'. This is convenient if you are converting a C
15649 program to C++. This option applies only to `protoize'.
15652 Add explicit global declarations. This means inserting explicit
15653 declarations at the beginning of each source file for each function
15654 that is called in the file and was not declared. These
15655 declarations precede the first function definition that contains a
15656 call to an undeclared function. This option applies only to
15660 Indent old-style parameter declarations with the string STRING.
15661 This option applies only to `protoize'.
15663 `unprotoize' converts prototyped function definitions to old-style
15664 function definitions, where the arguments are declared between the
15665 argument list and the initial `{'. By default, `unprotoize' uses
15666 five spaces as the indentation. If you want to indent with just
15667 one space instead, use `-i " "'.
15670 Keep the `.X' files. Normally, they are deleted after conversion
15674 Add explicit local declarations. `protoize' with `-l' inserts a
15675 prototype declaration for each function in each block which calls
15676 the function without any declaration. This option applies only to
15680 Make no real changes. This mode just prints information about the
15681 conversions that would have been done without `-n'.
15684 Make no `.save' files. The original files are simply deleted.
15685 Use this option with caution.
15688 Use the program PROGRAM as the compiler. Normally, the name `gcc'
15692 Work quietly. Most warnings are suppressed.
15695 Print the version number, just like `-v' for `gcc'.
15697 If you need special compiler options to compile one of your program's
15698 source files, then you should generate that file's `.X' file specially,
15699 by running `gcc' on that source file with the appropriate options and
15700 the option `-aux-info'. Then run `protoize' on the entire set of
15701 files. `protoize' will use the existing `.X' file because it is newer
15702 than the source file. For example:
15704 gcc -Dfoo=bar file1.c -aux-info file1.X
15707 You need to include the special files along with the rest in the
15708 `protoize' command, even though their `.X' files already exist, because
15709 otherwise they won't get converted.
15711 *Note Protoize Caveats::, for more information on how to use
15712 `protoize' successfully.
15715 File: gcc.info, Node: C Implementation, Next: C Extensions, Prev: Invoking GCC, Up: Top
15717 4 C Implementation-defined behavior
15718 ***********************************
15720 A conforming implementation of ISO C is required to document its choice
15721 of behavior in each of the areas that are designated "implementation
15722 defined". The following lists all such areas, along with the section
15723 numbers from the ISO/IEC 9899:1990 and ISO/IEC 9899:1999 standards.
15724 Some areas are only implementation-defined in one version of the
15727 Some choices depend on the externally determined ABI for the platform
15728 (including standard character encodings) which GCC follows; these are
15729 listed as "determined by ABI" below. *Note Binary Compatibility:
15730 Compatibility, and `http://gcc.gnu.org/readings.html'. Some choices
15731 are documented in the preprocessor manual. *Note
15732 Implementation-defined behavior: (cpp)Implementation-defined behavior.
15733 Some choices are made by the library and operating system (or other
15734 environment when compiling for a freestanding environment); refer to
15735 their documentation for details.
15739 * Translation implementation::
15740 * Environment implementation::
15741 * Identifiers implementation::
15742 * Characters implementation::
15743 * Integers implementation::
15744 * Floating point implementation::
15745 * Arrays and pointers implementation::
15746 * Hints implementation::
15747 * Structures unions enumerations and bit-fields implementation::
15748 * Qualifiers implementation::
15749 * Declarators implementation::
15750 * Statements implementation::
15751 * Preprocessing directives implementation::
15752 * Library functions implementation::
15753 * Architecture implementation::
15754 * Locale-specific behavior implementation::
15757 File: gcc.info, Node: Translation implementation, Next: Environment implementation, Up: C Implementation
15762 * `How a diagnostic is identified (C90 3.7, C99 3.10, C90 and C99
15765 Diagnostics consist of all the output sent to stderr by GCC.
15767 * `Whether each nonempty sequence of white-space characters other
15768 than new-line is retained or replaced by one space character in
15769 translation phase 3 (C90 and C99 5.1.1.2).'
15771 *Note Implementation-defined behavior: (cpp)Implementation-defined
15776 File: gcc.info, Node: Environment implementation, Next: Identifiers implementation, Prev: Translation implementation, Up: C Implementation
15781 The behavior of most of these points are dependent on the implementation
15782 of the C library, and are not defined by GCC itself.
15784 * `The mapping between physical source file multibyte characters and
15785 the source character set in translation phase 1 (C90 and C99
15788 *Note Implementation-defined behavior: (cpp)Implementation-defined
15793 File: gcc.info, Node: Identifiers implementation, Next: Characters implementation, Prev: Environment implementation, Up: C Implementation
15798 * `Which additional multibyte characters may appear in identifiers
15799 and their correspondence to universal character names (C99 6.4.2).'
15801 *Note Implementation-defined behavior: (cpp)Implementation-defined
15804 * `The number of significant initial characters in an identifier
15805 (C90 6.1.2, C90 and C99 5.2.4.1, C99 6.4.2).'
15807 For internal names, all characters are significant. For external
15808 names, the number of significant characters are defined by the
15809 linker; for almost all targets, all characters are significant.
15811 * `Whether case distinctions are significant in an identifier with
15812 external linkage (C90 6.1.2).'
15814 This is a property of the linker. C99 requires that case
15815 distinctions are always significant in identifiers with external
15816 linkage and systems without this property are not supported by GCC.
15820 File: gcc.info, Node: Characters implementation, Next: Integers implementation, Prev: Identifiers implementation, Up: C Implementation
15825 * `The number of bits in a byte (C90 3.4, C99 3.6).'
15829 * `The values of the members of the execution character set (C90 and
15834 * `The unique value of the member of the execution character set
15835 produced for each of the standard alphabetic escape sequences (C90
15840 * `The value of a `char' object into which has been stored any
15841 character other than a member of the basic execution character set
15842 (C90 6.1.2.5, C99 6.2.5).'
15846 * `Which of `signed char' or `unsigned char' has the same range,
15847 representation, and behavior as "plain" `char' (C90 6.1.2.5, C90
15848 6.2.1.1, C99 6.2.5, C99 6.3.1.1).'
15850 Determined by ABI. The options `-funsigned-char' and
15851 `-fsigned-char' change the default. *Note Options Controlling C
15852 Dialect: C Dialect Options.
15854 * `The mapping of members of the source character set (in character
15855 constants and string literals) to members of the execution
15856 character set (C90 6.1.3.4, C99 6.4.4.4, C90 and C99 5.1.1.2).'
15860 * `The value of an integer character constant containing more than
15861 one character or containing a character or escape sequence that
15862 does not map to a single-byte execution character (C90 6.1.3.4,
15865 *Note Implementation-defined behavior: (cpp)Implementation-defined
15868 * `The value of a wide character constant containing more than one
15869 multibyte character, or containing a multibyte character or escape
15870 sequence not represented in the extended execution character set
15871 (C90 6.1.3.4, C99 6.4.4.4).'
15873 *Note Implementation-defined behavior: (cpp)Implementation-defined
15876 * `The current locale used to convert a wide character constant
15877 consisting of a single multibyte character that maps to a member
15878 of the extended execution character set into a corresponding wide
15879 character code (C90 6.1.3.4, C99 6.4.4.4).'
15881 *Note Implementation-defined behavior: (cpp)Implementation-defined
15884 * `The current locale used to convert a wide string literal into
15885 corresponding wide character codes (C90 6.1.4, C99 6.4.5).'
15887 *Note Implementation-defined behavior: (cpp)Implementation-defined
15890 * `The value of a string literal containing a multibyte character or
15891 escape sequence not represented in the execution character set
15892 (C90 6.1.4, C99 6.4.5).'
15894 *Note Implementation-defined behavior: (cpp)Implementation-defined
15898 File: gcc.info, Node: Integers implementation, Next: Floating point implementation, Prev: Characters implementation, Up: C Implementation
15903 * `Any extended integer types that exist in the implementation (C99
15906 GCC does not support any extended integer types.
15908 * `Whether signed integer types are represented using sign and
15909 magnitude, two's complement, or one's complement, and whether the
15910 extraordinary value is a trap representation or an ordinary value
15913 GCC supports only two's complement integer types, and all bit
15914 patterns are ordinary values.
15916 * `The rank of any extended integer type relative to another extended
15917 integer type with the same precision (C99 6.3.1.1).'
15919 GCC does not support any extended integer types.
15921 * `The result of, or the signal raised by, converting an integer to a
15922 signed integer type when the value cannot be represented in an
15923 object of that type (C90 6.2.1.2, C99 6.3.1.3).'
15925 For conversion to a type of width N, the value is reduced modulo
15926 2^N to be within range of the type; no signal is raised.
15928 * `The results of some bitwise operations on signed integers (C90
15931 Bitwise operators act on the representation of the value including
15932 both the sign and value bits, where the sign bit is considered
15933 immediately above the highest-value value bit. Signed `>>' acts
15934 on negative numbers by sign extension.
15936 GCC does not use the latitude given in C99 only to treat certain
15937 aspects of signed `<<' as undefined, but this is subject to change.
15939 * `The sign of the remainder on integer division (C90 6.3.5).'
15941 GCC always follows the C99 requirement that the result of division
15942 is truncated towards zero.
15946 File: gcc.info, Node: Floating point implementation, Next: Arrays and pointers implementation, Prev: Integers implementation, Up: C Implementation
15951 * `The accuracy of the floating-point operations and of the library
15952 functions in `<math.h>' and `<complex.h>' that return
15953 floating-point results (C90 and C99 5.2.4.2.2).'
15955 The accuracy is unknown.
15957 * `The rounding behaviors characterized by non-standard values of
15958 `FLT_ROUNDS' (C90 and C99 5.2.4.2.2).'
15960 GCC does not use such values.
15962 * `The evaluation methods characterized by non-standard negative
15963 values of `FLT_EVAL_METHOD' (C99 5.2.4.2.2).'
15965 GCC does not use such values.
15967 * `The direction of rounding when an integer is converted to a
15968 floating-point number that cannot exactly represent the original
15969 value (C90 6.2.1.3, C99 6.3.1.4).'
15971 C99 Annex F is followed.
15973 * `The direction of rounding when a floating-point number is
15974 converted to a narrower floating-point number (C90 6.2.1.4, C99
15977 C99 Annex F is followed.
15979 * `How the nearest representable value or the larger or smaller
15980 representable value immediately adjacent to the nearest
15981 representable value is chosen for certain floating constants (C90
15982 6.1.3.1, C99 6.4.4.2).'
15984 C99 Annex F is followed.
15986 * `Whether and how floating expressions are contracted when not
15987 disallowed by the `FP_CONTRACT' pragma (C99 6.5).'
15989 Expressions are currently only contracted if
15990 `-funsafe-math-optimizations' or `-ffast-math' are used. This is
15993 * `The default state for the `FENV_ACCESS' pragma (C99 7.6.1).'
15995 This pragma is not implemented, but the default is to "off" unless
15996 `-frounding-math' is used in which case it is "on".
15998 * `Additional floating-point exceptions, rounding modes,
15999 environments, and classifications, and their macro names (C99 7.6,
16002 This is dependent on the implementation of the C library, and is
16003 not defined by GCC itself.
16005 * `The default state for the `FP_CONTRACT' pragma (C99 7.12.2).'
16007 This pragma is not implemented. Expressions are currently only
16008 contracted if `-funsafe-math-optimizations' or `-ffast-math' are
16009 used. This is subject to change.
16011 * `Whether the "inexact" floating-point exception can be raised when
16012 the rounded result actually does equal the mathematical result in
16013 an IEC 60559 conformant implementation (C99 F.9).'
16015 This is dependent on the implementation of the C library, and is
16016 not defined by GCC itself.
16018 * `Whether the "underflow" (and "inexact") floating-point exception
16019 can be raised when a result is tiny but not inexact in an IEC
16020 60559 conformant implementation (C99 F.9).'
16022 This is dependent on the implementation of the C library, and is
16023 not defined by GCC itself.
16027 File: gcc.info, Node: Arrays and pointers implementation, Next: Hints implementation, Prev: Floating point implementation, Up: C Implementation
16029 4.7 Arrays and pointers
16030 =======================
16032 * `The result of converting a pointer to an integer or vice versa
16033 (C90 6.3.4, C99 6.3.2.3).'
16035 A cast from pointer to integer discards most-significant bits if
16036 the pointer representation is larger than the integer type,
16037 sign-extends(1) if the pointer representation is smaller than the
16038 integer type, otherwise the bits are unchanged.
16040 A cast from integer to pointer discards most-significant bits if
16041 the pointer representation is smaller than the integer type,
16042 extends according to the signedness of the integer type if the
16043 pointer representation is larger than the integer type, otherwise
16044 the bits are unchanged.
16046 When casting from pointer to integer and back again, the resulting
16047 pointer must reference the same object as the original pointer,
16048 otherwise the behavior is undefined. That is, one may not use
16049 integer arithmetic to avoid the undefined behavior of pointer
16050 arithmetic as proscribed in C99 6.5.6/8.
16052 * `The size of the result of subtracting two pointers to elements of
16053 the same array (C90 6.3.6, C99 6.5.6).'
16055 The value is as specified in the standard and the type is
16056 determined by the ABI.
16059 ---------- Footnotes ----------
16061 (1) Future versions of GCC may zero-extend, or use a target-defined
16062 `ptr_extend' pattern. Do not rely on sign extension.
16065 File: gcc.info, Node: Hints implementation, Next: Structures unions enumerations and bit-fields implementation, Prev: Arrays and pointers implementation, Up: C Implementation
16070 * `The extent to which suggestions made by using the `register'
16071 storage-class specifier are effective (C90 6.5.1, C99 6.7.1).'
16073 The `register' specifier affects code generation only in these
16076 * When used as part of the register variable extension, see
16077 *note Explicit Reg Vars::.
16079 * When `-O0' is in use, the compiler allocates distinct stack
16080 memory for all variables that do not have the `register'
16081 storage-class specifier; if `register' is specified, the
16082 variable may have a shorter lifespan than the code would
16083 indicate and may never be placed in memory.
16085 * On some rare x86 targets, `setjmp' doesn't save the registers
16086 in all circumstances. In those cases, GCC doesn't allocate
16087 any variables in registers unless they are marked `register'.
16090 * `The extent to which suggestions made by using the inline function
16091 specifier are effective (C99 6.7.4).'
16093 GCC will not inline any functions if the `-fno-inline' option is
16094 used or if `-O0' is used. Otherwise, GCC may still be unable to
16095 inline a function for many reasons; the `-Winline' option may be
16096 used to determine if a function has not been inlined and why not.
16100 File: gcc.info, Node: Structures unions enumerations and bit-fields implementation, Next: Qualifiers implementation, Prev: Hints implementation, Up: C Implementation
16102 4.9 Structures, unions, enumerations, and bit-fields
16103 ====================================================
16105 * `A member of a union object is accessed using a member of a
16106 different type (C90 6.3.2.3).'
16108 The relevant bytes of the representation of the object are treated
16109 as an object of the type used for the access. *Note
16110 Type-punning::. This may be a trap representation.
16112 * `Whether a "plain" `int' bit-field is treated as a `signed int'
16113 bit-field or as an `unsigned int' bit-field (C90 6.5.2, C90
16114 6.5.2.1, C99 6.7.2, C99 6.7.2.1).'
16116 By default it is treated as `signed int' but this may be changed
16117 by the `-funsigned-bitfields' option.
16119 * `Allowable bit-field types other than `_Bool', `signed int', and
16120 `unsigned int' (C99 6.7.2.1).'
16122 No other types are permitted in strictly conforming mode.
16124 * `Whether a bit-field can straddle a storage-unit boundary (C90
16125 6.5.2.1, C99 6.7.2.1).'
16129 * `The order of allocation of bit-fields within a unit (C90 6.5.2.1,
16134 * `The alignment of non-bit-field members of structures (C90
16135 6.5.2.1, C99 6.7.2.1).'
16139 * `The integer type compatible with each enumerated type (C90
16140 6.5.2.2, C99 6.7.2.2).'
16142 Normally, the type is `unsigned int' if there are no negative
16143 values in the enumeration, otherwise `int'. If `-fshort-enums' is
16144 specified, then if there are negative values it is the first of
16145 `signed char', `short' and `int' that can represent all the
16146 values, otherwise it is the first of `unsigned char', `unsigned
16147 short' and `unsigned int' that can represent all the values.
16149 On some targets, `-fshort-enums' is the default; this is
16150 determined by the ABI.
16154 File: gcc.info, Node: Qualifiers implementation, Next: Declarators implementation, Prev: Structures unions enumerations and bit-fields implementation, Up: C Implementation
16159 * `What constitutes an access to an object that has
16160 volatile-qualified type (C90 6.5.3, C99 6.7.3).'
16162 Such an object is normally accessed by pointers and used for
16163 accessing hardware. In most expressions, it is intuitively
16164 obvious what is a read and what is a write. For example
16166 volatile int *dst = SOMEVALUE;
16167 volatile int *src = SOMEOTHERVALUE;
16170 will cause a read of the volatile object pointed to by SRC and
16171 store the value into the volatile object pointed to by DST. There
16172 is no guarantee that these reads and writes are atomic, especially
16173 for objects larger than `int'.
16175 However, if the volatile storage is not being modified, and the
16176 value of the volatile storage is not used, then the situation is
16177 less obvious. For example
16179 volatile int *src = SOMEVALUE;
16182 According to the C standard, such an expression is an rvalue whose
16183 type is the unqualified version of its original type, i.e. `int'.
16184 Whether GCC interprets this as a read of the volatile object being
16185 pointed to or only as a request to evaluate the expression for its
16186 side-effects depends on this type.
16188 If it is a scalar type, or on most targets an aggregate type whose
16189 only member object is of a scalar type, or a union type whose
16190 member objects are of scalar types, the expression is interpreted
16191 by GCC as a read of the volatile object; in the other cases, the
16192 expression is only evaluated for its side-effects.
16196 File: gcc.info, Node: Declarators implementation, Next: Statements implementation, Prev: Qualifiers implementation, Up: C Implementation
16201 * `The maximum number of declarators that may modify an arithmetic,
16202 structure or union type (C90 6.5.4).'
16204 GCC is only limited by available memory.
16208 File: gcc.info, Node: Statements implementation, Next: Preprocessing directives implementation, Prev: Declarators implementation, Up: C Implementation
16213 * `The maximum number of `case' values in a `switch' statement (C90
16216 GCC is only limited by available memory.
16220 File: gcc.info, Node: Preprocessing directives implementation, Next: Library functions implementation, Prev: Statements implementation, Up: C Implementation
16222 4.13 Preprocessing directives
16223 =============================
16225 *Note Implementation-defined behavior: (cpp)Implementation-defined
16226 behavior, for details of these aspects of implementation-defined
16229 * `How sequences in both forms of header names are mapped to headers
16230 or external source file names (C90 6.1.7, C99 6.4.7).'
16232 * `Whether the value of a character constant in a constant expression
16233 that controls conditional inclusion matches the value of the same
16234 character constant in the execution character set (C90 6.8.1, C99
16237 * `Whether the value of a single-character character constant in a
16238 constant expression that controls conditional inclusion may have a
16239 negative value (C90 6.8.1, C99 6.10.1).'
16241 * `The places that are searched for an included `<>' delimited
16242 header, and how the places are specified or the header is
16243 identified (C90 6.8.2, C99 6.10.2).'
16245 * `How the named source file is searched for in an included `""'
16246 delimited header (C90 6.8.2, C99 6.10.2).'
16248 * `The method by which preprocessing tokens (possibly resulting from
16249 macro expansion) in a `#include' directive are combined into a
16250 header name (C90 6.8.2, C99 6.10.2).'
16252 * `The nesting limit for `#include' processing (C90 6.8.2, C99
16255 * `Whether the `#' operator inserts a `\' character before the `\'
16256 character that begins a universal character name in a character
16257 constant or string literal (C99 6.10.3.2).'
16259 * `The behavior on each recognized non-`STDC #pragma' directive (C90
16260 6.8.6, C99 6.10.6).'
16262 *Note Pragmas: (cpp)Pragmas, for details of pragmas accepted by
16263 GCC on all targets. *Note Pragmas Accepted by GCC: Pragmas, for
16264 details of target-specific pragmas.
16266 * `The definitions for `__DATE__' and `__TIME__' when respectively,
16267 the date and time of translation are not available (C90 6.8.8, C99
16272 File: gcc.info, Node: Library functions implementation, Next: Architecture implementation, Prev: Preprocessing directives implementation, Up: C Implementation
16274 4.14 Library functions
16275 ======================
16277 The behavior of most of these points are dependent on the implementation
16278 of the C library, and are not defined by GCC itself.
16280 * `The null pointer constant to which the macro `NULL' expands (C90
16283 In `<stddef.h>', `NULL' expands to `((void *)0)'. GCC does not
16284 provide the other headers which define `NULL' and some library
16285 implementations may use other definitions in those headers.
16289 File: gcc.info, Node: Architecture implementation, Next: Locale-specific behavior implementation, Prev: Library functions implementation, Up: C Implementation
16294 * `The values or expressions assigned to the macros specified in the
16295 headers `<float.h>', `<limits.h>', and `<stdint.h>' (C90 and C99
16296 5.2.4.2, C99 7.18.2, C99 7.18.3).'
16300 * `The number, order, and encoding of bytes in any object (when not
16301 explicitly specified in this International Standard) (C99
16306 * `The value of the result of the `sizeof' operator (C90 6.3.3.4,
16313 File: gcc.info, Node: Locale-specific behavior implementation, Prev: Architecture implementation, Up: C Implementation
16315 4.16 Locale-specific behavior
16316 =============================
16318 The behavior of these points are dependent on the implementation of the
16319 C library, and are not defined by GCC itself.
16322 File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: C Implementation, Up: Top
16324 5 Extensions to the C Language Family
16325 *************************************
16327 GNU C provides several language features not found in ISO standard C.
16328 (The `-pedantic' option directs GCC to print a warning message if any
16329 of these features is used.) To test for the availability of these
16330 features in conditional compilation, check for a predefined macro
16331 `__GNUC__', which is always defined under GCC.
16333 These extensions are available in C and Objective-C. Most of them are
16334 also available in C++. *Note Extensions to the C++ Language: C++
16335 Extensions, for extensions that apply _only_ to C++.
16337 Some features that are in ISO C99 but not C89 or C++ are also, as
16338 extensions, accepted by GCC in C89 mode and in C++.
16342 * Statement Exprs:: Putting statements and declarations inside expressions.
16343 * Local Labels:: Labels local to a block.
16344 * Labels as Values:: Getting pointers to labels, and computed gotos.
16345 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
16346 * Constructing Calls:: Dispatching a call to another function.
16347 * Typeof:: `typeof': referring to the type of an expression.
16348 * Conditionals:: Omitting the middle operand of a `?:' expression.
16349 * Long Long:: Double-word integers---`long long int'.
16350 * Complex:: Data types for complex numbers.
16351 * Floating Types:: Additional Floating Types.
16352 * Decimal Float:: Decimal Floating Types.
16353 * Hex Floats:: Hexadecimal floating-point constants.
16354 * Fixed-Point:: Fixed-Point Types.
16355 * Zero Length:: Zero-length arrays.
16356 * Variable Length:: Arrays whose length is computed at run time.
16357 * Empty Structures:: Structures with no members.
16358 * Variadic Macros:: Macros with a variable number of arguments.
16359 * Escaped Newlines:: Slightly looser rules for escaped newlines.
16360 * Subscripting:: Any array can be subscripted, even if not an lvalue.
16361 * Pointer Arith:: Arithmetic on `void'-pointers and function pointers.
16362 * Initializers:: Non-constant initializers.
16363 * Compound Literals:: Compound literals give structures, unions
16364 or arrays as values.
16365 * Designated Inits:: Labeling elements of initializers.
16366 * Cast to Union:: Casting to union type from any member of the union.
16367 * Case Ranges:: `case 1 ... 9' and such.
16368 * Mixed Declarations:: Mixing declarations and code.
16369 * Function Attributes:: Declaring that functions have no side effects,
16370 or that they can never return.
16371 * Attribute Syntax:: Formal syntax for attributes.
16372 * Function Prototypes:: Prototype declarations and old-style definitions.
16373 * C++ Comments:: C++ comments are recognized.
16374 * Dollar Signs:: Dollar sign is allowed in identifiers.
16375 * Character Escapes:: `\e' stands for the character <ESC>.
16376 * Variable Attributes:: Specifying attributes of variables.
16377 * Type Attributes:: Specifying attributes of types.
16378 * Alignment:: Inquiring about the alignment of a type or variable.
16379 * Inline:: Defining inline functions (as fast as macros).
16380 * Extended Asm:: Assembler instructions with C expressions as operands.
16381 (With them you can define ``built-in'' functions.)
16382 * Constraints:: Constraints for asm operands
16383 * Asm Labels:: Specifying the assembler name to use for a C symbol.
16384 * Explicit Reg Vars:: Defining variables residing in specified registers.
16385 * Alternate Keywords:: `__const__', `__asm__', etc., for header files.
16386 * Incomplete Enums:: `enum foo;', with details to follow.
16387 * Function Names:: Printable strings which are the name of the current
16389 * Return Address:: Getting the return or frame address of a function.
16390 * Vector Extensions:: Using vector instructions through built-in functions.
16391 * Offsetof:: Special syntax for implementing `offsetof'.
16392 * Atomic Builtins:: Built-in functions for atomic memory access.
16393 * Object Size Checking:: Built-in functions for limited buffer overflow
16395 * Other Builtins:: Other built-in functions.
16396 * Target Builtins:: Built-in functions specific to particular targets.
16397 * Target Format Checks:: Format checks specific to particular targets.
16398 * Pragmas:: Pragmas accepted by GCC.
16399 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
16400 * Thread-Local:: Per-thread variables.
16401 * Binary constants:: Binary constants using the `0b' prefix.
16404 File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions
16406 5.1 Statements and Declarations in Expressions
16407 ==============================================
16409 A compound statement enclosed in parentheses may appear as an expression
16410 in GNU C. This allows you to use loops, switches, and local variables
16411 within an expression.
16413 Recall that a compound statement is a sequence of statements surrounded
16414 by braces; in this construct, parentheses go around the braces. For
16417 ({ int y = foo (); int z;
16422 is a valid (though slightly more complex than necessary) expression for
16423 the absolute value of `foo ()'.
16425 The last thing in the compound statement should be an expression
16426 followed by a semicolon; the value of this subexpression serves as the
16427 value of the entire construct. (If you use some other kind of statement
16428 last within the braces, the construct has type `void', and thus
16429 effectively no value.)
16431 This feature is especially useful in making macro definitions "safe"
16432 (so that they evaluate each operand exactly once). For example, the
16433 "maximum" function is commonly defined as a macro in standard C as
16436 #define max(a,b) ((a) > (b) ? (a) : (b))
16438 But this definition computes either A or B twice, with bad results if
16439 the operand has side effects. In GNU C, if you know the type of the
16440 operands (here taken as `int'), you can define the macro safely as
16443 #define maxint(a,b) \
16444 ({int _a = (a), _b = (b); _a > _b ? _a : _b; })
16446 Embedded statements are not allowed in constant expressions, such as
16447 the value of an enumeration constant, the width of a bit-field, or the
16448 initial value of a static variable.
16450 If you don't know the type of the operand, you can still do this, but
16451 you must use `typeof' (*note Typeof::).
16453 In G++, the result value of a statement expression undergoes array and
16454 function pointer decay, and is returned by value to the enclosing
16455 expression. For instance, if `A' is a class, then
16461 will construct a temporary `A' object to hold the result of the
16462 statement expression, and that will be used to invoke `Foo'. Therefore
16463 the `this' pointer observed by `Foo' will not be the address of `a'.
16465 Any temporaries created within a statement within a statement
16466 expression will be destroyed at the statement's end. This makes
16467 statement expressions inside macros slightly different from function
16468 calls. In the latter case temporaries introduced during argument
16469 evaluation will be destroyed at the end of the statement that includes
16470 the function call. In the statement expression case they will be
16471 destroyed during the statement expression. For instance,
16473 #define macro(a) ({__typeof__(a) b = (a); b + 3; })
16474 template<typename T> T function(T a) { T b = a; return b + 3; }
16482 will have different places where temporaries are destroyed. For the
16483 `macro' case, the temporary `X' will be destroyed just after the
16484 initialization of `b'. In the `function' case that temporary will be
16485 destroyed when the function returns.
16487 These considerations mean that it is probably a bad idea to use
16488 statement-expressions of this form in header files that are designed to
16489 work with C++. (Note that some versions of the GNU C Library contained
16490 header files using statement-expression that lead to precisely this
16493 Jumping into a statement expression with `goto' or using a `switch'
16494 statement outside the statement expression with a `case' or `default'
16495 label inside the statement expression is not permitted. Jumping into a
16496 statement expression with a computed `goto' (*note Labels as Values::)
16497 yields undefined behavior. Jumping out of a statement expression is
16498 permitted, but if the statement expression is part of a larger
16499 expression then it is unspecified which other subexpressions of that
16500 expression have been evaluated except where the language definition
16501 requires certain subexpressions to be evaluated before or after the
16502 statement expression. In any case, as with a function call the
16503 evaluation of a statement expression is not interleaved with the
16504 evaluation of other parts of the containing expression. For example,
16506 foo (), (({ bar1 (); goto a; 0; }) + bar2 ()), baz();
16508 will call `foo' and `bar1' and will not call `baz' but may or may not
16509 call `bar2'. If `bar2' is called, it will be called after `foo' and
16513 File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions
16515 5.2 Locally Declared Labels
16516 ===========================
16518 GCC allows you to declare "local labels" in any nested block scope. A
16519 local label is just like an ordinary label, but you can only reference
16520 it (with a `goto' statement, or by taking its address) within the block
16521 in which it was declared.
16523 A local label declaration looks like this:
16529 __label__ LABEL1, LABEL2, /* ... */;
16531 Local label declarations must come at the beginning of the block,
16532 before any ordinary declarations or statements.
16534 The label declaration defines the label _name_, but does not define
16535 the label itself. You must do this in the usual way, with `LABEL:',
16536 within the statements of the statement expression.
16538 The local label feature is useful for complex macros. If a macro
16539 contains nested loops, a `goto' can be useful for breaking out of them.
16540 However, an ordinary label whose scope is the whole function cannot be
16541 used: if the macro can be expanded several times in one function, the
16542 label will be multiply defined in that function. A local label avoids
16543 this problem. For example:
16545 #define SEARCH(value, array, target) \
16548 typeof (target) _SEARCH_target = (target); \
16549 typeof (*(array)) *_SEARCH_array = (array); \
16552 for (i = 0; i < max; i++) \
16553 for (j = 0; j < max; j++) \
16554 if (_SEARCH_array[i][j] == _SEARCH_target) \
16555 { (value) = i; goto found; } \
16560 This could also be written using a statement-expression:
16562 #define SEARCH(array, target) \
16565 typeof (target) _SEARCH_target = (target); \
16566 typeof (*(array)) *_SEARCH_array = (array); \
16569 for (i = 0; i < max; i++) \
16570 for (j = 0; j < max; j++) \
16571 if (_SEARCH_array[i][j] == _SEARCH_target) \
16572 { value = i; goto found; } \
16578 Local label declarations also make the labels they declare visible to
16579 nested functions, if there are any. *Note Nested Functions::, for
16583 File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions
16585 5.3 Labels as Values
16586 ====================
16588 You can get the address of a label defined in the current function (or
16589 a containing function) with the unary operator `&&'. The value has
16590 type `void *'. This value is a constant and can be used wherever a
16591 constant of that type is valid. For example:
16597 To use these values, you need to be able to jump to one. This is done
16598 with the computed goto statement(1), `goto *EXP;'. For example,
16602 Any expression of type `void *' is allowed.
16604 One way of using these constants is in initializing a static array that
16605 will serve as a jump table:
16607 static void *array[] = { &&foo, &&bar, &&hack };
16609 Then you can select a label with indexing, like this:
16613 Note that this does not check whether the subscript is in bounds--array
16614 indexing in C never does that.
16616 Such an array of label values serves a purpose much like that of the
16617 `switch' statement. The `switch' statement is cleaner, so use that
16618 rather than an array unless the problem does not fit a `switch'
16619 statement very well.
16621 Another use of label values is in an interpreter for threaded code.
16622 The labels within the interpreter function can be stored in the
16623 threaded code for super-fast dispatching.
16625 You may not use this mechanism to jump to code in a different function.
16626 If you do that, totally unpredictable things will happen. The best way
16627 to avoid this is to store the label address only in automatic variables
16628 and never pass it as an argument.
16630 An alternate way to write the above example is
16632 static const int array[] = { &&foo - &&foo, &&bar - &&foo,
16634 goto *(&&foo + array[i]);
16636 This is more friendly to code living in shared libraries, as it reduces
16637 the number of dynamic relocations that are needed, and by consequence,
16638 allows the data to be read-only.
16640 The `&&foo' expressions for the same label might have different values
16641 if the containing function is inlined or cloned. If a program relies on
16642 them being always the same, `__attribute__((__noinline__))' should be
16643 used to prevent inlining. If `&&foo' is used in a static variable
16644 initializer, inlining is forbidden.
16646 ---------- Footnotes ----------
16648 (1) The analogous feature in Fortran is called an assigned goto, but
16649 that name seems inappropriate in C, where one can do more than simply
16650 store label addresses in label variables.
16653 File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions
16655 5.4 Nested Functions
16656 ====================
16658 A "nested function" is a function defined inside another function.
16659 (Nested functions are not supported for GNU C++.) The nested function's
16660 name is local to the block where it is defined. For example, here we
16661 define a nested function named `square', and call it twice:
16663 foo (double a, double b)
16665 double square (double z) { return z * z; }
16667 return square (a) + square (b);
16670 The nested function can access all the variables of the containing
16671 function that are visible at the point of its definition. This is
16672 called "lexical scoping". For example, here we show a nested function
16673 which uses an inherited variable named `offset':
16675 bar (int *array, int offset, int size)
16677 int access (int *array, int index)
16678 { return array[index + offset]; }
16681 for (i = 0; i < size; i++)
16682 /* ... */ access (array, i) /* ... */
16685 Nested function definitions are permitted within functions in the
16686 places where variable definitions are allowed; that is, in any block,
16687 mixed with the other declarations and statements in the block.
16689 It is possible to call the nested function from outside the scope of
16690 its name by storing its address or passing the address to another
16693 hack (int *array, int size)
16695 void store (int index, int value)
16696 { array[index] = value; }
16698 intermediate (store, size);
16701 Here, the function `intermediate' receives the address of `store' as
16702 an argument. If `intermediate' calls `store', the arguments given to
16703 `store' are used to store into `array'. But this technique works only
16704 so long as the containing function (`hack', in this example) does not
16707 If you try to call the nested function through its address after the
16708 containing function has exited, all hell will break loose. If you try
16709 to call it after a containing scope level has exited, and if it refers
16710 to some of the variables that are no longer in scope, you may be lucky,
16711 but it's not wise to take the risk. If, however, the nested function
16712 does not refer to anything that has gone out of scope, you should be
16715 GCC implements taking the address of a nested function using a
16716 technique called "trampolines". A paper describing them is available as
16718 `http://people.debian.org/~aaronl/Usenix88-lexic.pdf'.
16720 A nested function can jump to a label inherited from a containing
16721 function, provided the label was explicitly declared in the containing
16722 function (*note Local Labels::). Such a jump returns instantly to the
16723 containing function, exiting the nested function which did the `goto'
16724 and any intermediate functions as well. Here is an example:
16726 bar (int *array, int offset, int size)
16729 int access (int *array, int index)
16733 return array[index + offset];
16737 for (i = 0; i < size; i++)
16738 /* ... */ access (array, i) /* ... */
16742 /* Control comes here from `access'
16743 if it detects an error. */
16748 A nested function always has no linkage. Declaring one with `extern'
16749 or `static' is erroneous. If you need to declare the nested function
16750 before its definition, use `auto' (which is otherwise meaningless for
16751 function declarations).
16753 bar (int *array, int offset, int size)
16756 auto int access (int *, int);
16758 int access (int *array, int index)
16762 return array[index + offset];
16768 File: gcc.info, Node: Constructing Calls, Next: Typeof, Prev: Nested Functions, Up: C Extensions
16770 5.5 Constructing Function Calls
16771 ===============================
16773 Using the built-in functions described below, you can record the
16774 arguments a function received, and call another function with the same
16775 arguments, without knowing the number or types of the arguments.
16777 You can also record the return value of that function call, and later
16778 return that value, without knowing what data type the function tried to
16779 return (as long as your caller expects that data type).
16781 However, these built-in functions may interact badly with some
16782 sophisticated features or other extensions of the language. It is,
16783 therefore, not recommended to use them outside very simple functions
16784 acting as mere forwarders for their arguments.
16786 -- Built-in Function: void * __builtin_apply_args ()
16787 This built-in function returns a pointer to data describing how to
16788 perform a call with the same arguments as were passed to the
16791 The function saves the arg pointer register, structure value
16792 address, and all registers that might be used to pass arguments to
16793 a function into a block of memory allocated on the stack. Then it
16794 returns the address of that block.
16796 -- Built-in Function: void * __builtin_apply (void (*FUNCTION)(), void
16797 *ARGUMENTS, size_t SIZE)
16798 This built-in function invokes FUNCTION with a copy of the
16799 parameters described by ARGUMENTS and SIZE.
16801 The value of ARGUMENTS should be the value returned by
16802 `__builtin_apply_args'. The argument SIZE specifies the size of
16803 the stack argument data, in bytes.
16805 This function returns a pointer to data describing how to return
16806 whatever value was returned by FUNCTION. The data is saved in a
16807 block of memory allocated on the stack.
16809 It is not always simple to compute the proper value for SIZE. The
16810 value is used by `__builtin_apply' to compute the amount of data
16811 that should be pushed on the stack and copied from the incoming
16814 -- Built-in Function: void __builtin_return (void *RESULT)
16815 This built-in function returns the value described by RESULT from
16816 the containing function. You should specify, for RESULT, a value
16817 returned by `__builtin_apply'.
16819 -- Built-in Function: __builtin_va_arg_pack ()
16820 This built-in function represents all anonymous arguments of an
16821 inline function. It can be used only in inline functions which
16822 will be always inlined, never compiled as a separate function,
16823 such as those using `__attribute__ ((__always_inline__))' or
16824 `__attribute__ ((__gnu_inline__))' extern inline functions. It
16825 must be only passed as last argument to some other function with
16826 variable arguments. This is useful for writing small wrapper
16827 inlines for variable argument functions, when using preprocessor
16828 macros is undesirable. For example:
16829 extern int myprintf (FILE *f, const char *format, ...);
16830 extern inline __attribute__ ((__gnu_inline__)) int
16831 myprintf (FILE *f, const char *format, ...)
16833 int r = fprintf (f, "myprintf: ");
16836 int s = fprintf (f, format, __builtin_va_arg_pack ());
16842 -- Built-in Function: __builtin_va_arg_pack_len ()
16843 This built-in function returns the number of anonymous arguments of
16844 an inline function. It can be used only in inline functions which
16845 will be always inlined, never compiled as a separate function, such
16846 as those using `__attribute__ ((__always_inline__))' or
16847 `__attribute__ ((__gnu_inline__))' extern inline functions. For
16848 example following will do link or runtime checking of open
16849 arguments for optimized code:
16850 #ifdef __OPTIMIZE__
16851 extern inline __attribute__((__gnu_inline__)) int
16852 myopen (const char *path, int oflag, ...)
16854 if (__builtin_va_arg_pack_len () > 1)
16855 warn_open_too_many_arguments ();
16857 if (__builtin_constant_p (oflag))
16859 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
16861 warn_open_missing_mode ();
16862 return __open_2 (path, oflag);
16864 return open (path, oflag, __builtin_va_arg_pack ());
16867 if (__builtin_va_arg_pack_len () < 1)
16868 return __open_2 (path, oflag);
16870 return open (path, oflag, __builtin_va_arg_pack ());
16875 File: gcc.info, Node: Typeof, Next: Conditionals, Prev: Constructing Calls, Up: C Extensions
16877 5.6 Referring to a Type with `typeof'
16878 =====================================
16880 Another way to refer to the type of an expression is with `typeof'.
16881 The syntax of using of this keyword looks like `sizeof', but the
16882 construct acts semantically like a type name defined with `typedef'.
16884 There are two ways of writing the argument to `typeof': with an
16885 expression or with a type. Here is an example with an expression:
16889 This assumes that `x' is an array of pointers to functions; the type
16890 described is that of the values of the functions.
16892 Here is an example with a typename as the argument:
16896 Here the type described is that of pointers to `int'.
16898 If you are writing a header file that must work when included in ISO C
16899 programs, write `__typeof__' instead of `typeof'. *Note Alternate
16902 A `typeof'-construct can be used anywhere a typedef name could be
16903 used. For example, you can use it in a declaration, in a cast, or
16904 inside of `sizeof' or `typeof'.
16906 `typeof' is often useful in conjunction with the
16907 statements-within-expressions feature. Here is how the two together can
16908 be used to define a safe "maximum" macro that operates on any
16909 arithmetic type and evaluates each of its arguments exactly once:
16912 ({ typeof (a) _a = (a); \
16913 typeof (b) _b = (b); \
16914 _a > _b ? _a : _b; })
16916 The reason for using names that start with underscores for the local
16917 variables is to avoid conflicts with variable names that occur within
16918 the expressions that are substituted for `a' and `b'. Eventually we
16919 hope to design a new form of declaration syntax that allows you to
16920 declare variables whose scopes start only after their initializers;
16921 this will be a more reliable way to prevent such conflicts.
16923 Some more examples of the use of `typeof':
16925 * This declares `y' with the type of what `x' points to.
16929 * This declares `y' as an array of such values.
16933 * This declares `y' as an array of pointers to characters:
16935 typeof (typeof (char *)[4]) y;
16937 It is equivalent to the following traditional C declaration:
16941 To see the meaning of the declaration using `typeof', and why it
16942 might be a useful way to write, rewrite it with these macros:
16944 #define pointer(T) typeof(T *)
16945 #define array(T, N) typeof(T [N])
16947 Now the declaration can be rewritten this way:
16949 array (pointer (char), 4) y;
16951 Thus, `array (pointer (char), 4)' is the type of arrays of 4
16952 pointers to `char'.
16954 _Compatibility Note:_ In addition to `typeof', GCC 2 supported a more
16955 limited extension which permitted one to write
16959 with the effect of declaring T to have the type of the expression EXPR.
16960 This extension does not work with GCC 3 (versions between 3.0 and 3.2
16961 will crash; 3.2.1 and later give an error). Code which relies on it
16962 should be rewritten to use `typeof':
16964 typedef typeof(EXPR) T;
16966 This will work with all versions of GCC.
16969 File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Typeof, Up: C Extensions
16971 5.7 Conditionals with Omitted Operands
16972 ======================================
16974 The middle operand in a conditional expression may be omitted. Then if
16975 the first operand is nonzero, its value is the value of the conditional
16978 Therefore, the expression
16982 has the value of `x' if that is nonzero; otherwise, the value of `y'.
16984 This example is perfectly equivalent to
16988 In this simple case, the ability to omit the middle operand is not
16989 especially useful. When it becomes useful is when the first operand
16990 does, or may (if it is a macro argument), contain a side effect. Then
16991 repeating the operand in the middle would perform the side effect
16992 twice. Omitting the middle operand uses the value already computed
16993 without the undesirable effects of recomputing it.
16996 File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions
16998 5.8 Double-Word Integers
16999 ========================
17001 ISO C99 supports data types for integers that are at least 64 bits wide,
17002 and as an extension GCC supports them in C89 mode and in C++. Simply
17003 write `long long int' for a signed integer, or `unsigned long long int'
17004 for an unsigned integer. To make an integer constant of type `long
17005 long int', add the suffix `LL' to the integer. To make an integer
17006 constant of type `unsigned long long int', add the suffix `ULL' to the
17009 You can use these types in arithmetic like any other integer types.
17010 Addition, subtraction, and bitwise boolean operations on these types
17011 are open-coded on all types of machines. Multiplication is open-coded
17012 if the machine supports fullword-to-doubleword a widening multiply
17013 instruction. Division and shifts are open-coded only on machines that
17014 provide special support. The operations that are not open-coded use
17015 special library routines that come with GCC.
17017 There may be pitfalls when you use `long long' types for function
17018 arguments, unless you declare function prototypes. If a function
17019 expects type `int' for its argument, and you pass a value of type `long
17020 long int', confusion will result because the caller and the subroutine
17021 will disagree about the number of bytes for the argument. Likewise, if
17022 the function expects `long long int' and you pass `int'. The best way
17023 to avoid such problems is to use prototypes.
17026 File: gcc.info, Node: Complex, Next: Floating Types, Prev: Long Long, Up: C Extensions
17028 5.9 Complex Numbers
17029 ===================
17031 ISO C99 supports complex floating data types, and as an extension GCC
17032 supports them in C89 mode and in C++, and supports complex integer data
17033 types which are not part of ISO C99. You can declare complex types
17034 using the keyword `_Complex'. As an extension, the older GNU keyword
17035 `__complex__' is also supported.
17037 For example, `_Complex double x;' declares `x' as a variable whose
17038 real part and imaginary part are both of type `double'. `_Complex
17039 short int y;' declares `y' to have real and imaginary parts of type
17040 `short int'; this is not likely to be useful, but it shows that the set
17041 of complex types is complete.
17043 To write a constant with a complex data type, use the suffix `i' or
17044 `j' (either one; they are equivalent). For example, `2.5fi' has type
17045 `_Complex float' and `3i' has type `_Complex int'. Such a constant
17046 always has a pure imaginary value, but you can form any complex value
17047 you like by adding one to a real constant. This is a GNU extension; if
17048 you have an ISO C99 conforming C library (such as GNU libc), and want
17049 to construct complex constants of floating type, you should include
17050 `<complex.h>' and use the macros `I' or `_Complex_I' instead.
17052 To extract the real part of a complex-valued expression EXP, write
17053 `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary
17054 part. This is a GNU extension; for values of floating type, you should
17055 use the ISO C99 functions `crealf', `creal', `creall', `cimagf',
17056 `cimag' and `cimagl', declared in `<complex.h>' and also provided as
17057 built-in functions by GCC.
17059 The operator `~' performs complex conjugation when used on a value
17060 with a complex type. This is a GNU extension; for values of floating
17061 type, you should use the ISO C99 functions `conjf', `conj' and `conjl',
17062 declared in `<complex.h>' and also provided as built-in functions by
17065 GCC can allocate complex automatic variables in a noncontiguous
17066 fashion; it's even possible for the real part to be in a register while
17067 the imaginary part is on the stack (or vice-versa). Only the DWARF2
17068 debug info format can represent this, so use of DWARF2 is recommended.
17069 If you are using the stabs debug info format, GCC describes a
17070 noncontiguous complex variable as if it were two separate variables of
17071 noncomplex type. If the variable's actual name is `foo', the two
17072 fictitious variables are named `foo$real' and `foo$imag'. You can
17073 examine and set these two fictitious variables with your debugger.
17076 File: gcc.info, Node: Floating Types, Next: Decimal Float, Prev: Complex, Up: C Extensions
17078 5.10 Additional Floating Types
17079 ==============================
17081 As an extension, the GNU C compiler supports additional floating types,
17082 `__float80' and `__float128' to support 80bit (`XFmode') and 128 bit
17083 (`TFmode') floating types. Support for additional types includes the
17084 arithmetic operators: add, subtract, multiply, divide; unary arithmetic
17085 operators; relational operators; equality operators; and conversions to
17086 and from integer and other floating types. Use a suffix `w' or `W' in
17087 a literal constant of type `__float80' and `q' or `Q' for `_float128'.
17088 You can declare complex types using the corresponding internal complex
17089 type, `XCmode' for `__float80' type and `TCmode' for `__float128' type:
17091 typedef _Complex float __attribute__((mode(TC))) _Complex128;
17092 typedef _Complex float __attribute__((mode(XC))) _Complex80;
17094 Not all targets support additional floating point types. `__float80'
17095 is supported on i386, x86_64 and ia64 targets and target `__float128'
17096 is supported on x86_64 and ia64 targets.
17099 File: gcc.info, Node: Decimal Float, Next: Hex Floats, Prev: Floating Types, Up: C Extensions
17101 5.11 Decimal Floating Types
17102 ===========================
17104 As an extension, the GNU C compiler supports decimal floating types as
17105 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
17106 floating types in GCC will evolve as the draft technical report changes.
17107 Calling conventions for any target might also change. Not all targets
17108 support decimal floating types.
17110 The decimal floating types are `_Decimal32', `_Decimal64', and
17111 `_Decimal128'. They use a radix of ten, unlike the floating types
17112 `float', `double', and `long double' whose radix is not specified by
17113 the C standard but is usually two.
17115 Support for decimal floating types includes the arithmetic operators
17116 add, subtract, multiply, divide; unary arithmetic operators; relational
17117 operators; equality operators; and conversions to and from integer and
17118 other floating types. Use a suffix `df' or `DF' in a literal constant
17119 of type `_Decimal32', `dd' or `DD' for `_Decimal64', and `dl' or `DL'
17122 GCC support of decimal float as specified by the draft technical report
17125 * Pragma `FLOAT_CONST_DECIMAL64' is not supported, nor is the `d'
17126 suffix for literal constants of type `double'.
17128 * When the value of a decimal floating type cannot be represented in
17129 the integer type to which it is being converted, the result is
17130 undefined rather than the result value specified by the draft
17133 * GCC does not provide the C library functionality associated with
17134 `math.h', `fenv.h', `stdio.h', `stdlib.h', and `wchar.h', which
17135 must come from a separate C library implementation. Because of
17136 this the GNU C compiler does not define macro `__STDC_DEC_FP__' to
17137 indicate that the implementation conforms to the technical report.
17139 Types `_Decimal32', `_Decimal64', and `_Decimal128' are supported by
17140 the DWARF2 debug information format.
17143 File: gcc.info, Node: Hex Floats, Next: Fixed-Point, Prev: Decimal Float, Up: C Extensions
17148 ISO C99 supports floating-point numbers written not only in the usual
17149 decimal notation, such as `1.55e1', but also numbers such as `0x1.fp3'
17150 written in hexadecimal format. As a GNU extension, GCC supports this
17151 in C89 mode (except in some cases when strictly conforming) and in C++.
17152 In that format the `0x' hex introducer and the `p' or `P' exponent
17153 field are mandatory. The exponent is a decimal number that indicates
17154 the power of 2 by which the significant part will be multiplied. Thus
17155 `0x1.f' is 1 15/16, `p3' multiplies it by 8, and the value of `0x1.fp3'
17156 is the same as `1.55e1'.
17158 Unlike for floating-point numbers in the decimal notation the exponent
17159 is always required in the hexadecimal notation. Otherwise the compiler
17160 would not be able to resolve the ambiguity of, e.g., `0x1.f'. This
17161 could mean `1.0f' or `1.9375' since `f' is also the extension for
17162 floating-point constants of type `float'.
17165 File: gcc.info, Node: Fixed-Point, Next: Zero Length, Prev: Hex Floats, Up: C Extensions
17167 5.13 Fixed-Point Types
17168 ======================
17170 As an extension, the GNU C compiler supports fixed-point types as
17171 defined in the N1169 draft of ISO/IEC DTR 18037. Support for
17172 fixed-point types in GCC will evolve as the draft technical report
17173 changes. Calling conventions for any target might also change. Not
17174 all targets support fixed-point types.
17176 The fixed-point types are `short _Fract', `_Fract', `long _Fract',
17177 `long long _Fract', `unsigned short _Fract', `unsigned _Fract',
17178 `unsigned long _Fract', `unsigned long long _Fract', `_Sat short
17179 _Fract', `_Sat _Fract', `_Sat long _Fract', `_Sat long long _Fract',
17180 `_Sat unsigned short _Fract', `_Sat unsigned _Fract', `_Sat unsigned
17181 long _Fract', `_Sat unsigned long long _Fract', `short _Accum',
17182 `_Accum', `long _Accum', `long long _Accum', `unsigned short _Accum',
17183 `unsigned _Accum', `unsigned long _Accum', `unsigned long long _Accum',
17184 `_Sat short _Accum', `_Sat _Accum', `_Sat long _Accum', `_Sat long long
17185 _Accum', `_Sat unsigned short _Accum', `_Sat unsigned _Accum', `_Sat
17186 unsigned long _Accum', `_Sat unsigned long long _Accum'.
17188 Fixed-point data values contain fractional and optional integral parts.
17189 The format of fixed-point data varies and depends on the target machine.
17191 Support for fixed-point types includes:
17192 * prefix and postfix increment and decrement operators (`++', `--')
17194 * unary arithmetic operators (`+', `-', `!')
17196 * binary arithmetic operators (`+', `-', `*', `/')
17198 * binary shift operators (`<<', `>>')
17200 * relational operators (`<', `<=', `>=', `>')
17202 * equality operators (`==', `!=')
17204 * assignment operators (`+=', `-=', `*=', `/=', `<<=', `>>=')
17206 * conversions to and from integer, floating-point, or fixed-point
17209 Use a suffix in a fixed-point literal constant:
17210 * `hr' or `HR' for `short _Fract' and `_Sat short _Fract'
17212 * `r' or `R' for `_Fract' and `_Sat _Fract'
17214 * `lr' or `LR' for `long _Fract' and `_Sat long _Fract'
17216 * `llr' or `LLR' for `long long _Fract' and `_Sat long long _Fract'
17218 * `uhr' or `UHR' for `unsigned short _Fract' and `_Sat unsigned
17221 * `ur' or `UR' for `unsigned _Fract' and `_Sat unsigned _Fract'
17223 * `ulr' or `ULR' for `unsigned long _Fract' and `_Sat unsigned long
17226 * `ullr' or `ULLR' for `unsigned long long _Fract' and `_Sat
17227 unsigned long long _Fract'
17229 * `hk' or `HK' for `short _Accum' and `_Sat short _Accum'
17231 * `k' or `K' for `_Accum' and `_Sat _Accum'
17233 * `lk' or `LK' for `long _Accum' and `_Sat long _Accum'
17235 * `llk' or `LLK' for `long long _Accum' and `_Sat long long _Accum'
17237 * `uhk' or `UHK' for `unsigned short _Accum' and `_Sat unsigned
17240 * `uk' or `UK' for `unsigned _Accum' and `_Sat unsigned _Accum'
17242 * `ulk' or `ULK' for `unsigned long _Accum' and `_Sat unsigned long
17245 * `ullk' or `ULLK' for `unsigned long long _Accum' and `_Sat
17246 unsigned long long _Accum'
17248 GCC support of fixed-point types as specified by the draft technical
17249 report is incomplete:
17251 * Pragmas to control overflow and rounding behaviors are not
17254 Fixed-point types are supported by the DWARF2 debug information format.
17257 File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Fixed-Point, Up: C Extensions
17259 5.14 Arrays of Length Zero
17260 ==========================
17262 Zero-length arrays are allowed in GNU C. They are very useful as the
17263 last element of a structure which is really a header for a
17264 variable-length object:
17271 struct line *thisline = (struct line *)
17272 malloc (sizeof (struct line) + this_length);
17273 thisline->length = this_length;
17275 In ISO C90, you would have to give `contents' a length of 1, which
17276 means either you waste space or complicate the argument to `malloc'.
17278 In ISO C99, you would use a "flexible array member", which is slightly
17279 different in syntax and semantics:
17281 * Flexible array members are written as `contents[]' without the `0'.
17283 * Flexible array members have incomplete type, and so the `sizeof'
17284 operator may not be applied. As a quirk of the original
17285 implementation of zero-length arrays, `sizeof' evaluates to zero.
17287 * Flexible array members may only appear as the last member of a
17288 `struct' that is otherwise non-empty.
17290 * A structure containing a flexible array member, or a union
17291 containing such a structure (possibly recursively), may not be a
17292 member of a structure or an element of an array. (However, these
17293 uses are permitted by GCC as extensions.)
17295 GCC versions before 3.0 allowed zero-length arrays to be statically
17296 initialized, as if they were flexible arrays. In addition to those
17297 cases that were useful, it also allowed initializations in situations
17298 that would corrupt later data. Non-empty initialization of zero-length
17299 arrays is now treated like any case where there are more initializer
17300 elements than the array holds, in that a suitable warning about "excess
17301 elements in array" is given, and the excess elements (all of them, in
17302 this case) are ignored.
17304 Instead GCC allows static initialization of flexible array members.
17305 This is equivalent to defining a new structure containing the original
17306 structure followed by an array of sufficient size to contain the data.
17307 I.e. in the following, `f1' is constructed as if it were declared like
17312 } f1 = { 1, { 2, 3, 4 } };
17315 struct f1 f1; int data[3];
17316 } f2 = { { 1 }, { 2, 3, 4 } };
17318 The convenience of this extension is that `f1' has the desired type,
17319 eliminating the need to consistently refer to `f2.f1'.
17321 This has symmetry with normal static arrays, in that an array of
17322 unknown size is also written with `[]'.
17324 Of course, this extension only makes sense if the extra data comes at
17325 the end of a top-level object, as otherwise we would be overwriting
17326 data at subsequent offsets. To avoid undue complication and confusion
17327 with initialization of deeply nested arrays, we simply disallow any
17328 non-empty initialization except when the structure is the top-level
17329 object. For example:
17331 struct foo { int x; int y[]; };
17332 struct bar { struct foo z; };
17334 struct foo a = { 1, { 2, 3, 4 } }; // Valid.
17335 struct bar b = { { 1, { 2, 3, 4 } } }; // Invalid.
17336 struct bar c = { { 1, { } } }; // Valid.
17337 struct foo d[1] = { { 1 { 2, 3, 4 } } }; // Invalid.
17340 File: gcc.info, Node: Empty Structures, Next: Variadic Macros, Prev: Variable Length, Up: C Extensions
17342 5.15 Structures With No Members
17343 ===============================
17345 GCC permits a C structure to have no members:
17350 The structure will have size zero. In C++, empty structures are part
17351 of the language. G++ treats empty structures as if they had a single
17352 member of type `char'.
17355 File: gcc.info, Node: Variable Length, Next: Empty Structures, Prev: Zero Length, Up: C Extensions
17357 5.16 Arrays of Variable Length
17358 ==============================
17360 Variable-length automatic arrays are allowed in ISO C99, and as an
17361 extension GCC accepts them in C89 mode and in C++. (However, GCC's
17362 implementation of variable-length arrays does not yet conform in detail
17363 to the ISO C99 standard.) These arrays are declared like any other
17364 automatic arrays, but with a length that is not a constant expression.
17365 The storage is allocated at the point of declaration and deallocated
17366 when the brace-level is exited. For example:
17369 concat_fopen (char *s1, char *s2, char *mode)
17371 char str[strlen (s1) + strlen (s2) + 1];
17374 return fopen (str, mode);
17377 Jumping or breaking out of the scope of the array name deallocates the
17378 storage. Jumping into the scope is not allowed; you get an error
17381 You can use the function `alloca' to get an effect much like
17382 variable-length arrays. The function `alloca' is available in many
17383 other C implementations (but not in all). On the other hand,
17384 variable-length arrays are more elegant.
17386 There are other differences between these two methods. Space allocated
17387 with `alloca' exists until the containing _function_ returns. The
17388 space for a variable-length array is deallocated as soon as the array
17389 name's scope ends. (If you use both variable-length arrays and
17390 `alloca' in the same function, deallocation of a variable-length array
17391 will also deallocate anything more recently allocated with `alloca'.)
17393 You can also use variable-length arrays as arguments to functions:
17396 tester (int len, char data[len][len])
17401 The length of an array is computed once when the storage is allocated
17402 and is remembered for the scope of the array in case you access it with
17405 If you want to pass the array first and the length afterward, you can
17406 use a forward declaration in the parameter list--another GNU extension.
17409 tester (int len; char data[len][len], int len)
17414 The `int len' before the semicolon is a "parameter forward
17415 declaration", and it serves the purpose of making the name `len' known
17416 when the declaration of `data' is parsed.
17418 You can write any number of such parameter forward declarations in the
17419 parameter list. They can be separated by commas or semicolons, but the
17420 last one must end with a semicolon, which is followed by the "real"
17421 parameter declarations. Each forward declaration must match a "real"
17422 declaration in parameter name and data type. ISO C99 does not support
17423 parameter forward declarations.
17426 File: gcc.info, Node: Variadic Macros, Next: Escaped Newlines, Prev: Empty Structures, Up: C Extensions
17428 5.17 Macros with a Variable Number of Arguments.
17429 ================================================
17431 In the ISO C standard of 1999, a macro can be declared to accept a
17432 variable number of arguments much as a function can. The syntax for
17433 defining the macro is similar to that of a function. Here is an
17436 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
17438 Here `...' is a "variable argument". In the invocation of such a
17439 macro, it represents the zero or more tokens until the closing
17440 parenthesis that ends the invocation, including any commas. This set of
17441 tokens replaces the identifier `__VA_ARGS__' in the macro body wherever
17442 it appears. See the CPP manual for more information.
17444 GCC has long supported variadic macros, and used a different syntax
17445 that allowed you to give a name to the variable arguments just like any
17446 other argument. Here is an example:
17448 #define debug(format, args...) fprintf (stderr, format, args)
17450 This is in all ways equivalent to the ISO C example above, but arguably
17451 more readable and descriptive.
17453 GNU CPP has two further variadic macro extensions, and permits them to
17454 be used with either of the above forms of macro definition.
17456 In standard C, you are not allowed to leave the variable argument out
17457 entirely; but you are allowed to pass an empty argument. For example,
17458 this invocation is invalid in ISO C, because there is no comma after
17461 debug ("A message")
17463 GNU CPP permits you to completely omit the variable arguments in this
17464 way. In the above examples, the compiler would complain, though since
17465 the expansion of the macro still has the extra comma after the format
17468 To help solve this problem, CPP behaves specially for variable
17469 arguments used with the token paste operator, `##'. If instead you
17472 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
17474 and if the variable arguments are omitted or empty, the `##' operator
17475 causes the preprocessor to remove the comma before it. If you do
17476 provide some variable arguments in your macro invocation, GNU CPP does
17477 not complain about the paste operation and instead places the variable
17478 arguments after the comma. Just like any other pasted macro argument,
17479 these arguments are not macro expanded.
17482 File: gcc.info, Node: Escaped Newlines, Next: Subscripting, Prev: Variadic Macros, Up: C Extensions
17484 5.18 Slightly Looser Rules for Escaped Newlines
17485 ===============================================
17487 Recently, the preprocessor has relaxed its treatment of escaped
17488 newlines. Previously, the newline had to immediately follow a
17489 backslash. The current implementation allows whitespace in the form of
17490 spaces, horizontal and vertical tabs, and form feeds between the
17491 backslash and the subsequent newline. The preprocessor issues a
17492 warning, but treats it as a valid escaped newline and combines the two
17493 lines to form a single logical line. This works within comments and
17494 tokens, as well as between tokens. Comments are _not_ treated as
17495 whitespace for the purposes of this relaxation, since they have not yet
17496 been replaced with spaces.
17499 File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Escaped Newlines, Up: C Extensions
17501 5.19 Non-Lvalue Arrays May Have Subscripts
17502 ==========================================
17504 In ISO C99, arrays that are not lvalues still decay to pointers, and
17505 may be subscripted, although they may not be modified or used after the
17506 next sequence point and the unary `&' operator may not be applied to
17507 them. As an extension, GCC allows such arrays to be subscripted in C89
17508 mode, though otherwise they do not decay to pointers outside C99 mode.
17509 For example, this is valid in GNU C though not valid in C89:
17511 struct foo {int a[4];};
17517 return f().a[index];
17521 File: gcc.info, Node: Pointer Arith, Next: Initializers, Prev: Subscripting, Up: C Extensions
17523 5.20 Arithmetic on `void'- and Function-Pointers
17524 ================================================
17526 In GNU C, addition and subtraction operations are supported on pointers
17527 to `void' and on pointers to functions. This is done by treating the
17528 size of a `void' or of a function as 1.
17530 A consequence of this is that `sizeof' is also allowed on `void' and
17531 on function types, and returns 1.
17533 The option `-Wpointer-arith' requests a warning if these extensions
17537 File: gcc.info, Node: Initializers, Next: Compound Literals, Prev: Pointer Arith, Up: C Extensions
17539 5.21 Non-Constant Initializers
17540 ==============================
17542 As in standard C++ and ISO C99, the elements of an aggregate
17543 initializer for an automatic variable are not required to be constant
17544 expressions in GNU C. Here is an example of an initializer with
17545 run-time varying elements:
17547 foo (float f, float g)
17549 float beat_freqs[2] = { f-g, f+g };
17554 File: gcc.info, Node: Compound Literals, Next: Designated Inits, Prev: Initializers, Up: C Extensions
17556 5.22 Compound Literals
17557 ======================
17559 ISO C99 supports compound literals. A compound literal looks like a
17560 cast containing an initializer. Its value is an object of the type
17561 specified in the cast, containing the elements specified in the
17562 initializer; it is an lvalue. As an extension, GCC supports compound
17563 literals in C89 mode and in C++.
17565 Usually, the specified type is a structure. Assume that `struct foo'
17566 and `structure' are declared as shown:
17568 struct foo {int a; char b[2];} structure;
17570 Here is an example of constructing a `struct foo' with a compound
17573 structure = ((struct foo) {x + y, 'a', 0});
17575 This is equivalent to writing the following:
17578 struct foo temp = {x + y, 'a', 0};
17582 You can also construct an array. If all the elements of the compound
17583 literal are (made up of) simple constant expressions, suitable for use
17584 in initializers of objects of static storage duration, then the compound
17585 literal can be coerced to a pointer to its first element and used in
17586 such an initializer, as shown here:
17588 char **foo = (char *[]) { "x", "y", "z" };
17590 Compound literals for scalar types and union types are is also
17591 allowed, but then the compound literal is equivalent to a cast.
17593 As a GNU extension, GCC allows initialization of objects with static
17594 storage duration by compound literals (which is not possible in ISO
17595 C99, because the initializer is not a constant). It is handled as if
17596 the object was initialized only with the bracket enclosed list if the
17597 types of the compound literal and the object match. The initializer
17598 list of the compound literal must be constant. If the object being
17599 initialized has array type of unknown size, the size is determined by
17600 compound literal size.
17602 static struct foo x = (struct foo) {1, 'a', 'b'};
17603 static int y[] = (int []) {1, 2, 3};
17604 static int z[] = (int [3]) {1};
17606 The above lines are equivalent to the following:
17607 static struct foo x = {1, 'a', 'b'};
17608 static int y[] = {1, 2, 3};
17609 static int z[] = {1, 0, 0};
17612 File: gcc.info, Node: Designated Inits, Next: Cast to Union, Prev: Compound Literals, Up: C Extensions
17614 5.23 Designated Initializers
17615 ============================
17617 Standard C89 requires the elements of an initializer to appear in a
17618 fixed order, the same as the order of the elements in the array or
17619 structure being initialized.
17621 In ISO C99 you can give the elements in any order, specifying the array
17622 indices or structure field names they apply to, and GNU C allows this as
17623 an extension in C89 mode as well. This extension is not implemented in
17626 To specify an array index, write `[INDEX] =' before the element value.
17629 int a[6] = { [4] = 29, [2] = 15 };
17633 int a[6] = { 0, 0, 15, 0, 29, 0 };
17635 The index values must be constant expressions, even if the array being
17636 initialized is automatic.
17638 An alternative syntax for this which has been obsolete since GCC 2.5
17639 but GCC still accepts is to write `[INDEX]' before the element value,
17642 To initialize a range of elements to the same value, write `[FIRST ...
17643 LAST] = VALUE'. This is a GNU extension. For example,
17645 int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
17647 If the value in it has side-effects, the side-effects will happen only
17648 once, not for each initialized field by the range initializer.
17650 Note that the length of the array is the highest value specified plus
17653 In a structure initializer, specify the name of a field to initialize
17654 with `.FIELDNAME =' before the element value. For example, given the
17655 following structure,
17657 struct point { int x, y; };
17659 the following initialization
17661 struct point p = { .y = yvalue, .x = xvalue };
17665 struct point p = { xvalue, yvalue };
17667 Another syntax which has the same meaning, obsolete since GCC 2.5, is
17668 `FIELDNAME:', as shown here:
17670 struct point p = { y: yvalue, x: xvalue };
17672 The `[INDEX]' or `.FIELDNAME' is known as a "designator". You can
17673 also use a designator (or the obsolete colon syntax) when initializing
17674 a union, to specify which element of the union should be used. For
17677 union foo { int i; double d; };
17679 union foo f = { .d = 4 };
17681 will convert 4 to a `double' to store it in the union using the second
17682 element. By contrast, casting 4 to type `union foo' would store it
17683 into the union as the integer `i', since it is an integer. (*Note Cast
17686 You can combine this technique of naming elements with ordinary C
17687 initialization of successive elements. Each initializer element that
17688 does not have a designator applies to the next consecutive element of
17689 the array or structure. For example,
17691 int a[6] = { [1] = v1, v2, [4] = v4 };
17695 int a[6] = { 0, v1, v2, 0, v4, 0 };
17697 Labeling the elements of an array initializer is especially useful
17698 when the indices are characters or belong to an `enum' type. For
17701 int whitespace[256]
17702 = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
17703 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
17705 You can also write a series of `.FIELDNAME' and `[INDEX]' designators
17706 before an `=' to specify a nested subobject to initialize; the list is
17707 taken relative to the subobject corresponding to the closest
17708 surrounding brace pair. For example, with the `struct point'
17711 struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
17713 If the same field is initialized multiple times, it will have value from
17714 the last initialization. If any such overridden initialization has
17715 side-effect, it is unspecified whether the side-effect happens or not.
17716 Currently, GCC will discard them and issue a warning.
17719 File: gcc.info, Node: Case Ranges, Next: Mixed Declarations, Prev: Cast to Union, Up: C Extensions
17724 You can specify a range of consecutive values in a single `case' label,
17729 This has the same effect as the proper number of individual `case'
17730 labels, one for each integer value from LOW to HIGH, inclusive.
17732 This feature is especially useful for ranges of ASCII character codes:
17736 *Be careful:* Write spaces around the `...', for otherwise it may be
17737 parsed wrong when you use it with integer values. For example, write
17747 File: gcc.info, Node: Cast to Union, Next: Case Ranges, Prev: Designated Inits, Up: C Extensions
17749 5.25 Cast to a Union Type
17750 =========================
17752 A cast to union type is similar to other casts, except that the type
17753 specified is a union type. You can specify the type either with `union
17754 TAG' or with a typedef name. A cast to union is actually a constructor
17755 though, not a cast, and hence does not yield an lvalue like normal
17756 casts. (*Note Compound Literals::.)
17758 The types that may be cast to the union type are those of the members
17759 of the union. Thus, given the following union and variables:
17761 union foo { int i; double d; };
17765 both `x' and `y' can be cast to type `union foo'.
17767 Using the cast as the right-hand side of an assignment to a variable of
17768 union type is equivalent to storing in a member of the union:
17772 u = (union foo) x == u.i = x
17773 u = (union foo) y == u.d = y
17775 You can also use the union cast as a function argument:
17777 void hack (union foo);
17779 hack ((union foo) x);
17782 File: gcc.info, Node: Mixed Declarations, Next: Function Attributes, Prev: Case Ranges, Up: C Extensions
17784 5.26 Mixed Declarations and Code
17785 ================================
17787 ISO C99 and ISO C++ allow declarations and code to be freely mixed
17788 within compound statements. As an extension, GCC also allows this in
17789 C89 mode. For example, you could do:
17796 Each identifier is visible from where it is declared until the end of
17797 the enclosing block.
17800 File: gcc.info, Node: Function Attributes, Next: Attribute Syntax, Prev: Mixed Declarations, Up: C Extensions
17802 5.27 Declaring Attributes of Functions
17803 ======================================
17805 In GNU C, you declare certain things about functions called in your
17806 program which help the compiler optimize function calls and check your
17807 code more carefully.
17809 The keyword `__attribute__' allows you to specify special attributes
17810 when making a declaration. This keyword is followed by an attribute
17811 specification inside double parentheses. The following attributes are
17812 currently defined for functions on all targets: `aligned',
17813 `alloc_size', `noreturn', `returns_twice', `noinline', `always_inline',
17814 `flatten', `pure', `const', `nothrow', `sentinel', `format',
17815 `format_arg', `no_instrument_function', `section', `constructor',
17816 `destructor', `used', `unused', `deprecated', `weak', `malloc',
17817 `alias', `warn_unused_result', `nonnull', `gnu_inline',
17818 `externally_visible', `hot', `cold', `artificial', `error' and
17819 `warning'. Several other attributes are defined for functions on
17820 particular target systems. Other attributes, including `section' are
17821 supported for variables declarations (*note Variable Attributes::) and
17822 for types (*note Type Attributes::).
17824 You may also specify attributes with `__' preceding and following each
17825 keyword. This allows you to use them in header files without being
17826 concerned about a possible macro of the same name. For example, you
17827 may use `__noreturn__' instead of `noreturn'.
17829 *Note Attribute Syntax::, for details of the exact syntax for using
17833 The `alias' attribute causes the declaration to be emitted as an
17834 alias for another symbol, which must be specified. For instance,
17836 void __f () { /* Do something. */; }
17837 void f () __attribute__ ((weak, alias ("__f")));
17839 defines `f' to be a weak alias for `__f'. In C++, the mangled
17840 name for the target must be used. It is an error if `__f' is not
17841 defined in the same translation unit.
17843 Not all target machines support this attribute.
17845 `aligned (ALIGNMENT)'
17846 This attribute specifies a minimum alignment for the function,
17849 You cannot use this attribute to decrease the alignment of a
17850 function, only to increase it. However, when you explicitly
17851 specify a function alignment this will override the effect of the
17852 `-falign-functions' (*note Optimize Options::) option for this
17855 Note that the effectiveness of `aligned' attributes may be limited
17856 by inherent limitations in your linker. On many systems, the
17857 linker is only able to arrange for functions to be aligned up to a
17858 certain maximum alignment. (For some linkers, the maximum
17859 supported alignment may be very very small.) See your linker
17860 documentation for further information.
17862 The `aligned' attribute can also be used for variables and fields
17863 (*note Variable Attributes::.)
17866 The `alloc_size' attribute is used to tell the compiler that the
17867 function return value points to memory, where the size is given by
17868 one or two of the functions parameters. GCC uses this information
17869 to improve the correctness of `__builtin_object_size'.
17871 The function parameter(s) denoting the allocated size are
17872 specified by one or two integer arguments supplied to the
17873 attribute. The allocated size is either the value of the single
17874 function argument specified or the product of the two function
17875 arguments specified. Argument numbering starts at one.
17879 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
17880 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
17882 declares that my_calloc will return memory of the size given by
17883 the product of parameter 1 and 2 and that my_realloc will return
17884 memory of the size given by parameter 2.
17887 Generally, functions are not inlined unless optimization is
17888 specified. For functions declared inline, this attribute inlines
17889 the function even if no optimization level was specified.
17892 This attribute should be used with a function which is also
17893 declared with the `inline' keyword. It directs GCC to treat the
17894 function as if it were defined in gnu89 mode even when compiling
17895 in C99 or gnu99 mode.
17897 If the function is declared `extern', then this definition of the
17898 function is used only for inlining. In no case is the function
17899 compiled as a standalone function, not even if you take its address
17900 explicitly. Such an address becomes an external reference, as if
17901 you had only declared the function, and had not defined it. This
17902 has almost the effect of a macro. The way to use this is to put a
17903 function definition in a header file with this attribute, and put
17904 another copy of the function, without `extern', in a library file.
17905 The definition in the header file will cause most calls to the
17906 function to be inlined. If any uses of the function remain, they
17907 will refer to the single copy in the library. Note that the two
17908 definitions of the functions need not be precisely the same,
17909 although if they do not have the same effect your program may
17912 In C, if the function is neither `extern' nor `static', then the
17913 function is compiled as a standalone function, as well as being
17914 inlined where possible.
17916 This is how GCC traditionally handled functions declared `inline'.
17917 Since ISO C99 specifies a different semantics for `inline', this
17918 function attribute is provided as a transition measure and as a
17919 useful feature in its own right. This attribute is available in
17920 GCC 4.1.3 and later. It is available if either of the
17921 preprocessor macros `__GNUC_GNU_INLINE__' or
17922 `__GNUC_STDC_INLINE__' are defined. *Note An Inline Function is
17923 As Fast As a Macro: Inline.
17925 In C++, this attribute does not depend on `extern' in any way, but
17926 it still requires the `inline' keyword to enable its special
17930 This attribute is useful for small inline wrappers which if
17931 possible should appear during debugging as a unit, depending on
17932 the debug info format it will either mean marking the function as
17933 artificial or using the caller location for all instructions
17934 within the inlined body.
17937 Generally, inlining into a function is limited. For a function
17938 marked with this attribute, every call inside this function will
17939 be inlined, if possible. Whether the function itself is
17940 considered for inlining depends on its size and the current
17941 inlining parameters.
17943 `error ("MESSAGE")'
17944 If this attribute is used on a function declaration and a call to
17945 such a function is not eliminated through dead code elimination or
17946 other optimizations, an error which will include MESSAGE will be
17947 diagnosed. This is useful for compile time checking, especially
17948 together with `__builtin_constant_p' and inline functions where
17949 checking the inline function arguments is not possible through
17950 `extern char [(condition) ? 1 : -1];' tricks. While it is
17951 possible to leave the function undefined and thus invoke a link
17952 failure, when using this attribute the problem will be diagnosed
17953 earlier and with exact location of the call even in presence of
17954 inline functions or when not emitting debugging information.
17956 `warning ("MESSAGE")'
17957 If this attribute is used on a function declaration and a call to
17958 such a function is not eliminated through dead code elimination or
17959 other optimizations, a warning which will include MESSAGE will be
17960 diagnosed. This is useful for compile time checking, especially
17961 together with `__builtin_constant_p' and inline functions. While
17962 it is possible to define the function with a message in
17963 `.gnu.warning*' section, when using this attribute the problem
17964 will be diagnosed earlier and with exact location of the call even
17965 in presence of inline functions or when not emitting debugging
17969 On the Intel 386, the `cdecl' attribute causes the compiler to
17970 assume that the calling function will pop off the stack space used
17971 to pass arguments. This is useful to override the effects of the
17975 Many functions do not examine any values except their arguments,
17976 and have no effects except the return value. Basically this is
17977 just slightly more strict class than the `pure' attribute below,
17978 since function is not allowed to read global memory.
17980 Note that a function that has pointer arguments and examines the
17981 data pointed to must _not_ be declared `const'. Likewise, a
17982 function that calls a non-`const' function usually must not be
17983 `const'. It does not make sense for a `const' function to return
17986 The attribute `const' is not implemented in GCC versions earlier
17987 than 2.5. An alternative way to declare that a function has no
17988 side effects, which works in the current version and in some older
17989 versions, is as follows:
17991 typedef int intfn ();
17993 extern const intfn square;
17995 This approach does not work in GNU C++ from 2.6.0 on, since the
17996 language specifies that the `const' must be attached to the return
18001 `constructor (PRIORITY)'
18002 `destructor (PRIORITY)'
18003 The `constructor' attribute causes the function to be called
18004 automatically before execution enters `main ()'. Similarly, the
18005 `destructor' attribute causes the function to be called
18006 automatically after `main ()' has completed or `exit ()' has been
18007 called. Functions with these attributes are useful for
18008 initializing data that will be used implicitly during the
18009 execution of the program.
18011 You may provide an optional integer priority to control the order
18012 in which constructor and destructor functions are run. A
18013 constructor with a smaller priority number runs before a
18014 constructor with a larger priority number; the opposite
18015 relationship holds for destructors. So, if you have a constructor
18016 that allocates a resource and a destructor that deallocates the
18017 same resource, both functions typically have the same priority.
18018 The priorities for constructor and destructor functions are the
18019 same as those specified for namespace-scope C++ objects (*note C++
18022 These attributes are not currently implemented for Objective-C.
18025 The `deprecated' attribute results in a warning if the function is
18026 used anywhere in the source file. This is useful when identifying
18027 functions that are expected to be removed in a future version of a
18028 program. The warning also includes the location of the declaration
18029 of the deprecated function, to enable users to easily find further
18030 information about why the function is deprecated, or what they
18031 should do instead. Note that the warnings only occurs for uses:
18033 int old_fn () __attribute__ ((deprecated));
18035 int (*fn_ptr)() = old_fn;
18037 results in a warning on line 3 but not line 2.
18039 The `deprecated' attribute can also be used for variables and
18040 types (*note Variable Attributes::, *note Type Attributes::.)
18043 On Microsoft Windows targets and Symbian OS targets the
18044 `dllexport' attribute causes the compiler to provide a global
18045 pointer to a pointer in a DLL, so that it can be referenced with
18046 the `dllimport' attribute. On Microsoft Windows targets, the
18047 pointer name is formed by combining `_imp__' and the function or
18050 You can use `__declspec(dllexport)' as a synonym for
18051 `__attribute__ ((dllexport))' for compatibility with other
18054 On systems that support the `visibility' attribute, this attribute
18055 also implies "default" visibility. It is an error to explicitly
18056 specify any other visibility.
18058 Currently, the `dllexport' attribute is ignored for inlined
18059 functions, unless the `-fkeep-inline-functions' flag has been
18060 used. The attribute is also ignored for undefined symbols.
18062 When applied to C++ classes, the attribute marks defined
18063 non-inlined member functions and static data members as exports.
18064 Static consts initialized in-class are not marked unless they are
18065 also defined out-of-class.
18067 For Microsoft Windows targets there are alternative methods for
18068 including the symbol in the DLL's export table such as using a
18069 `.def' file with an `EXPORTS' section or, with GNU ld, using the
18070 `--export-all' linker flag.
18073 On Microsoft Windows and Symbian OS targets, the `dllimport'
18074 attribute causes the compiler to reference a function or variable
18075 via a global pointer to a pointer that is set up by the DLL
18076 exporting the symbol. The attribute implies `extern'. On
18077 Microsoft Windows targets, the pointer name is formed by combining
18078 `_imp__' and the function or variable name.
18080 You can use `__declspec(dllimport)' as a synonym for
18081 `__attribute__ ((dllimport))' for compatibility with other
18084 On systems that support the `visibility' attribute, this attribute
18085 also implies "default" visibility. It is an error to explicitly
18086 specify any other visibility.
18088 Currently, the attribute is ignored for inlined functions. If the
18089 attribute is applied to a symbol _definition_, an error is
18090 reported. If a symbol previously declared `dllimport' is later
18091 defined, the attribute is ignored in subsequent references, and a
18092 warning is emitted. The attribute is also overridden by a
18093 subsequent declaration as `dllexport'.
18095 When applied to C++ classes, the attribute marks non-inlined
18096 member functions and static data members as imports. However, the
18097 attribute is ignored for virtual methods to allow creation of
18098 vtables using thunks.
18100 On the SH Symbian OS target the `dllimport' attribute also has
18101 another affect--it can cause the vtable and run-time type
18102 information for a class to be exported. This happens when the
18103 class has a dllimport'ed constructor or a non-inline, non-pure
18104 virtual function and, for either of those two conditions, the
18105 class also has a inline constructor or destructor and has a key
18106 function that is defined in the current translation unit.
18108 For Microsoft Windows based targets the use of the `dllimport'
18109 attribute on functions is not necessary, but provides a small
18110 performance benefit by eliminating a thunk in the DLL. The use of
18111 the `dllimport' attribute on imported variables was required on
18112 older versions of the GNU linker, but can now be avoided by
18113 passing the `--enable-auto-import' switch to the GNU linker. As
18114 with functions, using the attribute for a variable eliminates a
18117 One drawback to using this attribute is that a pointer to a
18118 _variable_ marked as `dllimport' cannot be used as a constant
18119 address. However, a pointer to a _function_ with the `dllimport'
18120 attribute can be used as a constant initializer; in this case, the
18121 address of a stub function in the import lib is referenced. On
18122 Microsoft Windows targets, the attribute can be disabled for
18123 functions by setting the `-mnop-fun-dllimport' flag.
18126 Use this attribute on the H8/300, H8/300H, and H8S to indicate
18127 that the specified variable should be placed into the eight bit
18128 data section. The compiler will generate more efficient code for
18129 certain operations on data in the eight bit data area. Note the
18130 eight bit data area is limited to 256 bytes of data.
18132 You must use GAS and GLD from GNU binutils version 2.7 or later for
18133 this attribute to work correctly.
18135 `exception_handler'
18136 Use this attribute on the Blackfin to indicate that the specified
18137 function is an exception handler. The compiler will generate
18138 function entry and exit sequences suitable for use in an exception
18139 handler when this attribute is present.
18141 `externally_visible'
18142 This attribute, attached to a global variable or function,
18143 nullifies the effect of the `-fwhole-program' command-line option,
18144 so the object remains visible outside the current compilation unit.
18147 On 68HC11 and 68HC12 the `far' attribute causes the compiler to
18148 use a calling convention that takes care of switching memory banks
18149 when entering and leaving a function. This calling convention is
18150 also the default when using the `-mlong-calls' option.
18152 On 68HC12 the compiler will use the `call' and `rtc' instructions
18153 to call and return from a function.
18155 On 68HC11 the compiler will generate a sequence of instructions to
18156 invoke a board-specific routine to switch the memory bank and call
18157 the real function. The board-specific routine simulates a `call'.
18158 At the end of a function, it will jump to a board-specific routine
18159 instead of using `rts'. The board-specific return routine
18160 simulates the `rtc'.
18163 On the Intel 386, the `fastcall' attribute causes the compiler to
18164 pass the first argument (if of integral type) in the register ECX
18165 and the second argument (if of integral type) in the register EDX.
18166 Subsequent and other typed arguments are passed on the stack. The
18167 called function will pop the arguments off the stack. If the
18168 number of arguments is variable all arguments are pushed on the
18171 `format (ARCHETYPE, STRING-INDEX, FIRST-TO-CHECK)'
18172 The `format' attribute specifies that a function takes `printf',
18173 `scanf', `strftime' or `strfmon' style arguments which should be
18174 type-checked against a format string. For example, the
18178 my_printf (void *my_object, const char *my_format, ...)
18179 __attribute__ ((format (printf, 2, 3)));
18181 causes the compiler to check the arguments in calls to `my_printf'
18182 for consistency with the `printf' style format string argument
18185 The parameter ARCHETYPE determines how the format string is
18186 interpreted, and should be `printf', `scanf', `strftime',
18187 `gnu_printf', `gnu_scanf', `gnu_strftime' or `strfmon'. (You can
18188 also use `__printf__', `__scanf__', `__strftime__' or
18189 `__strfmon__'.) On MinGW targets, `ms_printf', `ms_scanf', and
18190 `ms_strftime' are also present. ARCHTYPE values such as `printf'
18191 refer to the formats accepted by the system's C run-time library,
18192 while `gnu_' values always refer to the formats accepted by the
18193 GNU C Library. On Microsoft Windows targets, `ms_' values refer
18194 to the formats accepted by the `msvcrt.dll' library. The
18195 parameter STRING-INDEX specifies which argument is the format
18196 string argument (starting from 1), while FIRST-TO-CHECK is the
18197 number of the first argument to check against the format string.
18198 For functions where the arguments are not available to be checked
18199 (such as `vprintf'), specify the third parameter as zero. In this
18200 case the compiler only checks the format string for consistency.
18201 For `strftime' formats, the third parameter is required to be zero.
18202 Since non-static C++ methods have an implicit `this' argument, the
18203 arguments of such methods should be counted from two, not one, when
18204 giving values for STRING-INDEX and FIRST-TO-CHECK.
18206 In the example above, the format string (`my_format') is the second
18207 argument of the function `my_print', and the arguments to check
18208 start with the third argument, so the correct parameters for the
18209 format attribute are 2 and 3.
18211 The `format' attribute allows you to identify your own functions
18212 which take format strings as arguments, so that GCC can check the
18213 calls to these functions for errors. The compiler always (unless
18214 `-ffreestanding' or `-fno-builtin' is used) checks formats for the
18215 standard library functions `printf', `fprintf', `sprintf',
18216 `scanf', `fscanf', `sscanf', `strftime', `vprintf', `vfprintf' and
18217 `vsprintf' whenever such warnings are requested (using
18218 `-Wformat'), so there is no need to modify the header file
18219 `stdio.h'. In C99 mode, the functions `snprintf', `vsnprintf',
18220 `vscanf', `vfscanf' and `vsscanf' are also checked. Except in
18221 strictly conforming C standard modes, the X/Open function
18222 `strfmon' is also checked as are `printf_unlocked' and
18223 `fprintf_unlocked'. *Note Options Controlling C Dialect: C
18226 The target may provide additional types of format checks. *Note
18227 Format Checks Specific to Particular Target Machines: Target
18230 `format_arg (STRING-INDEX)'
18231 The `format_arg' attribute specifies that a function takes a format
18232 string for a `printf', `scanf', `strftime' or `strfmon' style
18233 function and modifies it (for example, to translate it into
18234 another language), so the result can be passed to a `printf',
18235 `scanf', `strftime' or `strfmon' style function (with the
18236 remaining arguments to the format function the same as they would
18237 have been for the unmodified string). For example, the
18241 my_dgettext (char *my_domain, const char *my_format)
18242 __attribute__ ((format_arg (2)));
18244 causes the compiler to check the arguments in calls to a `printf',
18245 `scanf', `strftime' or `strfmon' type function, whose format
18246 string argument is a call to the `my_dgettext' function, for
18247 consistency with the format string argument `my_format'. If the
18248 `format_arg' attribute had not been specified, all the compiler
18249 could tell in such calls to format functions would be that the
18250 format string argument is not constant; this would generate a
18251 warning when `-Wformat-nonliteral' is used, but the calls could
18252 not be checked without the attribute.
18254 The parameter STRING-INDEX specifies which argument is the format
18255 string argument (starting from one). Since non-static C++ methods
18256 have an implicit `this' argument, the arguments of such methods
18257 should be counted from two.
18259 The `format-arg' attribute allows you to identify your own
18260 functions which modify format strings, so that GCC can check the
18261 calls to `printf', `scanf', `strftime' or `strfmon' type function
18262 whose operands are a call to one of your own function. The
18263 compiler always treats `gettext', `dgettext', and `dcgettext' in
18264 this manner except when strict ISO C support is requested by
18265 `-ansi' or an appropriate `-std' option, or `-ffreestanding' or
18266 `-fno-builtin' is used. *Note Options Controlling C Dialect: C
18270 Use this attribute on the H8/300, H8/300H, and H8S to indicate
18271 that the specified function should be called through the function
18272 vector. Calling a function through the function vector will
18273 reduce code size, however; the function vector has a limited size
18274 (maximum 128 entries on the H8/300 and 64 entries on the H8/300H
18275 and H8S) and shares space with the interrupt vector.
18277 In SH2A target, this attribute declares a function to be called
18278 using the TBR relative addressing mode. The argument to this
18279 attribute is the entry number of the same function in a vector
18280 table containing all the TBR relative addressable functions. For
18281 the successful jump, register TBR should contain the start address
18282 of this TBR relative vector table. In the startup routine of the
18283 user application, user needs to care of this TBR register
18284 initialization. The TBR relative vector table can have at max 256
18285 function entries. The jumps to these functions will be generated
18286 using a SH2A specific, non delayed branch instruction JSR/N
18287 @(disp8,TBR). You must use GAS and GLD from GNU binutils version
18288 2.7 or later for this attribute to work correctly.
18290 Please refer the example of M16C target, to see the use of this
18291 attribute while declaring a function,
18293 In an application, for a function being called once, this
18294 attribute will save at least 8 bytes of code; and if other
18295 successive calls are being made to the same function, it will save
18296 2 bytes of code per each of these calls.
18298 On M16C/M32C targets, the `function_vector' attribute declares a
18299 special page subroutine call function. Use of this attribute
18300 reduces the code size by 2 bytes for each call generated to the
18301 subroutine. The argument to the attribute is the vector number
18302 entry from the special page vector table which contains the 16
18303 low-order bits of the subroutine's entry address. Each vector
18304 table has special page number (18 to 255) which are used in `jsrs'
18305 instruction. Jump addresses of the routines are generated by
18306 adding 0x0F0000 (in case of M16C targets) or 0xFF0000 (in case of
18307 M32C targets), to the 2 byte addresses set in the vector table.
18308 Therefore you need to ensure that all the special page vector
18309 routines should get mapped within the address range 0x0F0000 to
18310 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF (for M32C).
18312 In the following example 2 bytes will be saved for each call to
18315 void foo (void) __attribute__((function_vector(0x18)));
18325 If functions are defined in one file and are called in another
18326 file, then be sure to write this declaration in both files.
18328 This attribute is ignored for R8C target.
18331 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k, and
18332 Xstormy16 ports to indicate that the specified function is an
18333 interrupt handler. The compiler will generate function entry and
18334 exit sequences suitable for use in an interrupt handler when this
18335 attribute is present.
18337 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S,
18338 and SH processors can be specified via the `interrupt_handler'
18341 Note, on the AVR, interrupts will be enabled inside the function.
18343 Note, for the ARM, you can specify the kind of interrupt to be
18344 handled by adding an optional parameter to the interrupt attribute
18347 void f () __attribute__ ((interrupt ("IRQ")));
18349 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT
18352 On ARMv7-M the interrupt type is ignored, and the attribute means
18353 the function may be called with a word aligned stack pointer.
18355 `interrupt_handler'
18356 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S,
18357 and SH to indicate that the specified function is an interrupt
18358 handler. The compiler will generate function entry and exit
18359 sequences suitable for use in an interrupt handler when this
18360 attribute is present.
18363 Use this attribute on fido, a subarchitecture of the m68k, to
18364 indicate that the specified function is an interrupt handler that
18365 is designed to run as a thread. The compiler omits generate
18366 prologue/epilogue sequences and replaces the return instruction
18367 with a `sleep' instruction. This attribute is available only on
18371 Use this attribute on ARM to write Interrupt Service Routines.
18372 This is an alias to the `interrupt' attribute above.
18375 When used together with `interrupt_handler', `exception_handler'
18376 or `nmi_handler', code will be generated to load the stack pointer
18377 from the USP register in the function prologue.
18380 This attribute specifies a function to be placed into L1
18381 Instruction SRAM. The function will be put into a specific section
18382 named `.l1.text'. With `-mfdpic', function calls with a such
18383 function as the callee or caller will use inlined PLT.
18385 `long_call/short_call'
18386 This attribute specifies how a particular function is called on
18387 ARM. Both attributes override the `-mlong-calls' (*note ARM
18388 Options::) command line switch and `#pragma long_calls' settings.
18389 The `long_call' attribute indicates that the function might be far
18390 away from the call site and require a different (more expensive)
18391 calling sequence. The `short_call' attribute always places the
18392 offset to the function from the call site into the `BL'
18393 instruction directly.
18395 `longcall/shortcall'
18396 On the Blackfin, RS/6000 and PowerPC, the `longcall' attribute
18397 indicates that the function might be far away from the call site
18398 and require a different (more expensive) calling sequence. The
18399 `shortcall' attribute indicates that the function is always close
18400 enough for the shorter calling sequence to be used. These
18401 attributes override both the `-mlongcall' switch and, on the
18402 RS/6000 and PowerPC, the `#pragma longcall' setting.
18404 *Note RS/6000 and PowerPC Options::, for more information on
18405 whether long calls are necessary.
18407 `long_call/near/far'
18408 These attributes specify how a particular function is called on
18409 MIPS. The attributes override the `-mlong-calls' (*note MIPS
18410 Options::) command-line switch. The `long_call' and `far'
18411 attributes are synonyms, and cause the compiler to always call the
18412 function by first loading its address into a register, and then
18413 using the contents of that register. The `near' attribute has the
18414 opposite effect; it specifies that non-PIC calls should be made
18415 using the more efficient `jal' instruction.
18418 The `malloc' attribute is used to tell the compiler that a function
18419 may be treated as if any non-`NULL' pointer it returns cannot
18420 alias any other pointer valid when the function returns. This
18421 will often improve optimization. Standard functions with this
18422 property include `malloc' and `calloc'. `realloc'-like functions
18423 have this property as long as the old pointer is never referred to
18424 (including comparing it to the new pointer) after the function
18425 returns a non-`NULL' value.
18428 On MIPS targets, you can use the `mips16' and `nomips16' function
18429 attributes to locally select or turn off MIPS16 code generation.
18430 A function with the `mips16' attribute is emitted as MIPS16 code,
18431 while MIPS16 code generation is disabled for functions with the
18432 `nomips16' attribute. These attributes override the `-mips16' and
18433 `-mno-mips16' options on the command line (*note MIPS Options::).
18435 When compiling files containing mixed MIPS16 and non-MIPS16 code,
18436 the preprocessor symbol `__mips16' reflects the setting on the
18437 command line, not that within individual functions. Mixed MIPS16
18438 and non-MIPS16 code may interact badly with some GCC extensions
18439 such as `__builtin_apply' (*note Constructing Calls::).
18441 `model (MODEL-NAME)'
18442 On the M32R/D, use this attribute to set the addressability of an
18443 object, and of the code generated for a function. The identifier
18444 MODEL-NAME is one of `small', `medium', or `large', representing
18445 each of the code models.
18447 Small model objects live in the lower 16MB of memory (so that their
18448 addresses can be loaded with the `ld24' instruction), and are
18449 callable with the `bl' instruction.
18451 Medium model objects may live anywhere in the 32-bit address space
18452 (the compiler will generate `seth/add3' instructions to load their
18453 addresses), and are callable with the `bl' instruction.
18455 Large model objects may live anywhere in the 32-bit address space
18456 (the compiler will generate `seth/add3' instructions to load their
18457 addresses), and may not be reachable with the `bl' instruction
18458 (the compiler will generate the much slower `seth/add3/jl'
18459 instruction sequence).
18461 On IA-64, use this attribute to set the addressability of an
18462 object. At present, the only supported identifier for MODEL-NAME
18463 is `small', indicating addressability via "small" (22-bit)
18464 addresses (so that their addresses can be loaded with the `addl'
18465 instruction). Caveat: such addressing is by definition not
18466 position independent and hence this attribute must not be used for
18467 objects defined by shared libraries.
18470 On 64-bit x86_64-*-* targets, you can use an ABI attribute to
18471 indicate which calling convention should be used for a function.
18472 The `ms_abi' attribute tells the compiler to use the Microsoft
18473 ABI, while the `sysv_abi' attribute tells the compiler to use the
18474 ABI used on GNU/Linux and other systems. The default is to use
18475 the Microsoft ABI when targeting Windows. On all other systems,
18476 the default is the AMD ABI.
18478 Note, This feature is currently sorried out for Windows targets
18482 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate
18483 that the specified function does not need prologue/epilogue
18484 sequences generated by the compiler. It is up to the programmer
18485 to provide these sequences. The only statements that can be safely
18486 included in naked functions are `asm' statements that do not have
18487 operands. All other statements, including declarations of local
18488 variables, `if' statements, and so forth, should be avoided.
18489 Naked functions should be used to implement the body of an
18490 assembly function, while allowing the compiler to construct the
18491 requisite function declaration for the assembler.
18494 On 68HC11 and 68HC12 the `near' attribute causes the compiler to
18495 use the normal calling convention based on `jsr' and `rts'. This
18496 attribute can be used to cancel the effect of the `-mlong-calls'
18500 Use this attribute together with `interrupt_handler',
18501 `exception_handler' or `nmi_handler' to indicate that the function
18502 entry code should enable nested interrupts or exceptions.
18505 Use this attribute on the Blackfin to indicate that the specified
18506 function is an NMI handler. The compiler will generate function
18507 entry and exit sequences suitable for use in an NMI handler when
18508 this attribute is present.
18510 `no_instrument_function'
18511 If `-finstrument-functions' is given, profiling function calls will
18512 be generated at entry and exit of most user-compiled functions.
18513 Functions with this attribute will not be so instrumented.
18516 This function attribute prevents a function from being considered
18517 for inlining. If the function does not have side-effects, there
18518 are optimizations other than inlining that causes function calls
18519 to be optimized away, although the function call is live. To keep
18520 such calls from being optimized away, put
18522 (*note Extended Asm::) in the called function, to serve as a
18523 special side-effect.
18525 `nonnull (ARG-INDEX, ...)'
18526 The `nonnull' attribute specifies that some function parameters
18527 should be non-null pointers. For instance, the declaration:
18530 my_memcpy (void *dest, const void *src, size_t len)
18531 __attribute__((nonnull (1, 2)));
18533 causes the compiler to check that, in calls to `my_memcpy',
18534 arguments DEST and SRC are non-null. If the compiler determines
18535 that a null pointer is passed in an argument slot marked as
18536 non-null, and the `-Wnonnull' option is enabled, a warning is
18537 issued. The compiler may also choose to make optimizations based
18538 on the knowledge that certain function arguments will not be null.
18540 If no argument index list is given to the `nonnull' attribute, all
18541 pointer arguments are marked as non-null. To illustrate, the
18542 following declaration is equivalent to the previous example:
18545 my_memcpy (void *dest, const void *src, size_t len)
18546 __attribute__((nonnull));
18549 A few standard library functions, such as `abort' and `exit',
18550 cannot return. GCC knows this automatically. Some programs define
18551 their own functions that never return. You can declare them
18552 `noreturn' to tell the compiler this fact. For example,
18554 void fatal () __attribute__ ((noreturn));
18559 /* ... */ /* Print error message. */ /* ... */
18563 The `noreturn' keyword tells the compiler to assume that `fatal'
18564 cannot return. It can then optimize without regard to what would
18565 happen if `fatal' ever did return. This makes slightly better
18566 code. More importantly, it helps avoid spurious warnings of
18567 uninitialized variables.
18569 The `noreturn' keyword does not affect the exceptional path when
18570 that applies: a `noreturn'-marked function may still return to the
18571 caller by throwing an exception or calling `longjmp'.
18573 Do not assume that registers saved by the calling function are
18574 restored before calling the `noreturn' function.
18576 It does not make sense for a `noreturn' function to have a return
18577 type other than `void'.
18579 The attribute `noreturn' is not implemented in GCC versions
18580 earlier than 2.5. An alternative way to declare that a function
18581 does not return, which works in the current version and in some
18582 older versions, is as follows:
18584 typedef void voidfn ();
18586 volatile voidfn fatal;
18588 This approach does not work in GNU C++.
18591 The `nothrow' attribute is used to inform the compiler that a
18592 function cannot throw an exception. For example, most functions in
18593 the standard C library can be guaranteed not to throw an exception
18594 with the notable exceptions of `qsort' and `bsearch' that take
18595 function pointer arguments. The `nothrow' attribute is not
18596 implemented in GCC versions earlier than 3.3.
18599 The `optimize' attribute is used to specify that a function is to
18600 be compiled with different optimization options than specified on
18601 the command line. Arguments can either be numbers or strings.
18602 Numbers are assumed to be an optimization level. Strings that
18603 begin with `O' are assumed to be an optimization option, while
18604 other options are assumed to be used with a `-f' prefix. You can
18605 also use the `#pragma GCC optimize' pragma to set the optimization
18606 options that affect more than one function. *Note Function
18607 Specific Option Pragmas::, for details about the `#pragma GCC
18610 This can be used for instance to have frequently executed functions
18611 compiled with more aggressive optimization options that produce
18612 faster and larger code, while other functions can be called with
18613 less aggressive options.
18616 Many functions have no effects except the return value and their
18617 return value depends only on the parameters and/or global
18618 variables. Such a function can be subject to common subexpression
18619 elimination and loop optimization just as an arithmetic operator
18620 would be. These functions should be declared with the attribute
18621 `pure'. For example,
18623 int square (int) __attribute__ ((pure));
18625 says that the hypothetical function `square' is safe to call fewer
18626 times than the program says.
18628 Some of common examples of pure functions are `strlen' or `memcmp'.
18629 Interesting non-pure functions are functions with infinite loops
18630 or those depending on volatile memory or other system resource,
18631 that may change between two consecutive calls (such as `feof' in a
18632 multithreading environment).
18634 The attribute `pure' is not implemented in GCC versions earlier
18638 The `hot' attribute is used to inform the compiler that a function
18639 is a hot spot of the compiled program. The function is optimized
18640 more aggressively and on many target it is placed into special
18641 subsection of the text section so all hot functions appears close
18642 together improving locality.
18644 When profile feedback is available, via `-fprofile-use', hot
18645 functions are automatically detected and this attribute is ignored.
18647 The `hot' attribute is not implemented in GCC versions earlier
18651 The `cold' attribute is used to inform the compiler that a
18652 function is unlikely executed. The function is optimized for size
18653 rather than speed and on many targets it is placed into special
18654 subsection of the text section so all cold functions appears close
18655 together improving code locality of non-cold parts of program.
18656 The paths leading to call of cold functions within code are marked
18657 as unlikely by the branch prediction mechanism. It is thus useful
18658 to mark functions used to handle unlikely conditions, such as
18659 `perror', as cold to improve optimization of hot functions that do
18660 call marked functions in rare occasions.
18662 When profile feedback is available, via `-fprofile-use', hot
18663 functions are automatically detected and this attribute is ignored.
18665 The `cold' attribute is not implemented in GCC versions earlier
18669 On the Intel 386, the `regparm' attribute causes the compiler to
18670 pass arguments number one to NUMBER if they are of integral type
18671 in registers EAX, EDX, and ECX instead of on the stack. Functions
18672 that take a variable number of arguments will continue to be
18673 passed all of their arguments on the stack.
18675 Beware that on some ELF systems this attribute is unsuitable for
18676 global functions in shared libraries with lazy binding (which is
18677 the default). Lazy binding will send the first call via resolving
18678 code in the loader, which might assume EAX, EDX and ECX can be
18679 clobbered, as per the standard calling conventions. Solaris 8 is
18680 affected by this. GNU systems with GLIBC 2.1 or higher, and
18681 FreeBSD, are believed to be safe since the loaders there save EAX,
18682 EDX and ECX. (Lazy binding can be disabled with the linker or the
18683 loader if desired, to avoid the problem.)
18686 On the Intel 386 with SSE support, the `sseregparm' attribute
18687 causes the compiler to pass up to 3 floating point arguments in
18688 SSE registers instead of on the stack. Functions that take a
18689 variable number of arguments will continue to pass all of their
18690 floating point arguments on the stack.
18692 `force_align_arg_pointer'
18693 On the Intel x86, the `force_align_arg_pointer' attribute may be
18694 applied to individual function definitions, generating an alternate
18695 prologue and epilogue that realigns the runtime stack if necessary.
18696 This supports mixing legacy codes that run with a 4-byte aligned
18697 stack with modern codes that keep a 16-byte stack for SSE
18701 On the SH2A target, this attribute enables the high-speed register
18702 saving and restoration using a register bank for
18703 `interrupt_handler' routines. Saving to the bank is performed
18704 automatically after the CPU accepts an interrupt that uses a
18707 The nineteen 32-bit registers comprising general register R0 to
18708 R14, control register GBR, and system registers MACH, MACL, and PR
18709 and the vector table address offset are saved into a register
18710 bank. Register banks are stacked in first-in last-out (FILO)
18711 sequence. Restoration from the bank is executed by issuing a
18712 RESBANK instruction.
18715 The `returns_twice' attribute tells the compiler that a function
18716 may return more than one time. The compiler will ensure that all
18717 registers are dead before calling such a function and will emit a
18718 warning about the variables that may be clobbered after the second
18719 return from the function. Examples of such functions are `setjmp'
18720 and `vfork'. The `longjmp'-like counterpart of such function, if
18721 any, might need to be marked with the `noreturn' attribute.
18724 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to
18725 indicate that all registers except the stack pointer should be
18726 saved in the prologue regardless of whether they are used or not.
18728 `section ("SECTION-NAME")'
18729 Normally, the compiler places the code it generates in the `text'
18730 section. Sometimes, however, you need additional sections, or you
18731 need certain particular functions to appear in special sections.
18732 The `section' attribute specifies that a function lives in a
18733 particular section. For example, the declaration:
18735 extern void foobar (void) __attribute__ ((section ("bar")));
18737 puts the function `foobar' in the `bar' section.
18739 Some file formats do not support arbitrary sections so the
18740 `section' attribute is not available on all platforms. If you
18741 need to map the entire contents of a module to a particular
18742 section, consider using the facilities of the linker instead.
18745 This function attribute ensures that a parameter in a function
18746 call is an explicit `NULL'. The attribute is only valid on
18747 variadic functions. By default, the sentinel is located at
18748 position zero, the last parameter of the function call. If an
18749 optional integer position argument P is supplied to the attribute,
18750 the sentinel must be located at position P counting backwards from
18751 the end of the argument list.
18753 __attribute__ ((sentinel))
18755 __attribute__ ((sentinel(0)))
18757 The attribute is automatically set with a position of 0 for the
18758 built-in functions `execl' and `execlp'. The built-in function
18759 `execle' has the attribute set with a position of 1.
18761 A valid `NULL' in this context is defined as zero with any pointer
18762 type. If your system defines the `NULL' macro with an integer type
18763 then you need to add an explicit cast. GCC replaces `stddef.h'
18764 with a copy that redefines NULL appropriately.
18766 The warnings for missing or incorrect sentinels are enabled with
18770 See long_call/short_call.
18773 See longcall/shortcall.
18776 Use this attribute on the AVR to indicate that the specified
18777 function is a signal handler. The compiler will generate function
18778 entry and exit sequences suitable for use in a signal handler when
18779 this attribute is present. Interrupts will be disabled inside the
18783 Use this attribute on the SH to indicate an `interrupt_handler'
18784 function should switch to an alternate stack. It expects a string
18785 argument that names a global variable holding the address of the
18789 void f () __attribute__ ((interrupt_handler,
18790 sp_switch ("alt_stack")));
18793 On the Intel 386, the `stdcall' attribute causes the compiler to
18794 assume that the called function will pop off the stack space used
18795 to pass arguments, unless it takes a variable number of arguments.
18798 This attribute is used to modify the IA64 calling convention by
18799 marking all input registers as live at all function exits. This
18800 makes it possible to restart a system call after an interrupt
18801 without having to save/restore the input registers. This also
18802 prevents kernel data from leaking into application code.
18805 The `target' attribute is used to specify that a function is to be
18806 compiled with different target options than specified on the
18807 command line. This can be used for instance to have functions
18808 compiled with a different ISA (instruction set architecture) than
18809 the default. You can also use the `#pragma GCC target' pragma to
18810 set more than one function to be compiled with specific target
18811 options. *Note Function Specific Option Pragmas::, for details
18812 about the `#pragma GCC target' pragma.
18814 For instance on a 386, you could compile one function with
18815 `target("sse4.1,arch=core2")' and another with
18816 `target("sse4a,arch=amdfam10")' that would be equivalent to
18817 compiling the first function with `-msse4.1' and `-march=core2'
18818 options, and the second function with `-msse4a' and
18819 `-march=amdfam10' options. It is up to the user to make sure that
18820 a function is only invoked on a machine that supports the
18821 particular ISA it was compiled for (for example by using `cpuid'
18822 on 386 to determine what feature bits and architecture family are
18825 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
18826 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
18828 On the 386, the following options are allowed:
18832 Enable/disable the generation of the advanced bit
18837 Enable/disable the generation of the AES instructions.
18841 Enable/disable the generation of the MMX instructions.
18845 Enable/disable the generation of the PCLMUL instructions.
18849 Enable/disable the generation of the POPCNT instruction.
18853 Enable/disable the generation of the SSE instructions.
18857 Enable/disable the generation of the SSE2 instructions.
18861 Enable/disable the generation of the SSE3 instructions.
18865 Enable/disable the generation of the SSE4 instructions (both
18866 SSE4.1 and SSE4.2).
18870 Enable/disable the generation of the sse4.1 instructions.
18874 Enable/disable the generation of the sse4.2 instructions.
18878 Enable/disable the generation of the SSE4A instructions.
18882 Enable/disable the generation of the SSE5 instructions.
18886 Enable/disable the generation of the SSSE3 instructions.
18890 Enable/disable the generation of the CLD before string moves.
18893 `no-fancy-math-387'
18894 Enable/disable the generation of the `sin', `cos', and `sqrt'
18895 instructions on the 387 floating point unit.
18899 Enable/disable the generation of the fused multiply/add
18904 Enable/disable the generation of floating point that depends
18905 on IEEE arithmetic.
18907 `inline-all-stringops'
18908 `no-inline-all-stringops'
18909 Enable/disable inlining of string operations.
18911 `inline-stringops-dynamically'
18912 `no-inline-stringops-dynamically'
18913 Enable/disable the generation of the inline code to do small
18914 string operations and calling the library routines for large
18918 `no-align-stringops'
18919 Do/do not align destination of inlined string operations.
18923 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and
18924 RSQRTPS instructions followed an additional Newton-Raphson
18925 step instead of doing a floating point division.
18928 Specify the architecture to generate code for in compiling
18932 Specify the architecture to tune for in compiling the
18936 Specify which floating point unit to use. The
18937 `target("fpmath=sse,387")' option must be specified as
18938 `target("fpmath=sse+387")' because the comma would separate
18941 On the 386, you can use either multiple strings to specify multiple
18942 options, or you can separate the option with a comma (`,').
18944 On the 386, the inliner will not inline a function that has
18945 different target options than the caller, unless the callee has a
18946 subset of the target options of the caller. For example a
18947 function declared with `target("sse5")' can inline a function with
18948 `target("sse2")', since `-msse5' implies `-msse2'.
18950 The `target' attribute is not implemented in GCC versions earlier
18951 than 4.4, and at present only the 386 uses it.
18954 Use this attribute on the H8/300H and H8S to indicate that the
18955 specified variable should be placed into the tiny data section.
18956 The compiler will generate more efficient code for loads and stores
18957 on data in the tiny data section. Note the tiny data area is
18958 limited to slightly under 32kbytes of data.
18961 Use this attribute on the SH for an `interrupt_handler' to return
18962 using `trapa' instead of `rte'. This attribute expects an integer
18963 argument specifying the trap number to be used.
18966 This attribute, attached to a function, means that the function is
18967 meant to be possibly unused. GCC will not produce a warning for
18971 This attribute, attached to a function, means that code must be
18972 emitted for the function even if it appears that the function is
18973 not referenced. This is useful, for example, when the function is
18974 referenced only in inline assembly.
18977 This IA64 HP-UX attribute, attached to a global variable or
18978 function, renames a symbol to contain a version string, thus
18979 allowing for function level versioning. HP-UX system header files
18980 may use version level functioning for some system calls.
18982 extern int foo () __attribute__((version_id ("20040821")));
18984 Calls to FOO will be mapped to calls to FOO{20040821}.
18986 `visibility ("VISIBILITY_TYPE")'
18987 This attribute affects the linkage of the declaration to which it
18988 is attached. There are four supported VISIBILITY_TYPE values:
18989 default, hidden, protected or internal visibility.
18991 void __attribute__ ((visibility ("protected")))
18992 f () { /* Do something. */; }
18993 int i __attribute__ ((visibility ("hidden")));
18995 The possible values of VISIBILITY_TYPE correspond to the
18996 visibility settings in the ELF gABI.
18999 Default visibility is the normal case for the object file
19000 format. This value is available for the visibility attribute
19001 to override other options that may change the assumed
19002 visibility of entities.
19004 On ELF, default visibility means that the declaration is
19005 visible to other modules and, in shared libraries, means that
19006 the declared entity may be overridden.
19008 On Darwin, default visibility means that the declaration is
19009 visible to other modules.
19011 Default visibility corresponds to "external linkage" in the
19015 Hidden visibility indicates that the entity declared will
19016 have a new form of linkage, which we'll call "hidden
19017 linkage". Two declarations of an object with hidden linkage
19018 refer to the same object if they are in the same shared
19022 Internal visibility is like hidden visibility, but with
19023 additional processor specific semantics. Unless otherwise
19024 specified by the psABI, GCC defines internal visibility to
19025 mean that a function is _never_ called from another module.
19026 Compare this with hidden functions which, while they cannot
19027 be referenced directly by other modules, can be referenced
19028 indirectly via function pointers. By indicating that a
19029 function cannot be called from outside the module, GCC may
19030 for instance omit the load of a PIC register since it is known
19031 that the calling function loaded the correct value.
19034 Protected visibility is like default visibility except that it
19035 indicates that references within the defining module will
19036 bind to the definition in that module. That is, the declared
19037 entity cannot be overridden by another module.
19040 All visibilities are supported on many, but not all, ELF targets
19041 (supported when the assembler supports the `.visibility'
19042 pseudo-op). Default visibility is supported everywhere. Hidden
19043 visibility is supported on Darwin targets.
19045 The visibility attribute should be applied only to declarations
19046 which would otherwise have external linkage. The attribute should
19047 be applied consistently, so that the same entity should not be
19048 declared with different settings of the attribute.
19050 In C++, the visibility attribute applies to types as well as
19051 functions and objects, because in C++ types have linkage. A class
19052 must not have greater visibility than its non-static data member
19053 types and bases, and class members default to the visibility of
19054 their class. Also, a declaration without explicit visibility is
19055 limited to the visibility of its type.
19057 In C++, you can mark member functions and static member variables
19058 of a class with the visibility attribute. This is useful if you
19059 know a particular method or static member variable should only be
19060 used from one shared object; then you can mark it hidden while the
19061 rest of the class has default visibility. Care must be taken to
19062 avoid breaking the One Definition Rule; for example, it is usually
19063 not useful to mark an inline method as hidden without marking the
19064 whole class as hidden.
19066 A C++ namespace declaration can also have the visibility attribute.
19067 This attribute applies only to the particular namespace body, not
19068 to other definitions of the same namespace; it is equivalent to
19069 using `#pragma GCC visibility' before and after the namespace
19070 definition (*note Visibility Pragmas::).
19072 In C++, if a template argument has limited visibility, this
19073 restriction is implicitly propagated to the template instantiation.
19074 Otherwise, template instantiations and specializations default to
19075 the visibility of their template.
19077 If both the template and enclosing class have explicit visibility,
19078 the visibility from the template is used.
19080 `warn_unused_result'
19081 The `warn_unused_result' attribute causes a warning to be emitted
19082 if a caller of the function with this attribute does not use its
19083 return value. This is useful for functions where not checking the
19084 result is either a security problem or always a bug, such as
19087 int fn () __attribute__ ((warn_unused_result));
19090 if (fn () < 0) return -1;
19095 results in warning on line 5.
19098 The `weak' attribute causes the declaration to be emitted as a weak
19099 symbol rather than a global. This is primarily useful in defining
19100 library functions which can be overridden in user code, though it
19101 can also be used with non-function declarations. Weak symbols are
19102 supported for ELF targets, and also for a.out targets when using
19103 the GNU assembler and linker.
19106 `weakref ("TARGET")'
19107 The `weakref' attribute marks a declaration as a weak reference.
19108 Without arguments, it should be accompanied by an `alias' attribute
19109 naming the target symbol. Optionally, the TARGET may be given as
19110 an argument to `weakref' itself. In either case, `weakref'
19111 implicitly marks the declaration as `weak'. Without a TARGET,
19112 given as an argument to `weakref' or to `alias', `weakref' is
19113 equivalent to `weak'.
19115 static int x() __attribute__ ((weakref ("y")));
19116 /* is equivalent to... */
19117 static int x() __attribute__ ((weak, weakref, alias ("y")));
19119 static int x() __attribute__ ((weakref));
19120 static int x() __attribute__ ((alias ("y")));
19122 A weak reference is an alias that does not by itself require a
19123 definition to be given for the target symbol. If the target
19124 symbol is only referenced through weak references, then the
19125 becomes a `weak' undefined symbol. If it is directly referenced,
19126 however, then such strong references prevail, and a definition
19127 will be required for the symbol, not necessarily in the same
19130 The effect is equivalent to moving all references to the alias to a
19131 separate translation unit, renaming the alias to the aliased
19132 symbol, declaring it as weak, compiling the two separate
19133 translation units and performing a reloadable link on them.
19135 At present, a declaration to which `weakref' is attached can only
19139 You can specify multiple attributes in a declaration by separating them
19140 by commas within the double parentheses or by immediately following an
19141 attribute declaration with another attribute declaration.
19143 Some people object to the `__attribute__' feature, suggesting that ISO
19144 C's `#pragma' should be used instead. At the time `__attribute__' was
19145 designed, there were two reasons for not doing this.
19147 1. It is impossible to generate `#pragma' commands from a macro.
19149 2. There is no telling what the same `#pragma' might mean in another
19152 These two reasons applied to almost any application that might have
19153 been proposed for `#pragma'. It was basically a mistake to use
19154 `#pragma' for _anything_.
19156 The ISO C99 standard includes `_Pragma', which now allows pragmas to
19157 be generated from macros. In addition, a `#pragma GCC' namespace is
19158 now in use for GCC-specific pragmas. However, it has been found
19159 convenient to use `__attribute__' to achieve a natural attachment of
19160 attributes to their corresponding declarations, whereas `#pragma GCC'
19161 is of use for constructs that do not naturally form part of the
19162 grammar. *Note Miscellaneous Preprocessing Directives: (cpp)Other
19166 File: gcc.info, Node: Attribute Syntax, Next: Function Prototypes, Prev: Function Attributes, Up: C Extensions
19168 5.28 Attribute Syntax
19169 =====================
19171 This section describes the syntax with which `__attribute__' may be
19172 used, and the constructs to which attribute specifiers bind, for the C
19173 language. Some details may vary for C++ and Objective-C. Because of
19174 infelicities in the grammar for attributes, some forms described here
19175 may not be successfully parsed in all cases.
19177 There are some problems with the semantics of attributes in C++. For
19178 example, there are no manglings for attributes, although they may affect
19179 code generation, so problems may arise when attributed types are used in
19180 conjunction with templates or overloading. Similarly, `typeid' does
19181 not distinguish between types with different attributes. Support for
19182 attributes in C++ may be restricted in future to attributes on
19183 declarations only, but not on nested declarators.
19185 *Note Function Attributes::, for details of the semantics of attributes
19186 applying to functions. *Note Variable Attributes::, for details of the
19187 semantics of attributes applying to variables. *Note Type Attributes::,
19188 for details of the semantics of attributes applying to structure, union
19189 and enumerated types.
19191 An "attribute specifier" is of the form `__attribute__
19192 ((ATTRIBUTE-LIST))'. An "attribute list" is a possibly empty
19193 comma-separated sequence of "attributes", where each attribute is one
19196 * Empty. Empty attributes are ignored.
19198 * A word (which may be an identifier such as `unused', or a reserved
19199 word such as `const').
19201 * A word, followed by, in parentheses, parameters for the attribute.
19202 These parameters take one of the following forms:
19204 * An identifier. For example, `mode' attributes use this form.
19206 * An identifier followed by a comma and a non-empty
19207 comma-separated list of expressions. For example, `format'
19208 attributes use this form.
19210 * A possibly empty comma-separated list of expressions. For
19211 example, `format_arg' attributes use this form with the list
19212 being a single integer constant expression, and `alias'
19213 attributes use this form with the list being a single string
19216 An "attribute specifier list" is a sequence of one or more attribute
19217 specifiers, not separated by any other tokens.
19219 In GNU C, an attribute specifier list may appear after the colon
19220 following a label, other than a `case' or `default' label. The only
19221 attribute it makes sense to use after a label is `unused'. This
19222 feature is intended for code generated by programs which contains labels
19223 that may be unused but which is compiled with `-Wall'. It would not
19224 normally be appropriate to use in it human-written code, though it
19225 could be useful in cases where the code that jumps to the label is
19226 contained within an `#ifdef' conditional. GNU C++ does not permit such
19227 placement of attribute lists, as it is permissible for a declaration,
19228 which could begin with an attribute list, to be labelled in C++.
19229 Declarations cannot be labelled in C90 or C99, so the ambiguity does
19232 An attribute specifier list may appear as part of a `struct', `union'
19233 or `enum' specifier. It may go either immediately after the `struct',
19234 `union' or `enum' keyword, or after the closing brace. The former
19235 syntax is preferred. Where attribute specifiers follow the closing
19236 brace, they are considered to relate to the structure, union or
19237 enumerated type defined, not to any enclosing declaration the type
19238 specifier appears in, and the type defined is not complete until after
19239 the attribute specifiers.
19241 Otherwise, an attribute specifier appears as part of a declaration,
19242 counting declarations of unnamed parameters and type names, and relates
19243 to that declaration (which may be nested in another declaration, for
19244 example in the case of a parameter declaration), or to a particular
19245 declarator within a declaration. Where an attribute specifier is
19246 applied to a parameter declared as a function or an array, it should
19247 apply to the function or array rather than the pointer to which the
19248 parameter is implicitly converted, but this is not yet correctly
19251 Any list of specifiers and qualifiers at the start of a declaration may
19252 contain attribute specifiers, whether or not such a list may in that
19253 context contain storage class specifiers. (Some attributes, however,
19254 are essentially in the nature of storage class specifiers, and only make
19255 sense where storage class specifiers may be used; for example,
19256 `section'.) There is one necessary limitation to this syntax: the
19257 first old-style parameter declaration in a function definition cannot
19258 begin with an attribute specifier, because such an attribute applies to
19259 the function instead by syntax described below (which, however, is not
19260 yet implemented in this case). In some other cases, attribute
19261 specifiers are permitted by this grammar but not yet supported by the
19262 compiler. All attribute specifiers in this place relate to the
19263 declaration as a whole. In the obsolescent usage where a type of `int'
19264 is implied by the absence of type specifiers, such a list of specifiers
19265 and qualifiers may be an attribute specifier list with no other
19266 specifiers or qualifiers.
19268 At present, the first parameter in a function prototype must have some
19269 type specifier which is not an attribute specifier; this resolves an
19270 ambiguity in the interpretation of `void f(int (__attribute__((foo))
19271 x))', but is subject to change. At present, if the parentheses of a
19272 function declarator contain only attributes then those attributes are
19273 ignored, rather than yielding an error or warning or implying a single
19274 parameter of type int, but this is subject to change.
19276 An attribute specifier list may appear immediately before a declarator
19277 (other than the first) in a comma-separated list of declarators in a
19278 declaration of more than one identifier using a single list of
19279 specifiers and qualifiers. Such attribute specifiers apply only to the
19280 identifier before whose declarator they appear. For example, in
19282 __attribute__((noreturn)) void d0 (void),
19283 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
19286 the `noreturn' attribute applies to all the functions declared; the
19287 `format' attribute only applies to `d1'.
19289 An attribute specifier list may appear immediately before the comma,
19290 `=' or semicolon terminating the declaration of an identifier other
19291 than a function definition. Such attribute specifiers apply to the
19292 declared object or function. Where an assembler name for an object or
19293 function is specified (*note Asm Labels::), the attribute must follow
19294 the `asm' specification.
19296 An attribute specifier list may, in future, be permitted to appear
19297 after the declarator in a function definition (before any old-style
19298 parameter declarations or the function body).
19300 Attribute specifiers may be mixed with type qualifiers appearing inside
19301 the `[]' of a parameter array declarator, in the C99 construct by which
19302 such qualifiers are applied to the pointer to which the array is
19303 implicitly converted. Such attribute specifiers apply to the pointer,
19304 not to the array, but at present this is not implemented and they are
19307 An attribute specifier list may appear at the start of a nested
19308 declarator. At present, there are some limitations in this usage: the
19309 attributes correctly apply to the declarator, but for most individual
19310 attributes the semantics this implies are not implemented. When
19311 attribute specifiers follow the `*' of a pointer declarator, they may
19312 be mixed with any type qualifiers present. The following describes the
19313 formal semantics of this syntax. It will make the most sense if you
19314 are familiar with the formal specification of declarators in the ISO C
19317 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration `T D1',
19318 where `T' contains declaration specifiers that specify a type TYPE
19319 (such as `int') and `D1' is a declarator that contains an identifier
19320 IDENT. The type specified for IDENT for derived declarators whose type
19321 does not include an attribute specifier is as in the ISO C standard.
19323 If `D1' has the form `( ATTRIBUTE-SPECIFIER-LIST D )', and the
19324 declaration `T D' specifies the type "DERIVED-DECLARATOR-TYPE-LIST
19325 TYPE" for IDENT, then `T D1' specifies the type
19326 "DERIVED-DECLARATOR-TYPE-LIST ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
19328 If `D1' has the form `* TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST
19329 D', and the declaration `T D' specifies the type
19330 "DERIVED-DECLARATOR-TYPE-LIST TYPE" for IDENT, then `T D1' specifies
19331 the type "DERIVED-DECLARATOR-TYPE-LIST
19332 TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
19336 void (__attribute__((noreturn)) ****f) (void);
19338 specifies the type "pointer to pointer to pointer to pointer to
19339 non-returning function returning `void'". As another example,
19341 char *__attribute__((aligned(8))) *f;
19343 specifies the type "pointer to 8-byte-aligned pointer to `char'". Note
19344 again that this does not work with most attributes; for example, the
19345 usage of `aligned' and `noreturn' attributes given above is not yet
19348 For compatibility with existing code written for compiler versions that
19349 did not implement attributes on nested declarators, some laxity is
19350 allowed in the placing of attributes. If an attribute that only applies
19351 to types is applied to a declaration, it will be treated as applying to
19352 the type of that declaration. If an attribute that only applies to
19353 declarations is applied to the type of a declaration, it will be treated
19354 as applying to that declaration; and, for compatibility with code
19355 placing the attributes immediately before the identifier declared, such
19356 an attribute applied to a function return type will be treated as
19357 applying to the function type, and such an attribute applied to an array
19358 element type will be treated as applying to the array type. If an
19359 attribute that only applies to function types is applied to a
19360 pointer-to-function type, it will be treated as applying to the pointer
19361 target type; if such an attribute is applied to a function return type
19362 that is not a pointer-to-function type, it will be treated as applying
19363 to the function type.
19366 File: gcc.info, Node: Function Prototypes, Next: C++ Comments, Prev: Attribute Syntax, Up: C Extensions
19368 5.29 Prototypes and Old-Style Function Definitions
19369 ==================================================
19371 GNU C extends ISO C to allow a function prototype to override a later
19372 old-style non-prototype definition. Consider the following example:
19374 /* Use prototypes unless the compiler is old-fashioned. */
19381 /* Prototype function declaration. */
19382 int isroot P((uid_t));
19384 /* Old-style function definition. */
19386 isroot (x) /* ??? lossage here ??? */
19392 Suppose the type `uid_t' happens to be `short'. ISO C does not allow
19393 this example, because subword arguments in old-style non-prototype
19394 definitions are promoted. Therefore in this example the function
19395 definition's argument is really an `int', which does not match the
19396 prototype argument type of `short'.
19398 This restriction of ISO C makes it hard to write code that is portable
19399 to traditional C compilers, because the programmer does not know
19400 whether the `uid_t' type is `short', `int', or `long'. Therefore, in
19401 cases like these GNU C allows a prototype to override a later old-style
19402 definition. More precisely, in GNU C, a function prototype argument
19403 type overrides the argument type specified by a later old-style
19404 definition if the former type is the same as the latter type before
19405 promotion. Thus in GNU C the above example is equivalent to the
19408 int isroot (uid_t);
19416 GNU C++ does not support old-style function definitions, so this
19417 extension is irrelevant.
19420 File: gcc.info, Node: C++ Comments, Next: Dollar Signs, Prev: Function Prototypes, Up: C Extensions
19422 5.30 C++ Style Comments
19423 =======================
19425 In GNU C, you may use C++ style comments, which start with `//' and
19426 continue until the end of the line. Many other C implementations allow
19427 such comments, and they are included in the 1999 C standard. However,
19428 C++ style comments are not recognized if you specify an `-std' option
19429 specifying a version of ISO C before C99, or `-ansi' (equivalent to
19433 File: gcc.info, Node: Dollar Signs, Next: Character Escapes, Prev: C++ Comments, Up: C Extensions
19435 5.31 Dollar Signs in Identifier Names
19436 =====================================
19438 In GNU C, you may normally use dollar signs in identifier names. This
19439 is because many traditional C implementations allow such identifiers.
19440 However, dollar signs in identifiers are not supported on a few target
19441 machines, typically because the target assembler does not allow them.
19444 File: gcc.info, Node: Character Escapes, Next: Variable Attributes, Prev: Dollar Signs, Up: C Extensions
19446 5.32 The Character <ESC> in Constants
19447 =====================================
19449 You can use the sequence `\e' in a string or character constant to
19450 stand for the ASCII character <ESC>.
19453 File: gcc.info, Node: Alignment, Next: Inline, Prev: Type Attributes, Up: C Extensions
19455 5.33 Inquiring on Alignment of Types or Variables
19456 =================================================
19458 The keyword `__alignof__' allows you to inquire about how an object is
19459 aligned, or the minimum alignment usually required by a type. Its
19460 syntax is just like `sizeof'.
19462 For example, if the target machine requires a `double' value to be
19463 aligned on an 8-byte boundary, then `__alignof__ (double)' is 8. This
19464 is true on many RISC machines. On more traditional machine designs,
19465 `__alignof__ (double)' is 4 or even 2.
19467 Some machines never actually require alignment; they allow reference
19468 to any data type even at an odd address. For these machines,
19469 `__alignof__' reports the smallest alignment that GCC will give the
19470 data type, usually as mandated by the target ABI.
19472 If the operand of `__alignof__' is an lvalue rather than a type, its
19473 value is the required alignment for its type, taking into account any
19474 minimum alignment specified with GCC's `__attribute__' extension (*note
19475 Variable Attributes::). For example, after this declaration:
19477 struct foo { int x; char y; } foo1;
19479 the value of `__alignof__ (foo1.y)' is 1, even though its actual
19480 alignment is probably 2 or 4, the same as `__alignof__ (int)'.
19482 It is an error to ask for the alignment of an incomplete type.
19485 File: gcc.info, Node: Variable Attributes, Next: Type Attributes, Prev: Character Escapes, Up: C Extensions
19487 5.34 Specifying Attributes of Variables
19488 =======================================
19490 The keyword `__attribute__' allows you to specify special attributes of
19491 variables or structure fields. This keyword is followed by an
19492 attribute specification inside double parentheses. Some attributes are
19493 currently defined generically for variables. Other attributes are
19494 defined for variables on particular target systems. Other attributes
19495 are available for functions (*note Function Attributes::) and for types
19496 (*note Type Attributes::). Other front ends might define more
19497 attributes (*note Extensions to the C++ Language: C++ Extensions.).
19499 You may also specify attributes with `__' preceding and following each
19500 keyword. This allows you to use them in header files without being
19501 concerned about a possible macro of the same name. For example, you
19502 may use `__aligned__' instead of `aligned'.
19504 *Note Attribute Syntax::, for details of the exact syntax for using
19507 `aligned (ALIGNMENT)'
19508 This attribute specifies a minimum alignment for the variable or
19509 structure field, measured in bytes. For example, the declaration:
19511 int x __attribute__ ((aligned (16))) = 0;
19513 causes the compiler to allocate the global variable `x' on a
19514 16-byte boundary. On a 68040, this could be used in conjunction
19515 with an `asm' expression to access the `move16' instruction which
19516 requires 16-byte aligned operands.
19518 You can also specify the alignment of structure fields. For
19519 example, to create a double-word aligned `int' pair, you could
19522 struct foo { int x[2] __attribute__ ((aligned (8))); };
19524 This is an alternative to creating a union with a `double' member
19525 that forces the union to be double-word aligned.
19527 As in the preceding examples, you can explicitly specify the
19528 alignment (in bytes) that you wish the compiler to use for a given
19529 variable or structure field. Alternatively, you can leave out the
19530 alignment factor and just ask the compiler to align a variable or
19531 field to the default alignment for the target architecture you are
19532 compiling for. The default alignment is sufficient for all scalar
19533 types, but may not be enough for all vector types on a target
19534 which supports vector operations. The default alignment is fixed
19535 for a particular target ABI.
19537 Gcc also provides a target specific macro `__BIGGEST_ALIGNMENT__',
19538 which is the largest alignment ever used for any data type on the
19539 target machine you are compiling for. For example, you could
19542 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
19544 The compiler automatically sets the alignment for the declared
19545 variable or field to `__BIGGEST_ALIGNMENT__'. Doing this can
19546 often make copy operations more efficient, because the compiler can
19547 use whatever instructions copy the biggest chunks of memory when
19548 performing copies to or from the variables or fields that you have
19549 aligned this way. Note that the value of `__BIGGEST_ALIGNMENT__'
19550 may change depending on command line options.
19552 When used on a struct, or struct member, the `aligned' attribute
19553 can only increase the alignment; in order to decrease it, the
19554 `packed' attribute must be specified as well. When used as part
19555 of a typedef, the `aligned' attribute can both increase and
19556 decrease alignment, and specifying the `packed' attribute will
19557 generate a warning.
19559 Note that the effectiveness of `aligned' attributes may be limited
19560 by inherent limitations in your linker. On many systems, the
19561 linker is only able to arrange for variables to be aligned up to a
19562 certain maximum alignment. (For some linkers, the maximum
19563 supported alignment may be very very small.) If your linker is
19564 only able to align variables up to a maximum of 8 byte alignment,
19565 then specifying `aligned(16)' in an `__attribute__' will still
19566 only provide you with 8 byte alignment. See your linker
19567 documentation for further information.
19569 The `aligned' attribute can also be used for functions (*note
19570 Function Attributes::.)
19572 `cleanup (CLEANUP_FUNCTION)'
19573 The `cleanup' attribute runs a function when the variable goes out
19574 of scope. This attribute can only be applied to auto function
19575 scope variables; it may not be applied to parameters or variables
19576 with static storage duration. The function must take one
19577 parameter, a pointer to a type compatible with the variable. The
19578 return value of the function (if any) is ignored.
19580 If `-fexceptions' is enabled, then CLEANUP_FUNCTION will be run
19581 during the stack unwinding that happens during the processing of
19582 the exception. Note that the `cleanup' attribute does not allow
19583 the exception to be caught, only to perform an action. It is
19584 undefined what happens if CLEANUP_FUNCTION does not return
19589 The `common' attribute requests GCC to place a variable in
19590 "common" storage. The `nocommon' attribute requests the
19591 opposite--to allocate space for it directly.
19593 These attributes override the default chosen by the `-fno-common'
19594 and `-fcommon' flags respectively.
19597 The `deprecated' attribute results in a warning if the variable is
19598 used anywhere in the source file. This is useful when identifying
19599 variables that are expected to be removed in a future version of a
19600 program. The warning also includes the location of the declaration
19601 of the deprecated variable, to enable users to easily find further
19602 information about why the variable is deprecated, or what they
19603 should do instead. Note that the warning only occurs for uses:
19605 extern int old_var __attribute__ ((deprecated));
19606 extern int old_var;
19607 int new_fn () { return old_var; }
19609 results in a warning on line 3 but not line 2.
19611 The `deprecated' attribute can also be used for functions and
19612 types (*note Function Attributes::, *note Type Attributes::.)
19615 This attribute specifies the data type for the
19616 declaration--whichever type corresponds to the mode MODE. This in
19617 effect lets you request an integer or floating point type
19618 according to its width.
19620 You may also specify a mode of `byte' or `__byte__' to indicate
19621 the mode corresponding to a one-byte integer, `word' or `__word__'
19622 for the mode of a one-word integer, and `pointer' or `__pointer__'
19623 for the mode used to represent pointers.
19626 The `packed' attribute specifies that a variable or structure field
19627 should have the smallest possible alignment--one byte for a
19628 variable, and one bit for a field, unless you specify a larger
19629 value with the `aligned' attribute.
19631 Here is a structure in which the field `x' is packed, so that it
19632 immediately follows `a':
19637 int x[2] __attribute__ ((packed));
19640 _Note:_ The 4.1, 4.2 and 4.3 series of GCC ignore the `packed'
19641 attribute on bit-fields of type `char'. This has been fixed in
19642 GCC 4.4 but the change can lead to differences in the structure
19643 layout. See the documentation of `-Wpacked-bitfield-compat' for
19646 `section ("SECTION-NAME")'
19647 Normally, the compiler places the objects it generates in sections
19648 like `data' and `bss'. Sometimes, however, you need additional
19649 sections, or you need certain particular variables to appear in
19650 special sections, for example to map to special hardware. The
19651 `section' attribute specifies that a variable (or function) lives
19652 in a particular section. For example, this small program uses
19653 several specific section names:
19655 struct duart a __attribute__ ((section ("DUART_A"))) = { 0 };
19656 struct duart b __attribute__ ((section ("DUART_B"))) = { 0 };
19657 char stack[10000] __attribute__ ((section ("STACK"))) = { 0 };
19658 int init_data __attribute__ ((section ("INITDATA")));
19662 /* Initialize stack pointer */
19663 init_sp (stack + sizeof (stack));
19665 /* Initialize initialized data */
19666 memcpy (&init_data, &data, &edata - &data);
19668 /* Turn on the serial ports */
19673 Use the `section' attribute with _global_ variables and not
19674 _local_ variables, as shown in the example.
19676 You may use the `section' attribute with initialized or
19677 uninitialized global variables but the linker requires each object
19678 be defined once, with the exception that uninitialized variables
19679 tentatively go in the `common' (or `bss') section and can be
19680 multiply "defined". Using the `section' attribute will change
19681 what section the variable goes into and may cause the linker to
19682 issue an error if an uninitialized variable has multiple
19683 definitions. You can force a variable to be initialized with the
19684 `-fno-common' flag or the `nocommon' attribute.
19686 Some file formats do not support arbitrary sections so the
19687 `section' attribute is not available on all platforms. If you
19688 need to map the entire contents of a module to a particular
19689 section, consider using the facilities of the linker instead.
19692 On Microsoft Windows, in addition to putting variable definitions
19693 in a named section, the section can also be shared among all
19694 running copies of an executable or DLL. For example, this small
19695 program defines shared data by putting it in a named section
19696 `shared' and marking the section shareable:
19698 int foo __attribute__((section ("shared"), shared)) = 0;
19703 /* Read and write foo. All running
19704 copies see the same value. */
19708 You may only use the `shared' attribute along with `section'
19709 attribute with a fully initialized global definition because of
19710 the way linkers work. See `section' attribute for more
19713 The `shared' attribute is only available on Microsoft Windows.
19715 `tls_model ("TLS_MODEL")'
19716 The `tls_model' attribute sets thread-local storage model (*note
19717 Thread-Local::) of a particular `__thread' variable, overriding
19718 `-ftls-model=' command line switch on a per-variable basis. The
19719 TLS_MODEL argument should be one of `global-dynamic',
19720 `local-dynamic', `initial-exec' or `local-exec'.
19722 Not all targets support this attribute.
19725 This attribute, attached to a variable, means that the variable is
19726 meant to be possibly unused. GCC will not produce a warning for
19730 This attribute, attached to a variable, means that the variable
19731 must be emitted even if it appears that the variable is not
19734 `vector_size (BYTES)'
19735 This attribute specifies the vector size for the variable,
19736 measured in bytes. For example, the declaration:
19738 int foo __attribute__ ((vector_size (16)));
19740 causes the compiler to set the mode for `foo', to be 16 bytes,
19741 divided into `int' sized units. Assuming a 32-bit int (a vector of
19742 4 units of 4 bytes), the corresponding mode of `foo' will be V4SI.
19744 This attribute is only applicable to integral and float scalars,
19745 although arrays, pointers, and function return values are allowed
19746 in conjunction with this construct.
19748 Aggregates with this attribute are invalid, even if they are of
19749 the same size as a corresponding scalar. For example, the
19752 struct S { int a; };
19753 struct S __attribute__ ((vector_size (16))) foo;
19755 is invalid even if the size of the structure is the same as the
19759 The `selectany' attribute causes an initialized global variable to
19760 have link-once semantics. When multiple definitions of the
19761 variable are encountered by the linker, the first is selected and
19762 the remainder are discarded. Following usage by the Microsoft
19763 compiler, the linker is told _not_ to warn about size or content
19764 differences of the multiple definitions.
19766 Although the primary usage of this attribute is for POD types, the
19767 attribute can also be applied to global C++ objects that are
19768 initialized by a constructor. In this case, the static
19769 initialization and destruction code for the object is emitted in
19770 each translation defining the object, but the calls to the
19771 constructor and destructor are protected by a link-once guard
19774 The `selectany' attribute is only available on Microsoft Windows
19775 targets. You can use `__declspec (selectany)' as a synonym for
19776 `__attribute__ ((selectany))' for compatibility with other
19780 The `weak' attribute is described in *note Function Attributes::.
19783 The `dllimport' attribute is described in *note Function
19787 The `dllexport' attribute is described in *note Function
19791 5.34.1 Blackfin Variable Attributes
19792 -----------------------------------
19794 Three attributes are currently defined for the Blackfin.
19801 Use these attributes on the Blackfin to place the variable into L1
19802 Data SRAM. Variables with `l1_data' attribute will be put into
19803 the specific section named `.l1.data'. Those with `l1_data_A'
19804 attribute will be put into the specific section named
19805 `.l1.data.A'. Those with `l1_data_B' attribute will be put into
19806 the specific section named `.l1.data.B'.
19808 5.34.2 M32R/D Variable Attributes
19809 ---------------------------------
19811 One attribute is currently defined for the M32R/D.
19813 `model (MODEL-NAME)'
19814 Use this attribute on the M32R/D to set the addressability of an
19815 object. The identifier MODEL-NAME is one of `small', `medium', or
19816 `large', representing each of the code models.
19818 Small model objects live in the lower 16MB of memory (so that their
19819 addresses can be loaded with the `ld24' instruction).
19821 Medium and large model objects may live anywhere in the 32-bit
19822 address space (the compiler will generate `seth/add3' instructions
19823 to load their addresses).
19825 5.34.3 i386 Variable Attributes
19826 -------------------------------
19828 Two attributes are currently defined for i386 configurations:
19829 `ms_struct' and `gcc_struct'
19833 If `packed' is used on a structure, or if bit-fields are used it
19834 may be that the Microsoft ABI packs them differently than GCC
19835 would normally pack them. Particularly when moving packed data
19836 between functions compiled with GCC and the native Microsoft
19837 compiler (either via function call or as data in a file), it may
19838 be necessary to access either format.
19840 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
19841 Windows X86 compilers to match the native Microsoft compiler.
19843 The Microsoft structure layout algorithm is fairly simple with the
19844 exception of the bitfield packing:
19846 The padding and alignment of members of structures and whether a
19847 bit field can straddle a storage-unit boundary
19849 1. Structure members are stored sequentially in the order in
19850 which they are declared: the first member has the lowest
19851 memory address and the last member the highest.
19853 2. Every data object has an alignment-requirement. The
19854 alignment-requirement for all data except structures, unions,
19855 and arrays is either the size of the object or the current
19856 packing size (specified with either the aligned attribute or
19857 the pack pragma), whichever is less. For structures, unions,
19858 and arrays, the alignment-requirement is the largest
19859 alignment-requirement of its members. Every object is
19860 allocated an offset so that:
19862 offset % alignment-requirement == 0
19864 3. Adjacent bit fields are packed into the same 1-, 2-, or
19865 4-byte allocation unit if the integral types are the same
19866 size and if the next bit field fits into the current
19867 allocation unit without crossing the boundary imposed by the
19868 common alignment requirements of the bit fields.
19870 Handling of zero-length bitfields:
19872 MSVC interprets zero-length bitfields in the following ways:
19874 1. If a zero-length bitfield is inserted between two bitfields
19875 that would normally be coalesced, the bitfields will not be
19882 unsigned long bf_1 : 12;
19884 unsigned long bf_2 : 12;
19887 The size of `t1' would be 8 bytes with the zero-length
19888 bitfield. If the zero-length bitfield were removed, `t1''s
19889 size would be 4 bytes.
19891 2. If a zero-length bitfield is inserted after a bitfield,
19892 `foo', and the alignment of the zero-length bitfield is
19893 greater than the member that follows it, `bar', `bar' will be
19894 aligned as the type of the zero-length bitfield.
19912 For `t2', `bar' will be placed at offset 2, rather than
19913 offset 1. Accordingly, the size of `t2' will be 4. For
19914 `t3', the zero-length bitfield will not affect the alignment
19915 of `bar' or, as a result, the size of the structure.
19917 Taking this into account, it is important to note the
19920 1. If a zero-length bitfield follows a normal bitfield, the
19921 type of the zero-length bitfield may affect the
19922 alignment of the structure as whole. For example, `t2'
19923 has a size of 4 bytes, since the zero-length bitfield
19924 follows a normal bitfield, and is of type short.
19926 2. Even if a zero-length bitfield is not followed by a
19927 normal bitfield, it may still affect the alignment of
19936 Here, `t4' will take up 4 bytes.
19938 3. Zero-length bitfields following non-bitfield members are
19948 Here, `t5' will take up 2 bytes.
19950 5.34.4 PowerPC Variable Attributes
19951 ----------------------------------
19953 Three attributes currently are defined for PowerPC configurations:
19954 `altivec', `ms_struct' and `gcc_struct'.
19956 For full documentation of the struct attributes please see the
19957 documentation in *note i386 Variable Attributes::.
19959 For documentation of `altivec' attribute please see the documentation
19960 in *note PowerPC Type Attributes::.
19962 5.34.5 SPU Variable Attributes
19963 ------------------------------
19965 The SPU supports the `spu_vector' attribute for variables. For
19966 documentation of this attribute please see the documentation in *note
19967 SPU Type Attributes::.
19969 5.34.6 Xstormy16 Variable Attributes
19970 ------------------------------------
19972 One attribute is currently defined for xstormy16 configurations:
19976 If a variable has the `below100' attribute (`BELOW100' is allowed
19977 also), GCC will place the variable in the first 0x100 bytes of
19978 memory and use special opcodes to access it. Such variables will
19979 be placed in either the `.bss_below100' section or the
19980 `.data_below100' section.
19983 5.34.7 AVR Variable Attributes
19984 ------------------------------
19987 The `progmem' attribute is used on the AVR to place data in the
19988 Program Memory address space. The AVR is a Harvard Architecture
19989 processor and data normally resides in the Data Memory address
19993 File: gcc.info, Node: Type Attributes, Next: Alignment, Prev: Variable Attributes, Up: C Extensions
19995 5.35 Specifying Attributes of Types
19996 ===================================
19998 The keyword `__attribute__' allows you to specify special attributes of
19999 `struct' and `union' types when you define such types. This keyword is
20000 followed by an attribute specification inside double parentheses.
20001 Seven attributes are currently defined for types: `aligned', `packed',
20002 `transparent_union', `unused', `deprecated', `visibility', and
20003 `may_alias'. Other attributes are defined for functions (*note
20004 Function Attributes::) and for variables (*note Variable Attributes::).
20006 You may also specify any one of these attributes with `__' preceding
20007 and following its keyword. This allows you to use these attributes in
20008 header files without being concerned about a possible macro of the same
20009 name. For example, you may use `__aligned__' instead of `aligned'.
20011 You may specify type attributes in an enum, struct or union type
20012 declaration or definition, or for other types in a `typedef'
20015 For an enum, struct or union type, you may specify attributes either
20016 between the enum, struct or union tag and the name of the type, or just
20017 past the closing curly brace of the _definition_. The former syntax is
20020 *Note Attribute Syntax::, for details of the exact syntax for using
20023 `aligned (ALIGNMENT)'
20024 This attribute specifies a minimum alignment (in bytes) for
20025 variables of the specified type. For example, the declarations:
20027 struct S { short f[3]; } __attribute__ ((aligned (8)));
20028 typedef int more_aligned_int __attribute__ ((aligned (8)));
20030 force the compiler to insure (as far as it can) that each variable
20031 whose type is `struct S' or `more_aligned_int' will be allocated
20032 and aligned _at least_ on a 8-byte boundary. On a SPARC, having
20033 all variables of type `struct S' aligned to 8-byte boundaries
20034 allows the compiler to use the `ldd' and `std' (doubleword load and
20035 store) instructions when copying one variable of type `struct S' to
20036 another, thus improving run-time efficiency.
20038 Note that the alignment of any given `struct' or `union' type is
20039 required by the ISO C standard to be at least a perfect multiple of
20040 the lowest common multiple of the alignments of all of the members
20041 of the `struct' or `union' in question. This means that you _can_
20042 effectively adjust the alignment of a `struct' or `union' type by
20043 attaching an `aligned' attribute to any one of the members of such
20044 a type, but the notation illustrated in the example above is a
20045 more obvious, intuitive, and readable way to request the compiler
20046 to adjust the alignment of an entire `struct' or `union' type.
20048 As in the preceding example, you can explicitly specify the
20049 alignment (in bytes) that you wish the compiler to use for a given
20050 `struct' or `union' type. Alternatively, you can leave out the
20051 alignment factor and just ask the compiler to align a type to the
20052 maximum useful alignment for the target machine you are compiling
20053 for. For example, you could write:
20055 struct S { short f[3]; } __attribute__ ((aligned));
20057 Whenever you leave out the alignment factor in an `aligned'
20058 attribute specification, the compiler automatically sets the
20059 alignment for the type to the largest alignment which is ever used
20060 for any data type on the target machine you are compiling for.
20061 Doing this can often make copy operations more efficient, because
20062 the compiler can use whatever instructions copy the biggest chunks
20063 of memory when performing copies to or from the variables which
20064 have types that you have aligned this way.
20066 In the example above, if the size of each `short' is 2 bytes, then
20067 the size of the entire `struct S' type is 6 bytes. The smallest
20068 power of two which is greater than or equal to that is 8, so the
20069 compiler sets the alignment for the entire `struct S' type to 8
20072 Note that although you can ask the compiler to select a
20073 time-efficient alignment for a given type and then declare only
20074 individual stand-alone objects of that type, the compiler's
20075 ability to select a time-efficient alignment is primarily useful
20076 only when you plan to create arrays of variables having the
20077 relevant (efficiently aligned) type. If you declare or use arrays
20078 of variables of an efficiently-aligned type, then it is likely
20079 that your program will also be doing pointer arithmetic (or
20080 subscripting, which amounts to the same thing) on pointers to the
20081 relevant type, and the code that the compiler generates for these
20082 pointer arithmetic operations will often be more efficient for
20083 efficiently-aligned types than for other types.
20085 The `aligned' attribute can only increase the alignment; but you
20086 can decrease it by specifying `packed' as well. See below.
20088 Note that the effectiveness of `aligned' attributes may be limited
20089 by inherent limitations in your linker. On many systems, the
20090 linker is only able to arrange for variables to be aligned up to a
20091 certain maximum alignment. (For some linkers, the maximum
20092 supported alignment may be very very small.) If your linker is
20093 only able to align variables up to a maximum of 8 byte alignment,
20094 then specifying `aligned(16)' in an `__attribute__' will still
20095 only provide you with 8 byte alignment. See your linker
20096 documentation for further information.
20099 This attribute, attached to `struct' or `union' type definition,
20100 specifies that each member (other than zero-width bitfields) of
20101 the structure or union is placed to minimize the memory required.
20102 When attached to an `enum' definition, it indicates that the
20103 smallest integral type should be used.
20105 Specifying this attribute for `struct' and `union' types is
20106 equivalent to specifying the `packed' attribute on each of the
20107 structure or union members. Specifying the `-fshort-enums' flag
20108 on the line is equivalent to specifying the `packed' attribute on
20109 all `enum' definitions.
20111 In the following example `struct my_packed_struct''s members are
20112 packed closely together, but the internal layout of its `s' member
20113 is not packed--to do that, `struct my_unpacked_struct' would need
20116 struct my_unpacked_struct
20122 struct __attribute__ ((__packed__)) my_packed_struct
20126 struct my_unpacked_struct s;
20129 You may only specify this attribute on the definition of a `enum',
20130 `struct' or `union', not on a `typedef' which does not also define
20131 the enumerated type, structure or union.
20133 `transparent_union'
20134 This attribute, attached to a `union' type definition, indicates
20135 that any function parameter having that union type causes calls to
20136 that function to be treated in a special way.
20138 First, the argument corresponding to a transparent union type can
20139 be of any type in the union; no cast is required. Also, if the
20140 union contains a pointer type, the corresponding argument can be a
20141 null pointer constant or a void pointer expression; and if the
20142 union contains a void pointer type, the corresponding argument can
20143 be any pointer expression. If the union member type is a pointer,
20144 qualifiers like `const' on the referenced type must be respected,
20145 just as with normal pointer conversions.
20147 Second, the argument is passed to the function using the calling
20148 conventions of the first member of the transparent union, not the
20149 calling conventions of the union itself. All members of the union
20150 must have the same machine representation; this is necessary for
20151 this argument passing to work properly.
20153 Transparent unions are designed for library functions that have
20154 multiple interfaces for compatibility reasons. For example,
20155 suppose the `wait' function must accept either a value of type
20156 `int *' to comply with Posix, or a value of type `union wait *' to
20157 comply with the 4.1BSD interface. If `wait''s parameter were
20158 `void *', `wait' would accept both kinds of arguments, but it
20159 would also accept any other pointer type and this would make
20160 argument type checking less useful. Instead, `<sys/wait.h>' might
20161 define the interface as follows:
20163 typedef union __attribute__ ((__transparent_union__))
20167 } wait_status_ptr_t;
20169 pid_t wait (wait_status_ptr_t);
20171 This interface allows either `int *' or `union wait *' arguments
20172 to be passed, using the `int *' calling convention. The program
20173 can call `wait' with arguments of either type:
20175 int w1 () { int w; return wait (&w); }
20176 int w2 () { union wait w; return wait (&w); }
20178 With this interface, `wait''s implementation might look like this:
20180 pid_t wait (wait_status_ptr_t p)
20182 return waitpid (-1, p.__ip, 0);
20186 When attached to a type (including a `union' or a `struct'), this
20187 attribute means that variables of that type are meant to appear
20188 possibly unused. GCC will not produce a warning for any variables
20189 of that type, even if the variable appears to do nothing. This is
20190 often the case with lock or thread classes, which are usually
20191 defined and then not referenced, but contain constructors and
20192 destructors that have nontrivial bookkeeping functions.
20195 The `deprecated' attribute results in a warning if the type is
20196 used anywhere in the source file. This is useful when identifying
20197 types that are expected to be removed in a future version of a
20198 program. If possible, the warning also includes the location of
20199 the declaration of the deprecated type, to enable users to easily
20200 find further information about why the type is deprecated, or what
20201 they should do instead. Note that the warnings only occur for
20202 uses and then only if the type is being applied to an identifier
20203 that itself is not being declared as deprecated.
20205 typedef int T1 __attribute__ ((deprecated));
20209 typedef T1 T3 __attribute__ ((deprecated));
20210 T3 z __attribute__ ((deprecated));
20212 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
20213 warning is issued for line 4 because T2 is not explicitly
20214 deprecated. Line 5 has no warning because T3 is explicitly
20215 deprecated. Similarly for line 6.
20217 The `deprecated' attribute can also be used for functions and
20218 variables (*note Function Attributes::, *note Variable
20222 Accesses through pointers to types with this attribute are not
20223 subject to type-based alias analysis, but are instead assumed to
20224 be able to alias any other type of objects. In the context of
20225 6.5/7 an lvalue expression dereferencing such a pointer is treated
20226 like having a character type. See `-fstrict-aliasing' for more
20227 information on aliasing issues. This extension exists to support
20228 some vector APIs, in which pointers to one vector type are
20229 permitted to alias pointers to a different vector type.
20231 Note that an object of a type with this attribute does not have any
20236 typedef short __attribute__((__may_alias__)) short_a;
20241 int a = 0x12345678;
20242 short_a *b = (short_a *) &a;
20246 if (a == 0x12345678)
20252 If you replaced `short_a' with `short' in the variable
20253 declaration, the above program would abort when compiled with
20254 `-fstrict-aliasing', which is on by default at `-O2' or above in
20255 recent GCC versions.
20258 In C++, attribute visibility (*note Function Attributes::) can
20259 also be applied to class, struct, union and enum types. Unlike
20260 other type attributes, the attribute must appear between the
20261 initial keyword and the name of the type; it cannot appear after
20262 the body of the type.
20264 Note that the type visibility is applied to vague linkage entities
20265 associated with the class (vtable, typeinfo node, etc.). In
20266 particular, if a class is thrown as an exception in one shared
20267 object and caught in another, the class must have default
20268 visibility. Otherwise the two shared objects will be unable to
20269 use the same typeinfo node and exception handling will break.
20272 5.35.1 ARM Type Attributes
20273 --------------------------
20275 On those ARM targets that support `dllimport' (such as Symbian OS), you
20276 can use the `notshared' attribute to indicate that the virtual table
20277 and other similar data for a class should not be exported from a DLL.
20280 class __declspec(notshared) C {
20282 __declspec(dllimport) C();
20286 __declspec(dllexport)
20289 In this code, `C::C' is exported from the current DLL, but the virtual
20290 table for `C' is not exported. (You can use `__attribute__' instead of
20291 `__declspec' if you prefer, but most Symbian OS code uses `__declspec'.)
20293 5.35.2 i386 Type Attributes
20294 ---------------------------
20296 Two attributes are currently defined for i386 configurations:
20297 `ms_struct' and `gcc_struct'.
20301 If `packed' is used on a structure, or if bit-fields are used it
20302 may be that the Microsoft ABI packs them differently than GCC
20303 would normally pack them. Particularly when moving packed data
20304 between functions compiled with GCC and the native Microsoft
20305 compiler (either via function call or as data in a file), it may
20306 be necessary to access either format.
20308 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
20309 Windows X86 compilers to match the native Microsoft compiler.
20311 To specify multiple attributes, separate them by commas within the
20312 double parentheses: for example, `__attribute__ ((aligned (16),
20315 5.35.3 PowerPC Type Attributes
20316 ------------------------------
20318 Three attributes currently are defined for PowerPC configurations:
20319 `altivec', `ms_struct' and `gcc_struct'.
20321 For full documentation of the `ms_struct' and `gcc_struct' attributes
20322 please see the documentation in *note i386 Type Attributes::.
20324 The `altivec' attribute allows one to declare AltiVec vector data
20325 types supported by the AltiVec Programming Interface Manual. The
20326 attribute requires an argument to specify one of three vector types:
20327 `vector__', `pixel__' (always followed by unsigned short), and `bool__'
20328 (always followed by unsigned).
20330 __attribute__((altivec(vector__)))
20331 __attribute__((altivec(pixel__))) unsigned short
20332 __attribute__((altivec(bool__))) unsigned
20334 These attributes mainly are intended to support the `__vector',
20335 `__pixel', and `__bool' AltiVec keywords.
20337 5.35.4 SPU Type Attributes
20338 --------------------------
20340 The SPU supports the `spu_vector' attribute for types. This attribute
20341 allows one to declare vector data types supported by the
20342 Sony/Toshiba/IBM SPU Language Extensions Specification. It is intended
20343 to support the `__vector' keyword.
20346 File: gcc.info, Node: Inline, Next: Extended Asm, Prev: Alignment, Up: C Extensions
20348 5.36 An Inline Function is As Fast As a Macro
20349 =============================================
20351 By declaring a function inline, you can direct GCC to make calls to
20352 that function faster. One way GCC can achieve this is to integrate
20353 that function's code into the code for its callers. This makes
20354 execution faster by eliminating the function-call overhead; in
20355 addition, if any of the actual argument values are constant, their
20356 known values may permit simplifications at compile time so that not all
20357 of the inline function's code needs to be included. The effect on code
20358 size is less predictable; object code may be larger or smaller with
20359 function inlining, depending on the particular case. You can also
20360 direct GCC to try to integrate all "simple enough" functions into their
20361 callers with the option `-finline-functions'.
20363 GCC implements three different semantics of declaring a function
20364 inline. One is available with `-std=gnu89' or `-fgnu89-inline' or when
20365 `gnu_inline' attribute is present on all inline declarations, another
20366 when `-std=c99' or `-std=gnu99' (without `-fgnu89-inline'), and the
20367 third is used when compiling C++.
20369 To declare a function inline, use the `inline' keyword in its
20370 declaration, like this:
20378 If you are writing a header file to be included in ISO C89 programs,
20379 write `__inline__' instead of `inline'. *Note Alternate Keywords::.
20381 The three types of inlining behave similarly in two important cases:
20382 when the `inline' keyword is used on a `static' function, like the
20383 example above, and when a function is first declared without using the
20384 `inline' keyword and then is defined with `inline', like this:
20386 extern int inc (int *a);
20393 In both of these common cases, the program behaves the same as if you
20394 had not used the `inline' keyword, except for its speed.
20396 When a function is both inline and `static', if all calls to the
20397 function are integrated into the caller, and the function's address is
20398 never used, then the function's own assembler code is never referenced.
20399 In this case, GCC does not actually output assembler code for the
20400 function, unless you specify the option `-fkeep-inline-functions'.
20401 Some calls cannot be integrated for various reasons (in particular,
20402 calls that precede the function's definition cannot be integrated, and
20403 neither can recursive calls within the definition). If there is a
20404 nonintegrated call, then the function is compiled to assembler code as
20405 usual. The function must also be compiled as usual if the program
20406 refers to its address, because that can't be inlined.
20408 Note that certain usages in a function definition can make it
20409 unsuitable for inline substitution. Among these usages are: use of
20410 varargs, use of alloca, use of variable sized data types (*note
20411 Variable Length::), use of computed goto (*note Labels as Values::),
20412 use of nonlocal goto, and nested functions (*note Nested Functions::).
20413 Using `-Winline' will warn when a function marked `inline' could not be
20414 substituted, and will give the reason for the failure.
20416 As required by ISO C++, GCC considers member functions defined within
20417 the body of a class to be marked inline even if they are not explicitly
20418 declared with the `inline' keyword. You can override this with
20419 `-fno-default-inline'; *note Options Controlling C++ Dialect: C++
20422 GCC does not inline any functions when not optimizing unless you
20423 specify the `always_inline' attribute for the function, like this:
20426 inline void foo (const char) __attribute__((always_inline));
20428 The remainder of this section is specific to GNU C89 inlining.
20430 When an inline function is not `static', then the compiler must assume
20431 that there may be calls from other source files; since a global symbol
20432 can be defined only once in any program, the function must not be
20433 defined in the other source files, so the calls therein cannot be
20434 integrated. Therefore, a non-`static' inline function is always
20435 compiled on its own in the usual fashion.
20437 If you specify both `inline' and `extern' in the function definition,
20438 then the definition is used only for inlining. In no case is the
20439 function compiled on its own, not even if you refer to its address
20440 explicitly. Such an address becomes an external reference, as if you
20441 had only declared the function, and had not defined it.
20443 This combination of `inline' and `extern' has almost the effect of a
20444 macro. The way to use it is to put a function definition in a header
20445 file with these keywords, and put another copy of the definition
20446 (lacking `inline' and `extern') in a library file. The definition in
20447 the header file will cause most calls to the function to be inlined.
20448 If any uses of the function remain, they will refer to the single copy
20452 File: gcc.info, Node: Extended Asm, Next: Constraints, Prev: Inline, Up: C Extensions
20454 5.37 Assembler Instructions with C Expression Operands
20455 ======================================================
20457 In an assembler instruction using `asm', you can specify the operands
20458 of the instruction using C expressions. This means you need not guess
20459 which registers or memory locations will contain the data you want to
20462 You must specify an assembler instruction template much like what
20463 appears in a machine description, plus an operand constraint string for
20466 For example, here is how to use the 68881's `fsinx' instruction:
20468 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
20470 Here `angle' is the C expression for the input operand while `result'
20471 is that of the output operand. Each has `"f"' as its operand
20472 constraint, saying that a floating point register is required. The `='
20473 in `=f' indicates that the operand is an output; all output operands'
20474 constraints must use `='. The constraints use the same language used
20475 in the machine description (*note Constraints::).
20477 Each operand is described by an operand-constraint string followed by
20478 the C expression in parentheses. A colon separates the assembler
20479 template from the first output operand and another separates the last
20480 output operand from the first input, if any. Commas separate the
20481 operands within each group. The total number of operands is currently
20482 limited to 30; this limitation may be lifted in some future version of
20485 If there are no output operands but there are input operands, you must
20486 place two consecutive colons surrounding the place where the output
20489 As of GCC version 3.1, it is also possible to specify input and output
20490 operands using symbolic names which can be referenced within the
20491 assembler code. These names are specified inside square brackets
20492 preceding the constraint string, and can be referenced inside the
20493 assembler code using `%[NAME]' instead of a percentage sign followed by
20494 the operand number. Using named operands the above example could look
20497 asm ("fsinx %[angle],%[output]"
20498 : [output] "=f" (result)
20499 : [angle] "f" (angle));
20501 Note that the symbolic operand names have no relation whatsoever to
20502 other C identifiers. You may use any name you like, even those of
20503 existing C symbols, but you must ensure that no two operands within the
20504 same assembler construct use the same symbolic name.
20506 Output operand expressions must be lvalues; the compiler can check
20507 this. The input operands need not be lvalues. The compiler cannot
20508 check whether the operands have data types that are reasonable for the
20509 instruction being executed. It does not parse the assembler instruction
20510 template and does not know what it means or even whether it is valid
20511 assembler input. The extended `asm' feature is most often used for
20512 machine instructions the compiler itself does not know exist. If the
20513 output expression cannot be directly addressed (for example, it is a
20514 bit-field), your constraint must allow a register. In that case, GCC
20515 will use the register as the output of the `asm', and then store that
20516 register into the output.
20518 The ordinary output operands must be write-only; GCC will assume that
20519 the values in these operands before the instruction are dead and need
20520 not be generated. Extended asm supports input-output or read-write
20521 operands. Use the constraint character `+' to indicate such an operand
20522 and list it with the output operands. You should only use read-write
20523 operands when the constraints for the operand (or the operand in which
20524 only some of the bits are to be changed) allow a register.
20526 You may, as an alternative, logically split its function into two
20527 separate operands, one input operand and one write-only output operand.
20528 The connection between them is expressed by constraints which say they
20529 need to be in the same location when the instruction executes. You can
20530 use the same C expression for both operands, or different expressions.
20531 For example, here we write the (fictitious) `combine' instruction with
20532 `bar' as its read-only source operand and `foo' as its read-write
20535 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
20537 The constraint `"0"' for operand 1 says that it must occupy the same
20538 location as operand 0. A number in constraint is allowed only in an
20539 input operand and it must refer to an output operand.
20541 Only a number in the constraint can guarantee that one operand will be
20542 in the same place as another. The mere fact that `foo' is the value of
20543 both operands is not enough to guarantee that they will be in the same
20544 place in the generated assembler code. The following would not work
20547 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
20549 Various optimizations or reloading could cause operands 0 and 1 to be
20550 in different registers; GCC knows no reason not to do so. For example,
20551 the compiler might find a copy of the value of `foo' in one register and
20552 use it for operand 1, but generate the output operand 0 in a different
20553 register (copying it afterward to `foo''s own address). Of course,
20554 since the register for operand 1 is not even mentioned in the assembler
20555 code, the result will not work, but GCC can't tell that.
20557 As of GCC version 3.1, one may write `[NAME]' instead of the operand
20558 number for a matching constraint. For example:
20560 asm ("cmoveq %1,%2,%[result]"
20561 : [result] "=r"(result)
20562 : "r" (test), "r"(new), "[result]"(old));
20564 Sometimes you need to make an `asm' operand be a specific register,
20565 but there's no matching constraint letter for that register _by
20566 itself_. To force the operand into that register, use a local variable
20567 for the operand and specify the register in the variable declaration.
20568 *Note Explicit Reg Vars::. Then for the `asm' operand, use any
20569 register constraint letter that matches the register:
20571 register int *p1 asm ("r0") = ...;
20572 register int *p2 asm ("r1") = ...;
20573 register int *result asm ("r0");
20574 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
20576 In the above example, beware that a register that is call-clobbered by
20577 the target ABI will be overwritten by any function call in the
20578 assignment, including library calls for arithmetic operators. Also a
20579 register may be clobbered when generating some operations, like
20580 variable shift, memory copy or memory move on x86. Assuming it is a
20581 call-clobbered register, this may happen to `r0' above by the
20582 assignment to `p2'. If you have to use such a register, use temporary
20583 variables for expressions between the register assignment and use:
20586 register int *p1 asm ("r0") = ...;
20587 register int *p2 asm ("r1") = t1;
20588 register int *result asm ("r0");
20589 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
20591 Some instructions clobber specific hard registers. To describe this,
20592 write a third colon after the input operands, followed by the names of
20593 the clobbered hard registers (given as strings). Here is a realistic
20594 example for the VAX:
20596 asm volatile ("movc3 %0,%1,%2"
20598 : "g" (from), "g" (to), "g" (count)
20599 : "r0", "r1", "r2", "r3", "r4", "r5");
20601 You may not write a clobber description in a way that overlaps with an
20602 input or output operand. For example, you may not have an operand
20603 describing a register class with one member if you mention that register
20604 in the clobber list. Variables declared to live in specific registers
20605 (*note Explicit Reg Vars::), and used as asm input or output operands
20606 must have no part mentioned in the clobber description. There is no
20607 way for you to specify that an input operand is modified without also
20608 specifying it as an output operand. Note that if all the output
20609 operands you specify are for this purpose (and hence unused), you will
20610 then also need to specify `volatile' for the `asm' construct, as
20611 described below, to prevent GCC from deleting the `asm' statement as
20614 If you refer to a particular hardware register from the assembler code,
20615 you will probably have to list the register after the third colon to
20616 tell the compiler the register's value is modified. In some assemblers,
20617 the register names begin with `%'; to produce one `%' in the assembler
20618 code, you must write `%%' in the input.
20620 If your assembler instruction can alter the condition code register,
20621 add `cc' to the list of clobbered registers. GCC on some machines
20622 represents the condition codes as a specific hardware register; `cc'
20623 serves to name this register. On other machines, the condition code is
20624 handled differently, and specifying `cc' has no effect. But it is
20625 valid no matter what the machine.
20627 If your assembler instructions access memory in an unpredictable
20628 fashion, add `memory' to the list of clobbered registers. This will
20629 cause GCC to not keep memory values cached in registers across the
20630 assembler instruction and not optimize stores or loads to that memory.
20631 You will also want to add the `volatile' keyword if the memory affected
20632 is not listed in the inputs or outputs of the `asm', as the `memory'
20633 clobber does not count as a side-effect of the `asm'. If you know how
20634 large the accessed memory is, you can add it as input or output but if
20635 this is not known, you should add `memory'. As an example, if you
20636 access ten bytes of a string, you can use a memory input like:
20638 {"m"( ({ struct { char x[10]; } *p = (void *)ptr ; *p; }) )}.
20640 Note that in the following example the memory input is necessary,
20641 otherwise GCC might optimize the store to `x' away:
20647 asm ("magic stuff accessing an 'int' pointed to by '%1'"
20648 "=&d" (r) : "a" (y), "m" (*y));
20652 You can put multiple assembler instructions together in a single `asm'
20653 template, separated by the characters normally used in assembly code
20654 for the system. A combination that works in most places is a newline
20655 to break the line, plus a tab character to move to the instruction field
20656 (written as `\n\t'). Sometimes semicolons can be used, if the
20657 assembler allows semicolons as a line-breaking character. Note that
20658 some assembler dialects use semicolons to start a comment. The input
20659 operands are guaranteed not to use any of the clobbered registers, and
20660 neither will the output operands' addresses, so you can read and write
20661 the clobbered registers as many times as you like. Here is an example
20662 of multiple instructions in a template; it assumes the subroutine
20663 `_foo' accepts arguments in registers 9 and 10:
20665 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
20667 : "g" (from), "g" (to)
20670 Unless an output operand has the `&' constraint modifier, GCC may
20671 allocate it in the same register as an unrelated input operand, on the
20672 assumption the inputs are consumed before the outputs are produced.
20673 This assumption may be false if the assembler code actually consists of
20674 more than one instruction. In such a case, use `&' for each output
20675 operand that may not overlap an input. *Note Modifiers::.
20677 If you want to test the condition code produced by an assembler
20678 instruction, you must include a branch and a label in the `asm'
20679 construct, as follows:
20681 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
20685 This assumes your assembler supports local labels, as the GNU assembler
20686 and most Unix assemblers do.
20688 Speaking of labels, jumps from one `asm' to another are not supported.
20689 The compiler's optimizers do not know about these jumps, and therefore
20690 they cannot take account of them when deciding how to optimize.
20692 Usually the most convenient way to use these `asm' instructions is to
20693 encapsulate them in macros that look like functions. For example,
20696 ({ double __value, __arg = (x); \
20697 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
20700 Here the variable `__arg' is used to make sure that the instruction
20701 operates on a proper `double' value, and to accept only those arguments
20702 `x' which can convert automatically to a `double'.
20704 Another way to make sure the instruction operates on the correct data
20705 type is to use a cast in the `asm'. This is different from using a
20706 variable `__arg' in that it converts more different types. For
20707 example, if the desired type were `int', casting the argument to `int'
20708 would accept a pointer with no complaint, while assigning the argument
20709 to an `int' variable named `__arg' would warn about using a pointer
20710 unless the caller explicitly casts it.
20712 If an `asm' has output operands, GCC assumes for optimization purposes
20713 the instruction has no side effects except to change the output
20714 operands. This does not mean instructions with a side effect cannot be
20715 used, but you must be careful, because the compiler may eliminate them
20716 if the output operands aren't used, or move them out of loops, or
20717 replace two with one if they constitute a common subexpression. Also,
20718 if your instruction does have a side effect on a variable that otherwise
20719 appears not to change, the old value of the variable may be reused later
20720 if it happens to be found in a register.
20722 You can prevent an `asm' instruction from being deleted by writing the
20723 keyword `volatile' after the `asm'. For example:
20725 #define get_and_set_priority(new) \
20727 asm volatile ("get_and_set_priority %0, %1" \
20728 : "=g" (__old) : "g" (new)); \
20731 The `volatile' keyword indicates that the instruction has important
20732 side-effects. GCC will not delete a volatile `asm' if it is reachable.
20733 (The instruction can still be deleted if GCC can prove that
20734 control-flow will never reach the location of the instruction.) Note
20735 that even a volatile `asm' instruction can be moved relative to other
20736 code, including across jump instructions. For example, on many targets
20737 there is a system register which can be set to control the rounding
20738 mode of floating point operations. You might try setting it with a
20739 volatile `asm', like this PowerPC example:
20741 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
20744 This will not work reliably, as the compiler may move the addition back
20745 before the volatile `asm'. To make it work you need to add an
20746 artificial dependency to the `asm' referencing a variable in the code
20747 you don't want moved, for example:
20749 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
20752 Similarly, you can't expect a sequence of volatile `asm' instructions
20753 to remain perfectly consecutive. If you want consecutive output, use a
20754 single `asm'. Also, GCC will perform some optimizations across a
20755 volatile `asm' instruction; GCC does not "forget everything" when it
20756 encounters a volatile `asm' instruction the way some other compilers do.
20758 An `asm' instruction without any output operands will be treated
20759 identically to a volatile `asm' instruction.
20761 It is a natural idea to look for a way to give access to the condition
20762 code left by the assembler instruction. However, when we attempted to
20763 implement this, we found no way to make it work reliably. The problem
20764 is that output operands might need reloading, which would result in
20765 additional following "store" instructions. On most machines, these
20766 instructions would alter the condition code before there was time to
20767 test it. This problem doesn't arise for ordinary "test" and "compare"
20768 instructions because they don't have any output operands.
20770 For reasons similar to those described above, it is not possible to
20771 give an assembler instruction access to the condition code left by
20772 previous instructions.
20774 If you are writing a header file that should be includable in ISO C
20775 programs, write `__asm__' instead of `asm'. *Note Alternate Keywords::.
20777 5.37.1 Size of an `asm'
20778 -----------------------
20780 Some targets require that GCC track the size of each instruction used in
20781 order to generate correct code. Because the final length of an `asm'
20782 is only known by the assembler, GCC must make an estimate as to how big
20783 it will be. The estimate is formed by counting the number of
20784 statements in the pattern of the `asm' and multiplying that by the
20785 length of the longest instruction on that processor. Statements in the
20786 `asm' are identified by newline characters and whatever statement
20787 separator characters are supported by the assembler; on most processors
20788 this is the ``;'' character.
20790 Normally, GCC's estimate is perfectly adequate to ensure that correct
20791 code is generated, but it is possible to confuse the compiler if you use
20792 pseudo instructions or assembler macros that expand into multiple real
20793 instructions or if you use assembler directives that expand to more
20794 space in the object file than would be needed for a single instruction.
20795 If this happens then the assembler will produce a diagnostic saying that
20796 a label is unreachable.
20798 5.37.2 i386 floating point asm operands
20799 ---------------------------------------
20801 There are several rules on the usage of stack-like regs in asm_operands
20802 insns. These rules apply only to the operands that are stack-like regs:
20804 1. Given a set of input regs that die in an asm_operands, it is
20805 necessary to know which are implicitly popped by the asm, and
20806 which must be explicitly popped by gcc.
20808 An input reg that is implicitly popped by the asm must be
20809 explicitly clobbered, unless it is constrained to match an output
20812 2. For any input reg that is implicitly popped by an asm, it is
20813 necessary to know how to adjust the stack to compensate for the
20814 pop. If any non-popped input is closer to the top of the
20815 reg-stack than the implicitly popped reg, it would not be possible
20816 to know what the stack looked like--it's not clear how the rest of
20817 the stack "slides up".
20819 All implicitly popped input regs must be closer to the top of the
20820 reg-stack than any input that is not implicitly popped.
20822 It is possible that if an input dies in an insn, reload might use
20823 the input reg for an output reload. Consider this example:
20825 asm ("foo" : "=t" (a) : "f" (b));
20827 This asm says that input B is not popped by the asm, and that the
20828 asm pushes a result onto the reg-stack, i.e., the stack is one
20829 deeper after the asm than it was before. But, it is possible that
20830 reload will think that it can use the same reg for both the input
20831 and the output, if input B dies in this insn.
20833 If any input operand uses the `f' constraint, all output reg
20834 constraints must use the `&' earlyclobber.
20836 The asm above would be written as
20838 asm ("foo" : "=&t" (a) : "f" (b));
20840 3. Some operands need to be in particular places on the stack. All
20841 output operands fall in this category--there is no other way to
20842 know which regs the outputs appear in unless the user indicates
20843 this in the constraints.
20845 Output operands must specifically indicate which reg an output
20846 appears in after an asm. `=f' is not allowed: the operand
20847 constraints must select a class with a single reg.
20849 4. Output operands may not be "inserted" between existing stack regs.
20850 Since no 387 opcode uses a read/write operand, all output operands
20851 are dead before the asm_operands, and are pushed by the
20852 asm_operands. It makes no sense to push anywhere but the top of
20855 Output operands must start at the top of the reg-stack: output
20856 operands may not "skip" a reg.
20858 5. Some asm statements may need extra stack space for internal
20859 calculations. This can be guaranteed by clobbering stack registers
20860 unrelated to the inputs and outputs.
20863 Here are a couple of reasonable asms to want to write. This asm takes
20864 one input, which is internally popped, and produces two outputs.
20866 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
20868 This asm takes two inputs, which are popped by the `fyl2xp1' opcode,
20869 and replaces them with one output. The user must code the `st(1)'
20870 clobber for reg-stack.c to know that `fyl2xp1' pops both inputs.
20872 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
20875 File: gcc.info, Node: Constraints, Next: Asm Labels, Prev: Extended Asm, Up: C Extensions
20877 5.38 Constraints for `asm' Operands
20878 ===================================
20880 Here are specific details on what constraint letters you can use with
20881 `asm' operands. Constraints can say whether an operand may be in a
20882 register, and which kinds of register; whether the operand can be a
20883 memory reference, and which kinds of address; whether the operand may
20884 be an immediate constant, and which possible values it may have.
20885 Constraints can also require two operands to match.
20889 * Simple Constraints:: Basic use of constraints.
20890 * Multi-Alternative:: When an insn has two alternative constraint-patterns.
20891 * Modifiers:: More precise control over effects of constraints.
20892 * Machine Constraints:: Special constraints for some particular machines.
20895 File: gcc.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints
20897 5.38.1 Simple Constraints
20898 -------------------------
20900 The simplest kind of constraint is a string full of letters, each of
20901 which describes one kind of operand that is permitted. Here are the
20902 letters that are allowed:
20905 Whitespace characters are ignored and can be inserted at any
20906 position except the first. This enables each alternative for
20907 different operands to be visually aligned in the machine
20908 description even if they have different number of constraints and
20912 A memory operand is allowed, with any kind of address that the
20913 machine supports in general. Note that the letter used for the
20914 general memory constraint can be re-defined by a back end using
20915 the `TARGET_MEM_CONSTRAINT' macro.
20918 A memory operand is allowed, but only if the address is
20919 "offsettable". This means that adding a small integer (actually,
20920 the width in bytes of the operand, as determined by its machine
20921 mode) may be added to the address and the result is also a valid
20924 For example, an address which is constant is offsettable; so is an
20925 address that is the sum of a register and a constant (as long as a
20926 slightly larger constant is also within the range of
20927 address-offsets supported by the machine); but an autoincrement or
20928 autodecrement address is not offsettable. More complicated
20929 indirect/indexed addresses may or may not be offsettable depending
20930 on the other addressing modes that the machine supports.
20932 Note that in an output operand which can be matched by another
20933 operand, the constraint letter `o' is valid only when accompanied
20934 by both `<' (if the target machine has predecrement addressing)
20935 and `>' (if the target machine has preincrement addressing).
20938 A memory operand that is not offsettable. In other words,
20939 anything that would fit the `m' constraint but not the `o'
20943 A memory operand with autodecrement addressing (either
20944 predecrement or postdecrement) is allowed.
20947 A memory operand with autoincrement addressing (either
20948 preincrement or postincrement) is allowed.
20951 A register operand is allowed provided that it is in a general
20955 An immediate integer operand (one with constant value) is allowed.
20956 This includes symbolic constants whose values will be known only at
20957 assembly time or later.
20960 An immediate integer operand with a known numeric value is allowed.
20961 Many systems cannot support assembly-time constants for operands
20962 less than a word wide. Constraints for these operands should use
20963 `n' rather than `i'.
20965 `I', `J', `K', ... `P'
20966 Other letters in the range `I' through `P' may be defined in a
20967 machine-dependent fashion to permit immediate integer operands with
20968 explicit integer values in specified ranges. For example, on the
20969 68000, `I' is defined to stand for the range of values 1 to 8.
20970 This is the range permitted as a shift count in the shift
20974 An immediate floating operand (expression code `const_double') is
20975 allowed, but only if the target floating point format is the same
20976 as that of the host machine (on which the compiler is running).
20979 An immediate floating operand (expression code `const_double' or
20980 `const_vector') is allowed.
20983 `G' and `H' may be defined in a machine-dependent fashion to
20984 permit immediate floating operands in particular ranges of values.
20987 An immediate integer operand whose value is not an explicit
20988 integer is allowed.
20990 This might appear strange; if an insn allows a constant operand
20991 with a value not known at compile time, it certainly must allow
20992 any known value. So why use `s' instead of `i'? Sometimes it
20993 allows better code to be generated.
20995 For example, on the 68000 in a fullword instruction it is possible
20996 to use an immediate operand; but if the immediate value is between
20997 -128 and 127, better code results from loading the value into a
20998 register and using the register. This is because the load into
20999 the register can be done with a `moveq' instruction. We arrange
21000 for this to happen by defining the letter `K' to mean "any integer
21001 outside the range -128 to 127", and then specifying `Ks' in the
21002 operand constraints.
21005 Any register, memory or immediate integer operand is allowed,
21006 except for registers that are not general registers.
21009 Any operand whatsoever is allowed.
21011 `0', `1', `2', ... `9'
21012 An operand that matches the specified operand number is allowed.
21013 If a digit is used together with letters within the same
21014 alternative, the digit should come last.
21016 This number is allowed to be more than a single digit. If multiple
21017 digits are encountered consecutively, they are interpreted as a
21018 single decimal integer. There is scant chance for ambiguity,
21019 since to-date it has never been desirable that `10' be interpreted
21020 as matching either operand 1 _or_ operand 0. Should this be
21021 desired, one can use multiple alternatives instead.
21023 This is called a "matching constraint" and what it really means is
21024 that the assembler has only a single operand that fills two roles
21025 which `asm' distinguishes. For example, an add instruction uses
21026 two input operands and an output operand, but on most CISC
21027 machines an add instruction really has only two operands, one of
21028 them an input-output operand:
21032 Matching constraints are used in these circumstances. More
21033 precisely, the two operands that match must include one input-only
21034 operand and one output-only operand. Moreover, the digit must be a
21035 smaller number than the number of the operand that uses it in the
21039 An operand that is a valid memory address is allowed. This is for
21040 "load address" and "push address" instructions.
21042 `p' in the constraint must be accompanied by `address_operand' as
21043 the predicate in the `match_operand'. This predicate interprets
21044 the mode specified in the `match_operand' as the mode of the memory
21045 reference for which the address would be valid.
21048 Other letters can be defined in machine-dependent fashion to stand
21049 for particular classes of registers or other arbitrary operand
21050 types. `d', `a' and `f' are defined on the 68000/68020 to stand
21051 for data, address and floating point registers.
21054 File: gcc.info, Node: Multi-Alternative, Next: Modifiers, Prev: Simple Constraints, Up: Constraints
21056 5.38.2 Multiple Alternative Constraints
21057 ---------------------------------------
21059 Sometimes a single instruction has multiple alternative sets of possible
21060 operands. For example, on the 68000, a logical-or instruction can
21061 combine register or an immediate value into memory, or it can combine
21062 any kind of operand into a register; but it cannot combine one memory
21063 location into another.
21065 These constraints are represented as multiple alternatives. An
21066 alternative can be described by a series of letters for each operand.
21067 The overall constraint for an operand is made from the letters for this
21068 operand from the first alternative, a comma, the letters for this
21069 operand from the second alternative, a comma, and so on until the last
21072 If all the operands fit any one alternative, the instruction is valid.
21073 Otherwise, for each alternative, the compiler counts how many
21074 instructions must be added to copy the operands so that that
21075 alternative applies. The alternative requiring the least copying is
21076 chosen. If two alternatives need the same amount of copying, the one
21077 that comes first is chosen. These choices can be altered with the `?'
21078 and `!' characters:
21081 Disparage slightly the alternative that the `?' appears in, as a
21082 choice when no alternative applies exactly. The compiler regards
21083 this alternative as one unit more costly for each `?' that appears
21087 Disparage severely the alternative that the `!' appears in. This
21088 alternative can still be used if it fits without reloading, but if
21089 reloading is needed, some other alternative will be used.
21092 File: gcc.info, Node: Modifiers, Next: Machine Constraints, Prev: Multi-Alternative, Up: Constraints
21094 5.38.3 Constraint Modifier Characters
21095 -------------------------------------
21097 Here are constraint modifier characters.
21100 Means that this operand is write-only for this instruction: the
21101 previous value is discarded and replaced by output data.
21104 Means that this operand is both read and written by the
21107 When the compiler fixes up the operands to satisfy the constraints,
21108 it needs to know which operands are inputs to the instruction and
21109 which are outputs from it. `=' identifies an output; `+'
21110 identifies an operand that is both input and output; all other
21111 operands are assumed to be input only.
21113 If you specify `=' or `+' in a constraint, you put it in the first
21114 character of the constraint string.
21117 Means (in a particular alternative) that this operand is an
21118 "earlyclobber" operand, which is modified before the instruction is
21119 finished using the input operands. Therefore, this operand may
21120 not lie in a register that is used as an input operand or as part
21121 of any memory address.
21123 `&' applies only to the alternative in which it is written. In
21124 constraints with multiple alternatives, sometimes one alternative
21125 requires `&' while others do not. See, for example, the `movdf'
21128 An input operand can be tied to an earlyclobber operand if its only
21129 use as an input occurs before the early result is written. Adding
21130 alternatives of this form often allows GCC to produce better code
21131 when only some of the inputs can be affected by the earlyclobber.
21132 See, for example, the `mulsi3' insn of the ARM.
21134 `&' does not obviate the need to write `='.
21137 Declares the instruction to be commutative for this operand and the
21138 following operand. This means that the compiler may interchange
21139 the two operands if that is the cheapest way to make all operands
21140 fit the constraints. GCC can only handle one commutative pair in
21141 an asm; if you use more, the compiler may fail. Note that you
21142 need not use the modifier if the two alternatives are strictly
21143 identical; this would only waste time in the reload pass. The
21144 modifier is not operational after register allocation, so the
21145 result of `define_peephole2' and `define_split's performed after
21146 reload cannot rely on `%' to make the intended insn match.
21149 Says that all following characters, up to the next comma, are to be
21150 ignored as a constraint. They are significant only for choosing
21151 register preferences.
21154 Says that the following character should be ignored when choosing
21155 register preferences. `*' has no effect on the meaning of the
21156 constraint as a constraint, and no effect on reloading.
21160 File: gcc.info, Node: Machine Constraints, Prev: Modifiers, Up: Constraints
21162 5.38.4 Constraints for Particular Machines
21163 ------------------------------------------
21165 Whenever possible, you should use the general-purpose constraint letters
21166 in `asm' arguments, since they will convey meaning more readily to
21167 people reading your code. Failing that, use the constraint letters
21168 that usually have very similar meanings across architectures. The most
21169 commonly used constraints are `m' and `r' (for memory and
21170 general-purpose registers respectively; *note Simple Constraints::), and
21171 `I', usually the letter indicating the most common immediate-constant
21174 Each architecture defines additional constraints. These constraints
21175 are used by the compiler itself for instruction generation, as well as
21176 for `asm' statements; therefore, some of the constraints are not
21177 particularly useful for `asm'. Here is a summary of some of the
21178 machine-dependent constraints available on some particular machines; it
21179 includes both constraints that are useful for `asm' and constraints
21180 that aren't. The compiler source file mentioned in the table heading
21181 for each architecture is the definitive reference for the meanings of
21182 that architecture's constraints.
21184 _ARM family--`config/arm/arm.h'_
21187 Floating-point register
21190 VFP floating-point register
21193 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0,
21197 Floating-point constant that would satisfy the constraint `F'
21201 Integer that is valid as an immediate operand in a data
21202 processing instruction. That is, an integer in the range 0
21203 to 255 rotated by a multiple of 2
21206 Integer in the range -4095 to 4095
21209 Integer that satisfies constraint `I' when inverted (ones
21213 Integer that satisfies constraint `I' when negated (twos
21217 Integer in the range 0 to 32
21220 A memory reference where the exact address is in a single
21221 register (``m'' is preferable for `asm' statements)
21224 An item in the constant pool
21227 A symbol in the text segment of the current file
21230 A memory reference suitable for VFP load/store insns
21231 (reg+constant offset)
21234 A memory reference suitable for iWMMXt load/store
21238 A memory reference suitable for the ARMv4 ldrsb instruction.
21240 _AVR family--`config/avr/constraints.md'_
21243 Registers from r0 to r15
21246 Registers from r16 to r23
21249 Registers from r16 to r31
21252 Registers from r24 to r31. These registers can be used in
21256 Pointer register (r26-r31)
21259 Base pointer register (r28-r31)
21262 Stack pointer register (SPH:SPL)
21265 Temporary register r0
21268 Register pair X (r27:r26)
21271 Register pair Y (r29:r28)
21274 Register pair Z (r31:r30)
21277 Constant greater than -1, less than 64
21280 Constant greater than -64, less than 1
21289 Constant that fits in 8 bits
21292 Constant integer -1
21295 Constant integer 8, 16, or 24
21301 A floating point constant 0.0
21304 Integer constant in the range -6 ... 5.
21307 A memory address based on Y or Z pointer with displacement.
21309 _CRX Architecture--`config/crx/crx.h'_
21312 Registers from r0 to r14 (registers without stack pointer)
21315 Register r16 (64-bit accumulator lo register)
21318 Register r17 (64-bit accumulator hi register)
21321 Register pair r16-r17. (64-bit accumulator lo-hi pair)
21324 Constant that fits in 3 bits
21327 Constant that fits in 4 bits
21330 Constant that fits in 5 bits
21333 Constant that is one of -1, 4, -4, 7, 8, 12, 16, 20, 32, 48
21336 Floating point constant that is legal for store immediate
21338 _Hewlett-Packard PA-RISC--`config/pa/pa.h'_
21344 Floating point register
21347 Shift amount register
21350 Floating point register (deprecated)
21353 Upper floating point register (32-bit), floating point
21360 Signed 11-bit integer constant
21363 Signed 14-bit integer constant
21366 Integer constant that can be deposited with a `zdepi'
21370 Signed 5-bit integer constant
21376 Integer constant that can be loaded with a `ldil' instruction
21379 Integer constant whose value plus one is a power of 2
21382 Integer constant that can be used for `and' operations in
21383 `depi' and `extru' instructions
21386 Integer constant 31
21389 Integer constant 63
21392 Floating-point constant 0.0
21395 A `lo_sum' data-linkage-table memory operand
21398 A memory operand that can be used as the destination operand
21399 of an integer store instruction
21402 A scaled or unscaled indexed memory operand
21405 A memory operand for floating-point loads and stores
21408 A register indirect memory operand
21410 _picoChip family--`picochip.h'_
21416 Pointer register. A register which can be used to access
21417 memory without supplying an offset. Any other register can
21418 be used to access memory, but will need a constant offset.
21419 In the case of the offset being zero, it is more efficient to
21420 use a pointer register, since this reduces code size.
21423 A twin register. A register which may be paired with an
21424 adjacent register to create a 32-bit register.
21427 Any absolute memory address (e.g., symbolic constant, symbolic
21428 constant + offset).
21431 4-bit signed integer.
21434 4-bit unsigned integer.
21437 8-bit signed integer.
21440 Any constant whose absolute value is no greater than 4-bits.
21443 10-bit signed integer
21446 16-bit signed integer.
21449 _PowerPC and IBM RS6000--`config/rs6000/rs6000.h'_
21452 Address base register
21455 Floating point register
21461 `MQ', `CTR', or `LINK' register
21473 `CR' register (condition register) number 0
21476 `CR' register (condition register)
21479 `FPMEM' stack memory for FPR-GPR transfers
21482 Signed 16-bit constant
21485 Unsigned 16-bit constant shifted left 16 bits (use `L'
21486 instead for `SImode' constants)
21489 Unsigned 16-bit constant
21492 Signed 16-bit constant shifted left 16 bits
21495 Constant larger than 31
21504 Constant whose negation is a signed 16-bit constant
21507 Floating point constant that can be loaded into a register
21508 with one instruction per word
21511 Integer/Floating point constant that can be loaded into a
21512 register using three instructions
21515 Memory operand that is an offset from a register (`m' is
21516 preferable for `asm' statements)
21519 Memory operand that is an indexed or indirect from a register
21520 (`m' is preferable for `asm' statements)
21526 Address operand that is an indexed or indirect from a
21527 register (`p' is preferable for `asm' statements)
21530 Constant suitable as a 64-bit mask operand
21533 Constant suitable as a 32-bit mask operand
21536 System V Release 4 small data area reference
21539 AND masks that can be performed by two rldic{l, r}
21543 Vector constant that does not require memory
21546 _Intel 386--`config/i386/constraints.md'_
21549 Legacy register--the eight integer registers available on all
21550 i386 processors (`a', `b', `c', `d', `si', `di', `bp', `sp').
21553 Any register accessible as `Rl'. In 32-bit mode, `a', `b',
21554 `c', and `d'; in 64-bit mode, any integer register.
21557 Any register accessible as `Rh': `a', `b', `c', and `d'.
21578 The `a' and `d' registers, as a pair (for instructions that
21579 return half the result in one and half in the other).
21582 Any 80387 floating-point (stack) register.
21585 Top of 80387 floating-point stack (`%st(0)').
21588 Second from top of 80387 floating-point stack (`%st(1)').
21597 First SSE register (`%xmm0').
21600 Integer constant in the range 0 ... 31, for 32-bit shifts.
21603 Integer constant in the range 0 ... 63, for 64-bit shifts.
21606 Signed 8-bit integer constant.
21609 `0xFF' or `0xFFFF', for andsi as a zero-extending move.
21612 0, 1, 2, or 3 (shifts for the `lea' instruction).
21615 Unsigned 8-bit integer constant (for `in' and `out'
21619 Standard 80387 floating point constant.
21622 Standard SSE floating point constant.
21625 32-bit signed integer constant, or a symbolic reference known
21626 to fit that range (for immediate operands in sign-extending
21627 x86-64 instructions).
21630 32-bit unsigned integer constant, or a symbolic reference
21631 known to fit that range (for immediate operands in
21632 zero-extending x86-64 instructions).
21635 _Intel IA-64--`config/ia64/ia64.h'_
21638 General register `r0' to `r3' for `addl' instruction
21644 Predicate register (`c' as in "conditional")
21647 Application register residing in M-unit
21650 Application register residing in I-unit
21653 Floating-point register
21656 Memory operand. Remember that `m' allows postincrement and
21657 postdecrement which require printing with `%Pn' on IA-64.
21658 Use `S' to disallow postincrement and postdecrement.
21661 Floating-point constant 0.0 or 1.0
21664 14-bit signed integer constant
21667 22-bit signed integer constant
21670 8-bit signed integer constant for logical instructions
21673 8-bit adjusted signed integer constant for compare pseudo-ops
21676 6-bit unsigned integer constant for shift counts
21679 9-bit signed integer constant for load and store
21686 0 or -1 for `dep' instruction
21689 Non-volatile memory for floating-point loads and stores
21692 Integer constant in the range 1 to 4 for `shladd' instruction
21695 Memory operand except postincrement and postdecrement
21697 _FRV--`config/frv/frv.h'_
21700 Register in the class `ACC_REGS' (`acc0' to `acc7').
21703 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7').
21706 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0'
21710 Register in the class `GPR_REGS' (`gr0' to `gr63').
21713 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd
21714 registers are excluded not in the class but through the use
21715 of a machine mode larger than 4 bytes.
21718 Register in the class `FPR_REGS' (`fr0' to `fr63').
21721 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd
21722 registers are excluded not in the class but through the use
21723 of a machine mode larger than 4 bytes.
21726 Register in the class `LR_REG' (the `lr' register).
21729 Register in the class `QUAD_REGS' (`gr2' to `gr63').
21730 Register numbers not divisible by 4 are excluded not in the
21731 class but through the use of a machine mode larger than 8
21735 Register in the class `ICC_REGS' (`icc0' to `icc3').
21738 Register in the class `FCC_REGS' (`fcc0' to `fcc3').
21741 Register in the class `ICR_REGS' (`cc4' to `cc7').
21744 Register in the class `FCR_REGS' (`cc0' to `cc3').
21747 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63').
21748 Register numbers not divisible by 4 are excluded not in the
21749 class but through the use of a machine mode larger than 8
21753 Register in the class `SPR_REGS' (`lcr' and `lr').
21756 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7').
21759 Register in the class `ACCG_REGS' (`accg0' to `accg7').
21762 Register in the class `CR_REGS' (`cc0' to `cc7').
21765 Floating point constant zero
21768 6-bit signed integer constant
21771 10-bit signed integer constant
21774 16-bit signed integer constant
21777 16-bit unsigned integer constant
21780 12-bit signed integer constant that is negative--i.e. in the
21781 range of -2048 to -1
21787 12-bit signed integer constant that is greater than
21788 zero--i.e. in the range of 1 to 2047.
21791 _Blackfin family--`config/bfin/constraints.md'_
21800 A call clobbered P register.
21803 A single register. If N is in the range 0 to 7, the
21804 corresponding D register. If it is `A', then the register P0.
21807 Even-numbered D register
21810 Odd-numbered D register
21813 Accumulator register.
21816 Even-numbered accumulator register.
21819 Odd-numbered accumulator register.
21831 Registers used for circular buffering, i.e. I, B, or L
21847 Any D, P, B, M, I or L register.
21850 Additional registers typically used only in prologues and
21851 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and
21855 Any register except accumulators or CC.
21858 Signed 16 bit integer (in the range -32768 to 32767)
21861 Unsigned 16 bit integer (in the range 0 to 65535)
21864 Signed 7 bit integer (in the range -64 to 63)
21867 Unsigned 7 bit integer (in the range 0 to 127)
21870 Unsigned 5 bit integer (in the range 0 to 31)
21873 Signed 4 bit integer (in the range -8 to 7)
21876 Signed 3 bit integer (in the range -3 to 4)
21879 Unsigned 3 bit integer (in the range 0 to 7)
21882 Constant N, where N is a single-digit constant in the range 0
21886 An integer equal to one of the MACFLAG_XXX constants that is
21887 suitable for use with either accumulator.
21890 An integer equal to one of the MACFLAG_XXX constants that is
21891 suitable for use only with accumulator A1.
21900 An integer constant with exactly a single bit set.
21903 An integer constant with all bits set except exactly one.
21910 _M32C--`config/m32c/m32c.c'_
21915 `$sp', `$fb', `$sb'.
21918 Any control register, when they're 16 bits wide (nothing if
21919 control registers are 24 bits wide)
21922 Any control register, when they're 24 bits wide.
21928 $r0, $r1, $r2, $r3.
21931 $r0 or $r2, or $r2r0 for 32 bit values.
21934 $r1 or $r3, or $r3r1 for 32 bit values.
21937 A register that can hold a 64 bit value.
21940 $r0 or $r1 (registers with addressable high/low bytes)
21949 Address registers when they're 16 bits wide.
21952 Address registers when they're 24 bits wide.
21955 Registers that can hold QI values.
21958 Registers that can be used with displacements ($a0, $a1, $sb).
21961 Registers that can hold 32 bit values.
21964 Registers that can hold 16 bit values.
21967 Registers chat can hold 16 bit values, including all control
21971 $r0 through R1, plus $a0 and $a1.
21974 The flags register.
21977 The memory-based pseudo-registers $mem0 through $mem15.
21980 Registers that can hold pointers (16 bit registers for r8c,
21981 m16c; 24 bit registers for m32cm, m32c).
21984 Matches multiple registers in a PARALLEL to form a larger
21985 register. Used to match function return values.
22000 -8 ... -1 or 1 ... 8
22003 -16 ... -1 or 1 ... 16
22006 -32 ... -1 or 1 ... 32
22012 An 8 bit value with exactly one bit set.
22015 A 16 bit value with exactly one bit set.
22018 The common src/dest memory addressing modes.
22021 Memory addressed using $a0 or $a1.
22024 Memory addressed with immediate addresses.
22027 Memory addressed using the stack pointer ($sp).
22030 Memory addressed using the frame base register ($fb).
22033 Memory addressed using the small base register ($sb).
22038 _MIPS--`config/mips/constraints.md'_
22041 An address register. This is equivalent to `r' unless
22042 generating MIPS16 code.
22045 A floating-point register (if available).
22048 Formerly the `hi' register. This constraint is no longer
22052 The `lo' register. Use this register to store values that are
22053 no bigger than a word.
22056 The concatenated `hi' and `lo' registers. Use this register
22057 to store doubleword values.
22060 A register suitable for use in an indirect jump. This will
22061 always be `$25' for `-mabicalls'.
22064 Register `$3'. Do not use this constraint in new code; it is
22065 retained only for compatibility with glibc.
22068 Equivalent to `r'; retained for backwards compatibility.
22071 A floating-point condition code register.
22074 A signed 16-bit constant (for arithmetic instructions).
22080 An unsigned 16-bit constant (for logic instructions).
22083 A signed 32-bit constant in which the lower 16 bits are zero.
22084 Such constants can be loaded using `lui'.
22087 A constant that cannot be loaded using `lui', `addiu' or
22091 A constant in the range -65535 to -1 (inclusive).
22094 A signed 15-bit constant.
22097 A constant in the range 1 to 65535 (inclusive).
22100 Floating-point zero.
22103 An address that can be used in a non-macro load or store.
22105 _Motorola 680x0--`config/m68k/constraints.md'_
22114 68881 floating-point register, if available
22117 Integer in the range 1 to 8
22120 16-bit signed number
22123 Signed number whose magnitude is greater than 0x80
22126 Integer in the range -8 to -1
22129 Signed number whose magnitude is greater than 0x100
22132 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate
22135 16 (for rotate using swap)
22138 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate
22141 Numbers that mov3q can handle
22144 Floating point constant that is not a 68881 constant
22147 Operands that satisfy 'm' when -mpcrel is in effect
22150 Operands that satisfy 's' when -mpcrel is not in effect
22153 Address register indirect addressing mode
22156 Register offset addressing
22162 symbol_ref or const
22171 Range of signed numbers that don't fit in 16 bits
22174 Integers valid for mvq
22177 Integers valid for a moveq followed by a swap
22180 Integers valid for mvz
22183 Integers valid for mvs
22189 Non-register operands allowed in clr
22192 _Motorola 68HC11 & 68HC12 families--`config/m68hc11/m68hc11.h'_
22207 Temporary soft register _.tmp
22210 A soft register _.d1 to _.d31
22213 Stack pointer register
22222 Pseudo register `z' (replaced by `x' or `y' at the end)
22225 An address register: x, y or z
22228 An address register: x or y
22231 Register pair (x:d) to form a 32-bit value
22234 Constants in the range -65536 to 65535
22237 Constants whose 16-bit low part is zero
22240 Constant integer 1 or -1
22243 Constant integer 16
22246 Constants in the range -8 to 2
22249 _SPARC--`config/sparc/sparc.h'_
22252 Floating-point register on the SPARC-V8 architecture and
22253 lower floating-point register on the SPARC-V9 architecture.
22256 Floating-point register. It is equivalent to `f' on the
22257 SPARC-V8 architecture and contains both lower and upper
22258 floating-point registers on the SPARC-V9 architecture.
22261 Floating-point condition code register.
22264 Lower floating-point register. It is only valid on the
22265 SPARC-V9 architecture when the Visual Instruction Set is
22269 Floating-point register. It is only valid on the SPARC-V9
22270 architecture when the Visual Instruction Set is available.
22273 64-bit global or out register for the SPARC-V8+ architecture.
22279 Signed 13-bit constant
22285 32-bit constant with the low 12 bits clear (a constant that
22286 can be loaded with the `sethi' instruction)
22289 A constant in the range supported by `movcc' instructions
22292 A constant in the range supported by `movrcc' instructions
22295 Same as `K', except that it verifies that bits that are not
22296 in the lower 32-bit range are all zero. Must be used instead
22297 of `K' for modes wider than `SImode'
22303 Floating-point zero
22306 Signed 13-bit constant, sign-extended to 32 or 64 bits
22309 Floating-point constant whose integral representation can be
22310 moved into an integer register using a single sethi
22314 Floating-point constant whose integral representation can be
22315 moved into an integer register using a single mov instruction
22318 Floating-point constant whose integral representation can be
22319 moved into an integer register using a high/lo_sum
22320 instruction sequence
22323 Memory address aligned to an 8-byte boundary
22329 Memory address for `e' constraint registers
22335 _SPU--`config/spu/spu.h'_
22338 An immediate which can be loaded with the il/ila/ilh/ilhu
22339 instructions. const_int is treated as a 64 bit value.
22342 An immediate for and/xor/or instructions. const_int is
22343 treated as a 64 bit value.
22346 An immediate for the `iohl' instruction. const_int is
22347 treated as a 64 bit value.
22350 An immediate which can be loaded with `fsmbi'.
22353 An immediate which can be loaded with the il/ila/ilh/ilhu
22354 instructions. const_int is treated as a 32 bit value.
22357 An immediate for most arithmetic instructions. const_int is
22358 treated as a 32 bit value.
22361 An immediate for and/xor/or instructions. const_int is
22362 treated as a 32 bit value.
22365 An immediate for the `iohl' instruction. const_int is
22366 treated as a 32 bit value.
22369 A constant in the range [-64, 63] for shift/rotate
22373 An unsigned 7-bit constant for conversion/nop/channel
22377 A signed 10-bit constant for most arithmetic instructions.
22380 A signed 16 bit immediate for `stop'.
22383 An unsigned 16-bit constant for `iohl' and `fsmbi'.
22386 An unsigned 7-bit constant whose 3 least significant bits are
22390 An unsigned 3-bit constant for 16-byte rotates and shifts
22393 Call operand, reg, for indirect calls
22396 Call operand, symbol, for relative calls.
22399 Call operand, const_int, for absolute calls.
22402 An immediate which can be loaded with the il/ila/ilh/ilhu
22403 instructions. const_int is sign extended to 128 bit.
22406 An immediate for shift and rotate instructions. const_int is
22407 treated as a 32 bit value.
22410 An immediate for and/xor/or instructions. const_int is sign
22411 extended as a 128 bit.
22414 An immediate for the `iohl' instruction. const_int is sign
22415 extended to 128 bit.
22418 _S/390 and zSeries--`config/s390/s390.h'_
22421 Address register (general purpose register except r0)
22424 Condition code register
22427 Data register (arbitrary general purpose register)
22430 Floating-point register
22433 Unsigned 8-bit constant (0-255)
22436 Unsigned 12-bit constant (0-4095)
22439 Signed 16-bit constant (-32768-32767)
22442 Value appropriate as displacement.
22444 for short displacement
22446 `(-524288..524287)'
22447 for long displacement
22450 Constant integer with a value of 0x7fffffff.
22453 Multiple letter constraint followed by 4 parameter letters.
22455 number of the part counting from most to least
22462 mode of the containing operand
22465 value of the other parts (F--all bits set)
22466 The constraint matches if the specified part of a constant
22467 has a value different from its other parts.
22470 Memory reference without index register and with short
22474 Memory reference with index register and short displacement.
22477 Memory reference without index register but with long
22481 Memory reference with index register and long displacement.
22484 Pointer with short displacement.
22487 Pointer with long displacement.
22490 Shift count operand.
22493 _Score family--`config/score/score.h'_
22496 Registers from r0 to r32.
22499 Registers from r0 to r16.
22502 r8--r11 or r22--r27 registers.
22523 cnt + lcb + scb register.
22526 cr0--cr15 register.
22538 cp1 + cp2 + cp3 registers.
22541 High 16-bit constant (32-bit constant with 16 LSBs zero).
22544 Unsigned 5 bit integer (in the range 0 to 31).
22547 Unsigned 16 bit integer (in the range 0 to 65535).
22550 Signed 16 bit integer (in the range -32768 to 32767).
22553 Unsigned 14 bit integer (in the range 0 to 16383).
22556 Signed 14 bit integer (in the range -8192 to 8191).
22561 _Xstormy16--`config/stormy16/stormy16.h'_
22576 Registers r0 through r7.
22579 Registers r0 and r1.
22582 The carry register.
22585 Registers r8 and r9.
22588 A constant between 0 and 3 inclusive.
22591 A constant that has exactly one bit set.
22594 A constant that has exactly one bit clear.
22597 A constant between 0 and 255 inclusive.
22600 A constant between -255 and 0 inclusive.
22603 A constant between -3 and 0 inclusive.
22606 A constant between 1 and 4 inclusive.
22609 A constant between -4 and -1 inclusive.
22612 A memory reference that is a stack push.
22615 A memory reference that is a stack pop.
22618 A memory reference that refers to a constant address of known
22622 The register indicated by Rx (not implemented yet).
22625 A constant that is not between 2 and 15 inclusive.
22631 _Xtensa--`config/xtensa/constraints.md'_
22634 General-purpose 32-bit register
22637 One-bit boolean register
22640 MAC16 40-bit accumulator register
22643 Signed 12-bit integer constant, for use in MOVI instructions
22646 Signed 8-bit integer constant, for use in ADDI instructions
22649 Integer constant valid for BccI instructions
22652 Unsigned constant valid for BccUI instructions
22657 File: gcc.info, Node: Asm Labels, Next: Explicit Reg Vars, Prev: Constraints, Up: C Extensions
22659 5.39 Controlling Names Used in Assembler Code
22660 =============================================
22662 You can specify the name to be used in the assembler code for a C
22663 function or variable by writing the `asm' (or `__asm__') keyword after
22664 the declarator as follows:
22666 int foo asm ("myfoo") = 2;
22668 This specifies that the name to be used for the variable `foo' in the
22669 assembler code should be `myfoo' rather than the usual `_foo'.
22671 On systems where an underscore is normally prepended to the name of a C
22672 function or variable, this feature allows you to define names for the
22673 linker that do not start with an underscore.
22675 It does not make sense to use this feature with a non-static local
22676 variable since such variables do not have assembler names. If you are
22677 trying to put the variable in a particular register, see *note Explicit
22678 Reg Vars::. GCC presently accepts such code with a warning, but will
22679 probably be changed to issue an error, rather than a warning, in the
22682 You cannot use `asm' in this way in a function _definition_; but you
22683 can get the same effect by writing a declaration for the function
22684 before its definition and putting `asm' there, like this:
22686 extern func () asm ("FUNC");
22692 It is up to you to make sure that the assembler names you choose do not
22693 conflict with any other assembler symbols. Also, you must not use a
22694 register name; that would produce completely invalid assembler code.
22695 GCC does not as yet have the ability to store static variables in
22696 registers. Perhaps that will be added.
22699 File: gcc.info, Node: Explicit Reg Vars, Next: Alternate Keywords, Prev: Asm Labels, Up: C Extensions
22701 5.40 Variables in Specified Registers
22702 =====================================
22704 GNU C allows you to put a few global variables into specified hardware
22705 registers. You can also specify the register in which an ordinary
22706 register variable should be allocated.
22708 * Global register variables reserve registers throughout the program.
22709 This may be useful in programs such as programming language
22710 interpreters which have a couple of global variables that are
22711 accessed very often.
22713 * Local register variables in specific registers do not reserve the
22714 registers, except at the point where they are used as input or
22715 output operands in an `asm' statement and the `asm' statement
22716 itself is not deleted. The compiler's data flow analysis is
22717 capable of determining where the specified registers contain live
22718 values, and where they are available for other uses. Stores into
22719 local register variables may be deleted when they appear to be
22720 dead according to dataflow analysis. References to local register
22721 variables may be deleted or moved or simplified.
22723 These local variables are sometimes convenient for use with the
22724 extended `asm' feature (*note Extended Asm::), if you want to
22725 write one output of the assembler instruction directly into a
22726 particular register. (This will work provided the register you
22727 specify fits the constraints specified for that operand in the
22732 * Global Reg Vars::
22736 File: gcc.info, Node: Global Reg Vars, Next: Local Reg Vars, Up: Explicit Reg Vars
22738 5.40.1 Defining Global Register Variables
22739 -----------------------------------------
22741 You can define a global register variable in GNU C like this:
22743 register int *foo asm ("a5");
22745 Here `a5' is the name of the register which should be used. Choose a
22746 register which is normally saved and restored by function calls on your
22747 machine, so that library routines will not clobber it.
22749 Naturally the register name is cpu-dependent, so you would need to
22750 conditionalize your program according to cpu type. The register `a5'
22751 would be a good choice on a 68000 for a variable of pointer type. On
22752 machines with register windows, be sure to choose a "global" register
22753 that is not affected magically by the function call mechanism.
22755 In addition, operating systems on one type of cpu may differ in how
22756 they name the registers; then you would need additional conditionals.
22757 For example, some 68000 operating systems call this register `%a5'.
22759 Eventually there may be a way of asking the compiler to choose a
22760 register automatically, but first we need to figure out how it should
22761 choose and how to enable you to guide the choice. No solution is
22764 Defining a global register variable in a certain register reserves that
22765 register entirely for this use, at least within the current compilation.
22766 The register will not be allocated for any other purpose in the
22767 functions in the current compilation. The register will not be saved
22768 and restored by these functions. Stores into this register are never
22769 deleted even if they would appear to be dead, but references may be
22770 deleted or moved or simplified.
22772 It is not safe to access the global register variables from signal
22773 handlers, or from more than one thread of control, because the system
22774 library routines may temporarily use the register for other things
22775 (unless you recompile them specially for the task at hand).
22777 It is not safe for one function that uses a global register variable to
22778 call another such function `foo' by way of a third function `lose' that
22779 was compiled without knowledge of this variable (i.e. in a different
22780 source file in which the variable wasn't declared). This is because
22781 `lose' might save the register and put some other value there. For
22782 example, you can't expect a global register variable to be available in
22783 the comparison-function that you pass to `qsort', since `qsort' might
22784 have put something else in that register. (If you are prepared to
22785 recompile `qsort' with the same global register variable, you can solve
22788 If you want to recompile `qsort' or other source files which do not
22789 actually use your global register variable, so that they will not use
22790 that register for any other purpose, then it suffices to specify the
22791 compiler option `-ffixed-REG'. You need not actually add a global
22792 register declaration to their source code.
22794 A function which can alter the value of a global register variable
22795 cannot safely be called from a function compiled without this variable,
22796 because it could clobber the value the caller expects to find there on
22797 return. Therefore, the function which is the entry point into the part
22798 of the program that uses the global register variable must explicitly
22799 save and restore the value which belongs to its caller.
22801 On most machines, `longjmp' will restore to each global register
22802 variable the value it had at the time of the `setjmp'. On some
22803 machines, however, `longjmp' will not change the value of global
22804 register variables. To be portable, the function that called `setjmp'
22805 should make other arrangements to save the values of the global register
22806 variables, and to restore them in a `longjmp'. This way, the same
22807 thing will happen regardless of what `longjmp' does.
22809 All global register variable declarations must precede all function
22810 definitions. If such a declaration could appear after function
22811 definitions, the declaration would be too late to prevent the register
22812 from being used for other purposes in the preceding functions.
22814 Global register variables may not have initial values, because an
22815 executable file has no means to supply initial contents for a register.
22817 On the SPARC, there are reports that g3 ... g7 are suitable registers,
22818 but certain library functions, such as `getwd', as well as the
22819 subroutines for division and remainder, modify g3 and g4. g1 and g2
22820 are local temporaries.
22822 On the 68000, a2 ... a5 should be suitable, as should d2 ... d7. Of
22823 course, it will not do to use more than a few of those.
22826 File: gcc.info, Node: Local Reg Vars, Prev: Global Reg Vars, Up: Explicit Reg Vars
22828 5.40.2 Specifying Registers for Local Variables
22829 -----------------------------------------------
22831 You can define a local register variable with a specified register like
22834 register int *foo asm ("a5");
22836 Here `a5' is the name of the register which should be used. Note that
22837 this is the same syntax used for defining global register variables,
22838 but for a local variable it would appear within a function.
22840 Naturally the register name is cpu-dependent, but this is not a
22841 problem, since specific registers are most often useful with explicit
22842 assembler instructions (*note Extended Asm::). Both of these things
22843 generally require that you conditionalize your program according to cpu
22846 In addition, operating systems on one type of cpu may differ in how
22847 they name the registers; then you would need additional conditionals.
22848 For example, some 68000 operating systems call this register `%a5'.
22850 Defining such a register variable does not reserve the register; it
22851 remains available for other uses in places where flow control determines
22852 the variable's value is not live.
22854 This option does not guarantee that GCC will generate code that has
22855 this variable in the register you specify at all times. You may not
22856 code an explicit reference to this register in the _assembler
22857 instruction template_ part of an `asm' statement and assume it will
22858 always refer to this variable. However, using the variable as an `asm'
22859 _operand_ guarantees that the specified register is used for the
22862 Stores into local register variables may be deleted when they appear
22863 to be dead according to dataflow analysis. References to local
22864 register variables may be deleted or moved or simplified.
22866 As for global register variables, it's recommended that you choose a
22867 register which is normally saved and restored by function calls on your
22868 machine, so that library routines will not clobber it. A common
22869 pitfall is to initialize multiple call-clobbered registers with
22870 arbitrary expressions, where a function call or library call for an
22871 arithmetic operator will overwrite a register value from a previous
22872 assignment, for example `r0' below:
22873 register int *p1 asm ("r0") = ...;
22874 register int *p2 asm ("r1") = ...;
22875 In those cases, a solution is to use a temporary variable for each
22876 arbitrary expression. *Note Example of asm with clobbered asm reg::.
22879 File: gcc.info, Node: Alternate Keywords, Next: Incomplete Enums, Prev: Explicit Reg Vars, Up: C Extensions
22881 5.41 Alternate Keywords
22882 =======================
22884 `-ansi' and the various `-std' options disable certain keywords. This
22885 causes trouble when you want to use GNU C extensions, or a
22886 general-purpose header file that should be usable by all programs,
22887 including ISO C programs. The keywords `asm', `typeof' and `inline'
22888 are not available in programs compiled with `-ansi' or `-std' (although
22889 `inline' can be used in a program compiled with `-std=c99'). The ISO
22890 C99 keyword `restrict' is only available when `-std=gnu99' (which will
22891 eventually be the default) or `-std=c99' (or the equivalent
22892 `-std=iso9899:1999') is used.
22894 The way to solve these problems is to put `__' at the beginning and
22895 end of each problematical keyword. For example, use `__asm__' instead
22896 of `asm', and `__inline__' instead of `inline'.
22898 Other C compilers won't accept these alternative keywords; if you want
22899 to compile with another compiler, you can define the alternate keywords
22900 as macros to replace them with the customary keywords. It looks like
22904 #define __asm__ asm
22907 `-pedantic' and other options cause warnings for many GNU C extensions.
22908 You can prevent such warnings within one expression by writing
22909 `__extension__' before the expression. `__extension__' has no effect
22913 File: gcc.info, Node: Incomplete Enums, Next: Function Names, Prev: Alternate Keywords, Up: C Extensions
22915 5.42 Incomplete `enum' Types
22916 ============================
22918 You can define an `enum' tag without specifying its possible values.
22919 This results in an incomplete type, much like what you get if you write
22920 `struct foo' without describing the elements. A later declaration
22921 which does specify the possible values completes the type.
22923 You can't allocate variables or storage using the type while it is
22924 incomplete. However, you can work with pointers to that type.
22926 This extension may not be very useful, but it makes the handling of
22927 `enum' more consistent with the way `struct' and `union' are handled.
22929 This extension is not supported by GNU C++.
22932 File: gcc.info, Node: Function Names, Next: Return Address, Prev: Incomplete Enums, Up: C Extensions
22934 5.43 Function Names as Strings
22935 ==============================
22937 GCC provides three magic variables which hold the name of the current
22938 function, as a string. The first of these is `__func__', which is part
22939 of the C99 standard:
22941 The identifier `__func__' is implicitly declared by the translator as
22942 if, immediately following the opening brace of each function
22943 definition, the declaration
22945 static const char __func__[] = "function-name";
22947 appeared, where function-name is the name of the lexically-enclosing
22948 function. This name is the unadorned name of the function.
22950 `__FUNCTION__' is another name for `__func__'. Older versions of GCC
22951 recognize only this name. However, it is not standardized. For
22952 maximum portability, we recommend you use `__func__', but provide a
22953 fallback definition with the preprocessor:
22955 #if __STDC_VERSION__ < 199901L
22957 # define __func__ __FUNCTION__
22959 # define __func__ "<unknown>"
22963 In C, `__PRETTY_FUNCTION__' is yet another name for `__func__'.
22964 However, in C++, `__PRETTY_FUNCTION__' contains the type signature of
22965 the function as well as its bare name. For example, this program:
22968 extern int printf (char *, ...);
22975 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
22976 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
22991 __PRETTY_FUNCTION__ = void a::sub(int)
22993 These identifiers are not preprocessor macros. In GCC 3.3 and
22994 earlier, in C only, `__FUNCTION__' and `__PRETTY_FUNCTION__' were
22995 treated as string literals; they could be used to initialize `char'
22996 arrays, and they could be concatenated with other string literals. GCC
22997 3.4 and later treat them as variables, like `__func__'. In C++,
22998 `__FUNCTION__' and `__PRETTY_FUNCTION__' have always been variables.
23001 File: gcc.info, Node: Return Address, Next: Vector Extensions, Prev: Function Names, Up: C Extensions
23003 5.44 Getting the Return or Frame Address of a Function
23004 ======================================================
23006 These functions may be used to get information about the callers of a
23009 -- Built-in Function: void * __builtin_return_address (unsigned int
23011 This function returns the return address of the current function,
23012 or of one of its callers. The LEVEL argument is number of frames
23013 to scan up the call stack. A value of `0' yields the return
23014 address of the current function, a value of `1' yields the return
23015 address of the caller of the current function, and so forth. When
23016 inlining the expected behavior is that the function will return
23017 the address of the function that will be returned to. To work
23018 around this behavior use the `noinline' function attribute.
23020 The LEVEL argument must be a constant integer.
23022 On some machines it may be impossible to determine the return
23023 address of any function other than the current one; in such cases,
23024 or when the top of the stack has been reached, this function will
23025 return `0' or a random value. In addition,
23026 `__builtin_frame_address' may be used to determine if the top of
23027 the stack has been reached.
23029 This function should only be used with a nonzero argument for
23030 debugging purposes.
23032 -- Built-in Function: void * __builtin_frame_address (unsigned int
23034 This function is similar to `__builtin_return_address', but it
23035 returns the address of the function frame rather than the return
23036 address of the function. Calling `__builtin_frame_address' with a
23037 value of `0' yields the frame address of the current function, a
23038 value of `1' yields the frame address of the caller of the current
23039 function, and so forth.
23041 The frame is the area on the stack which holds local variables and
23042 saved registers. The frame address is normally the address of the
23043 first word pushed on to the stack by the function. However, the
23044 exact definition depends upon the processor and the calling
23045 convention. If the processor has a dedicated frame pointer
23046 register, and the function has a frame, then
23047 `__builtin_frame_address' will return the value of the frame
23050 On some machines it may be impossible to determine the frame
23051 address of any function other than the current one; in such cases,
23052 or when the top of the stack has been reached, this function will
23053 return `0' if the first frame pointer is properly initialized by
23056 This function should only be used with a nonzero argument for
23057 debugging purposes.
23060 File: gcc.info, Node: Vector Extensions, Next: Offsetof, Prev: Return Address, Up: C Extensions
23062 5.45 Using vector instructions through built-in functions
23063 =========================================================
23065 On some targets, the instruction set contains SIMD vector instructions
23066 that operate on multiple values contained in one large register at the
23067 same time. For example, on the i386 the MMX, 3Dnow! and SSE extensions
23068 can be used this way.
23070 The first step in using these extensions is to provide the necessary
23071 data types. This should be done using an appropriate `typedef':
23073 typedef int v4si __attribute__ ((vector_size (16)));
23075 The `int' type specifies the base type, while the attribute specifies
23076 the vector size for the variable, measured in bytes. For example, the
23077 declaration above causes the compiler to set the mode for the `v4si'
23078 type to be 16 bytes wide and divided into `int' sized units. For a
23079 32-bit `int' this means a vector of 4 units of 4 bytes, and the
23080 corresponding mode of `foo' will be V4SI.
23082 The `vector_size' attribute is only applicable to integral and float
23083 scalars, although arrays, pointers, and function return values are
23084 allowed in conjunction with this construct.
23086 All the basic integer types can be used as base types, both as signed
23087 and as unsigned: `char', `short', `int', `long', `long long'. In
23088 addition, `float' and `double' can be used to build floating-point
23091 Specifying a combination that is not valid for the current architecture
23092 will cause GCC to synthesize the instructions using a narrower mode.
23093 For example, if you specify a variable of type `V4SI' and your
23094 architecture does not allow for this specific SIMD type, GCC will
23095 produce code that uses 4 `SIs'.
23097 The types defined in this manner can be used with a subset of normal C
23098 operations. Currently, GCC will allow using the following operators on
23099 these types: `+, -, *, /, unary minus, ^, |, &, ~'.
23101 The operations behave like C++ `valarrays'. Addition is defined as
23102 the addition of the corresponding elements of the operands. For
23103 example, in the code below, each of the 4 elements in A will be added
23104 to the corresponding 4 elements in B and the resulting vector will be
23107 typedef int v4si __attribute__ ((vector_size (16)));
23113 Subtraction, multiplication, division, and the logical operations
23114 operate in a similar manner. Likewise, the result of using the unary
23115 minus or complement operators on a vector type is a vector whose
23116 elements are the negative or complemented values of the corresponding
23117 elements in the operand.
23119 You can declare variables and use them in function calls and returns,
23120 as well as in assignments and some casts. You can specify a vector
23121 type as a return type for a function. Vector types can also be used as
23122 function arguments. It is possible to cast from one vector type to
23123 another, provided they are of the same size (in fact, you can also cast
23124 vectors to and from other datatypes of the same size).
23126 You cannot operate between vectors of different lengths or different
23127 signedness without a cast.
23129 A port that supports hardware vector operations, usually provides a set
23130 of built-in functions that can be used to operate on vectors. For
23131 example, a function to add two vectors and multiply the result by a
23132 third could look like this:
23134 v4si f (v4si a, v4si b, v4si c)
23136 v4si tmp = __builtin_addv4si (a, b);
23137 return __builtin_mulv4si (tmp, c);
23141 File: gcc.info, Node: Offsetof, Next: Atomic Builtins, Prev: Vector Extensions, Up: C Extensions
23146 GCC implements for both C and C++ a syntactic extension to implement
23147 the `offsetof' macro.
23150 "__builtin_offsetof" "(" `typename' "," offsetof_member_designator ")"
23152 offsetof_member_designator:
23154 | offsetof_member_designator "." `identifier'
23155 | offsetof_member_designator "[" `expr' "]"
23157 This extension is sufficient such that
23159 #define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER)
23161 is a suitable definition of the `offsetof' macro. In C++, TYPE may be
23162 dependent. In either case, MEMBER may consist of a single identifier,
23163 or a sequence of member accesses and array references.
23166 File: gcc.info, Node: Atomic Builtins, Next: Object Size Checking, Prev: Offsetof, Up: C Extensions
23168 5.47 Built-in functions for atomic memory access
23169 ================================================
23171 The following builtins are intended to be compatible with those
23172 described in the `Intel Itanium Processor-specific Application Binary
23173 Interface', section 7.4. As such, they depart from the normal GCC
23174 practice of using the "__builtin_" prefix, and further that they are
23175 overloaded such that they work on multiple types.
23177 The definition given in the Intel documentation allows only for the
23178 use of the types `int', `long', `long long' as well as their unsigned
23179 counterparts. GCC will allow any integral scalar or pointer type that
23180 is 1, 2, 4 or 8 bytes in length.
23182 Not all operations are supported by all target processors. If a
23183 particular operation cannot be implemented on the target processor, a
23184 warning will be generated and a call an external function will be
23185 generated. The external function will carry the same name as the
23186 builtin, with an additional suffix `_N' where N is the size of the data
23189 In most cases, these builtins are considered a "full barrier". That
23190 is, no memory operand will be moved across the operation, either
23191 forward or backward. Further, instructions will be issued as necessary
23192 to prevent the processor from speculating loads across the operation
23193 and from queuing stores after the operation.
23195 All of the routines are described in the Intel documentation to take
23196 "an optional list of variables protected by the memory barrier". It's
23197 not clear what is meant by that; it could mean that _only_ the
23198 following variables are protected, or it could mean that these variables
23199 should in addition be protected. At present GCC ignores this list and
23200 protects all variables which are globally accessible. If in the future
23201 we make some use of this list, an empty list will continue to mean all
23202 globally accessible variables.
23204 `TYPE __sync_fetch_and_add (TYPE *ptr, TYPE value, ...)'
23205 `TYPE __sync_fetch_and_sub (TYPE *ptr, TYPE value, ...)'
23206 `TYPE __sync_fetch_and_or (TYPE *ptr, TYPE value, ...)'
23207 `TYPE __sync_fetch_and_and (TYPE *ptr, TYPE value, ...)'
23208 `TYPE __sync_fetch_and_xor (TYPE *ptr, TYPE value, ...)'
23209 `TYPE __sync_fetch_and_nand (TYPE *ptr, TYPE value, ...)'
23210 These builtins perform the operation suggested by the name, and
23211 returns the value that had previously been in memory. That is,
23213 { tmp = *ptr; *ptr OP= value; return tmp; }
23214 { tmp = *ptr; *ptr = ~(tmp & value); return tmp; } // nand
23216 _Note:_ GCC 4.4 and later implement `__sync_fetch_and_nand'
23217 builtin as `*ptr = ~(tmp & value)' instead of `*ptr = ~tmp &
23220 `TYPE __sync_add_and_fetch (TYPE *ptr, TYPE value, ...)'
23221 `TYPE __sync_sub_and_fetch (TYPE *ptr, TYPE value, ...)'
23222 `TYPE __sync_or_and_fetch (TYPE *ptr, TYPE value, ...)'
23223 `TYPE __sync_and_and_fetch (TYPE *ptr, TYPE value, ...)'
23224 `TYPE __sync_xor_and_fetch (TYPE *ptr, TYPE value, ...)'
23225 `TYPE __sync_nand_and_fetch (TYPE *ptr, TYPE value, ...)'
23226 These builtins perform the operation suggested by the name, and
23227 return the new value. That is,
23229 { *ptr OP= value; return *ptr; }
23230 { *ptr = ~(*ptr & value); return *ptr; } // nand
23232 _Note:_ GCC 4.4 and later implement `__sync_nand_and_fetch'
23233 builtin as `*ptr = ~(*ptr & value)' instead of `*ptr = ~*ptr &
23236 `bool __sync_bool_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
23237 `TYPE __sync_val_compare_and_swap (TYPE *ptr, TYPE oldval TYPE newval, ...)'
23238 These builtins perform an atomic compare and swap. That is, if
23239 the current value of `*PTR' is OLDVAL, then write NEWVAL into
23242 The "bool" version returns true if the comparison is successful and
23243 NEWVAL was written. The "val" version returns the contents of
23244 `*PTR' before the operation.
23246 `__sync_synchronize (...)'
23247 This builtin issues a full memory barrier.
23249 `TYPE __sync_lock_test_and_set (TYPE *ptr, TYPE value, ...)'
23250 This builtin, as described by Intel, is not a traditional
23251 test-and-set operation, but rather an atomic exchange operation.
23252 It writes VALUE into `*PTR', and returns the previous contents of
23255 Many targets have only minimal support for such locks, and do not
23256 support a full exchange operation. In this case, a target may
23257 support reduced functionality here by which the _only_ valid value
23258 to store is the immediate constant 1. The exact value actually
23259 stored in `*PTR' is implementation defined.
23261 This builtin is not a full barrier, but rather an "acquire
23262 barrier". This means that references after the builtin cannot
23263 move to (or be speculated to) before the builtin, but previous
23264 memory stores may not be globally visible yet, and previous memory
23265 loads may not yet be satisfied.
23267 `void __sync_lock_release (TYPE *ptr, ...)'
23268 This builtin releases the lock acquired by
23269 `__sync_lock_test_and_set'. Normally this means writing the
23270 constant 0 to `*PTR'.
23272 This builtin is not a full barrier, but rather a "release barrier".
23273 This means that all previous memory stores are globally visible,
23274 and all previous memory loads have been satisfied, but following
23275 memory reads are not prevented from being speculated to before the
23279 File: gcc.info, Node: Object Size Checking, Next: Other Builtins, Prev: Atomic Builtins, Up: C Extensions
23281 5.48 Object Size Checking Builtins
23282 ==================================
23284 GCC implements a limited buffer overflow protection mechanism that can
23285 prevent some buffer overflow attacks.
23287 -- Built-in Function: size_t __builtin_object_size (void * PTR, int
23289 is a built-in construct that returns a constant number of bytes
23290 from PTR to the end of the object PTR pointer points to (if known
23291 at compile time). `__builtin_object_size' never evaluates its
23292 arguments for side-effects. If there are any side-effects in
23293 them, it returns `(size_t) -1' for TYPE 0 or 1 and `(size_t) 0'
23294 for TYPE 2 or 3. If there are multiple objects PTR can point to
23295 and all of them are known at compile time, the returned number is
23296 the maximum of remaining byte counts in those objects if TYPE & 2
23297 is 0 and minimum if nonzero. If it is not possible to determine
23298 which objects PTR points to at compile time,
23299 `__builtin_object_size' should return `(size_t) -1' for TYPE 0 or
23300 1 and `(size_t) 0' for TYPE 2 or 3.
23302 TYPE is an integer constant from 0 to 3. If the least significant
23303 bit is clear, objects are whole variables, if it is set, a closest
23304 surrounding subobject is considered the object a pointer points to.
23305 The second bit determines if maximum or minimum of remaining bytes
23308 struct V { char buf1[10]; int b; char buf2[10]; } var;
23309 char *p = &var.buf1[1], *q = &var.b;
23311 /* Here the object p points to is var. */
23312 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
23313 /* The subobject p points to is var.buf1. */
23314 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
23315 /* The object q points to is var. */
23316 assert (__builtin_object_size (q, 0)
23317 == (char *) (&var + 1) - (char *) &var.b);
23318 /* The subobject q points to is var.b. */
23319 assert (__builtin_object_size (q, 1) == sizeof (var.b));
23321 There are built-in functions added for many common string operation
23322 functions, e.g., for `memcpy' `__builtin___memcpy_chk' built-in is
23323 provided. This built-in has an additional last argument, which is the
23324 number of bytes remaining in object the DEST argument points to or
23325 `(size_t) -1' if the size is not known.
23327 The built-in functions are optimized into the normal string functions
23328 like `memcpy' if the last argument is `(size_t) -1' or if it is known
23329 at compile time that the destination object will not be overflown. If
23330 the compiler can determine at compile time the object will be always
23331 overflown, it issues a warning.
23333 The intended use can be e.g.
23336 #define bos0(dest) __builtin_object_size (dest, 0)
23337 #define memcpy(dest, src, n) \
23338 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
23342 /* It is unknown what object p points to, so this is optimized
23343 into plain memcpy - no checking is possible. */
23344 memcpy (p, "abcde", n);
23345 /* Destination is known and length too. It is known at compile
23346 time there will be no overflow. */
23347 memcpy (&buf[5], "abcde", 5);
23348 /* Destination is known, but the length is not known at compile time.
23349 This will result in __memcpy_chk call that can check for overflow
23351 memcpy (&buf[5], "abcde", n);
23352 /* Destination is known and it is known at compile time there will
23353 be overflow. There will be a warning and __memcpy_chk call that
23354 will abort the program at runtime. */
23355 memcpy (&buf[6], "abcde", 5);
23357 Such built-in functions are provided for `memcpy', `mempcpy',
23358 `memmove', `memset', `strcpy', `stpcpy', `strncpy', `strcat' and
23361 There are also checking built-in functions for formatted output
23363 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
23364 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
23365 const char *fmt, ...);
23366 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
23368 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
23369 const char *fmt, va_list ap);
23371 The added FLAG argument is passed unchanged to `__sprintf_chk' etc.
23372 functions and can contain implementation specific flags on what
23373 additional security measures the checking function might take, such as
23374 handling `%n' differently.
23376 The OS argument is the object size S points to, like in the other
23377 built-in functions. There is a small difference in the behavior
23378 though, if OS is `(size_t) -1', the built-in functions are optimized
23379 into the non-checking functions only if FLAG is 0, otherwise the
23380 checking function is called with OS argument set to `(size_t) -1'.
23382 In addition to this, there are checking built-in functions
23383 `__builtin___printf_chk', `__builtin___vprintf_chk',
23384 `__builtin___fprintf_chk' and `__builtin___vfprintf_chk'. These have
23385 just one additional argument, FLAG, right before format string FMT. If
23386 the compiler is able to optimize them to `fputc' etc. functions, it
23387 will, otherwise the checking function should be called and the FLAG
23388 argument passed to it.
23391 File: gcc.info, Node: Other Builtins, Next: Target Builtins, Prev: Object Size Checking, Up: C Extensions
23393 5.49 Other built-in functions provided by GCC
23394 =============================================
23396 GCC provides a large number of built-in functions other than the ones
23397 mentioned above. Some of these are for internal use in the processing
23398 of exceptions or variable-length argument lists and will not be
23399 documented here because they may change from time to time; we do not
23400 recommend general use of these functions.
23402 The remaining functions are provided for optimization purposes.
23404 GCC includes built-in versions of many of the functions in the standard
23405 C library. The versions prefixed with `__builtin_' will always be
23406 treated as having the same meaning as the C library function even if you
23407 specify the `-fno-builtin' option. (*note C Dialect Options::) Many of
23408 these functions are only optimized in certain cases; if they are not
23409 optimized in a particular case, a call to the library function will be
23412 Outside strict ISO C mode (`-ansi', `-std=c89' or `-std=c99'), the
23413 functions `_exit', `alloca', `bcmp', `bzero', `dcgettext', `dgettext',
23414 `dremf', `dreml', `drem', `exp10f', `exp10l', `exp10', `ffsll', `ffsl',
23415 `ffs', `fprintf_unlocked', `fputs_unlocked', `gammaf', `gammal',
23416 `gamma', `gammaf_r', `gammal_r', `gamma_r', `gettext', `index',
23417 `isascii', `j0f', `j0l', `j0', `j1f', `j1l', `j1', `jnf', `jnl', `jn',
23418 `lgammaf_r', `lgammal_r', `lgamma_r', `mempcpy', `pow10f', `pow10l',
23419 `pow10', `printf_unlocked', `rindex', `scalbf', `scalbl', `scalb',
23420 `signbit', `signbitf', `signbitl', `signbitd32', `signbitd64',
23421 `signbitd128', `significandf', `significandl', `significand', `sincosf',
23422 `sincosl', `sincos', `stpcpy', `stpncpy', `strcasecmp', `strdup',
23423 `strfmon', `strncasecmp', `strndup', `toascii', `y0f', `y0l', `y0',
23424 `y1f', `y1l', `y1', `ynf', `ynl' and `yn' may be handled as built-in
23425 functions. All these functions have corresponding versions prefixed
23426 with `__builtin_', which may be used even in strict C89 mode.
23428 The ISO C99 functions `_Exit', `acoshf', `acoshl', `acosh', `asinhf',
23429 `asinhl', `asinh', `atanhf', `atanhl', `atanh', `cabsf', `cabsl',
23430 `cabs', `cacosf', `cacoshf', `cacoshl', `cacosh', `cacosl', `cacos',
23431 `cargf', `cargl', `carg', `casinf', `casinhf', `casinhl', `casinh',
23432 `casinl', `casin', `catanf', `catanhf', `catanhl', `catanh', `catanl',
23433 `catan', `cbrtf', `cbrtl', `cbrt', `ccosf', `ccoshf', `ccoshl',
23434 `ccosh', `ccosl', `ccos', `cexpf', `cexpl', `cexp', `cimagf', `cimagl',
23435 `cimag', `clogf', `clogl', `clog', `conjf', `conjl', `conj',
23436 `copysignf', `copysignl', `copysign', `cpowf', `cpowl', `cpow',
23437 `cprojf', `cprojl', `cproj', `crealf', `creall', `creal', `csinf',
23438 `csinhf', `csinhl', `csinh', `csinl', `csin', `csqrtf', `csqrtl',
23439 `csqrt', `ctanf', `ctanhf', `ctanhl', `ctanh', `ctanl', `ctan',
23440 `erfcf', `erfcl', `erfc', `erff', `erfl', `erf', `exp2f', `exp2l',
23441 `exp2', `expm1f', `expm1l', `expm1', `fdimf', `fdiml', `fdim', `fmaf',
23442 `fmal', `fmaxf', `fmaxl', `fmax', `fma', `fminf', `fminl', `fmin',
23443 `hypotf', `hypotl', `hypot', `ilogbf', `ilogbl', `ilogb', `imaxabs',
23444 `isblank', `iswblank', `lgammaf', `lgammal', `lgamma', `llabs',
23445 `llrintf', `llrintl', `llrint', `llroundf', `llroundl', `llround',
23446 `log1pf', `log1pl', `log1p', `log2f', `log2l', `log2', `logbf',
23447 `logbl', `logb', `lrintf', `lrintl', `lrint', `lroundf', `lroundl',
23448 `lround', `nearbyintf', `nearbyintl', `nearbyint', `nextafterf',
23449 `nextafterl', `nextafter', `nexttowardf', `nexttowardl', `nexttoward',
23450 `remainderf', `remainderl', `remainder', `remquof', `remquol',
23451 `remquo', `rintf', `rintl', `rint', `roundf', `roundl', `round',
23452 `scalblnf', `scalblnl', `scalbln', `scalbnf', `scalbnl', `scalbn',
23453 `snprintf', `tgammaf', `tgammal', `tgamma', `truncf', `truncl', `trunc',
23454 `vfscanf', `vscanf', `vsnprintf' and `vsscanf' are handled as built-in
23455 functions except in strict ISO C90 mode (`-ansi' or `-std=c89').
23457 There are also built-in versions of the ISO C99 functions `acosf',
23458 `acosl', `asinf', `asinl', `atan2f', `atan2l', `atanf', `atanl',
23459 `ceilf', `ceill', `cosf', `coshf', `coshl', `cosl', `expf', `expl',
23460 `fabsf', `fabsl', `floorf', `floorl', `fmodf', `fmodl', `frexpf',
23461 `frexpl', `ldexpf', `ldexpl', `log10f', `log10l', `logf', `logl',
23462 `modfl', `modf', `powf', `powl', `sinf', `sinhf', `sinhl', `sinl',
23463 `sqrtf', `sqrtl', `tanf', `tanhf', `tanhl' and `tanl' that are
23464 recognized in any mode since ISO C90 reserves these names for the
23465 purpose to which ISO C99 puts them. All these functions have
23466 corresponding versions prefixed with `__builtin_'.
23468 The ISO C94 functions `iswalnum', `iswalpha', `iswcntrl', `iswdigit',
23469 `iswgraph', `iswlower', `iswprint', `iswpunct', `iswspace', `iswupper',
23470 `iswxdigit', `towlower' and `towupper' are handled as built-in functions
23471 except in strict ISO C90 mode (`-ansi' or `-std=c89').
23473 The ISO C90 functions `abort', `abs', `acos', `asin', `atan2', `atan',
23474 `calloc', `ceil', `cosh', `cos', `exit', `exp', `fabs', `floor', `fmod',
23475 `fprintf', `fputs', `frexp', `fscanf', `isalnum', `isalpha', `iscntrl',
23476 `isdigit', `isgraph', `islower', `isprint', `ispunct', `isspace',
23477 `isupper', `isxdigit', `tolower', `toupper', `labs', `ldexp', `log10',
23478 `log', `malloc', `memchr', `memcmp', `memcpy', `memset', `modf', `pow',
23479 `printf', `putchar', `puts', `scanf', `sinh', `sin', `snprintf',
23480 `sprintf', `sqrt', `sscanf', `strcat', `strchr', `strcmp', `strcpy',
23481 `strcspn', `strlen', `strncat', `strncmp', `strncpy', `strpbrk',
23482 `strrchr', `strspn', `strstr', `tanh', `tan', `vfprintf', `vprintf' and
23483 `vsprintf' are all recognized as built-in functions unless
23484 `-fno-builtin' is specified (or `-fno-builtin-FUNCTION' is specified
23485 for an individual function). All of these functions have corresponding
23486 versions prefixed with `__builtin_'.
23488 GCC provides built-in versions of the ISO C99 floating point comparison
23489 macros that avoid raising exceptions for unordered operands. They have
23490 the same names as the standard macros ( `isgreater', `isgreaterequal',
23491 `isless', `islessequal', `islessgreater', and `isunordered') , with
23492 `__builtin_' prefixed. We intend for a library implementor to be able
23493 to simply `#define' each standard macro to its built-in equivalent. In
23494 the same fashion, GCC provides `fpclassify', `isfinite', `isinf_sign'
23495 and `isnormal' built-ins used with `__builtin_' prefixed. The `isinf'
23496 and `isnan' builtins appear both with and without the `__builtin_'
23499 -- Built-in Function: int __builtin_types_compatible_p (TYPE1, TYPE2)
23500 You can use the built-in function `__builtin_types_compatible_p' to
23501 determine whether two types are the same.
23503 This built-in function returns 1 if the unqualified versions of the
23504 types TYPE1 and TYPE2 (which are types, not expressions) are
23505 compatible, 0 otherwise. The result of this built-in function can
23506 be used in integer constant expressions.
23508 This built-in function ignores top level qualifiers (e.g., `const',
23509 `volatile'). For example, `int' is equivalent to `const int'.
23511 The type `int[]' and `int[5]' are compatible. On the other hand,
23512 `int' and `char *' are not compatible, even if the size of their
23513 types, on the particular architecture are the same. Also, the
23514 amount of pointer indirection is taken into account when
23515 determining similarity. Consequently, `short *' is not similar to
23516 `short **'. Furthermore, two types that are typedefed are
23517 considered compatible if their underlying types are compatible.
23519 An `enum' type is not considered to be compatible with another
23520 `enum' type even if both are compatible with the same integer
23521 type; this is what the C standard specifies. For example, `enum
23522 {foo, bar}' is not similar to `enum {hot, dog}'.
23524 You would typically use this function in code whose execution
23525 varies depending on the arguments' types. For example:
23529 typeof (x) tmp = (x); \
23530 if (__builtin_types_compatible_p (typeof (x), long double)) \
23531 tmp = foo_long_double (tmp); \
23532 else if (__builtin_types_compatible_p (typeof (x), double)) \
23533 tmp = foo_double (tmp); \
23534 else if (__builtin_types_compatible_p (typeof (x), float)) \
23535 tmp = foo_float (tmp); \
23541 _Note:_ This construct is only available for C.
23544 -- Built-in Function: TYPE __builtin_choose_expr (CONST_EXP, EXP1,
23546 You can use the built-in function `__builtin_choose_expr' to
23547 evaluate code depending on the value of a constant expression.
23548 This built-in function returns EXP1 if CONST_EXP, which is a
23549 constant expression that must be able to be determined at compile
23550 time, is nonzero. Otherwise it returns 0.
23552 This built-in function is analogous to the `? :' operator in C,
23553 except that the expression returned has its type unaltered by
23554 promotion rules. Also, the built-in function does not evaluate
23555 the expression that was not chosen. For example, if CONST_EXP
23556 evaluates to true, EXP2 is not evaluated even if it has
23559 This built-in function can return an lvalue if the chosen argument
23562 If EXP1 is returned, the return type is the same as EXP1's type.
23563 Similarly, if EXP2 is returned, its return type is the same as
23569 __builtin_choose_expr ( \
23570 __builtin_types_compatible_p (typeof (x), double), \
23572 __builtin_choose_expr ( \
23573 __builtin_types_compatible_p (typeof (x), float), \
23575 /* The void expression results in a compile-time error \
23576 when assigning the result to something. */ \
23579 _Note:_ This construct is only available for C. Furthermore, the
23580 unused expression (EXP1 or EXP2 depending on the value of
23581 CONST_EXP) may still generate syntax errors. This may change in
23585 -- Built-in Function: int __builtin_constant_p (EXP)
23586 You can use the built-in function `__builtin_constant_p' to
23587 determine if a value is known to be constant at compile-time and
23588 hence that GCC can perform constant-folding on expressions
23589 involving that value. The argument of the function is the value
23590 to test. The function returns the integer 1 if the argument is
23591 known to be a compile-time constant and 0 if it is not known to be
23592 a compile-time constant. A return of 0 does not indicate that the
23593 value is _not_ a constant, but merely that GCC cannot prove it is
23594 a constant with the specified value of the `-O' option.
23596 You would typically use this function in an embedded application
23597 where memory was a critical resource. If you have some complex
23598 calculation, you may want it to be folded if it involves
23599 constants, but need to call a function if it does not. For
23602 #define Scale_Value(X) \
23603 (__builtin_constant_p (X) \
23604 ? ((X) * SCALE + OFFSET) : Scale (X))
23606 You may use this built-in function in either a macro or an inline
23607 function. However, if you use it in an inlined function and pass
23608 an argument of the function as the argument to the built-in, GCC
23609 will never return 1 when you call the inline function with a
23610 string constant or compound literal (*note Compound Literals::)
23611 and will not return 1 when you pass a constant numeric value to
23612 the inline function unless you specify the `-O' option.
23614 You may also use `__builtin_constant_p' in initializers for static
23615 data. For instance, you can write
23617 static const int table[] = {
23618 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
23622 This is an acceptable initializer even if EXPRESSION is not a
23623 constant expression. GCC must be more conservative about
23624 evaluating the built-in in this case, because it has no
23625 opportunity to perform optimization.
23627 Previous versions of GCC did not accept this built-in in data
23628 initializers. The earliest version where it is completely safe is
23631 -- Built-in Function: long __builtin_expect (long EXP, long C)
23632 You may use `__builtin_expect' to provide the compiler with branch
23633 prediction information. In general, you should prefer to use
23634 actual profile feedback for this (`-fprofile-arcs'), as
23635 programmers are notoriously bad at predicting how their programs
23636 actually perform. However, there are applications in which this
23637 data is hard to collect.
23639 The return value is the value of EXP, which should be an integral
23640 expression. The semantics of the built-in are that it is expected
23641 that EXP == C. For example:
23643 if (__builtin_expect (x, 0))
23646 would indicate that we do not expect to call `foo', since we
23647 expect `x' to be zero. Since you are limited to integral
23648 expressions for EXP, you should use constructions such as
23650 if (__builtin_expect (ptr != NULL, 1))
23653 when testing pointer or floating-point values.
23655 -- Built-in Function: void __builtin_trap (void)
23656 This function causes the program to exit abnormally. GCC
23657 implements this function by using a target-dependent mechanism
23658 (such as intentionally executing an illegal instruction) or by
23659 calling `abort'. The mechanism used may vary from release to
23660 release so you should not rely on any particular implementation.
23662 -- Built-in Function: void __builtin___clear_cache (char *BEGIN, char
23664 This function is used to flush the processor's instruction cache
23665 for the region of memory between BEGIN inclusive and END
23666 exclusive. Some targets require that the instruction cache be
23667 flushed, after modifying memory containing code, in order to obtain
23668 deterministic behavior.
23670 If the target does not require instruction cache flushes,
23671 `__builtin___clear_cache' has no effect. Otherwise either
23672 instructions are emitted in-line to clear the instruction cache or
23673 a call to the `__clear_cache' function in libgcc is made.
23675 -- Built-in Function: void __builtin_prefetch (const void *ADDR, ...)
23676 This function is used to minimize cache-miss latency by moving
23677 data into a cache before it is accessed. You can insert calls to
23678 `__builtin_prefetch' into code for which you know addresses of
23679 data in memory that is likely to be accessed soon. If the target
23680 supports them, data prefetch instructions will be generated. If
23681 the prefetch is done early enough before the access then the data
23682 will be in the cache by the time it is accessed.
23684 The value of ADDR is the address of the memory to prefetch. There
23685 are two optional arguments, RW and LOCALITY. The value of RW is a
23686 compile-time constant one or zero; one means that the prefetch is
23687 preparing for a write to the memory address and zero, the default,
23688 means that the prefetch is preparing for a read. The value
23689 LOCALITY must be a compile-time constant integer between zero and
23690 three. A value of zero means that the data has no temporal
23691 locality, so it need not be left in the cache after the access. A
23692 value of three means that the data has a high degree of temporal
23693 locality and should be left in all levels of cache possible.
23694 Values of one and two mean, respectively, a low or moderate degree
23695 of temporal locality. The default is three.
23697 for (i = 0; i < n; i++)
23699 a[i] = a[i] + b[i];
23700 __builtin_prefetch (&a[i+j], 1, 1);
23701 __builtin_prefetch (&b[i+j], 0, 1);
23705 Data prefetch does not generate faults if ADDR is invalid, but the
23706 address expression itself must be valid. For example, a prefetch
23707 of `p->next' will not fault if `p->next' is not a valid address,
23708 but evaluation will fault if `p' is not a valid address.
23710 If the target does not support data prefetch, the address
23711 expression is evaluated if it includes side effects but no other
23712 code is generated and GCC does not issue a warning.
23714 -- Built-in Function: double __builtin_huge_val (void)
23715 Returns a positive infinity, if supported by the floating-point
23716 format, else `DBL_MAX'. This function is suitable for
23717 implementing the ISO C macro `HUGE_VAL'.
23719 -- Built-in Function: float __builtin_huge_valf (void)
23720 Similar to `__builtin_huge_val', except the return type is `float'.
23722 -- Built-in Function: long double __builtin_huge_vall (void)
23723 Similar to `__builtin_huge_val', except the return type is `long
23726 -- Built-in Function: int __builtin_fpclassify (int, int, int, int,
23728 This built-in implements the C99 fpclassify functionality. The
23729 first five int arguments should be the target library's notion of
23730 the possible FP classes and are used for return values. They must
23731 be constant values and they must appear in this order: `FP_NAN',
23732 `FP_INFINITE', `FP_NORMAL', `FP_SUBNORMAL' and `FP_ZERO'. The
23733 ellipsis is for exactly one floating point value to classify. GCC
23734 treats the last argument as type-generic, which means it does not
23735 do default promotion from float to double.
23737 -- Built-in Function: double __builtin_inf (void)
23738 Similar to `__builtin_huge_val', except a warning is generated if
23739 the target floating-point format does not support infinities.
23741 -- Built-in Function: _Decimal32 __builtin_infd32 (void)
23742 Similar to `__builtin_inf', except the return type is `_Decimal32'.
23744 -- Built-in Function: _Decimal64 __builtin_infd64 (void)
23745 Similar to `__builtin_inf', except the return type is `_Decimal64'.
23747 -- Built-in Function: _Decimal128 __builtin_infd128 (void)
23748 Similar to `__builtin_inf', except the return type is
23751 -- Built-in Function: float __builtin_inff (void)
23752 Similar to `__builtin_inf', except the return type is `float'.
23753 This function is suitable for implementing the ISO C99 macro
23756 -- Built-in Function: long double __builtin_infl (void)
23757 Similar to `__builtin_inf', except the return type is `long
23760 -- Built-in Function: int __builtin_isinf_sign (...)
23761 Similar to `isinf', except the return value will be negative for
23762 an argument of `-Inf'. Note while the parameter list is an
23763 ellipsis, this function only accepts exactly one floating point
23764 argument. GCC treats this parameter as type-generic, which means
23765 it does not do default promotion from float to double.
23767 -- Built-in Function: double __builtin_nan (const char *str)
23768 This is an implementation of the ISO C99 function `nan'.
23770 Since ISO C99 defines this function in terms of `strtod', which we
23771 do not implement, a description of the parsing is in order. The
23772 string is parsed as by `strtol'; that is, the base is recognized by
23773 leading `0' or `0x' prefixes. The number parsed is placed in the
23774 significand such that the least significant bit of the number is
23775 at the least significant bit of the significand. The number is
23776 truncated to fit the significand field provided. The significand
23777 is forced to be a quiet NaN.
23779 This function, if given a string literal all of which would have
23780 been consumed by strtol, is evaluated early enough that it is
23781 considered a compile-time constant.
23783 -- Built-in Function: _Decimal32 __builtin_nand32 (const char *str)
23784 Similar to `__builtin_nan', except the return type is `_Decimal32'.
23786 -- Built-in Function: _Decimal64 __builtin_nand64 (const char *str)
23787 Similar to `__builtin_nan', except the return type is `_Decimal64'.
23789 -- Built-in Function: _Decimal128 __builtin_nand128 (const char *str)
23790 Similar to `__builtin_nan', except the return type is
23793 -- Built-in Function: float __builtin_nanf (const char *str)
23794 Similar to `__builtin_nan', except the return type is `float'.
23796 -- Built-in Function: long double __builtin_nanl (const char *str)
23797 Similar to `__builtin_nan', except the return type is `long
23800 -- Built-in Function: double __builtin_nans (const char *str)
23801 Similar to `__builtin_nan', except the significand is forced to be
23802 a signaling NaN. The `nans' function is proposed by WG14 N965.
23804 -- Built-in Function: float __builtin_nansf (const char *str)
23805 Similar to `__builtin_nans', except the return type is `float'.
23807 -- Built-in Function: long double __builtin_nansl (const char *str)
23808 Similar to `__builtin_nans', except the return type is `long
23811 -- Built-in Function: int __builtin_ffs (unsigned int x)
23812 Returns one plus the index of the least significant 1-bit of X, or
23813 if X is zero, returns zero.
23815 -- Built-in Function: int __builtin_clz (unsigned int x)
23816 Returns the number of leading 0-bits in X, starting at the most
23817 significant bit position. If X is 0, the result is undefined.
23819 -- Built-in Function: int __builtin_ctz (unsigned int x)
23820 Returns the number of trailing 0-bits in X, starting at the least
23821 significant bit position. If X is 0, the result is undefined.
23823 -- Built-in Function: int __builtin_popcount (unsigned int x)
23824 Returns the number of 1-bits in X.
23826 -- Built-in Function: int __builtin_parity (unsigned int x)
23827 Returns the parity of X, i.e. the number of 1-bits in X modulo 2.
23829 -- Built-in Function: int __builtin_ffsl (unsigned long)
23830 Similar to `__builtin_ffs', except the argument type is `unsigned
23833 -- Built-in Function: int __builtin_clzl (unsigned long)
23834 Similar to `__builtin_clz', except the argument type is `unsigned
23837 -- Built-in Function: int __builtin_ctzl (unsigned long)
23838 Similar to `__builtin_ctz', except the argument type is `unsigned
23841 -- Built-in Function: int __builtin_popcountl (unsigned long)
23842 Similar to `__builtin_popcount', except the argument type is
23845 -- Built-in Function: int __builtin_parityl (unsigned long)
23846 Similar to `__builtin_parity', except the argument type is
23849 -- Built-in Function: int __builtin_ffsll (unsigned long long)
23850 Similar to `__builtin_ffs', except the argument type is `unsigned
23853 -- Built-in Function: int __builtin_clzll (unsigned long long)
23854 Similar to `__builtin_clz', except the argument type is `unsigned
23857 -- Built-in Function: int __builtin_ctzll (unsigned long long)
23858 Similar to `__builtin_ctz', except the argument type is `unsigned
23861 -- Built-in Function: int __builtin_popcountll (unsigned long long)
23862 Similar to `__builtin_popcount', except the argument type is
23863 `unsigned long long'.
23865 -- Built-in Function: int __builtin_parityll (unsigned long long)
23866 Similar to `__builtin_parity', except the argument type is
23867 `unsigned long long'.
23869 -- Built-in Function: double __builtin_powi (double, int)
23870 Returns the first argument raised to the power of the second.
23871 Unlike the `pow' function no guarantees about precision and
23874 -- Built-in Function: float __builtin_powif (float, int)
23875 Similar to `__builtin_powi', except the argument and return types
23878 -- Built-in Function: long double __builtin_powil (long double, int)
23879 Similar to `__builtin_powi', except the argument and return types
23882 -- Built-in Function: int32_t __builtin_bswap32 (int32_t x)
23883 Returns X with the order of the bytes reversed; for example,
23884 `0xaabbccdd' becomes `0xddccbbaa'. Byte here always means exactly
23887 -- Built-in Function: int64_t __builtin_bswap64 (int64_t x)
23888 Similar to `__builtin_bswap32', except the argument and return
23892 File: gcc.info, Node: Target Builtins, Next: Target Format Checks, Prev: Other Builtins, Up: C Extensions
23894 5.50 Built-in Functions Specific to Particular Target Machines
23895 ==============================================================
23897 On some target machines, GCC supports many built-in functions specific
23898 to those machines. Generally these generate calls to specific machine
23899 instructions, but allow the compiler to schedule those calls.
23903 * Alpha Built-in Functions::
23904 * ARM iWMMXt Built-in Functions::
23905 * ARM NEON Intrinsics::
23906 * Blackfin Built-in Functions::
23907 * FR-V Built-in Functions::
23908 * X86 Built-in Functions::
23909 * MIPS DSP Built-in Functions::
23910 * MIPS Paired-Single Support::
23911 * MIPS Loongson Built-in Functions::
23912 * Other MIPS Built-in Functions::
23913 * picoChip Built-in Functions::
23914 * PowerPC AltiVec Built-in Functions::
23915 * SPARC VIS Built-in Functions::
23916 * SPU Built-in Functions::
23919 File: gcc.info, Node: Alpha Built-in Functions, Next: ARM iWMMXt Built-in Functions, Up: Target Builtins
23921 5.50.1 Alpha Built-in Functions
23922 -------------------------------
23924 These built-in functions are available for the Alpha family of
23925 processors, depending on the command-line switches used.
23927 The following built-in functions are always available. They all
23928 generate the machine instruction that is part of the name.
23930 long __builtin_alpha_implver (void)
23931 long __builtin_alpha_rpcc (void)
23932 long __builtin_alpha_amask (long)
23933 long __builtin_alpha_cmpbge (long, long)
23934 long __builtin_alpha_extbl (long, long)
23935 long __builtin_alpha_extwl (long, long)
23936 long __builtin_alpha_extll (long, long)
23937 long __builtin_alpha_extql (long, long)
23938 long __builtin_alpha_extwh (long, long)
23939 long __builtin_alpha_extlh (long, long)
23940 long __builtin_alpha_extqh (long, long)
23941 long __builtin_alpha_insbl (long, long)
23942 long __builtin_alpha_inswl (long, long)
23943 long __builtin_alpha_insll (long, long)
23944 long __builtin_alpha_insql (long, long)
23945 long __builtin_alpha_inswh (long, long)
23946 long __builtin_alpha_inslh (long, long)
23947 long __builtin_alpha_insqh (long, long)
23948 long __builtin_alpha_mskbl (long, long)
23949 long __builtin_alpha_mskwl (long, long)
23950 long __builtin_alpha_mskll (long, long)
23951 long __builtin_alpha_mskql (long, long)
23952 long __builtin_alpha_mskwh (long, long)
23953 long __builtin_alpha_msklh (long, long)
23954 long __builtin_alpha_mskqh (long, long)
23955 long __builtin_alpha_umulh (long, long)
23956 long __builtin_alpha_zap (long, long)
23957 long __builtin_alpha_zapnot (long, long)
23959 The following built-in functions are always with `-mmax' or
23960 `-mcpu=CPU' where CPU is `pca56' or later. They all generate the
23961 machine instruction that is part of the name.
23963 long __builtin_alpha_pklb (long)
23964 long __builtin_alpha_pkwb (long)
23965 long __builtin_alpha_unpkbl (long)
23966 long __builtin_alpha_unpkbw (long)
23967 long __builtin_alpha_minub8 (long, long)
23968 long __builtin_alpha_minsb8 (long, long)
23969 long __builtin_alpha_minuw4 (long, long)
23970 long __builtin_alpha_minsw4 (long, long)
23971 long __builtin_alpha_maxub8 (long, long)
23972 long __builtin_alpha_maxsb8 (long, long)
23973 long __builtin_alpha_maxuw4 (long, long)
23974 long __builtin_alpha_maxsw4 (long, long)
23975 long __builtin_alpha_perr (long, long)
23977 The following built-in functions are always with `-mcix' or
23978 `-mcpu=CPU' where CPU is `ev67' or later. They all generate the
23979 machine instruction that is part of the name.
23981 long __builtin_alpha_cttz (long)
23982 long __builtin_alpha_ctlz (long)
23983 long __builtin_alpha_ctpop (long)
23985 The following builtins are available on systems that use the OSF/1
23986 PALcode. Normally they invoke the `rduniq' and `wruniq' PAL calls, but
23987 when invoked with `-mtls-kernel', they invoke `rdval' and `wrval'.
23989 void *__builtin_thread_pointer (void)
23990 void __builtin_set_thread_pointer (void *)
23993 File: gcc.info, Node: ARM iWMMXt Built-in Functions, Next: ARM NEON Intrinsics, Prev: Alpha Built-in Functions, Up: Target Builtins
23995 5.50.2 ARM iWMMXt Built-in Functions
23996 ------------------------------------
23998 These built-in functions are available for the ARM family of processors
23999 when the `-mcpu=iwmmxt' switch is used:
24001 typedef int v2si __attribute__ ((vector_size (8)));
24002 typedef short v4hi __attribute__ ((vector_size (8)));
24003 typedef char v8qi __attribute__ ((vector_size (8)));
24005 int __builtin_arm_getwcx (int)
24006 void __builtin_arm_setwcx (int, int)
24007 int __builtin_arm_textrmsb (v8qi, int)
24008 int __builtin_arm_textrmsh (v4hi, int)
24009 int __builtin_arm_textrmsw (v2si, int)
24010 int __builtin_arm_textrmub (v8qi, int)
24011 int __builtin_arm_textrmuh (v4hi, int)
24012 int __builtin_arm_textrmuw (v2si, int)
24013 v8qi __builtin_arm_tinsrb (v8qi, int)
24014 v4hi __builtin_arm_tinsrh (v4hi, int)
24015 v2si __builtin_arm_tinsrw (v2si, int)
24016 long long __builtin_arm_tmia (long long, int, int)
24017 long long __builtin_arm_tmiabb (long long, int, int)
24018 long long __builtin_arm_tmiabt (long long, int, int)
24019 long long __builtin_arm_tmiaph (long long, int, int)
24020 long long __builtin_arm_tmiatb (long long, int, int)
24021 long long __builtin_arm_tmiatt (long long, int, int)
24022 int __builtin_arm_tmovmskb (v8qi)
24023 int __builtin_arm_tmovmskh (v4hi)
24024 int __builtin_arm_tmovmskw (v2si)
24025 long long __builtin_arm_waccb (v8qi)
24026 long long __builtin_arm_wacch (v4hi)
24027 long long __builtin_arm_waccw (v2si)
24028 v8qi __builtin_arm_waddb (v8qi, v8qi)
24029 v8qi __builtin_arm_waddbss (v8qi, v8qi)
24030 v8qi __builtin_arm_waddbus (v8qi, v8qi)
24031 v4hi __builtin_arm_waddh (v4hi, v4hi)
24032 v4hi __builtin_arm_waddhss (v4hi, v4hi)
24033 v4hi __builtin_arm_waddhus (v4hi, v4hi)
24034 v2si __builtin_arm_waddw (v2si, v2si)
24035 v2si __builtin_arm_waddwss (v2si, v2si)
24036 v2si __builtin_arm_waddwus (v2si, v2si)
24037 v8qi __builtin_arm_walign (v8qi, v8qi, int)
24038 long long __builtin_arm_wand(long long, long long)
24039 long long __builtin_arm_wandn (long long, long long)
24040 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
24041 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
24042 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
24043 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
24044 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
24045 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
24046 v2si __builtin_arm_wcmpeqw (v2si, v2si)
24047 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
24048 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
24049 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
24050 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
24051 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
24052 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
24053 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
24054 long long __builtin_arm_wmacsz (v4hi, v4hi)
24055 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
24056 long long __builtin_arm_wmacuz (v4hi, v4hi)
24057 v4hi __builtin_arm_wmadds (v4hi, v4hi)
24058 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
24059 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
24060 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
24061 v2si __builtin_arm_wmaxsw (v2si, v2si)
24062 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
24063 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
24064 v2si __builtin_arm_wmaxuw (v2si, v2si)
24065 v8qi __builtin_arm_wminsb (v8qi, v8qi)
24066 v4hi __builtin_arm_wminsh (v4hi, v4hi)
24067 v2si __builtin_arm_wminsw (v2si, v2si)
24068 v8qi __builtin_arm_wminub (v8qi, v8qi)
24069 v4hi __builtin_arm_wminuh (v4hi, v4hi)
24070 v2si __builtin_arm_wminuw (v2si, v2si)
24071 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
24072 v4hi __builtin_arm_wmulul (v4hi, v4hi)
24073 v4hi __builtin_arm_wmulum (v4hi, v4hi)
24074 long long __builtin_arm_wor (long long, long long)
24075 v2si __builtin_arm_wpackdss (long long, long long)
24076 v2si __builtin_arm_wpackdus (long long, long long)
24077 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
24078 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
24079 v4hi __builtin_arm_wpackwss (v2si, v2si)
24080 v4hi __builtin_arm_wpackwus (v2si, v2si)
24081 long long __builtin_arm_wrord (long long, long long)
24082 long long __builtin_arm_wrordi (long long, int)
24083 v4hi __builtin_arm_wrorh (v4hi, long long)
24084 v4hi __builtin_arm_wrorhi (v4hi, int)
24085 v2si __builtin_arm_wrorw (v2si, long long)
24086 v2si __builtin_arm_wrorwi (v2si, int)
24087 v2si __builtin_arm_wsadb (v8qi, v8qi)
24088 v2si __builtin_arm_wsadbz (v8qi, v8qi)
24089 v2si __builtin_arm_wsadh (v4hi, v4hi)
24090 v2si __builtin_arm_wsadhz (v4hi, v4hi)
24091 v4hi __builtin_arm_wshufh (v4hi, int)
24092 long long __builtin_arm_wslld (long long, long long)
24093 long long __builtin_arm_wslldi (long long, int)
24094 v4hi __builtin_arm_wsllh (v4hi, long long)
24095 v4hi __builtin_arm_wsllhi (v4hi, int)
24096 v2si __builtin_arm_wsllw (v2si, long long)
24097 v2si __builtin_arm_wsllwi (v2si, int)
24098 long long __builtin_arm_wsrad (long long, long long)
24099 long long __builtin_arm_wsradi (long long, int)
24100 v4hi __builtin_arm_wsrah (v4hi, long long)
24101 v4hi __builtin_arm_wsrahi (v4hi, int)
24102 v2si __builtin_arm_wsraw (v2si, long long)
24103 v2si __builtin_arm_wsrawi (v2si, int)
24104 long long __builtin_arm_wsrld (long long, long long)
24105 long long __builtin_arm_wsrldi (long long, int)
24106 v4hi __builtin_arm_wsrlh (v4hi, long long)
24107 v4hi __builtin_arm_wsrlhi (v4hi, int)
24108 v2si __builtin_arm_wsrlw (v2si, long long)
24109 v2si __builtin_arm_wsrlwi (v2si, int)
24110 v8qi __builtin_arm_wsubb (v8qi, v8qi)
24111 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
24112 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
24113 v4hi __builtin_arm_wsubh (v4hi, v4hi)
24114 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
24115 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
24116 v2si __builtin_arm_wsubw (v2si, v2si)
24117 v2si __builtin_arm_wsubwss (v2si, v2si)
24118 v2si __builtin_arm_wsubwus (v2si, v2si)
24119 v4hi __builtin_arm_wunpckehsb (v8qi)
24120 v2si __builtin_arm_wunpckehsh (v4hi)
24121 long long __builtin_arm_wunpckehsw (v2si)
24122 v4hi __builtin_arm_wunpckehub (v8qi)
24123 v2si __builtin_arm_wunpckehuh (v4hi)
24124 long long __builtin_arm_wunpckehuw (v2si)
24125 v4hi __builtin_arm_wunpckelsb (v8qi)
24126 v2si __builtin_arm_wunpckelsh (v4hi)
24127 long long __builtin_arm_wunpckelsw (v2si)
24128 v4hi __builtin_arm_wunpckelub (v8qi)
24129 v2si __builtin_arm_wunpckeluh (v4hi)
24130 long long __builtin_arm_wunpckeluw (v2si)
24131 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
24132 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
24133 v2si __builtin_arm_wunpckihw (v2si, v2si)
24134 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
24135 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
24136 v2si __builtin_arm_wunpckilw (v2si, v2si)
24137 long long __builtin_arm_wxor (long long, long long)
24138 long long __builtin_arm_wzero ()
24141 File: gcc.info, Node: ARM NEON Intrinsics, Next: Blackfin Built-in Functions, Prev: ARM iWMMXt Built-in Functions, Up: Target Builtins
24143 5.50.3 ARM NEON Intrinsics
24144 --------------------------
24146 These built-in intrinsics for the ARM Advanced SIMD extension are
24147 available when the `-mfpu=neon' switch is used:
24152 * uint32x2_t vadd_u32 (uint32x2_t, uint32x2_t)
24153 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
24155 * uint16x4_t vadd_u16 (uint16x4_t, uint16x4_t)
24156 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
24158 * uint8x8_t vadd_u8 (uint8x8_t, uint8x8_t)
24159 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
24161 * int32x2_t vadd_s32 (int32x2_t, int32x2_t)
24162 _Form of expected instruction(s):_ `vadd.i32 D0, D0, D0'
24164 * int16x4_t vadd_s16 (int16x4_t, int16x4_t)
24165 _Form of expected instruction(s):_ `vadd.i16 D0, D0, D0'
24167 * int8x8_t vadd_s8 (int8x8_t, int8x8_t)
24168 _Form of expected instruction(s):_ `vadd.i8 D0, D0, D0'
24170 * uint64x1_t vadd_u64 (uint64x1_t, uint64x1_t)
24171 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
24173 * int64x1_t vadd_s64 (int64x1_t, int64x1_t)
24174 _Form of expected instruction(s):_ `vadd.i64 D0, D0, D0'
24176 * float32x2_t vadd_f32 (float32x2_t, float32x2_t)
24177 _Form of expected instruction(s):_ `vadd.f32 D0, D0, D0'
24179 * uint32x4_t vaddq_u32 (uint32x4_t, uint32x4_t)
24180 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
24182 * uint16x8_t vaddq_u16 (uint16x8_t, uint16x8_t)
24183 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
24185 * uint8x16_t vaddq_u8 (uint8x16_t, uint8x16_t)
24186 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
24188 * int32x4_t vaddq_s32 (int32x4_t, int32x4_t)
24189 _Form of expected instruction(s):_ `vadd.i32 Q0, Q0, Q0'
24191 * int16x8_t vaddq_s16 (int16x8_t, int16x8_t)
24192 _Form of expected instruction(s):_ `vadd.i16 Q0, Q0, Q0'
24194 * int8x16_t vaddq_s8 (int8x16_t, int8x16_t)
24195 _Form of expected instruction(s):_ `vadd.i8 Q0, Q0, Q0'
24197 * uint64x2_t vaddq_u64 (uint64x2_t, uint64x2_t)
24198 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
24200 * int64x2_t vaddq_s64 (int64x2_t, int64x2_t)
24201 _Form of expected instruction(s):_ `vadd.i64 Q0, Q0, Q0'
24203 * float32x4_t vaddq_f32 (float32x4_t, float32x4_t)
24204 _Form of expected instruction(s):_ `vadd.f32 Q0, Q0, Q0'
24206 * uint64x2_t vaddl_u32 (uint32x2_t, uint32x2_t)
24207 _Form of expected instruction(s):_ `vaddl.u32 Q0, D0, D0'
24209 * uint32x4_t vaddl_u16 (uint16x4_t, uint16x4_t)
24210 _Form of expected instruction(s):_ `vaddl.u16 Q0, D0, D0'
24212 * uint16x8_t vaddl_u8 (uint8x8_t, uint8x8_t)
24213 _Form of expected instruction(s):_ `vaddl.u8 Q0, D0, D0'
24215 * int64x2_t vaddl_s32 (int32x2_t, int32x2_t)
24216 _Form of expected instruction(s):_ `vaddl.s32 Q0, D0, D0'
24218 * int32x4_t vaddl_s16 (int16x4_t, int16x4_t)
24219 _Form of expected instruction(s):_ `vaddl.s16 Q0, D0, D0'
24221 * int16x8_t vaddl_s8 (int8x8_t, int8x8_t)
24222 _Form of expected instruction(s):_ `vaddl.s8 Q0, D0, D0'
24224 * uint64x2_t vaddw_u32 (uint64x2_t, uint32x2_t)
24225 _Form of expected instruction(s):_ `vaddw.u32 Q0, Q0, D0'
24227 * uint32x4_t vaddw_u16 (uint32x4_t, uint16x4_t)
24228 _Form of expected instruction(s):_ `vaddw.u16 Q0, Q0, D0'
24230 * uint16x8_t vaddw_u8 (uint16x8_t, uint8x8_t)
24231 _Form of expected instruction(s):_ `vaddw.u8 Q0, Q0, D0'
24233 * int64x2_t vaddw_s32 (int64x2_t, int32x2_t)
24234 _Form of expected instruction(s):_ `vaddw.s32 Q0, Q0, D0'
24236 * int32x4_t vaddw_s16 (int32x4_t, int16x4_t)
24237 _Form of expected instruction(s):_ `vaddw.s16 Q0, Q0, D0'
24239 * int16x8_t vaddw_s8 (int16x8_t, int8x8_t)
24240 _Form of expected instruction(s):_ `vaddw.s8 Q0, Q0, D0'
24242 * uint32x2_t vhadd_u32 (uint32x2_t, uint32x2_t)
24243 _Form of expected instruction(s):_ `vhadd.u32 D0, D0, D0'
24245 * uint16x4_t vhadd_u16 (uint16x4_t, uint16x4_t)
24246 _Form of expected instruction(s):_ `vhadd.u16 D0, D0, D0'
24248 * uint8x8_t vhadd_u8 (uint8x8_t, uint8x8_t)
24249 _Form of expected instruction(s):_ `vhadd.u8 D0, D0, D0'
24251 * int32x2_t vhadd_s32 (int32x2_t, int32x2_t)
24252 _Form of expected instruction(s):_ `vhadd.s32 D0, D0, D0'
24254 * int16x4_t vhadd_s16 (int16x4_t, int16x4_t)
24255 _Form of expected instruction(s):_ `vhadd.s16 D0, D0, D0'
24257 * int8x8_t vhadd_s8 (int8x8_t, int8x8_t)
24258 _Form of expected instruction(s):_ `vhadd.s8 D0, D0, D0'
24260 * uint32x4_t vhaddq_u32 (uint32x4_t, uint32x4_t)
24261 _Form of expected instruction(s):_ `vhadd.u32 Q0, Q0, Q0'
24263 * uint16x8_t vhaddq_u16 (uint16x8_t, uint16x8_t)
24264 _Form of expected instruction(s):_ `vhadd.u16 Q0, Q0, Q0'
24266 * uint8x16_t vhaddq_u8 (uint8x16_t, uint8x16_t)
24267 _Form of expected instruction(s):_ `vhadd.u8 Q0, Q0, Q0'
24269 * int32x4_t vhaddq_s32 (int32x4_t, int32x4_t)
24270 _Form of expected instruction(s):_ `vhadd.s32 Q0, Q0, Q0'
24272 * int16x8_t vhaddq_s16 (int16x8_t, int16x8_t)
24273 _Form of expected instruction(s):_ `vhadd.s16 Q0, Q0, Q0'
24275 * int8x16_t vhaddq_s8 (int8x16_t, int8x16_t)
24276 _Form of expected instruction(s):_ `vhadd.s8 Q0, Q0, Q0'
24278 * uint32x2_t vrhadd_u32 (uint32x2_t, uint32x2_t)
24279 _Form of expected instruction(s):_ `vrhadd.u32 D0, D0, D0'
24281 * uint16x4_t vrhadd_u16 (uint16x4_t, uint16x4_t)
24282 _Form of expected instruction(s):_ `vrhadd.u16 D0, D0, D0'
24284 * uint8x8_t vrhadd_u8 (uint8x8_t, uint8x8_t)
24285 _Form of expected instruction(s):_ `vrhadd.u8 D0, D0, D0'
24287 * int32x2_t vrhadd_s32 (int32x2_t, int32x2_t)
24288 _Form of expected instruction(s):_ `vrhadd.s32 D0, D0, D0'
24290 * int16x4_t vrhadd_s16 (int16x4_t, int16x4_t)
24291 _Form of expected instruction(s):_ `vrhadd.s16 D0, D0, D0'
24293 * int8x8_t vrhadd_s8 (int8x8_t, int8x8_t)
24294 _Form of expected instruction(s):_ `vrhadd.s8 D0, D0, D0'
24296 * uint32x4_t vrhaddq_u32 (uint32x4_t, uint32x4_t)
24297 _Form of expected instruction(s):_ `vrhadd.u32 Q0, Q0, Q0'
24299 * uint16x8_t vrhaddq_u16 (uint16x8_t, uint16x8_t)
24300 _Form of expected instruction(s):_ `vrhadd.u16 Q0, Q0, Q0'
24302 * uint8x16_t vrhaddq_u8 (uint8x16_t, uint8x16_t)
24303 _Form of expected instruction(s):_ `vrhadd.u8 Q0, Q0, Q0'
24305 * int32x4_t vrhaddq_s32 (int32x4_t, int32x4_t)
24306 _Form of expected instruction(s):_ `vrhadd.s32 Q0, Q0, Q0'
24308 * int16x8_t vrhaddq_s16 (int16x8_t, int16x8_t)
24309 _Form of expected instruction(s):_ `vrhadd.s16 Q0, Q0, Q0'
24311 * int8x16_t vrhaddq_s8 (int8x16_t, int8x16_t)
24312 _Form of expected instruction(s):_ `vrhadd.s8 Q0, Q0, Q0'
24314 * uint32x2_t vqadd_u32 (uint32x2_t, uint32x2_t)
24315 _Form of expected instruction(s):_ `vqadd.u32 D0, D0, D0'
24317 * uint16x4_t vqadd_u16 (uint16x4_t, uint16x4_t)
24318 _Form of expected instruction(s):_ `vqadd.u16 D0, D0, D0'
24320 * uint8x8_t vqadd_u8 (uint8x8_t, uint8x8_t)
24321 _Form of expected instruction(s):_ `vqadd.u8 D0, D0, D0'
24323 * int32x2_t vqadd_s32 (int32x2_t, int32x2_t)
24324 _Form of expected instruction(s):_ `vqadd.s32 D0, D0, D0'
24326 * int16x4_t vqadd_s16 (int16x4_t, int16x4_t)
24327 _Form of expected instruction(s):_ `vqadd.s16 D0, D0, D0'
24329 * int8x8_t vqadd_s8 (int8x8_t, int8x8_t)
24330 _Form of expected instruction(s):_ `vqadd.s8 D0, D0, D0'
24332 * uint64x1_t vqadd_u64 (uint64x1_t, uint64x1_t)
24333 _Form of expected instruction(s):_ `vqadd.u64 D0, D0, D0'
24335 * int64x1_t vqadd_s64 (int64x1_t, int64x1_t)
24336 _Form of expected instruction(s):_ `vqadd.s64 D0, D0, D0'
24338 * uint32x4_t vqaddq_u32 (uint32x4_t, uint32x4_t)
24339 _Form of expected instruction(s):_ `vqadd.u32 Q0, Q0, Q0'
24341 * uint16x8_t vqaddq_u16 (uint16x8_t, uint16x8_t)
24342 _Form of expected instruction(s):_ `vqadd.u16 Q0, Q0, Q0'
24344 * uint8x16_t vqaddq_u8 (uint8x16_t, uint8x16_t)
24345 _Form of expected instruction(s):_ `vqadd.u8 Q0, Q0, Q0'
24347 * int32x4_t vqaddq_s32 (int32x4_t, int32x4_t)
24348 _Form of expected instruction(s):_ `vqadd.s32 Q0, Q0, Q0'
24350 * int16x8_t vqaddq_s16 (int16x8_t, int16x8_t)
24351 _Form of expected instruction(s):_ `vqadd.s16 Q0, Q0, Q0'
24353 * int8x16_t vqaddq_s8 (int8x16_t, int8x16_t)
24354 _Form of expected instruction(s):_ `vqadd.s8 Q0, Q0, Q0'
24356 * uint64x2_t vqaddq_u64 (uint64x2_t, uint64x2_t)
24357 _Form of expected instruction(s):_ `vqadd.u64 Q0, Q0, Q0'
24359 * int64x2_t vqaddq_s64 (int64x2_t, int64x2_t)
24360 _Form of expected instruction(s):_ `vqadd.s64 Q0, Q0, Q0'
24362 * uint32x2_t vaddhn_u64 (uint64x2_t, uint64x2_t)
24363 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
24365 * uint16x4_t vaddhn_u32 (uint32x4_t, uint32x4_t)
24366 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
24368 * uint8x8_t vaddhn_u16 (uint16x8_t, uint16x8_t)
24369 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
24371 * int32x2_t vaddhn_s64 (int64x2_t, int64x2_t)
24372 _Form of expected instruction(s):_ `vaddhn.i64 D0, Q0, Q0'
24374 * int16x4_t vaddhn_s32 (int32x4_t, int32x4_t)
24375 _Form of expected instruction(s):_ `vaddhn.i32 D0, Q0, Q0'
24377 * int8x8_t vaddhn_s16 (int16x8_t, int16x8_t)
24378 _Form of expected instruction(s):_ `vaddhn.i16 D0, Q0, Q0'
24380 * uint32x2_t vraddhn_u64 (uint64x2_t, uint64x2_t)
24381 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
24383 * uint16x4_t vraddhn_u32 (uint32x4_t, uint32x4_t)
24384 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
24386 * uint8x8_t vraddhn_u16 (uint16x8_t, uint16x8_t)
24387 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
24389 * int32x2_t vraddhn_s64 (int64x2_t, int64x2_t)
24390 _Form of expected instruction(s):_ `vraddhn.i64 D0, Q0, Q0'
24392 * int16x4_t vraddhn_s32 (int32x4_t, int32x4_t)
24393 _Form of expected instruction(s):_ `vraddhn.i32 D0, Q0, Q0'
24395 * int8x8_t vraddhn_s16 (int16x8_t, int16x8_t)
24396 _Form of expected instruction(s):_ `vraddhn.i16 D0, Q0, Q0'
24398 5.50.3.2 Multiplication
24399 .......................
24401 * uint32x2_t vmul_u32 (uint32x2_t, uint32x2_t)
24402 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
24404 * uint16x4_t vmul_u16 (uint16x4_t, uint16x4_t)
24405 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
24407 * uint8x8_t vmul_u8 (uint8x8_t, uint8x8_t)
24408 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
24410 * int32x2_t vmul_s32 (int32x2_t, int32x2_t)
24411 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0'
24413 * int16x4_t vmul_s16 (int16x4_t, int16x4_t)
24414 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0'
24416 * int8x8_t vmul_s8 (int8x8_t, int8x8_t)
24417 _Form of expected instruction(s):_ `vmul.i8 D0, D0, D0'
24419 * float32x2_t vmul_f32 (float32x2_t, float32x2_t)
24420 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0'
24422 * poly8x8_t vmul_p8 (poly8x8_t, poly8x8_t)
24423 _Form of expected instruction(s):_ `vmul.p8 D0, D0, D0'
24425 * uint32x4_t vmulq_u32 (uint32x4_t, uint32x4_t)
24426 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
24428 * uint16x8_t vmulq_u16 (uint16x8_t, uint16x8_t)
24429 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
24431 * uint8x16_t vmulq_u8 (uint8x16_t, uint8x16_t)
24432 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
24434 * int32x4_t vmulq_s32 (int32x4_t, int32x4_t)
24435 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, Q0'
24437 * int16x8_t vmulq_s16 (int16x8_t, int16x8_t)
24438 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, Q0'
24440 * int8x16_t vmulq_s8 (int8x16_t, int8x16_t)
24441 _Form of expected instruction(s):_ `vmul.i8 Q0, Q0, Q0'
24443 * float32x4_t vmulq_f32 (float32x4_t, float32x4_t)
24444 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, Q0'
24446 * poly8x16_t vmulq_p8 (poly8x16_t, poly8x16_t)
24447 _Form of expected instruction(s):_ `vmul.p8 Q0, Q0, Q0'
24449 * int32x2_t vqdmulh_s32 (int32x2_t, int32x2_t)
24450 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0'
24452 * int16x4_t vqdmulh_s16 (int16x4_t, int16x4_t)
24453 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0'
24455 * int32x4_t vqdmulhq_s32 (int32x4_t, int32x4_t)
24456 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, Q0'
24458 * int16x8_t vqdmulhq_s16 (int16x8_t, int16x8_t)
24459 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, Q0'
24461 * int32x2_t vqrdmulh_s32 (int32x2_t, int32x2_t)
24462 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0'
24464 * int16x4_t vqrdmulh_s16 (int16x4_t, int16x4_t)
24465 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0'
24467 * int32x4_t vqrdmulhq_s32 (int32x4_t, int32x4_t)
24468 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, Q0'
24470 * int16x8_t vqrdmulhq_s16 (int16x8_t, int16x8_t)
24471 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, Q0'
24473 * uint64x2_t vmull_u32 (uint32x2_t, uint32x2_t)
24474 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0'
24476 * uint32x4_t vmull_u16 (uint16x4_t, uint16x4_t)
24477 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0'
24479 * uint16x8_t vmull_u8 (uint8x8_t, uint8x8_t)
24480 _Form of expected instruction(s):_ `vmull.u8 Q0, D0, D0'
24482 * int64x2_t vmull_s32 (int32x2_t, int32x2_t)
24483 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0'
24485 * int32x4_t vmull_s16 (int16x4_t, int16x4_t)
24486 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0'
24488 * int16x8_t vmull_s8 (int8x8_t, int8x8_t)
24489 _Form of expected instruction(s):_ `vmull.s8 Q0, D0, D0'
24491 * poly16x8_t vmull_p8 (poly8x8_t, poly8x8_t)
24492 _Form of expected instruction(s):_ `vmull.p8 Q0, D0, D0'
24494 * int64x2_t vqdmull_s32 (int32x2_t, int32x2_t)
24495 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0'
24497 * int32x4_t vqdmull_s16 (int16x4_t, int16x4_t)
24498 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0'
24500 5.50.3.3 Multiply-accumulate
24501 ............................
24503 * uint32x2_t vmla_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
24504 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
24506 * uint16x4_t vmla_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
24507 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
24509 * uint8x8_t vmla_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
24510 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
24512 * int32x2_t vmla_s32 (int32x2_t, int32x2_t, int32x2_t)
24513 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0'
24515 * int16x4_t vmla_s16 (int16x4_t, int16x4_t, int16x4_t)
24516 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0'
24518 * int8x8_t vmla_s8 (int8x8_t, int8x8_t, int8x8_t)
24519 _Form of expected instruction(s):_ `vmla.i8 D0, D0, D0'
24521 * float32x2_t vmla_f32 (float32x2_t, float32x2_t, float32x2_t)
24522 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0'
24524 * uint32x4_t vmlaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
24525 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
24527 * uint16x8_t vmlaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
24528 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
24530 * uint8x16_t vmlaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
24531 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
24533 * int32x4_t vmlaq_s32 (int32x4_t, int32x4_t, int32x4_t)
24534 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, Q0'
24536 * int16x8_t vmlaq_s16 (int16x8_t, int16x8_t, int16x8_t)
24537 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, Q0'
24539 * int8x16_t vmlaq_s8 (int8x16_t, int8x16_t, int8x16_t)
24540 _Form of expected instruction(s):_ `vmla.i8 Q0, Q0, Q0'
24542 * float32x4_t vmlaq_f32 (float32x4_t, float32x4_t, float32x4_t)
24543 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, Q0'
24545 * uint64x2_t vmlal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
24546 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0'
24548 * uint32x4_t vmlal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
24549 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0'
24551 * uint16x8_t vmlal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
24552 _Form of expected instruction(s):_ `vmlal.u8 Q0, D0, D0'
24554 * int64x2_t vmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
24555 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0'
24557 * int32x4_t vmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
24558 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0'
24560 * int16x8_t vmlal_s8 (int16x8_t, int8x8_t, int8x8_t)
24561 _Form of expected instruction(s):_ `vmlal.s8 Q0, D0, D0'
24563 * int64x2_t vqdmlal_s32 (int64x2_t, int32x2_t, int32x2_t)
24564 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0'
24566 * int32x4_t vqdmlal_s16 (int32x4_t, int16x4_t, int16x4_t)
24567 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0'
24569 5.50.3.4 Multiply-subtract
24570 ..........................
24572 * uint32x2_t vmls_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
24573 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
24575 * uint16x4_t vmls_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
24576 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
24578 * uint8x8_t vmls_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
24579 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
24581 * int32x2_t vmls_s32 (int32x2_t, int32x2_t, int32x2_t)
24582 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0'
24584 * int16x4_t vmls_s16 (int16x4_t, int16x4_t, int16x4_t)
24585 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0'
24587 * int8x8_t vmls_s8 (int8x8_t, int8x8_t, int8x8_t)
24588 _Form of expected instruction(s):_ `vmls.i8 D0, D0, D0'
24590 * float32x2_t vmls_f32 (float32x2_t, float32x2_t, float32x2_t)
24591 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0'
24593 * uint32x4_t vmlsq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
24594 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
24596 * uint16x8_t vmlsq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
24597 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
24599 * uint8x16_t vmlsq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
24600 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
24602 * int32x4_t vmlsq_s32 (int32x4_t, int32x4_t, int32x4_t)
24603 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, Q0'
24605 * int16x8_t vmlsq_s16 (int16x8_t, int16x8_t, int16x8_t)
24606 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, Q0'
24608 * int8x16_t vmlsq_s8 (int8x16_t, int8x16_t, int8x16_t)
24609 _Form of expected instruction(s):_ `vmls.i8 Q0, Q0, Q0'
24611 * float32x4_t vmlsq_f32 (float32x4_t, float32x4_t, float32x4_t)
24612 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, Q0'
24614 * uint64x2_t vmlsl_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
24615 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0'
24617 * uint32x4_t vmlsl_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
24618 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0'
24620 * uint16x8_t vmlsl_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
24621 _Form of expected instruction(s):_ `vmlsl.u8 Q0, D0, D0'
24623 * int64x2_t vmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
24624 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0'
24626 * int32x4_t vmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
24627 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0'
24629 * int16x8_t vmlsl_s8 (int16x8_t, int8x8_t, int8x8_t)
24630 _Form of expected instruction(s):_ `vmlsl.s8 Q0, D0, D0'
24632 * int64x2_t vqdmlsl_s32 (int64x2_t, int32x2_t, int32x2_t)
24633 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0'
24635 * int32x4_t vqdmlsl_s16 (int32x4_t, int16x4_t, int16x4_t)
24636 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0'
24638 5.50.3.5 Subtraction
24639 ....................
24641 * uint32x2_t vsub_u32 (uint32x2_t, uint32x2_t)
24642 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
24644 * uint16x4_t vsub_u16 (uint16x4_t, uint16x4_t)
24645 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
24647 * uint8x8_t vsub_u8 (uint8x8_t, uint8x8_t)
24648 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
24650 * int32x2_t vsub_s32 (int32x2_t, int32x2_t)
24651 _Form of expected instruction(s):_ `vsub.i32 D0, D0, D0'
24653 * int16x4_t vsub_s16 (int16x4_t, int16x4_t)
24654 _Form of expected instruction(s):_ `vsub.i16 D0, D0, D0'
24656 * int8x8_t vsub_s8 (int8x8_t, int8x8_t)
24657 _Form of expected instruction(s):_ `vsub.i8 D0, D0, D0'
24659 * uint64x1_t vsub_u64 (uint64x1_t, uint64x1_t)
24660 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
24662 * int64x1_t vsub_s64 (int64x1_t, int64x1_t)
24663 _Form of expected instruction(s):_ `vsub.i64 D0, D0, D0'
24665 * float32x2_t vsub_f32 (float32x2_t, float32x2_t)
24666 _Form of expected instruction(s):_ `vsub.f32 D0, D0, D0'
24668 * uint32x4_t vsubq_u32 (uint32x4_t, uint32x4_t)
24669 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
24671 * uint16x8_t vsubq_u16 (uint16x8_t, uint16x8_t)
24672 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
24674 * uint8x16_t vsubq_u8 (uint8x16_t, uint8x16_t)
24675 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
24677 * int32x4_t vsubq_s32 (int32x4_t, int32x4_t)
24678 _Form of expected instruction(s):_ `vsub.i32 Q0, Q0, Q0'
24680 * int16x8_t vsubq_s16 (int16x8_t, int16x8_t)
24681 _Form of expected instruction(s):_ `vsub.i16 Q0, Q0, Q0'
24683 * int8x16_t vsubq_s8 (int8x16_t, int8x16_t)
24684 _Form of expected instruction(s):_ `vsub.i8 Q0, Q0, Q0'
24686 * uint64x2_t vsubq_u64 (uint64x2_t, uint64x2_t)
24687 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
24689 * int64x2_t vsubq_s64 (int64x2_t, int64x2_t)
24690 _Form of expected instruction(s):_ `vsub.i64 Q0, Q0, Q0'
24692 * float32x4_t vsubq_f32 (float32x4_t, float32x4_t)
24693 _Form of expected instruction(s):_ `vsub.f32 Q0, Q0, Q0'
24695 * uint64x2_t vsubl_u32 (uint32x2_t, uint32x2_t)
24696 _Form of expected instruction(s):_ `vsubl.u32 Q0, D0, D0'
24698 * uint32x4_t vsubl_u16 (uint16x4_t, uint16x4_t)
24699 _Form of expected instruction(s):_ `vsubl.u16 Q0, D0, D0'
24701 * uint16x8_t vsubl_u8 (uint8x8_t, uint8x8_t)
24702 _Form of expected instruction(s):_ `vsubl.u8 Q0, D0, D0'
24704 * int64x2_t vsubl_s32 (int32x2_t, int32x2_t)
24705 _Form of expected instruction(s):_ `vsubl.s32 Q0, D0, D0'
24707 * int32x4_t vsubl_s16 (int16x4_t, int16x4_t)
24708 _Form of expected instruction(s):_ `vsubl.s16 Q0, D0, D0'
24710 * int16x8_t vsubl_s8 (int8x8_t, int8x8_t)
24711 _Form of expected instruction(s):_ `vsubl.s8 Q0, D0, D0'
24713 * uint64x2_t vsubw_u32 (uint64x2_t, uint32x2_t)
24714 _Form of expected instruction(s):_ `vsubw.u32 Q0, Q0, D0'
24716 * uint32x4_t vsubw_u16 (uint32x4_t, uint16x4_t)
24717 _Form of expected instruction(s):_ `vsubw.u16 Q0, Q0, D0'
24719 * uint16x8_t vsubw_u8 (uint16x8_t, uint8x8_t)
24720 _Form of expected instruction(s):_ `vsubw.u8 Q0, Q0, D0'
24722 * int64x2_t vsubw_s32 (int64x2_t, int32x2_t)
24723 _Form of expected instruction(s):_ `vsubw.s32 Q0, Q0, D0'
24725 * int32x4_t vsubw_s16 (int32x4_t, int16x4_t)
24726 _Form of expected instruction(s):_ `vsubw.s16 Q0, Q0, D0'
24728 * int16x8_t vsubw_s8 (int16x8_t, int8x8_t)
24729 _Form of expected instruction(s):_ `vsubw.s8 Q0, Q0, D0'
24731 * uint32x2_t vhsub_u32 (uint32x2_t, uint32x2_t)
24732 _Form of expected instruction(s):_ `vhsub.u32 D0, D0, D0'
24734 * uint16x4_t vhsub_u16 (uint16x4_t, uint16x4_t)
24735 _Form of expected instruction(s):_ `vhsub.u16 D0, D0, D0'
24737 * uint8x8_t vhsub_u8 (uint8x8_t, uint8x8_t)
24738 _Form of expected instruction(s):_ `vhsub.u8 D0, D0, D0'
24740 * int32x2_t vhsub_s32 (int32x2_t, int32x2_t)
24741 _Form of expected instruction(s):_ `vhsub.s32 D0, D0, D0'
24743 * int16x4_t vhsub_s16 (int16x4_t, int16x4_t)
24744 _Form of expected instruction(s):_ `vhsub.s16 D0, D0, D0'
24746 * int8x8_t vhsub_s8 (int8x8_t, int8x8_t)
24747 _Form of expected instruction(s):_ `vhsub.s8 D0, D0, D0'
24749 * uint32x4_t vhsubq_u32 (uint32x4_t, uint32x4_t)
24750 _Form of expected instruction(s):_ `vhsub.u32 Q0, Q0, Q0'
24752 * uint16x8_t vhsubq_u16 (uint16x8_t, uint16x8_t)
24753 _Form of expected instruction(s):_ `vhsub.u16 Q0, Q0, Q0'
24755 * uint8x16_t vhsubq_u8 (uint8x16_t, uint8x16_t)
24756 _Form of expected instruction(s):_ `vhsub.u8 Q0, Q0, Q0'
24758 * int32x4_t vhsubq_s32 (int32x4_t, int32x4_t)
24759 _Form of expected instruction(s):_ `vhsub.s32 Q0, Q0, Q0'
24761 * int16x8_t vhsubq_s16 (int16x8_t, int16x8_t)
24762 _Form of expected instruction(s):_ `vhsub.s16 Q0, Q0, Q0'
24764 * int8x16_t vhsubq_s8 (int8x16_t, int8x16_t)
24765 _Form of expected instruction(s):_ `vhsub.s8 Q0, Q0, Q0'
24767 * uint32x2_t vqsub_u32 (uint32x2_t, uint32x2_t)
24768 _Form of expected instruction(s):_ `vqsub.u32 D0, D0, D0'
24770 * uint16x4_t vqsub_u16 (uint16x4_t, uint16x4_t)
24771 _Form of expected instruction(s):_ `vqsub.u16 D0, D0, D0'
24773 * uint8x8_t vqsub_u8 (uint8x8_t, uint8x8_t)
24774 _Form of expected instruction(s):_ `vqsub.u8 D0, D0, D0'
24776 * int32x2_t vqsub_s32 (int32x2_t, int32x2_t)
24777 _Form of expected instruction(s):_ `vqsub.s32 D0, D0, D0'
24779 * int16x4_t vqsub_s16 (int16x4_t, int16x4_t)
24780 _Form of expected instruction(s):_ `vqsub.s16 D0, D0, D0'
24782 * int8x8_t vqsub_s8 (int8x8_t, int8x8_t)
24783 _Form of expected instruction(s):_ `vqsub.s8 D0, D0, D0'
24785 * uint64x1_t vqsub_u64 (uint64x1_t, uint64x1_t)
24786 _Form of expected instruction(s):_ `vqsub.u64 D0, D0, D0'
24788 * int64x1_t vqsub_s64 (int64x1_t, int64x1_t)
24789 _Form of expected instruction(s):_ `vqsub.s64 D0, D0, D0'
24791 * uint32x4_t vqsubq_u32 (uint32x4_t, uint32x4_t)
24792 _Form of expected instruction(s):_ `vqsub.u32 Q0, Q0, Q0'
24794 * uint16x8_t vqsubq_u16 (uint16x8_t, uint16x8_t)
24795 _Form of expected instruction(s):_ `vqsub.u16 Q0, Q0, Q0'
24797 * uint8x16_t vqsubq_u8 (uint8x16_t, uint8x16_t)
24798 _Form of expected instruction(s):_ `vqsub.u8 Q0, Q0, Q0'
24800 * int32x4_t vqsubq_s32 (int32x4_t, int32x4_t)
24801 _Form of expected instruction(s):_ `vqsub.s32 Q0, Q0, Q0'
24803 * int16x8_t vqsubq_s16 (int16x8_t, int16x8_t)
24804 _Form of expected instruction(s):_ `vqsub.s16 Q0, Q0, Q0'
24806 * int8x16_t vqsubq_s8 (int8x16_t, int8x16_t)
24807 _Form of expected instruction(s):_ `vqsub.s8 Q0, Q0, Q0'
24809 * uint64x2_t vqsubq_u64 (uint64x2_t, uint64x2_t)
24810 _Form of expected instruction(s):_ `vqsub.u64 Q0, Q0, Q0'
24812 * int64x2_t vqsubq_s64 (int64x2_t, int64x2_t)
24813 _Form of expected instruction(s):_ `vqsub.s64 Q0, Q0, Q0'
24815 * uint32x2_t vsubhn_u64 (uint64x2_t, uint64x2_t)
24816 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
24818 * uint16x4_t vsubhn_u32 (uint32x4_t, uint32x4_t)
24819 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
24821 * uint8x8_t vsubhn_u16 (uint16x8_t, uint16x8_t)
24822 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
24824 * int32x2_t vsubhn_s64 (int64x2_t, int64x2_t)
24825 _Form of expected instruction(s):_ `vsubhn.i64 D0, Q0, Q0'
24827 * int16x4_t vsubhn_s32 (int32x4_t, int32x4_t)
24828 _Form of expected instruction(s):_ `vsubhn.i32 D0, Q0, Q0'
24830 * int8x8_t vsubhn_s16 (int16x8_t, int16x8_t)
24831 _Form of expected instruction(s):_ `vsubhn.i16 D0, Q0, Q0'
24833 * uint32x2_t vrsubhn_u64 (uint64x2_t, uint64x2_t)
24834 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
24836 * uint16x4_t vrsubhn_u32 (uint32x4_t, uint32x4_t)
24837 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
24839 * uint8x8_t vrsubhn_u16 (uint16x8_t, uint16x8_t)
24840 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
24842 * int32x2_t vrsubhn_s64 (int64x2_t, int64x2_t)
24843 _Form of expected instruction(s):_ `vrsubhn.i64 D0, Q0, Q0'
24845 * int16x4_t vrsubhn_s32 (int32x4_t, int32x4_t)
24846 _Form of expected instruction(s):_ `vrsubhn.i32 D0, Q0, Q0'
24848 * int8x8_t vrsubhn_s16 (int16x8_t, int16x8_t)
24849 _Form of expected instruction(s):_ `vrsubhn.i16 D0, Q0, Q0'
24851 5.50.3.6 Comparison (equal-to)
24852 ..............................
24854 * uint32x2_t vceq_u32 (uint32x2_t, uint32x2_t)
24855 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
24857 * uint16x4_t vceq_u16 (uint16x4_t, uint16x4_t)
24858 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
24860 * uint8x8_t vceq_u8 (uint8x8_t, uint8x8_t)
24861 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24863 * uint32x2_t vceq_s32 (int32x2_t, int32x2_t)
24864 _Form of expected instruction(s):_ `vceq.i32 D0, D0, D0'
24866 * uint16x4_t vceq_s16 (int16x4_t, int16x4_t)
24867 _Form of expected instruction(s):_ `vceq.i16 D0, D0, D0'
24869 * uint8x8_t vceq_s8 (int8x8_t, int8x8_t)
24870 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24872 * uint32x2_t vceq_f32 (float32x2_t, float32x2_t)
24873 _Form of expected instruction(s):_ `vceq.f32 D0, D0, D0'
24875 * uint8x8_t vceq_p8 (poly8x8_t, poly8x8_t)
24876 _Form of expected instruction(s):_ `vceq.i8 D0, D0, D0'
24878 * uint32x4_t vceqq_u32 (uint32x4_t, uint32x4_t)
24879 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
24881 * uint16x8_t vceqq_u16 (uint16x8_t, uint16x8_t)
24882 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
24884 * uint8x16_t vceqq_u8 (uint8x16_t, uint8x16_t)
24885 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24887 * uint32x4_t vceqq_s32 (int32x4_t, int32x4_t)
24888 _Form of expected instruction(s):_ `vceq.i32 Q0, Q0, Q0'
24890 * uint16x8_t vceqq_s16 (int16x8_t, int16x8_t)
24891 _Form of expected instruction(s):_ `vceq.i16 Q0, Q0, Q0'
24893 * uint8x16_t vceqq_s8 (int8x16_t, int8x16_t)
24894 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24896 * uint32x4_t vceqq_f32 (float32x4_t, float32x4_t)
24897 _Form of expected instruction(s):_ `vceq.f32 Q0, Q0, Q0'
24899 * uint8x16_t vceqq_p8 (poly8x16_t, poly8x16_t)
24900 _Form of expected instruction(s):_ `vceq.i8 Q0, Q0, Q0'
24902 5.50.3.7 Comparison (greater-than-or-equal-to)
24903 ..............................................
24905 * uint32x2_t vcge_u32 (uint32x2_t, uint32x2_t)
24906 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
24908 * uint16x4_t vcge_u16 (uint16x4_t, uint16x4_t)
24909 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
24911 * uint8x8_t vcge_u8 (uint8x8_t, uint8x8_t)
24912 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
24914 * uint32x2_t vcge_s32 (int32x2_t, int32x2_t)
24915 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
24917 * uint16x4_t vcge_s16 (int16x4_t, int16x4_t)
24918 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
24920 * uint8x8_t vcge_s8 (int8x8_t, int8x8_t)
24921 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
24923 * uint32x2_t vcge_f32 (float32x2_t, float32x2_t)
24924 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
24926 * uint32x4_t vcgeq_u32 (uint32x4_t, uint32x4_t)
24927 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
24929 * uint16x8_t vcgeq_u16 (uint16x8_t, uint16x8_t)
24930 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
24932 * uint8x16_t vcgeq_u8 (uint8x16_t, uint8x16_t)
24933 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
24935 * uint32x4_t vcgeq_s32 (int32x4_t, int32x4_t)
24936 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
24938 * uint16x8_t vcgeq_s16 (int16x8_t, int16x8_t)
24939 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
24941 * uint8x16_t vcgeq_s8 (int8x16_t, int8x16_t)
24942 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
24944 * uint32x4_t vcgeq_f32 (float32x4_t, float32x4_t)
24945 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
24947 5.50.3.8 Comparison (less-than-or-equal-to)
24948 ...........................................
24950 * uint32x2_t vcle_u32 (uint32x2_t, uint32x2_t)
24951 _Form of expected instruction(s):_ `vcge.u32 D0, D0, D0'
24953 * uint16x4_t vcle_u16 (uint16x4_t, uint16x4_t)
24954 _Form of expected instruction(s):_ `vcge.u16 D0, D0, D0'
24956 * uint8x8_t vcle_u8 (uint8x8_t, uint8x8_t)
24957 _Form of expected instruction(s):_ `vcge.u8 D0, D0, D0'
24959 * uint32x2_t vcle_s32 (int32x2_t, int32x2_t)
24960 _Form of expected instruction(s):_ `vcge.s32 D0, D0, D0'
24962 * uint16x4_t vcle_s16 (int16x4_t, int16x4_t)
24963 _Form of expected instruction(s):_ `vcge.s16 D0, D0, D0'
24965 * uint8x8_t vcle_s8 (int8x8_t, int8x8_t)
24966 _Form of expected instruction(s):_ `vcge.s8 D0, D0, D0'
24968 * uint32x2_t vcle_f32 (float32x2_t, float32x2_t)
24969 _Form of expected instruction(s):_ `vcge.f32 D0, D0, D0'
24971 * uint32x4_t vcleq_u32 (uint32x4_t, uint32x4_t)
24972 _Form of expected instruction(s):_ `vcge.u32 Q0, Q0, Q0'
24974 * uint16x8_t vcleq_u16 (uint16x8_t, uint16x8_t)
24975 _Form of expected instruction(s):_ `vcge.u16 Q0, Q0, Q0'
24977 * uint8x16_t vcleq_u8 (uint8x16_t, uint8x16_t)
24978 _Form of expected instruction(s):_ `vcge.u8 Q0, Q0, Q0'
24980 * uint32x4_t vcleq_s32 (int32x4_t, int32x4_t)
24981 _Form of expected instruction(s):_ `vcge.s32 Q0, Q0, Q0'
24983 * uint16x8_t vcleq_s16 (int16x8_t, int16x8_t)
24984 _Form of expected instruction(s):_ `vcge.s16 Q0, Q0, Q0'
24986 * uint8x16_t vcleq_s8 (int8x16_t, int8x16_t)
24987 _Form of expected instruction(s):_ `vcge.s8 Q0, Q0, Q0'
24989 * uint32x4_t vcleq_f32 (float32x4_t, float32x4_t)
24990 _Form of expected instruction(s):_ `vcge.f32 Q0, Q0, Q0'
24992 5.50.3.9 Comparison (greater-than)
24993 ..................................
24995 * uint32x2_t vcgt_u32 (uint32x2_t, uint32x2_t)
24996 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
24998 * uint16x4_t vcgt_u16 (uint16x4_t, uint16x4_t)
24999 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
25001 * uint8x8_t vcgt_u8 (uint8x8_t, uint8x8_t)
25002 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
25004 * uint32x2_t vcgt_s32 (int32x2_t, int32x2_t)
25005 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
25007 * uint16x4_t vcgt_s16 (int16x4_t, int16x4_t)
25008 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
25010 * uint8x8_t vcgt_s8 (int8x8_t, int8x8_t)
25011 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
25013 * uint32x2_t vcgt_f32 (float32x2_t, float32x2_t)
25014 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
25016 * uint32x4_t vcgtq_u32 (uint32x4_t, uint32x4_t)
25017 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
25019 * uint16x8_t vcgtq_u16 (uint16x8_t, uint16x8_t)
25020 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
25022 * uint8x16_t vcgtq_u8 (uint8x16_t, uint8x16_t)
25023 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
25025 * uint32x4_t vcgtq_s32 (int32x4_t, int32x4_t)
25026 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
25028 * uint16x8_t vcgtq_s16 (int16x8_t, int16x8_t)
25029 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
25031 * uint8x16_t vcgtq_s8 (int8x16_t, int8x16_t)
25032 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
25034 * uint32x4_t vcgtq_f32 (float32x4_t, float32x4_t)
25035 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
25037 5.50.3.10 Comparison (less-than)
25038 ................................
25040 * uint32x2_t vclt_u32 (uint32x2_t, uint32x2_t)
25041 _Form of expected instruction(s):_ `vcgt.u32 D0, D0, D0'
25043 * uint16x4_t vclt_u16 (uint16x4_t, uint16x4_t)
25044 _Form of expected instruction(s):_ `vcgt.u16 D0, D0, D0'
25046 * uint8x8_t vclt_u8 (uint8x8_t, uint8x8_t)
25047 _Form of expected instruction(s):_ `vcgt.u8 D0, D0, D0'
25049 * uint32x2_t vclt_s32 (int32x2_t, int32x2_t)
25050 _Form of expected instruction(s):_ `vcgt.s32 D0, D0, D0'
25052 * uint16x4_t vclt_s16 (int16x4_t, int16x4_t)
25053 _Form of expected instruction(s):_ `vcgt.s16 D0, D0, D0'
25055 * uint8x8_t vclt_s8 (int8x8_t, int8x8_t)
25056 _Form of expected instruction(s):_ `vcgt.s8 D0, D0, D0'
25058 * uint32x2_t vclt_f32 (float32x2_t, float32x2_t)
25059 _Form of expected instruction(s):_ `vcgt.f32 D0, D0, D0'
25061 * uint32x4_t vcltq_u32 (uint32x4_t, uint32x4_t)
25062 _Form of expected instruction(s):_ `vcgt.u32 Q0, Q0, Q0'
25064 * uint16x8_t vcltq_u16 (uint16x8_t, uint16x8_t)
25065 _Form of expected instruction(s):_ `vcgt.u16 Q0, Q0, Q0'
25067 * uint8x16_t vcltq_u8 (uint8x16_t, uint8x16_t)
25068 _Form of expected instruction(s):_ `vcgt.u8 Q0, Q0, Q0'
25070 * uint32x4_t vcltq_s32 (int32x4_t, int32x4_t)
25071 _Form of expected instruction(s):_ `vcgt.s32 Q0, Q0, Q0'
25073 * uint16x8_t vcltq_s16 (int16x8_t, int16x8_t)
25074 _Form of expected instruction(s):_ `vcgt.s16 Q0, Q0, Q0'
25076 * uint8x16_t vcltq_s8 (int8x16_t, int8x16_t)
25077 _Form of expected instruction(s):_ `vcgt.s8 Q0, Q0, Q0'
25079 * uint32x4_t vcltq_f32 (float32x4_t, float32x4_t)
25080 _Form of expected instruction(s):_ `vcgt.f32 Q0, Q0, Q0'
25082 5.50.3.11 Comparison (absolute greater-than-or-equal-to)
25083 ........................................................
25085 * uint32x2_t vcage_f32 (float32x2_t, float32x2_t)
25086 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
25088 * uint32x4_t vcageq_f32 (float32x4_t, float32x4_t)
25089 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
25091 5.50.3.12 Comparison (absolute less-than-or-equal-to)
25092 .....................................................
25094 * uint32x2_t vcale_f32 (float32x2_t, float32x2_t)
25095 _Form of expected instruction(s):_ `vacge.f32 D0, D0, D0'
25097 * uint32x4_t vcaleq_f32 (float32x4_t, float32x4_t)
25098 _Form of expected instruction(s):_ `vacge.f32 Q0, Q0, Q0'
25100 5.50.3.13 Comparison (absolute greater-than)
25101 ............................................
25103 * uint32x2_t vcagt_f32 (float32x2_t, float32x2_t)
25104 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
25106 * uint32x4_t vcagtq_f32 (float32x4_t, float32x4_t)
25107 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
25109 5.50.3.14 Comparison (absolute less-than)
25110 .........................................
25112 * uint32x2_t vcalt_f32 (float32x2_t, float32x2_t)
25113 _Form of expected instruction(s):_ `vacgt.f32 D0, D0, D0'
25115 * uint32x4_t vcaltq_f32 (float32x4_t, float32x4_t)
25116 _Form of expected instruction(s):_ `vacgt.f32 Q0, Q0, Q0'
25118 5.50.3.15 Test bits
25119 ...................
25121 * uint32x2_t vtst_u32 (uint32x2_t, uint32x2_t)
25122 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
25124 * uint16x4_t vtst_u16 (uint16x4_t, uint16x4_t)
25125 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
25127 * uint8x8_t vtst_u8 (uint8x8_t, uint8x8_t)
25128 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25130 * uint32x2_t vtst_s32 (int32x2_t, int32x2_t)
25131 _Form of expected instruction(s):_ `vtst.32 D0, D0, D0'
25133 * uint16x4_t vtst_s16 (int16x4_t, int16x4_t)
25134 _Form of expected instruction(s):_ `vtst.16 D0, D0, D0'
25136 * uint8x8_t vtst_s8 (int8x8_t, int8x8_t)
25137 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25139 * uint8x8_t vtst_p8 (poly8x8_t, poly8x8_t)
25140 _Form of expected instruction(s):_ `vtst.8 D0, D0, D0'
25142 * uint32x4_t vtstq_u32 (uint32x4_t, uint32x4_t)
25143 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
25145 * uint16x8_t vtstq_u16 (uint16x8_t, uint16x8_t)
25146 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
25148 * uint8x16_t vtstq_u8 (uint8x16_t, uint8x16_t)
25149 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25151 * uint32x4_t vtstq_s32 (int32x4_t, int32x4_t)
25152 _Form of expected instruction(s):_ `vtst.32 Q0, Q0, Q0'
25154 * uint16x8_t vtstq_s16 (int16x8_t, int16x8_t)
25155 _Form of expected instruction(s):_ `vtst.16 Q0, Q0, Q0'
25157 * uint8x16_t vtstq_s8 (int8x16_t, int8x16_t)
25158 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25160 * uint8x16_t vtstq_p8 (poly8x16_t, poly8x16_t)
25161 _Form of expected instruction(s):_ `vtst.8 Q0, Q0, Q0'
25163 5.50.3.16 Absolute difference
25164 .............................
25166 * uint32x2_t vabd_u32 (uint32x2_t, uint32x2_t)
25167 _Form of expected instruction(s):_ `vabd.u32 D0, D0, D0'
25169 * uint16x4_t vabd_u16 (uint16x4_t, uint16x4_t)
25170 _Form of expected instruction(s):_ `vabd.u16 D0, D0, D0'
25172 * uint8x8_t vabd_u8 (uint8x8_t, uint8x8_t)
25173 _Form of expected instruction(s):_ `vabd.u8 D0, D0, D0'
25175 * int32x2_t vabd_s32 (int32x2_t, int32x2_t)
25176 _Form of expected instruction(s):_ `vabd.s32 D0, D0, D0'
25178 * int16x4_t vabd_s16 (int16x4_t, int16x4_t)
25179 _Form of expected instruction(s):_ `vabd.s16 D0, D0, D0'
25181 * int8x8_t vabd_s8 (int8x8_t, int8x8_t)
25182 _Form of expected instruction(s):_ `vabd.s8 D0, D0, D0'
25184 * float32x2_t vabd_f32 (float32x2_t, float32x2_t)
25185 _Form of expected instruction(s):_ `vabd.f32 D0, D0, D0'
25187 * uint32x4_t vabdq_u32 (uint32x4_t, uint32x4_t)
25188 _Form of expected instruction(s):_ `vabd.u32 Q0, Q0, Q0'
25190 * uint16x8_t vabdq_u16 (uint16x8_t, uint16x8_t)
25191 _Form of expected instruction(s):_ `vabd.u16 Q0, Q0, Q0'
25193 * uint8x16_t vabdq_u8 (uint8x16_t, uint8x16_t)
25194 _Form of expected instruction(s):_ `vabd.u8 Q0, Q0, Q0'
25196 * int32x4_t vabdq_s32 (int32x4_t, int32x4_t)
25197 _Form of expected instruction(s):_ `vabd.s32 Q0, Q0, Q0'
25199 * int16x8_t vabdq_s16 (int16x8_t, int16x8_t)
25200 _Form of expected instruction(s):_ `vabd.s16 Q0, Q0, Q0'
25202 * int8x16_t vabdq_s8 (int8x16_t, int8x16_t)
25203 _Form of expected instruction(s):_ `vabd.s8 Q0, Q0, Q0'
25205 * float32x4_t vabdq_f32 (float32x4_t, float32x4_t)
25206 _Form of expected instruction(s):_ `vabd.f32 Q0, Q0, Q0'
25208 * uint64x2_t vabdl_u32 (uint32x2_t, uint32x2_t)
25209 _Form of expected instruction(s):_ `vabdl.u32 Q0, D0, D0'
25211 * uint32x4_t vabdl_u16 (uint16x4_t, uint16x4_t)
25212 _Form of expected instruction(s):_ `vabdl.u16 Q0, D0, D0'
25214 * uint16x8_t vabdl_u8 (uint8x8_t, uint8x8_t)
25215 _Form of expected instruction(s):_ `vabdl.u8 Q0, D0, D0'
25217 * int64x2_t vabdl_s32 (int32x2_t, int32x2_t)
25218 _Form of expected instruction(s):_ `vabdl.s32 Q0, D0, D0'
25220 * int32x4_t vabdl_s16 (int16x4_t, int16x4_t)
25221 _Form of expected instruction(s):_ `vabdl.s16 Q0, D0, D0'
25223 * int16x8_t vabdl_s8 (int8x8_t, int8x8_t)
25224 _Form of expected instruction(s):_ `vabdl.s8 Q0, D0, D0'
25226 5.50.3.17 Absolute difference and accumulate
25227 ............................................
25229 * uint32x2_t vaba_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
25230 _Form of expected instruction(s):_ `vaba.u32 D0, D0, D0'
25232 * uint16x4_t vaba_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
25233 _Form of expected instruction(s):_ `vaba.u16 D0, D0, D0'
25235 * uint8x8_t vaba_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
25236 _Form of expected instruction(s):_ `vaba.u8 D0, D0, D0'
25238 * int32x2_t vaba_s32 (int32x2_t, int32x2_t, int32x2_t)
25239 _Form of expected instruction(s):_ `vaba.s32 D0, D0, D0'
25241 * int16x4_t vaba_s16 (int16x4_t, int16x4_t, int16x4_t)
25242 _Form of expected instruction(s):_ `vaba.s16 D0, D0, D0'
25244 * int8x8_t vaba_s8 (int8x8_t, int8x8_t, int8x8_t)
25245 _Form of expected instruction(s):_ `vaba.s8 D0, D0, D0'
25247 * uint32x4_t vabaq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
25248 _Form of expected instruction(s):_ `vaba.u32 Q0, Q0, Q0'
25250 * uint16x8_t vabaq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
25251 _Form of expected instruction(s):_ `vaba.u16 Q0, Q0, Q0'
25253 * uint8x16_t vabaq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
25254 _Form of expected instruction(s):_ `vaba.u8 Q0, Q0, Q0'
25256 * int32x4_t vabaq_s32 (int32x4_t, int32x4_t, int32x4_t)
25257 _Form of expected instruction(s):_ `vaba.s32 Q0, Q0, Q0'
25259 * int16x8_t vabaq_s16 (int16x8_t, int16x8_t, int16x8_t)
25260 _Form of expected instruction(s):_ `vaba.s16 Q0, Q0, Q0'
25262 * int8x16_t vabaq_s8 (int8x16_t, int8x16_t, int8x16_t)
25263 _Form of expected instruction(s):_ `vaba.s8 Q0, Q0, Q0'
25265 * uint64x2_t vabal_u32 (uint64x2_t, uint32x2_t, uint32x2_t)
25266 _Form of expected instruction(s):_ `vabal.u32 Q0, D0, D0'
25268 * uint32x4_t vabal_u16 (uint32x4_t, uint16x4_t, uint16x4_t)
25269 _Form of expected instruction(s):_ `vabal.u16 Q0, D0, D0'
25271 * uint16x8_t vabal_u8 (uint16x8_t, uint8x8_t, uint8x8_t)
25272 _Form of expected instruction(s):_ `vabal.u8 Q0, D0, D0'
25274 * int64x2_t vabal_s32 (int64x2_t, int32x2_t, int32x2_t)
25275 _Form of expected instruction(s):_ `vabal.s32 Q0, D0, D0'
25277 * int32x4_t vabal_s16 (int32x4_t, int16x4_t, int16x4_t)
25278 _Form of expected instruction(s):_ `vabal.s16 Q0, D0, D0'
25280 * int16x8_t vabal_s8 (int16x8_t, int8x8_t, int8x8_t)
25281 _Form of expected instruction(s):_ `vabal.s8 Q0, D0, D0'
25286 * uint32x2_t vmax_u32 (uint32x2_t, uint32x2_t)
25287 _Form of expected instruction(s):_ `vmax.u32 D0, D0, D0'
25289 * uint16x4_t vmax_u16 (uint16x4_t, uint16x4_t)
25290 _Form of expected instruction(s):_ `vmax.u16 D0, D0, D0'
25292 * uint8x8_t vmax_u8 (uint8x8_t, uint8x8_t)
25293 _Form of expected instruction(s):_ `vmax.u8 D0, D0, D0'
25295 * int32x2_t vmax_s32 (int32x2_t, int32x2_t)
25296 _Form of expected instruction(s):_ `vmax.s32 D0, D0, D0'
25298 * int16x4_t vmax_s16 (int16x4_t, int16x4_t)
25299 _Form of expected instruction(s):_ `vmax.s16 D0, D0, D0'
25301 * int8x8_t vmax_s8 (int8x8_t, int8x8_t)
25302 _Form of expected instruction(s):_ `vmax.s8 D0, D0, D0'
25304 * float32x2_t vmax_f32 (float32x2_t, float32x2_t)
25305 _Form of expected instruction(s):_ `vmax.f32 D0, D0, D0'
25307 * uint32x4_t vmaxq_u32 (uint32x4_t, uint32x4_t)
25308 _Form of expected instruction(s):_ `vmax.u32 Q0, Q0, Q0'
25310 * uint16x8_t vmaxq_u16 (uint16x8_t, uint16x8_t)
25311 _Form of expected instruction(s):_ `vmax.u16 Q0, Q0, Q0'
25313 * uint8x16_t vmaxq_u8 (uint8x16_t, uint8x16_t)
25314 _Form of expected instruction(s):_ `vmax.u8 Q0, Q0, Q0'
25316 * int32x4_t vmaxq_s32 (int32x4_t, int32x4_t)
25317 _Form of expected instruction(s):_ `vmax.s32 Q0, Q0, Q0'
25319 * int16x8_t vmaxq_s16 (int16x8_t, int16x8_t)
25320 _Form of expected instruction(s):_ `vmax.s16 Q0, Q0, Q0'
25322 * int8x16_t vmaxq_s8 (int8x16_t, int8x16_t)
25323 _Form of expected instruction(s):_ `vmax.s8 Q0, Q0, Q0'
25325 * float32x4_t vmaxq_f32 (float32x4_t, float32x4_t)
25326 _Form of expected instruction(s):_ `vmax.f32 Q0, Q0, Q0'
25331 * uint32x2_t vmin_u32 (uint32x2_t, uint32x2_t)
25332 _Form of expected instruction(s):_ `vmin.u32 D0, D0, D0'
25334 * uint16x4_t vmin_u16 (uint16x4_t, uint16x4_t)
25335 _Form of expected instruction(s):_ `vmin.u16 D0, D0, D0'
25337 * uint8x8_t vmin_u8 (uint8x8_t, uint8x8_t)
25338 _Form of expected instruction(s):_ `vmin.u8 D0, D0, D0'
25340 * int32x2_t vmin_s32 (int32x2_t, int32x2_t)
25341 _Form of expected instruction(s):_ `vmin.s32 D0, D0, D0'
25343 * int16x4_t vmin_s16 (int16x4_t, int16x4_t)
25344 _Form of expected instruction(s):_ `vmin.s16 D0, D0, D0'
25346 * int8x8_t vmin_s8 (int8x8_t, int8x8_t)
25347 _Form of expected instruction(s):_ `vmin.s8 D0, D0, D0'
25349 * float32x2_t vmin_f32 (float32x2_t, float32x2_t)
25350 _Form of expected instruction(s):_ `vmin.f32 D0, D0, D0'
25352 * uint32x4_t vminq_u32 (uint32x4_t, uint32x4_t)
25353 _Form of expected instruction(s):_ `vmin.u32 Q0, Q0, Q0'
25355 * uint16x8_t vminq_u16 (uint16x8_t, uint16x8_t)
25356 _Form of expected instruction(s):_ `vmin.u16 Q0, Q0, Q0'
25358 * uint8x16_t vminq_u8 (uint8x16_t, uint8x16_t)
25359 _Form of expected instruction(s):_ `vmin.u8 Q0, Q0, Q0'
25361 * int32x4_t vminq_s32 (int32x4_t, int32x4_t)
25362 _Form of expected instruction(s):_ `vmin.s32 Q0, Q0, Q0'
25364 * int16x8_t vminq_s16 (int16x8_t, int16x8_t)
25365 _Form of expected instruction(s):_ `vmin.s16 Q0, Q0, Q0'
25367 * int8x16_t vminq_s8 (int8x16_t, int8x16_t)
25368 _Form of expected instruction(s):_ `vmin.s8 Q0, Q0, Q0'
25370 * float32x4_t vminq_f32 (float32x4_t, float32x4_t)
25371 _Form of expected instruction(s):_ `vmin.f32 Q0, Q0, Q0'
25373 5.50.3.20 Pairwise add
25374 ......................
25376 * uint32x2_t vpadd_u32 (uint32x2_t, uint32x2_t)
25377 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
25379 * uint16x4_t vpadd_u16 (uint16x4_t, uint16x4_t)
25380 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
25382 * uint8x8_t vpadd_u8 (uint8x8_t, uint8x8_t)
25383 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
25385 * int32x2_t vpadd_s32 (int32x2_t, int32x2_t)
25386 _Form of expected instruction(s):_ `vpadd.i32 D0, D0, D0'
25388 * int16x4_t vpadd_s16 (int16x4_t, int16x4_t)
25389 _Form of expected instruction(s):_ `vpadd.i16 D0, D0, D0'
25391 * int8x8_t vpadd_s8 (int8x8_t, int8x8_t)
25392 _Form of expected instruction(s):_ `vpadd.i8 D0, D0, D0'
25394 * float32x2_t vpadd_f32 (float32x2_t, float32x2_t)
25395 _Form of expected instruction(s):_ `vpadd.f32 D0, D0, D0'
25397 * uint64x1_t vpaddl_u32 (uint32x2_t)
25398 _Form of expected instruction(s):_ `vpaddl.u32 D0, D0'
25400 * uint32x2_t vpaddl_u16 (uint16x4_t)
25401 _Form of expected instruction(s):_ `vpaddl.u16 D0, D0'
25403 * uint16x4_t vpaddl_u8 (uint8x8_t)
25404 _Form of expected instruction(s):_ `vpaddl.u8 D0, D0'
25406 * int64x1_t vpaddl_s32 (int32x2_t)
25407 _Form of expected instruction(s):_ `vpaddl.s32 D0, D0'
25409 * int32x2_t vpaddl_s16 (int16x4_t)
25410 _Form of expected instruction(s):_ `vpaddl.s16 D0, D0'
25412 * int16x4_t vpaddl_s8 (int8x8_t)
25413 _Form of expected instruction(s):_ `vpaddl.s8 D0, D0'
25415 * uint64x2_t vpaddlq_u32 (uint32x4_t)
25416 _Form of expected instruction(s):_ `vpaddl.u32 Q0, Q0'
25418 * uint32x4_t vpaddlq_u16 (uint16x8_t)
25419 _Form of expected instruction(s):_ `vpaddl.u16 Q0, Q0'
25421 * uint16x8_t vpaddlq_u8 (uint8x16_t)
25422 _Form of expected instruction(s):_ `vpaddl.u8 Q0, Q0'
25424 * int64x2_t vpaddlq_s32 (int32x4_t)
25425 _Form of expected instruction(s):_ `vpaddl.s32 Q0, Q0'
25427 * int32x4_t vpaddlq_s16 (int16x8_t)
25428 _Form of expected instruction(s):_ `vpaddl.s16 Q0, Q0'
25430 * int16x8_t vpaddlq_s8 (int8x16_t)
25431 _Form of expected instruction(s):_ `vpaddl.s8 Q0, Q0'
25433 5.50.3.21 Pairwise add, single_opcode widen and accumulate
25434 ..........................................................
25436 * uint64x1_t vpadal_u32 (uint64x1_t, uint32x2_t)
25437 _Form of expected instruction(s):_ `vpadal.u32 D0, D0'
25439 * uint32x2_t vpadal_u16 (uint32x2_t, uint16x4_t)
25440 _Form of expected instruction(s):_ `vpadal.u16 D0, D0'
25442 * uint16x4_t vpadal_u8 (uint16x4_t, uint8x8_t)
25443 _Form of expected instruction(s):_ `vpadal.u8 D0, D0'
25445 * int64x1_t vpadal_s32 (int64x1_t, int32x2_t)
25446 _Form of expected instruction(s):_ `vpadal.s32 D0, D0'
25448 * int32x2_t vpadal_s16 (int32x2_t, int16x4_t)
25449 _Form of expected instruction(s):_ `vpadal.s16 D0, D0'
25451 * int16x4_t vpadal_s8 (int16x4_t, int8x8_t)
25452 _Form of expected instruction(s):_ `vpadal.s8 D0, D0'
25454 * uint64x2_t vpadalq_u32 (uint64x2_t, uint32x4_t)
25455 _Form of expected instruction(s):_ `vpadal.u32 Q0, Q0'
25457 * uint32x4_t vpadalq_u16 (uint32x4_t, uint16x8_t)
25458 _Form of expected instruction(s):_ `vpadal.u16 Q0, Q0'
25460 * uint16x8_t vpadalq_u8 (uint16x8_t, uint8x16_t)
25461 _Form of expected instruction(s):_ `vpadal.u8 Q0, Q0'
25463 * int64x2_t vpadalq_s32 (int64x2_t, int32x4_t)
25464 _Form of expected instruction(s):_ `vpadal.s32 Q0, Q0'
25466 * int32x4_t vpadalq_s16 (int32x4_t, int16x8_t)
25467 _Form of expected instruction(s):_ `vpadal.s16 Q0, Q0'
25469 * int16x8_t vpadalq_s8 (int16x8_t, int8x16_t)
25470 _Form of expected instruction(s):_ `vpadal.s8 Q0, Q0'
25472 5.50.3.22 Folding maximum
25473 .........................
25475 * uint32x2_t vpmax_u32 (uint32x2_t, uint32x2_t)
25476 _Form of expected instruction(s):_ `vpmax.u32 D0, D0, D0'
25478 * uint16x4_t vpmax_u16 (uint16x4_t, uint16x4_t)
25479 _Form of expected instruction(s):_ `vpmax.u16 D0, D0, D0'
25481 * uint8x8_t vpmax_u8 (uint8x8_t, uint8x8_t)
25482 _Form of expected instruction(s):_ `vpmax.u8 D0, D0, D0'
25484 * int32x2_t vpmax_s32 (int32x2_t, int32x2_t)
25485 _Form of expected instruction(s):_ `vpmax.s32 D0, D0, D0'
25487 * int16x4_t vpmax_s16 (int16x4_t, int16x4_t)
25488 _Form of expected instruction(s):_ `vpmax.s16 D0, D0, D0'
25490 * int8x8_t vpmax_s8 (int8x8_t, int8x8_t)
25491 _Form of expected instruction(s):_ `vpmax.s8 D0, D0, D0'
25493 * float32x2_t vpmax_f32 (float32x2_t, float32x2_t)
25494 _Form of expected instruction(s):_ `vpmax.f32 D0, D0, D0'
25496 5.50.3.23 Folding minimum
25497 .........................
25499 * uint32x2_t vpmin_u32 (uint32x2_t, uint32x2_t)
25500 _Form of expected instruction(s):_ `vpmin.u32 D0, D0, D0'
25502 * uint16x4_t vpmin_u16 (uint16x4_t, uint16x4_t)
25503 _Form of expected instruction(s):_ `vpmin.u16 D0, D0, D0'
25505 * uint8x8_t vpmin_u8 (uint8x8_t, uint8x8_t)
25506 _Form of expected instruction(s):_ `vpmin.u8 D0, D0, D0'
25508 * int32x2_t vpmin_s32 (int32x2_t, int32x2_t)
25509 _Form of expected instruction(s):_ `vpmin.s32 D0, D0, D0'
25511 * int16x4_t vpmin_s16 (int16x4_t, int16x4_t)
25512 _Form of expected instruction(s):_ `vpmin.s16 D0, D0, D0'
25514 * int8x8_t vpmin_s8 (int8x8_t, int8x8_t)
25515 _Form of expected instruction(s):_ `vpmin.s8 D0, D0, D0'
25517 * float32x2_t vpmin_f32 (float32x2_t, float32x2_t)
25518 _Form of expected instruction(s):_ `vpmin.f32 D0, D0, D0'
25520 5.50.3.24 Reciprocal step
25521 .........................
25523 * float32x2_t vrecps_f32 (float32x2_t, float32x2_t)
25524 _Form of expected instruction(s):_ `vrecps.f32 D0, D0, D0'
25526 * float32x4_t vrecpsq_f32 (float32x4_t, float32x4_t)
25527 _Form of expected instruction(s):_ `vrecps.f32 Q0, Q0, Q0'
25529 * float32x2_t vrsqrts_f32 (float32x2_t, float32x2_t)
25530 _Form of expected instruction(s):_ `vrsqrts.f32 D0, D0, D0'
25532 * float32x4_t vrsqrtsq_f32 (float32x4_t, float32x4_t)
25533 _Form of expected instruction(s):_ `vrsqrts.f32 Q0, Q0, Q0'
25535 5.50.3.25 Vector shift left
25536 ...........................
25538 * uint32x2_t vshl_u32 (uint32x2_t, int32x2_t)
25539 _Form of expected instruction(s):_ `vshl.u32 D0, D0, D0'
25541 * uint16x4_t vshl_u16 (uint16x4_t, int16x4_t)
25542 _Form of expected instruction(s):_ `vshl.u16 D0, D0, D0'
25544 * uint8x8_t vshl_u8 (uint8x8_t, int8x8_t)
25545 _Form of expected instruction(s):_ `vshl.u8 D0, D0, D0'
25547 * int32x2_t vshl_s32 (int32x2_t, int32x2_t)
25548 _Form of expected instruction(s):_ `vshl.s32 D0, D0, D0'
25550 * int16x4_t vshl_s16 (int16x4_t, int16x4_t)
25551 _Form of expected instruction(s):_ `vshl.s16 D0, D0, D0'
25553 * int8x8_t vshl_s8 (int8x8_t, int8x8_t)
25554 _Form of expected instruction(s):_ `vshl.s8 D0, D0, D0'
25556 * uint64x1_t vshl_u64 (uint64x1_t, int64x1_t)
25557 _Form of expected instruction(s):_ `vshl.u64 D0, D0, D0'
25559 * int64x1_t vshl_s64 (int64x1_t, int64x1_t)
25560 _Form of expected instruction(s):_ `vshl.s64 D0, D0, D0'
25562 * uint32x4_t vshlq_u32 (uint32x4_t, int32x4_t)
25563 _Form of expected instruction(s):_ `vshl.u32 Q0, Q0, Q0'
25565 * uint16x8_t vshlq_u16 (uint16x8_t, int16x8_t)
25566 _Form of expected instruction(s):_ `vshl.u16 Q0, Q0, Q0'
25568 * uint8x16_t vshlq_u8 (uint8x16_t, int8x16_t)
25569 _Form of expected instruction(s):_ `vshl.u8 Q0, Q0, Q0'
25571 * int32x4_t vshlq_s32 (int32x4_t, int32x4_t)
25572 _Form of expected instruction(s):_ `vshl.s32 Q0, Q0, Q0'
25574 * int16x8_t vshlq_s16 (int16x8_t, int16x8_t)
25575 _Form of expected instruction(s):_ `vshl.s16 Q0, Q0, Q0'
25577 * int8x16_t vshlq_s8 (int8x16_t, int8x16_t)
25578 _Form of expected instruction(s):_ `vshl.s8 Q0, Q0, Q0'
25580 * uint64x2_t vshlq_u64 (uint64x2_t, int64x2_t)
25581 _Form of expected instruction(s):_ `vshl.u64 Q0, Q0, Q0'
25583 * int64x2_t vshlq_s64 (int64x2_t, int64x2_t)
25584 _Form of expected instruction(s):_ `vshl.s64 Q0, Q0, Q0'
25586 * uint32x2_t vrshl_u32 (uint32x2_t, int32x2_t)
25587 _Form of expected instruction(s):_ `vrshl.u32 D0, D0, D0'
25589 * uint16x4_t vrshl_u16 (uint16x4_t, int16x4_t)
25590 _Form of expected instruction(s):_ `vrshl.u16 D0, D0, D0'
25592 * uint8x8_t vrshl_u8 (uint8x8_t, int8x8_t)
25593 _Form of expected instruction(s):_ `vrshl.u8 D0, D0, D0'
25595 * int32x2_t vrshl_s32 (int32x2_t, int32x2_t)
25596 _Form of expected instruction(s):_ `vrshl.s32 D0, D0, D0'
25598 * int16x4_t vrshl_s16 (int16x4_t, int16x4_t)
25599 _Form of expected instruction(s):_ `vrshl.s16 D0, D0, D0'
25601 * int8x8_t vrshl_s8 (int8x8_t, int8x8_t)
25602 _Form of expected instruction(s):_ `vrshl.s8 D0, D0, D0'
25604 * uint64x1_t vrshl_u64 (uint64x1_t, int64x1_t)
25605 _Form of expected instruction(s):_ `vrshl.u64 D0, D0, D0'
25607 * int64x1_t vrshl_s64 (int64x1_t, int64x1_t)
25608 _Form of expected instruction(s):_ `vrshl.s64 D0, D0, D0'
25610 * uint32x4_t vrshlq_u32 (uint32x4_t, int32x4_t)
25611 _Form of expected instruction(s):_ `vrshl.u32 Q0, Q0, Q0'
25613 * uint16x8_t vrshlq_u16 (uint16x8_t, int16x8_t)
25614 _Form of expected instruction(s):_ `vrshl.u16 Q0, Q0, Q0'
25616 * uint8x16_t vrshlq_u8 (uint8x16_t, int8x16_t)
25617 _Form of expected instruction(s):_ `vrshl.u8 Q0, Q0, Q0'
25619 * int32x4_t vrshlq_s32 (int32x4_t, int32x4_t)
25620 _Form of expected instruction(s):_ `vrshl.s32 Q0, Q0, Q0'
25622 * int16x8_t vrshlq_s16 (int16x8_t, int16x8_t)
25623 _Form of expected instruction(s):_ `vrshl.s16 Q0, Q0, Q0'
25625 * int8x16_t vrshlq_s8 (int8x16_t, int8x16_t)
25626 _Form of expected instruction(s):_ `vrshl.s8 Q0, Q0, Q0'
25628 * uint64x2_t vrshlq_u64 (uint64x2_t, int64x2_t)
25629 _Form of expected instruction(s):_ `vrshl.u64 Q0, Q0, Q0'
25631 * int64x2_t vrshlq_s64 (int64x2_t, int64x2_t)
25632 _Form of expected instruction(s):_ `vrshl.s64 Q0, Q0, Q0'
25634 * uint32x2_t vqshl_u32 (uint32x2_t, int32x2_t)
25635 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, D0'
25637 * uint16x4_t vqshl_u16 (uint16x4_t, int16x4_t)
25638 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, D0'
25640 * uint8x8_t vqshl_u8 (uint8x8_t, int8x8_t)
25641 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, D0'
25643 * int32x2_t vqshl_s32 (int32x2_t, int32x2_t)
25644 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, D0'
25646 * int16x4_t vqshl_s16 (int16x4_t, int16x4_t)
25647 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, D0'
25649 * int8x8_t vqshl_s8 (int8x8_t, int8x8_t)
25650 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, D0'
25652 * uint64x1_t vqshl_u64 (uint64x1_t, int64x1_t)
25653 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, D0'
25655 * int64x1_t vqshl_s64 (int64x1_t, int64x1_t)
25656 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, D0'
25658 * uint32x4_t vqshlq_u32 (uint32x4_t, int32x4_t)
25659 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, Q0'
25661 * uint16x8_t vqshlq_u16 (uint16x8_t, int16x8_t)
25662 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, Q0'
25664 * uint8x16_t vqshlq_u8 (uint8x16_t, int8x16_t)
25665 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, Q0'
25667 * int32x4_t vqshlq_s32 (int32x4_t, int32x4_t)
25668 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, Q0'
25670 * int16x8_t vqshlq_s16 (int16x8_t, int16x8_t)
25671 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, Q0'
25673 * int8x16_t vqshlq_s8 (int8x16_t, int8x16_t)
25674 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, Q0'
25676 * uint64x2_t vqshlq_u64 (uint64x2_t, int64x2_t)
25677 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, Q0'
25679 * int64x2_t vqshlq_s64 (int64x2_t, int64x2_t)
25680 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, Q0'
25682 * uint32x2_t vqrshl_u32 (uint32x2_t, int32x2_t)
25683 _Form of expected instruction(s):_ `vqrshl.u32 D0, D0, D0'
25685 * uint16x4_t vqrshl_u16 (uint16x4_t, int16x4_t)
25686 _Form of expected instruction(s):_ `vqrshl.u16 D0, D0, D0'
25688 * uint8x8_t vqrshl_u8 (uint8x8_t, int8x8_t)
25689 _Form of expected instruction(s):_ `vqrshl.u8 D0, D0, D0'
25691 * int32x2_t vqrshl_s32 (int32x2_t, int32x2_t)
25692 _Form of expected instruction(s):_ `vqrshl.s32 D0, D0, D0'
25694 * int16x4_t vqrshl_s16 (int16x4_t, int16x4_t)
25695 _Form of expected instruction(s):_ `vqrshl.s16 D0, D0, D0'
25697 * int8x8_t vqrshl_s8 (int8x8_t, int8x8_t)
25698 _Form of expected instruction(s):_ `vqrshl.s8 D0, D0, D0'
25700 * uint64x1_t vqrshl_u64 (uint64x1_t, int64x1_t)
25701 _Form of expected instruction(s):_ `vqrshl.u64 D0, D0, D0'
25703 * int64x1_t vqrshl_s64 (int64x1_t, int64x1_t)
25704 _Form of expected instruction(s):_ `vqrshl.s64 D0, D0, D0'
25706 * uint32x4_t vqrshlq_u32 (uint32x4_t, int32x4_t)
25707 _Form of expected instruction(s):_ `vqrshl.u32 Q0, Q0, Q0'
25709 * uint16x8_t vqrshlq_u16 (uint16x8_t, int16x8_t)
25710 _Form of expected instruction(s):_ `vqrshl.u16 Q0, Q0, Q0'
25712 * uint8x16_t vqrshlq_u8 (uint8x16_t, int8x16_t)
25713 _Form of expected instruction(s):_ `vqrshl.u8 Q0, Q0, Q0'
25715 * int32x4_t vqrshlq_s32 (int32x4_t, int32x4_t)
25716 _Form of expected instruction(s):_ `vqrshl.s32 Q0, Q0, Q0'
25718 * int16x8_t vqrshlq_s16 (int16x8_t, int16x8_t)
25719 _Form of expected instruction(s):_ `vqrshl.s16 Q0, Q0, Q0'
25721 * int8x16_t vqrshlq_s8 (int8x16_t, int8x16_t)
25722 _Form of expected instruction(s):_ `vqrshl.s8 Q0, Q0, Q0'
25724 * uint64x2_t vqrshlq_u64 (uint64x2_t, int64x2_t)
25725 _Form of expected instruction(s):_ `vqrshl.u64 Q0, Q0, Q0'
25727 * int64x2_t vqrshlq_s64 (int64x2_t, int64x2_t)
25728 _Form of expected instruction(s):_ `vqrshl.s64 Q0, Q0, Q0'
25730 5.50.3.26 Vector shift left by constant
25731 .......................................
25733 * uint32x2_t vshl_n_u32 (uint32x2_t, const int)
25734 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
25736 * uint16x4_t vshl_n_u16 (uint16x4_t, const int)
25737 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
25739 * uint8x8_t vshl_n_u8 (uint8x8_t, const int)
25740 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
25742 * int32x2_t vshl_n_s32 (int32x2_t, const int)
25743 _Form of expected instruction(s):_ `vshl.i32 D0, D0, #0'
25745 * int16x4_t vshl_n_s16 (int16x4_t, const int)
25746 _Form of expected instruction(s):_ `vshl.i16 D0, D0, #0'
25748 * int8x8_t vshl_n_s8 (int8x8_t, const int)
25749 _Form of expected instruction(s):_ `vshl.i8 D0, D0, #0'
25751 * uint64x1_t vshl_n_u64 (uint64x1_t, const int)
25752 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
25754 * int64x1_t vshl_n_s64 (int64x1_t, const int)
25755 _Form of expected instruction(s):_ `vshl.i64 D0, D0, #0'
25757 * uint32x4_t vshlq_n_u32 (uint32x4_t, const int)
25758 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
25760 * uint16x8_t vshlq_n_u16 (uint16x8_t, const int)
25761 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
25763 * uint8x16_t vshlq_n_u8 (uint8x16_t, const int)
25764 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
25766 * int32x4_t vshlq_n_s32 (int32x4_t, const int)
25767 _Form of expected instruction(s):_ `vshl.i32 Q0, Q0, #0'
25769 * int16x8_t vshlq_n_s16 (int16x8_t, const int)
25770 _Form of expected instruction(s):_ `vshl.i16 Q0, Q0, #0'
25772 * int8x16_t vshlq_n_s8 (int8x16_t, const int)
25773 _Form of expected instruction(s):_ `vshl.i8 Q0, Q0, #0'
25775 * uint64x2_t vshlq_n_u64 (uint64x2_t, const int)
25776 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
25778 * int64x2_t vshlq_n_s64 (int64x2_t, const int)
25779 _Form of expected instruction(s):_ `vshl.i64 Q0, Q0, #0'
25781 * uint32x2_t vqshl_n_u32 (uint32x2_t, const int)
25782 _Form of expected instruction(s):_ `vqshl.u32 D0, D0, #0'
25784 * uint16x4_t vqshl_n_u16 (uint16x4_t, const int)
25785 _Form of expected instruction(s):_ `vqshl.u16 D0, D0, #0'
25787 * uint8x8_t vqshl_n_u8 (uint8x8_t, const int)
25788 _Form of expected instruction(s):_ `vqshl.u8 D0, D0, #0'
25790 * int32x2_t vqshl_n_s32 (int32x2_t, const int)
25791 _Form of expected instruction(s):_ `vqshl.s32 D0, D0, #0'
25793 * int16x4_t vqshl_n_s16 (int16x4_t, const int)
25794 _Form of expected instruction(s):_ `vqshl.s16 D0, D0, #0'
25796 * int8x8_t vqshl_n_s8 (int8x8_t, const int)
25797 _Form of expected instruction(s):_ `vqshl.s8 D0, D0, #0'
25799 * uint64x1_t vqshl_n_u64 (uint64x1_t, const int)
25800 _Form of expected instruction(s):_ `vqshl.u64 D0, D0, #0'
25802 * int64x1_t vqshl_n_s64 (int64x1_t, const int)
25803 _Form of expected instruction(s):_ `vqshl.s64 D0, D0, #0'
25805 * uint32x4_t vqshlq_n_u32 (uint32x4_t, const int)
25806 _Form of expected instruction(s):_ `vqshl.u32 Q0, Q0, #0'
25808 * uint16x8_t vqshlq_n_u16 (uint16x8_t, const int)
25809 _Form of expected instruction(s):_ `vqshl.u16 Q0, Q0, #0'
25811 * uint8x16_t vqshlq_n_u8 (uint8x16_t, const int)
25812 _Form of expected instruction(s):_ `vqshl.u8 Q0, Q0, #0'
25814 * int32x4_t vqshlq_n_s32 (int32x4_t, const int)
25815 _Form of expected instruction(s):_ `vqshl.s32 Q0, Q0, #0'
25817 * int16x8_t vqshlq_n_s16 (int16x8_t, const int)
25818 _Form of expected instruction(s):_ `vqshl.s16 Q0, Q0, #0'
25820 * int8x16_t vqshlq_n_s8 (int8x16_t, const int)
25821 _Form of expected instruction(s):_ `vqshl.s8 Q0, Q0, #0'
25823 * uint64x2_t vqshlq_n_u64 (uint64x2_t, const int)
25824 _Form of expected instruction(s):_ `vqshl.u64 Q0, Q0, #0'
25826 * int64x2_t vqshlq_n_s64 (int64x2_t, const int)
25827 _Form of expected instruction(s):_ `vqshl.s64 Q0, Q0, #0'
25829 * uint64x1_t vqshlu_n_s64 (int64x1_t, const int)
25830 _Form of expected instruction(s):_ `vqshlu.s64 D0, D0, #0'
25832 * uint32x2_t vqshlu_n_s32 (int32x2_t, const int)
25833 _Form of expected instruction(s):_ `vqshlu.s32 D0, D0, #0'
25835 * uint16x4_t vqshlu_n_s16 (int16x4_t, const int)
25836 _Form of expected instruction(s):_ `vqshlu.s16 D0, D0, #0'
25838 * uint8x8_t vqshlu_n_s8 (int8x8_t, const int)
25839 _Form of expected instruction(s):_ `vqshlu.s8 D0, D0, #0'
25841 * uint64x2_t vqshluq_n_s64 (int64x2_t, const int)
25842 _Form of expected instruction(s):_ `vqshlu.s64 Q0, Q0, #0'
25844 * uint32x4_t vqshluq_n_s32 (int32x4_t, const int)
25845 _Form of expected instruction(s):_ `vqshlu.s32 Q0, Q0, #0'
25847 * uint16x8_t vqshluq_n_s16 (int16x8_t, const int)
25848 _Form of expected instruction(s):_ `vqshlu.s16 Q0, Q0, #0'
25850 * uint8x16_t vqshluq_n_s8 (int8x16_t, const int)
25851 _Form of expected instruction(s):_ `vqshlu.s8 Q0, Q0, #0'
25853 * uint64x2_t vshll_n_u32 (uint32x2_t, const int)
25854 _Form of expected instruction(s):_ `vshll.u32 Q0, D0, #0'
25856 * uint32x4_t vshll_n_u16 (uint16x4_t, const int)
25857 _Form of expected instruction(s):_ `vshll.u16 Q0, D0, #0'
25859 * uint16x8_t vshll_n_u8 (uint8x8_t, const int)
25860 _Form of expected instruction(s):_ `vshll.u8 Q0, D0, #0'
25862 * int64x2_t vshll_n_s32 (int32x2_t, const int)
25863 _Form of expected instruction(s):_ `vshll.s32 Q0, D0, #0'
25865 * int32x4_t vshll_n_s16 (int16x4_t, const int)
25866 _Form of expected instruction(s):_ `vshll.s16 Q0, D0, #0'
25868 * int16x8_t vshll_n_s8 (int8x8_t, const int)
25869 _Form of expected instruction(s):_ `vshll.s8 Q0, D0, #0'
25871 5.50.3.27 Vector shift right by constant
25872 ........................................
25874 * uint32x2_t vshr_n_u32 (uint32x2_t, const int)
25875 _Form of expected instruction(s):_ `vshr.u32 D0, D0, #0'
25877 * uint16x4_t vshr_n_u16 (uint16x4_t, const int)
25878 _Form of expected instruction(s):_ `vshr.u16 D0, D0, #0'
25880 * uint8x8_t vshr_n_u8 (uint8x8_t, const int)
25881 _Form of expected instruction(s):_ `vshr.u8 D0, D0, #0'
25883 * int32x2_t vshr_n_s32 (int32x2_t, const int)
25884 _Form of expected instruction(s):_ `vshr.s32 D0, D0, #0'
25886 * int16x4_t vshr_n_s16 (int16x4_t, const int)
25887 _Form of expected instruction(s):_ `vshr.s16 D0, D0, #0'
25889 * int8x8_t vshr_n_s8 (int8x8_t, const int)
25890 _Form of expected instruction(s):_ `vshr.s8 D0, D0, #0'
25892 * uint64x1_t vshr_n_u64 (uint64x1_t, const int)
25893 _Form of expected instruction(s):_ `vshr.u64 D0, D0, #0'
25895 * int64x1_t vshr_n_s64 (int64x1_t, const int)
25896 _Form of expected instruction(s):_ `vshr.s64 D0, D0, #0'
25898 * uint32x4_t vshrq_n_u32 (uint32x4_t, const int)
25899 _Form of expected instruction(s):_ `vshr.u32 Q0, Q0, #0'
25901 * uint16x8_t vshrq_n_u16 (uint16x8_t, const int)
25902 _Form of expected instruction(s):_ `vshr.u16 Q0, Q0, #0'
25904 * uint8x16_t vshrq_n_u8 (uint8x16_t, const int)
25905 _Form of expected instruction(s):_ `vshr.u8 Q0, Q0, #0'
25907 * int32x4_t vshrq_n_s32 (int32x4_t, const int)
25908 _Form of expected instruction(s):_ `vshr.s32 Q0, Q0, #0'
25910 * int16x8_t vshrq_n_s16 (int16x8_t, const int)
25911 _Form of expected instruction(s):_ `vshr.s16 Q0, Q0, #0'
25913 * int8x16_t vshrq_n_s8 (int8x16_t, const int)
25914 _Form of expected instruction(s):_ `vshr.s8 Q0, Q0, #0'
25916 * uint64x2_t vshrq_n_u64 (uint64x2_t, const int)
25917 _Form of expected instruction(s):_ `vshr.u64 Q0, Q0, #0'
25919 * int64x2_t vshrq_n_s64 (int64x2_t, const int)
25920 _Form of expected instruction(s):_ `vshr.s64 Q0, Q0, #0'
25922 * uint32x2_t vrshr_n_u32 (uint32x2_t, const int)
25923 _Form of expected instruction(s):_ `vrshr.u32 D0, D0, #0'
25925 * uint16x4_t vrshr_n_u16 (uint16x4_t, const int)
25926 _Form of expected instruction(s):_ `vrshr.u16 D0, D0, #0'
25928 * uint8x8_t vrshr_n_u8 (uint8x8_t, const int)
25929 _Form of expected instruction(s):_ `vrshr.u8 D0, D0, #0'
25931 * int32x2_t vrshr_n_s32 (int32x2_t, const int)
25932 _Form of expected instruction(s):_ `vrshr.s32 D0, D0, #0'
25934 * int16x4_t vrshr_n_s16 (int16x4_t, const int)
25935 _Form of expected instruction(s):_ `vrshr.s16 D0, D0, #0'
25937 * int8x8_t vrshr_n_s8 (int8x8_t, const int)
25938 _Form of expected instruction(s):_ `vrshr.s8 D0, D0, #0'
25940 * uint64x1_t vrshr_n_u64 (uint64x1_t, const int)
25941 _Form of expected instruction(s):_ `vrshr.u64 D0, D0, #0'
25943 * int64x1_t vrshr_n_s64 (int64x1_t, const int)
25944 _Form of expected instruction(s):_ `vrshr.s64 D0, D0, #0'
25946 * uint32x4_t vrshrq_n_u32 (uint32x4_t, const int)
25947 _Form of expected instruction(s):_ `vrshr.u32 Q0, Q0, #0'
25949 * uint16x8_t vrshrq_n_u16 (uint16x8_t, const int)
25950 _Form of expected instruction(s):_ `vrshr.u16 Q0, Q0, #0'
25952 * uint8x16_t vrshrq_n_u8 (uint8x16_t, const int)
25953 _Form of expected instruction(s):_ `vrshr.u8 Q0, Q0, #0'
25955 * int32x4_t vrshrq_n_s32 (int32x4_t, const int)
25956 _Form of expected instruction(s):_ `vrshr.s32 Q0, Q0, #0'
25958 * int16x8_t vrshrq_n_s16 (int16x8_t, const int)
25959 _Form of expected instruction(s):_ `vrshr.s16 Q0, Q0, #0'
25961 * int8x16_t vrshrq_n_s8 (int8x16_t, const int)
25962 _Form of expected instruction(s):_ `vrshr.s8 Q0, Q0, #0'
25964 * uint64x2_t vrshrq_n_u64 (uint64x2_t, const int)
25965 _Form of expected instruction(s):_ `vrshr.u64 Q0, Q0, #0'
25967 * int64x2_t vrshrq_n_s64 (int64x2_t, const int)
25968 _Form of expected instruction(s):_ `vrshr.s64 Q0, Q0, #0'
25970 * uint32x2_t vshrn_n_u64 (uint64x2_t, const int)
25971 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
25973 * uint16x4_t vshrn_n_u32 (uint32x4_t, const int)
25974 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
25976 * uint8x8_t vshrn_n_u16 (uint16x8_t, const int)
25977 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
25979 * int32x2_t vshrn_n_s64 (int64x2_t, const int)
25980 _Form of expected instruction(s):_ `vshrn.i64 D0, Q0, #0'
25982 * int16x4_t vshrn_n_s32 (int32x4_t, const int)
25983 _Form of expected instruction(s):_ `vshrn.i32 D0, Q0, #0'
25985 * int8x8_t vshrn_n_s16 (int16x8_t, const int)
25986 _Form of expected instruction(s):_ `vshrn.i16 D0, Q0, #0'
25988 * uint32x2_t vrshrn_n_u64 (uint64x2_t, const int)
25989 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
25991 * uint16x4_t vrshrn_n_u32 (uint32x4_t, const int)
25992 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
25994 * uint8x8_t vrshrn_n_u16 (uint16x8_t, const int)
25995 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
25997 * int32x2_t vrshrn_n_s64 (int64x2_t, const int)
25998 _Form of expected instruction(s):_ `vrshrn.i64 D0, Q0, #0'
26000 * int16x4_t vrshrn_n_s32 (int32x4_t, const int)
26001 _Form of expected instruction(s):_ `vrshrn.i32 D0, Q0, #0'
26003 * int8x8_t vrshrn_n_s16 (int16x8_t, const int)
26004 _Form of expected instruction(s):_ `vrshrn.i16 D0, Q0, #0'
26006 * uint32x2_t vqshrn_n_u64 (uint64x2_t, const int)
26007 _Form of expected instruction(s):_ `vqshrn.u64 D0, Q0, #0'
26009 * uint16x4_t vqshrn_n_u32 (uint32x4_t, const int)
26010 _Form of expected instruction(s):_ `vqshrn.u32 D0, Q0, #0'
26012 * uint8x8_t vqshrn_n_u16 (uint16x8_t, const int)
26013 _Form of expected instruction(s):_ `vqshrn.u16 D0, Q0, #0'
26015 * int32x2_t vqshrn_n_s64 (int64x2_t, const int)
26016 _Form of expected instruction(s):_ `vqshrn.s64 D0, Q0, #0'
26018 * int16x4_t vqshrn_n_s32 (int32x4_t, const int)
26019 _Form of expected instruction(s):_ `vqshrn.s32 D0, Q0, #0'
26021 * int8x8_t vqshrn_n_s16 (int16x8_t, const int)
26022 _Form of expected instruction(s):_ `vqshrn.s16 D0, Q0, #0'
26024 * uint32x2_t vqrshrn_n_u64 (uint64x2_t, const int)
26025 _Form of expected instruction(s):_ `vqrshrn.u64 D0, Q0, #0'
26027 * uint16x4_t vqrshrn_n_u32 (uint32x4_t, const int)
26028 _Form of expected instruction(s):_ `vqrshrn.u32 D0, Q0, #0'
26030 * uint8x8_t vqrshrn_n_u16 (uint16x8_t, const int)
26031 _Form of expected instruction(s):_ `vqrshrn.u16 D0, Q0, #0'
26033 * int32x2_t vqrshrn_n_s64 (int64x2_t, const int)
26034 _Form of expected instruction(s):_ `vqrshrn.s64 D0, Q0, #0'
26036 * int16x4_t vqrshrn_n_s32 (int32x4_t, const int)
26037 _Form of expected instruction(s):_ `vqrshrn.s32 D0, Q0, #0'
26039 * int8x8_t vqrshrn_n_s16 (int16x8_t, const int)
26040 _Form of expected instruction(s):_ `vqrshrn.s16 D0, Q0, #0'
26042 * uint32x2_t vqshrun_n_s64 (int64x2_t, const int)
26043 _Form of expected instruction(s):_ `vqshrun.s64 D0, Q0, #0'
26045 * uint16x4_t vqshrun_n_s32 (int32x4_t, const int)
26046 _Form of expected instruction(s):_ `vqshrun.s32 D0, Q0, #0'
26048 * uint8x8_t vqshrun_n_s16 (int16x8_t, const int)
26049 _Form of expected instruction(s):_ `vqshrun.s16 D0, Q0, #0'
26051 * uint32x2_t vqrshrun_n_s64 (int64x2_t, const int)
26052 _Form of expected instruction(s):_ `vqrshrun.s64 D0, Q0, #0'
26054 * uint16x4_t vqrshrun_n_s32 (int32x4_t, const int)
26055 _Form of expected instruction(s):_ `vqrshrun.s32 D0, Q0, #0'
26057 * uint8x8_t vqrshrun_n_s16 (int16x8_t, const int)
26058 _Form of expected instruction(s):_ `vqrshrun.s16 D0, Q0, #0'
26060 5.50.3.28 Vector shift right by constant and accumulate
26061 .......................................................
26063 * uint32x2_t vsra_n_u32 (uint32x2_t, uint32x2_t, const int)
26064 _Form of expected instruction(s):_ `vsra.u32 D0, D0, #0'
26066 * uint16x4_t vsra_n_u16 (uint16x4_t, uint16x4_t, const int)
26067 _Form of expected instruction(s):_ `vsra.u16 D0, D0, #0'
26069 * uint8x8_t vsra_n_u8 (uint8x8_t, uint8x8_t, const int)
26070 _Form of expected instruction(s):_ `vsra.u8 D0, D0, #0'
26072 * int32x2_t vsra_n_s32 (int32x2_t, int32x2_t, const int)
26073 _Form of expected instruction(s):_ `vsra.s32 D0, D0, #0'
26075 * int16x4_t vsra_n_s16 (int16x4_t, int16x4_t, const int)
26076 _Form of expected instruction(s):_ `vsra.s16 D0, D0, #0'
26078 * int8x8_t vsra_n_s8 (int8x8_t, int8x8_t, const int)
26079 _Form of expected instruction(s):_ `vsra.s8 D0, D0, #0'
26081 * uint64x1_t vsra_n_u64 (uint64x1_t, uint64x1_t, const int)
26082 _Form of expected instruction(s):_ `vsra.u64 D0, D0, #0'
26084 * int64x1_t vsra_n_s64 (int64x1_t, int64x1_t, const int)
26085 _Form of expected instruction(s):_ `vsra.s64 D0, D0, #0'
26087 * uint32x4_t vsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
26088 _Form of expected instruction(s):_ `vsra.u32 Q0, Q0, #0'
26090 * uint16x8_t vsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
26091 _Form of expected instruction(s):_ `vsra.u16 Q0, Q0, #0'
26093 * uint8x16_t vsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
26094 _Form of expected instruction(s):_ `vsra.u8 Q0, Q0, #0'
26096 * int32x4_t vsraq_n_s32 (int32x4_t, int32x4_t, const int)
26097 _Form of expected instruction(s):_ `vsra.s32 Q0, Q0, #0'
26099 * int16x8_t vsraq_n_s16 (int16x8_t, int16x8_t, const int)
26100 _Form of expected instruction(s):_ `vsra.s16 Q0, Q0, #0'
26102 * int8x16_t vsraq_n_s8 (int8x16_t, int8x16_t, const int)
26103 _Form of expected instruction(s):_ `vsra.s8 Q0, Q0, #0'
26105 * uint64x2_t vsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
26106 _Form of expected instruction(s):_ `vsra.u64 Q0, Q0, #0'
26108 * int64x2_t vsraq_n_s64 (int64x2_t, int64x2_t, const int)
26109 _Form of expected instruction(s):_ `vsra.s64 Q0, Q0, #0'
26111 * uint32x2_t vrsra_n_u32 (uint32x2_t, uint32x2_t, const int)
26112 _Form of expected instruction(s):_ `vrsra.u32 D0, D0, #0'
26114 * uint16x4_t vrsra_n_u16 (uint16x4_t, uint16x4_t, const int)
26115 _Form of expected instruction(s):_ `vrsra.u16 D0, D0, #0'
26117 * uint8x8_t vrsra_n_u8 (uint8x8_t, uint8x8_t, const int)
26118 _Form of expected instruction(s):_ `vrsra.u8 D0, D0, #0'
26120 * int32x2_t vrsra_n_s32 (int32x2_t, int32x2_t, const int)
26121 _Form of expected instruction(s):_ `vrsra.s32 D0, D0, #0'
26123 * int16x4_t vrsra_n_s16 (int16x4_t, int16x4_t, const int)
26124 _Form of expected instruction(s):_ `vrsra.s16 D0, D0, #0'
26126 * int8x8_t vrsra_n_s8 (int8x8_t, int8x8_t, const int)
26127 _Form of expected instruction(s):_ `vrsra.s8 D0, D0, #0'
26129 * uint64x1_t vrsra_n_u64 (uint64x1_t, uint64x1_t, const int)
26130 _Form of expected instruction(s):_ `vrsra.u64 D0, D0, #0'
26132 * int64x1_t vrsra_n_s64 (int64x1_t, int64x1_t, const int)
26133 _Form of expected instruction(s):_ `vrsra.s64 D0, D0, #0'
26135 * uint32x4_t vrsraq_n_u32 (uint32x4_t, uint32x4_t, const int)
26136 _Form of expected instruction(s):_ `vrsra.u32 Q0, Q0, #0'
26138 * uint16x8_t vrsraq_n_u16 (uint16x8_t, uint16x8_t, const int)
26139 _Form of expected instruction(s):_ `vrsra.u16 Q0, Q0, #0'
26141 * uint8x16_t vrsraq_n_u8 (uint8x16_t, uint8x16_t, const int)
26142 _Form of expected instruction(s):_ `vrsra.u8 Q0, Q0, #0'
26144 * int32x4_t vrsraq_n_s32 (int32x4_t, int32x4_t, const int)
26145 _Form of expected instruction(s):_ `vrsra.s32 Q0, Q0, #0'
26147 * int16x8_t vrsraq_n_s16 (int16x8_t, int16x8_t, const int)
26148 _Form of expected instruction(s):_ `vrsra.s16 Q0, Q0, #0'
26150 * int8x16_t vrsraq_n_s8 (int8x16_t, int8x16_t, const int)
26151 _Form of expected instruction(s):_ `vrsra.s8 Q0, Q0, #0'
26153 * uint64x2_t vrsraq_n_u64 (uint64x2_t, uint64x2_t, const int)
26154 _Form of expected instruction(s):_ `vrsra.u64 Q0, Q0, #0'
26156 * int64x2_t vrsraq_n_s64 (int64x2_t, int64x2_t, const int)
26157 _Form of expected instruction(s):_ `vrsra.s64 Q0, Q0, #0'
26159 5.50.3.29 Vector shift right and insert
26160 .......................................
26162 * uint32x2_t vsri_n_u32 (uint32x2_t, uint32x2_t, const int)
26163 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
26165 * uint16x4_t vsri_n_u16 (uint16x4_t, uint16x4_t, const int)
26166 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26168 * uint8x8_t vsri_n_u8 (uint8x8_t, uint8x8_t, const int)
26169 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26171 * int32x2_t vsri_n_s32 (int32x2_t, int32x2_t, const int)
26172 _Form of expected instruction(s):_ `vsri.32 D0, D0, #0'
26174 * int16x4_t vsri_n_s16 (int16x4_t, int16x4_t, const int)
26175 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26177 * int8x8_t vsri_n_s8 (int8x8_t, int8x8_t, const int)
26178 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26180 * uint64x1_t vsri_n_u64 (uint64x1_t, uint64x1_t, const int)
26181 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
26183 * int64x1_t vsri_n_s64 (int64x1_t, int64x1_t, const int)
26184 _Form of expected instruction(s):_ `vsri.64 D0, D0, #0'
26186 * poly16x4_t vsri_n_p16 (poly16x4_t, poly16x4_t, const int)
26187 _Form of expected instruction(s):_ `vsri.16 D0, D0, #0'
26189 * poly8x8_t vsri_n_p8 (poly8x8_t, poly8x8_t, const int)
26190 _Form of expected instruction(s):_ `vsri.8 D0, D0, #0'
26192 * uint32x4_t vsriq_n_u32 (uint32x4_t, uint32x4_t, const int)
26193 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
26195 * uint16x8_t vsriq_n_u16 (uint16x8_t, uint16x8_t, const int)
26196 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26198 * uint8x16_t vsriq_n_u8 (uint8x16_t, uint8x16_t, const int)
26199 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26201 * int32x4_t vsriq_n_s32 (int32x4_t, int32x4_t, const int)
26202 _Form of expected instruction(s):_ `vsri.32 Q0, Q0, #0'
26204 * int16x8_t vsriq_n_s16 (int16x8_t, int16x8_t, const int)
26205 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26207 * int8x16_t vsriq_n_s8 (int8x16_t, int8x16_t, const int)
26208 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26210 * uint64x2_t vsriq_n_u64 (uint64x2_t, uint64x2_t, const int)
26211 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
26213 * int64x2_t vsriq_n_s64 (int64x2_t, int64x2_t, const int)
26214 _Form of expected instruction(s):_ `vsri.64 Q0, Q0, #0'
26216 * poly16x8_t vsriq_n_p16 (poly16x8_t, poly16x8_t, const int)
26217 _Form of expected instruction(s):_ `vsri.16 Q0, Q0, #0'
26219 * poly8x16_t vsriq_n_p8 (poly8x16_t, poly8x16_t, const int)
26220 _Form of expected instruction(s):_ `vsri.8 Q0, Q0, #0'
26222 5.50.3.30 Vector shift left and insert
26223 ......................................
26225 * uint32x2_t vsli_n_u32 (uint32x2_t, uint32x2_t, const int)
26226 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
26228 * uint16x4_t vsli_n_u16 (uint16x4_t, uint16x4_t, const int)
26229 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26231 * uint8x8_t vsli_n_u8 (uint8x8_t, uint8x8_t, const int)
26232 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26234 * int32x2_t vsli_n_s32 (int32x2_t, int32x2_t, const int)
26235 _Form of expected instruction(s):_ `vsli.32 D0, D0, #0'
26237 * int16x4_t vsli_n_s16 (int16x4_t, int16x4_t, const int)
26238 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26240 * int8x8_t vsli_n_s8 (int8x8_t, int8x8_t, const int)
26241 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26243 * uint64x1_t vsli_n_u64 (uint64x1_t, uint64x1_t, const int)
26244 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
26246 * int64x1_t vsli_n_s64 (int64x1_t, int64x1_t, const int)
26247 _Form of expected instruction(s):_ `vsli.64 D0, D0, #0'
26249 * poly16x4_t vsli_n_p16 (poly16x4_t, poly16x4_t, const int)
26250 _Form of expected instruction(s):_ `vsli.16 D0, D0, #0'
26252 * poly8x8_t vsli_n_p8 (poly8x8_t, poly8x8_t, const int)
26253 _Form of expected instruction(s):_ `vsli.8 D0, D0, #0'
26255 * uint32x4_t vsliq_n_u32 (uint32x4_t, uint32x4_t, const int)
26256 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
26258 * uint16x8_t vsliq_n_u16 (uint16x8_t, uint16x8_t, const int)
26259 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26261 * uint8x16_t vsliq_n_u8 (uint8x16_t, uint8x16_t, const int)
26262 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26264 * int32x4_t vsliq_n_s32 (int32x4_t, int32x4_t, const int)
26265 _Form of expected instruction(s):_ `vsli.32 Q0, Q0, #0'
26267 * int16x8_t vsliq_n_s16 (int16x8_t, int16x8_t, const int)
26268 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26270 * int8x16_t vsliq_n_s8 (int8x16_t, int8x16_t, const int)
26271 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26273 * uint64x2_t vsliq_n_u64 (uint64x2_t, uint64x2_t, const int)
26274 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
26276 * int64x2_t vsliq_n_s64 (int64x2_t, int64x2_t, const int)
26277 _Form of expected instruction(s):_ `vsli.64 Q0, Q0, #0'
26279 * poly16x8_t vsliq_n_p16 (poly16x8_t, poly16x8_t, const int)
26280 _Form of expected instruction(s):_ `vsli.16 Q0, Q0, #0'
26282 * poly8x16_t vsliq_n_p8 (poly8x16_t, poly8x16_t, const int)
26283 _Form of expected instruction(s):_ `vsli.8 Q0, Q0, #0'
26285 5.50.3.31 Absolute value
26286 ........................
26288 * float32x2_t vabs_f32 (float32x2_t)
26289 _Form of expected instruction(s):_ `vabs.f32 D0, D0'
26291 * int32x2_t vabs_s32 (int32x2_t)
26292 _Form of expected instruction(s):_ `vabs.s32 D0, D0'
26294 * int16x4_t vabs_s16 (int16x4_t)
26295 _Form of expected instruction(s):_ `vabs.s16 D0, D0'
26297 * int8x8_t vabs_s8 (int8x8_t)
26298 _Form of expected instruction(s):_ `vabs.s8 D0, D0'
26300 * float32x4_t vabsq_f32 (float32x4_t)
26301 _Form of expected instruction(s):_ `vabs.f32 Q0, Q0'
26303 * int32x4_t vabsq_s32 (int32x4_t)
26304 _Form of expected instruction(s):_ `vabs.s32 Q0, Q0'
26306 * int16x8_t vabsq_s16 (int16x8_t)
26307 _Form of expected instruction(s):_ `vabs.s16 Q0, Q0'
26309 * int8x16_t vabsq_s8 (int8x16_t)
26310 _Form of expected instruction(s):_ `vabs.s8 Q0, Q0'
26312 * int32x2_t vqabs_s32 (int32x2_t)
26313 _Form of expected instruction(s):_ `vqabs.s32 D0, D0'
26315 * int16x4_t vqabs_s16 (int16x4_t)
26316 _Form of expected instruction(s):_ `vqabs.s16 D0, D0'
26318 * int8x8_t vqabs_s8 (int8x8_t)
26319 _Form of expected instruction(s):_ `vqabs.s8 D0, D0'
26321 * int32x4_t vqabsq_s32 (int32x4_t)
26322 _Form of expected instruction(s):_ `vqabs.s32 Q0, Q0'
26324 * int16x8_t vqabsq_s16 (int16x8_t)
26325 _Form of expected instruction(s):_ `vqabs.s16 Q0, Q0'
26327 * int8x16_t vqabsq_s8 (int8x16_t)
26328 _Form of expected instruction(s):_ `vqabs.s8 Q0, Q0'
26333 * float32x2_t vneg_f32 (float32x2_t)
26334 _Form of expected instruction(s):_ `vneg.f32 D0, D0'
26336 * int32x2_t vneg_s32 (int32x2_t)
26337 _Form of expected instruction(s):_ `vneg.s32 D0, D0'
26339 * int16x4_t vneg_s16 (int16x4_t)
26340 _Form of expected instruction(s):_ `vneg.s16 D0, D0'
26342 * int8x8_t vneg_s8 (int8x8_t)
26343 _Form of expected instruction(s):_ `vneg.s8 D0, D0'
26345 * float32x4_t vnegq_f32 (float32x4_t)
26346 _Form of expected instruction(s):_ `vneg.f32 Q0, Q0'
26348 * int32x4_t vnegq_s32 (int32x4_t)
26349 _Form of expected instruction(s):_ `vneg.s32 Q0, Q0'
26351 * int16x8_t vnegq_s16 (int16x8_t)
26352 _Form of expected instruction(s):_ `vneg.s16 Q0, Q0'
26354 * int8x16_t vnegq_s8 (int8x16_t)
26355 _Form of expected instruction(s):_ `vneg.s8 Q0, Q0'
26357 * int32x2_t vqneg_s32 (int32x2_t)
26358 _Form of expected instruction(s):_ `vqneg.s32 D0, D0'
26360 * int16x4_t vqneg_s16 (int16x4_t)
26361 _Form of expected instruction(s):_ `vqneg.s16 D0, D0'
26363 * int8x8_t vqneg_s8 (int8x8_t)
26364 _Form of expected instruction(s):_ `vqneg.s8 D0, D0'
26366 * int32x4_t vqnegq_s32 (int32x4_t)
26367 _Form of expected instruction(s):_ `vqneg.s32 Q0, Q0'
26369 * int16x8_t vqnegq_s16 (int16x8_t)
26370 _Form of expected instruction(s):_ `vqneg.s16 Q0, Q0'
26372 * int8x16_t vqnegq_s8 (int8x16_t)
26373 _Form of expected instruction(s):_ `vqneg.s8 Q0, Q0'
26375 5.50.3.33 Bitwise not
26376 .....................
26378 * uint32x2_t vmvn_u32 (uint32x2_t)
26379 _Form of expected instruction(s):_ `vmvn D0, D0'
26381 * uint16x4_t vmvn_u16 (uint16x4_t)
26382 _Form of expected instruction(s):_ `vmvn D0, D0'
26384 * uint8x8_t vmvn_u8 (uint8x8_t)
26385 _Form of expected instruction(s):_ `vmvn D0, D0'
26387 * int32x2_t vmvn_s32 (int32x2_t)
26388 _Form of expected instruction(s):_ `vmvn D0, D0'
26390 * int16x4_t vmvn_s16 (int16x4_t)
26391 _Form of expected instruction(s):_ `vmvn D0, D0'
26393 * int8x8_t vmvn_s8 (int8x8_t)
26394 _Form of expected instruction(s):_ `vmvn D0, D0'
26396 * poly8x8_t vmvn_p8 (poly8x8_t)
26397 _Form of expected instruction(s):_ `vmvn D0, D0'
26399 * uint32x4_t vmvnq_u32 (uint32x4_t)
26400 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26402 * uint16x8_t vmvnq_u16 (uint16x8_t)
26403 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26405 * uint8x16_t vmvnq_u8 (uint8x16_t)
26406 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26408 * int32x4_t vmvnq_s32 (int32x4_t)
26409 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26411 * int16x8_t vmvnq_s16 (int16x8_t)
26412 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26414 * int8x16_t vmvnq_s8 (int8x16_t)
26415 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26417 * poly8x16_t vmvnq_p8 (poly8x16_t)
26418 _Form of expected instruction(s):_ `vmvn Q0, Q0'
26420 5.50.3.34 Count leading sign bits
26421 .................................
26423 * int32x2_t vcls_s32 (int32x2_t)
26424 _Form of expected instruction(s):_ `vcls.s32 D0, D0'
26426 * int16x4_t vcls_s16 (int16x4_t)
26427 _Form of expected instruction(s):_ `vcls.s16 D0, D0'
26429 * int8x8_t vcls_s8 (int8x8_t)
26430 _Form of expected instruction(s):_ `vcls.s8 D0, D0'
26432 * int32x4_t vclsq_s32 (int32x4_t)
26433 _Form of expected instruction(s):_ `vcls.s32 Q0, Q0'
26435 * int16x8_t vclsq_s16 (int16x8_t)
26436 _Form of expected instruction(s):_ `vcls.s16 Q0, Q0'
26438 * int8x16_t vclsq_s8 (int8x16_t)
26439 _Form of expected instruction(s):_ `vcls.s8 Q0, Q0'
26441 5.50.3.35 Count leading zeros
26442 .............................
26444 * uint32x2_t vclz_u32 (uint32x2_t)
26445 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
26447 * uint16x4_t vclz_u16 (uint16x4_t)
26448 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
26450 * uint8x8_t vclz_u8 (uint8x8_t)
26451 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
26453 * int32x2_t vclz_s32 (int32x2_t)
26454 _Form of expected instruction(s):_ `vclz.i32 D0, D0'
26456 * int16x4_t vclz_s16 (int16x4_t)
26457 _Form of expected instruction(s):_ `vclz.i16 D0, D0'
26459 * int8x8_t vclz_s8 (int8x8_t)
26460 _Form of expected instruction(s):_ `vclz.i8 D0, D0'
26462 * uint32x4_t vclzq_u32 (uint32x4_t)
26463 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
26465 * uint16x8_t vclzq_u16 (uint16x8_t)
26466 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
26468 * uint8x16_t vclzq_u8 (uint8x16_t)
26469 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
26471 * int32x4_t vclzq_s32 (int32x4_t)
26472 _Form of expected instruction(s):_ `vclz.i32 Q0, Q0'
26474 * int16x8_t vclzq_s16 (int16x8_t)
26475 _Form of expected instruction(s):_ `vclz.i16 Q0, Q0'
26477 * int8x16_t vclzq_s8 (int8x16_t)
26478 _Form of expected instruction(s):_ `vclz.i8 Q0, Q0'
26480 5.50.3.36 Count number of set bits
26481 ..................................
26483 * uint8x8_t vcnt_u8 (uint8x8_t)
26484 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26486 * int8x8_t vcnt_s8 (int8x8_t)
26487 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26489 * poly8x8_t vcnt_p8 (poly8x8_t)
26490 _Form of expected instruction(s):_ `vcnt.8 D0, D0'
26492 * uint8x16_t vcntq_u8 (uint8x16_t)
26493 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26495 * int8x16_t vcntq_s8 (int8x16_t)
26496 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26498 * poly8x16_t vcntq_p8 (poly8x16_t)
26499 _Form of expected instruction(s):_ `vcnt.8 Q0, Q0'
26501 5.50.3.37 Reciprocal estimate
26502 .............................
26504 * float32x2_t vrecpe_f32 (float32x2_t)
26505 _Form of expected instruction(s):_ `vrecpe.f32 D0, D0'
26507 * uint32x2_t vrecpe_u32 (uint32x2_t)
26508 _Form of expected instruction(s):_ `vrecpe.u32 D0, D0'
26510 * float32x4_t vrecpeq_f32 (float32x4_t)
26511 _Form of expected instruction(s):_ `vrecpe.f32 Q0, Q0'
26513 * uint32x4_t vrecpeq_u32 (uint32x4_t)
26514 _Form of expected instruction(s):_ `vrecpe.u32 Q0, Q0'
26516 5.50.3.38 Reciprocal square-root estimate
26517 .........................................
26519 * float32x2_t vrsqrte_f32 (float32x2_t)
26520 _Form of expected instruction(s):_ `vrsqrte.f32 D0, D0'
26522 * uint32x2_t vrsqrte_u32 (uint32x2_t)
26523 _Form of expected instruction(s):_ `vrsqrte.u32 D0, D0'
26525 * float32x4_t vrsqrteq_f32 (float32x4_t)
26526 _Form of expected instruction(s):_ `vrsqrte.f32 Q0, Q0'
26528 * uint32x4_t vrsqrteq_u32 (uint32x4_t)
26529 _Form of expected instruction(s):_ `vrsqrte.u32 Q0, Q0'
26531 5.50.3.39 Get lanes from a vector
26532 .................................
26534 * uint32_t vget_lane_u32 (uint32x2_t, const int)
26535 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
26537 * uint16_t vget_lane_u16 (uint16x4_t, const int)
26538 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26540 * uint8_t vget_lane_u8 (uint8x8_t, const int)
26541 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26543 * int32_t vget_lane_s32 (int32x2_t, const int)
26544 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
26546 * int16_t vget_lane_s16 (int16x4_t, const int)
26547 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
26549 * int8_t vget_lane_s8 (int8x8_t, const int)
26550 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
26552 * float32_t vget_lane_f32 (float32x2_t, const int)
26553 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
26555 * poly16_t vget_lane_p16 (poly16x4_t, const int)
26556 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26558 * poly8_t vget_lane_p8 (poly8x8_t, const int)
26559 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26561 * uint64_t vget_lane_u64 (uint64x1_t, const int)
26562 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26564 * int64_t vget_lane_s64 (int64x1_t, const int)
26565 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26567 * uint32_t vgetq_lane_u32 (uint32x4_t, const int)
26568 _Form of expected instruction(s):_ `vmov.u32 R0, D0[0]'
26570 * uint16_t vgetq_lane_u16 (uint16x8_t, const int)
26571 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26573 * uint8_t vgetq_lane_u8 (uint8x16_t, const int)
26574 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26576 * int32_t vgetq_lane_s32 (int32x4_t, const int)
26577 _Form of expected instruction(s):_ `vmov.s32 R0, D0[0]'
26579 * int16_t vgetq_lane_s16 (int16x8_t, const int)
26580 _Form of expected instruction(s):_ `vmov.s16 R0, D0[0]'
26582 * int8_t vgetq_lane_s8 (int8x16_t, const int)
26583 _Form of expected instruction(s):_ `vmov.s8 R0, D0[0]'
26585 * float32_t vgetq_lane_f32 (float32x4_t, const int)
26586 _Form of expected instruction(s):_ `vmov.f32 R0, D0[0]'
26588 * poly16_t vgetq_lane_p16 (poly16x8_t, const int)
26589 _Form of expected instruction(s):_ `vmov.u16 R0, D0[0]'
26591 * poly8_t vgetq_lane_p8 (poly8x16_t, const int)
26592 _Form of expected instruction(s):_ `vmov.u8 R0, D0[0]'
26594 * uint64_t vgetq_lane_u64 (uint64x2_t, const int)
26595 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26597 * int64_t vgetq_lane_s64 (int64x2_t, const int)
26598 _Form of expected instruction(s):_ `vmov R0, R0, D0'
26600 5.50.3.40 Set lanes in a vector
26601 ...............................
26603 * uint32x2_t vset_lane_u32 (uint32_t, uint32x2_t, const int)
26604 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26606 * uint16x4_t vset_lane_u16 (uint16_t, uint16x4_t, const int)
26607 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26609 * uint8x8_t vset_lane_u8 (uint8_t, uint8x8_t, const int)
26610 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26612 * int32x2_t vset_lane_s32 (int32_t, int32x2_t, const int)
26613 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26615 * int16x4_t vset_lane_s16 (int16_t, int16x4_t, const int)
26616 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26618 * int8x8_t vset_lane_s8 (int8_t, int8x8_t, const int)
26619 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26621 * float32x2_t vset_lane_f32 (float32_t, float32x2_t, const int)
26622 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26624 * poly16x4_t vset_lane_p16 (poly16_t, poly16x4_t, const int)
26625 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26627 * poly8x8_t vset_lane_p8 (poly8_t, poly8x8_t, const int)
26628 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26630 * uint64x1_t vset_lane_u64 (uint64_t, uint64x1_t, const int)
26631 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26633 * int64x1_t vset_lane_s64 (int64_t, int64x1_t, const int)
26634 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26636 * uint32x4_t vsetq_lane_u32 (uint32_t, uint32x4_t, const int)
26637 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26639 * uint16x8_t vsetq_lane_u16 (uint16_t, uint16x8_t, const int)
26640 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26642 * uint8x16_t vsetq_lane_u8 (uint8_t, uint8x16_t, const int)
26643 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26645 * int32x4_t vsetq_lane_s32 (int32_t, int32x4_t, const int)
26646 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26648 * int16x8_t vsetq_lane_s16 (int16_t, int16x8_t, const int)
26649 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26651 * int8x16_t vsetq_lane_s8 (int8_t, int8x16_t, const int)
26652 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26654 * float32x4_t vsetq_lane_f32 (float32_t, float32x4_t, const int)
26655 _Form of expected instruction(s):_ `vmov.32 D0[0], R0'
26657 * poly16x8_t vsetq_lane_p16 (poly16_t, poly16x8_t, const int)
26658 _Form of expected instruction(s):_ `vmov.16 D0[0], R0'
26660 * poly8x16_t vsetq_lane_p8 (poly8_t, poly8x16_t, const int)
26661 _Form of expected instruction(s):_ `vmov.8 D0[0], R0'
26663 * uint64x2_t vsetq_lane_u64 (uint64_t, uint64x2_t, const int)
26664 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26666 * int64x2_t vsetq_lane_s64 (int64_t, int64x2_t, const int)
26667 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26669 5.50.3.41 Create vector from literal bit pattern
26670 ................................................
26672 * uint32x2_t vcreate_u32 (uint64_t)
26674 * uint16x4_t vcreate_u16 (uint64_t)
26676 * uint8x8_t vcreate_u8 (uint64_t)
26678 * int32x2_t vcreate_s32 (uint64_t)
26680 * int16x4_t vcreate_s16 (uint64_t)
26682 * int8x8_t vcreate_s8 (uint64_t)
26684 * uint64x1_t vcreate_u64 (uint64_t)
26686 * int64x1_t vcreate_s64 (uint64_t)
26688 * float32x2_t vcreate_f32 (uint64_t)
26690 * poly16x4_t vcreate_p16 (uint64_t)
26692 * poly8x8_t vcreate_p8 (uint64_t)
26694 5.50.3.42 Set all lanes to the same value
26695 .........................................
26697 * uint32x2_t vdup_n_u32 (uint32_t)
26698 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26700 * uint16x4_t vdup_n_u16 (uint16_t)
26701 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26703 * uint8x8_t vdup_n_u8 (uint8_t)
26704 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26706 * int32x2_t vdup_n_s32 (int32_t)
26707 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26709 * int16x4_t vdup_n_s16 (int16_t)
26710 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26712 * int8x8_t vdup_n_s8 (int8_t)
26713 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26715 * float32x2_t vdup_n_f32 (float32_t)
26716 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26718 * poly16x4_t vdup_n_p16 (poly16_t)
26719 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26721 * poly8x8_t vdup_n_p8 (poly8_t)
26722 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26724 * uint64x1_t vdup_n_u64 (uint64_t)
26725 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26727 * int64x1_t vdup_n_s64 (int64_t)
26728 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26730 * uint32x4_t vdupq_n_u32 (uint32_t)
26731 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26733 * uint16x8_t vdupq_n_u16 (uint16_t)
26734 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26736 * uint8x16_t vdupq_n_u8 (uint8_t)
26737 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26739 * int32x4_t vdupq_n_s32 (int32_t)
26740 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26742 * int16x8_t vdupq_n_s16 (int16_t)
26743 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26745 * int8x16_t vdupq_n_s8 (int8_t)
26746 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26748 * float32x4_t vdupq_n_f32 (float32_t)
26749 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26751 * poly16x8_t vdupq_n_p16 (poly16_t)
26752 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26754 * poly8x16_t vdupq_n_p8 (poly8_t)
26755 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26757 * uint64x2_t vdupq_n_u64 (uint64_t)
26758 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26760 * int64x2_t vdupq_n_s64 (int64_t)
26761 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26763 * uint32x2_t vmov_n_u32 (uint32_t)
26764 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26766 * uint16x4_t vmov_n_u16 (uint16_t)
26767 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26769 * uint8x8_t vmov_n_u8 (uint8_t)
26770 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26772 * int32x2_t vmov_n_s32 (int32_t)
26773 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26775 * int16x4_t vmov_n_s16 (int16_t)
26776 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26778 * int8x8_t vmov_n_s8 (int8_t)
26779 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26781 * float32x2_t vmov_n_f32 (float32_t)
26782 _Form of expected instruction(s):_ `vdup.32 D0, R0'
26784 * poly16x4_t vmov_n_p16 (poly16_t)
26785 _Form of expected instruction(s):_ `vdup.16 D0, R0'
26787 * poly8x8_t vmov_n_p8 (poly8_t)
26788 _Form of expected instruction(s):_ `vdup.8 D0, R0'
26790 * uint64x1_t vmov_n_u64 (uint64_t)
26791 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26793 * int64x1_t vmov_n_s64 (int64_t)
26794 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26796 * uint32x4_t vmovq_n_u32 (uint32_t)
26797 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26799 * uint16x8_t vmovq_n_u16 (uint16_t)
26800 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26802 * uint8x16_t vmovq_n_u8 (uint8_t)
26803 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26805 * int32x4_t vmovq_n_s32 (int32_t)
26806 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26808 * int16x8_t vmovq_n_s16 (int16_t)
26809 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26811 * int8x16_t vmovq_n_s8 (int8_t)
26812 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26814 * float32x4_t vmovq_n_f32 (float32_t)
26815 _Form of expected instruction(s):_ `vdup.32 Q0, R0'
26817 * poly16x8_t vmovq_n_p16 (poly16_t)
26818 _Form of expected instruction(s):_ `vdup.16 Q0, R0'
26820 * poly8x16_t vmovq_n_p8 (poly8_t)
26821 _Form of expected instruction(s):_ `vdup.8 Q0, R0'
26823 * uint64x2_t vmovq_n_u64 (uint64_t)
26824 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26826 * int64x2_t vmovq_n_s64 (int64_t)
26827 _Form of expected instruction(s):_ `vmov D0, R0, R0'
26829 * uint32x2_t vdup_lane_u32 (uint32x2_t, const int)
26830 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26832 * uint16x4_t vdup_lane_u16 (uint16x4_t, const int)
26833 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26835 * uint8x8_t vdup_lane_u8 (uint8x8_t, const int)
26836 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26838 * int32x2_t vdup_lane_s32 (int32x2_t, const int)
26839 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26841 * int16x4_t vdup_lane_s16 (int16x4_t, const int)
26842 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26844 * int8x8_t vdup_lane_s8 (int8x8_t, const int)
26845 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26847 * float32x2_t vdup_lane_f32 (float32x2_t, const int)
26848 _Form of expected instruction(s):_ `vdup.32 D0, D0[0]'
26850 * poly16x4_t vdup_lane_p16 (poly16x4_t, const int)
26851 _Form of expected instruction(s):_ `vdup.16 D0, D0[0]'
26853 * poly8x8_t vdup_lane_p8 (poly8x8_t, const int)
26854 _Form of expected instruction(s):_ `vdup.8 D0, D0[0]'
26856 * uint64x1_t vdup_lane_u64 (uint64x1_t, const int)
26858 * int64x1_t vdup_lane_s64 (int64x1_t, const int)
26860 * uint32x4_t vdupq_lane_u32 (uint32x2_t, const int)
26861 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26863 * uint16x8_t vdupq_lane_u16 (uint16x4_t, const int)
26864 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26866 * uint8x16_t vdupq_lane_u8 (uint8x8_t, const int)
26867 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26869 * int32x4_t vdupq_lane_s32 (int32x2_t, const int)
26870 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26872 * int16x8_t vdupq_lane_s16 (int16x4_t, const int)
26873 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26875 * int8x16_t vdupq_lane_s8 (int8x8_t, const int)
26876 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26878 * float32x4_t vdupq_lane_f32 (float32x2_t, const int)
26879 _Form of expected instruction(s):_ `vdup.32 Q0, D0[0]'
26881 * poly16x8_t vdupq_lane_p16 (poly16x4_t, const int)
26882 _Form of expected instruction(s):_ `vdup.16 Q0, D0[0]'
26884 * poly8x16_t vdupq_lane_p8 (poly8x8_t, const int)
26885 _Form of expected instruction(s):_ `vdup.8 Q0, D0[0]'
26887 * uint64x2_t vdupq_lane_u64 (uint64x1_t, const int)
26889 * int64x2_t vdupq_lane_s64 (int64x1_t, const int)
26891 5.50.3.43 Combining vectors
26892 ...........................
26894 * uint32x4_t vcombine_u32 (uint32x2_t, uint32x2_t)
26896 * uint16x8_t vcombine_u16 (uint16x4_t, uint16x4_t)
26898 * uint8x16_t vcombine_u8 (uint8x8_t, uint8x8_t)
26900 * int32x4_t vcombine_s32 (int32x2_t, int32x2_t)
26902 * int16x8_t vcombine_s16 (int16x4_t, int16x4_t)
26904 * int8x16_t vcombine_s8 (int8x8_t, int8x8_t)
26906 * uint64x2_t vcombine_u64 (uint64x1_t, uint64x1_t)
26908 * int64x2_t vcombine_s64 (int64x1_t, int64x1_t)
26910 * float32x4_t vcombine_f32 (float32x2_t, float32x2_t)
26912 * poly16x8_t vcombine_p16 (poly16x4_t, poly16x4_t)
26914 * poly8x16_t vcombine_p8 (poly8x8_t, poly8x8_t)
26916 5.50.3.44 Splitting vectors
26917 ...........................
26919 * uint32x2_t vget_high_u32 (uint32x4_t)
26921 * uint16x4_t vget_high_u16 (uint16x8_t)
26923 * uint8x8_t vget_high_u8 (uint8x16_t)
26925 * int32x2_t vget_high_s32 (int32x4_t)
26927 * int16x4_t vget_high_s16 (int16x8_t)
26929 * int8x8_t vget_high_s8 (int8x16_t)
26931 * uint64x1_t vget_high_u64 (uint64x2_t)
26933 * int64x1_t vget_high_s64 (int64x2_t)
26935 * float32x2_t vget_high_f32 (float32x4_t)
26937 * poly16x4_t vget_high_p16 (poly16x8_t)
26939 * poly8x8_t vget_high_p8 (poly8x16_t)
26941 * uint32x2_t vget_low_u32 (uint32x4_t)
26942 _Form of expected instruction(s):_ `vmov D0, D0'
26944 * uint16x4_t vget_low_u16 (uint16x8_t)
26945 _Form of expected instruction(s):_ `vmov D0, D0'
26947 * uint8x8_t vget_low_u8 (uint8x16_t)
26948 _Form of expected instruction(s):_ `vmov D0, D0'
26950 * int32x2_t vget_low_s32 (int32x4_t)
26951 _Form of expected instruction(s):_ `vmov D0, D0'
26953 * int16x4_t vget_low_s16 (int16x8_t)
26954 _Form of expected instruction(s):_ `vmov D0, D0'
26956 * int8x8_t vget_low_s8 (int8x16_t)
26957 _Form of expected instruction(s):_ `vmov D0, D0'
26959 * uint64x1_t vget_low_u64 (uint64x2_t)
26960 _Form of expected instruction(s):_ `vmov D0, D0'
26962 * int64x1_t vget_low_s64 (int64x2_t)
26963 _Form of expected instruction(s):_ `vmov D0, D0'
26965 * float32x2_t vget_low_f32 (float32x4_t)
26966 _Form of expected instruction(s):_ `vmov D0, D0'
26968 * poly16x4_t vget_low_p16 (poly16x8_t)
26969 _Form of expected instruction(s):_ `vmov D0, D0'
26971 * poly8x8_t vget_low_p8 (poly8x16_t)
26972 _Form of expected instruction(s):_ `vmov D0, D0'
26974 5.50.3.45 Conversions
26975 .....................
26977 * float32x2_t vcvt_f32_u32 (uint32x2_t)
26978 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0'
26980 * float32x2_t vcvt_f32_s32 (int32x2_t)
26981 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0'
26983 * uint32x2_t vcvt_u32_f32 (float32x2_t)
26984 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0'
26986 * int32x2_t vcvt_s32_f32 (float32x2_t)
26987 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0'
26989 * float32x4_t vcvtq_f32_u32 (uint32x4_t)
26990 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0'
26992 * float32x4_t vcvtq_f32_s32 (int32x4_t)
26993 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0'
26995 * uint32x4_t vcvtq_u32_f32 (float32x4_t)
26996 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0'
26998 * int32x4_t vcvtq_s32_f32 (float32x4_t)
26999 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0'
27001 * float32x2_t vcvt_n_f32_u32 (uint32x2_t, const int)
27002 _Form of expected instruction(s):_ `vcvt.f32.u32 D0, D0, #0'
27004 * float32x2_t vcvt_n_f32_s32 (int32x2_t, const int)
27005 _Form of expected instruction(s):_ `vcvt.f32.s32 D0, D0, #0'
27007 * uint32x2_t vcvt_n_u32_f32 (float32x2_t, const int)
27008 _Form of expected instruction(s):_ `vcvt.u32.f32 D0, D0, #0'
27010 * int32x2_t vcvt_n_s32_f32 (float32x2_t, const int)
27011 _Form of expected instruction(s):_ `vcvt.s32.f32 D0, D0, #0'
27013 * float32x4_t vcvtq_n_f32_u32 (uint32x4_t, const int)
27014 _Form of expected instruction(s):_ `vcvt.f32.u32 Q0, Q0, #0'
27016 * float32x4_t vcvtq_n_f32_s32 (int32x4_t, const int)
27017 _Form of expected instruction(s):_ `vcvt.f32.s32 Q0, Q0, #0'
27019 * uint32x4_t vcvtq_n_u32_f32 (float32x4_t, const int)
27020 _Form of expected instruction(s):_ `vcvt.u32.f32 Q0, Q0, #0'
27022 * int32x4_t vcvtq_n_s32_f32 (float32x4_t, const int)
27023 _Form of expected instruction(s):_ `vcvt.s32.f32 Q0, Q0, #0'
27025 5.50.3.46 Move, single_opcode narrowing
27026 .......................................
27028 * uint32x2_t vmovn_u64 (uint64x2_t)
27029 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
27031 * uint16x4_t vmovn_u32 (uint32x4_t)
27032 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
27034 * uint8x8_t vmovn_u16 (uint16x8_t)
27035 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
27037 * int32x2_t vmovn_s64 (int64x2_t)
27038 _Form of expected instruction(s):_ `vmovn.i64 D0, Q0'
27040 * int16x4_t vmovn_s32 (int32x4_t)
27041 _Form of expected instruction(s):_ `vmovn.i32 D0, Q0'
27043 * int8x8_t vmovn_s16 (int16x8_t)
27044 _Form of expected instruction(s):_ `vmovn.i16 D0, Q0'
27046 * uint32x2_t vqmovn_u64 (uint64x2_t)
27047 _Form of expected instruction(s):_ `vqmovn.u64 D0, Q0'
27049 * uint16x4_t vqmovn_u32 (uint32x4_t)
27050 _Form of expected instruction(s):_ `vqmovn.u32 D0, Q0'
27052 * uint8x8_t vqmovn_u16 (uint16x8_t)
27053 _Form of expected instruction(s):_ `vqmovn.u16 D0, Q0'
27055 * int32x2_t vqmovn_s64 (int64x2_t)
27056 _Form of expected instruction(s):_ `vqmovn.s64 D0, Q0'
27058 * int16x4_t vqmovn_s32 (int32x4_t)
27059 _Form of expected instruction(s):_ `vqmovn.s32 D0, Q0'
27061 * int8x8_t vqmovn_s16 (int16x8_t)
27062 _Form of expected instruction(s):_ `vqmovn.s16 D0, Q0'
27064 * uint32x2_t vqmovun_s64 (int64x2_t)
27065 _Form of expected instruction(s):_ `vqmovun.s64 D0, Q0'
27067 * uint16x4_t vqmovun_s32 (int32x4_t)
27068 _Form of expected instruction(s):_ `vqmovun.s32 D0, Q0'
27070 * uint8x8_t vqmovun_s16 (int16x8_t)
27071 _Form of expected instruction(s):_ `vqmovun.s16 D0, Q0'
27073 5.50.3.47 Move, single_opcode long
27074 ..................................
27076 * uint64x2_t vmovl_u32 (uint32x2_t)
27077 _Form of expected instruction(s):_ `vmovl.u32 Q0, D0'
27079 * uint32x4_t vmovl_u16 (uint16x4_t)
27080 _Form of expected instruction(s):_ `vmovl.u16 Q0, D0'
27082 * uint16x8_t vmovl_u8 (uint8x8_t)
27083 _Form of expected instruction(s):_ `vmovl.u8 Q0, D0'
27085 * int64x2_t vmovl_s32 (int32x2_t)
27086 _Form of expected instruction(s):_ `vmovl.s32 Q0, D0'
27088 * int32x4_t vmovl_s16 (int16x4_t)
27089 _Form of expected instruction(s):_ `vmovl.s16 Q0, D0'
27091 * int16x8_t vmovl_s8 (int8x8_t)
27092 _Form of expected instruction(s):_ `vmovl.s8 Q0, D0'
27094 5.50.3.48 Table lookup
27095 ......................
27097 * poly8x8_t vtbl1_p8 (poly8x8_t, uint8x8_t)
27098 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27100 * int8x8_t vtbl1_s8 (int8x8_t, int8x8_t)
27101 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27103 * uint8x8_t vtbl1_u8 (uint8x8_t, uint8x8_t)
27104 _Form of expected instruction(s):_ `vtbl.8 D0, {D0}, D0'
27106 * poly8x8_t vtbl2_p8 (poly8x8x2_t, uint8x8_t)
27107 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27109 * int8x8_t vtbl2_s8 (int8x8x2_t, int8x8_t)
27110 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27112 * uint8x8_t vtbl2_u8 (uint8x8x2_t, uint8x8_t)
27113 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1}, D0'
27115 * poly8x8_t vtbl3_p8 (poly8x8x3_t, uint8x8_t)
27116 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27118 * int8x8_t vtbl3_s8 (int8x8x3_t, int8x8_t)
27119 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27121 * uint8x8_t vtbl3_u8 (uint8x8x3_t, uint8x8_t)
27122 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2}, D0'
27124 * poly8x8_t vtbl4_p8 (poly8x8x4_t, uint8x8_t)
27125 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27128 * int8x8_t vtbl4_s8 (int8x8x4_t, int8x8_t)
27129 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27132 * uint8x8_t vtbl4_u8 (uint8x8x4_t, uint8x8_t)
27133 _Form of expected instruction(s):_ `vtbl.8 D0, {D0, D1, D2, D3},
27136 5.50.3.49 Extended table lookup
27137 ...............................
27139 * poly8x8_t vtbx1_p8 (poly8x8_t, poly8x8_t, uint8x8_t)
27140 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27142 * int8x8_t vtbx1_s8 (int8x8_t, int8x8_t, int8x8_t)
27143 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27145 * uint8x8_t vtbx1_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
27146 _Form of expected instruction(s):_ `vtbx.8 D0, {D0}, D0'
27148 * poly8x8_t vtbx2_p8 (poly8x8_t, poly8x8x2_t, uint8x8_t)
27149 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27151 * int8x8_t vtbx2_s8 (int8x8_t, int8x8x2_t, int8x8_t)
27152 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27154 * uint8x8_t vtbx2_u8 (uint8x8_t, uint8x8x2_t, uint8x8_t)
27155 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1}, D0'
27157 * poly8x8_t vtbx3_p8 (poly8x8_t, poly8x8x3_t, uint8x8_t)
27158 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27160 * int8x8_t vtbx3_s8 (int8x8_t, int8x8x3_t, int8x8_t)
27161 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27163 * uint8x8_t vtbx3_u8 (uint8x8_t, uint8x8x3_t, uint8x8_t)
27164 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2}, D0'
27166 * poly8x8_t vtbx4_p8 (poly8x8_t, poly8x8x4_t, uint8x8_t)
27167 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27170 * int8x8_t vtbx4_s8 (int8x8_t, int8x8x4_t, int8x8_t)
27171 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27174 * uint8x8_t vtbx4_u8 (uint8x8_t, uint8x8x4_t, uint8x8_t)
27175 _Form of expected instruction(s):_ `vtbx.8 D0, {D0, D1, D2, D3},
27178 5.50.3.50 Multiply, lane
27179 ........................
27181 * float32x2_t vmul_lane_f32 (float32x2_t, float32x2_t, const int)
27182 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
27184 * uint32x2_t vmul_lane_u32 (uint32x2_t, uint32x2_t, const int)
27185 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27187 * uint16x4_t vmul_lane_u16 (uint16x4_t, uint16x4_t, const int)
27188 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27190 * int32x2_t vmul_lane_s32 (int32x2_t, int32x2_t, const int)
27191 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27193 * int16x4_t vmul_lane_s16 (int16x4_t, int16x4_t, const int)
27194 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27196 * float32x4_t vmulq_lane_f32 (float32x4_t, float32x2_t, const int)
27197 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
27199 * uint32x4_t vmulq_lane_u32 (uint32x4_t, uint32x2_t, const int)
27200 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27202 * uint16x8_t vmulq_lane_u16 (uint16x8_t, uint16x4_t, const int)
27203 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27205 * int32x4_t vmulq_lane_s32 (int32x4_t, int32x2_t, const int)
27206 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27208 * int16x8_t vmulq_lane_s16 (int16x8_t, int16x4_t, const int)
27209 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27211 5.50.3.51 Long multiply, lane
27212 .............................
27214 * uint64x2_t vmull_lane_u32 (uint32x2_t, uint32x2_t, const int)
27215 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
27217 * uint32x4_t vmull_lane_u16 (uint16x4_t, uint16x4_t, const int)
27218 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
27220 * int64x2_t vmull_lane_s32 (int32x2_t, int32x2_t, const int)
27221 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
27223 * int32x4_t vmull_lane_s16 (int16x4_t, int16x4_t, const int)
27224 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
27226 5.50.3.52 Saturating doubling long multiply, lane
27227 .................................................
27229 * int64x2_t vqdmull_lane_s32 (int32x2_t, int32x2_t, const int)
27230 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
27232 * int32x4_t vqdmull_lane_s16 (int16x4_t, int16x4_t, const int)
27233 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
27235 5.50.3.53 Saturating doubling multiply high, lane
27236 .................................................
27238 * int32x4_t vqdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
27239 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
27241 * int16x8_t vqdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
27242 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
27244 * int32x2_t vqdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
27245 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
27247 * int16x4_t vqdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
27248 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
27250 * int32x4_t vqrdmulhq_lane_s32 (int32x4_t, int32x2_t, const int)
27251 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
27253 * int16x8_t vqrdmulhq_lane_s16 (int16x8_t, int16x4_t, const int)
27254 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
27256 * int32x2_t vqrdmulh_lane_s32 (int32x2_t, int32x2_t, const int)
27257 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
27259 * int16x4_t vqrdmulh_lane_s16 (int16x4_t, int16x4_t, const int)
27260 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
27262 5.50.3.54 Multiply-accumulate, lane
27263 ...................................
27265 * float32x2_t vmla_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
27267 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
27269 * uint32x2_t vmla_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
27271 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27273 * uint16x4_t vmla_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
27275 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27277 * int32x2_t vmla_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
27279 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27281 * int16x4_t vmla_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
27283 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27285 * float32x4_t vmlaq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
27287 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
27289 * uint32x4_t vmlaq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
27291 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27293 * uint16x8_t vmlaq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
27295 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27297 * int32x4_t vmlaq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
27299 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27301 * int16x8_t vmlaq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
27303 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27305 * uint64x2_t vmlal_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
27307 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
27309 * uint32x4_t vmlal_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
27311 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
27313 * int64x2_t vmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27315 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
27317 * int32x4_t vmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27319 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
27321 * int64x2_t vqdmlal_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27323 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
27325 * int32x4_t vqdmlal_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27327 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
27329 5.50.3.55 Multiply-subtract, lane
27330 .................................
27332 * float32x2_t vmls_lane_f32 (float32x2_t, float32x2_t, float32x2_t,
27334 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
27336 * uint32x2_t vmls_lane_u32 (uint32x2_t, uint32x2_t, uint32x2_t,
27338 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27340 * uint16x4_t vmls_lane_u16 (uint16x4_t, uint16x4_t, uint16x4_t,
27342 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27344 * int32x2_t vmls_lane_s32 (int32x2_t, int32x2_t, int32x2_t, const
27346 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27348 * int16x4_t vmls_lane_s16 (int16x4_t, int16x4_t, int16x4_t, const
27350 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27352 * float32x4_t vmlsq_lane_f32 (float32x4_t, float32x4_t, float32x2_t,
27354 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
27356 * uint32x4_t vmlsq_lane_u32 (uint32x4_t, uint32x4_t, uint32x2_t,
27358 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27360 * uint16x8_t vmlsq_lane_u16 (uint16x8_t, uint16x8_t, uint16x4_t,
27362 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27364 * int32x4_t vmlsq_lane_s32 (int32x4_t, int32x4_t, int32x2_t, const
27366 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27368 * int16x8_t vmlsq_lane_s16 (int16x8_t, int16x8_t, int16x4_t, const
27370 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27372 * uint64x2_t vmlsl_lane_u32 (uint64x2_t, uint32x2_t, uint32x2_t,
27374 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
27376 * uint32x4_t vmlsl_lane_u16 (uint32x4_t, uint16x4_t, uint16x4_t,
27378 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
27380 * int64x2_t vmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27382 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
27384 * int32x4_t vmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27386 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
27388 * int64x2_t vqdmlsl_lane_s32 (int64x2_t, int32x2_t, int32x2_t, const
27390 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
27392 * int32x4_t vqdmlsl_lane_s16 (int32x4_t, int16x4_t, int16x4_t, const
27394 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
27396 5.50.3.56 Vector multiply by scalar
27397 ...................................
27399 * float32x2_t vmul_n_f32 (float32x2_t, float32_t)
27400 _Form of expected instruction(s):_ `vmul.f32 D0, D0, D0[0]'
27402 * uint32x2_t vmul_n_u32 (uint32x2_t, uint32_t)
27403 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27405 * uint16x4_t vmul_n_u16 (uint16x4_t, uint16_t)
27406 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27408 * int32x2_t vmul_n_s32 (int32x2_t, int32_t)
27409 _Form of expected instruction(s):_ `vmul.i32 D0, D0, D0[0]'
27411 * int16x4_t vmul_n_s16 (int16x4_t, int16_t)
27412 _Form of expected instruction(s):_ `vmul.i16 D0, D0, D0[0]'
27414 * float32x4_t vmulq_n_f32 (float32x4_t, float32_t)
27415 _Form of expected instruction(s):_ `vmul.f32 Q0, Q0, D0[0]'
27417 * uint32x4_t vmulq_n_u32 (uint32x4_t, uint32_t)
27418 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27420 * uint16x8_t vmulq_n_u16 (uint16x8_t, uint16_t)
27421 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27423 * int32x4_t vmulq_n_s32 (int32x4_t, int32_t)
27424 _Form of expected instruction(s):_ `vmul.i32 Q0, Q0, D0[0]'
27426 * int16x8_t vmulq_n_s16 (int16x8_t, int16_t)
27427 _Form of expected instruction(s):_ `vmul.i16 Q0, Q0, D0[0]'
27429 5.50.3.57 Vector long multiply by scalar
27430 ........................................
27432 * uint64x2_t vmull_n_u32 (uint32x2_t, uint32_t)
27433 _Form of expected instruction(s):_ `vmull.u32 Q0, D0, D0[0]'
27435 * uint32x4_t vmull_n_u16 (uint16x4_t, uint16_t)
27436 _Form of expected instruction(s):_ `vmull.u16 Q0, D0, D0[0]'
27438 * int64x2_t vmull_n_s32 (int32x2_t, int32_t)
27439 _Form of expected instruction(s):_ `vmull.s32 Q0, D0, D0[0]'
27441 * int32x4_t vmull_n_s16 (int16x4_t, int16_t)
27442 _Form of expected instruction(s):_ `vmull.s16 Q0, D0, D0[0]'
27444 5.50.3.58 Vector saturating doubling long multiply by scalar
27445 ............................................................
27447 * int64x2_t vqdmull_n_s32 (int32x2_t, int32_t)
27448 _Form of expected instruction(s):_ `vqdmull.s32 Q0, D0, D0[0]'
27450 * int32x4_t vqdmull_n_s16 (int16x4_t, int16_t)
27451 _Form of expected instruction(s):_ `vqdmull.s16 Q0, D0, D0[0]'
27453 5.50.3.59 Vector saturating doubling multiply high by scalar
27454 ............................................................
27456 * int32x4_t vqdmulhq_n_s32 (int32x4_t, int32_t)
27457 _Form of expected instruction(s):_ `vqdmulh.s32 Q0, Q0, D0[0]'
27459 * int16x8_t vqdmulhq_n_s16 (int16x8_t, int16_t)
27460 _Form of expected instruction(s):_ `vqdmulh.s16 Q0, Q0, D0[0]'
27462 * int32x2_t vqdmulh_n_s32 (int32x2_t, int32_t)
27463 _Form of expected instruction(s):_ `vqdmulh.s32 D0, D0, D0[0]'
27465 * int16x4_t vqdmulh_n_s16 (int16x4_t, int16_t)
27466 _Form of expected instruction(s):_ `vqdmulh.s16 D0, D0, D0[0]'
27468 * int32x4_t vqrdmulhq_n_s32 (int32x4_t, int32_t)
27469 _Form of expected instruction(s):_ `vqrdmulh.s32 Q0, Q0, D0[0]'
27471 * int16x8_t vqrdmulhq_n_s16 (int16x8_t, int16_t)
27472 _Form of expected instruction(s):_ `vqrdmulh.s16 Q0, Q0, D0[0]'
27474 * int32x2_t vqrdmulh_n_s32 (int32x2_t, int32_t)
27475 _Form of expected instruction(s):_ `vqrdmulh.s32 D0, D0, D0[0]'
27477 * int16x4_t vqrdmulh_n_s16 (int16x4_t, int16_t)
27478 _Form of expected instruction(s):_ `vqrdmulh.s16 D0, D0, D0[0]'
27480 5.50.3.60 Vector multiply-accumulate by scalar
27481 ..............................................
27483 * float32x2_t vmla_n_f32 (float32x2_t, float32x2_t, float32_t)
27484 _Form of expected instruction(s):_ `vmla.f32 D0, D0, D0[0]'
27486 * uint32x2_t vmla_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
27487 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27489 * uint16x4_t vmla_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
27490 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27492 * int32x2_t vmla_n_s32 (int32x2_t, int32x2_t, int32_t)
27493 _Form of expected instruction(s):_ `vmla.i32 D0, D0, D0[0]'
27495 * int16x4_t vmla_n_s16 (int16x4_t, int16x4_t, int16_t)
27496 _Form of expected instruction(s):_ `vmla.i16 D0, D0, D0[0]'
27498 * float32x4_t vmlaq_n_f32 (float32x4_t, float32x4_t, float32_t)
27499 _Form of expected instruction(s):_ `vmla.f32 Q0, Q0, D0[0]'
27501 * uint32x4_t vmlaq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
27502 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27504 * uint16x8_t vmlaq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
27505 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27507 * int32x4_t vmlaq_n_s32 (int32x4_t, int32x4_t, int32_t)
27508 _Form of expected instruction(s):_ `vmla.i32 Q0, Q0, D0[0]'
27510 * int16x8_t vmlaq_n_s16 (int16x8_t, int16x8_t, int16_t)
27511 _Form of expected instruction(s):_ `vmla.i16 Q0, Q0, D0[0]'
27513 * uint64x2_t vmlal_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
27514 _Form of expected instruction(s):_ `vmlal.u32 Q0, D0, D0[0]'
27516 * uint32x4_t vmlal_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
27517 _Form of expected instruction(s):_ `vmlal.u16 Q0, D0, D0[0]'
27519 * int64x2_t vmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
27520 _Form of expected instruction(s):_ `vmlal.s32 Q0, D0, D0[0]'
27522 * int32x4_t vmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
27523 _Form of expected instruction(s):_ `vmlal.s16 Q0, D0, D0[0]'
27525 * int64x2_t vqdmlal_n_s32 (int64x2_t, int32x2_t, int32_t)
27526 _Form of expected instruction(s):_ `vqdmlal.s32 Q0, D0, D0[0]'
27528 * int32x4_t vqdmlal_n_s16 (int32x4_t, int16x4_t, int16_t)
27529 _Form of expected instruction(s):_ `vqdmlal.s16 Q0, D0, D0[0]'
27531 5.50.3.61 Vector multiply-subtract by scalar
27532 ............................................
27534 * float32x2_t vmls_n_f32 (float32x2_t, float32x2_t, float32_t)
27535 _Form of expected instruction(s):_ `vmls.f32 D0, D0, D0[0]'
27537 * uint32x2_t vmls_n_u32 (uint32x2_t, uint32x2_t, uint32_t)
27538 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27540 * uint16x4_t vmls_n_u16 (uint16x4_t, uint16x4_t, uint16_t)
27541 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27543 * int32x2_t vmls_n_s32 (int32x2_t, int32x2_t, int32_t)
27544 _Form of expected instruction(s):_ `vmls.i32 D0, D0, D0[0]'
27546 * int16x4_t vmls_n_s16 (int16x4_t, int16x4_t, int16_t)
27547 _Form of expected instruction(s):_ `vmls.i16 D0, D0, D0[0]'
27549 * float32x4_t vmlsq_n_f32 (float32x4_t, float32x4_t, float32_t)
27550 _Form of expected instruction(s):_ `vmls.f32 Q0, Q0, D0[0]'
27552 * uint32x4_t vmlsq_n_u32 (uint32x4_t, uint32x4_t, uint32_t)
27553 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27555 * uint16x8_t vmlsq_n_u16 (uint16x8_t, uint16x8_t, uint16_t)
27556 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27558 * int32x4_t vmlsq_n_s32 (int32x4_t, int32x4_t, int32_t)
27559 _Form of expected instruction(s):_ `vmls.i32 Q0, Q0, D0[0]'
27561 * int16x8_t vmlsq_n_s16 (int16x8_t, int16x8_t, int16_t)
27562 _Form of expected instruction(s):_ `vmls.i16 Q0, Q0, D0[0]'
27564 * uint64x2_t vmlsl_n_u32 (uint64x2_t, uint32x2_t, uint32_t)
27565 _Form of expected instruction(s):_ `vmlsl.u32 Q0, D0, D0[0]'
27567 * uint32x4_t vmlsl_n_u16 (uint32x4_t, uint16x4_t, uint16_t)
27568 _Form of expected instruction(s):_ `vmlsl.u16 Q0, D0, D0[0]'
27570 * int64x2_t vmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
27571 _Form of expected instruction(s):_ `vmlsl.s32 Q0, D0, D0[0]'
27573 * int32x4_t vmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
27574 _Form of expected instruction(s):_ `vmlsl.s16 Q0, D0, D0[0]'
27576 * int64x2_t vqdmlsl_n_s32 (int64x2_t, int32x2_t, int32_t)
27577 _Form of expected instruction(s):_ `vqdmlsl.s32 Q0, D0, D0[0]'
27579 * int32x4_t vqdmlsl_n_s16 (int32x4_t, int16x4_t, int16_t)
27580 _Form of expected instruction(s):_ `vqdmlsl.s16 Q0, D0, D0[0]'
27582 5.50.3.62 Vector extract
27583 ........................
27585 * uint32x2_t vext_u32 (uint32x2_t, uint32x2_t, const int)
27586 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27588 * uint16x4_t vext_u16 (uint16x4_t, uint16x4_t, const int)
27589 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27591 * uint8x8_t vext_u8 (uint8x8_t, uint8x8_t, const int)
27592 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27594 * int32x2_t vext_s32 (int32x2_t, int32x2_t, const int)
27595 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27597 * int16x4_t vext_s16 (int16x4_t, int16x4_t, const int)
27598 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27600 * int8x8_t vext_s8 (int8x8_t, int8x8_t, const int)
27601 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27603 * uint64x1_t vext_u64 (uint64x1_t, uint64x1_t, const int)
27604 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
27606 * int64x1_t vext_s64 (int64x1_t, int64x1_t, const int)
27607 _Form of expected instruction(s):_ `vext.64 D0, D0, D0, #0'
27609 * float32x2_t vext_f32 (float32x2_t, float32x2_t, const int)
27610 _Form of expected instruction(s):_ `vext.32 D0, D0, D0, #0'
27612 * poly16x4_t vext_p16 (poly16x4_t, poly16x4_t, const int)
27613 _Form of expected instruction(s):_ `vext.16 D0, D0, D0, #0'
27615 * poly8x8_t vext_p8 (poly8x8_t, poly8x8_t, const int)
27616 _Form of expected instruction(s):_ `vext.8 D0, D0, D0, #0'
27618 * uint32x4_t vextq_u32 (uint32x4_t, uint32x4_t, const int)
27619 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27621 * uint16x8_t vextq_u16 (uint16x8_t, uint16x8_t, const int)
27622 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27624 * uint8x16_t vextq_u8 (uint8x16_t, uint8x16_t, const int)
27625 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27627 * int32x4_t vextq_s32 (int32x4_t, int32x4_t, const int)
27628 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27630 * int16x8_t vextq_s16 (int16x8_t, int16x8_t, const int)
27631 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27633 * int8x16_t vextq_s8 (int8x16_t, int8x16_t, const int)
27634 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27636 * uint64x2_t vextq_u64 (uint64x2_t, uint64x2_t, const int)
27637 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
27639 * int64x2_t vextq_s64 (int64x2_t, int64x2_t, const int)
27640 _Form of expected instruction(s):_ `vext.64 Q0, Q0, Q0, #0'
27642 * float32x4_t vextq_f32 (float32x4_t, float32x4_t, const int)
27643 _Form of expected instruction(s):_ `vext.32 Q0, Q0, Q0, #0'
27645 * poly16x8_t vextq_p16 (poly16x8_t, poly16x8_t, const int)
27646 _Form of expected instruction(s):_ `vext.16 Q0, Q0, Q0, #0'
27648 * poly8x16_t vextq_p8 (poly8x16_t, poly8x16_t, const int)
27649 _Form of expected instruction(s):_ `vext.8 Q0, Q0, Q0, #0'
27651 5.50.3.63 Reverse elements
27652 ..........................
27654 * uint32x2_t vrev64_u32 (uint32x2_t)
27655 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27657 * uint16x4_t vrev64_u16 (uint16x4_t)
27658 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27660 * uint8x8_t vrev64_u8 (uint8x8_t)
27661 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27663 * int32x2_t vrev64_s32 (int32x2_t)
27664 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27666 * int16x4_t vrev64_s16 (int16x4_t)
27667 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27669 * int8x8_t vrev64_s8 (int8x8_t)
27670 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27672 * float32x2_t vrev64_f32 (float32x2_t)
27673 _Form of expected instruction(s):_ `vrev64.32 D0, D0'
27675 * poly16x4_t vrev64_p16 (poly16x4_t)
27676 _Form of expected instruction(s):_ `vrev64.16 D0, D0'
27678 * poly8x8_t vrev64_p8 (poly8x8_t)
27679 _Form of expected instruction(s):_ `vrev64.8 D0, D0'
27681 * uint32x4_t vrev64q_u32 (uint32x4_t)
27682 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27684 * uint16x8_t vrev64q_u16 (uint16x8_t)
27685 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27687 * uint8x16_t vrev64q_u8 (uint8x16_t)
27688 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27690 * int32x4_t vrev64q_s32 (int32x4_t)
27691 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27693 * int16x8_t vrev64q_s16 (int16x8_t)
27694 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27696 * int8x16_t vrev64q_s8 (int8x16_t)
27697 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27699 * float32x4_t vrev64q_f32 (float32x4_t)
27700 _Form of expected instruction(s):_ `vrev64.32 Q0, Q0'
27702 * poly16x8_t vrev64q_p16 (poly16x8_t)
27703 _Form of expected instruction(s):_ `vrev64.16 Q0, Q0'
27705 * poly8x16_t vrev64q_p8 (poly8x16_t)
27706 _Form of expected instruction(s):_ `vrev64.8 Q0, Q0'
27708 * uint16x4_t vrev32_u16 (uint16x4_t)
27709 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27711 * int16x4_t vrev32_s16 (int16x4_t)
27712 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27714 * uint8x8_t vrev32_u8 (uint8x8_t)
27715 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27717 * int8x8_t vrev32_s8 (int8x8_t)
27718 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27720 * poly16x4_t vrev32_p16 (poly16x4_t)
27721 _Form of expected instruction(s):_ `vrev32.16 D0, D0'
27723 * poly8x8_t vrev32_p8 (poly8x8_t)
27724 _Form of expected instruction(s):_ `vrev32.8 D0, D0'
27726 * uint16x8_t vrev32q_u16 (uint16x8_t)
27727 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27729 * int16x8_t vrev32q_s16 (int16x8_t)
27730 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27732 * uint8x16_t vrev32q_u8 (uint8x16_t)
27733 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27735 * int8x16_t vrev32q_s8 (int8x16_t)
27736 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27738 * poly16x8_t vrev32q_p16 (poly16x8_t)
27739 _Form of expected instruction(s):_ `vrev32.16 Q0, Q0'
27741 * poly8x16_t vrev32q_p8 (poly8x16_t)
27742 _Form of expected instruction(s):_ `vrev32.8 Q0, Q0'
27744 * uint8x8_t vrev16_u8 (uint8x8_t)
27745 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27747 * int8x8_t vrev16_s8 (int8x8_t)
27748 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27750 * poly8x8_t vrev16_p8 (poly8x8_t)
27751 _Form of expected instruction(s):_ `vrev16.8 D0, D0'
27753 * uint8x16_t vrev16q_u8 (uint8x16_t)
27754 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27756 * int8x16_t vrev16q_s8 (int8x16_t)
27757 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27759 * poly8x16_t vrev16q_p8 (poly8x16_t)
27760 _Form of expected instruction(s):_ `vrev16.8 Q0, Q0'
27762 5.50.3.64 Bit selection
27763 .......................
27765 * uint32x2_t vbsl_u32 (uint32x2_t, uint32x2_t, uint32x2_t)
27766 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27767 D0, D0, D0' _or_ `vbif D0, D0, D0'
27769 * uint16x4_t vbsl_u16 (uint16x4_t, uint16x4_t, uint16x4_t)
27770 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27771 D0, D0, D0' _or_ `vbif D0, D0, D0'
27773 * uint8x8_t vbsl_u8 (uint8x8_t, uint8x8_t, uint8x8_t)
27774 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27775 D0, D0, D0' _or_ `vbif D0, D0, D0'
27777 * int32x2_t vbsl_s32 (uint32x2_t, int32x2_t, int32x2_t)
27778 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27779 D0, D0, D0' _or_ `vbif D0, D0, D0'
27781 * int16x4_t vbsl_s16 (uint16x4_t, int16x4_t, int16x4_t)
27782 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27783 D0, D0, D0' _or_ `vbif D0, D0, D0'
27785 * int8x8_t vbsl_s8 (uint8x8_t, int8x8_t, int8x8_t)
27786 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27787 D0, D0, D0' _or_ `vbif D0, D0, D0'
27789 * uint64x1_t vbsl_u64 (uint64x1_t, uint64x1_t, uint64x1_t)
27790 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27791 D0, D0, D0' _or_ `vbif D0, D0, D0'
27793 * int64x1_t vbsl_s64 (uint64x1_t, int64x1_t, int64x1_t)
27794 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27795 D0, D0, D0' _or_ `vbif D0, D0, D0'
27797 * float32x2_t vbsl_f32 (uint32x2_t, float32x2_t, float32x2_t)
27798 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27799 D0, D0, D0' _or_ `vbif D0, D0, D0'
27801 * poly16x4_t vbsl_p16 (uint16x4_t, poly16x4_t, poly16x4_t)
27802 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27803 D0, D0, D0' _or_ `vbif D0, D0, D0'
27805 * poly8x8_t vbsl_p8 (uint8x8_t, poly8x8_t, poly8x8_t)
27806 _Form of expected instruction(s):_ `vbsl D0, D0, D0' _or_ `vbit
27807 D0, D0, D0' _or_ `vbif D0, D0, D0'
27809 * uint32x4_t vbslq_u32 (uint32x4_t, uint32x4_t, uint32x4_t)
27810 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27811 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27813 * uint16x8_t vbslq_u16 (uint16x8_t, uint16x8_t, uint16x8_t)
27814 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27815 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27817 * uint8x16_t vbslq_u8 (uint8x16_t, uint8x16_t, uint8x16_t)
27818 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27819 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27821 * int32x4_t vbslq_s32 (uint32x4_t, int32x4_t, int32x4_t)
27822 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27823 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27825 * int16x8_t vbslq_s16 (uint16x8_t, int16x8_t, int16x8_t)
27826 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27827 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27829 * int8x16_t vbslq_s8 (uint8x16_t, int8x16_t, int8x16_t)
27830 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27831 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27833 * uint64x2_t vbslq_u64 (uint64x2_t, uint64x2_t, uint64x2_t)
27834 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27835 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27837 * int64x2_t vbslq_s64 (uint64x2_t, int64x2_t, int64x2_t)
27838 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27839 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27841 * float32x4_t vbslq_f32 (uint32x4_t, float32x4_t, float32x4_t)
27842 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27843 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27845 * poly16x8_t vbslq_p16 (uint16x8_t, poly16x8_t, poly16x8_t)
27846 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27847 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27849 * poly8x16_t vbslq_p8 (uint8x16_t, poly8x16_t, poly8x16_t)
27850 _Form of expected instruction(s):_ `vbsl Q0, Q0, Q0' _or_ `vbit
27851 Q0, Q0, Q0' _or_ `vbif Q0, Q0, Q0'
27853 5.50.3.65 Transpose elements
27854 ............................
27856 * uint32x2x2_t vtrn_u32 (uint32x2_t, uint32x2_t)
27857 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27859 * uint16x4x2_t vtrn_u16 (uint16x4_t, uint16x4_t)
27860 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27862 * uint8x8x2_t vtrn_u8 (uint8x8_t, uint8x8_t)
27863 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27865 * int32x2x2_t vtrn_s32 (int32x2_t, int32x2_t)
27866 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27868 * int16x4x2_t vtrn_s16 (int16x4_t, int16x4_t)
27869 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27871 * int8x8x2_t vtrn_s8 (int8x8_t, int8x8_t)
27872 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27874 * float32x2x2_t vtrn_f32 (float32x2_t, float32x2_t)
27875 _Form of expected instruction(s):_ `vtrn.32 D0, D1'
27877 * poly16x4x2_t vtrn_p16 (poly16x4_t, poly16x4_t)
27878 _Form of expected instruction(s):_ `vtrn.16 D0, D1'
27880 * poly8x8x2_t vtrn_p8 (poly8x8_t, poly8x8_t)
27881 _Form of expected instruction(s):_ `vtrn.8 D0, D1'
27883 * uint32x4x2_t vtrnq_u32 (uint32x4_t, uint32x4_t)
27884 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27886 * uint16x8x2_t vtrnq_u16 (uint16x8_t, uint16x8_t)
27887 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27889 * uint8x16x2_t vtrnq_u8 (uint8x16_t, uint8x16_t)
27890 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27892 * int32x4x2_t vtrnq_s32 (int32x4_t, int32x4_t)
27893 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27895 * int16x8x2_t vtrnq_s16 (int16x8_t, int16x8_t)
27896 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27898 * int8x16x2_t vtrnq_s8 (int8x16_t, int8x16_t)
27899 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27901 * float32x4x2_t vtrnq_f32 (float32x4_t, float32x4_t)
27902 _Form of expected instruction(s):_ `vtrn.32 Q0, Q1'
27904 * poly16x8x2_t vtrnq_p16 (poly16x8_t, poly16x8_t)
27905 _Form of expected instruction(s):_ `vtrn.16 Q0, Q1'
27907 * poly8x16x2_t vtrnq_p8 (poly8x16_t, poly8x16_t)
27908 _Form of expected instruction(s):_ `vtrn.8 Q0, Q1'
27910 5.50.3.66 Zip elements
27911 ......................
27913 * uint32x2x2_t vzip_u32 (uint32x2_t, uint32x2_t)
27914 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27916 * uint16x4x2_t vzip_u16 (uint16x4_t, uint16x4_t)
27917 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27919 * uint8x8x2_t vzip_u8 (uint8x8_t, uint8x8_t)
27920 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27922 * int32x2x2_t vzip_s32 (int32x2_t, int32x2_t)
27923 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27925 * int16x4x2_t vzip_s16 (int16x4_t, int16x4_t)
27926 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27928 * int8x8x2_t vzip_s8 (int8x8_t, int8x8_t)
27929 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27931 * float32x2x2_t vzip_f32 (float32x2_t, float32x2_t)
27932 _Form of expected instruction(s):_ `vzip.32 D0, D1'
27934 * poly16x4x2_t vzip_p16 (poly16x4_t, poly16x4_t)
27935 _Form of expected instruction(s):_ `vzip.16 D0, D1'
27937 * poly8x8x2_t vzip_p8 (poly8x8_t, poly8x8_t)
27938 _Form of expected instruction(s):_ `vzip.8 D0, D1'
27940 * uint32x4x2_t vzipq_u32 (uint32x4_t, uint32x4_t)
27941 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
27943 * uint16x8x2_t vzipq_u16 (uint16x8_t, uint16x8_t)
27944 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
27946 * uint8x16x2_t vzipq_u8 (uint8x16_t, uint8x16_t)
27947 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
27949 * int32x4x2_t vzipq_s32 (int32x4_t, int32x4_t)
27950 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
27952 * int16x8x2_t vzipq_s16 (int16x8_t, int16x8_t)
27953 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
27955 * int8x16x2_t vzipq_s8 (int8x16_t, int8x16_t)
27956 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
27958 * float32x4x2_t vzipq_f32 (float32x4_t, float32x4_t)
27959 _Form of expected instruction(s):_ `vzip.32 Q0, Q1'
27961 * poly16x8x2_t vzipq_p16 (poly16x8_t, poly16x8_t)
27962 _Form of expected instruction(s):_ `vzip.16 Q0, Q1'
27964 * poly8x16x2_t vzipq_p8 (poly8x16_t, poly8x16_t)
27965 _Form of expected instruction(s):_ `vzip.8 Q0, Q1'
27967 5.50.3.67 Unzip elements
27968 ........................
27970 * uint32x2x2_t vuzp_u32 (uint32x2_t, uint32x2_t)
27971 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
27973 * uint16x4x2_t vuzp_u16 (uint16x4_t, uint16x4_t)
27974 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
27976 * uint8x8x2_t vuzp_u8 (uint8x8_t, uint8x8_t)
27977 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
27979 * int32x2x2_t vuzp_s32 (int32x2_t, int32x2_t)
27980 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
27982 * int16x4x2_t vuzp_s16 (int16x4_t, int16x4_t)
27983 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
27985 * int8x8x2_t vuzp_s8 (int8x8_t, int8x8_t)
27986 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
27988 * float32x2x2_t vuzp_f32 (float32x2_t, float32x2_t)
27989 _Form of expected instruction(s):_ `vuzp.32 D0, D1'
27991 * poly16x4x2_t vuzp_p16 (poly16x4_t, poly16x4_t)
27992 _Form of expected instruction(s):_ `vuzp.16 D0, D1'
27994 * poly8x8x2_t vuzp_p8 (poly8x8_t, poly8x8_t)
27995 _Form of expected instruction(s):_ `vuzp.8 D0, D1'
27997 * uint32x4x2_t vuzpq_u32 (uint32x4_t, uint32x4_t)
27998 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28000 * uint16x8x2_t vuzpq_u16 (uint16x8_t, uint16x8_t)
28001 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28003 * uint8x16x2_t vuzpq_u8 (uint8x16_t, uint8x16_t)
28004 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28006 * int32x4x2_t vuzpq_s32 (int32x4_t, int32x4_t)
28007 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28009 * int16x8x2_t vuzpq_s16 (int16x8_t, int16x8_t)
28010 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28012 * int8x16x2_t vuzpq_s8 (int8x16_t, int8x16_t)
28013 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28015 * float32x4x2_t vuzpq_f32 (float32x4_t, float32x4_t)
28016 _Form of expected instruction(s):_ `vuzp.32 Q0, Q1'
28018 * poly16x8x2_t vuzpq_p16 (poly16x8_t, poly16x8_t)
28019 _Form of expected instruction(s):_ `vuzp.16 Q0, Q1'
28021 * poly8x16x2_t vuzpq_p8 (poly8x16_t, poly8x16_t)
28022 _Form of expected instruction(s):_ `vuzp.8 Q0, Q1'
28024 5.50.3.68 Element/structure loads, VLD1 variants
28025 ................................................
28027 * uint32x2_t vld1_u32 (const uint32_t *)
28028 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28030 * uint16x4_t vld1_u16 (const uint16_t *)
28031 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28033 * uint8x8_t vld1_u8 (const uint8_t *)
28034 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28036 * int32x2_t vld1_s32 (const int32_t *)
28037 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28039 * int16x4_t vld1_s16 (const int16_t *)
28040 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28042 * int8x8_t vld1_s8 (const int8_t *)
28043 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28045 * uint64x1_t vld1_u64 (const uint64_t *)
28046 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28048 * int64x1_t vld1_s64 (const int64_t *)
28049 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28051 * float32x2_t vld1_f32 (const float32_t *)
28052 _Form of expected instruction(s):_ `vld1.32 {D0}, [R0]'
28054 * poly16x4_t vld1_p16 (const poly16_t *)
28055 _Form of expected instruction(s):_ `vld1.16 {D0}, [R0]'
28057 * poly8x8_t vld1_p8 (const poly8_t *)
28058 _Form of expected instruction(s):_ `vld1.8 {D0}, [R0]'
28060 * uint32x4_t vld1q_u32 (const uint32_t *)
28061 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28063 * uint16x8_t vld1q_u16 (const uint16_t *)
28064 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28066 * uint8x16_t vld1q_u8 (const uint8_t *)
28067 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28069 * int32x4_t vld1q_s32 (const int32_t *)
28070 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28072 * int16x8_t vld1q_s16 (const int16_t *)
28073 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28075 * int8x16_t vld1q_s8 (const int8_t *)
28076 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28078 * uint64x2_t vld1q_u64 (const uint64_t *)
28079 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28081 * int64x2_t vld1q_s64 (const int64_t *)
28082 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28084 * float32x4_t vld1q_f32 (const float32_t *)
28085 _Form of expected instruction(s):_ `vld1.32 {D0, D1}, [R0]'
28087 * poly16x8_t vld1q_p16 (const poly16_t *)
28088 _Form of expected instruction(s):_ `vld1.16 {D0, D1}, [R0]'
28090 * poly8x16_t vld1q_p8 (const poly8_t *)
28091 _Form of expected instruction(s):_ `vld1.8 {D0, D1}, [R0]'
28093 * uint32x2_t vld1_lane_u32 (const uint32_t *, uint32x2_t, const int)
28094 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28096 * uint16x4_t vld1_lane_u16 (const uint16_t *, uint16x4_t, const int)
28097 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28099 * uint8x8_t vld1_lane_u8 (const uint8_t *, uint8x8_t, const int)
28100 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28102 * int32x2_t vld1_lane_s32 (const int32_t *, int32x2_t, const int)
28103 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28105 * int16x4_t vld1_lane_s16 (const int16_t *, int16x4_t, const int)
28106 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28108 * int8x8_t vld1_lane_s8 (const int8_t *, int8x8_t, const int)
28109 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28111 * float32x2_t vld1_lane_f32 (const float32_t *, float32x2_t, const
28113 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28115 * poly16x4_t vld1_lane_p16 (const poly16_t *, poly16x4_t, const int)
28116 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28118 * poly8x8_t vld1_lane_p8 (const poly8_t *, poly8x8_t, const int)
28119 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28121 * uint64x1_t vld1_lane_u64 (const uint64_t *, uint64x1_t, const int)
28122 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28124 * int64x1_t vld1_lane_s64 (const int64_t *, int64x1_t, const int)
28125 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28127 * uint32x4_t vld1q_lane_u32 (const uint32_t *, uint32x4_t, const int)
28128 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28130 * uint16x8_t vld1q_lane_u16 (const uint16_t *, uint16x8_t, const int)
28131 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28133 * uint8x16_t vld1q_lane_u8 (const uint8_t *, uint8x16_t, const int)
28134 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28136 * int32x4_t vld1q_lane_s32 (const int32_t *, int32x4_t, const int)
28137 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28139 * int16x8_t vld1q_lane_s16 (const int16_t *, int16x8_t, const int)
28140 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28142 * int8x16_t vld1q_lane_s8 (const int8_t *, int8x16_t, const int)
28143 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28145 * float32x4_t vld1q_lane_f32 (const float32_t *, float32x4_t, const
28147 _Form of expected instruction(s):_ `vld1.32 {D0[0]}, [R0]'
28149 * poly16x8_t vld1q_lane_p16 (const poly16_t *, poly16x8_t, const int)
28150 _Form of expected instruction(s):_ `vld1.16 {D0[0]}, [R0]'
28152 * poly8x16_t vld1q_lane_p8 (const poly8_t *, poly8x16_t, const int)
28153 _Form of expected instruction(s):_ `vld1.8 {D0[0]}, [R0]'
28155 * uint64x2_t vld1q_lane_u64 (const uint64_t *, uint64x2_t, const int)
28156 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28158 * int64x2_t vld1q_lane_s64 (const int64_t *, int64x2_t, const int)
28159 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28161 * uint32x2_t vld1_dup_u32 (const uint32_t *)
28162 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28164 * uint16x4_t vld1_dup_u16 (const uint16_t *)
28165 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28167 * uint8x8_t vld1_dup_u8 (const uint8_t *)
28168 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28170 * int32x2_t vld1_dup_s32 (const int32_t *)
28171 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28173 * int16x4_t vld1_dup_s16 (const int16_t *)
28174 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28176 * int8x8_t vld1_dup_s8 (const int8_t *)
28177 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28179 * float32x2_t vld1_dup_f32 (const float32_t *)
28180 _Form of expected instruction(s):_ `vld1.32 {D0[]}, [R0]'
28182 * poly16x4_t vld1_dup_p16 (const poly16_t *)
28183 _Form of expected instruction(s):_ `vld1.16 {D0[]}, [R0]'
28185 * poly8x8_t vld1_dup_p8 (const poly8_t *)
28186 _Form of expected instruction(s):_ `vld1.8 {D0[]}, [R0]'
28188 * uint64x1_t vld1_dup_u64 (const uint64_t *)
28189 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28191 * int64x1_t vld1_dup_s64 (const int64_t *)
28192 _Form of expected instruction(s):_ `vld1.64 {D0}, [R0]'
28194 * uint32x4_t vld1q_dup_u32 (const uint32_t *)
28195 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28197 * uint16x8_t vld1q_dup_u16 (const uint16_t *)
28198 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28200 * uint8x16_t vld1q_dup_u8 (const uint8_t *)
28201 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28203 * int32x4_t vld1q_dup_s32 (const int32_t *)
28204 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28206 * int16x8_t vld1q_dup_s16 (const int16_t *)
28207 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28209 * int8x16_t vld1q_dup_s8 (const int8_t *)
28210 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28212 * float32x4_t vld1q_dup_f32 (const float32_t *)
28213 _Form of expected instruction(s):_ `vld1.32 {D0[], D1[]}, [R0]'
28215 * poly16x8_t vld1q_dup_p16 (const poly16_t *)
28216 _Form of expected instruction(s):_ `vld1.16 {D0[], D1[]}, [R0]'
28218 * poly8x16_t vld1q_dup_p8 (const poly8_t *)
28219 _Form of expected instruction(s):_ `vld1.8 {D0[], D1[]}, [R0]'
28221 * uint64x2_t vld1q_dup_u64 (const uint64_t *)
28222 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28224 * int64x2_t vld1q_dup_s64 (const int64_t *)
28225 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28227 5.50.3.69 Element/structure stores, VST1 variants
28228 .................................................
28230 * void vst1_u32 (uint32_t *, uint32x2_t)
28231 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28233 * void vst1_u16 (uint16_t *, uint16x4_t)
28234 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28236 * void vst1_u8 (uint8_t *, uint8x8_t)
28237 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28239 * void vst1_s32 (int32_t *, int32x2_t)
28240 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28242 * void vst1_s16 (int16_t *, int16x4_t)
28243 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28245 * void vst1_s8 (int8_t *, int8x8_t)
28246 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28248 * void vst1_u64 (uint64_t *, uint64x1_t)
28249 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28251 * void vst1_s64 (int64_t *, int64x1_t)
28252 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28254 * void vst1_f32 (float32_t *, float32x2_t)
28255 _Form of expected instruction(s):_ `vst1.32 {D0}, [R0]'
28257 * void vst1_p16 (poly16_t *, poly16x4_t)
28258 _Form of expected instruction(s):_ `vst1.16 {D0}, [R0]'
28260 * void vst1_p8 (poly8_t *, poly8x8_t)
28261 _Form of expected instruction(s):_ `vst1.8 {D0}, [R0]'
28263 * void vst1q_u32 (uint32_t *, uint32x4_t)
28264 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28266 * void vst1q_u16 (uint16_t *, uint16x8_t)
28267 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28269 * void vst1q_u8 (uint8_t *, uint8x16_t)
28270 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28272 * void vst1q_s32 (int32_t *, int32x4_t)
28273 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28275 * void vst1q_s16 (int16_t *, int16x8_t)
28276 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28278 * void vst1q_s8 (int8_t *, int8x16_t)
28279 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28281 * void vst1q_u64 (uint64_t *, uint64x2_t)
28282 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28284 * void vst1q_s64 (int64_t *, int64x2_t)
28285 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28287 * void vst1q_f32 (float32_t *, float32x4_t)
28288 _Form of expected instruction(s):_ `vst1.32 {D0, D1}, [R0]'
28290 * void vst1q_p16 (poly16_t *, poly16x8_t)
28291 _Form of expected instruction(s):_ `vst1.16 {D0, D1}, [R0]'
28293 * void vst1q_p8 (poly8_t *, poly8x16_t)
28294 _Form of expected instruction(s):_ `vst1.8 {D0, D1}, [R0]'
28296 * void vst1_lane_u32 (uint32_t *, uint32x2_t, const int)
28297 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28299 * void vst1_lane_u16 (uint16_t *, uint16x4_t, const int)
28300 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28302 * void vst1_lane_u8 (uint8_t *, uint8x8_t, const int)
28303 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28305 * void vst1_lane_s32 (int32_t *, int32x2_t, const int)
28306 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28308 * void vst1_lane_s16 (int16_t *, int16x4_t, const int)
28309 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28311 * void vst1_lane_s8 (int8_t *, int8x8_t, const int)
28312 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28314 * void vst1_lane_f32 (float32_t *, float32x2_t, const int)
28315 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28317 * void vst1_lane_p16 (poly16_t *, poly16x4_t, const int)
28318 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28320 * void vst1_lane_p8 (poly8_t *, poly8x8_t, const int)
28321 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28323 * void vst1_lane_s64 (int64_t *, int64x1_t, const int)
28324 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28326 * void vst1_lane_u64 (uint64_t *, uint64x1_t, const int)
28327 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28329 * void vst1q_lane_u32 (uint32_t *, uint32x4_t, const int)
28330 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28332 * void vst1q_lane_u16 (uint16_t *, uint16x8_t, const int)
28333 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28335 * void vst1q_lane_u8 (uint8_t *, uint8x16_t, const int)
28336 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28338 * void vst1q_lane_s32 (int32_t *, int32x4_t, const int)
28339 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28341 * void vst1q_lane_s16 (int16_t *, int16x8_t, const int)
28342 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28344 * void vst1q_lane_s8 (int8_t *, int8x16_t, const int)
28345 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28347 * void vst1q_lane_f32 (float32_t *, float32x4_t, const int)
28348 _Form of expected instruction(s):_ `vst1.32 {D0[0]}, [R0]'
28350 * void vst1q_lane_p16 (poly16_t *, poly16x8_t, const int)
28351 _Form of expected instruction(s):_ `vst1.16 {D0[0]}, [R0]'
28353 * void vst1q_lane_p8 (poly8_t *, poly8x16_t, const int)
28354 _Form of expected instruction(s):_ `vst1.8 {D0[0]}, [R0]'
28356 * void vst1q_lane_s64 (int64_t *, int64x2_t, const int)
28357 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28359 * void vst1q_lane_u64 (uint64_t *, uint64x2_t, const int)
28360 _Form of expected instruction(s):_ `vst1.64 {D0}, [R0]'
28362 5.50.3.70 Element/structure loads, VLD2 variants
28363 ................................................
28365 * uint32x2x2_t vld2_u32 (const uint32_t *)
28366 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28368 * uint16x4x2_t vld2_u16 (const uint16_t *)
28369 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28371 * uint8x8x2_t vld2_u8 (const uint8_t *)
28372 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28374 * int32x2x2_t vld2_s32 (const int32_t *)
28375 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28377 * int16x4x2_t vld2_s16 (const int16_t *)
28378 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28380 * int8x8x2_t vld2_s8 (const int8_t *)
28381 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28383 * float32x2x2_t vld2_f32 (const float32_t *)
28384 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28386 * poly16x4x2_t vld2_p16 (const poly16_t *)
28387 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28389 * poly8x8x2_t vld2_p8 (const poly8_t *)
28390 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28392 * uint64x1x2_t vld2_u64 (const uint64_t *)
28393 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28395 * int64x1x2_t vld2_s64 (const int64_t *)
28396 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28398 * uint32x4x2_t vld2q_u32 (const uint32_t *)
28399 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28401 * uint16x8x2_t vld2q_u16 (const uint16_t *)
28402 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28404 * uint8x16x2_t vld2q_u8 (const uint8_t *)
28405 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28407 * int32x4x2_t vld2q_s32 (const int32_t *)
28408 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28410 * int16x8x2_t vld2q_s16 (const int16_t *)
28411 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28413 * int8x16x2_t vld2q_s8 (const int8_t *)
28414 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28416 * float32x4x2_t vld2q_f32 (const float32_t *)
28417 _Form of expected instruction(s):_ `vld2.32 {D0, D1}, [R0]'
28419 * poly16x8x2_t vld2q_p16 (const poly16_t *)
28420 _Form of expected instruction(s):_ `vld2.16 {D0, D1}, [R0]'
28422 * poly8x16x2_t vld2q_p8 (const poly8_t *)
28423 _Form of expected instruction(s):_ `vld2.8 {D0, D1}, [R0]'
28425 * uint32x2x2_t vld2_lane_u32 (const uint32_t *, uint32x2x2_t, const
28427 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28429 * uint16x4x2_t vld2_lane_u16 (const uint16_t *, uint16x4x2_t, const
28431 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28433 * uint8x8x2_t vld2_lane_u8 (const uint8_t *, uint8x8x2_t, const int)
28434 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28436 * int32x2x2_t vld2_lane_s32 (const int32_t *, int32x2x2_t, const int)
28437 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28439 * int16x4x2_t vld2_lane_s16 (const int16_t *, int16x4x2_t, const int)
28440 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28442 * int8x8x2_t vld2_lane_s8 (const int8_t *, int8x8x2_t, const int)
28443 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28445 * float32x2x2_t vld2_lane_f32 (const float32_t *, float32x2x2_t,
28447 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28449 * poly16x4x2_t vld2_lane_p16 (const poly16_t *, poly16x4x2_t, const
28451 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28453 * poly8x8x2_t vld2_lane_p8 (const poly8_t *, poly8x8x2_t, const int)
28454 _Form of expected instruction(s):_ `vld2.8 {D0[0], D1[0]}, [R0]'
28456 * int32x4x2_t vld2q_lane_s32 (const int32_t *, int32x4x2_t, const
28458 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28460 * int16x8x2_t vld2q_lane_s16 (const int16_t *, int16x8x2_t, const
28462 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28464 * uint32x4x2_t vld2q_lane_u32 (const uint32_t *, uint32x4x2_t, const
28466 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28468 * uint16x8x2_t vld2q_lane_u16 (const uint16_t *, uint16x8x2_t, const
28470 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28472 * float32x4x2_t vld2q_lane_f32 (const float32_t *, float32x4x2_t,
28474 _Form of expected instruction(s):_ `vld2.32 {D0[0], D1[0]}, [R0]'
28476 * poly16x8x2_t vld2q_lane_p16 (const poly16_t *, poly16x8x2_t, const
28478 _Form of expected instruction(s):_ `vld2.16 {D0[0], D1[0]}, [R0]'
28480 * uint32x2x2_t vld2_dup_u32 (const uint32_t *)
28481 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28483 * uint16x4x2_t vld2_dup_u16 (const uint16_t *)
28484 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28486 * uint8x8x2_t vld2_dup_u8 (const uint8_t *)
28487 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28489 * int32x2x2_t vld2_dup_s32 (const int32_t *)
28490 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28492 * int16x4x2_t vld2_dup_s16 (const int16_t *)
28493 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28495 * int8x8x2_t vld2_dup_s8 (const int8_t *)
28496 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28498 * float32x2x2_t vld2_dup_f32 (const float32_t *)
28499 _Form of expected instruction(s):_ `vld2.32 {D0[], D1[]}, [R0]'
28501 * poly16x4x2_t vld2_dup_p16 (const poly16_t *)
28502 _Form of expected instruction(s):_ `vld2.16 {D0[], D1[]}, [R0]'
28504 * poly8x8x2_t vld2_dup_p8 (const poly8_t *)
28505 _Form of expected instruction(s):_ `vld2.8 {D0[], D1[]}, [R0]'
28507 * uint64x1x2_t vld2_dup_u64 (const uint64_t *)
28508 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28510 * int64x1x2_t vld2_dup_s64 (const int64_t *)
28511 _Form of expected instruction(s):_ `vld1.64 {D0, D1}, [R0]'
28513 5.50.3.71 Element/structure stores, VST2 variants
28514 .................................................
28516 * void vst2_u32 (uint32_t *, uint32x2x2_t)
28517 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28519 * void vst2_u16 (uint16_t *, uint16x4x2_t)
28520 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28522 * void vst2_u8 (uint8_t *, uint8x8x2_t)
28523 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28525 * void vst2_s32 (int32_t *, int32x2x2_t)
28526 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28528 * void vst2_s16 (int16_t *, int16x4x2_t)
28529 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28531 * void vst2_s8 (int8_t *, int8x8x2_t)
28532 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28534 * void vst2_f32 (float32_t *, float32x2x2_t)
28535 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28537 * void vst2_p16 (poly16_t *, poly16x4x2_t)
28538 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28540 * void vst2_p8 (poly8_t *, poly8x8x2_t)
28541 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28543 * void vst2_u64 (uint64_t *, uint64x1x2_t)
28544 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28546 * void vst2_s64 (int64_t *, int64x1x2_t)
28547 _Form of expected instruction(s):_ `vst1.64 {D0, D1}, [R0]'
28549 * void vst2q_u32 (uint32_t *, uint32x4x2_t)
28550 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28552 * void vst2q_u16 (uint16_t *, uint16x8x2_t)
28553 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28555 * void vst2q_u8 (uint8_t *, uint8x16x2_t)
28556 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28558 * void vst2q_s32 (int32_t *, int32x4x2_t)
28559 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28561 * void vst2q_s16 (int16_t *, int16x8x2_t)
28562 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28564 * void vst2q_s8 (int8_t *, int8x16x2_t)
28565 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28567 * void vst2q_f32 (float32_t *, float32x4x2_t)
28568 _Form of expected instruction(s):_ `vst2.32 {D0, D1}, [R0]'
28570 * void vst2q_p16 (poly16_t *, poly16x8x2_t)
28571 _Form of expected instruction(s):_ `vst2.16 {D0, D1}, [R0]'
28573 * void vst2q_p8 (poly8_t *, poly8x16x2_t)
28574 _Form of expected instruction(s):_ `vst2.8 {D0, D1}, [R0]'
28576 * void vst2_lane_u32 (uint32_t *, uint32x2x2_t, const int)
28577 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28579 * void vst2_lane_u16 (uint16_t *, uint16x4x2_t, const int)
28580 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28582 * void vst2_lane_u8 (uint8_t *, uint8x8x2_t, const int)
28583 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28585 * void vst2_lane_s32 (int32_t *, int32x2x2_t, const int)
28586 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28588 * void vst2_lane_s16 (int16_t *, int16x4x2_t, const int)
28589 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28591 * void vst2_lane_s8 (int8_t *, int8x8x2_t, const int)
28592 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28594 * void vst2_lane_f32 (float32_t *, float32x2x2_t, const int)
28595 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28597 * void vst2_lane_p16 (poly16_t *, poly16x4x2_t, const int)
28598 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28600 * void vst2_lane_p8 (poly8_t *, poly8x8x2_t, const int)
28601 _Form of expected instruction(s):_ `vst2.8 {D0[0], D1[0]}, [R0]'
28603 * void vst2q_lane_s32 (int32_t *, int32x4x2_t, const int)
28604 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28606 * void vst2q_lane_s16 (int16_t *, int16x8x2_t, const int)
28607 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28609 * void vst2q_lane_u32 (uint32_t *, uint32x4x2_t, const int)
28610 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28612 * void vst2q_lane_u16 (uint16_t *, uint16x8x2_t, const int)
28613 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28615 * void vst2q_lane_f32 (float32_t *, float32x4x2_t, const int)
28616 _Form of expected instruction(s):_ `vst2.32 {D0[0], D1[0]}, [R0]'
28618 * void vst2q_lane_p16 (poly16_t *, poly16x8x2_t, const int)
28619 _Form of expected instruction(s):_ `vst2.16 {D0[0], D1[0]}, [R0]'
28621 5.50.3.72 Element/structure loads, VLD3 variants
28622 ................................................
28624 * uint32x2x3_t vld3_u32 (const uint32_t *)
28625 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28627 * uint16x4x3_t vld3_u16 (const uint16_t *)
28628 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28630 * uint8x8x3_t vld3_u8 (const uint8_t *)
28631 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28633 * int32x2x3_t vld3_s32 (const int32_t *)
28634 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28636 * int16x4x3_t vld3_s16 (const int16_t *)
28637 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28639 * int8x8x3_t vld3_s8 (const int8_t *)
28640 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28642 * float32x2x3_t vld3_f32 (const float32_t *)
28643 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28645 * poly16x4x3_t vld3_p16 (const poly16_t *)
28646 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28648 * poly8x8x3_t vld3_p8 (const poly8_t *)
28649 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28651 * uint64x1x3_t vld3_u64 (const uint64_t *)
28652 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28654 * int64x1x3_t vld3_s64 (const int64_t *)
28655 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28657 * uint32x4x3_t vld3q_u32 (const uint32_t *)
28658 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28660 * uint16x8x3_t vld3q_u16 (const uint16_t *)
28661 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28663 * uint8x16x3_t vld3q_u8 (const uint8_t *)
28664 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28666 * int32x4x3_t vld3q_s32 (const int32_t *)
28667 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28669 * int16x8x3_t vld3q_s16 (const int16_t *)
28670 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28672 * int8x16x3_t vld3q_s8 (const int8_t *)
28673 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28675 * float32x4x3_t vld3q_f32 (const float32_t *)
28676 _Form of expected instruction(s):_ `vld3.32 {D0, D1, D2}, [R0]'
28678 * poly16x8x3_t vld3q_p16 (const poly16_t *)
28679 _Form of expected instruction(s):_ `vld3.16 {D0, D1, D2}, [R0]'
28681 * poly8x16x3_t vld3q_p8 (const poly8_t *)
28682 _Form of expected instruction(s):_ `vld3.8 {D0, D1, D2}, [R0]'
28684 * uint32x2x3_t vld3_lane_u32 (const uint32_t *, uint32x2x3_t, const
28686 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28689 * uint16x4x3_t vld3_lane_u16 (const uint16_t *, uint16x4x3_t, const
28691 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28694 * uint8x8x3_t vld3_lane_u8 (const uint8_t *, uint8x8x3_t, const int)
28695 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28698 * int32x2x3_t vld3_lane_s32 (const int32_t *, int32x2x3_t, const int)
28699 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28702 * int16x4x3_t vld3_lane_s16 (const int16_t *, int16x4x3_t, const int)
28703 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28706 * int8x8x3_t vld3_lane_s8 (const int8_t *, int8x8x3_t, const int)
28707 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28710 * float32x2x3_t vld3_lane_f32 (const float32_t *, float32x2x3_t,
28712 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28715 * poly16x4x3_t vld3_lane_p16 (const poly16_t *, poly16x4x3_t, const
28717 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28720 * poly8x8x3_t vld3_lane_p8 (const poly8_t *, poly8x8x3_t, const int)
28721 _Form of expected instruction(s):_ `vld3.8 {D0[0], D1[0], D2[0]},
28724 * int32x4x3_t vld3q_lane_s32 (const int32_t *, int32x4x3_t, const
28726 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28729 * int16x8x3_t vld3q_lane_s16 (const int16_t *, int16x8x3_t, const
28731 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28734 * uint32x4x3_t vld3q_lane_u32 (const uint32_t *, uint32x4x3_t, const
28736 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28739 * uint16x8x3_t vld3q_lane_u16 (const uint16_t *, uint16x8x3_t, const
28741 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28744 * float32x4x3_t vld3q_lane_f32 (const float32_t *, float32x4x3_t,
28746 _Form of expected instruction(s):_ `vld3.32 {D0[0], D1[0], D2[0]},
28749 * poly16x8x3_t vld3q_lane_p16 (const poly16_t *, poly16x8x3_t, const
28751 _Form of expected instruction(s):_ `vld3.16 {D0[0], D1[0], D2[0]},
28754 * uint32x2x3_t vld3_dup_u32 (const uint32_t *)
28755 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28758 * uint16x4x3_t vld3_dup_u16 (const uint16_t *)
28759 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28762 * uint8x8x3_t vld3_dup_u8 (const uint8_t *)
28763 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28766 * int32x2x3_t vld3_dup_s32 (const int32_t *)
28767 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28770 * int16x4x3_t vld3_dup_s16 (const int16_t *)
28771 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28774 * int8x8x3_t vld3_dup_s8 (const int8_t *)
28775 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28778 * float32x2x3_t vld3_dup_f32 (const float32_t *)
28779 _Form of expected instruction(s):_ `vld3.32 {D0[], D1[], D2[]},
28782 * poly16x4x3_t vld3_dup_p16 (const poly16_t *)
28783 _Form of expected instruction(s):_ `vld3.16 {D0[], D1[], D2[]},
28786 * poly8x8x3_t vld3_dup_p8 (const poly8_t *)
28787 _Form of expected instruction(s):_ `vld3.8 {D0[], D1[], D2[]},
28790 * uint64x1x3_t vld3_dup_u64 (const uint64_t *)
28791 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28793 * int64x1x3_t vld3_dup_s64 (const int64_t *)
28794 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2}, [R0]'
28796 5.50.3.73 Element/structure stores, VST3 variants
28797 .................................................
28799 * void vst3_u32 (uint32_t *, uint32x2x3_t)
28800 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28802 * void vst3_u16 (uint16_t *, uint16x4x3_t)
28803 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28805 * void vst3_u8 (uint8_t *, uint8x8x3_t)
28806 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28808 * void vst3_s32 (int32_t *, int32x2x3_t)
28809 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28811 * void vst3_s16 (int16_t *, int16x4x3_t)
28812 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28814 * void vst3_s8 (int8_t *, int8x8x3_t)
28815 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28817 * void vst3_f32 (float32_t *, float32x2x3_t)
28818 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2, D3}, [R0]'
28820 * void vst3_p16 (poly16_t *, poly16x4x3_t)
28821 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2, D3}, [R0]'
28823 * void vst3_p8 (poly8_t *, poly8x8x3_t)
28824 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2, D3}, [R0]'
28826 * void vst3_u64 (uint64_t *, uint64x1x3_t)
28827 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
28829 * void vst3_s64 (int64_t *, int64x1x3_t)
28830 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
28832 * void vst3q_u32 (uint32_t *, uint32x4x3_t)
28833 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28835 * void vst3q_u16 (uint16_t *, uint16x8x3_t)
28836 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28838 * void vst3q_u8 (uint8_t *, uint8x16x3_t)
28839 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28841 * void vst3q_s32 (int32_t *, int32x4x3_t)
28842 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28844 * void vst3q_s16 (int16_t *, int16x8x3_t)
28845 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28847 * void vst3q_s8 (int8_t *, int8x16x3_t)
28848 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28850 * void vst3q_f32 (float32_t *, float32x4x3_t)
28851 _Form of expected instruction(s):_ `vst3.32 {D0, D1, D2}, [R0]'
28853 * void vst3q_p16 (poly16_t *, poly16x8x3_t)
28854 _Form of expected instruction(s):_ `vst3.16 {D0, D1, D2}, [R0]'
28856 * void vst3q_p8 (poly8_t *, poly8x16x3_t)
28857 _Form of expected instruction(s):_ `vst3.8 {D0, D1, D2}, [R0]'
28859 * void vst3_lane_u32 (uint32_t *, uint32x2x3_t, const int)
28860 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28863 * void vst3_lane_u16 (uint16_t *, uint16x4x3_t, const int)
28864 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28867 * void vst3_lane_u8 (uint8_t *, uint8x8x3_t, const int)
28868 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28871 * void vst3_lane_s32 (int32_t *, int32x2x3_t, const int)
28872 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28875 * void vst3_lane_s16 (int16_t *, int16x4x3_t, const int)
28876 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28879 * void vst3_lane_s8 (int8_t *, int8x8x3_t, const int)
28880 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28883 * void vst3_lane_f32 (float32_t *, float32x2x3_t, const int)
28884 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28887 * void vst3_lane_p16 (poly16_t *, poly16x4x3_t, const int)
28888 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28891 * void vst3_lane_p8 (poly8_t *, poly8x8x3_t, const int)
28892 _Form of expected instruction(s):_ `vst3.8 {D0[0], D1[0], D2[0]},
28895 * void vst3q_lane_s32 (int32_t *, int32x4x3_t, const int)
28896 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28899 * void vst3q_lane_s16 (int16_t *, int16x8x3_t, const int)
28900 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28903 * void vst3q_lane_u32 (uint32_t *, uint32x4x3_t, const int)
28904 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28907 * void vst3q_lane_u16 (uint16_t *, uint16x8x3_t, const int)
28908 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28911 * void vst3q_lane_f32 (float32_t *, float32x4x3_t, const int)
28912 _Form of expected instruction(s):_ `vst3.32 {D0[0], D1[0], D2[0]},
28915 * void vst3q_lane_p16 (poly16_t *, poly16x8x3_t, const int)
28916 _Form of expected instruction(s):_ `vst3.16 {D0[0], D1[0], D2[0]},
28919 5.50.3.74 Element/structure loads, VLD4 variants
28920 ................................................
28922 * uint32x2x4_t vld4_u32 (const uint32_t *)
28923 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28925 * uint16x4x4_t vld4_u16 (const uint16_t *)
28926 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28928 * uint8x8x4_t vld4_u8 (const uint8_t *)
28929 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28931 * int32x2x4_t vld4_s32 (const int32_t *)
28932 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28934 * int16x4x4_t vld4_s16 (const int16_t *)
28935 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28937 * int8x8x4_t vld4_s8 (const int8_t *)
28938 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28940 * float32x2x4_t vld4_f32 (const float32_t *)
28941 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28943 * poly16x4x4_t vld4_p16 (const poly16_t *)
28944 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28946 * poly8x8x4_t vld4_p8 (const poly8_t *)
28947 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28949 * uint64x1x4_t vld4_u64 (const uint64_t *)
28950 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
28952 * int64x1x4_t vld4_s64 (const int64_t *)
28953 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
28955 * uint32x4x4_t vld4q_u32 (const uint32_t *)
28956 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28958 * uint16x8x4_t vld4q_u16 (const uint16_t *)
28959 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28961 * uint8x16x4_t vld4q_u8 (const uint8_t *)
28962 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28964 * int32x4x4_t vld4q_s32 (const int32_t *)
28965 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28967 * int16x8x4_t vld4q_s16 (const int16_t *)
28968 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28970 * int8x16x4_t vld4q_s8 (const int8_t *)
28971 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28973 * float32x4x4_t vld4q_f32 (const float32_t *)
28974 _Form of expected instruction(s):_ `vld4.32 {D0, D1, D2, D3}, [R0]'
28976 * poly16x8x4_t vld4q_p16 (const poly16_t *)
28977 _Form of expected instruction(s):_ `vld4.16 {D0, D1, D2, D3}, [R0]'
28979 * poly8x16x4_t vld4q_p8 (const poly8_t *)
28980 _Form of expected instruction(s):_ `vld4.8 {D0, D1, D2, D3}, [R0]'
28982 * uint32x2x4_t vld4_lane_u32 (const uint32_t *, uint32x2x4_t, const
28984 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
28987 * uint16x4x4_t vld4_lane_u16 (const uint16_t *, uint16x4x4_t, const
28989 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
28992 * uint8x8x4_t vld4_lane_u8 (const uint8_t *, uint8x8x4_t, const int)
28993 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
28996 * int32x2x4_t vld4_lane_s32 (const int32_t *, int32x2x4_t, const int)
28997 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29000 * int16x4x4_t vld4_lane_s16 (const int16_t *, int16x4x4_t, const int)
29001 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29004 * int8x8x4_t vld4_lane_s8 (const int8_t *, int8x8x4_t, const int)
29005 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
29008 * float32x2x4_t vld4_lane_f32 (const float32_t *, float32x2x4_t,
29010 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29013 * poly16x4x4_t vld4_lane_p16 (const poly16_t *, poly16x4x4_t, const
29015 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29018 * poly8x8x4_t vld4_lane_p8 (const poly8_t *, poly8x8x4_t, const int)
29019 _Form of expected instruction(s):_ `vld4.8 {D0[0], D1[0], D2[0],
29022 * int32x4x4_t vld4q_lane_s32 (const int32_t *, int32x4x4_t, const
29024 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29027 * int16x8x4_t vld4q_lane_s16 (const int16_t *, int16x8x4_t, const
29029 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29032 * uint32x4x4_t vld4q_lane_u32 (const uint32_t *, uint32x4x4_t, const
29034 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29037 * uint16x8x4_t vld4q_lane_u16 (const uint16_t *, uint16x8x4_t, const
29039 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29042 * float32x4x4_t vld4q_lane_f32 (const float32_t *, float32x4x4_t,
29044 _Form of expected instruction(s):_ `vld4.32 {D0[0], D1[0], D2[0],
29047 * poly16x8x4_t vld4q_lane_p16 (const poly16_t *, poly16x8x4_t, const
29049 _Form of expected instruction(s):_ `vld4.16 {D0[0], D1[0], D2[0],
29052 * uint32x2x4_t vld4_dup_u32 (const uint32_t *)
29053 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29056 * uint16x4x4_t vld4_dup_u16 (const uint16_t *)
29057 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29060 * uint8x8x4_t vld4_dup_u8 (const uint8_t *)
29061 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29064 * int32x2x4_t vld4_dup_s32 (const int32_t *)
29065 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29068 * int16x4x4_t vld4_dup_s16 (const int16_t *)
29069 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29072 * int8x8x4_t vld4_dup_s8 (const int8_t *)
29073 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29076 * float32x2x4_t vld4_dup_f32 (const float32_t *)
29077 _Form of expected instruction(s):_ `vld4.32 {D0[], D1[], D2[],
29080 * poly16x4x4_t vld4_dup_p16 (const poly16_t *)
29081 _Form of expected instruction(s):_ `vld4.16 {D0[], D1[], D2[],
29084 * poly8x8x4_t vld4_dup_p8 (const poly8_t *)
29085 _Form of expected instruction(s):_ `vld4.8 {D0[], D1[], D2[],
29088 * uint64x1x4_t vld4_dup_u64 (const uint64_t *)
29089 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
29091 * int64x1x4_t vld4_dup_s64 (const int64_t *)
29092 _Form of expected instruction(s):_ `vld1.64 {D0, D1, D2, D3}, [R0]'
29094 5.50.3.75 Element/structure stores, VST4 variants
29095 .................................................
29097 * void vst4_u32 (uint32_t *, uint32x2x4_t)
29098 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29100 * void vst4_u16 (uint16_t *, uint16x4x4_t)
29101 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29103 * void vst4_u8 (uint8_t *, uint8x8x4_t)
29104 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29106 * void vst4_s32 (int32_t *, int32x2x4_t)
29107 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29109 * void vst4_s16 (int16_t *, int16x4x4_t)
29110 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29112 * void vst4_s8 (int8_t *, int8x8x4_t)
29113 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29115 * void vst4_f32 (float32_t *, float32x2x4_t)
29116 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29118 * void vst4_p16 (poly16_t *, poly16x4x4_t)
29119 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29121 * void vst4_p8 (poly8_t *, poly8x8x4_t)
29122 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29124 * void vst4_u64 (uint64_t *, uint64x1x4_t)
29125 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
29127 * void vst4_s64 (int64_t *, int64x1x4_t)
29128 _Form of expected instruction(s):_ `vst1.64 {D0, D1, D2, D3}, [R0]'
29130 * void vst4q_u32 (uint32_t *, uint32x4x4_t)
29131 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29133 * void vst4q_u16 (uint16_t *, uint16x8x4_t)
29134 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29136 * void vst4q_u8 (uint8_t *, uint8x16x4_t)
29137 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29139 * void vst4q_s32 (int32_t *, int32x4x4_t)
29140 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29142 * void vst4q_s16 (int16_t *, int16x8x4_t)
29143 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29145 * void vst4q_s8 (int8_t *, int8x16x4_t)
29146 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29148 * void vst4q_f32 (float32_t *, float32x4x4_t)
29149 _Form of expected instruction(s):_ `vst4.32 {D0, D1, D2, D3}, [R0]'
29151 * void vst4q_p16 (poly16_t *, poly16x8x4_t)
29152 _Form of expected instruction(s):_ `vst4.16 {D0, D1, D2, D3}, [R0]'
29154 * void vst4q_p8 (poly8_t *, poly8x16x4_t)
29155 _Form of expected instruction(s):_ `vst4.8 {D0, D1, D2, D3}, [R0]'
29157 * void vst4_lane_u32 (uint32_t *, uint32x2x4_t, const int)
29158 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29161 * void vst4_lane_u16 (uint16_t *, uint16x4x4_t, const int)
29162 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29165 * void vst4_lane_u8 (uint8_t *, uint8x8x4_t, const int)
29166 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29169 * void vst4_lane_s32 (int32_t *, int32x2x4_t, const int)
29170 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29173 * void vst4_lane_s16 (int16_t *, int16x4x4_t, const int)
29174 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29177 * void vst4_lane_s8 (int8_t *, int8x8x4_t, const int)
29178 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29181 * void vst4_lane_f32 (float32_t *, float32x2x4_t, const int)
29182 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29185 * void vst4_lane_p16 (poly16_t *, poly16x4x4_t, const int)
29186 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29189 * void vst4_lane_p8 (poly8_t *, poly8x8x4_t, const int)
29190 _Form of expected instruction(s):_ `vst4.8 {D0[0], D1[0], D2[0],
29193 * void vst4q_lane_s32 (int32_t *, int32x4x4_t, const int)
29194 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29197 * void vst4q_lane_s16 (int16_t *, int16x8x4_t, const int)
29198 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29201 * void vst4q_lane_u32 (uint32_t *, uint32x4x4_t, const int)
29202 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29205 * void vst4q_lane_u16 (uint16_t *, uint16x8x4_t, const int)
29206 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29209 * void vst4q_lane_f32 (float32_t *, float32x4x4_t, const int)
29210 _Form of expected instruction(s):_ `vst4.32 {D0[0], D1[0], D2[0],
29213 * void vst4q_lane_p16 (poly16_t *, poly16x8x4_t, const int)
29214 _Form of expected instruction(s):_ `vst4.16 {D0[0], D1[0], D2[0],
29217 5.50.3.76 Logical operations (AND)
29218 ..................................
29220 * uint32x2_t vand_u32 (uint32x2_t, uint32x2_t)
29221 _Form of expected instruction(s):_ `vand D0, D0, D0'
29223 * uint16x4_t vand_u16 (uint16x4_t, uint16x4_t)
29224 _Form of expected instruction(s):_ `vand D0, D0, D0'
29226 * uint8x8_t vand_u8 (uint8x8_t, uint8x8_t)
29227 _Form of expected instruction(s):_ `vand D0, D0, D0'
29229 * int32x2_t vand_s32 (int32x2_t, int32x2_t)
29230 _Form of expected instruction(s):_ `vand D0, D0, D0'
29232 * int16x4_t vand_s16 (int16x4_t, int16x4_t)
29233 _Form of expected instruction(s):_ `vand D0, D0, D0'
29235 * int8x8_t vand_s8 (int8x8_t, int8x8_t)
29236 _Form of expected instruction(s):_ `vand D0, D0, D0'
29238 * uint64x1_t vand_u64 (uint64x1_t, uint64x1_t)
29239 _Form of expected instruction(s):_ `vand D0, D0, D0'
29241 * int64x1_t vand_s64 (int64x1_t, int64x1_t)
29242 _Form of expected instruction(s):_ `vand D0, D0, D0'
29244 * uint32x4_t vandq_u32 (uint32x4_t, uint32x4_t)
29245 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29247 * uint16x8_t vandq_u16 (uint16x8_t, uint16x8_t)
29248 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29250 * uint8x16_t vandq_u8 (uint8x16_t, uint8x16_t)
29251 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29253 * int32x4_t vandq_s32 (int32x4_t, int32x4_t)
29254 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29256 * int16x8_t vandq_s16 (int16x8_t, int16x8_t)
29257 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29259 * int8x16_t vandq_s8 (int8x16_t, int8x16_t)
29260 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29262 * uint64x2_t vandq_u64 (uint64x2_t, uint64x2_t)
29263 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29265 * int64x2_t vandq_s64 (int64x2_t, int64x2_t)
29266 _Form of expected instruction(s):_ `vand Q0, Q0, Q0'
29268 5.50.3.77 Logical operations (OR)
29269 .................................
29271 * uint32x2_t vorr_u32 (uint32x2_t, uint32x2_t)
29272 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29274 * uint16x4_t vorr_u16 (uint16x4_t, uint16x4_t)
29275 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29277 * uint8x8_t vorr_u8 (uint8x8_t, uint8x8_t)
29278 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29280 * int32x2_t vorr_s32 (int32x2_t, int32x2_t)
29281 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29283 * int16x4_t vorr_s16 (int16x4_t, int16x4_t)
29284 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29286 * int8x8_t vorr_s8 (int8x8_t, int8x8_t)
29287 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29289 * uint64x1_t vorr_u64 (uint64x1_t, uint64x1_t)
29290 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29292 * int64x1_t vorr_s64 (int64x1_t, int64x1_t)
29293 _Form of expected instruction(s):_ `vorr D0, D0, D0'
29295 * uint32x4_t vorrq_u32 (uint32x4_t, uint32x4_t)
29296 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29298 * uint16x8_t vorrq_u16 (uint16x8_t, uint16x8_t)
29299 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29301 * uint8x16_t vorrq_u8 (uint8x16_t, uint8x16_t)
29302 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29304 * int32x4_t vorrq_s32 (int32x4_t, int32x4_t)
29305 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29307 * int16x8_t vorrq_s16 (int16x8_t, int16x8_t)
29308 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29310 * int8x16_t vorrq_s8 (int8x16_t, int8x16_t)
29311 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29313 * uint64x2_t vorrq_u64 (uint64x2_t, uint64x2_t)
29314 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29316 * int64x2_t vorrq_s64 (int64x2_t, int64x2_t)
29317 _Form of expected instruction(s):_ `vorr Q0, Q0, Q0'
29319 5.50.3.78 Logical operations (exclusive OR)
29320 ...........................................
29322 * uint32x2_t veor_u32 (uint32x2_t, uint32x2_t)
29323 _Form of expected instruction(s):_ `veor D0, D0, D0'
29325 * uint16x4_t veor_u16 (uint16x4_t, uint16x4_t)
29326 _Form of expected instruction(s):_ `veor D0, D0, D0'
29328 * uint8x8_t veor_u8 (uint8x8_t, uint8x8_t)
29329 _Form of expected instruction(s):_ `veor D0, D0, D0'
29331 * int32x2_t veor_s32 (int32x2_t, int32x2_t)
29332 _Form of expected instruction(s):_ `veor D0, D0, D0'
29334 * int16x4_t veor_s16 (int16x4_t, int16x4_t)
29335 _Form of expected instruction(s):_ `veor D0, D0, D0'
29337 * int8x8_t veor_s8 (int8x8_t, int8x8_t)
29338 _Form of expected instruction(s):_ `veor D0, D0, D0'
29340 * uint64x1_t veor_u64 (uint64x1_t, uint64x1_t)
29341 _Form of expected instruction(s):_ `veor D0, D0, D0'
29343 * int64x1_t veor_s64 (int64x1_t, int64x1_t)
29344 _Form of expected instruction(s):_ `veor D0, D0, D0'
29346 * uint32x4_t veorq_u32 (uint32x4_t, uint32x4_t)
29347 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29349 * uint16x8_t veorq_u16 (uint16x8_t, uint16x8_t)
29350 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29352 * uint8x16_t veorq_u8 (uint8x16_t, uint8x16_t)
29353 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29355 * int32x4_t veorq_s32 (int32x4_t, int32x4_t)
29356 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29358 * int16x8_t veorq_s16 (int16x8_t, int16x8_t)
29359 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29361 * int8x16_t veorq_s8 (int8x16_t, int8x16_t)
29362 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29364 * uint64x2_t veorq_u64 (uint64x2_t, uint64x2_t)
29365 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29367 * int64x2_t veorq_s64 (int64x2_t, int64x2_t)
29368 _Form of expected instruction(s):_ `veor Q0, Q0, Q0'
29370 5.50.3.79 Logical operations (AND-NOT)
29371 ......................................
29373 * uint32x2_t vbic_u32 (uint32x2_t, uint32x2_t)
29374 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29376 * uint16x4_t vbic_u16 (uint16x4_t, uint16x4_t)
29377 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29379 * uint8x8_t vbic_u8 (uint8x8_t, uint8x8_t)
29380 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29382 * int32x2_t vbic_s32 (int32x2_t, int32x2_t)
29383 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29385 * int16x4_t vbic_s16 (int16x4_t, int16x4_t)
29386 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29388 * int8x8_t vbic_s8 (int8x8_t, int8x8_t)
29389 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29391 * uint64x1_t vbic_u64 (uint64x1_t, uint64x1_t)
29392 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29394 * int64x1_t vbic_s64 (int64x1_t, int64x1_t)
29395 _Form of expected instruction(s):_ `vbic D0, D0, D0'
29397 * uint32x4_t vbicq_u32 (uint32x4_t, uint32x4_t)
29398 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29400 * uint16x8_t vbicq_u16 (uint16x8_t, uint16x8_t)
29401 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29403 * uint8x16_t vbicq_u8 (uint8x16_t, uint8x16_t)
29404 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29406 * int32x4_t vbicq_s32 (int32x4_t, int32x4_t)
29407 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29409 * int16x8_t vbicq_s16 (int16x8_t, int16x8_t)
29410 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29412 * int8x16_t vbicq_s8 (int8x16_t, int8x16_t)
29413 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29415 * uint64x2_t vbicq_u64 (uint64x2_t, uint64x2_t)
29416 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29418 * int64x2_t vbicq_s64 (int64x2_t, int64x2_t)
29419 _Form of expected instruction(s):_ `vbic Q0, Q0, Q0'
29421 5.50.3.80 Logical operations (OR-NOT)
29422 .....................................
29424 * uint32x2_t vorn_u32 (uint32x2_t, uint32x2_t)
29425 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29427 * uint16x4_t vorn_u16 (uint16x4_t, uint16x4_t)
29428 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29430 * uint8x8_t vorn_u8 (uint8x8_t, uint8x8_t)
29431 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29433 * int32x2_t vorn_s32 (int32x2_t, int32x2_t)
29434 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29436 * int16x4_t vorn_s16 (int16x4_t, int16x4_t)
29437 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29439 * int8x8_t vorn_s8 (int8x8_t, int8x8_t)
29440 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29442 * uint64x1_t vorn_u64 (uint64x1_t, uint64x1_t)
29443 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29445 * int64x1_t vorn_s64 (int64x1_t, int64x1_t)
29446 _Form of expected instruction(s):_ `vorn D0, D0, D0'
29448 * uint32x4_t vornq_u32 (uint32x4_t, uint32x4_t)
29449 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29451 * uint16x8_t vornq_u16 (uint16x8_t, uint16x8_t)
29452 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29454 * uint8x16_t vornq_u8 (uint8x16_t, uint8x16_t)
29455 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29457 * int32x4_t vornq_s32 (int32x4_t, int32x4_t)
29458 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29460 * int16x8_t vornq_s16 (int16x8_t, int16x8_t)
29461 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29463 * int8x16_t vornq_s8 (int8x16_t, int8x16_t)
29464 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29466 * uint64x2_t vornq_u64 (uint64x2_t, uint64x2_t)
29467 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29469 * int64x2_t vornq_s64 (int64x2_t, int64x2_t)
29470 _Form of expected instruction(s):_ `vorn Q0, Q0, Q0'
29472 5.50.3.81 Reinterpret casts
29473 ...........................
29475 * poly8x8_t vreinterpret_p8_u32 (uint32x2_t)
29477 * poly8x8_t vreinterpret_p8_u16 (uint16x4_t)
29479 * poly8x8_t vreinterpret_p8_u8 (uint8x8_t)
29481 * poly8x8_t vreinterpret_p8_s32 (int32x2_t)
29483 * poly8x8_t vreinterpret_p8_s16 (int16x4_t)
29485 * poly8x8_t vreinterpret_p8_s8 (int8x8_t)
29487 * poly8x8_t vreinterpret_p8_u64 (uint64x1_t)
29489 * poly8x8_t vreinterpret_p8_s64 (int64x1_t)
29491 * poly8x8_t vreinterpret_p8_f32 (float32x2_t)
29493 * poly8x8_t vreinterpret_p8_p16 (poly16x4_t)
29495 * poly8x16_t vreinterpretq_p8_u32 (uint32x4_t)
29497 * poly8x16_t vreinterpretq_p8_u16 (uint16x8_t)
29499 * poly8x16_t vreinterpretq_p8_u8 (uint8x16_t)
29501 * poly8x16_t vreinterpretq_p8_s32 (int32x4_t)
29503 * poly8x16_t vreinterpretq_p8_s16 (int16x8_t)
29505 * poly8x16_t vreinterpretq_p8_s8 (int8x16_t)
29507 * poly8x16_t vreinterpretq_p8_u64 (uint64x2_t)
29509 * poly8x16_t vreinterpretq_p8_s64 (int64x2_t)
29511 * poly8x16_t vreinterpretq_p8_f32 (float32x4_t)
29513 * poly8x16_t vreinterpretq_p8_p16 (poly16x8_t)
29515 * poly16x4_t vreinterpret_p16_u32 (uint32x2_t)
29517 * poly16x4_t vreinterpret_p16_u16 (uint16x4_t)
29519 * poly16x4_t vreinterpret_p16_u8 (uint8x8_t)
29521 * poly16x4_t vreinterpret_p16_s32 (int32x2_t)
29523 * poly16x4_t vreinterpret_p16_s16 (int16x4_t)
29525 * poly16x4_t vreinterpret_p16_s8 (int8x8_t)
29527 * poly16x4_t vreinterpret_p16_u64 (uint64x1_t)
29529 * poly16x4_t vreinterpret_p16_s64 (int64x1_t)
29531 * poly16x4_t vreinterpret_p16_f32 (float32x2_t)
29533 * poly16x4_t vreinterpret_p16_p8 (poly8x8_t)
29535 * poly16x8_t vreinterpretq_p16_u32 (uint32x4_t)
29537 * poly16x8_t vreinterpretq_p16_u16 (uint16x8_t)
29539 * poly16x8_t vreinterpretq_p16_u8 (uint8x16_t)
29541 * poly16x8_t vreinterpretq_p16_s32 (int32x4_t)
29543 * poly16x8_t vreinterpretq_p16_s16 (int16x8_t)
29545 * poly16x8_t vreinterpretq_p16_s8 (int8x16_t)
29547 * poly16x8_t vreinterpretq_p16_u64 (uint64x2_t)
29549 * poly16x8_t vreinterpretq_p16_s64 (int64x2_t)
29551 * poly16x8_t vreinterpretq_p16_f32 (float32x4_t)
29553 * poly16x8_t vreinterpretq_p16_p8 (poly8x16_t)
29555 * float32x2_t vreinterpret_f32_u32 (uint32x2_t)
29557 * float32x2_t vreinterpret_f32_u16 (uint16x4_t)
29559 * float32x2_t vreinterpret_f32_u8 (uint8x8_t)
29561 * float32x2_t vreinterpret_f32_s32 (int32x2_t)
29563 * float32x2_t vreinterpret_f32_s16 (int16x4_t)
29565 * float32x2_t vreinterpret_f32_s8 (int8x8_t)
29567 * float32x2_t vreinterpret_f32_u64 (uint64x1_t)
29569 * float32x2_t vreinterpret_f32_s64 (int64x1_t)
29571 * float32x2_t vreinterpret_f32_p16 (poly16x4_t)
29573 * float32x2_t vreinterpret_f32_p8 (poly8x8_t)
29575 * float32x4_t vreinterpretq_f32_u32 (uint32x4_t)
29577 * float32x4_t vreinterpretq_f32_u16 (uint16x8_t)
29579 * float32x4_t vreinterpretq_f32_u8 (uint8x16_t)
29581 * float32x4_t vreinterpretq_f32_s32 (int32x4_t)
29583 * float32x4_t vreinterpretq_f32_s16 (int16x8_t)
29585 * float32x4_t vreinterpretq_f32_s8 (int8x16_t)
29587 * float32x4_t vreinterpretq_f32_u64 (uint64x2_t)
29589 * float32x4_t vreinterpretq_f32_s64 (int64x2_t)
29591 * float32x4_t vreinterpretq_f32_p16 (poly16x8_t)
29593 * float32x4_t vreinterpretq_f32_p8 (poly8x16_t)
29595 * int64x1_t vreinterpret_s64_u32 (uint32x2_t)
29597 * int64x1_t vreinterpret_s64_u16 (uint16x4_t)
29599 * int64x1_t vreinterpret_s64_u8 (uint8x8_t)
29601 * int64x1_t vreinterpret_s64_s32 (int32x2_t)
29603 * int64x1_t vreinterpret_s64_s16 (int16x4_t)
29605 * int64x1_t vreinterpret_s64_s8 (int8x8_t)
29607 * int64x1_t vreinterpret_s64_u64 (uint64x1_t)
29609 * int64x1_t vreinterpret_s64_f32 (float32x2_t)
29611 * int64x1_t vreinterpret_s64_p16 (poly16x4_t)
29613 * int64x1_t vreinterpret_s64_p8 (poly8x8_t)
29615 * int64x2_t vreinterpretq_s64_u32 (uint32x4_t)
29617 * int64x2_t vreinterpretq_s64_u16 (uint16x8_t)
29619 * int64x2_t vreinterpretq_s64_u8 (uint8x16_t)
29621 * int64x2_t vreinterpretq_s64_s32 (int32x4_t)
29623 * int64x2_t vreinterpretq_s64_s16 (int16x8_t)
29625 * int64x2_t vreinterpretq_s64_s8 (int8x16_t)
29627 * int64x2_t vreinterpretq_s64_u64 (uint64x2_t)
29629 * int64x2_t vreinterpretq_s64_f32 (float32x4_t)
29631 * int64x2_t vreinterpretq_s64_p16 (poly16x8_t)
29633 * int64x2_t vreinterpretq_s64_p8 (poly8x16_t)
29635 * uint64x1_t vreinterpret_u64_u32 (uint32x2_t)
29637 * uint64x1_t vreinterpret_u64_u16 (uint16x4_t)
29639 * uint64x1_t vreinterpret_u64_u8 (uint8x8_t)
29641 * uint64x1_t vreinterpret_u64_s32 (int32x2_t)
29643 * uint64x1_t vreinterpret_u64_s16 (int16x4_t)
29645 * uint64x1_t vreinterpret_u64_s8 (int8x8_t)
29647 * uint64x1_t vreinterpret_u64_s64 (int64x1_t)
29649 * uint64x1_t vreinterpret_u64_f32 (float32x2_t)
29651 * uint64x1_t vreinterpret_u64_p16 (poly16x4_t)
29653 * uint64x1_t vreinterpret_u64_p8 (poly8x8_t)
29655 * uint64x2_t vreinterpretq_u64_u32 (uint32x4_t)
29657 * uint64x2_t vreinterpretq_u64_u16 (uint16x8_t)
29659 * uint64x2_t vreinterpretq_u64_u8 (uint8x16_t)
29661 * uint64x2_t vreinterpretq_u64_s32 (int32x4_t)
29663 * uint64x2_t vreinterpretq_u64_s16 (int16x8_t)
29665 * uint64x2_t vreinterpretq_u64_s8 (int8x16_t)
29667 * uint64x2_t vreinterpretq_u64_s64 (int64x2_t)
29669 * uint64x2_t vreinterpretq_u64_f32 (float32x4_t)
29671 * uint64x2_t vreinterpretq_u64_p16 (poly16x8_t)
29673 * uint64x2_t vreinterpretq_u64_p8 (poly8x16_t)
29675 * int8x8_t vreinterpret_s8_u32 (uint32x2_t)
29677 * int8x8_t vreinterpret_s8_u16 (uint16x4_t)
29679 * int8x8_t vreinterpret_s8_u8 (uint8x8_t)
29681 * int8x8_t vreinterpret_s8_s32 (int32x2_t)
29683 * int8x8_t vreinterpret_s8_s16 (int16x4_t)
29685 * int8x8_t vreinterpret_s8_u64 (uint64x1_t)
29687 * int8x8_t vreinterpret_s8_s64 (int64x1_t)
29689 * int8x8_t vreinterpret_s8_f32 (float32x2_t)
29691 * int8x8_t vreinterpret_s8_p16 (poly16x4_t)
29693 * int8x8_t vreinterpret_s8_p8 (poly8x8_t)
29695 * int8x16_t vreinterpretq_s8_u32 (uint32x4_t)
29697 * int8x16_t vreinterpretq_s8_u16 (uint16x8_t)
29699 * int8x16_t vreinterpretq_s8_u8 (uint8x16_t)
29701 * int8x16_t vreinterpretq_s8_s32 (int32x4_t)
29703 * int8x16_t vreinterpretq_s8_s16 (int16x8_t)
29705 * int8x16_t vreinterpretq_s8_u64 (uint64x2_t)
29707 * int8x16_t vreinterpretq_s8_s64 (int64x2_t)
29709 * int8x16_t vreinterpretq_s8_f32 (float32x4_t)
29711 * int8x16_t vreinterpretq_s8_p16 (poly16x8_t)
29713 * int8x16_t vreinterpretq_s8_p8 (poly8x16_t)
29715 * int16x4_t vreinterpret_s16_u32 (uint32x2_t)
29717 * int16x4_t vreinterpret_s16_u16 (uint16x4_t)
29719 * int16x4_t vreinterpret_s16_u8 (uint8x8_t)
29721 * int16x4_t vreinterpret_s16_s32 (int32x2_t)
29723 * int16x4_t vreinterpret_s16_s8 (int8x8_t)
29725 * int16x4_t vreinterpret_s16_u64 (uint64x1_t)
29727 * int16x4_t vreinterpret_s16_s64 (int64x1_t)
29729 * int16x4_t vreinterpret_s16_f32 (float32x2_t)
29731 * int16x4_t vreinterpret_s16_p16 (poly16x4_t)
29733 * int16x4_t vreinterpret_s16_p8 (poly8x8_t)
29735 * int16x8_t vreinterpretq_s16_u32 (uint32x4_t)
29737 * int16x8_t vreinterpretq_s16_u16 (uint16x8_t)
29739 * int16x8_t vreinterpretq_s16_u8 (uint8x16_t)
29741 * int16x8_t vreinterpretq_s16_s32 (int32x4_t)
29743 * int16x8_t vreinterpretq_s16_s8 (int8x16_t)
29745 * int16x8_t vreinterpretq_s16_u64 (uint64x2_t)
29747 * int16x8_t vreinterpretq_s16_s64 (int64x2_t)
29749 * int16x8_t vreinterpretq_s16_f32 (float32x4_t)
29751 * int16x8_t vreinterpretq_s16_p16 (poly16x8_t)
29753 * int16x8_t vreinterpretq_s16_p8 (poly8x16_t)
29755 * int32x2_t vreinterpret_s32_u32 (uint32x2_t)
29757 * int32x2_t vreinterpret_s32_u16 (uint16x4_t)
29759 * int32x2_t vreinterpret_s32_u8 (uint8x8_t)
29761 * int32x2_t vreinterpret_s32_s16 (int16x4_t)
29763 * int32x2_t vreinterpret_s32_s8 (int8x8_t)
29765 * int32x2_t vreinterpret_s32_u64 (uint64x1_t)
29767 * int32x2_t vreinterpret_s32_s64 (int64x1_t)
29769 * int32x2_t vreinterpret_s32_f32 (float32x2_t)
29771 * int32x2_t vreinterpret_s32_p16 (poly16x4_t)
29773 * int32x2_t vreinterpret_s32_p8 (poly8x8_t)
29775 * int32x4_t vreinterpretq_s32_u32 (uint32x4_t)
29777 * int32x4_t vreinterpretq_s32_u16 (uint16x8_t)
29779 * int32x4_t vreinterpretq_s32_u8 (uint8x16_t)
29781 * int32x4_t vreinterpretq_s32_s16 (int16x8_t)
29783 * int32x4_t vreinterpretq_s32_s8 (int8x16_t)
29785 * int32x4_t vreinterpretq_s32_u64 (uint64x2_t)
29787 * int32x4_t vreinterpretq_s32_s64 (int64x2_t)
29789 * int32x4_t vreinterpretq_s32_f32 (float32x4_t)
29791 * int32x4_t vreinterpretq_s32_p16 (poly16x8_t)
29793 * int32x4_t vreinterpretq_s32_p8 (poly8x16_t)
29795 * uint8x8_t vreinterpret_u8_u32 (uint32x2_t)
29797 * uint8x8_t vreinterpret_u8_u16 (uint16x4_t)
29799 * uint8x8_t vreinterpret_u8_s32 (int32x2_t)
29801 * uint8x8_t vreinterpret_u8_s16 (int16x4_t)
29803 * uint8x8_t vreinterpret_u8_s8 (int8x8_t)
29805 * uint8x8_t vreinterpret_u8_u64 (uint64x1_t)
29807 * uint8x8_t vreinterpret_u8_s64 (int64x1_t)
29809 * uint8x8_t vreinterpret_u8_f32 (float32x2_t)
29811 * uint8x8_t vreinterpret_u8_p16 (poly16x4_t)
29813 * uint8x8_t vreinterpret_u8_p8 (poly8x8_t)
29815 * uint8x16_t vreinterpretq_u8_u32 (uint32x4_t)
29817 * uint8x16_t vreinterpretq_u8_u16 (uint16x8_t)
29819 * uint8x16_t vreinterpretq_u8_s32 (int32x4_t)
29821 * uint8x16_t vreinterpretq_u8_s16 (int16x8_t)
29823 * uint8x16_t vreinterpretq_u8_s8 (int8x16_t)
29825 * uint8x16_t vreinterpretq_u8_u64 (uint64x2_t)
29827 * uint8x16_t vreinterpretq_u8_s64 (int64x2_t)
29829 * uint8x16_t vreinterpretq_u8_f32 (float32x4_t)
29831 * uint8x16_t vreinterpretq_u8_p16 (poly16x8_t)
29833 * uint8x16_t vreinterpretq_u8_p8 (poly8x16_t)
29835 * uint16x4_t vreinterpret_u16_u32 (uint32x2_t)
29837 * uint16x4_t vreinterpret_u16_u8 (uint8x8_t)
29839 * uint16x4_t vreinterpret_u16_s32 (int32x2_t)
29841 * uint16x4_t vreinterpret_u16_s16 (int16x4_t)
29843 * uint16x4_t vreinterpret_u16_s8 (int8x8_t)
29845 * uint16x4_t vreinterpret_u16_u64 (uint64x1_t)
29847 * uint16x4_t vreinterpret_u16_s64 (int64x1_t)
29849 * uint16x4_t vreinterpret_u16_f32 (float32x2_t)
29851 * uint16x4_t vreinterpret_u16_p16 (poly16x4_t)
29853 * uint16x4_t vreinterpret_u16_p8 (poly8x8_t)
29855 * uint16x8_t vreinterpretq_u16_u32 (uint32x4_t)
29857 * uint16x8_t vreinterpretq_u16_u8 (uint8x16_t)
29859 * uint16x8_t vreinterpretq_u16_s32 (int32x4_t)
29861 * uint16x8_t vreinterpretq_u16_s16 (int16x8_t)
29863 * uint16x8_t vreinterpretq_u16_s8 (int8x16_t)
29865 * uint16x8_t vreinterpretq_u16_u64 (uint64x2_t)
29867 * uint16x8_t vreinterpretq_u16_s64 (int64x2_t)
29869 * uint16x8_t vreinterpretq_u16_f32 (float32x4_t)
29871 * uint16x8_t vreinterpretq_u16_p16 (poly16x8_t)
29873 * uint16x8_t vreinterpretq_u16_p8 (poly8x16_t)
29875 * uint32x2_t vreinterpret_u32_u16 (uint16x4_t)
29877 * uint32x2_t vreinterpret_u32_u8 (uint8x8_t)
29879 * uint32x2_t vreinterpret_u32_s32 (int32x2_t)
29881 * uint32x2_t vreinterpret_u32_s16 (int16x4_t)
29883 * uint32x2_t vreinterpret_u32_s8 (int8x8_t)
29885 * uint32x2_t vreinterpret_u32_u64 (uint64x1_t)
29887 * uint32x2_t vreinterpret_u32_s64 (int64x1_t)
29889 * uint32x2_t vreinterpret_u32_f32 (float32x2_t)
29891 * uint32x2_t vreinterpret_u32_p16 (poly16x4_t)
29893 * uint32x2_t vreinterpret_u32_p8 (poly8x8_t)
29895 * uint32x4_t vreinterpretq_u32_u16 (uint16x8_t)
29897 * uint32x4_t vreinterpretq_u32_u8 (uint8x16_t)
29899 * uint32x4_t vreinterpretq_u32_s32 (int32x4_t)
29901 * uint32x4_t vreinterpretq_u32_s16 (int16x8_t)
29903 * uint32x4_t vreinterpretq_u32_s8 (int8x16_t)
29905 * uint32x4_t vreinterpretq_u32_u64 (uint64x2_t)
29907 * uint32x4_t vreinterpretq_u32_s64 (int64x2_t)
29909 * uint32x4_t vreinterpretq_u32_f32 (float32x4_t)
29911 * uint32x4_t vreinterpretq_u32_p16 (poly16x8_t)
29913 * uint32x4_t vreinterpretq_u32_p8 (poly8x16_t)
29916 File: gcc.info, Node: Blackfin Built-in Functions, Next: FR-V Built-in Functions, Prev: ARM NEON Intrinsics, Up: Target Builtins
29918 5.50.4 Blackfin Built-in Functions
29919 ----------------------------------
29921 Currently, there are two Blackfin-specific built-in functions. These
29922 are used for generating `CSYNC' and `SSYNC' machine insns without using
29923 inline assembly; by using these built-in functions the compiler can
29924 automatically add workarounds for hardware errata involving these
29925 instructions. These functions are named as follows:
29927 void __builtin_bfin_csync (void)
29928 void __builtin_bfin_ssync (void)
29931 File: gcc.info, Node: FR-V Built-in Functions, Next: X86 Built-in Functions, Prev: Blackfin Built-in Functions, Up: Target Builtins
29933 5.50.5 FR-V Built-in Functions
29934 ------------------------------
29936 GCC provides many FR-V-specific built-in functions. In general, these
29937 functions are intended to be compatible with those described by `FR-V
29938 Family, Softune C/C++ Compiler Manual (V6), Fujitsu Semiconductor'.
29939 The two exceptions are `__MDUNPACKH' and `__MBTOHE', the gcc forms of
29940 which pass 128-bit values by pointer rather than by value.
29942 Most of the functions are named after specific FR-V instructions.
29943 Such functions are said to be "directly mapped" and are summarized here
29949 * Directly-mapped Integer Functions::
29950 * Directly-mapped Media Functions::
29951 * Raw read/write Functions::
29952 * Other Built-in Functions::
29955 File: gcc.info, Node: Argument Types, Next: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
29957 5.50.5.1 Argument Types
29958 .......................
29960 The arguments to the built-in functions can be divided into three
29961 groups: register numbers, compile-time constants and run-time values.
29962 In order to make this classification clear at a glance, the arguments
29963 and return values are given the following pseudo types:
29965 Pseudo type Real C type Constant? Description
29966 `uh' `unsigned short' No an unsigned halfword
29967 `uw1' `unsigned int' No an unsigned word
29968 `sw1' `int' No a signed word
29969 `uw2' `unsigned long long' No an unsigned doubleword
29970 `sw2' `long long' No a signed doubleword
29971 `const' `int' Yes an integer constant
29972 `acc' `int' Yes an ACC register number
29973 `iacc' `int' Yes an IACC register number
29975 These pseudo types are not defined by GCC, they are simply a notational
29976 convenience used in this manual.
29978 Arguments of type `uh', `uw1', `sw1', `uw2' and `sw2' are evaluated at
29979 run time. They correspond to register operands in the underlying FR-V
29982 `const' arguments represent immediate operands in the underlying FR-V
29983 instructions. They must be compile-time constants.
29985 `acc' arguments are evaluated at compile time and specify the number
29986 of an accumulator register. For example, an `acc' argument of 2 will
29987 select the ACC2 register.
29989 `iacc' arguments are similar to `acc' arguments but specify the number
29990 of an IACC register. See *note Other Built-in Functions:: for more
29994 File: gcc.info, Node: Directly-mapped Integer Functions, Next: Directly-mapped Media Functions, Prev: Argument Types, Up: FR-V Built-in Functions
29996 5.50.5.2 Directly-mapped Integer Functions
29997 ..........................................
29999 The functions listed below map directly to FR-V I-type instructions.
30001 Function prototype Example usage Assembly output
30002 `sw1 __ADDSS (sw1, sw1)' `C = __ADDSS (A, B)' `ADDSS A,B,C'
30003 `sw1 __SCAN (sw1, sw1)' `C = __SCAN (A, B)' `SCAN A,B,C'
30004 `sw1 __SCUTSS (sw1)' `B = __SCUTSS (A)' `SCUTSS A,B'
30005 `sw1 __SLASS (sw1, sw1)' `C = __SLASS (A, B)' `SLASS A,B,C'
30006 `void __SMASS (sw1, sw1)' `__SMASS (A, B)' `SMASS A,B'
30007 `void __SMSSS (sw1, sw1)' `__SMSSS (A, B)' `SMSSS A,B'
30008 `void __SMU (sw1, sw1)' `__SMU (A, B)' `SMU A,B'
30009 `sw2 __SMUL (sw1, sw1)' `C = __SMUL (A, B)' `SMUL A,B,C'
30010 `sw1 __SUBSS (sw1, sw1)' `C = __SUBSS (A, B)' `SUBSS A,B,C'
30011 `uw2 __UMUL (uw1, uw1)' `C = __UMUL (A, B)' `UMUL A,B,C'
30014 File: gcc.info, Node: Directly-mapped Media Functions, Next: Raw read/write Functions, Prev: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
30016 5.50.5.3 Directly-mapped Media Functions
30017 ........................................
30019 The functions listed below map directly to FR-V M-type instructions.
30021 Function prototype Example usage Assembly output
30022 `uw1 __MABSHS (sw1)' `B = __MABSHS (A)' `MABSHS A,B'
30023 `void __MADDACCS (acc, acc)' `__MADDACCS (B, A)' `MADDACCS A,B'
30024 `sw1 __MADDHSS (sw1, sw1)' `C = __MADDHSS (A, B)' `MADDHSS A,B,C'
30025 `uw1 __MADDHUS (uw1, uw1)' `C = __MADDHUS (A, B)' `MADDHUS A,B,C'
30026 `uw1 __MAND (uw1, uw1)' `C = __MAND (A, B)' `MAND A,B,C'
30027 `void __MASACCS (acc, acc)' `__MASACCS (B, A)' `MASACCS A,B'
30028 `uw1 __MAVEH (uw1, uw1)' `C = __MAVEH (A, B)' `MAVEH A,B,C'
30029 `uw2 __MBTOH (uw1)' `B = __MBTOH (A)' `MBTOH A,B'
30030 `void __MBTOHE (uw1 *, uw1)' `__MBTOHE (&B, A)' `MBTOHE A,B'
30031 `void __MCLRACC (acc)' `__MCLRACC (A)' `MCLRACC A'
30032 `void __MCLRACCA (void)' `__MCLRACCA ()' `MCLRACCA'
30033 `uw1 __Mcop1 (uw1, uw1)' `C = __Mcop1 (A, B)' `Mcop1 A,B,C'
30034 `uw1 __Mcop2 (uw1, uw1)' `C = __Mcop2 (A, B)' `Mcop2 A,B,C'
30035 `uw1 __MCPLHI (uw2, const)' `C = __MCPLHI (A, B)' `MCPLHI A,#B,C'
30036 `uw1 __MCPLI (uw2, const)' `C = __MCPLI (A, B)' `MCPLI A,#B,C'
30037 `void __MCPXIS (acc, sw1, sw1)' `__MCPXIS (C, A, B)' `MCPXIS A,B,C'
30038 `void __MCPXIU (acc, uw1, uw1)' `__MCPXIU (C, A, B)' `MCPXIU A,B,C'
30039 `void __MCPXRS (acc, sw1, sw1)' `__MCPXRS (C, A, B)' `MCPXRS A,B,C'
30040 `void __MCPXRU (acc, uw1, uw1)' `__MCPXRU (C, A, B)' `MCPXRU A,B,C'
30041 `uw1 __MCUT (acc, uw1)' `C = __MCUT (A, B)' `MCUT A,B,C'
30042 `uw1 __MCUTSS (acc, sw1)' `C = __MCUTSS (A, B)' `MCUTSS A,B,C'
30043 `void __MDADDACCS (acc, acc)' `__MDADDACCS (B, A)' `MDADDACCS A,B'
30044 `void __MDASACCS (acc, acc)' `__MDASACCS (B, A)' `MDASACCS A,B'
30045 `uw2 __MDCUTSSI (acc, const)' `C = __MDCUTSSI (A, B)' `MDCUTSSI A,#B,C'
30046 `uw2 __MDPACKH (uw2, uw2)' `C = __MDPACKH (A, B)' `MDPACKH A,B,C'
30047 `uw2 __MDROTLI (uw2, const)' `C = __MDROTLI (A, B)' `MDROTLI A,#B,C'
30048 `void __MDSUBACCS (acc, acc)' `__MDSUBACCS (B, A)' `MDSUBACCS A,B'
30049 `void __MDUNPACKH (uw1 *, uw2)' `__MDUNPACKH (&B, A)' `MDUNPACKH A,B'
30050 `uw2 __MEXPDHD (uw1, const)' `C = __MEXPDHD (A, B)' `MEXPDHD A,#B,C'
30051 `uw1 __MEXPDHW (uw1, const)' `C = __MEXPDHW (A, B)' `MEXPDHW A,#B,C'
30052 `uw1 __MHDSETH (uw1, const)' `C = __MHDSETH (A, B)' `MHDSETH A,#B,C'
30053 `sw1 __MHDSETS (const)' `B = __MHDSETS (A)' `MHDSETS #A,B'
30054 `uw1 __MHSETHIH (uw1, const)' `B = __MHSETHIH (B, A)' `MHSETHIH #A,B'
30055 `sw1 __MHSETHIS (sw1, const)' `B = __MHSETHIS (B, A)' `MHSETHIS #A,B'
30056 `uw1 __MHSETLOH (uw1, const)' `B = __MHSETLOH (B, A)' `MHSETLOH #A,B'
30057 `sw1 __MHSETLOS (sw1, const)' `B = __MHSETLOS (B, A)' `MHSETLOS #A,B'
30058 `uw1 __MHTOB (uw2)' `B = __MHTOB (A)' `MHTOB A,B'
30059 `void __MMACHS (acc, sw1, sw1)' `__MMACHS (C, A, B)' `MMACHS A,B,C'
30060 `void __MMACHU (acc, uw1, uw1)' `__MMACHU (C, A, B)' `MMACHU A,B,C'
30061 `void __MMRDHS (acc, sw1, sw1)' `__MMRDHS (C, A, B)' `MMRDHS A,B,C'
30062 `void __MMRDHU (acc, uw1, uw1)' `__MMRDHU (C, A, B)' `MMRDHU A,B,C'
30063 `void __MMULHS (acc, sw1, sw1)' `__MMULHS (C, A, B)' `MMULHS A,B,C'
30064 `void __MMULHU (acc, uw1, uw1)' `__MMULHU (C, A, B)' `MMULHU A,B,C'
30065 `void __MMULXHS (acc, sw1, sw1)' `__MMULXHS (C, A, B)' `MMULXHS A,B,C'
30066 `void __MMULXHU (acc, uw1, uw1)' `__MMULXHU (C, A, B)' `MMULXHU A,B,C'
30067 `uw1 __MNOT (uw1)' `B = __MNOT (A)' `MNOT A,B'
30068 `uw1 __MOR (uw1, uw1)' `C = __MOR (A, B)' `MOR A,B,C'
30069 `uw1 __MPACKH (uh, uh)' `C = __MPACKH (A, B)' `MPACKH A,B,C'
30070 `sw2 __MQADDHSS (sw2, sw2)' `C = __MQADDHSS (A, B)' `MQADDHSS A,B,C'
30071 `uw2 __MQADDHUS (uw2, uw2)' `C = __MQADDHUS (A, B)' `MQADDHUS A,B,C'
30072 `void __MQCPXIS (acc, sw2, sw2)' `__MQCPXIS (C, A, B)' `MQCPXIS A,B,C'
30073 `void __MQCPXIU (acc, uw2, uw2)' `__MQCPXIU (C, A, B)' `MQCPXIU A,B,C'
30074 `void __MQCPXRS (acc, sw2, sw2)' `__MQCPXRS (C, A, B)' `MQCPXRS A,B,C'
30075 `void __MQCPXRU (acc, uw2, uw2)' `__MQCPXRU (C, A, B)' `MQCPXRU A,B,C'
30076 `sw2 __MQLCLRHS (sw2, sw2)' `C = __MQLCLRHS (A, B)' `MQLCLRHS A,B,C'
30077 `sw2 __MQLMTHS (sw2, sw2)' `C = __MQLMTHS (A, B)' `MQLMTHS A,B,C'
30078 `void __MQMACHS (acc, sw2, sw2)' `__MQMACHS (C, A, B)' `MQMACHS A,B,C'
30079 `void __MQMACHU (acc, uw2, uw2)' `__MQMACHU (C, A, B)' `MQMACHU A,B,C'
30080 `void __MQMACXHS (acc, sw2, `__MQMACXHS (C, A, B)' `MQMACXHS A,B,C'
30082 `void __MQMULHS (acc, sw2, sw2)' `__MQMULHS (C, A, B)' `MQMULHS A,B,C'
30083 `void __MQMULHU (acc, uw2, uw2)' `__MQMULHU (C, A, B)' `MQMULHU A,B,C'
30084 `void __MQMULXHS (acc, sw2, `__MQMULXHS (C, A, B)' `MQMULXHS A,B,C'
30086 `void __MQMULXHU (acc, uw2, `__MQMULXHU (C, A, B)' `MQMULXHU A,B,C'
30088 `sw2 __MQSATHS (sw2, sw2)' `C = __MQSATHS (A, B)' `MQSATHS A,B,C'
30089 `uw2 __MQSLLHI (uw2, int)' `C = __MQSLLHI (A, B)' `MQSLLHI A,B,C'
30090 `sw2 __MQSRAHI (sw2, int)' `C = __MQSRAHI (A, B)' `MQSRAHI A,B,C'
30091 `sw2 __MQSUBHSS (sw2, sw2)' `C = __MQSUBHSS (A, B)' `MQSUBHSS A,B,C'
30092 `uw2 __MQSUBHUS (uw2, uw2)' `C = __MQSUBHUS (A, B)' `MQSUBHUS A,B,C'
30093 `void __MQXMACHS (acc, sw2, `__MQXMACHS (C, A, B)' `MQXMACHS A,B,C'
30095 `void __MQXMACXHS (acc, sw2, `__MQXMACXHS (C, A, B)' `MQXMACXHS A,B,C'
30097 `uw1 __MRDACC (acc)' `B = __MRDACC (A)' `MRDACC A,B'
30098 `uw1 __MRDACCG (acc)' `B = __MRDACCG (A)' `MRDACCG A,B'
30099 `uw1 __MROTLI (uw1, const)' `C = __MROTLI (A, B)' `MROTLI A,#B,C'
30100 `uw1 __MROTRI (uw1, const)' `C = __MROTRI (A, B)' `MROTRI A,#B,C'
30101 `sw1 __MSATHS (sw1, sw1)' `C = __MSATHS (A, B)' `MSATHS A,B,C'
30102 `uw1 __MSATHU (uw1, uw1)' `C = __MSATHU (A, B)' `MSATHU A,B,C'
30103 `uw1 __MSLLHI (uw1, const)' `C = __MSLLHI (A, B)' `MSLLHI A,#B,C'
30104 `sw1 __MSRAHI (sw1, const)' `C = __MSRAHI (A, B)' `MSRAHI A,#B,C'
30105 `uw1 __MSRLHI (uw1, const)' `C = __MSRLHI (A, B)' `MSRLHI A,#B,C'
30106 `void __MSUBACCS (acc, acc)' `__MSUBACCS (B, A)' `MSUBACCS A,B'
30107 `sw1 __MSUBHSS (sw1, sw1)' `C = __MSUBHSS (A, B)' `MSUBHSS A,B,C'
30108 `uw1 __MSUBHUS (uw1, uw1)' `C = __MSUBHUS (A, B)' `MSUBHUS A,B,C'
30109 `void __MTRAP (void)' `__MTRAP ()' `MTRAP'
30110 `uw2 __MUNPACKH (uw1)' `B = __MUNPACKH (A)' `MUNPACKH A,B'
30111 `uw1 __MWCUT (uw2, uw1)' `C = __MWCUT (A, B)' `MWCUT A,B,C'
30112 `void __MWTACC (acc, uw1)' `__MWTACC (B, A)' `MWTACC A,B'
30113 `void __MWTACCG (acc, uw1)' `__MWTACCG (B, A)' `MWTACCG A,B'
30114 `uw1 __MXOR (uw1, uw1)' `C = __MXOR (A, B)' `MXOR A,B,C'
30117 File: gcc.info, Node: Raw read/write Functions, Next: Other Built-in Functions, Prev: Directly-mapped Media Functions, Up: FR-V Built-in Functions
30119 5.50.5.4 Raw read/write Functions
30120 .................................
30122 This sections describes built-in functions related to read and write
30123 instructions to access memory. These functions generate `membar'
30124 instructions to flush the I/O load and stores where appropriate, as
30125 described in Fujitsu's manual described above.
30127 `unsigned char __builtin_read8 (void *DATA)'
30129 `unsigned short __builtin_read16 (void *DATA)'
30131 `unsigned long __builtin_read32 (void *DATA)'
30133 `unsigned long long __builtin_read64 (void *DATA)'
30135 `void __builtin_write8 (void *DATA, unsigned char DATUM)'
30137 `void __builtin_write16 (void *DATA, unsigned short DATUM)'
30139 `void __builtin_write32 (void *DATA, unsigned long DATUM)'
30141 `void __builtin_write64 (void *DATA, unsigned long long DATUM)'
30144 File: gcc.info, Node: Other Built-in Functions, Prev: Raw read/write Functions, Up: FR-V Built-in Functions
30146 5.50.5.5 Other Built-in Functions
30147 .................................
30149 This section describes built-in functions that are not named after a
30150 specific FR-V instruction.
30152 `sw2 __IACCreadll (iacc REG)'
30153 Return the full 64-bit value of IACC0. The REG argument is
30154 reserved for future expansion and must be 0.
30156 `sw1 __IACCreadl (iacc REG)'
30157 Return the value of IACC0H if REG is 0 and IACC0L if REG is 1.
30158 Other values of REG are rejected as invalid.
30160 `void __IACCsetll (iacc REG, sw2 X)'
30161 Set the full 64-bit value of IACC0 to X. The REG argument is
30162 reserved for future expansion and must be 0.
30164 `void __IACCsetl (iacc REG, sw1 X)'
30165 Set IACC0H to X if REG is 0 and IACC0L to X if REG is 1. Other
30166 values of REG are rejected as invalid.
30168 `void __data_prefetch0 (const void *X)'
30169 Use the `dcpl' instruction to load the contents of address X into
30172 `void __data_prefetch (const void *X)'
30173 Use the `nldub' instruction to load the contents of address X into
30174 the data cache. The instruction will be issued in slot I1.
30177 File: gcc.info, Node: X86 Built-in Functions, Next: MIPS DSP Built-in Functions, Prev: FR-V Built-in Functions, Up: Target Builtins
30179 5.50.6 X86 Built-in Functions
30180 -----------------------------
30182 These built-in functions are available for the i386 and x86-64 family
30183 of computers, depending on the command-line switches used.
30185 Note that, if you specify command-line switches such as `-msse', the
30186 compiler could use the extended instruction sets even if the built-ins
30187 are not used explicitly in the program. For this reason, applications
30188 which perform runtime CPU detection must compile separate files for each
30189 supported architecture, using the appropriate flags. In particular,
30190 the file containing the CPU detection code should be compiled without
30193 The following machine modes are available for use with MMX built-in
30194 functions (*note Vector Extensions::): `V2SI' for a vector of two
30195 32-bit integers, `V4HI' for a vector of four 16-bit integers, and
30196 `V8QI' for a vector of eight 8-bit integers. Some of the built-in
30197 functions operate on MMX registers as a whole 64-bit entity, these use
30198 `V1DI' as their mode.
30200 If 3Dnow extensions are enabled, `V2SF' is used as a mode for a vector
30201 of two 32-bit floating point values.
30203 If SSE extensions are enabled, `V4SF' is used for a vector of four
30204 32-bit floating point values. Some instructions use a vector of four
30205 32-bit integers, these use `V4SI'. Finally, some instructions operate
30206 on an entire vector register, interpreting it as a 128-bit integer,
30207 these use mode `TI'.
30209 In 64-bit mode, the x86-64 family of processors uses additional
30210 built-in functions for efficient use of `TF' (`__float128') 128-bit
30211 floating point and `TC' 128-bit complex floating point values.
30213 The following floating point built-in functions are available in 64-bit
30214 mode. All of them implement the function that is part of the name.
30216 __float128 __builtin_fabsq (__float128)
30217 __float128 __builtin_copysignq (__float128, __float128)
30219 The following floating point built-in functions are made available in
30222 `__float128 __builtin_infq (void)'
30223 Similar to `__builtin_inf', except the return type is `__float128'.
30225 The following built-in functions are made available by `-mmmx'. All
30226 of them generate the machine instruction that is part of the name.
30228 v8qi __builtin_ia32_paddb (v8qi, v8qi)
30229 v4hi __builtin_ia32_paddw (v4hi, v4hi)
30230 v2si __builtin_ia32_paddd (v2si, v2si)
30231 v8qi __builtin_ia32_psubb (v8qi, v8qi)
30232 v4hi __builtin_ia32_psubw (v4hi, v4hi)
30233 v2si __builtin_ia32_psubd (v2si, v2si)
30234 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
30235 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
30236 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
30237 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
30238 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
30239 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
30240 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
30241 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
30242 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
30243 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
30244 di __builtin_ia32_pand (di, di)
30245 di __builtin_ia32_pandn (di,di)
30246 di __builtin_ia32_por (di, di)
30247 di __builtin_ia32_pxor (di, di)
30248 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
30249 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
30250 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
30251 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
30252 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
30253 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
30254 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
30255 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
30256 v2si __builtin_ia32_punpckhdq (v2si, v2si)
30257 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
30258 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
30259 v2si __builtin_ia32_punpckldq (v2si, v2si)
30260 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
30261 v4hi __builtin_ia32_packssdw (v2si, v2si)
30262 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
30264 v4hi __builtin_ia32_psllw (v4hi, v4hi)
30265 v2si __builtin_ia32_pslld (v2si, v2si)
30266 v1di __builtin_ia32_psllq (v1di, v1di)
30267 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
30268 v2si __builtin_ia32_psrld (v2si, v2si)
30269 v1di __builtin_ia32_psrlq (v1di, v1di)
30270 v4hi __builtin_ia32_psraw (v4hi, v4hi)
30271 v2si __builtin_ia32_psrad (v2si, v2si)
30272 v4hi __builtin_ia32_psllwi (v4hi, int)
30273 v2si __builtin_ia32_pslldi (v2si, int)
30274 v1di __builtin_ia32_psllqi (v1di, int)
30275 v4hi __builtin_ia32_psrlwi (v4hi, int)
30276 v2si __builtin_ia32_psrldi (v2si, int)
30277 v1di __builtin_ia32_psrlqi (v1di, int)
30278 v4hi __builtin_ia32_psrawi (v4hi, int)
30279 v2si __builtin_ia32_psradi (v2si, int)
30281 The following built-in functions are made available either with
30282 `-msse', or with a combination of `-m3dnow' and `-march=athlon'. All
30283 of them generate the machine instruction that is part of the name.
30285 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
30286 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
30287 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
30288 v1di __builtin_ia32_psadbw (v8qi, v8qi)
30289 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
30290 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
30291 v8qi __builtin_ia32_pminub (v8qi, v8qi)
30292 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
30293 int __builtin_ia32_pextrw (v4hi, int)
30294 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
30295 int __builtin_ia32_pmovmskb (v8qi)
30296 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
30297 void __builtin_ia32_movntq (di *, di)
30298 void __builtin_ia32_sfence (void)
30300 The following built-in functions are available when `-msse' is used.
30301 All of them generate the machine instruction that is part of the name.
30303 int __builtin_ia32_comieq (v4sf, v4sf)
30304 int __builtin_ia32_comineq (v4sf, v4sf)
30305 int __builtin_ia32_comilt (v4sf, v4sf)
30306 int __builtin_ia32_comile (v4sf, v4sf)
30307 int __builtin_ia32_comigt (v4sf, v4sf)
30308 int __builtin_ia32_comige (v4sf, v4sf)
30309 int __builtin_ia32_ucomieq (v4sf, v4sf)
30310 int __builtin_ia32_ucomineq (v4sf, v4sf)
30311 int __builtin_ia32_ucomilt (v4sf, v4sf)
30312 int __builtin_ia32_ucomile (v4sf, v4sf)
30313 int __builtin_ia32_ucomigt (v4sf, v4sf)
30314 int __builtin_ia32_ucomige (v4sf, v4sf)
30315 v4sf __builtin_ia32_addps (v4sf, v4sf)
30316 v4sf __builtin_ia32_subps (v4sf, v4sf)
30317 v4sf __builtin_ia32_mulps (v4sf, v4sf)
30318 v4sf __builtin_ia32_divps (v4sf, v4sf)
30319 v4sf __builtin_ia32_addss (v4sf, v4sf)
30320 v4sf __builtin_ia32_subss (v4sf, v4sf)
30321 v4sf __builtin_ia32_mulss (v4sf, v4sf)
30322 v4sf __builtin_ia32_divss (v4sf, v4sf)
30323 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
30324 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
30325 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
30326 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
30327 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
30328 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
30329 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
30330 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
30331 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
30332 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
30333 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
30334 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
30335 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
30336 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
30337 v4si __builtin_ia32_cmpless (v4sf, v4sf)
30338 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
30339 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
30340 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
30341 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
30342 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
30343 v4sf __builtin_ia32_maxps (v4sf, v4sf)
30344 v4sf __builtin_ia32_maxss (v4sf, v4sf)
30345 v4sf __builtin_ia32_minps (v4sf, v4sf)
30346 v4sf __builtin_ia32_minss (v4sf, v4sf)
30347 v4sf __builtin_ia32_andps (v4sf, v4sf)
30348 v4sf __builtin_ia32_andnps (v4sf, v4sf)
30349 v4sf __builtin_ia32_orps (v4sf, v4sf)
30350 v4sf __builtin_ia32_xorps (v4sf, v4sf)
30351 v4sf __builtin_ia32_movss (v4sf, v4sf)
30352 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
30353 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
30354 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
30355 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
30356 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
30357 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
30358 v2si __builtin_ia32_cvtps2pi (v4sf)
30359 int __builtin_ia32_cvtss2si (v4sf)
30360 v2si __builtin_ia32_cvttps2pi (v4sf)
30361 int __builtin_ia32_cvttss2si (v4sf)
30362 v4sf __builtin_ia32_rcpps (v4sf)
30363 v4sf __builtin_ia32_rsqrtps (v4sf)
30364 v4sf __builtin_ia32_sqrtps (v4sf)
30365 v4sf __builtin_ia32_rcpss (v4sf)
30366 v4sf __builtin_ia32_rsqrtss (v4sf)
30367 v4sf __builtin_ia32_sqrtss (v4sf)
30368 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
30369 void __builtin_ia32_movntps (float *, v4sf)
30370 int __builtin_ia32_movmskps (v4sf)
30372 The following built-in functions are available when `-msse' is used.
30374 `v4sf __builtin_ia32_loadaps (float *)'
30375 Generates the `movaps' machine instruction as a load from memory.
30377 `void __builtin_ia32_storeaps (float *, v4sf)'
30378 Generates the `movaps' machine instruction as a store to memory.
30380 `v4sf __builtin_ia32_loadups (float *)'
30381 Generates the `movups' machine instruction as a load from memory.
30383 `void __builtin_ia32_storeups (float *, v4sf)'
30384 Generates the `movups' machine instruction as a store to memory.
30386 `v4sf __builtin_ia32_loadsss (float *)'
30387 Generates the `movss' machine instruction as a load from memory.
30389 `void __builtin_ia32_storess (float *, v4sf)'
30390 Generates the `movss' machine instruction as a store to memory.
30392 `v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)'
30393 Generates the `movhps' machine instruction as a load from memory.
30395 `v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)'
30396 Generates the `movlps' machine instruction as a load from memory
30398 `void __builtin_ia32_storehps (v2sf *, v4sf)'
30399 Generates the `movhps' machine instruction as a store to memory.
30401 `void __builtin_ia32_storelps (v2sf *, v4sf)'
30402 Generates the `movlps' machine instruction as a store to memory.
30404 The following built-in functions are available when `-msse2' is used.
30405 All of them generate the machine instruction that is part of the name.
30407 int __builtin_ia32_comisdeq (v2df, v2df)
30408 int __builtin_ia32_comisdlt (v2df, v2df)
30409 int __builtin_ia32_comisdle (v2df, v2df)
30410 int __builtin_ia32_comisdgt (v2df, v2df)
30411 int __builtin_ia32_comisdge (v2df, v2df)
30412 int __builtin_ia32_comisdneq (v2df, v2df)
30413 int __builtin_ia32_ucomisdeq (v2df, v2df)
30414 int __builtin_ia32_ucomisdlt (v2df, v2df)
30415 int __builtin_ia32_ucomisdle (v2df, v2df)
30416 int __builtin_ia32_ucomisdgt (v2df, v2df)
30417 int __builtin_ia32_ucomisdge (v2df, v2df)
30418 int __builtin_ia32_ucomisdneq (v2df, v2df)
30419 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
30420 v2df __builtin_ia32_cmpltpd (v2df, v2df)
30421 v2df __builtin_ia32_cmplepd (v2df, v2df)
30422 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
30423 v2df __builtin_ia32_cmpgepd (v2df, v2df)
30424 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
30425 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
30426 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
30427 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
30428 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
30429 v2df __builtin_ia32_cmpngepd (v2df, v2df)
30430 v2df __builtin_ia32_cmpordpd (v2df, v2df)
30431 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
30432 v2df __builtin_ia32_cmpltsd (v2df, v2df)
30433 v2df __builtin_ia32_cmplesd (v2df, v2df)
30434 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
30435 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
30436 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
30437 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
30438 v2df __builtin_ia32_cmpordsd (v2df, v2df)
30439 v2di __builtin_ia32_paddq (v2di, v2di)
30440 v2di __builtin_ia32_psubq (v2di, v2di)
30441 v2df __builtin_ia32_addpd (v2df, v2df)
30442 v2df __builtin_ia32_subpd (v2df, v2df)
30443 v2df __builtin_ia32_mulpd (v2df, v2df)
30444 v2df __builtin_ia32_divpd (v2df, v2df)
30445 v2df __builtin_ia32_addsd (v2df, v2df)
30446 v2df __builtin_ia32_subsd (v2df, v2df)
30447 v2df __builtin_ia32_mulsd (v2df, v2df)
30448 v2df __builtin_ia32_divsd (v2df, v2df)
30449 v2df __builtin_ia32_minpd (v2df, v2df)
30450 v2df __builtin_ia32_maxpd (v2df, v2df)
30451 v2df __builtin_ia32_minsd (v2df, v2df)
30452 v2df __builtin_ia32_maxsd (v2df, v2df)
30453 v2df __builtin_ia32_andpd (v2df, v2df)
30454 v2df __builtin_ia32_andnpd (v2df, v2df)
30455 v2df __builtin_ia32_orpd (v2df, v2df)
30456 v2df __builtin_ia32_xorpd (v2df, v2df)
30457 v2df __builtin_ia32_movsd (v2df, v2df)
30458 v2df __builtin_ia32_unpckhpd (v2df, v2df)
30459 v2df __builtin_ia32_unpcklpd (v2df, v2df)
30460 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
30461 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
30462 v4si __builtin_ia32_paddd128 (v4si, v4si)
30463 v2di __builtin_ia32_paddq128 (v2di, v2di)
30464 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
30465 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
30466 v4si __builtin_ia32_psubd128 (v4si, v4si)
30467 v2di __builtin_ia32_psubq128 (v2di, v2di)
30468 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
30469 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
30470 v2di __builtin_ia32_pand128 (v2di, v2di)
30471 v2di __builtin_ia32_pandn128 (v2di, v2di)
30472 v2di __builtin_ia32_por128 (v2di, v2di)
30473 v2di __builtin_ia32_pxor128 (v2di, v2di)
30474 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
30475 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
30476 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
30477 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
30478 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
30479 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
30480 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
30481 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
30482 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
30483 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
30484 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
30485 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
30486 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
30487 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
30488 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
30489 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
30490 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
30491 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
30492 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
30493 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
30494 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
30495 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
30496 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
30497 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
30498 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
30499 v2df __builtin_ia32_loadupd (double *)
30500 void __builtin_ia32_storeupd (double *, v2df)
30501 v2df __builtin_ia32_loadhpd (v2df, double const *)
30502 v2df __builtin_ia32_loadlpd (v2df, double const *)
30503 int __builtin_ia32_movmskpd (v2df)
30504 int __builtin_ia32_pmovmskb128 (v16qi)
30505 void __builtin_ia32_movnti (int *, int)
30506 void __builtin_ia32_movntpd (double *, v2df)
30507 void __builtin_ia32_movntdq (v2df *, v2df)
30508 v4si __builtin_ia32_pshufd (v4si, int)
30509 v8hi __builtin_ia32_pshuflw (v8hi, int)
30510 v8hi __builtin_ia32_pshufhw (v8hi, int)
30511 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
30512 v2df __builtin_ia32_sqrtpd (v2df)
30513 v2df __builtin_ia32_sqrtsd (v2df)
30514 v2df __builtin_ia32_shufpd (v2df, v2df, int)
30515 v2df __builtin_ia32_cvtdq2pd (v4si)
30516 v4sf __builtin_ia32_cvtdq2ps (v4si)
30517 v4si __builtin_ia32_cvtpd2dq (v2df)
30518 v2si __builtin_ia32_cvtpd2pi (v2df)
30519 v4sf __builtin_ia32_cvtpd2ps (v2df)
30520 v4si __builtin_ia32_cvttpd2dq (v2df)
30521 v2si __builtin_ia32_cvttpd2pi (v2df)
30522 v2df __builtin_ia32_cvtpi2pd (v2si)
30523 int __builtin_ia32_cvtsd2si (v2df)
30524 int __builtin_ia32_cvttsd2si (v2df)
30525 long long __builtin_ia32_cvtsd2si64 (v2df)
30526 long long __builtin_ia32_cvttsd2si64 (v2df)
30527 v4si __builtin_ia32_cvtps2dq (v4sf)
30528 v2df __builtin_ia32_cvtps2pd (v4sf)
30529 v4si __builtin_ia32_cvttps2dq (v4sf)
30530 v2df __builtin_ia32_cvtsi2sd (v2df, int)
30531 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
30532 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
30533 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
30534 void __builtin_ia32_clflush (const void *)
30535 void __builtin_ia32_lfence (void)
30536 void __builtin_ia32_mfence (void)
30537 v16qi __builtin_ia32_loaddqu (const char *)
30538 void __builtin_ia32_storedqu (char *, v16qi)
30539 v1di __builtin_ia32_pmuludq (v2si, v2si)
30540 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
30541 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
30542 v4si __builtin_ia32_pslld128 (v4si, v4si)
30543 v2di __builtin_ia32_psllq128 (v2di, v2di)
30544 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
30545 v4si __builtin_ia32_psrld128 (v4si, v4si)
30546 v2di __builtin_ia32_psrlq128 (v2di, v2di)
30547 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
30548 v4si __builtin_ia32_psrad128 (v4si, v4si)
30549 v2di __builtin_ia32_pslldqi128 (v2di, int)
30550 v8hi __builtin_ia32_psllwi128 (v8hi, int)
30551 v4si __builtin_ia32_pslldi128 (v4si, int)
30552 v2di __builtin_ia32_psllqi128 (v2di, int)
30553 v2di __builtin_ia32_psrldqi128 (v2di, int)
30554 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
30555 v4si __builtin_ia32_psrldi128 (v4si, int)
30556 v2di __builtin_ia32_psrlqi128 (v2di, int)
30557 v8hi __builtin_ia32_psrawi128 (v8hi, int)
30558 v4si __builtin_ia32_psradi128 (v4si, int)
30559 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
30560 v2di __builtin_ia32_movq128 (v2di)
30562 The following built-in functions are available when `-msse3' is used.
30563 All of them generate the machine instruction that is part of the name.
30565 v2df __builtin_ia32_addsubpd (v2df, v2df)
30566 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
30567 v2df __builtin_ia32_haddpd (v2df, v2df)
30568 v4sf __builtin_ia32_haddps (v4sf, v4sf)
30569 v2df __builtin_ia32_hsubpd (v2df, v2df)
30570 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
30571 v16qi __builtin_ia32_lddqu (char const *)
30572 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
30573 v2df __builtin_ia32_movddup (v2df)
30574 v4sf __builtin_ia32_movshdup (v4sf)
30575 v4sf __builtin_ia32_movsldup (v4sf)
30576 void __builtin_ia32_mwait (unsigned int, unsigned int)
30578 The following built-in functions are available when `-msse3' is used.
30580 `v2df __builtin_ia32_loadddup (double const *)'
30581 Generates the `movddup' machine instruction as a load from memory.
30583 The following built-in functions are available when `-mssse3' is used.
30584 All of them generate the machine instruction that is part of the name
30585 with MMX registers.
30587 v2si __builtin_ia32_phaddd (v2si, v2si)
30588 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
30589 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
30590 v2si __builtin_ia32_phsubd (v2si, v2si)
30591 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
30592 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
30593 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
30594 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
30595 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
30596 v8qi __builtin_ia32_psignb (v8qi, v8qi)
30597 v2si __builtin_ia32_psignd (v2si, v2si)
30598 v4hi __builtin_ia32_psignw (v4hi, v4hi)
30599 v1di __builtin_ia32_palignr (v1di, v1di, int)
30600 v8qi __builtin_ia32_pabsb (v8qi)
30601 v2si __builtin_ia32_pabsd (v2si)
30602 v4hi __builtin_ia32_pabsw (v4hi)
30604 The following built-in functions are available when `-mssse3' is used.
30605 All of them generate the machine instruction that is part of the name
30606 with SSE registers.
30608 v4si __builtin_ia32_phaddd128 (v4si, v4si)
30609 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
30610 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
30611 v4si __builtin_ia32_phsubd128 (v4si, v4si)
30612 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
30613 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
30614 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
30615 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
30616 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
30617 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
30618 v4si __builtin_ia32_psignd128 (v4si, v4si)
30619 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
30620 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
30621 v16qi __builtin_ia32_pabsb128 (v16qi)
30622 v4si __builtin_ia32_pabsd128 (v4si)
30623 v8hi __builtin_ia32_pabsw128 (v8hi)
30625 The following built-in functions are available when `-msse4.1' is
30626 used. All of them generate the machine instruction that is part of the
30629 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
30630 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
30631 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
30632 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
30633 v2df __builtin_ia32_dppd (v2df, v2df, const int)
30634 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
30635 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
30636 v2di __builtin_ia32_movntdqa (v2di *);
30637 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
30638 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
30639 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
30640 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
30641 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
30642 v8hi __builtin_ia32_phminposuw128 (v8hi)
30643 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
30644 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
30645 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
30646 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
30647 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
30648 v4si __builtin_ia32_pminsd128 (v4si, v4si)
30649 v4si __builtin_ia32_pminud128 (v4si, v4si)
30650 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
30651 v4si __builtin_ia32_pmovsxbd128 (v16qi)
30652 v2di __builtin_ia32_pmovsxbq128 (v16qi)
30653 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
30654 v2di __builtin_ia32_pmovsxdq128 (v4si)
30655 v4si __builtin_ia32_pmovsxwd128 (v8hi)
30656 v2di __builtin_ia32_pmovsxwq128 (v8hi)
30657 v4si __builtin_ia32_pmovzxbd128 (v16qi)
30658 v2di __builtin_ia32_pmovzxbq128 (v16qi)
30659 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
30660 v2di __builtin_ia32_pmovzxdq128 (v4si)
30661 v4si __builtin_ia32_pmovzxwd128 (v8hi)
30662 v2di __builtin_ia32_pmovzxwq128 (v8hi)
30663 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
30664 v4si __builtin_ia32_pmulld128 (v4si, v4si)
30665 int __builtin_ia32_ptestc128 (v2di, v2di)
30666 int __builtin_ia32_ptestnzc128 (v2di, v2di)
30667 int __builtin_ia32_ptestz128 (v2di, v2di)
30668 v2df __builtin_ia32_roundpd (v2df, const int)
30669 v4sf __builtin_ia32_roundps (v4sf, const int)
30670 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
30671 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
30673 The following built-in functions are available when `-msse4.1' is used.
30675 `v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)'
30676 Generates the `insertps' machine instruction.
30678 `int __builtin_ia32_vec_ext_v16qi (v16qi, const int)'
30679 Generates the `pextrb' machine instruction.
30681 `v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)'
30682 Generates the `pinsrb' machine instruction.
30684 `v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)'
30685 Generates the `pinsrd' machine instruction.
30687 `v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)'
30688 Generates the `pinsrq' machine instruction in 64bit mode.
30690 The following built-in functions are changed to generate new SSE4.1
30691 instructions when `-msse4.1' is used.
30693 `float __builtin_ia32_vec_ext_v4sf (v4sf, const int)'
30694 Generates the `extractps' machine instruction.
30696 `int __builtin_ia32_vec_ext_v4si (v4si, const int)'
30697 Generates the `pextrd' machine instruction.
30699 `long long __builtin_ia32_vec_ext_v2di (v2di, const int)'
30700 Generates the `pextrq' machine instruction in 64bit mode.
30702 The following built-in functions are available when `-msse4.2' is
30703 used. All of them generate the machine instruction that is part of the
30706 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
30707 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
30708 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
30709 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
30710 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
30711 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
30712 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
30713 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
30714 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
30715 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
30716 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
30717 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
30718 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
30719 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
30720 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
30722 The following built-in functions are available when `-msse4.2' is used.
30724 `unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)'
30725 Generates the `crc32b' machine instruction.
30727 `unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)'
30728 Generates the `crc32w' machine instruction.
30730 `unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)'
30731 Generates the `crc32l' machine instruction.
30733 `unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)'
30735 The following built-in functions are changed to generate new SSE4.2
30736 instructions when `-msse4.2' is used.
30738 `int __builtin_popcount (unsigned int)'
30739 Generates the `popcntl' machine instruction.
30741 `int __builtin_popcountl (unsigned long)'
30742 Generates the `popcntl' or `popcntq' machine instruction,
30743 depending on the size of `unsigned long'.
30745 `int __builtin_popcountll (unsigned long long)'
30746 Generates the `popcntq' machine instruction.
30748 The following built-in functions are available when `-mavx' is used.
30749 All of them generate the machine instruction that is part of the name.
30751 v4df __builtin_ia32_addpd256 (v4df,v4df)
30752 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
30753 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
30754 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
30755 v4df __builtin_ia32_andnpd256 (v4df,v4df)
30756 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
30757 v4df __builtin_ia32_andpd256 (v4df,v4df)
30758 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
30759 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
30760 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
30761 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
30762 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
30763 v2df __builtin_ia32_cmppd (v2df,v2df,int)
30764 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
30765 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
30766 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
30767 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
30768 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
30769 v4df __builtin_ia32_cvtdq2pd256 (v4si)
30770 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
30771 v4si __builtin_ia32_cvtpd2dq256 (v4df)
30772 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
30773 v8si __builtin_ia32_cvtps2dq256 (v8sf)
30774 v4df __builtin_ia32_cvtps2pd256 (v4sf)
30775 v4si __builtin_ia32_cvttpd2dq256 (v4df)
30776 v8si __builtin_ia32_cvttps2dq256 (v8sf)
30777 v4df __builtin_ia32_divpd256 (v4df,v4df)
30778 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
30779 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
30780 v4df __builtin_ia32_haddpd256 (v4df,v4df)
30781 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
30782 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
30783 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
30784 v32qi __builtin_ia32_lddqu256 (pcchar)
30785 v32qi __builtin_ia32_loaddqu256 (pcchar)
30786 v4df __builtin_ia32_loadupd256 (pcdouble)
30787 v8sf __builtin_ia32_loadups256 (pcfloat)
30788 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
30789 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
30790 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
30791 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
30792 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
30793 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
30794 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
30795 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
30796 v4df __builtin_ia32_maxpd256 (v4df,v4df)
30797 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
30798 v4df __builtin_ia32_minpd256 (v4df,v4df)
30799 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
30800 v4df __builtin_ia32_movddup256 (v4df)
30801 int __builtin_ia32_movmskpd256 (v4df)
30802 int __builtin_ia32_movmskps256 (v8sf)
30803 v8sf __builtin_ia32_movshdup256 (v8sf)
30804 v8sf __builtin_ia32_movsldup256 (v8sf)
30805 v4df __builtin_ia32_mulpd256 (v4df,v4df)
30806 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
30807 v4df __builtin_ia32_orpd256 (v4df,v4df)
30808 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
30809 v2df __builtin_ia32_pd_pd256 (v4df)
30810 v4df __builtin_ia32_pd256_pd (v2df)
30811 v4sf __builtin_ia32_ps_ps256 (v8sf)
30812 v8sf __builtin_ia32_ps256_ps (v4sf)
30813 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
30814 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
30815 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
30816 v8sf __builtin_ia32_rcpps256 (v8sf)
30817 v4df __builtin_ia32_roundpd256 (v4df,int)
30818 v8sf __builtin_ia32_roundps256 (v8sf,int)
30819 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
30820 v8sf __builtin_ia32_rsqrtps256 (v8sf)
30821 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
30822 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
30823 v4si __builtin_ia32_si_si256 (v8si)
30824 v8si __builtin_ia32_si256_si (v4si)
30825 v4df __builtin_ia32_sqrtpd256 (v4df)
30826 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
30827 v8sf __builtin_ia32_sqrtps256 (v8sf)
30828 void __builtin_ia32_storedqu256 (pchar,v32qi)
30829 void __builtin_ia32_storeupd256 (pdouble,v4df)
30830 void __builtin_ia32_storeups256 (pfloat,v8sf)
30831 v4df __builtin_ia32_subpd256 (v4df,v4df)
30832 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
30833 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
30834 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
30835 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
30836 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
30837 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
30838 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
30839 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
30840 v4sf __builtin_ia32_vbroadcastss (pcfloat)
30841 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
30842 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
30843 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
30844 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
30845 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
30846 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
30847 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
30848 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
30849 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
30850 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
30851 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
30852 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
30853 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
30854 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
30855 v2df __builtin_ia32_vpermilpd (v2df,int)
30856 v4df __builtin_ia32_vpermilpd256 (v4df,int)
30857 v4sf __builtin_ia32_vpermilps (v4sf,int)
30858 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
30859 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
30860 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
30861 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
30862 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
30863 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
30864 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
30865 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
30866 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
30867 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
30868 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
30869 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
30870 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
30871 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
30872 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
30873 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
30874 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
30875 void __builtin_ia32_vzeroall (void)
30876 void __builtin_ia32_vzeroupper (void)
30877 v4df __builtin_ia32_xorpd256 (v4df,v4df)
30878 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
30880 The following built-in functions are available when `-maes' is used.
30881 All of them generate the machine instruction that is part of the name.
30883 v2di __builtin_ia32_aesenc128 (v2di, v2di)
30884 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
30885 v2di __builtin_ia32_aesdec128 (v2di, v2di)
30886 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
30887 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
30888 v2di __builtin_ia32_aesimc128 (v2di)
30890 The following built-in function is available when `-mpclmul' is used.
30892 `v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)'
30893 Generates the `pclmulqdq' machine instruction.
30895 The following built-in functions are available when `-msse4a' is used.
30896 All of them generate the machine instruction that is part of the name.
30898 void __builtin_ia32_movntsd (double *, v2df)
30899 void __builtin_ia32_movntss (float *, v4sf)
30900 v2di __builtin_ia32_extrq (v2di, v16qi)
30901 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
30902 v2di __builtin_ia32_insertq (v2di, v2di)
30903 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
30905 The following built-in functions are available when `-msse5' is used.
30906 All of them generate the machine instruction that is part of the name
30907 with MMX registers.
30909 v2df __builtin_ia32_comeqpd (v2df, v2df)
30910 v2df __builtin_ia32_comeqps (v2df, v2df)
30911 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
30912 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
30913 v2df __builtin_ia32_comfalsepd (v2df, v2df)
30914 v2df __builtin_ia32_comfalseps (v2df, v2df)
30915 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
30916 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
30917 v2df __builtin_ia32_comgepd (v2df, v2df)
30918 v2df __builtin_ia32_comgeps (v2df, v2df)
30919 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
30920 v4sf __builtin_ia32_comgess (v4sf, v4sf)
30921 v2df __builtin_ia32_comgtpd (v2df, v2df)
30922 v2df __builtin_ia32_comgtps (v2df, v2df)
30923 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
30924 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
30925 v2df __builtin_ia32_comlepd (v2df, v2df)
30926 v2df __builtin_ia32_comleps (v2df, v2df)
30927 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
30928 v4sf __builtin_ia32_comless (v4sf, v4sf)
30929 v2df __builtin_ia32_comltpd (v2df, v2df)
30930 v2df __builtin_ia32_comltps (v2df, v2df)
30931 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
30932 v4sf __builtin_ia32_comltss (v4sf, v4sf)
30933 v2df __builtin_ia32_comnepd (v2df, v2df)
30934 v2df __builtin_ia32_comneps (v2df, v2df)
30935 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
30936 v4sf __builtin_ia32_comness (v4sf, v4sf)
30937 v2df __builtin_ia32_comordpd (v2df, v2df)
30938 v2df __builtin_ia32_comordps (v2df, v2df)
30939 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
30940 v4sf __builtin_ia32_comordss (v4sf, v4sf)
30941 v2df __builtin_ia32_comtruepd (v2df, v2df)
30942 v2df __builtin_ia32_comtrueps (v2df, v2df)
30943 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
30944 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
30945 v2df __builtin_ia32_comueqpd (v2df, v2df)
30946 v2df __builtin_ia32_comueqps (v2df, v2df)
30947 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
30948 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
30949 v2df __builtin_ia32_comugepd (v2df, v2df)
30950 v2df __builtin_ia32_comugeps (v2df, v2df)
30951 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
30952 v4sf __builtin_ia32_comugess (v4sf, v4sf)
30953 v2df __builtin_ia32_comugtpd (v2df, v2df)
30954 v2df __builtin_ia32_comugtps (v2df, v2df)
30955 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
30956 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
30957 v2df __builtin_ia32_comulepd (v2df, v2df)
30958 v2df __builtin_ia32_comuleps (v2df, v2df)
30959 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
30960 v4sf __builtin_ia32_comuless (v4sf, v4sf)
30961 v2df __builtin_ia32_comultpd (v2df, v2df)
30962 v2df __builtin_ia32_comultps (v2df, v2df)
30963 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
30964 v4sf __builtin_ia32_comultss (v4sf, v4sf)
30965 v2df __builtin_ia32_comunepd (v2df, v2df)
30966 v2df __builtin_ia32_comuneps (v2df, v2df)
30967 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
30968 v4sf __builtin_ia32_comuness (v4sf, v4sf)
30969 v2df __builtin_ia32_comunordpd (v2df, v2df)
30970 v2df __builtin_ia32_comunordps (v2df, v2df)
30971 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
30972 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
30973 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
30974 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
30975 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
30976 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
30977 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
30978 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
30979 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
30980 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
30981 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
30982 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
30983 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
30984 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
30985 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
30986 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
30987 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
30988 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
30989 v2df __builtin_ia32_frczpd (v2df)
30990 v4sf __builtin_ia32_frczps (v4sf)
30991 v2df __builtin_ia32_frczsd (v2df, v2df)
30992 v4sf __builtin_ia32_frczss (v4sf, v4sf)
30993 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
30994 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
30995 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
30996 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
30997 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
30998 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
30999 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
31000 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
31001 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
31002 v4si __builtin_ia32_pcomeqd (v4si, v4si)
31003 v2di __builtin_ia32_pcomeqq (v2di, v2di)
31004 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
31005 v4si __builtin_ia32_pcomequd (v4si, v4si)
31006 v2di __builtin_ia32_pcomequq (v2di, v2di)
31007 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
31008 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
31009 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
31010 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
31011 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
31012 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
31013 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
31014 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
31015 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
31016 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
31017 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
31018 v4si __builtin_ia32_pcomged (v4si, v4si)
31019 v2di __builtin_ia32_pcomgeq (v2di, v2di)
31020 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
31021 v4si __builtin_ia32_pcomgeud (v4si, v4si)
31022 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
31023 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
31024 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
31025 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
31026 v4si __builtin_ia32_pcomgtd (v4si, v4si)
31027 v2di __builtin_ia32_pcomgtq (v2di, v2di)
31028 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
31029 v4si __builtin_ia32_pcomgtud (v4si, v4si)
31030 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
31031 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
31032 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
31033 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
31034 v4si __builtin_ia32_pcomled (v4si, v4si)
31035 v2di __builtin_ia32_pcomleq (v2di, v2di)
31036 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
31037 v4si __builtin_ia32_pcomleud (v4si, v4si)
31038 v2di __builtin_ia32_pcomleuq (v2di, v2di)
31039 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
31040 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
31041 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
31042 v4si __builtin_ia32_pcomltd (v4si, v4si)
31043 v2di __builtin_ia32_pcomltq (v2di, v2di)
31044 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
31045 v4si __builtin_ia32_pcomltud (v4si, v4si)
31046 v2di __builtin_ia32_pcomltuq (v2di, v2di)
31047 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
31048 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
31049 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
31050 v4si __builtin_ia32_pcomned (v4si, v4si)
31051 v2di __builtin_ia32_pcomneq (v2di, v2di)
31052 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
31053 v4si __builtin_ia32_pcomneud (v4si, v4si)
31054 v2di __builtin_ia32_pcomneuq (v2di, v2di)
31055 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
31056 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
31057 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
31058 v4si __builtin_ia32_pcomtrued (v4si, v4si)
31059 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
31060 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
31061 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
31062 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
31063 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
31064 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
31065 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
31066 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
31067 v4si __builtin_ia32_phaddbd (v16qi)
31068 v2di __builtin_ia32_phaddbq (v16qi)
31069 v8hi __builtin_ia32_phaddbw (v16qi)
31070 v2di __builtin_ia32_phadddq (v4si)
31071 v4si __builtin_ia32_phaddubd (v16qi)
31072 v2di __builtin_ia32_phaddubq (v16qi)
31073 v8hi __builtin_ia32_phaddubw (v16qi)
31074 v2di __builtin_ia32_phaddudq (v4si)
31075 v4si __builtin_ia32_phadduwd (v8hi)
31076 v2di __builtin_ia32_phadduwq (v8hi)
31077 v4si __builtin_ia32_phaddwd (v8hi)
31078 v2di __builtin_ia32_phaddwq (v8hi)
31079 v8hi __builtin_ia32_phsubbw (v16qi)
31080 v2di __builtin_ia32_phsubdq (v4si)
31081 v4si __builtin_ia32_phsubwd (v8hi)
31082 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
31083 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
31084 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
31085 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
31086 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
31087 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
31088 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
31089 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
31090 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
31091 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
31092 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
31093 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
31094 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
31095 v16qi __builtin_ia32_protb (v16qi, v16qi)
31096 v4si __builtin_ia32_protd (v4si, v4si)
31097 v2di __builtin_ia32_protq (v2di, v2di)
31098 v8hi __builtin_ia32_protw (v8hi, v8hi)
31099 v16qi __builtin_ia32_pshab (v16qi, v16qi)
31100 v4si __builtin_ia32_pshad (v4si, v4si)
31101 v2di __builtin_ia32_pshaq (v2di, v2di)
31102 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
31103 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
31104 v4si __builtin_ia32_pshld (v4si, v4si)
31105 v2di __builtin_ia32_pshlq (v2di, v2di)
31106 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
31108 The following builtin-in functions are available when `-msse5' is
31109 used. The second argument must be an integer constant and generate the
31110 machine instruction that is part of the name with the `_imm' suffix
31113 v16qi __builtin_ia32_protb_imm (v16qi, int)
31114 v4si __builtin_ia32_protd_imm (v4si, int)
31115 v2di __builtin_ia32_protq_imm (v2di, int)
31116 v8hi __builtin_ia32_protw_imm (v8hi, int)
31118 The following built-in functions are available when `-m3dnow' is used.
31119 All of them generate the machine instruction that is part of the name.
31121 void __builtin_ia32_femms (void)
31122 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
31123 v2si __builtin_ia32_pf2id (v2sf)
31124 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
31125 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
31126 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
31127 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
31128 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
31129 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
31130 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
31131 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
31132 v2sf __builtin_ia32_pfrcp (v2sf)
31133 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
31134 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
31135 v2sf __builtin_ia32_pfrsqrt (v2sf)
31136 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
31137 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
31138 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
31139 v2sf __builtin_ia32_pi2fd (v2si)
31140 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
31142 The following built-in functions are available when both `-m3dnow' and
31143 `-march=athlon' are used. All of them generate the machine instruction
31144 that is part of the name.
31146 v2si __builtin_ia32_pf2iw (v2sf)
31147 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
31148 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
31149 v2sf __builtin_ia32_pi2fw (v2si)
31150 v2sf __builtin_ia32_pswapdsf (v2sf)
31151 v2si __builtin_ia32_pswapdsi (v2si)
31154 File: gcc.info, Node: MIPS DSP Built-in Functions, Next: MIPS Paired-Single Support, Prev: X86 Built-in Functions, Up: Target Builtins
31156 5.50.7 MIPS DSP Built-in Functions
31157 ----------------------------------
31159 The MIPS DSP Application-Specific Extension (ASE) includes new
31160 instructions that are designed to improve the performance of DSP and
31161 media applications. It provides instructions that operate on packed
31162 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
31164 GCC supports MIPS DSP operations using both the generic vector
31165 extensions (*note Vector Extensions::) and a collection of
31166 MIPS-specific built-in functions. Both kinds of support are enabled by
31167 the `-mdsp' command-line option.
31169 Revision 2 of the ASE was introduced in the second half of 2006. This
31170 revision adds extra instructions to the original ASE, but is otherwise
31171 backwards-compatible with it. You can select revision 2 using the
31172 command-line option `-mdspr2'; this option implies `-mdsp'.
31174 The SCOUNT and POS bits of the DSP control register are global. The
31175 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
31176 POS bits. During optimization, the compiler will not delete these
31177 instructions and it will not delete calls to functions containing these
31180 At present, GCC only provides support for operations on 32-bit
31181 vectors. The vector type associated with 8-bit integer data is usually
31182 called `v4i8', the vector type associated with Q7 is usually called
31183 `v4q7', the vector type associated with 16-bit integer data is usually
31184 called `v2i16', and the vector type associated with Q15 is usually
31185 called `v2q15'. They can be defined in C as follows:
31187 typedef signed char v4i8 __attribute__ ((vector_size(4)));
31188 typedef signed char v4q7 __attribute__ ((vector_size(4)));
31189 typedef short v2i16 __attribute__ ((vector_size(4)));
31190 typedef short v2q15 __attribute__ ((vector_size(4)));
31192 `v4i8', `v4q7', `v2i16' and `v2q15' values are initialized in the same
31193 way as aggregates. For example:
31195 v4i8 a = {1, 2, 3, 4};
31197 b = (v4i8) {5, 6, 7, 8};
31199 v2q15 c = {0x0fcb, 0x3a75};
31201 d = (v2q15) {0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15};
31203 _Note:_ The CPU's endianness determines the order in which values are
31204 packed. On little-endian targets, the first value is the least
31205 significant and the last value is the most significant. The opposite
31206 order applies to big-endian targets. For example, the code above will
31207 set the lowest byte of `a' to `1' on little-endian targets and `4' on
31208 big-endian targets.
31210 _Note:_ Q7, Q15 and Q31 values must be initialized with their integer
31211 representation. As shown in this example, the integer representation
31212 of a Q7 value can be obtained by multiplying the fractional value by
31213 `0x1.0p7'. The equivalent for Q15 values is to multiply by `0x1.0p15'.
31214 The equivalent for Q31 values is to multiply by `0x1.0p31'.
31216 The table below lists the `v4i8' and `v2q15' operations for which
31217 hardware support exists. `a' and `b' are `v4i8' values, and `c' and
31218 `d' are `v2q15' values.
31220 C code MIPS instruction
31226 The table below lists the `v2i16' operation for which hardware support
31227 exists for the DSP ASE REV 2. `e' and `f' are `v2i16' values.
31229 C code MIPS instruction
31232 It is easier to describe the DSP built-in functions if we first define
31233 the following types:
31237 typedef unsigned int ui32;
31238 typedef long long a64;
31240 `q31' and `i32' are actually the same as `int', but we use `q31' to
31241 indicate a Q31 fractional value and `i32' to indicate a 32-bit integer
31242 value. Similarly, `a64' is the same as `long long', but we use `a64'
31243 to indicate values that will be placed in one of the four DSP
31244 accumulators (`$ac0', `$ac1', `$ac2' or `$ac3').
31246 Also, some built-in functions prefer or require immediate numbers as
31247 parameters, because the corresponding DSP instructions accept both
31248 immediate numbers and register operands, or accept immediate numbers
31249 only. The immediate parameters are listed as follows.
31256 imm0_255: 0 to 255.
31257 imm_n32_31: -32 to 31.
31258 imm_n512_511: -512 to 511.
31260 The following built-in functions map directly to a particular MIPS DSP
31261 instruction. Please refer to the architecture specification for
31262 details on what each instruction does.
31264 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
31265 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
31266 q31 __builtin_mips_addq_s_w (q31, q31)
31267 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
31268 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
31269 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
31270 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
31271 q31 __builtin_mips_subq_s_w (q31, q31)
31272 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
31273 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
31274 i32 __builtin_mips_addsc (i32, i32)
31275 i32 __builtin_mips_addwc (i32, i32)
31276 i32 __builtin_mips_modsub (i32, i32)
31277 i32 __builtin_mips_raddu_w_qb (v4i8)
31278 v2q15 __builtin_mips_absq_s_ph (v2q15)
31279 q31 __builtin_mips_absq_s_w (q31)
31280 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
31281 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
31282 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
31283 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
31284 q31 __builtin_mips_preceq_w_phl (v2q15)
31285 q31 __builtin_mips_preceq_w_phr (v2q15)
31286 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
31287 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
31288 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
31289 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
31290 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
31291 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
31292 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
31293 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
31294 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
31295 v4i8 __builtin_mips_shll_qb (v4i8, i32)
31296 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
31297 v2q15 __builtin_mips_shll_ph (v2q15, i32)
31298 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
31299 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
31300 q31 __builtin_mips_shll_s_w (q31, imm0_31)
31301 q31 __builtin_mips_shll_s_w (q31, i32)
31302 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
31303 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
31304 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
31305 v2q15 __builtin_mips_shra_ph (v2q15, i32)
31306 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
31307 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
31308 q31 __builtin_mips_shra_r_w (q31, imm0_31)
31309 q31 __builtin_mips_shra_r_w (q31, i32)
31310 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
31311 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
31312 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
31313 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
31314 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
31315 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
31316 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
31317 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
31318 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
31319 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
31320 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
31321 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
31322 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
31323 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
31324 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
31325 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
31326 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
31327 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
31328 i32 __builtin_mips_bitrev (i32)
31329 i32 __builtin_mips_insv (i32, i32)
31330 v4i8 __builtin_mips_repl_qb (imm0_255)
31331 v4i8 __builtin_mips_repl_qb (i32)
31332 v2q15 __builtin_mips_repl_ph (imm_n512_511)
31333 v2q15 __builtin_mips_repl_ph (i32)
31334 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
31335 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
31336 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
31337 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
31338 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
31339 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
31340 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
31341 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
31342 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
31343 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
31344 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
31345 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
31346 i32 __builtin_mips_extr_w (a64, imm0_31)
31347 i32 __builtin_mips_extr_w (a64, i32)
31348 i32 __builtin_mips_extr_r_w (a64, imm0_31)
31349 i32 __builtin_mips_extr_s_h (a64, i32)
31350 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
31351 i32 __builtin_mips_extr_rs_w (a64, i32)
31352 i32 __builtin_mips_extr_s_h (a64, imm0_31)
31353 i32 __builtin_mips_extr_r_w (a64, i32)
31354 i32 __builtin_mips_extp (a64, imm0_31)
31355 i32 __builtin_mips_extp (a64, i32)
31356 i32 __builtin_mips_extpdp (a64, imm0_31)
31357 i32 __builtin_mips_extpdp (a64, i32)
31358 a64 __builtin_mips_shilo (a64, imm_n32_31)
31359 a64 __builtin_mips_shilo (a64, i32)
31360 a64 __builtin_mips_mthlip (a64, i32)
31361 void __builtin_mips_wrdsp (i32, imm0_63)
31362 i32 __builtin_mips_rddsp (imm0_63)
31363 i32 __builtin_mips_lbux (void *, i32)
31364 i32 __builtin_mips_lhx (void *, i32)
31365 i32 __builtin_mips_lwx (void *, i32)
31366 i32 __builtin_mips_bposge32 (void)
31368 The following built-in functions map directly to a particular MIPS DSP
31369 REV 2 instruction. Please refer to the architecture specification for
31370 details on what each instruction does.
31372 v4q7 __builtin_mips_absq_s_qb (v4q7);
31373 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
31374 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
31375 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
31376 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
31377 i32 __builtin_mips_append (i32, i32, imm0_31);
31378 i32 __builtin_mips_balign (i32, i32, imm0_3);
31379 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
31380 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
31381 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
31382 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
31383 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
31384 a64 __builtin_mips_madd (a64, i32, i32);
31385 a64 __builtin_mips_maddu (a64, ui32, ui32);
31386 a64 __builtin_mips_msub (a64, i32, i32);
31387 a64 __builtin_mips_msubu (a64, ui32, ui32);
31388 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
31389 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
31390 q31 __builtin_mips_mulq_rs_w (q31, q31);
31391 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
31392 q31 __builtin_mips_mulq_s_w (q31, q31);
31393 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
31394 a64 __builtin_mips_mult (i32, i32);
31395 a64 __builtin_mips_multu (ui32, ui32);
31396 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
31397 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
31398 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
31399 i32 __builtin_mips_prepend (i32, i32, imm0_31);
31400 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
31401 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
31402 v4i8 __builtin_mips_shra_qb (v4i8, i32);
31403 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
31404 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
31405 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
31406 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
31407 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
31408 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
31409 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
31410 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
31411 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
31412 q31 __builtin_mips_addqh_w (q31, q31);
31413 q31 __builtin_mips_addqh_r_w (q31, q31);
31414 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
31415 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
31416 q31 __builtin_mips_subqh_w (q31, q31);
31417 q31 __builtin_mips_subqh_r_w (q31, q31);
31418 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
31419 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
31420 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
31421 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
31422 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
31423 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
31426 File: gcc.info, Node: MIPS Paired-Single Support, Next: MIPS Loongson Built-in Functions, Prev: MIPS DSP Built-in Functions, Up: Target Builtins
31428 5.50.8 MIPS Paired-Single Support
31429 ---------------------------------
31431 The MIPS64 architecture includes a number of instructions that operate
31432 on pairs of single-precision floating-point values. Each pair is
31433 packed into a 64-bit floating-point register, with one element being
31434 designated the "upper half" and the other being designated the "lower
31437 GCC supports paired-single operations using both the generic vector
31438 extensions (*note Vector Extensions::) and a collection of
31439 MIPS-specific built-in functions. Both kinds of support are enabled by
31440 the `-mpaired-single' command-line option.
31442 The vector type associated with paired-single values is usually called
31443 `v2sf'. It can be defined in C as follows:
31445 typedef float v2sf __attribute__ ((vector_size (8)));
31447 `v2sf' values are initialized in the same way as aggregates. For
31450 v2sf a = {1.5, 9.1};
31455 _Note:_ The CPU's endianness determines which value is stored in the
31456 upper half of a register and which value is stored in the lower half.
31457 On little-endian targets, the first value is the lower one and the
31458 second value is the upper one. The opposite order applies to
31459 big-endian targets. For example, the code above will set the lower
31460 half of `a' to `1.5' on little-endian targets and `9.1' on big-endian
31464 File: gcc.info, Node: MIPS Loongson Built-in Functions, Next: Other MIPS Built-in Functions, Prev: MIPS Paired-Single Support, Up: Target Builtins
31466 5.50.9 MIPS Loongson Built-in Functions
31467 ---------------------------------------
31469 GCC provides intrinsics to access the SIMD instructions provided by the
31470 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
31471 available after inclusion of the `loongson.h' header file, operate on
31472 the following 64-bit vector types:
31474 * `uint8x8_t', a vector of eight unsigned 8-bit integers;
31476 * `uint16x4_t', a vector of four unsigned 16-bit integers;
31478 * `uint32x2_t', a vector of two unsigned 32-bit integers;
31480 * `int8x8_t', a vector of eight signed 8-bit integers;
31482 * `int16x4_t', a vector of four signed 16-bit integers;
31484 * `int32x2_t', a vector of two signed 32-bit integers.
31486 The intrinsics provided are listed below; each is named after the
31487 machine instruction to which it corresponds, with suffixes added as
31488 appropriate to distinguish intrinsics that expand to the same machine
31489 instruction yet have different argument types. Refer to the
31490 architecture documentation for a description of the functionality of
31493 int16x4_t packsswh (int32x2_t s, int32x2_t t);
31494 int8x8_t packsshb (int16x4_t s, int16x4_t t);
31495 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
31496 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
31497 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
31498 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
31499 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
31500 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
31501 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
31502 uint64_t paddd_u (uint64_t s, uint64_t t);
31503 int64_t paddd_s (int64_t s, int64_t t);
31504 int16x4_t paddsh (int16x4_t s, int16x4_t t);
31505 int8x8_t paddsb (int8x8_t s, int8x8_t t);
31506 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
31507 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
31508 uint64_t pandn_ud (uint64_t s, uint64_t t);
31509 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
31510 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
31511 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
31512 int64_t pandn_sd (int64_t s, int64_t t);
31513 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
31514 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
31515 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
31516 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
31517 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
31518 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
31519 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
31520 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
31521 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
31522 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
31523 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
31524 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
31525 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
31526 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
31527 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
31528 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
31529 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
31530 uint16x4_t pextrh_u (uint16x4_t s, int field);
31531 int16x4_t pextrh_s (int16x4_t s, int field);
31532 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
31533 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
31534 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
31535 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
31536 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
31537 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
31538 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
31539 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
31540 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
31541 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
31542 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
31543 int16x4_t pminsh (int16x4_t s, int16x4_t t);
31544 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
31545 uint8x8_t pmovmskb_u (uint8x8_t s);
31546 int8x8_t pmovmskb_s (int8x8_t s);
31547 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
31548 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
31549 int16x4_t pmullh (int16x4_t s, int16x4_t t);
31550 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
31551 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
31552 uint16x4_t biadd (uint8x8_t s);
31553 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
31554 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
31555 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
31556 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
31557 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
31558 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
31559 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
31560 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
31561 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
31562 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
31563 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
31564 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
31565 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
31566 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
31567 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
31568 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
31569 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
31570 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
31571 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
31572 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
31573 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
31574 uint64_t psubd_u (uint64_t s, uint64_t t);
31575 int64_t psubd_s (int64_t s, int64_t t);
31576 int16x4_t psubsh (int16x4_t s, int16x4_t t);
31577 int8x8_t psubsb (int8x8_t s, int8x8_t t);
31578 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
31579 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
31580 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
31581 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
31582 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
31583 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
31584 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
31585 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
31586 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
31587 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
31588 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
31589 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
31590 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
31591 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
31595 * Paired-Single Arithmetic::
31596 * Paired-Single Built-in Functions::
31597 * MIPS-3D Built-in Functions::
31600 File: gcc.info, Node: Paired-Single Arithmetic, Next: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions
31602 5.50.9.1 Paired-Single Arithmetic
31603 .................................
31605 The table below lists the `v2sf' operations for which hardware support
31606 exists. `a', `b' and `c' are `v2sf' values and `x' is an integral
31609 C code MIPS instruction
31614 `a * b + c' `madd.ps'
31615 `a * b - c' `msub.ps'
31616 `-(a * b + c)' `nmadd.ps'
31617 `-(a * b - c)' `nmsub.ps'
31618 `x ? a : b' `movn.ps'/`movz.ps'
31620 Note that the multiply-accumulate instructions can be disabled using
31621 the command-line option `-mno-fused-madd'.
31624 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
31626 5.50.9.2 Paired-Single Built-in Functions
31627 .........................................
31629 The following paired-single functions map directly to a particular MIPS
31630 instruction. Please refer to the architecture specification for
31631 details on what each instruction does.
31633 `v2sf __builtin_mips_pll_ps (v2sf, v2sf)'
31634 Pair lower lower (`pll.ps').
31636 `v2sf __builtin_mips_pul_ps (v2sf, v2sf)'
31637 Pair upper lower (`pul.ps').
31639 `v2sf __builtin_mips_plu_ps (v2sf, v2sf)'
31640 Pair lower upper (`plu.ps').
31642 `v2sf __builtin_mips_puu_ps (v2sf, v2sf)'
31643 Pair upper upper (`puu.ps').
31645 `v2sf __builtin_mips_cvt_ps_s (float, float)'
31646 Convert pair to paired single (`cvt.ps.s').
31648 `float __builtin_mips_cvt_s_pl (v2sf)'
31649 Convert pair lower to single (`cvt.s.pl').
31651 `float __builtin_mips_cvt_s_pu (v2sf)'
31652 Convert pair upper to single (`cvt.s.pu').
31654 `v2sf __builtin_mips_abs_ps (v2sf)'
31655 Absolute value (`abs.ps').
31657 `v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)'
31658 Align variable (`alnv.ps').
31660 _Note:_ The value of the third parameter must be 0 or 4 modulo 8,
31661 otherwise the result will be unpredictable. Please read the
31662 instruction description for details.
31664 The following multi-instruction functions are also available. In each
31665 case, COND can be any of the 16 floating-point conditions: `f', `un',
31666 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
31667 `lt', `nge', `le' or `ngt'.
31669 `v2sf __builtin_mips_movt_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31670 `v2sf __builtin_mips_movf_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31671 Conditional move based on floating point comparison (`c.COND.ps',
31672 `movt.ps'/`movf.ps').
31674 The `movt' functions return the value X computed by:
31680 The `movf' functions are similar but use `movf.ps' instead of
31683 `int __builtin_mips_upper_c_COND_ps (v2sf A, v2sf B)'
31684 `int __builtin_mips_lower_c_COND_ps (v2sf A, v2sf B)'
31685 Comparison of two paired-single values (`c.COND.ps',
31688 These functions compare A and B using `c.COND.ps' and return
31689 either the upper or lower half of the result. For example:
31692 if (__builtin_mips_upper_c_eq_ps (a, b))
31693 upper_halves_are_equal ();
31695 upper_halves_are_unequal ();
31697 if (__builtin_mips_lower_c_eq_ps (a, b))
31698 lower_halves_are_equal ();
31700 lower_halves_are_unequal ();
31703 File: gcc.info, Node: MIPS-3D Built-in Functions, Prev: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions
31705 5.50.9.3 MIPS-3D Built-in Functions
31706 ...................................
31708 The MIPS-3D Application-Specific Extension (ASE) includes additional
31709 paired-single instructions that are designed to improve the performance
31710 of 3D graphics operations. Support for these instructions is controlled
31711 by the `-mips3d' command-line option.
31713 The functions listed below map directly to a particular MIPS-3D
31714 instruction. Please refer to the architecture specification for more
31715 details on what each instruction does.
31717 `v2sf __builtin_mips_addr_ps (v2sf, v2sf)'
31718 Reduction add (`addr.ps').
31720 `v2sf __builtin_mips_mulr_ps (v2sf, v2sf)'
31721 Reduction multiply (`mulr.ps').
31723 `v2sf __builtin_mips_cvt_pw_ps (v2sf)'
31724 Convert paired single to paired word (`cvt.pw.ps').
31726 `v2sf __builtin_mips_cvt_ps_pw (v2sf)'
31727 Convert paired word to paired single (`cvt.ps.pw').
31729 `float __builtin_mips_recip1_s (float)'
31730 `double __builtin_mips_recip1_d (double)'
31731 `v2sf __builtin_mips_recip1_ps (v2sf)'
31732 Reduced precision reciprocal (sequence step 1) (`recip1.FMT').
31734 `float __builtin_mips_recip2_s (float, float)'
31735 `double __builtin_mips_recip2_d (double, double)'
31736 `v2sf __builtin_mips_recip2_ps (v2sf, v2sf)'
31737 Reduced precision reciprocal (sequence step 2) (`recip2.FMT').
31739 `float __builtin_mips_rsqrt1_s (float)'
31740 `double __builtin_mips_rsqrt1_d (double)'
31741 `v2sf __builtin_mips_rsqrt1_ps (v2sf)'
31742 Reduced precision reciprocal square root (sequence step 1)
31745 `float __builtin_mips_rsqrt2_s (float, float)'
31746 `double __builtin_mips_rsqrt2_d (double, double)'
31747 `v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)'
31748 Reduced precision reciprocal square root (sequence step 2)
31751 The following multi-instruction functions are also available. In each
31752 case, COND can be any of the 16 floating-point conditions: `f', `un',
31753 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
31754 `lt', `nge', `le' or `ngt'.
31756 `int __builtin_mips_cabs_COND_s (float A, float B)'
31757 `int __builtin_mips_cabs_COND_d (double A, double B)'
31758 Absolute comparison of two scalar values (`cabs.COND.FMT',
31761 These functions compare A and B using `cabs.COND.s' or
31762 `cabs.COND.d' and return the result as a boolean value. For
31766 if (__builtin_mips_cabs_eq_s (a, b))
31771 `int __builtin_mips_upper_cabs_COND_ps (v2sf A, v2sf B)'
31772 `int __builtin_mips_lower_cabs_COND_ps (v2sf A, v2sf B)'
31773 Absolute comparison of two paired-single values (`cabs.COND.ps',
31776 These functions compare A and B using `cabs.COND.ps' and return
31777 either the upper or lower half of the result. For example:
31780 if (__builtin_mips_upper_cabs_eq_ps (a, b))
31781 upper_halves_are_equal ();
31783 upper_halves_are_unequal ();
31785 if (__builtin_mips_lower_cabs_eq_ps (a, b))
31786 lower_halves_are_equal ();
31788 lower_halves_are_unequal ();
31790 `v2sf __builtin_mips_movt_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31791 `v2sf __builtin_mips_movf_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
31792 Conditional move based on absolute comparison (`cabs.COND.ps',
31793 `movt.ps'/`movf.ps').
31795 The `movt' functions return the value X computed by:
31797 cabs.COND.ps CC,A,B
31801 The `movf' functions are similar but use `movf.ps' instead of
31804 `int __builtin_mips_any_c_COND_ps (v2sf A, v2sf B)'
31805 `int __builtin_mips_all_c_COND_ps (v2sf A, v2sf B)'
31806 `int __builtin_mips_any_cabs_COND_ps (v2sf A, v2sf B)'
31807 `int __builtin_mips_all_cabs_COND_ps (v2sf A, v2sf B)'
31808 Comparison of two paired-single values (`c.COND.ps'/`cabs.COND.ps',
31809 `bc1any2t'/`bc1any2f').
31811 These functions compare A and B using `c.COND.ps' or
31812 `cabs.COND.ps'. The `any' forms return true if either result is
31813 true and the `all' forms return true if both results are true.
31817 if (__builtin_mips_any_c_eq_ps (a, b))
31822 if (__builtin_mips_all_c_eq_ps (a, b))
31827 `int __builtin_mips_any_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31828 `int __builtin_mips_all_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31829 `int __builtin_mips_any_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31830 `int __builtin_mips_all_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
31831 Comparison of four paired-single values
31832 (`c.COND.ps'/`cabs.COND.ps', `bc1any4t'/`bc1any4f').
31834 These functions use `c.COND.ps' or `cabs.COND.ps' to compare A
31835 with B and to compare C with D. The `any' forms return true if
31836 any of the four results are true and the `all' forms return true
31837 if all four results are true. For example:
31840 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
31845 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
31851 File: gcc.info, Node: picoChip Built-in Functions, Next: PowerPC AltiVec Built-in Functions, Prev: Other MIPS Built-in Functions, Up: Target Builtins
31853 5.50.10 picoChip Built-in Functions
31854 -----------------------------------
31856 GCC provides an interface to selected machine instructions from the
31857 picoChip instruction set.
31859 `int __builtin_sbc (int VALUE)'
31860 Sign bit count. Return the number of consecutive bits in VALUE
31861 which have the same value as the sign-bit. The result is the
31862 number of leading sign bits minus one, giving the number of
31863 redundant sign bits in VALUE.
31865 `int __builtin_byteswap (int VALUE)'
31866 Byte swap. Return the result of swapping the upper and lower
31869 `int __builtin_brev (int VALUE)'
31870 Bit reversal. Return the result of reversing the bits in VALUE.
31871 Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1, and so
31874 `int __builtin_adds (int X, int Y)'
31875 Saturating addition. Return the result of adding X and Y, storing
31876 the value 32767 if the result overflows.
31878 `int __builtin_subs (int X, int Y)'
31879 Saturating subtraction. Return the result of subtracting Y from
31880 X, storing the value -32768 if the result overflows.
31882 `void __builtin_halt (void)'
31883 Halt. The processor will stop execution. This built-in is useful
31884 for implementing assertions.
31888 File: gcc.info, Node: Other MIPS Built-in Functions, Next: picoChip Built-in Functions, Prev: MIPS Loongson Built-in Functions, Up: Target Builtins
31890 5.50.11 Other MIPS Built-in Functions
31891 -------------------------------------
31893 GCC provides other MIPS-specific built-in functions:
31895 `void __builtin_mips_cache (int OP, const volatile void *ADDR)'
31896 Insert a `cache' instruction with operands OP and ADDR. GCC
31897 defines the preprocessor macro `___GCC_HAVE_BUILTIN_MIPS_CACHE'
31898 when this function is available.
31901 File: gcc.info, Node: PowerPC AltiVec Built-in Functions, Next: SPARC VIS Built-in Functions, Prev: picoChip Built-in Functions, Up: Target Builtins
31903 5.50.12 PowerPC AltiVec Built-in Functions
31904 ------------------------------------------
31906 GCC provides an interface for the PowerPC family of processors to access
31907 the AltiVec operations described in Motorola's AltiVec Programming
31908 Interface Manual. The interface is made available by including
31909 `<altivec.h>' and using `-maltivec' and `-mabi=altivec'. The interface
31910 supports the following vector types.
31912 vector unsigned char
31916 vector unsigned short
31917 vector signed short
31921 vector unsigned int
31926 GCC's implementation of the high-level language interface available
31927 from C and C++ code differs from Motorola's documentation in several
31930 * A vector constant is a list of constant expressions within curly
31933 * A vector initializer requires no cast if the vector constant is of
31934 the same type as the variable it is initializing.
31936 * If `signed' or `unsigned' is omitted, the signedness of the vector
31937 type is the default signedness of the base type. The default
31938 varies depending on the operating system, so a portable program
31939 should always specify the signedness.
31941 * Compiling with `-maltivec' adds keywords `__vector', `vector',
31942 `__pixel', `pixel', `__bool' and `bool'. When compiling ISO C,
31943 the context-sensitive substitution of the keywords `vector',
31944 `pixel' and `bool' is disabled. To use them, you must include
31945 `<altivec.h>' instead.
31947 * GCC allows using a `typedef' name as the type specifier for a
31950 * For C, overloaded functions are implemented with macros so the
31951 following does not work:
31953 vec_add ((vector signed int){1, 2, 3, 4}, foo);
31955 Since `vec_add' is a macro, the vector constant in the example is
31956 treated as four separate arguments. Wrap the entire argument in
31957 parentheses for this to work.
31959 _Note:_ Only the `<altivec.h>' interface is supported. Internally,
31960 GCC uses built-in functions to achieve the functionality in the
31961 aforementioned header file, but they are not supported and are subject
31962 to change without notice.
31964 The following interfaces are supported for the generic and specific
31965 AltiVec operations and the AltiVec predicates. In cases where there is
31966 a direct mapping between generic and specific operations, only the
31967 generic names are shown here, although the specific operations can also
31970 Arguments that are documented as `const int' require literal integral
31971 values within the range required for that operation.
31973 vector signed char vec_abs (vector signed char);
31974 vector signed short vec_abs (vector signed short);
31975 vector signed int vec_abs (vector signed int);
31976 vector float vec_abs (vector float);
31978 vector signed char vec_abss (vector signed char);
31979 vector signed short vec_abss (vector signed short);
31980 vector signed int vec_abss (vector signed int);
31982 vector signed char vec_add (vector bool char, vector signed char);
31983 vector signed char vec_add (vector signed char, vector bool char);
31984 vector signed char vec_add (vector signed char, vector signed char);
31985 vector unsigned char vec_add (vector bool char, vector unsigned char);
31986 vector unsigned char vec_add (vector unsigned char, vector bool char);
31987 vector unsigned char vec_add (vector unsigned char,
31988 vector unsigned char);
31989 vector signed short vec_add (vector bool short, vector signed short);
31990 vector signed short vec_add (vector signed short, vector bool short);
31991 vector signed short vec_add (vector signed short, vector signed short);
31992 vector unsigned short vec_add (vector bool short,
31993 vector unsigned short);
31994 vector unsigned short vec_add (vector unsigned short,
31995 vector bool short);
31996 vector unsigned short vec_add (vector unsigned short,
31997 vector unsigned short);
31998 vector signed int vec_add (vector bool int, vector signed int);
31999 vector signed int vec_add (vector signed int, vector bool int);
32000 vector signed int vec_add (vector signed int, vector signed int);
32001 vector unsigned int vec_add (vector bool int, vector unsigned int);
32002 vector unsigned int vec_add (vector unsigned int, vector bool int);
32003 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
32004 vector float vec_add (vector float, vector float);
32006 vector float vec_vaddfp (vector float, vector float);
32008 vector signed int vec_vadduwm (vector bool int, vector signed int);
32009 vector signed int vec_vadduwm (vector signed int, vector bool int);
32010 vector signed int vec_vadduwm (vector signed int, vector signed int);
32011 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
32012 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
32013 vector unsigned int vec_vadduwm (vector unsigned int,
32014 vector unsigned int);
32016 vector signed short vec_vadduhm (vector bool short,
32017 vector signed short);
32018 vector signed short vec_vadduhm (vector signed short,
32019 vector bool short);
32020 vector signed short vec_vadduhm (vector signed short,
32021 vector signed short);
32022 vector unsigned short vec_vadduhm (vector bool short,
32023 vector unsigned short);
32024 vector unsigned short vec_vadduhm (vector unsigned short,
32025 vector bool short);
32026 vector unsigned short vec_vadduhm (vector unsigned short,
32027 vector unsigned short);
32029 vector signed char vec_vaddubm (vector bool char, vector signed char);
32030 vector signed char vec_vaddubm (vector signed char, vector bool char);
32031 vector signed char vec_vaddubm (vector signed char, vector signed char);
32032 vector unsigned char vec_vaddubm (vector bool char,
32033 vector unsigned char);
32034 vector unsigned char vec_vaddubm (vector unsigned char,
32036 vector unsigned char vec_vaddubm (vector unsigned char,
32037 vector unsigned char);
32039 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
32041 vector unsigned char vec_adds (vector bool char, vector unsigned char);
32042 vector unsigned char vec_adds (vector unsigned char, vector bool char);
32043 vector unsigned char vec_adds (vector unsigned char,
32044 vector unsigned char);
32045 vector signed char vec_adds (vector bool char, vector signed char);
32046 vector signed char vec_adds (vector signed char, vector bool char);
32047 vector signed char vec_adds (vector signed char, vector signed char);
32048 vector unsigned short vec_adds (vector bool short,
32049 vector unsigned short);
32050 vector unsigned short vec_adds (vector unsigned short,
32051 vector bool short);
32052 vector unsigned short vec_adds (vector unsigned short,
32053 vector unsigned short);
32054 vector signed short vec_adds (vector bool short, vector signed short);
32055 vector signed short vec_adds (vector signed short, vector bool short);
32056 vector signed short vec_adds (vector signed short, vector signed short);
32057 vector unsigned int vec_adds (vector bool int, vector unsigned int);
32058 vector unsigned int vec_adds (vector unsigned int, vector bool int);
32059 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
32060 vector signed int vec_adds (vector bool int, vector signed int);
32061 vector signed int vec_adds (vector signed int, vector bool int);
32062 vector signed int vec_adds (vector signed int, vector signed int);
32064 vector signed int vec_vaddsws (vector bool int, vector signed int);
32065 vector signed int vec_vaddsws (vector signed int, vector bool int);
32066 vector signed int vec_vaddsws (vector signed int, vector signed int);
32068 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
32069 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
32070 vector unsigned int vec_vadduws (vector unsigned int,
32071 vector unsigned int);
32073 vector signed short vec_vaddshs (vector bool short,
32074 vector signed short);
32075 vector signed short vec_vaddshs (vector signed short,
32076 vector bool short);
32077 vector signed short vec_vaddshs (vector signed short,
32078 vector signed short);
32080 vector unsigned short vec_vadduhs (vector bool short,
32081 vector unsigned short);
32082 vector unsigned short vec_vadduhs (vector unsigned short,
32083 vector bool short);
32084 vector unsigned short vec_vadduhs (vector unsigned short,
32085 vector unsigned short);
32087 vector signed char vec_vaddsbs (vector bool char, vector signed char);
32088 vector signed char vec_vaddsbs (vector signed char, vector bool char);
32089 vector signed char vec_vaddsbs (vector signed char, vector signed char);
32091 vector unsigned char vec_vaddubs (vector bool char,
32092 vector unsigned char);
32093 vector unsigned char vec_vaddubs (vector unsigned char,
32095 vector unsigned char vec_vaddubs (vector unsigned char,
32096 vector unsigned char);
32098 vector float vec_and (vector float, vector float);
32099 vector float vec_and (vector float, vector bool int);
32100 vector float vec_and (vector bool int, vector float);
32101 vector bool int vec_and (vector bool int, vector bool int);
32102 vector signed int vec_and (vector bool int, vector signed int);
32103 vector signed int vec_and (vector signed int, vector bool int);
32104 vector signed int vec_and (vector signed int, vector signed int);
32105 vector unsigned int vec_and (vector bool int, vector unsigned int);
32106 vector unsigned int vec_and (vector unsigned int, vector bool int);
32107 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
32108 vector bool short vec_and (vector bool short, vector bool short);
32109 vector signed short vec_and (vector bool short, vector signed short);
32110 vector signed short vec_and (vector signed short, vector bool short);
32111 vector signed short vec_and (vector signed short, vector signed short);
32112 vector unsigned short vec_and (vector bool short,
32113 vector unsigned short);
32114 vector unsigned short vec_and (vector unsigned short,
32115 vector bool short);
32116 vector unsigned short vec_and (vector unsigned short,
32117 vector unsigned short);
32118 vector signed char vec_and (vector bool char, vector signed char);
32119 vector bool char vec_and (vector bool char, vector bool char);
32120 vector signed char vec_and (vector signed char, vector bool char);
32121 vector signed char vec_and (vector signed char, vector signed char);
32122 vector unsigned char vec_and (vector bool char, vector unsigned char);
32123 vector unsigned char vec_and (vector unsigned char, vector bool char);
32124 vector unsigned char vec_and (vector unsigned char,
32125 vector unsigned char);
32127 vector float vec_andc (vector float, vector float);
32128 vector float vec_andc (vector float, vector bool int);
32129 vector float vec_andc (vector bool int, vector float);
32130 vector bool int vec_andc (vector bool int, vector bool int);
32131 vector signed int vec_andc (vector bool int, vector signed int);
32132 vector signed int vec_andc (vector signed int, vector bool int);
32133 vector signed int vec_andc (vector signed int, vector signed int);
32134 vector unsigned int vec_andc (vector bool int, vector unsigned int);
32135 vector unsigned int vec_andc (vector unsigned int, vector bool int);
32136 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
32137 vector bool short vec_andc (vector bool short, vector bool short);
32138 vector signed short vec_andc (vector bool short, vector signed short);
32139 vector signed short vec_andc (vector signed short, vector bool short);
32140 vector signed short vec_andc (vector signed short, vector signed short);
32141 vector unsigned short vec_andc (vector bool short,
32142 vector unsigned short);
32143 vector unsigned short vec_andc (vector unsigned short,
32144 vector bool short);
32145 vector unsigned short vec_andc (vector unsigned short,
32146 vector unsigned short);
32147 vector signed char vec_andc (vector bool char, vector signed char);
32148 vector bool char vec_andc (vector bool char, vector bool char);
32149 vector signed char vec_andc (vector signed char, vector bool char);
32150 vector signed char vec_andc (vector signed char, vector signed char);
32151 vector unsigned char vec_andc (vector bool char, vector unsigned char);
32152 vector unsigned char vec_andc (vector unsigned char, vector bool char);
32153 vector unsigned char vec_andc (vector unsigned char,
32154 vector unsigned char);
32156 vector unsigned char vec_avg (vector unsigned char,
32157 vector unsigned char);
32158 vector signed char vec_avg (vector signed char, vector signed char);
32159 vector unsigned short vec_avg (vector unsigned short,
32160 vector unsigned short);
32161 vector signed short vec_avg (vector signed short, vector signed short);
32162 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
32163 vector signed int vec_avg (vector signed int, vector signed int);
32165 vector signed int vec_vavgsw (vector signed int, vector signed int);
32167 vector unsigned int vec_vavguw (vector unsigned int,
32168 vector unsigned int);
32170 vector signed short vec_vavgsh (vector signed short,
32171 vector signed short);
32173 vector unsigned short vec_vavguh (vector unsigned short,
32174 vector unsigned short);
32176 vector signed char vec_vavgsb (vector signed char, vector signed char);
32178 vector unsigned char vec_vavgub (vector unsigned char,
32179 vector unsigned char);
32181 vector float vec_ceil (vector float);
32183 vector signed int vec_cmpb (vector float, vector float);
32185 vector bool char vec_cmpeq (vector signed char, vector signed char);
32186 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
32187 vector bool short vec_cmpeq (vector signed short, vector signed short);
32188 vector bool short vec_cmpeq (vector unsigned short,
32189 vector unsigned short);
32190 vector bool int vec_cmpeq (vector signed int, vector signed int);
32191 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
32192 vector bool int vec_cmpeq (vector float, vector float);
32194 vector bool int vec_vcmpeqfp (vector float, vector float);
32196 vector bool int vec_vcmpequw (vector signed int, vector signed int);
32197 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
32199 vector bool short vec_vcmpequh (vector signed short,
32200 vector signed short);
32201 vector bool short vec_vcmpequh (vector unsigned short,
32202 vector unsigned short);
32204 vector bool char vec_vcmpequb (vector signed char, vector signed char);
32205 vector bool char vec_vcmpequb (vector unsigned char,
32206 vector unsigned char);
32208 vector bool int vec_cmpge (vector float, vector float);
32210 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
32211 vector bool char vec_cmpgt (vector signed char, vector signed char);
32212 vector bool short vec_cmpgt (vector unsigned short,
32213 vector unsigned short);
32214 vector bool short vec_cmpgt (vector signed short, vector signed short);
32215 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
32216 vector bool int vec_cmpgt (vector signed int, vector signed int);
32217 vector bool int vec_cmpgt (vector float, vector float);
32219 vector bool int vec_vcmpgtfp (vector float, vector float);
32221 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
32223 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
32225 vector bool short vec_vcmpgtsh (vector signed short,
32226 vector signed short);
32228 vector bool short vec_vcmpgtuh (vector unsigned short,
32229 vector unsigned short);
32231 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
32233 vector bool char vec_vcmpgtub (vector unsigned char,
32234 vector unsigned char);
32236 vector bool int vec_cmple (vector float, vector float);
32238 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
32239 vector bool char vec_cmplt (vector signed char, vector signed char);
32240 vector bool short vec_cmplt (vector unsigned short,
32241 vector unsigned short);
32242 vector bool short vec_cmplt (vector signed short, vector signed short);
32243 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
32244 vector bool int vec_cmplt (vector signed int, vector signed int);
32245 vector bool int vec_cmplt (vector float, vector float);
32247 vector float vec_ctf (vector unsigned int, const int);
32248 vector float vec_ctf (vector signed int, const int);
32250 vector float vec_vcfsx (vector signed int, const int);
32252 vector float vec_vcfux (vector unsigned int, const int);
32254 vector signed int vec_cts (vector float, const int);
32256 vector unsigned int vec_ctu (vector float, const int);
32258 void vec_dss (const int);
32260 void vec_dssall (void);
32262 void vec_dst (const vector unsigned char *, int, const int);
32263 void vec_dst (const vector signed char *, int, const int);
32264 void vec_dst (const vector bool char *, int, const int);
32265 void vec_dst (const vector unsigned short *, int, const int);
32266 void vec_dst (const vector signed short *, int, const int);
32267 void vec_dst (const vector bool short *, int, const int);
32268 void vec_dst (const vector pixel *, int, const int);
32269 void vec_dst (const vector unsigned int *, int, const int);
32270 void vec_dst (const vector signed int *, int, const int);
32271 void vec_dst (const vector bool int *, int, const int);
32272 void vec_dst (const vector float *, int, const int);
32273 void vec_dst (const unsigned char *, int, const int);
32274 void vec_dst (const signed char *, int, const int);
32275 void vec_dst (const unsigned short *, int, const int);
32276 void vec_dst (const short *, int, const int);
32277 void vec_dst (const unsigned int *, int, const int);
32278 void vec_dst (const int *, int, const int);
32279 void vec_dst (const unsigned long *, int, const int);
32280 void vec_dst (const long *, int, const int);
32281 void vec_dst (const float *, int, const int);
32283 void vec_dstst (const vector unsigned char *, int, const int);
32284 void vec_dstst (const vector signed char *, int, const int);
32285 void vec_dstst (const vector bool char *, int, const int);
32286 void vec_dstst (const vector unsigned short *, int, const int);
32287 void vec_dstst (const vector signed short *, int, const int);
32288 void vec_dstst (const vector bool short *, int, const int);
32289 void vec_dstst (const vector pixel *, int, const int);
32290 void vec_dstst (const vector unsigned int *, int, const int);
32291 void vec_dstst (const vector signed int *, int, const int);
32292 void vec_dstst (const vector bool int *, int, const int);
32293 void vec_dstst (const vector float *, int, const int);
32294 void vec_dstst (const unsigned char *, int, const int);
32295 void vec_dstst (const signed char *, int, const int);
32296 void vec_dstst (const unsigned short *, int, const int);
32297 void vec_dstst (const short *, int, const int);
32298 void vec_dstst (const unsigned int *, int, const int);
32299 void vec_dstst (const int *, int, const int);
32300 void vec_dstst (const unsigned long *, int, const int);
32301 void vec_dstst (const long *, int, const int);
32302 void vec_dstst (const float *, int, const int);
32304 void vec_dststt (const vector unsigned char *, int, const int);
32305 void vec_dststt (const vector signed char *, int, const int);
32306 void vec_dststt (const vector bool char *, int, const int);
32307 void vec_dststt (const vector unsigned short *, int, const int);
32308 void vec_dststt (const vector signed short *, int, const int);
32309 void vec_dststt (const vector bool short *, int, const int);
32310 void vec_dststt (const vector pixel *, int, const int);
32311 void vec_dststt (const vector unsigned int *, int, const int);
32312 void vec_dststt (const vector signed int *, int, const int);
32313 void vec_dststt (const vector bool int *, int, const int);
32314 void vec_dststt (const vector float *, int, const int);
32315 void vec_dststt (const unsigned char *, int, const int);
32316 void vec_dststt (const signed char *, int, const int);
32317 void vec_dststt (const unsigned short *, int, const int);
32318 void vec_dststt (const short *, int, const int);
32319 void vec_dststt (const unsigned int *, int, const int);
32320 void vec_dststt (const int *, int, const int);
32321 void vec_dststt (const unsigned long *, int, const int);
32322 void vec_dststt (const long *, int, const int);
32323 void vec_dststt (const float *, int, const int);
32325 void vec_dstt (const vector unsigned char *, int, const int);
32326 void vec_dstt (const vector signed char *, int, const int);
32327 void vec_dstt (const vector bool char *, int, const int);
32328 void vec_dstt (const vector unsigned short *, int, const int);
32329 void vec_dstt (const vector signed short *, int, const int);
32330 void vec_dstt (const vector bool short *, int, const int);
32331 void vec_dstt (const vector pixel *, int, const int);
32332 void vec_dstt (const vector unsigned int *, int, const int);
32333 void vec_dstt (const vector signed int *, int, const int);
32334 void vec_dstt (const vector bool int *, int, const int);
32335 void vec_dstt (const vector float *, int, const int);
32336 void vec_dstt (const unsigned char *, int, const int);
32337 void vec_dstt (const signed char *, int, const int);
32338 void vec_dstt (const unsigned short *, int, const int);
32339 void vec_dstt (const short *, int, const int);
32340 void vec_dstt (const unsigned int *, int, const int);
32341 void vec_dstt (const int *, int, const int);
32342 void vec_dstt (const unsigned long *, int, const int);
32343 void vec_dstt (const long *, int, const int);
32344 void vec_dstt (const float *, int, const int);
32346 vector float vec_expte (vector float);
32348 vector float vec_floor (vector float);
32350 vector float vec_ld (int, const vector float *);
32351 vector float vec_ld (int, const float *);
32352 vector bool int vec_ld (int, const vector bool int *);
32353 vector signed int vec_ld (int, const vector signed int *);
32354 vector signed int vec_ld (int, const int *);
32355 vector signed int vec_ld (int, const long *);
32356 vector unsigned int vec_ld (int, const vector unsigned int *);
32357 vector unsigned int vec_ld (int, const unsigned int *);
32358 vector unsigned int vec_ld (int, const unsigned long *);
32359 vector bool short vec_ld (int, const vector bool short *);
32360 vector pixel vec_ld (int, const vector pixel *);
32361 vector signed short vec_ld (int, const vector signed short *);
32362 vector signed short vec_ld (int, const short *);
32363 vector unsigned short vec_ld (int, const vector unsigned short *);
32364 vector unsigned short vec_ld (int, const unsigned short *);
32365 vector bool char vec_ld (int, const vector bool char *);
32366 vector signed char vec_ld (int, const vector signed char *);
32367 vector signed char vec_ld (int, const signed char *);
32368 vector unsigned char vec_ld (int, const vector unsigned char *);
32369 vector unsigned char vec_ld (int, const unsigned char *);
32371 vector signed char vec_lde (int, const signed char *);
32372 vector unsigned char vec_lde (int, const unsigned char *);
32373 vector signed short vec_lde (int, const short *);
32374 vector unsigned short vec_lde (int, const unsigned short *);
32375 vector float vec_lde (int, const float *);
32376 vector signed int vec_lde (int, const int *);
32377 vector unsigned int vec_lde (int, const unsigned int *);
32378 vector signed int vec_lde (int, const long *);
32379 vector unsigned int vec_lde (int, const unsigned long *);
32381 vector float vec_lvewx (int, float *);
32382 vector signed int vec_lvewx (int, int *);
32383 vector unsigned int vec_lvewx (int, unsigned int *);
32384 vector signed int vec_lvewx (int, long *);
32385 vector unsigned int vec_lvewx (int, unsigned long *);
32387 vector signed short vec_lvehx (int, short *);
32388 vector unsigned short vec_lvehx (int, unsigned short *);
32390 vector signed char vec_lvebx (int, char *);
32391 vector unsigned char vec_lvebx (int, unsigned char *);
32393 vector float vec_ldl (int, const vector float *);
32394 vector float vec_ldl (int, const float *);
32395 vector bool int vec_ldl (int, const vector bool int *);
32396 vector signed int vec_ldl (int, const vector signed int *);
32397 vector signed int vec_ldl (int, const int *);
32398 vector signed int vec_ldl (int, const long *);
32399 vector unsigned int vec_ldl (int, const vector unsigned int *);
32400 vector unsigned int vec_ldl (int, const unsigned int *);
32401 vector unsigned int vec_ldl (int, const unsigned long *);
32402 vector bool short vec_ldl (int, const vector bool short *);
32403 vector pixel vec_ldl (int, const vector pixel *);
32404 vector signed short vec_ldl (int, const vector signed short *);
32405 vector signed short vec_ldl (int, const short *);
32406 vector unsigned short vec_ldl (int, const vector unsigned short *);
32407 vector unsigned short vec_ldl (int, const unsigned short *);
32408 vector bool char vec_ldl (int, const vector bool char *);
32409 vector signed char vec_ldl (int, const vector signed char *);
32410 vector signed char vec_ldl (int, const signed char *);
32411 vector unsigned char vec_ldl (int, const vector unsigned char *);
32412 vector unsigned char vec_ldl (int, const unsigned char *);
32414 vector float vec_loge (vector float);
32416 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
32417 vector unsigned char vec_lvsl (int, const volatile signed char *);
32418 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
32419 vector unsigned char vec_lvsl (int, const volatile short *);
32420 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
32421 vector unsigned char vec_lvsl (int, const volatile int *);
32422 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
32423 vector unsigned char vec_lvsl (int, const volatile long *);
32424 vector unsigned char vec_lvsl (int, const volatile float *);
32426 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
32427 vector unsigned char vec_lvsr (int, const volatile signed char *);
32428 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
32429 vector unsigned char vec_lvsr (int, const volatile short *);
32430 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
32431 vector unsigned char vec_lvsr (int, const volatile int *);
32432 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
32433 vector unsigned char vec_lvsr (int, const volatile long *);
32434 vector unsigned char vec_lvsr (int, const volatile float *);
32436 vector float vec_madd (vector float, vector float, vector float);
32438 vector signed short vec_madds (vector signed short,
32439 vector signed short,
32440 vector signed short);
32442 vector unsigned char vec_max (vector bool char, vector unsigned char);
32443 vector unsigned char vec_max (vector unsigned char, vector bool char);
32444 vector unsigned char vec_max (vector unsigned char,
32445 vector unsigned char);
32446 vector signed char vec_max (vector bool char, vector signed char);
32447 vector signed char vec_max (vector signed char, vector bool char);
32448 vector signed char vec_max (vector signed char, vector signed char);
32449 vector unsigned short vec_max (vector bool short,
32450 vector unsigned short);
32451 vector unsigned short vec_max (vector unsigned short,
32452 vector bool short);
32453 vector unsigned short vec_max (vector unsigned short,
32454 vector unsigned short);
32455 vector signed short vec_max (vector bool short, vector signed short);
32456 vector signed short vec_max (vector signed short, vector bool short);
32457 vector signed short vec_max (vector signed short, vector signed short);
32458 vector unsigned int vec_max (vector bool int, vector unsigned int);
32459 vector unsigned int vec_max (vector unsigned int, vector bool int);
32460 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
32461 vector signed int vec_max (vector bool int, vector signed int);
32462 vector signed int vec_max (vector signed int, vector bool int);
32463 vector signed int vec_max (vector signed int, vector signed int);
32464 vector float vec_max (vector float, vector float);
32466 vector float vec_vmaxfp (vector float, vector float);
32468 vector signed int vec_vmaxsw (vector bool int, vector signed int);
32469 vector signed int vec_vmaxsw (vector signed int, vector bool int);
32470 vector signed int vec_vmaxsw (vector signed int, vector signed int);
32472 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
32473 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
32474 vector unsigned int vec_vmaxuw (vector unsigned int,
32475 vector unsigned int);
32477 vector signed short vec_vmaxsh (vector bool short, vector signed short);
32478 vector signed short vec_vmaxsh (vector signed short, vector bool short);
32479 vector signed short vec_vmaxsh (vector signed short,
32480 vector signed short);
32482 vector unsigned short vec_vmaxuh (vector bool short,
32483 vector unsigned short);
32484 vector unsigned short vec_vmaxuh (vector unsigned short,
32485 vector bool short);
32486 vector unsigned short vec_vmaxuh (vector unsigned short,
32487 vector unsigned short);
32489 vector signed char vec_vmaxsb (vector bool char, vector signed char);
32490 vector signed char vec_vmaxsb (vector signed char, vector bool char);
32491 vector signed char vec_vmaxsb (vector signed char, vector signed char);
32493 vector unsigned char vec_vmaxub (vector bool char,
32494 vector unsigned char);
32495 vector unsigned char vec_vmaxub (vector unsigned char,
32497 vector unsigned char vec_vmaxub (vector unsigned char,
32498 vector unsigned char);
32500 vector bool char vec_mergeh (vector bool char, vector bool char);
32501 vector signed char vec_mergeh (vector signed char, vector signed char);
32502 vector unsigned char vec_mergeh (vector unsigned char,
32503 vector unsigned char);
32504 vector bool short vec_mergeh (vector bool short, vector bool short);
32505 vector pixel vec_mergeh (vector pixel, vector pixel);
32506 vector signed short vec_mergeh (vector signed short,
32507 vector signed short);
32508 vector unsigned short vec_mergeh (vector unsigned short,
32509 vector unsigned short);
32510 vector float vec_mergeh (vector float, vector float);
32511 vector bool int vec_mergeh (vector bool int, vector bool int);
32512 vector signed int vec_mergeh (vector signed int, vector signed int);
32513 vector unsigned int vec_mergeh (vector unsigned int,
32514 vector unsigned int);
32516 vector float vec_vmrghw (vector float, vector float);
32517 vector bool int vec_vmrghw (vector bool int, vector bool int);
32518 vector signed int vec_vmrghw (vector signed int, vector signed int);
32519 vector unsigned int vec_vmrghw (vector unsigned int,
32520 vector unsigned int);
32522 vector bool short vec_vmrghh (vector bool short, vector bool short);
32523 vector signed short vec_vmrghh (vector signed short,
32524 vector signed short);
32525 vector unsigned short vec_vmrghh (vector unsigned short,
32526 vector unsigned short);
32527 vector pixel vec_vmrghh (vector pixel, vector pixel);
32529 vector bool char vec_vmrghb (vector bool char, vector bool char);
32530 vector signed char vec_vmrghb (vector signed char, vector signed char);
32531 vector unsigned char vec_vmrghb (vector unsigned char,
32532 vector unsigned char);
32534 vector bool char vec_mergel (vector bool char, vector bool char);
32535 vector signed char vec_mergel (vector signed char, vector signed char);
32536 vector unsigned char vec_mergel (vector unsigned char,
32537 vector unsigned char);
32538 vector bool short vec_mergel (vector bool short, vector bool short);
32539 vector pixel vec_mergel (vector pixel, vector pixel);
32540 vector signed short vec_mergel (vector signed short,
32541 vector signed short);
32542 vector unsigned short vec_mergel (vector unsigned short,
32543 vector unsigned short);
32544 vector float vec_mergel (vector float, vector float);
32545 vector bool int vec_mergel (vector bool int, vector bool int);
32546 vector signed int vec_mergel (vector signed int, vector signed int);
32547 vector unsigned int vec_mergel (vector unsigned int,
32548 vector unsigned int);
32550 vector float vec_vmrglw (vector float, vector float);
32551 vector signed int vec_vmrglw (vector signed int, vector signed int);
32552 vector unsigned int vec_vmrglw (vector unsigned int,
32553 vector unsigned int);
32554 vector bool int vec_vmrglw (vector bool int, vector bool int);
32556 vector bool short vec_vmrglh (vector bool short, vector bool short);
32557 vector signed short vec_vmrglh (vector signed short,
32558 vector signed short);
32559 vector unsigned short vec_vmrglh (vector unsigned short,
32560 vector unsigned short);
32561 vector pixel vec_vmrglh (vector pixel, vector pixel);
32563 vector bool char vec_vmrglb (vector bool char, vector bool char);
32564 vector signed char vec_vmrglb (vector signed char, vector signed char);
32565 vector unsigned char vec_vmrglb (vector unsigned char,
32566 vector unsigned char);
32568 vector unsigned short vec_mfvscr (void);
32570 vector unsigned char vec_min (vector bool char, vector unsigned char);
32571 vector unsigned char vec_min (vector unsigned char, vector bool char);
32572 vector unsigned char vec_min (vector unsigned char,
32573 vector unsigned char);
32574 vector signed char vec_min (vector bool char, vector signed char);
32575 vector signed char vec_min (vector signed char, vector bool char);
32576 vector signed char vec_min (vector signed char, vector signed char);
32577 vector unsigned short vec_min (vector bool short,
32578 vector unsigned short);
32579 vector unsigned short vec_min (vector unsigned short,
32580 vector bool short);
32581 vector unsigned short vec_min (vector unsigned short,
32582 vector unsigned short);
32583 vector signed short vec_min (vector bool short, vector signed short);
32584 vector signed short vec_min (vector signed short, vector bool short);
32585 vector signed short vec_min (vector signed short, vector signed short);
32586 vector unsigned int vec_min (vector bool int, vector unsigned int);
32587 vector unsigned int vec_min (vector unsigned int, vector bool int);
32588 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
32589 vector signed int vec_min (vector bool int, vector signed int);
32590 vector signed int vec_min (vector signed int, vector bool int);
32591 vector signed int vec_min (vector signed int, vector signed int);
32592 vector float vec_min (vector float, vector float);
32594 vector float vec_vminfp (vector float, vector float);
32596 vector signed int vec_vminsw (vector bool int, vector signed int);
32597 vector signed int vec_vminsw (vector signed int, vector bool int);
32598 vector signed int vec_vminsw (vector signed int, vector signed int);
32600 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
32601 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
32602 vector unsigned int vec_vminuw (vector unsigned int,
32603 vector unsigned int);
32605 vector signed short vec_vminsh (vector bool short, vector signed short);
32606 vector signed short vec_vminsh (vector signed short, vector bool short);
32607 vector signed short vec_vminsh (vector signed short,
32608 vector signed short);
32610 vector unsigned short vec_vminuh (vector bool short,
32611 vector unsigned short);
32612 vector unsigned short vec_vminuh (vector unsigned short,
32613 vector bool short);
32614 vector unsigned short vec_vminuh (vector unsigned short,
32615 vector unsigned short);
32617 vector signed char vec_vminsb (vector bool char, vector signed char);
32618 vector signed char vec_vminsb (vector signed char, vector bool char);
32619 vector signed char vec_vminsb (vector signed char, vector signed char);
32621 vector unsigned char vec_vminub (vector bool char,
32622 vector unsigned char);
32623 vector unsigned char vec_vminub (vector unsigned char,
32625 vector unsigned char vec_vminub (vector unsigned char,
32626 vector unsigned char);
32628 vector signed short vec_mladd (vector signed short,
32629 vector signed short,
32630 vector signed short);
32631 vector signed short vec_mladd (vector signed short,
32632 vector unsigned short,
32633 vector unsigned short);
32634 vector signed short vec_mladd (vector unsigned short,
32635 vector signed short,
32636 vector signed short);
32637 vector unsigned short vec_mladd (vector unsigned short,
32638 vector unsigned short,
32639 vector unsigned short);
32641 vector signed short vec_mradds (vector signed short,
32642 vector signed short,
32643 vector signed short);
32645 vector unsigned int vec_msum (vector unsigned char,
32646 vector unsigned char,
32647 vector unsigned int);
32648 vector signed int vec_msum (vector signed char,
32649 vector unsigned char,
32650 vector signed int);
32651 vector unsigned int vec_msum (vector unsigned short,
32652 vector unsigned short,
32653 vector unsigned int);
32654 vector signed int vec_msum (vector signed short,
32655 vector signed short,
32656 vector signed int);
32658 vector signed int vec_vmsumshm (vector signed short,
32659 vector signed short,
32660 vector signed int);
32662 vector unsigned int vec_vmsumuhm (vector unsigned short,
32663 vector unsigned short,
32664 vector unsigned int);
32666 vector signed int vec_vmsummbm (vector signed char,
32667 vector unsigned char,
32668 vector signed int);
32670 vector unsigned int vec_vmsumubm (vector unsigned char,
32671 vector unsigned char,
32672 vector unsigned int);
32674 vector unsigned int vec_msums (vector unsigned short,
32675 vector unsigned short,
32676 vector unsigned int);
32677 vector signed int vec_msums (vector signed short,
32678 vector signed short,
32679 vector signed int);
32681 vector signed int vec_vmsumshs (vector signed short,
32682 vector signed short,
32683 vector signed int);
32685 vector unsigned int vec_vmsumuhs (vector unsigned short,
32686 vector unsigned short,
32687 vector unsigned int);
32689 void vec_mtvscr (vector signed int);
32690 void vec_mtvscr (vector unsigned int);
32691 void vec_mtvscr (vector bool int);
32692 void vec_mtvscr (vector signed short);
32693 void vec_mtvscr (vector unsigned short);
32694 void vec_mtvscr (vector bool short);
32695 void vec_mtvscr (vector pixel);
32696 void vec_mtvscr (vector signed char);
32697 void vec_mtvscr (vector unsigned char);
32698 void vec_mtvscr (vector bool char);
32700 vector unsigned short vec_mule (vector unsigned char,
32701 vector unsigned char);
32702 vector signed short vec_mule (vector signed char,
32703 vector signed char);
32704 vector unsigned int vec_mule (vector unsigned short,
32705 vector unsigned short);
32706 vector signed int vec_mule (vector signed short, vector signed short);
32708 vector signed int vec_vmulesh (vector signed short,
32709 vector signed short);
32711 vector unsigned int vec_vmuleuh (vector unsigned short,
32712 vector unsigned short);
32714 vector signed short vec_vmulesb (vector signed char,
32715 vector signed char);
32717 vector unsigned short vec_vmuleub (vector unsigned char,
32718 vector unsigned char);
32720 vector unsigned short vec_mulo (vector unsigned char,
32721 vector unsigned char);
32722 vector signed short vec_mulo (vector signed char, vector signed char);
32723 vector unsigned int vec_mulo (vector unsigned short,
32724 vector unsigned short);
32725 vector signed int vec_mulo (vector signed short, vector signed short);
32727 vector signed int vec_vmulosh (vector signed short,
32728 vector signed short);
32730 vector unsigned int vec_vmulouh (vector unsigned short,
32731 vector unsigned short);
32733 vector signed short vec_vmulosb (vector signed char,
32734 vector signed char);
32736 vector unsigned short vec_vmuloub (vector unsigned char,
32737 vector unsigned char);
32739 vector float vec_nmsub (vector float, vector float, vector float);
32741 vector float vec_nor (vector float, vector float);
32742 vector signed int vec_nor (vector signed int, vector signed int);
32743 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
32744 vector bool int vec_nor (vector bool int, vector bool int);
32745 vector signed short vec_nor (vector signed short, vector signed short);
32746 vector unsigned short vec_nor (vector unsigned short,
32747 vector unsigned short);
32748 vector bool short vec_nor (vector bool short, vector bool short);
32749 vector signed char vec_nor (vector signed char, vector signed char);
32750 vector unsigned char vec_nor (vector unsigned char,
32751 vector unsigned char);
32752 vector bool char vec_nor (vector bool char, vector bool char);
32754 vector float vec_or (vector float, vector float);
32755 vector float vec_or (vector float, vector bool int);
32756 vector float vec_or (vector bool int, vector float);
32757 vector bool int vec_or (vector bool int, vector bool int);
32758 vector signed int vec_or (vector bool int, vector signed int);
32759 vector signed int vec_or (vector signed int, vector bool int);
32760 vector signed int vec_or (vector signed int, vector signed int);
32761 vector unsigned int vec_or (vector bool int, vector unsigned int);
32762 vector unsigned int vec_or (vector unsigned int, vector bool int);
32763 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
32764 vector bool short vec_or (vector bool short, vector bool short);
32765 vector signed short vec_or (vector bool short, vector signed short);
32766 vector signed short vec_or (vector signed short, vector bool short);
32767 vector signed short vec_or (vector signed short, vector signed short);
32768 vector unsigned short vec_or (vector bool short, vector unsigned short);
32769 vector unsigned short vec_or (vector unsigned short, vector bool short);
32770 vector unsigned short vec_or (vector unsigned short,
32771 vector unsigned short);
32772 vector signed char vec_or (vector bool char, vector signed char);
32773 vector bool char vec_or (vector bool char, vector bool char);
32774 vector signed char vec_or (vector signed char, vector bool char);
32775 vector signed char vec_or (vector signed char, vector signed char);
32776 vector unsigned char vec_or (vector bool char, vector unsigned char);
32777 vector unsigned char vec_or (vector unsigned char, vector bool char);
32778 vector unsigned char vec_or (vector unsigned char,
32779 vector unsigned char);
32781 vector signed char vec_pack (vector signed short, vector signed short);
32782 vector unsigned char vec_pack (vector unsigned short,
32783 vector unsigned short);
32784 vector bool char vec_pack (vector bool short, vector bool short);
32785 vector signed short vec_pack (vector signed int, vector signed int);
32786 vector unsigned short vec_pack (vector unsigned int,
32787 vector unsigned int);
32788 vector bool short vec_pack (vector bool int, vector bool int);
32790 vector bool short vec_vpkuwum (vector bool int, vector bool int);
32791 vector signed short vec_vpkuwum (vector signed int, vector signed int);
32792 vector unsigned short vec_vpkuwum (vector unsigned int,
32793 vector unsigned int);
32795 vector bool char vec_vpkuhum (vector bool short, vector bool short);
32796 vector signed char vec_vpkuhum (vector signed short,
32797 vector signed short);
32798 vector unsigned char vec_vpkuhum (vector unsigned short,
32799 vector unsigned short);
32801 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
32803 vector unsigned char vec_packs (vector unsigned short,
32804 vector unsigned short);
32805 vector signed char vec_packs (vector signed short, vector signed short);
32806 vector unsigned short vec_packs (vector unsigned int,
32807 vector unsigned int);
32808 vector signed short vec_packs (vector signed int, vector signed int);
32810 vector signed short vec_vpkswss (vector signed int, vector signed int);
32812 vector unsigned short vec_vpkuwus (vector unsigned int,
32813 vector unsigned int);
32815 vector signed char vec_vpkshss (vector signed short,
32816 vector signed short);
32818 vector unsigned char vec_vpkuhus (vector unsigned short,
32819 vector unsigned short);
32821 vector unsigned char vec_packsu (vector unsigned short,
32822 vector unsigned short);
32823 vector unsigned char vec_packsu (vector signed short,
32824 vector signed short);
32825 vector unsigned short vec_packsu (vector unsigned int,
32826 vector unsigned int);
32827 vector unsigned short vec_packsu (vector signed int, vector signed int);
32829 vector unsigned short vec_vpkswus (vector signed int,
32830 vector signed int);
32832 vector unsigned char vec_vpkshus (vector signed short,
32833 vector signed short);
32835 vector float vec_perm (vector float,
32837 vector unsigned char);
32838 vector signed int vec_perm (vector signed int,
32840 vector unsigned char);
32841 vector unsigned int vec_perm (vector unsigned int,
32842 vector unsigned int,
32843 vector unsigned char);
32844 vector bool int vec_perm (vector bool int,
32846 vector unsigned char);
32847 vector signed short vec_perm (vector signed short,
32848 vector signed short,
32849 vector unsigned char);
32850 vector unsigned short vec_perm (vector unsigned short,
32851 vector unsigned short,
32852 vector unsigned char);
32853 vector bool short vec_perm (vector bool short,
32855 vector unsigned char);
32856 vector pixel vec_perm (vector pixel,
32858 vector unsigned char);
32859 vector signed char vec_perm (vector signed char,
32860 vector signed char,
32861 vector unsigned char);
32862 vector unsigned char vec_perm (vector unsigned char,
32863 vector unsigned char,
32864 vector unsigned char);
32865 vector bool char vec_perm (vector bool char,
32867 vector unsigned char);
32869 vector float vec_re (vector float);
32871 vector signed char vec_rl (vector signed char,
32872 vector unsigned char);
32873 vector unsigned char vec_rl (vector unsigned char,
32874 vector unsigned char);
32875 vector signed short vec_rl (vector signed short, vector unsigned short);
32876 vector unsigned short vec_rl (vector unsigned short,
32877 vector unsigned short);
32878 vector signed int vec_rl (vector signed int, vector unsigned int);
32879 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
32881 vector signed int vec_vrlw (vector signed int, vector unsigned int);
32882 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
32884 vector signed short vec_vrlh (vector signed short,
32885 vector unsigned short);
32886 vector unsigned short vec_vrlh (vector unsigned short,
32887 vector unsigned short);
32889 vector signed char vec_vrlb (vector signed char, vector unsigned char);
32890 vector unsigned char vec_vrlb (vector unsigned char,
32891 vector unsigned char);
32893 vector float vec_round (vector float);
32895 vector float vec_rsqrte (vector float);
32897 vector float vec_sel (vector float, vector float, vector bool int);
32898 vector float vec_sel (vector float, vector float, vector unsigned int);
32899 vector signed int vec_sel (vector signed int,
32902 vector signed int vec_sel (vector signed int,
32904 vector unsigned int);
32905 vector unsigned int vec_sel (vector unsigned int,
32906 vector unsigned int,
32908 vector unsigned int vec_sel (vector unsigned int,
32909 vector unsigned int,
32910 vector unsigned int);
32911 vector bool int vec_sel (vector bool int,
32914 vector bool int vec_sel (vector bool int,
32916 vector unsigned int);
32917 vector signed short vec_sel (vector signed short,
32918 vector signed short,
32919 vector bool short);
32920 vector signed short vec_sel (vector signed short,
32921 vector signed short,
32922 vector unsigned short);
32923 vector unsigned short vec_sel (vector unsigned short,
32924 vector unsigned short,
32925 vector bool short);
32926 vector unsigned short vec_sel (vector unsigned short,
32927 vector unsigned short,
32928 vector unsigned short);
32929 vector bool short vec_sel (vector bool short,
32931 vector bool short);
32932 vector bool short vec_sel (vector bool short,
32934 vector unsigned short);
32935 vector signed char vec_sel (vector signed char,
32936 vector signed char,
32938 vector signed char vec_sel (vector signed char,
32939 vector signed char,
32940 vector unsigned char);
32941 vector unsigned char vec_sel (vector unsigned char,
32942 vector unsigned char,
32944 vector unsigned char vec_sel (vector unsigned char,
32945 vector unsigned char,
32946 vector unsigned char);
32947 vector bool char vec_sel (vector bool char,
32950 vector bool char vec_sel (vector bool char,
32952 vector unsigned char);
32954 vector signed char vec_sl (vector signed char,
32955 vector unsigned char);
32956 vector unsigned char vec_sl (vector unsigned char,
32957 vector unsigned char);
32958 vector signed short vec_sl (vector signed short, vector unsigned short);
32959 vector unsigned short vec_sl (vector unsigned short,
32960 vector unsigned short);
32961 vector signed int vec_sl (vector signed int, vector unsigned int);
32962 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
32964 vector signed int vec_vslw (vector signed int, vector unsigned int);
32965 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
32967 vector signed short vec_vslh (vector signed short,
32968 vector unsigned short);
32969 vector unsigned short vec_vslh (vector unsigned short,
32970 vector unsigned short);
32972 vector signed char vec_vslb (vector signed char, vector unsigned char);
32973 vector unsigned char vec_vslb (vector unsigned char,
32974 vector unsigned char);
32976 vector float vec_sld (vector float, vector float, const int);
32977 vector signed int vec_sld (vector signed int,
32980 vector unsigned int vec_sld (vector unsigned int,
32981 vector unsigned int,
32983 vector bool int vec_sld (vector bool int,
32986 vector signed short vec_sld (vector signed short,
32987 vector signed short,
32989 vector unsigned short vec_sld (vector unsigned short,
32990 vector unsigned short,
32992 vector bool short vec_sld (vector bool short,
32995 vector pixel vec_sld (vector pixel,
32998 vector signed char vec_sld (vector signed char,
32999 vector signed char,
33001 vector unsigned char vec_sld (vector unsigned char,
33002 vector unsigned char,
33004 vector bool char vec_sld (vector bool char,
33008 vector signed int vec_sll (vector signed int,
33009 vector unsigned int);
33010 vector signed int vec_sll (vector signed int,
33011 vector unsigned short);
33012 vector signed int vec_sll (vector signed int,
33013 vector unsigned char);
33014 vector unsigned int vec_sll (vector unsigned int,
33015 vector unsigned int);
33016 vector unsigned int vec_sll (vector unsigned int,
33017 vector unsigned short);
33018 vector unsigned int vec_sll (vector unsigned int,
33019 vector unsigned char);
33020 vector bool int vec_sll (vector bool int,
33021 vector unsigned int);
33022 vector bool int vec_sll (vector bool int,
33023 vector unsigned short);
33024 vector bool int vec_sll (vector bool int,
33025 vector unsigned char);
33026 vector signed short vec_sll (vector signed short,
33027 vector unsigned int);
33028 vector signed short vec_sll (vector signed short,
33029 vector unsigned short);
33030 vector signed short vec_sll (vector signed short,
33031 vector unsigned char);
33032 vector unsigned short vec_sll (vector unsigned short,
33033 vector unsigned int);
33034 vector unsigned short vec_sll (vector unsigned short,
33035 vector unsigned short);
33036 vector unsigned short vec_sll (vector unsigned short,
33037 vector unsigned char);
33038 vector bool short vec_sll (vector bool short, vector unsigned int);
33039 vector bool short vec_sll (vector bool short, vector unsigned short);
33040 vector bool short vec_sll (vector bool short, vector unsigned char);
33041 vector pixel vec_sll (vector pixel, vector unsigned int);
33042 vector pixel vec_sll (vector pixel, vector unsigned short);
33043 vector pixel vec_sll (vector pixel, vector unsigned char);
33044 vector signed char vec_sll (vector signed char, vector unsigned int);
33045 vector signed char vec_sll (vector signed char, vector unsigned short);
33046 vector signed char vec_sll (vector signed char, vector unsigned char);
33047 vector unsigned char vec_sll (vector unsigned char,
33048 vector unsigned int);
33049 vector unsigned char vec_sll (vector unsigned char,
33050 vector unsigned short);
33051 vector unsigned char vec_sll (vector unsigned char,
33052 vector unsigned char);
33053 vector bool char vec_sll (vector bool char, vector unsigned int);
33054 vector bool char vec_sll (vector bool char, vector unsigned short);
33055 vector bool char vec_sll (vector bool char, vector unsigned char);
33057 vector float vec_slo (vector float, vector signed char);
33058 vector float vec_slo (vector float, vector unsigned char);
33059 vector signed int vec_slo (vector signed int, vector signed char);
33060 vector signed int vec_slo (vector signed int, vector unsigned char);
33061 vector unsigned int vec_slo (vector unsigned int, vector signed char);
33062 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
33063 vector signed short vec_slo (vector signed short, vector signed char);
33064 vector signed short vec_slo (vector signed short, vector unsigned char);
33065 vector unsigned short vec_slo (vector unsigned short,
33066 vector signed char);
33067 vector unsigned short vec_slo (vector unsigned short,
33068 vector unsigned char);
33069 vector pixel vec_slo (vector pixel, vector signed char);
33070 vector pixel vec_slo (vector pixel, vector unsigned char);
33071 vector signed char vec_slo (vector signed char, vector signed char);
33072 vector signed char vec_slo (vector signed char, vector unsigned char);
33073 vector unsigned char vec_slo (vector unsigned char, vector signed char);
33074 vector unsigned char vec_slo (vector unsigned char,
33075 vector unsigned char);
33077 vector signed char vec_splat (vector signed char, const int);
33078 vector unsigned char vec_splat (vector unsigned char, const int);
33079 vector bool char vec_splat (vector bool char, const int);
33080 vector signed short vec_splat (vector signed short, const int);
33081 vector unsigned short vec_splat (vector unsigned short, const int);
33082 vector bool short vec_splat (vector bool short, const int);
33083 vector pixel vec_splat (vector pixel, const int);
33084 vector float vec_splat (vector float, const int);
33085 vector signed int vec_splat (vector signed int, const int);
33086 vector unsigned int vec_splat (vector unsigned int, const int);
33087 vector bool int vec_splat (vector bool int, const int);
33089 vector float vec_vspltw (vector float, const int);
33090 vector signed int vec_vspltw (vector signed int, const int);
33091 vector unsigned int vec_vspltw (vector unsigned int, const int);
33092 vector bool int vec_vspltw (vector bool int, const int);
33094 vector bool short vec_vsplth (vector bool short, const int);
33095 vector signed short vec_vsplth (vector signed short, const int);
33096 vector unsigned short vec_vsplth (vector unsigned short, const int);
33097 vector pixel vec_vsplth (vector pixel, const int);
33099 vector signed char vec_vspltb (vector signed char, const int);
33100 vector unsigned char vec_vspltb (vector unsigned char, const int);
33101 vector bool char vec_vspltb (vector bool char, const int);
33103 vector signed char vec_splat_s8 (const int);
33105 vector signed short vec_splat_s16 (const int);
33107 vector signed int vec_splat_s32 (const int);
33109 vector unsigned char vec_splat_u8 (const int);
33111 vector unsigned short vec_splat_u16 (const int);
33113 vector unsigned int vec_splat_u32 (const int);
33115 vector signed char vec_sr (vector signed char, vector unsigned char);
33116 vector unsigned char vec_sr (vector unsigned char,
33117 vector unsigned char);
33118 vector signed short vec_sr (vector signed short,
33119 vector unsigned short);
33120 vector unsigned short vec_sr (vector unsigned short,
33121 vector unsigned short);
33122 vector signed int vec_sr (vector signed int, vector unsigned int);
33123 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
33125 vector signed int vec_vsrw (vector signed int, vector unsigned int);
33126 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
33128 vector signed short vec_vsrh (vector signed short,
33129 vector unsigned short);
33130 vector unsigned short vec_vsrh (vector unsigned short,
33131 vector unsigned short);
33133 vector signed char vec_vsrb (vector signed char, vector unsigned char);
33134 vector unsigned char vec_vsrb (vector unsigned char,
33135 vector unsigned char);
33137 vector signed char vec_sra (vector signed char, vector unsigned char);
33138 vector unsigned char vec_sra (vector unsigned char,
33139 vector unsigned char);
33140 vector signed short vec_sra (vector signed short,
33141 vector unsigned short);
33142 vector unsigned short vec_sra (vector unsigned short,
33143 vector unsigned short);
33144 vector signed int vec_sra (vector signed int, vector unsigned int);
33145 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
33147 vector signed int vec_vsraw (vector signed int, vector unsigned int);
33148 vector unsigned int vec_vsraw (vector unsigned int,
33149 vector unsigned int);
33151 vector signed short vec_vsrah (vector signed short,
33152 vector unsigned short);
33153 vector unsigned short vec_vsrah (vector unsigned short,
33154 vector unsigned short);
33156 vector signed char vec_vsrab (vector signed char, vector unsigned char);
33157 vector unsigned char vec_vsrab (vector unsigned char,
33158 vector unsigned char);
33160 vector signed int vec_srl (vector signed int, vector unsigned int);
33161 vector signed int vec_srl (vector signed int, vector unsigned short);
33162 vector signed int vec_srl (vector signed int, vector unsigned char);
33163 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
33164 vector unsigned int vec_srl (vector unsigned int,
33165 vector unsigned short);
33166 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
33167 vector bool int vec_srl (vector bool int, vector unsigned int);
33168 vector bool int vec_srl (vector bool int, vector unsigned short);
33169 vector bool int vec_srl (vector bool int, vector unsigned char);
33170 vector signed short vec_srl (vector signed short, vector unsigned int);
33171 vector signed short vec_srl (vector signed short,
33172 vector unsigned short);
33173 vector signed short vec_srl (vector signed short, vector unsigned char);
33174 vector unsigned short vec_srl (vector unsigned short,
33175 vector unsigned int);
33176 vector unsigned short vec_srl (vector unsigned short,
33177 vector unsigned short);
33178 vector unsigned short vec_srl (vector unsigned short,
33179 vector unsigned char);
33180 vector bool short vec_srl (vector bool short, vector unsigned int);
33181 vector bool short vec_srl (vector bool short, vector unsigned short);
33182 vector bool short vec_srl (vector bool short, vector unsigned char);
33183 vector pixel vec_srl (vector pixel, vector unsigned int);
33184 vector pixel vec_srl (vector pixel, vector unsigned short);
33185 vector pixel vec_srl (vector pixel, vector unsigned char);
33186 vector signed char vec_srl (vector signed char, vector unsigned int);
33187 vector signed char vec_srl (vector signed char, vector unsigned short);
33188 vector signed char vec_srl (vector signed char, vector unsigned char);
33189 vector unsigned char vec_srl (vector unsigned char,
33190 vector unsigned int);
33191 vector unsigned char vec_srl (vector unsigned char,
33192 vector unsigned short);
33193 vector unsigned char vec_srl (vector unsigned char,
33194 vector unsigned char);
33195 vector bool char vec_srl (vector bool char, vector unsigned int);
33196 vector bool char vec_srl (vector bool char, vector unsigned short);
33197 vector bool char vec_srl (vector bool char, vector unsigned char);
33199 vector float vec_sro (vector float, vector signed char);
33200 vector float vec_sro (vector float, vector unsigned char);
33201 vector signed int vec_sro (vector signed int, vector signed char);
33202 vector signed int vec_sro (vector signed int, vector unsigned char);
33203 vector unsigned int vec_sro (vector unsigned int, vector signed char);
33204 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
33205 vector signed short vec_sro (vector signed short, vector signed char);
33206 vector signed short vec_sro (vector signed short, vector unsigned char);
33207 vector unsigned short vec_sro (vector unsigned short,
33208 vector signed char);
33209 vector unsigned short vec_sro (vector unsigned short,
33210 vector unsigned char);
33211 vector pixel vec_sro (vector pixel, vector signed char);
33212 vector pixel vec_sro (vector pixel, vector unsigned char);
33213 vector signed char vec_sro (vector signed char, vector signed char);
33214 vector signed char vec_sro (vector signed char, vector unsigned char);
33215 vector unsigned char vec_sro (vector unsigned char, vector signed char);
33216 vector unsigned char vec_sro (vector unsigned char,
33217 vector unsigned char);
33219 void vec_st (vector float, int, vector float *);
33220 void vec_st (vector float, int, float *);
33221 void vec_st (vector signed int, int, vector signed int *);
33222 void vec_st (vector signed int, int, int *);
33223 void vec_st (vector unsigned int, int, vector unsigned int *);
33224 void vec_st (vector unsigned int, int, unsigned int *);
33225 void vec_st (vector bool int, int, vector bool int *);
33226 void vec_st (vector bool int, int, unsigned int *);
33227 void vec_st (vector bool int, int, int *);
33228 void vec_st (vector signed short, int, vector signed short *);
33229 void vec_st (vector signed short, int, short *);
33230 void vec_st (vector unsigned short, int, vector unsigned short *);
33231 void vec_st (vector unsigned short, int, unsigned short *);
33232 void vec_st (vector bool short, int, vector bool short *);
33233 void vec_st (vector bool short, int, unsigned short *);
33234 void vec_st (vector pixel, int, vector pixel *);
33235 void vec_st (vector pixel, int, unsigned short *);
33236 void vec_st (vector pixel, int, short *);
33237 void vec_st (vector bool short, int, short *);
33238 void vec_st (vector signed char, int, vector signed char *);
33239 void vec_st (vector signed char, int, signed char *);
33240 void vec_st (vector unsigned char, int, vector unsigned char *);
33241 void vec_st (vector unsigned char, int, unsigned char *);
33242 void vec_st (vector bool char, int, vector bool char *);
33243 void vec_st (vector bool char, int, unsigned char *);
33244 void vec_st (vector bool char, int, signed char *);
33246 void vec_ste (vector signed char, int, signed char *);
33247 void vec_ste (vector unsigned char, int, unsigned char *);
33248 void vec_ste (vector bool char, int, signed char *);
33249 void vec_ste (vector bool char, int, unsigned char *);
33250 void vec_ste (vector signed short, int, short *);
33251 void vec_ste (vector unsigned short, int, unsigned short *);
33252 void vec_ste (vector bool short, int, short *);
33253 void vec_ste (vector bool short, int, unsigned short *);
33254 void vec_ste (vector pixel, int, short *);
33255 void vec_ste (vector pixel, int, unsigned short *);
33256 void vec_ste (vector float, int, float *);
33257 void vec_ste (vector signed int, int, int *);
33258 void vec_ste (vector unsigned int, int, unsigned int *);
33259 void vec_ste (vector bool int, int, int *);
33260 void vec_ste (vector bool int, int, unsigned int *);
33262 void vec_stvewx (vector float, int, float *);
33263 void vec_stvewx (vector signed int, int, int *);
33264 void vec_stvewx (vector unsigned int, int, unsigned int *);
33265 void vec_stvewx (vector bool int, int, int *);
33266 void vec_stvewx (vector bool int, int, unsigned int *);
33268 void vec_stvehx (vector signed short, int, short *);
33269 void vec_stvehx (vector unsigned short, int, unsigned short *);
33270 void vec_stvehx (vector bool short, int, short *);
33271 void vec_stvehx (vector bool short, int, unsigned short *);
33272 void vec_stvehx (vector pixel, int, short *);
33273 void vec_stvehx (vector pixel, int, unsigned short *);
33275 void vec_stvebx (vector signed char, int, signed char *);
33276 void vec_stvebx (vector unsigned char, int, unsigned char *);
33277 void vec_stvebx (vector bool char, int, signed char *);
33278 void vec_stvebx (vector bool char, int, unsigned char *);
33280 void vec_stl (vector float, int, vector float *);
33281 void vec_stl (vector float, int, float *);
33282 void vec_stl (vector signed int, int, vector signed int *);
33283 void vec_stl (vector signed int, int, int *);
33284 void vec_stl (vector unsigned int, int, vector unsigned int *);
33285 void vec_stl (vector unsigned int, int, unsigned int *);
33286 void vec_stl (vector bool int, int, vector bool int *);
33287 void vec_stl (vector bool int, int, unsigned int *);
33288 void vec_stl (vector bool int, int, int *);
33289 void vec_stl (vector signed short, int, vector signed short *);
33290 void vec_stl (vector signed short, int, short *);
33291 void vec_stl (vector unsigned short, int, vector unsigned short *);
33292 void vec_stl (vector unsigned short, int, unsigned short *);
33293 void vec_stl (vector bool short, int, vector bool short *);
33294 void vec_stl (vector bool short, int, unsigned short *);
33295 void vec_stl (vector bool short, int, short *);
33296 void vec_stl (vector pixel, int, vector pixel *);
33297 void vec_stl (vector pixel, int, unsigned short *);
33298 void vec_stl (vector pixel, int, short *);
33299 void vec_stl (vector signed char, int, vector signed char *);
33300 void vec_stl (vector signed char, int, signed char *);
33301 void vec_stl (vector unsigned char, int, vector unsigned char *);
33302 void vec_stl (vector unsigned char, int, unsigned char *);
33303 void vec_stl (vector bool char, int, vector bool char *);
33304 void vec_stl (vector bool char, int, unsigned char *);
33305 void vec_stl (vector bool char, int, signed char *);
33307 vector signed char vec_sub (vector bool char, vector signed char);
33308 vector signed char vec_sub (vector signed char, vector bool char);
33309 vector signed char vec_sub (vector signed char, vector signed char);
33310 vector unsigned char vec_sub (vector bool char, vector unsigned char);
33311 vector unsigned char vec_sub (vector unsigned char, vector bool char);
33312 vector unsigned char vec_sub (vector unsigned char,
33313 vector unsigned char);
33314 vector signed short vec_sub (vector bool short, vector signed short);
33315 vector signed short vec_sub (vector signed short, vector bool short);
33316 vector signed short vec_sub (vector signed short, vector signed short);
33317 vector unsigned short vec_sub (vector bool short,
33318 vector unsigned short);
33319 vector unsigned short vec_sub (vector unsigned short,
33320 vector bool short);
33321 vector unsigned short vec_sub (vector unsigned short,
33322 vector unsigned short);
33323 vector signed int vec_sub (vector bool int, vector signed int);
33324 vector signed int vec_sub (vector signed int, vector bool int);
33325 vector signed int vec_sub (vector signed int, vector signed int);
33326 vector unsigned int vec_sub (vector bool int, vector unsigned int);
33327 vector unsigned int vec_sub (vector unsigned int, vector bool int);
33328 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
33329 vector float vec_sub (vector float, vector float);
33331 vector float vec_vsubfp (vector float, vector float);
33333 vector signed int vec_vsubuwm (vector bool int, vector signed int);
33334 vector signed int vec_vsubuwm (vector signed int, vector bool int);
33335 vector signed int vec_vsubuwm (vector signed int, vector signed int);
33336 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
33337 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
33338 vector unsigned int vec_vsubuwm (vector unsigned int,
33339 vector unsigned int);
33341 vector signed short vec_vsubuhm (vector bool short,
33342 vector signed short);
33343 vector signed short vec_vsubuhm (vector signed short,
33344 vector bool short);
33345 vector signed short vec_vsubuhm (vector signed short,
33346 vector signed short);
33347 vector unsigned short vec_vsubuhm (vector bool short,
33348 vector unsigned short);
33349 vector unsigned short vec_vsubuhm (vector unsigned short,
33350 vector bool short);
33351 vector unsigned short vec_vsubuhm (vector unsigned short,
33352 vector unsigned short);
33354 vector signed char vec_vsububm (vector bool char, vector signed char);
33355 vector signed char vec_vsububm (vector signed char, vector bool char);
33356 vector signed char vec_vsububm (vector signed char, vector signed char);
33357 vector unsigned char vec_vsububm (vector bool char,
33358 vector unsigned char);
33359 vector unsigned char vec_vsububm (vector unsigned char,
33361 vector unsigned char vec_vsububm (vector unsigned char,
33362 vector unsigned char);
33364 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
33366 vector unsigned char vec_subs (vector bool char, vector unsigned char);
33367 vector unsigned char vec_subs (vector unsigned char, vector bool char);
33368 vector unsigned char vec_subs (vector unsigned char,
33369 vector unsigned char);
33370 vector signed char vec_subs (vector bool char, vector signed char);
33371 vector signed char vec_subs (vector signed char, vector bool char);
33372 vector signed char vec_subs (vector signed char, vector signed char);
33373 vector unsigned short vec_subs (vector bool short,
33374 vector unsigned short);
33375 vector unsigned short vec_subs (vector unsigned short,
33376 vector bool short);
33377 vector unsigned short vec_subs (vector unsigned short,
33378 vector unsigned short);
33379 vector signed short vec_subs (vector bool short, vector signed short);
33380 vector signed short vec_subs (vector signed short, vector bool short);
33381 vector signed short vec_subs (vector signed short, vector signed short);
33382 vector unsigned int vec_subs (vector bool int, vector unsigned int);
33383 vector unsigned int vec_subs (vector unsigned int, vector bool int);
33384 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
33385 vector signed int vec_subs (vector bool int, vector signed int);
33386 vector signed int vec_subs (vector signed int, vector bool int);
33387 vector signed int vec_subs (vector signed int, vector signed int);
33389 vector signed int vec_vsubsws (vector bool int, vector signed int);
33390 vector signed int vec_vsubsws (vector signed int, vector bool int);
33391 vector signed int vec_vsubsws (vector signed int, vector signed int);
33393 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
33394 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
33395 vector unsigned int vec_vsubuws (vector unsigned int,
33396 vector unsigned int);
33398 vector signed short vec_vsubshs (vector bool short,
33399 vector signed short);
33400 vector signed short vec_vsubshs (vector signed short,
33401 vector bool short);
33402 vector signed short vec_vsubshs (vector signed short,
33403 vector signed short);
33405 vector unsigned short vec_vsubuhs (vector bool short,
33406 vector unsigned short);
33407 vector unsigned short vec_vsubuhs (vector unsigned short,
33408 vector bool short);
33409 vector unsigned short vec_vsubuhs (vector unsigned short,
33410 vector unsigned short);
33412 vector signed char vec_vsubsbs (vector bool char, vector signed char);
33413 vector signed char vec_vsubsbs (vector signed char, vector bool char);
33414 vector signed char vec_vsubsbs (vector signed char, vector signed char);
33416 vector unsigned char vec_vsububs (vector bool char,
33417 vector unsigned char);
33418 vector unsigned char vec_vsububs (vector unsigned char,
33420 vector unsigned char vec_vsububs (vector unsigned char,
33421 vector unsigned char);
33423 vector unsigned int vec_sum4s (vector unsigned char,
33424 vector unsigned int);
33425 vector signed int vec_sum4s (vector signed char, vector signed int);
33426 vector signed int vec_sum4s (vector signed short, vector signed int);
33428 vector signed int vec_vsum4shs (vector signed short, vector signed int);
33430 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
33432 vector unsigned int vec_vsum4ubs (vector unsigned char,
33433 vector unsigned int);
33435 vector signed int vec_sum2s (vector signed int, vector signed int);
33437 vector signed int vec_sums (vector signed int, vector signed int);
33439 vector float vec_trunc (vector float);
33441 vector signed short vec_unpackh (vector signed char);
33442 vector bool short vec_unpackh (vector bool char);
33443 vector signed int vec_unpackh (vector signed short);
33444 vector bool int vec_unpackh (vector bool short);
33445 vector unsigned int vec_unpackh (vector pixel);
33447 vector bool int vec_vupkhsh (vector bool short);
33448 vector signed int vec_vupkhsh (vector signed short);
33450 vector unsigned int vec_vupkhpx (vector pixel);
33452 vector bool short vec_vupkhsb (vector bool char);
33453 vector signed short vec_vupkhsb (vector signed char);
33455 vector signed short vec_unpackl (vector signed char);
33456 vector bool short vec_unpackl (vector bool char);
33457 vector unsigned int vec_unpackl (vector pixel);
33458 vector signed int vec_unpackl (vector signed short);
33459 vector bool int vec_unpackl (vector bool short);
33461 vector unsigned int vec_vupklpx (vector pixel);
33463 vector bool int vec_vupklsh (vector bool short);
33464 vector signed int vec_vupklsh (vector signed short);
33466 vector bool short vec_vupklsb (vector bool char);
33467 vector signed short vec_vupklsb (vector signed char);
33469 vector float vec_xor (vector float, vector float);
33470 vector float vec_xor (vector float, vector bool int);
33471 vector float vec_xor (vector bool int, vector float);
33472 vector bool int vec_xor (vector bool int, vector bool int);
33473 vector signed int vec_xor (vector bool int, vector signed int);
33474 vector signed int vec_xor (vector signed int, vector bool int);
33475 vector signed int vec_xor (vector signed int, vector signed int);
33476 vector unsigned int vec_xor (vector bool int, vector unsigned int);
33477 vector unsigned int vec_xor (vector unsigned int, vector bool int);
33478 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
33479 vector bool short vec_xor (vector bool short, vector bool short);
33480 vector signed short vec_xor (vector bool short, vector signed short);
33481 vector signed short vec_xor (vector signed short, vector bool short);
33482 vector signed short vec_xor (vector signed short, vector signed short);
33483 vector unsigned short vec_xor (vector bool short,
33484 vector unsigned short);
33485 vector unsigned short vec_xor (vector unsigned short,
33486 vector bool short);
33487 vector unsigned short vec_xor (vector unsigned short,
33488 vector unsigned short);
33489 vector signed char vec_xor (vector bool char, vector signed char);
33490 vector bool char vec_xor (vector bool char, vector bool char);
33491 vector signed char vec_xor (vector signed char, vector bool char);
33492 vector signed char vec_xor (vector signed char, vector signed char);
33493 vector unsigned char vec_xor (vector bool char, vector unsigned char);
33494 vector unsigned char vec_xor (vector unsigned char, vector bool char);
33495 vector unsigned char vec_xor (vector unsigned char,
33496 vector unsigned char);
33498 int vec_all_eq (vector signed char, vector bool char);
33499 int vec_all_eq (vector signed char, vector signed char);
33500 int vec_all_eq (vector unsigned char, vector bool char);
33501 int vec_all_eq (vector unsigned char, vector unsigned char);
33502 int vec_all_eq (vector bool char, vector bool char);
33503 int vec_all_eq (vector bool char, vector unsigned char);
33504 int vec_all_eq (vector bool char, vector signed char);
33505 int vec_all_eq (vector signed short, vector bool short);
33506 int vec_all_eq (vector signed short, vector signed short);
33507 int vec_all_eq (vector unsigned short, vector bool short);
33508 int vec_all_eq (vector unsigned short, vector unsigned short);
33509 int vec_all_eq (vector bool short, vector bool short);
33510 int vec_all_eq (vector bool short, vector unsigned short);
33511 int vec_all_eq (vector bool short, vector signed short);
33512 int vec_all_eq (vector pixel, vector pixel);
33513 int vec_all_eq (vector signed int, vector bool int);
33514 int vec_all_eq (vector signed int, vector signed int);
33515 int vec_all_eq (vector unsigned int, vector bool int);
33516 int vec_all_eq (vector unsigned int, vector unsigned int);
33517 int vec_all_eq (vector bool int, vector bool int);
33518 int vec_all_eq (vector bool int, vector unsigned int);
33519 int vec_all_eq (vector bool int, vector signed int);
33520 int vec_all_eq (vector float, vector float);
33522 int vec_all_ge (vector bool char, vector unsigned char);
33523 int vec_all_ge (vector unsigned char, vector bool char);
33524 int vec_all_ge (vector unsigned char, vector unsigned char);
33525 int vec_all_ge (vector bool char, vector signed char);
33526 int vec_all_ge (vector signed char, vector bool char);
33527 int vec_all_ge (vector signed char, vector signed char);
33528 int vec_all_ge (vector bool short, vector unsigned short);
33529 int vec_all_ge (vector unsigned short, vector bool short);
33530 int vec_all_ge (vector unsigned short, vector unsigned short);
33531 int vec_all_ge (vector signed short, vector signed short);
33532 int vec_all_ge (vector bool short, vector signed short);
33533 int vec_all_ge (vector signed short, vector bool short);
33534 int vec_all_ge (vector bool int, vector unsigned int);
33535 int vec_all_ge (vector unsigned int, vector bool int);
33536 int vec_all_ge (vector unsigned int, vector unsigned int);
33537 int vec_all_ge (vector bool int, vector signed int);
33538 int vec_all_ge (vector signed int, vector bool int);
33539 int vec_all_ge (vector signed int, vector signed int);
33540 int vec_all_ge (vector float, vector float);
33542 int vec_all_gt (vector bool char, vector unsigned char);
33543 int vec_all_gt (vector unsigned char, vector bool char);
33544 int vec_all_gt (vector unsigned char, vector unsigned char);
33545 int vec_all_gt (vector bool char, vector signed char);
33546 int vec_all_gt (vector signed char, vector bool char);
33547 int vec_all_gt (vector signed char, vector signed char);
33548 int vec_all_gt (vector bool short, vector unsigned short);
33549 int vec_all_gt (vector unsigned short, vector bool short);
33550 int vec_all_gt (vector unsigned short, vector unsigned short);
33551 int vec_all_gt (vector bool short, vector signed short);
33552 int vec_all_gt (vector signed short, vector bool short);
33553 int vec_all_gt (vector signed short, vector signed short);
33554 int vec_all_gt (vector bool int, vector unsigned int);
33555 int vec_all_gt (vector unsigned int, vector bool int);
33556 int vec_all_gt (vector unsigned int, vector unsigned int);
33557 int vec_all_gt (vector bool int, vector signed int);
33558 int vec_all_gt (vector signed int, vector bool int);
33559 int vec_all_gt (vector signed int, vector signed int);
33560 int vec_all_gt (vector float, vector float);
33562 int vec_all_in (vector float, vector float);
33564 int vec_all_le (vector bool char, vector unsigned char);
33565 int vec_all_le (vector unsigned char, vector bool char);
33566 int vec_all_le (vector unsigned char, vector unsigned char);
33567 int vec_all_le (vector bool char, vector signed char);
33568 int vec_all_le (vector signed char, vector bool char);
33569 int vec_all_le (vector signed char, vector signed char);
33570 int vec_all_le (vector bool short, vector unsigned short);
33571 int vec_all_le (vector unsigned short, vector bool short);
33572 int vec_all_le (vector unsigned short, vector unsigned short);
33573 int vec_all_le (vector bool short, vector signed short);
33574 int vec_all_le (vector signed short, vector bool short);
33575 int vec_all_le (vector signed short, vector signed short);
33576 int vec_all_le (vector bool int, vector unsigned int);
33577 int vec_all_le (vector unsigned int, vector bool int);
33578 int vec_all_le (vector unsigned int, vector unsigned int);
33579 int vec_all_le (vector bool int, vector signed int);
33580 int vec_all_le (vector signed int, vector bool int);
33581 int vec_all_le (vector signed int, vector signed int);
33582 int vec_all_le (vector float, vector float);
33584 int vec_all_lt (vector bool char, vector unsigned char);
33585 int vec_all_lt (vector unsigned char, vector bool char);
33586 int vec_all_lt (vector unsigned char, vector unsigned char);
33587 int vec_all_lt (vector bool char, vector signed char);
33588 int vec_all_lt (vector signed char, vector bool char);
33589 int vec_all_lt (vector signed char, vector signed char);
33590 int vec_all_lt (vector bool short, vector unsigned short);
33591 int vec_all_lt (vector unsigned short, vector bool short);
33592 int vec_all_lt (vector unsigned short, vector unsigned short);
33593 int vec_all_lt (vector bool short, vector signed short);
33594 int vec_all_lt (vector signed short, vector bool short);
33595 int vec_all_lt (vector signed short, vector signed short);
33596 int vec_all_lt (vector bool int, vector unsigned int);
33597 int vec_all_lt (vector unsigned int, vector bool int);
33598 int vec_all_lt (vector unsigned int, vector unsigned int);
33599 int vec_all_lt (vector bool int, vector signed int);
33600 int vec_all_lt (vector signed int, vector bool int);
33601 int vec_all_lt (vector signed int, vector signed int);
33602 int vec_all_lt (vector float, vector float);
33604 int vec_all_nan (vector float);
33606 int vec_all_ne (vector signed char, vector bool char);
33607 int vec_all_ne (vector signed char, vector signed char);
33608 int vec_all_ne (vector unsigned char, vector bool char);
33609 int vec_all_ne (vector unsigned char, vector unsigned char);
33610 int vec_all_ne (vector bool char, vector bool char);
33611 int vec_all_ne (vector bool char, vector unsigned char);
33612 int vec_all_ne (vector bool char, vector signed char);
33613 int vec_all_ne (vector signed short, vector bool short);
33614 int vec_all_ne (vector signed short, vector signed short);
33615 int vec_all_ne (vector unsigned short, vector bool short);
33616 int vec_all_ne (vector unsigned short, vector unsigned short);
33617 int vec_all_ne (vector bool short, vector bool short);
33618 int vec_all_ne (vector bool short, vector unsigned short);
33619 int vec_all_ne (vector bool short, vector signed short);
33620 int vec_all_ne (vector pixel, vector pixel);
33621 int vec_all_ne (vector signed int, vector bool int);
33622 int vec_all_ne (vector signed int, vector signed int);
33623 int vec_all_ne (vector unsigned int, vector bool int);
33624 int vec_all_ne (vector unsigned int, vector unsigned int);
33625 int vec_all_ne (vector bool int, vector bool int);
33626 int vec_all_ne (vector bool int, vector unsigned int);
33627 int vec_all_ne (vector bool int, vector signed int);
33628 int vec_all_ne (vector float, vector float);
33630 int vec_all_nge (vector float, vector float);
33632 int vec_all_ngt (vector float, vector float);
33634 int vec_all_nle (vector float, vector float);
33636 int vec_all_nlt (vector float, vector float);
33638 int vec_all_numeric (vector float);
33640 int vec_any_eq (vector signed char, vector bool char);
33641 int vec_any_eq (vector signed char, vector signed char);
33642 int vec_any_eq (vector unsigned char, vector bool char);
33643 int vec_any_eq (vector unsigned char, vector unsigned char);
33644 int vec_any_eq (vector bool char, vector bool char);
33645 int vec_any_eq (vector bool char, vector unsigned char);
33646 int vec_any_eq (vector bool char, vector signed char);
33647 int vec_any_eq (vector signed short, vector bool short);
33648 int vec_any_eq (vector signed short, vector signed short);
33649 int vec_any_eq (vector unsigned short, vector bool short);
33650 int vec_any_eq (vector unsigned short, vector unsigned short);
33651 int vec_any_eq (vector bool short, vector bool short);
33652 int vec_any_eq (vector bool short, vector unsigned short);
33653 int vec_any_eq (vector bool short, vector signed short);
33654 int vec_any_eq (vector pixel, vector pixel);
33655 int vec_any_eq (vector signed int, vector bool int);
33656 int vec_any_eq (vector signed int, vector signed int);
33657 int vec_any_eq (vector unsigned int, vector bool int);
33658 int vec_any_eq (vector unsigned int, vector unsigned int);
33659 int vec_any_eq (vector bool int, vector bool int);
33660 int vec_any_eq (vector bool int, vector unsigned int);
33661 int vec_any_eq (vector bool int, vector signed int);
33662 int vec_any_eq (vector float, vector float);
33664 int vec_any_ge (vector signed char, vector bool char);
33665 int vec_any_ge (vector unsigned char, vector bool char);
33666 int vec_any_ge (vector unsigned char, vector unsigned char);
33667 int vec_any_ge (vector signed char, vector signed char);
33668 int vec_any_ge (vector bool char, vector unsigned char);
33669 int vec_any_ge (vector bool char, vector signed char);
33670 int vec_any_ge (vector unsigned short, vector bool short);
33671 int vec_any_ge (vector unsigned short, vector unsigned short);
33672 int vec_any_ge (vector signed short, vector signed short);
33673 int vec_any_ge (vector signed short, vector bool short);
33674 int vec_any_ge (vector bool short, vector unsigned short);
33675 int vec_any_ge (vector bool short, vector signed short);
33676 int vec_any_ge (vector signed int, vector bool int);
33677 int vec_any_ge (vector unsigned int, vector bool int);
33678 int vec_any_ge (vector unsigned int, vector unsigned int);
33679 int vec_any_ge (vector signed int, vector signed int);
33680 int vec_any_ge (vector bool int, vector unsigned int);
33681 int vec_any_ge (vector bool int, vector signed int);
33682 int vec_any_ge (vector float, vector float);
33684 int vec_any_gt (vector bool char, vector unsigned char);
33685 int vec_any_gt (vector unsigned char, vector bool char);
33686 int vec_any_gt (vector unsigned char, vector unsigned char);
33687 int vec_any_gt (vector bool char, vector signed char);
33688 int vec_any_gt (vector signed char, vector bool char);
33689 int vec_any_gt (vector signed char, vector signed char);
33690 int vec_any_gt (vector bool short, vector unsigned short);
33691 int vec_any_gt (vector unsigned short, vector bool short);
33692 int vec_any_gt (vector unsigned short, vector unsigned short);
33693 int vec_any_gt (vector bool short, vector signed short);
33694 int vec_any_gt (vector signed short, vector bool short);
33695 int vec_any_gt (vector signed short, vector signed short);
33696 int vec_any_gt (vector bool int, vector unsigned int);
33697 int vec_any_gt (vector unsigned int, vector bool int);
33698 int vec_any_gt (vector unsigned int, vector unsigned int);
33699 int vec_any_gt (vector bool int, vector signed int);
33700 int vec_any_gt (vector signed int, vector bool int);
33701 int vec_any_gt (vector signed int, vector signed int);
33702 int vec_any_gt (vector float, vector float);
33704 int vec_any_le (vector bool char, vector unsigned char);
33705 int vec_any_le (vector unsigned char, vector bool char);
33706 int vec_any_le (vector unsigned char, vector unsigned char);
33707 int vec_any_le (vector bool char, vector signed char);
33708 int vec_any_le (vector signed char, vector bool char);
33709 int vec_any_le (vector signed char, vector signed char);
33710 int vec_any_le (vector bool short, vector unsigned short);
33711 int vec_any_le (vector unsigned short, vector bool short);
33712 int vec_any_le (vector unsigned short, vector unsigned short);
33713 int vec_any_le (vector bool short, vector signed short);
33714 int vec_any_le (vector signed short, vector bool short);
33715 int vec_any_le (vector signed short, vector signed short);
33716 int vec_any_le (vector bool int, vector unsigned int);
33717 int vec_any_le (vector unsigned int, vector bool int);
33718 int vec_any_le (vector unsigned int, vector unsigned int);
33719 int vec_any_le (vector bool int, vector signed int);
33720 int vec_any_le (vector signed int, vector bool int);
33721 int vec_any_le (vector signed int, vector signed int);
33722 int vec_any_le (vector float, vector float);
33724 int vec_any_lt (vector bool char, vector unsigned char);
33725 int vec_any_lt (vector unsigned char, vector bool char);
33726 int vec_any_lt (vector unsigned char, vector unsigned char);
33727 int vec_any_lt (vector bool char, vector signed char);
33728 int vec_any_lt (vector signed char, vector bool char);
33729 int vec_any_lt (vector signed char, vector signed char);
33730 int vec_any_lt (vector bool short, vector unsigned short);
33731 int vec_any_lt (vector unsigned short, vector bool short);
33732 int vec_any_lt (vector unsigned short, vector unsigned short);
33733 int vec_any_lt (vector bool short, vector signed short);
33734 int vec_any_lt (vector signed short, vector bool short);
33735 int vec_any_lt (vector signed short, vector signed short);
33736 int vec_any_lt (vector bool int, vector unsigned int);
33737 int vec_any_lt (vector unsigned int, vector bool int);
33738 int vec_any_lt (vector unsigned int, vector unsigned int);
33739 int vec_any_lt (vector bool int, vector signed int);
33740 int vec_any_lt (vector signed int, vector bool int);
33741 int vec_any_lt (vector signed int, vector signed int);
33742 int vec_any_lt (vector float, vector float);
33744 int vec_any_nan (vector float);
33746 int vec_any_ne (vector signed char, vector bool char);
33747 int vec_any_ne (vector signed char, vector signed char);
33748 int vec_any_ne (vector unsigned char, vector bool char);
33749 int vec_any_ne (vector unsigned char, vector unsigned char);
33750 int vec_any_ne (vector bool char, vector bool char);
33751 int vec_any_ne (vector bool char, vector unsigned char);
33752 int vec_any_ne (vector bool char, vector signed char);
33753 int vec_any_ne (vector signed short, vector bool short);
33754 int vec_any_ne (vector signed short, vector signed short);
33755 int vec_any_ne (vector unsigned short, vector bool short);
33756 int vec_any_ne (vector unsigned short, vector unsigned short);
33757 int vec_any_ne (vector bool short, vector bool short);
33758 int vec_any_ne (vector bool short, vector unsigned short);
33759 int vec_any_ne (vector bool short, vector signed short);
33760 int vec_any_ne (vector pixel, vector pixel);
33761 int vec_any_ne (vector signed int, vector bool int);
33762 int vec_any_ne (vector signed int, vector signed int);
33763 int vec_any_ne (vector unsigned int, vector bool int);
33764 int vec_any_ne (vector unsigned int, vector unsigned int);
33765 int vec_any_ne (vector bool int, vector bool int);
33766 int vec_any_ne (vector bool int, vector unsigned int);
33767 int vec_any_ne (vector bool int, vector signed int);
33768 int vec_any_ne (vector float, vector float);
33770 int vec_any_nge (vector float, vector float);
33772 int vec_any_ngt (vector float, vector float);
33774 int vec_any_nle (vector float, vector float);
33776 int vec_any_nlt (vector float, vector float);
33778 int vec_any_numeric (vector float);
33780 int vec_any_out (vector float, vector float);
33783 File: gcc.info, Node: SPARC VIS Built-in Functions, Next: SPU Built-in Functions, Prev: PowerPC AltiVec Built-in Functions, Up: Target Builtins
33785 5.50.13 SPARC VIS Built-in Functions
33786 ------------------------------------
33788 GCC supports SIMD operations on the SPARC using both the generic vector
33789 extensions (*note Vector Extensions::) as well as built-in functions for
33790 the SPARC Visual Instruction Set (VIS). When you use the `-mvis'
33791 switch, the VIS extension is exposed as the following built-in
33794 typedef int v2si __attribute__ ((vector_size (8)));
33795 typedef short v4hi __attribute__ ((vector_size (8)));
33796 typedef short v2hi __attribute__ ((vector_size (4)));
33797 typedef char v8qi __attribute__ ((vector_size (8)));
33798 typedef char v4qi __attribute__ ((vector_size (4)));
33800 void * __builtin_vis_alignaddr (void *, long);
33801 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
33802 v2si __builtin_vis_faligndatav2si (v2si, v2si);
33803 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
33804 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
33806 v4hi __builtin_vis_fexpand (v4qi);
33808 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
33809 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
33810 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
33811 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
33812 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
33813 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
33814 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
33816 v4qi __builtin_vis_fpack16 (v4hi);
33817 v8qi __builtin_vis_fpack32 (v2si, v2si);
33818 v2hi __builtin_vis_fpackfix (v2si);
33819 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
33821 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
33824 File: gcc.info, Node: SPU Built-in Functions, Prev: SPARC VIS Built-in Functions, Up: Target Builtins
33826 5.50.14 SPU Built-in Functions
33827 ------------------------------
33829 GCC provides extensions for the SPU processor as described in the
33830 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
33831 found at `http://cell.scei.co.jp/' or
33832 `http://www.ibm.com/developerworks/power/cell/'. GCC's implementation
33833 differs in several ways.
33835 * The optional extension of specifying vector constants in
33836 parentheses is not supported.
33838 * A vector initializer requires no cast if the vector constant is of
33839 the same type as the variable it is initializing.
33841 * If `signed' or `unsigned' is omitted, the signedness of the vector
33842 type is the default signedness of the base type. The default
33843 varies depending on the operating system, so a portable program
33844 should always specify the signedness.
33846 * By default, the keyword `__vector' is added. The macro `vector' is
33847 defined in `<spu_intrinsics.h>' and can be undefined.
33849 * GCC allows using a `typedef' name as the type specifier for a
33852 * For C, overloaded functions are implemented with macros so the
33853 following does not work:
33855 spu_add ((vector signed int){1, 2, 3, 4}, foo);
33857 Since `spu_add' is a macro, the vector constant in the example is
33858 treated as four separate arguments. Wrap the entire argument in
33859 parentheses for this to work.
33861 * The extended version of `__builtin_expect' is not supported.
33864 _Note:_ Only the interface described in the aforementioned
33865 specification is supported. Internally, GCC uses built-in functions to
33866 implement the required functionality, but these are not supported and
33867 are subject to change without notice.
33870 File: gcc.info, Node: Target Format Checks, Next: Pragmas, Prev: Target Builtins, Up: C Extensions
33872 5.51 Format Checks Specific to Particular Target Machines
33873 =========================================================
33875 For some target machines, GCC supports additional options to the format
33876 attribute (*note Declaring Attributes of Functions: Function
33881 * Solaris Format Checks::
33884 File: gcc.info, Node: Solaris Format Checks, Up: Target Format Checks
33886 5.51.1 Solaris Format Checks
33887 ----------------------------
33889 Solaris targets support the `cmn_err' (or `__cmn_err__') format check.
33890 `cmn_err' accepts a subset of the standard `printf' conversions, and
33891 the two-argument `%b' conversion for displaying bit-fields. See the
33892 Solaris man page for `cmn_err' for more information.
33895 File: gcc.info, Node: Pragmas, Next: Unnamed Fields, Prev: Target Format Checks, Up: C Extensions
33897 5.52 Pragmas Accepted by GCC
33898 ============================
33900 GCC supports several types of pragmas, primarily in order to compile
33901 code originally written for other compilers. Note that in general we
33902 do not recommend the use of pragmas; *Note Function Attributes::, for
33903 further explanation.
33909 * RS/6000 and PowerPC Pragmas::
33911 * Solaris Pragmas::
33912 * Symbol-Renaming Pragmas::
33913 * Structure-Packing Pragmas::
33915 * Diagnostic Pragmas::
33916 * Visibility Pragmas::
33917 * Push/Pop Macro Pragmas::
33918 * Function Specific Option Pragmas::
33921 File: gcc.info, Node: ARM Pragmas, Next: M32C Pragmas, Up: Pragmas
33926 The ARM target defines pragmas for controlling the default addition of
33927 `long_call' and `short_call' attributes to functions. *Note Function
33928 Attributes::, for information about the effects of these attributes.
33931 Set all subsequent functions to have the `long_call' attribute.
33934 Set all subsequent functions to have the `short_call' attribute.
33937 Do not affect the `long_call' or `short_call' attributes of
33938 subsequent functions.
33941 File: gcc.info, Node: M32C Pragmas, Next: RS/6000 and PowerPC Pragmas, Prev: ARM Pragmas, Up: Pragmas
33943 5.52.2 M32C Pragmas
33944 -------------------
33947 Overrides the command line option `-memregs=' for the current
33948 file. Use with care! This pragma must be before any function in
33949 the file, and mixing different memregs values in different objects
33950 may make them incompatible. This pragma is useful when a
33951 performance-critical function uses a memreg for temporary values,
33952 as it may allow you to reduce the number of memregs used.
33956 File: gcc.info, Node: RS/6000 and PowerPC Pragmas, Next: Darwin Pragmas, Prev: M32C Pragmas, Up: Pragmas
33958 5.52.3 RS/6000 and PowerPC Pragmas
33959 ----------------------------------
33961 The RS/6000 and PowerPC targets define one pragma for controlling
33962 whether or not the `longcall' attribute is added to function
33963 declarations by default. This pragma overrides the `-mlongcall'
33964 option, but not the `longcall' and `shortcall' attributes. *Note
33965 RS/6000 and PowerPC Options::, for more information about when long
33966 calls are and are not necessary.
33969 Apply the `longcall' attribute to all subsequent function
33973 Do not apply the `longcall' attribute to subsequent function
33977 File: gcc.info, Node: Darwin Pragmas, Next: Solaris Pragmas, Prev: RS/6000 and PowerPC Pragmas, Up: Pragmas
33979 5.52.4 Darwin Pragmas
33980 ---------------------
33982 The following pragmas are available for all architectures running the
33983 Darwin operating system. These are useful for compatibility with other
33987 This pragma is accepted, but has no effect.
33989 `options align=ALIGNMENT'
33990 This pragma sets the alignment of fields in structures. The
33991 values of ALIGNMENT may be `mac68k', to emulate m68k alignment, or
33992 `power', to emulate PowerPC alignment. Uses of this pragma nest
33993 properly; to restore the previous setting, use `reset' for the
33996 `segment TOKENS...'
33997 This pragma is accepted, but has no effect.
33999 `unused (VAR [, VAR]...)'
34000 This pragma declares variables to be possibly unused. GCC will not
34001 produce warnings for the listed variables. The effect is similar
34002 to that of the `unused' attribute, except that this pragma may
34003 appear anywhere within the variables' scopes.
34006 File: gcc.info, Node: Solaris Pragmas, Next: Symbol-Renaming Pragmas, Prev: Darwin Pragmas, Up: Pragmas
34008 5.52.5 Solaris Pragmas
34009 ----------------------
34011 The Solaris target supports `#pragma redefine_extname' (*note
34012 Symbol-Renaming Pragmas::). It also supports additional `#pragma'
34013 directives for compatibility with the system compiler.
34015 `align ALIGNMENT (VARIABLE [, VARIABLE]...)'
34016 Increase the minimum alignment of each VARIABLE to ALIGNMENT.
34017 This is the same as GCC's `aligned' attribute *note Variable
34018 Attributes::). Macro expansion occurs on the arguments to this
34019 pragma when compiling C and Objective-C. It does not currently
34020 occur when compiling C++, but this is a bug which may be fixed in
34023 `fini (FUNCTION [, FUNCTION]...)'
34024 This pragma causes each listed FUNCTION to be called after main,
34025 or during shared module unloading, by adding a call to the `.fini'
34028 `init (FUNCTION [, FUNCTION]...)'
34029 This pragma causes each listed FUNCTION to be called during
34030 initialization (before `main') or during shared module loading, by
34031 adding a call to the `.init' section.
34035 File: gcc.info, Node: Symbol-Renaming Pragmas, Next: Structure-Packing Pragmas, Prev: Solaris Pragmas, Up: Pragmas
34037 5.52.6 Symbol-Renaming Pragmas
34038 ------------------------------
34040 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
34041 supports two `#pragma' directives which change the name used in
34042 assembly for a given declaration. These pragmas are only available on
34043 platforms whose system headers need them. To get this effect on all
34044 platforms supported by GCC, use the asm labels extension (*note Asm
34047 `redefine_extname OLDNAME NEWNAME'
34048 This pragma gives the C function OLDNAME the assembly symbol
34049 NEWNAME. The preprocessor macro `__PRAGMA_REDEFINE_EXTNAME' will
34050 be defined if this pragma is available (currently only on Solaris).
34052 `extern_prefix STRING'
34053 This pragma causes all subsequent external function and variable
34054 declarations to have STRING prepended to their assembly symbols.
34055 This effect may be terminated with another `extern_prefix' pragma
34056 whose argument is an empty string. The preprocessor macro
34057 `__PRAGMA_EXTERN_PREFIX' will be defined if this pragma is
34058 available (currently only on Tru64 UNIX).
34060 These pragmas and the asm labels extension interact in a complicated
34061 manner. Here are some corner cases you may want to be aware of.
34063 1. Both pragmas silently apply only to declarations with external
34064 linkage. Asm labels do not have this restriction.
34066 2. In C++, both pragmas silently apply only to declarations with "C"
34067 linkage. Again, asm labels do not have this restriction.
34069 3. If any of the three ways of changing the assembly name of a
34070 declaration is applied to a declaration whose assembly name has
34071 already been determined (either by a previous use of one of these
34072 features, or because the compiler needed the assembly name in
34073 order to generate code), and the new name is different, a warning
34074 issues and the name does not change.
34076 4. The OLDNAME used by `#pragma redefine_extname' is always the
34079 5. If `#pragma extern_prefix' is in effect, and a declaration occurs
34080 with an asm label attached, the prefix is silently ignored for
34083 6. If `#pragma extern_prefix' and `#pragma redefine_extname' apply to
34084 the same declaration, whichever triggered first wins, and a
34085 warning issues if they contradict each other. (We would like to
34086 have `#pragma redefine_extname' always win, for consistency with
34087 asm labels, but if `#pragma extern_prefix' triggers first we have
34088 no way of knowing that that happened.)
34091 File: gcc.info, Node: Structure-Packing Pragmas, Next: Weak Pragmas, Prev: Symbol-Renaming Pragmas, Up: Pragmas
34093 5.52.7 Structure-Packing Pragmas
34094 --------------------------------
34096 For compatibility with Microsoft Windows compilers, GCC supports a set
34097 of `#pragma' directives which change the maximum alignment of members
34098 of structures (other than zero-width bitfields), unions, and classes
34099 subsequently defined. The N value below always is required to be a
34100 small power of two and specifies the new alignment in bytes.
34102 1. `#pragma pack(N)' simply sets the new alignment.
34104 2. `#pragma pack()' sets the alignment to the one that was in effect
34105 when compilation started (see also command line option
34106 `-fpack-struct[=<n>]' *note Code Gen Options::).
34108 3. `#pragma pack(push[,N])' pushes the current alignment setting on
34109 an internal stack and then optionally sets the new alignment.
34111 4. `#pragma pack(pop)' restores the alignment setting to the one
34112 saved at the top of the internal stack (and removes that stack
34113 entry). Note that `#pragma pack([N])' does not influence this
34114 internal stack; thus it is possible to have `#pragma pack(push)'
34115 followed by multiple `#pragma pack(N)' instances and finalized by
34116 a single `#pragma pack(pop)'.
34118 Some targets, e.g. i386 and powerpc, support the `ms_struct' `#pragma'
34119 which lays out a structure as the documented `__attribute__
34121 1. `#pragma ms_struct on' turns on the layout for structures declared.
34123 2. `#pragma ms_struct off' turns off the layout for structures
34126 3. `#pragma ms_struct reset' goes back to the default layout.
34129 File: gcc.info, Node: Weak Pragmas, Next: Diagnostic Pragmas, Prev: Structure-Packing Pragmas, Up: Pragmas
34131 5.52.8 Weak Pragmas
34132 -------------------
34134 For compatibility with SVR4, GCC supports a set of `#pragma' directives
34135 for declaring symbols to be weak, and defining weak aliases.
34137 `#pragma weak SYMBOL'
34138 This pragma declares SYMBOL to be weak, as if the declaration had
34139 the attribute of the same name. The pragma may appear before or
34140 after the declaration of SYMBOL, but must appear before either its
34141 first use or its definition. It is not an error for SYMBOL to
34142 never be defined at all.
34144 `#pragma weak SYMBOL1 = SYMBOL2'
34145 This pragma declares SYMBOL1 to be a weak alias of SYMBOL2. It is
34146 an error if SYMBOL2 is not defined in the current translation unit.
34149 File: gcc.info, Node: Diagnostic Pragmas, Next: Visibility Pragmas, Prev: Weak Pragmas, Up: Pragmas
34151 5.52.9 Diagnostic Pragmas
34152 -------------------------
34154 GCC allows the user to selectively enable or disable certain types of
34155 diagnostics, and change the kind of the diagnostic. For example, a
34156 project's policy might require that all sources compile with `-Werror'
34157 but certain files might have exceptions allowing specific types of
34158 warnings. Or, a project might selectively enable diagnostics and treat
34159 them as errors depending on which preprocessor macros are defined.
34161 `#pragma GCC diagnostic KIND OPTION'
34162 Modifies the disposition of a diagnostic. Note that not all
34163 diagnostics are modifiable; at the moment only warnings (normally
34164 controlled by `-W...') can be controlled, and not all of them.
34165 Use `-fdiagnostics-show-option' to determine which diagnostics are
34166 controllable and which option controls them.
34168 KIND is `error' to treat this diagnostic as an error, `warning' to
34169 treat it like a warning (even if `-Werror' is in effect), or
34170 `ignored' if the diagnostic is to be ignored. OPTION is a double
34171 quoted string which matches the command line option.
34173 #pragma GCC diagnostic warning "-Wformat"
34174 #pragma GCC diagnostic error "-Wformat"
34175 #pragma GCC diagnostic ignored "-Wformat"
34177 Note that these pragmas override any command line options. Also,
34178 while it is syntactically valid to put these pragmas anywhere in
34179 your sources, the only supported location for them is before any
34180 data or functions are defined. Doing otherwise may result in
34181 unpredictable results depending on how the optimizer manages your
34182 sources. If the same option is listed multiple times, the last
34183 one specified is the one that is in effect. This pragma is not
34184 intended to be a general purpose replacement for command line
34185 options, but for implementing strict control over project policies.
34188 GCC also offers a simple mechanism for printing messages during
34191 `#pragma message STRING'
34192 Prints STRING as a compiler message on compilation. The message
34193 is informational only, and is neither a compilation warning nor an
34196 #pragma message "Compiling " __FILE__ "..."
34198 STRING may be parenthesized, and is printed with location
34199 information. For example,
34201 #define DO_PRAGMA(x) _Pragma (#x)
34202 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
34204 TODO(Remember to fix this)
34206 prints `/tmp/file.c:4: note: #pragma message: TODO - Remember to
34211 File: gcc.info, Node: Visibility Pragmas, Next: Push/Pop Macro Pragmas, Prev: Diagnostic Pragmas, Up: Pragmas
34213 5.52.10 Visibility Pragmas
34214 --------------------------
34216 `#pragma GCC visibility push(VISIBILITY)'
34217 `#pragma GCC visibility pop'
34218 This pragma allows the user to set the visibility for multiple
34219 declarations without having to give each a visibility attribute
34220 *Note Function Attributes::, for more information about visibility
34221 and the attribute syntax.
34223 In C++, `#pragma GCC visibility' affects only namespace-scope
34224 declarations. Class members and template specializations are not
34225 affected; if you want to override the visibility for a particular
34226 member or instantiation, you must use an attribute.
34230 File: gcc.info, Node: Push/Pop Macro Pragmas, Next: Function Specific Option Pragmas, Prev: Visibility Pragmas, Up: Pragmas
34232 5.52.11 Push/Pop Macro Pragmas
34233 ------------------------------
34235 For compatibility with Microsoft Windows compilers, GCC supports
34236 `#pragma push_macro("MACRO_NAME")' and `#pragma
34237 pop_macro("MACRO_NAME")'.
34239 `#pragma push_macro("MACRO_NAME")'
34240 This pragma saves the value of the macro named as MACRO_NAME to
34241 the top of the stack for this macro.
34243 `#pragma pop_macro("MACRO_NAME")'
34244 This pragma sets the value of the macro named as MACRO_NAME to the
34245 value on top of the stack for this macro. If the stack for
34246 MACRO_NAME is empty, the value of the macro remains unchanged.
34251 #pragma push_macro("X")
34254 #pragma pop_macro("X")
34257 In this example, the definition of X as 1 is saved by `#pragma
34258 push_macro' and restored by `#pragma pop_macro'.
34261 File: gcc.info, Node: Function Specific Option Pragmas, Prev: Push/Pop Macro Pragmas, Up: Pragmas
34263 5.52.12 Function Specific Option Pragmas
34264 ----------------------------------------
34266 `#pragma GCC target ("STRING"...)'
34267 This pragma allows you to set target specific options for functions
34268 defined later in the source file. One or more strings can be
34269 specified. Each function that is defined after this point will be
34270 as if `attribute((target("STRING")))' was specified for that
34271 function. The parenthesis around the options is optional. *Note
34272 Function Attributes::, for more information about the `target'
34273 attribute and the attribute syntax.
34275 The `#pragma GCC target' pragma is not implemented in GCC versions
34276 earlier than 4.4, and is currently only implemented for the 386
34277 and x86_64 backends.
34279 `#pragma GCC optimize ("STRING"...)'
34280 This pragma allows you to set global optimization options for
34281 functions defined later in the source file. One or more strings
34282 can be specified. Each function that is defined after this point
34283 will be as if `attribute((optimize("STRING")))' was specified for
34284 that function. The parenthesis around the options is optional.
34285 *Note Function Attributes::, for more information about the
34286 `optimize' attribute and the attribute syntax.
34288 The `#pragma GCC optimize' pragma is not implemented in GCC
34289 versions earlier than 4.4.
34291 `#pragma GCC push_options'
34292 `#pragma GCC pop_options'
34293 These pragmas maintain a stack of the current target and
34294 optimization options. It is intended for include files where you
34295 temporarily want to switch to using a different `#pragma GCC
34296 target' or `#pragma GCC optimize' and then to pop back to the
34299 The `#pragma GCC push_options' and `#pragma GCC pop_options'
34300 pragmas are not implemented in GCC versions earlier than 4.4.
34302 `#pragma GCC reset_options'
34303 This pragma clears the current `#pragma GCC target' and `#pragma
34304 GCC optimize' to use the default switches as specified on the
34307 The `#pragma GCC reset_options' pragma is not implemented in GCC
34308 versions earlier than 4.4.
34311 File: gcc.info, Node: Unnamed Fields, Next: Thread-Local, Prev: Pragmas, Up: C Extensions
34313 5.53 Unnamed struct/union fields within structs/unions
34314 ======================================================
34316 For compatibility with other compilers, GCC allows you to define a
34317 structure or union that contains, as fields, structures and unions
34318 without names. For example:
34329 In this example, the user would be able to access members of the
34330 unnamed union with code like `foo.b'. Note that only unnamed structs
34331 and unions are allowed, you may not have, for example, an unnamed `int'.
34333 You must never create such structures that cause ambiguous field
34334 definitions. For example, this structure:
34343 It is ambiguous which `a' is being referred to with `foo.a'. Such
34344 constructs are not supported and must be avoided. In the future, such
34345 constructs may be detected and treated as compilation errors.
34347 Unless `-fms-extensions' is used, the unnamed field must be a
34348 structure or union definition without a tag (for example, `struct { int
34349 a; };'). If `-fms-extensions' is used, the field may also be a
34350 definition with a tag such as `struct foo { int a; };', a reference to
34351 a previously defined structure or union such as `struct foo;', or a
34352 reference to a `typedef' name for a previously defined structure or
34356 File: gcc.info, Node: Thread-Local, Next: Binary constants, Prev: Unnamed Fields, Up: C Extensions
34358 5.54 Thread-Local Storage
34359 =========================
34361 Thread-local storage (TLS) is a mechanism by which variables are
34362 allocated such that there is one instance of the variable per extant
34363 thread. The run-time model GCC uses to implement this originates in
34364 the IA-64 processor-specific ABI, but has since been migrated to other
34365 processors as well. It requires significant support from the linker
34366 (`ld'), dynamic linker (`ld.so'), and system libraries (`libc.so' and
34367 `libpthread.so'), so it is not available everywhere.
34369 At the user level, the extension is visible with a new storage class
34370 keyword: `__thread'. For example:
34373 extern __thread struct state s;
34374 static __thread char *p;
34376 The `__thread' specifier may be used alone, with the `extern' or
34377 `static' specifiers, but with no other storage class specifier. When
34378 used with `extern' or `static', `__thread' must appear immediately
34379 after the other storage class specifier.
34381 The `__thread' specifier may be applied to any global, file-scoped
34382 static, function-scoped static, or static data member of a class. It
34383 may not be applied to block-scoped automatic or non-static data member.
34385 When the address-of operator is applied to a thread-local variable, it
34386 is evaluated at run-time and returns the address of the current thread's
34387 instance of that variable. An address so obtained may be used by any
34388 thread. When a thread terminates, any pointers to thread-local
34389 variables in that thread become invalid.
34391 No static initialization may refer to the address of a thread-local
34394 In C++, if an initializer is present for a thread-local variable, it
34395 must be a CONSTANT-EXPRESSION, as defined in 5.19.2 of the ANSI/ISO C++
34398 See ELF Handling For Thread-Local Storage
34399 (http://people.redhat.com/drepper/tls.pdf) for a detailed explanation of
34400 the four thread-local storage addressing models, and how the run-time
34401 is expected to function.
34405 * C99 Thread-Local Edits::
34406 * C++98 Thread-Local Edits::
34409 File: gcc.info, Node: C99 Thread-Local Edits, Next: C++98 Thread-Local Edits, Up: Thread-Local
34411 5.54.1 ISO/IEC 9899:1999 Edits for Thread-Local Storage
34412 -------------------------------------------------------
34414 The following are a set of changes to ISO/IEC 9899:1999 (aka C99) that
34415 document the exact semantics of the language extension.
34417 * `5.1.2 Execution environments'
34419 Add new text after paragraph 1
34421 Within either execution environment, a "thread" is a flow of
34422 control within a program. It is implementation defined
34423 whether or not there may be more than one thread associated
34424 with a program. It is implementation defined how threads
34425 beyond the first are created, the name and type of the
34426 function called at thread startup, and how threads may be
34427 terminated. However, objects with thread storage duration
34428 shall be initialized before thread startup.
34430 * `6.2.4 Storage durations of objects'
34432 Add new text before paragraph 3
34434 An object whose identifier is declared with the storage-class
34435 specifier `__thread' has "thread storage duration". Its
34436 lifetime is the entire execution of the thread, and its
34437 stored value is initialized only once, prior to thread
34444 * `6.7.1 Storage-class specifiers'
34446 Add `__thread' to the list of storage class specifiers in
34449 Change paragraph 2 to
34451 With the exception of `__thread', at most one storage-class
34452 specifier may be given [...]. The `__thread' specifier may
34453 be used alone, or immediately following `extern' or `static'.
34455 Add new text after paragraph 6
34457 The declaration of an identifier for a variable that has
34458 block scope that specifies `__thread' shall also specify
34459 either `extern' or `static'.
34461 The `__thread' specifier shall be used only with variables.
34464 File: gcc.info, Node: C++98 Thread-Local Edits, Prev: C99 Thread-Local Edits, Up: Thread-Local
34466 5.54.2 ISO/IEC 14882:1998 Edits for Thread-Local Storage
34467 --------------------------------------------------------
34469 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
34470 that document the exact semantics of the language extension.
34472 * [intro.execution]
34474 New text after paragraph 4
34476 A "thread" is a flow of control within the abstract machine.
34477 It is implementation defined whether or not there may be more
34480 New text after paragraph 7
34482 It is unspecified whether additional action must be taken to
34483 ensure when and whether side effects are visible to other
34490 * [basic.start.main]
34492 Add after paragraph 5
34494 The thread that begins execution at the `main' function is
34495 called the "main thread". It is implementation defined how
34496 functions beginning threads other than the main thread are
34497 designated or typed. A function so designated, as well as
34498 the `main' function, is called a "thread startup function".
34499 It is implementation defined what happens if a thread startup
34500 function returns. It is implementation defined what happens
34501 to other threads when any thread calls `exit'.
34503 * [basic.start.init]
34505 Add after paragraph 4
34507 The storage for an object of thread storage duration shall be
34508 statically initialized before the first statement of the
34509 thread startup function. An object of thread storage
34510 duration shall not require dynamic initialization.
34512 * [basic.start.term]
34514 Add after paragraph 3
34516 The type of an object with thread storage duration shall not
34517 have a non-trivial destructor, nor shall it be an array type
34518 whose elements (directly or indirectly) have non-trivial
34523 Add "thread storage duration" to the list in paragraph 1.
34527 Thread, static, and automatic storage durations are
34528 associated with objects introduced by declarations [...].
34530 Add `__thread' to the list of specifiers in paragraph 3.
34532 * [basic.stc.thread]
34534 New section before [basic.stc.static]
34536 The keyword `__thread' applied to a non-local object gives the
34537 object thread storage duration.
34539 A local variable or class data member declared both `static'
34540 and `__thread' gives the variable or member thread storage
34543 * [basic.stc.static]
34547 All objects which have neither thread storage duration,
34548 dynamic storage duration nor are local [...].
34552 Add `__thread' to the list in paragraph 1.
34556 With the exception of `__thread', at most one
34557 STORAGE-CLASS-SPECIFIER shall appear in a given
34558 DECL-SPECIFIER-SEQ. The `__thread' specifier may be used
34559 alone, or immediately following the `extern' or `static'
34562 Add after paragraph 5
34564 The `__thread' specifier can be applied only to the names of
34565 objects and to anonymous unions.
34569 Add after paragraph 6
34571 Non-`static' members shall not be `__thread'.
34574 File: gcc.info, Node: Binary constants, Prev: Thread-Local, Up: C Extensions
34576 5.55 Binary constants using the `0b' prefix
34577 ===========================================
34579 Integer constants can be written as binary constants, consisting of a
34580 sequence of `0' and `1' digits, prefixed by `0b' or `0B'. This is
34581 particularly useful in environments that operate a lot on the bit-level
34582 (like microcontrollers).
34584 The following statements are identical:
34591 The type of these constants follows the same rules as for octal or
34592 hexadecimal integer constants, so suffixes like `L' or `UL' can be
34596 File: gcc.info, Node: C++ Extensions, Next: Objective-C, Prev: C Extensions, Up: Top
34598 6 Extensions to the C++ Language
34599 ********************************
34601 The GNU compiler provides these extensions to the C++ language (and you
34602 can also use most of the C language extensions in your C++ programs).
34603 If you want to write code that checks whether these features are
34604 available, you can test for the GNU compiler the same way as for C
34605 programs: check for a predefined macro `__GNUC__'. You can also use
34606 `__GNUG__' to test specifically for GNU C++ (*note Predefined Macros:
34607 (cpp)Common Predefined Macros.).
34611 * Volatiles:: What constitutes an access to a volatile object.
34612 * Restricted Pointers:: C99 restricted pointers and references.
34613 * Vague Linkage:: Where G++ puts inlines, vtables and such.
34614 * C++ Interface:: You can use a single C++ header file for both
34615 declarations and definitions.
34616 * Template Instantiation:: Methods for ensuring that exactly one copy of
34617 each needed template instantiation is emitted.
34618 * Bound member functions:: You can extract a function pointer to the
34619 method denoted by a `->*' or `.*' expression.
34620 * C++ Attributes:: Variable, function, and type attributes for C++ only.
34621 * Namespace Association:: Strong using-directives for namespace association.
34622 * Type Traits:: Compiler support for type traits
34623 * Java Exceptions:: Tweaking exception handling to work with Java.
34624 * Deprecated Features:: Things will disappear from g++.
34625 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
34628 File: gcc.info, Node: Volatiles, Next: Restricted Pointers, Up: C++ Extensions
34630 6.1 When is a Volatile Object Accessed?
34631 =======================================
34633 Both the C and C++ standard have the concept of volatile objects. These
34634 are normally accessed by pointers and used for accessing hardware. The
34635 standards encourage compilers to refrain from optimizations concerning
34636 accesses to volatile objects. The C standard leaves it implementation
34637 defined as to what constitutes a volatile access. The C++ standard
34638 omits to specify this, except to say that C++ should behave in a
34639 similar manner to C with respect to volatiles, where possible. The
34640 minimum either standard specifies is that at a sequence point all
34641 previous accesses to volatile objects have stabilized and no subsequent
34642 accesses have occurred. Thus an implementation is free to reorder and
34643 combine volatile accesses which occur between sequence points, but
34644 cannot do so for accesses across a sequence point. The use of
34645 volatiles does not allow you to violate the restriction on updating
34646 objects multiple times within a sequence point.
34648 *Note Volatile qualifier and the C compiler: Qualifiers implementation.
34650 The behavior differs slightly between C and C++ in the non-obvious
34653 volatile int *src = SOMEVALUE;
34656 With C, such expressions are rvalues, and GCC interprets this either
34657 as a read of the volatile object being pointed to or only as request to
34658 evaluate the side-effects. The C++ standard specifies that such
34659 expressions do not undergo lvalue to rvalue conversion, and that the
34660 type of the dereferenced object may be incomplete. The C++ standard
34661 does not specify explicitly that it is this lvalue to rvalue conversion
34662 which may be responsible for causing an access. However, there is
34663 reason to believe that it is, because otherwise certain simple
34664 expressions become undefined. However, because it would surprise most
34665 programmers, G++ treats dereferencing a pointer to volatile object of
34666 complete type when the value is unused as GCC would do for an
34667 equivalent type in C. When the object has incomplete type, G++ issues
34668 a warning; if you wish to force an error, you must force a conversion
34669 to rvalue with, for instance, a static cast.
34671 When using a reference to volatile, G++ does not treat equivalent
34672 expressions as accesses to volatiles, but instead issues a warning that
34673 no volatile is accessed. The rationale for this is that otherwise it
34674 becomes difficult to determine where volatile access occur, and not
34675 possible to ignore the return value from functions returning volatile
34676 references. Again, if you wish to force a read, cast the reference to
34680 File: gcc.info, Node: Restricted Pointers, Next: Vague Linkage, Prev: Volatiles, Up: C++ Extensions
34682 6.2 Restricting Pointer Aliasing
34683 ================================
34685 As with the C front end, G++ understands the C99 feature of restricted
34686 pointers, specified with the `__restrict__', or `__restrict' type
34687 qualifier. Because you cannot compile C++ by specifying the `-std=c99'
34688 language flag, `restrict' is not a keyword in C++.
34690 In addition to allowing restricted pointers, you can specify restricted
34691 references, which indicate that the reference is not aliased in the
34694 void fn (int *__restrict__ rptr, int &__restrict__ rref)
34699 In the body of `fn', RPTR points to an unaliased integer and RREF
34700 refers to a (different) unaliased integer.
34702 You may also specify whether a member function's THIS pointer is
34703 unaliased by using `__restrict__' as a member function qualifier.
34705 void T::fn () __restrict__
34710 Within the body of `T::fn', THIS will have the effective definition `T
34711 *__restrict__ const this'. Notice that the interpretation of a
34712 `__restrict__' member function qualifier is different to that of
34713 `const' or `volatile' qualifier, in that it is applied to the pointer
34714 rather than the object. This is consistent with other compilers which
34715 implement restricted pointers.
34717 As with all outermost parameter qualifiers, `__restrict__' is ignored
34718 in function definition matching. This means you only need to specify
34719 `__restrict__' in a function definition, rather than in a function
34723 File: gcc.info, Node: Vague Linkage, Next: C++ Interface, Prev: Restricted Pointers, Up: C++ Extensions
34728 There are several constructs in C++ which require space in the object
34729 file but are not clearly tied to a single translation unit. We say that
34730 these constructs have "vague linkage". Typically such constructs are
34731 emitted wherever they are needed, though sometimes we can be more
34735 Inline functions are typically defined in a header file which can
34736 be included in many different compilations. Hopefully they can
34737 usually be inlined, but sometimes an out-of-line copy is
34738 necessary, if the address of the function is taken or if inlining
34739 fails. In general, we emit an out-of-line copy in all translation
34740 units where one is needed. As an exception, we only emit inline
34741 virtual functions with the vtable, since it will always require a
34744 Local static variables and string constants used in an inline
34745 function are also considered to have vague linkage, since they
34746 must be shared between all inlined and out-of-line instances of
34750 C++ virtual functions are implemented in most compilers using a
34751 lookup table, known as a vtable. The vtable contains pointers to
34752 the virtual functions provided by a class, and each object of the
34753 class contains a pointer to its vtable (or vtables, in some
34754 multiple-inheritance situations). If the class declares any
34755 non-inline, non-pure virtual functions, the first one is chosen as
34756 the "key method" for the class, and the vtable is only emitted in
34757 the translation unit where the key method is defined.
34759 _Note:_ If the chosen key method is later defined as inline, the
34760 vtable will still be emitted in every translation unit which
34761 defines it. Make sure that any inline virtuals are declared
34762 inline in the class body, even if they are not defined there.
34765 C++ requires information about types to be written out in order to
34766 implement `dynamic_cast', `typeid' and exception handling. For
34767 polymorphic classes (classes with virtual functions), the type_info
34768 object is written out along with the vtable so that `dynamic_cast'
34769 can determine the dynamic type of a class object at runtime. For
34770 all other types, we write out the type_info object when it is
34771 used: when applying `typeid' to an expression, throwing an object,
34772 or referring to a type in a catch clause or exception
34775 Template Instantiations
34776 Most everything in this section also applies to template
34777 instantiations, but there are other options as well. *Note
34778 Where's the Template?: Template Instantiation.
34781 When used with GNU ld version 2.8 or later on an ELF system such as
34782 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
34783 these constructs will be discarded at link time. This is known as
34786 On targets that don't support COMDAT, but do support weak symbols, GCC
34787 will use them. This way one copy will override all the others, but the
34788 unused copies will still take up space in the executable.
34790 For targets which do not support either COMDAT or weak symbols, most
34791 entities with vague linkage will be emitted as local symbols to avoid
34792 duplicate definition errors from the linker. This will not happen for
34793 local statics in inlines, however, as having multiple copies will
34794 almost certainly break things.
34796 *Note Declarations and Definitions in One Header: C++ Interface, for
34797 another way to control placement of these constructs.
34800 File: gcc.info, Node: C++ Interface, Next: Template Instantiation, Prev: Vague Linkage, Up: C++ Extensions
34802 6.4 #pragma interface and implementation
34803 ========================================
34805 `#pragma interface' and `#pragma implementation' provide the user with
34806 a way of explicitly directing the compiler to emit entities with vague
34807 linkage (and debugging information) in a particular translation unit.
34809 _Note:_ As of GCC 2.7.2, these `#pragma's are not useful in most
34810 cases, because of COMDAT support and the "key method" heuristic
34811 mentioned in *note Vague Linkage::. Using them can actually cause your
34812 program to grow due to unnecessary out-of-line copies of inline
34813 functions. Currently (3.4) the only benefit of these `#pragma's is
34814 reduced duplication of debugging information, and that should be
34815 addressed soon on DWARF 2 targets with the use of COMDAT groups.
34817 `#pragma interface'
34818 `#pragma interface "SUBDIR/OBJECTS.h"'
34819 Use this directive in _header files_ that define object classes,
34820 to save space in most of the object files that use those classes.
34821 Normally, local copies of certain information (backup copies of
34822 inline member functions, debugging information, and the internal
34823 tables that implement virtual functions) must be kept in each
34824 object file that includes class definitions. You can use this
34825 pragma to avoid such duplication. When a header file containing
34826 `#pragma interface' is included in a compilation, this auxiliary
34827 information will not be generated (unless the main input source
34828 file itself uses `#pragma implementation'). Instead, the object
34829 files will contain references to be resolved at link time.
34831 The second form of this directive is useful for the case where you
34832 have multiple headers with the same name in different directories.
34833 If you use this form, you must specify the same string to `#pragma
34836 `#pragma implementation'
34837 `#pragma implementation "OBJECTS.h"'
34838 Use this pragma in a _main input file_, when you want full output
34839 from included header files to be generated (and made globally
34840 visible). The included header file, in turn, should use `#pragma
34841 interface'. Backup copies of inline member functions, debugging
34842 information, and the internal tables used to implement virtual
34843 functions are all generated in implementation files.
34845 If you use `#pragma implementation' with no argument, it applies to
34846 an include file with the same basename(1) as your source file.
34847 For example, in `allclass.cc', giving just `#pragma implementation'
34848 by itself is equivalent to `#pragma implementation "allclass.h"'.
34850 In versions of GNU C++ prior to 2.6.0 `allclass.h' was treated as
34851 an implementation file whenever you would include it from
34852 `allclass.cc' even if you never specified `#pragma
34853 implementation'. This was deemed to be more trouble than it was
34854 worth, however, and disabled.
34856 Use the string argument if you want a single implementation file to
34857 include code from multiple header files. (You must also use
34858 `#include' to include the header file; `#pragma implementation'
34859 only specifies how to use the file--it doesn't actually include
34862 There is no way to split up the contents of a single header file
34863 into multiple implementation files.
34865 `#pragma implementation' and `#pragma interface' also have an effect
34866 on function inlining.
34868 If you define a class in a header file marked with `#pragma
34869 interface', the effect on an inline function defined in that class is
34870 similar to an explicit `extern' declaration--the compiler emits no code
34871 at all to define an independent version of the function. Its
34872 definition is used only for inlining with its callers.
34874 Conversely, when you include the same header file in a main source file
34875 that declares it as `#pragma implementation', the compiler emits code
34876 for the function itself; this defines a version of the function that
34877 can be found via pointers (or by callers compiled without inlining).
34878 If all calls to the function can be inlined, you can avoid emitting the
34879 function by compiling with `-fno-implement-inlines'. If any calls were
34880 not inlined, you will get linker errors.
34882 ---------- Footnotes ----------
34884 (1) A file's "basename" was the name stripped of all leading path
34885 information and of trailing suffixes, such as `.h' or `.C' or `.cc'.
34888 File: gcc.info, Node: Template Instantiation, Next: Bound member functions, Prev: C++ Interface, Up: C++ Extensions
34890 6.5 Where's the Template?
34891 =========================
34893 C++ templates are the first language feature to require more
34894 intelligence from the environment than one usually finds on a UNIX
34895 system. Somehow the compiler and linker have to make sure that each
34896 template instance occurs exactly once in the executable if it is needed,
34897 and not at all otherwise. There are two basic approaches to this
34898 problem, which are referred to as the Borland model and the Cfront
34902 Borland C++ solved the template instantiation problem by adding
34903 the code equivalent of common blocks to their linker; the compiler
34904 emits template instances in each translation unit that uses them,
34905 and the linker collapses them together. The advantage of this
34906 model is that the linker only has to consider the object files
34907 themselves; there is no external complexity to worry about. This
34908 disadvantage is that compilation time is increased because the
34909 template code is being compiled repeatedly. Code written for this
34910 model tends to include definitions of all templates in the header
34911 file, since they must be seen to be instantiated.
34914 The AT&T C++ translator, Cfront, solved the template instantiation
34915 problem by creating the notion of a template repository, an
34916 automatically maintained place where template instances are
34917 stored. A more modern version of the repository works as follows:
34918 As individual object files are built, the compiler places any
34919 template definitions and instantiations encountered in the
34920 repository. At link time, the link wrapper adds in the objects in
34921 the repository and compiles any needed instances that were not
34922 previously emitted. The advantages of this model are more optimal
34923 compilation speed and the ability to use the system linker; to
34924 implement the Borland model a compiler vendor also needs to
34925 replace the linker. The disadvantages are vastly increased
34926 complexity, and thus potential for error; for some code this can be
34927 just as transparent, but in practice it can been very difficult to
34928 build multiple programs in one directory and one program in
34929 multiple directories. Code written for this model tends to
34930 separate definitions of non-inline member templates into a
34931 separate file, which should be compiled separately.
34933 When used with GNU ld version 2.8 or later on an ELF system such as
34934 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
34935 Borland model. On other systems, G++ implements neither automatic
34938 A future version of G++ will support a hybrid model whereby the
34939 compiler will emit any instantiations for which the template definition
34940 is included in the compile, and store template definitions and
34941 instantiation context information into the object file for the rest.
34942 The link wrapper will extract that information as necessary and invoke
34943 the compiler to produce the remaining instantiations. The linker will
34944 then combine duplicate instantiations.
34946 In the mean time, you have the following options for dealing with
34947 template instantiations:
34949 1. Compile your template-using code with `-frepo'. The compiler will
34950 generate files with the extension `.rpo' listing all of the
34951 template instantiations used in the corresponding object files
34952 which could be instantiated there; the link wrapper, `collect2',
34953 will then update the `.rpo' files to tell the compiler where to
34954 place those instantiations and rebuild any affected object files.
34955 The link-time overhead is negligible after the first pass, as the
34956 compiler will continue to place the instantiations in the same
34959 This is your best option for application code written for the
34960 Borland model, as it will just work. Code written for the Cfront
34961 model will need to be modified so that the template definitions
34962 are available at one or more points of instantiation; usually this
34963 is as simple as adding `#include <tmethods.cc>' to the end of each
34966 For library code, if you want the library to provide all of the
34967 template instantiations it needs, just try to link all of its
34968 object files together; the link will fail, but cause the
34969 instantiations to be generated as a side effect. Be warned,
34970 however, that this may cause conflicts if multiple libraries try
34971 to provide the same instantiations. For greater control, use
34972 explicit instantiation as described in the next option.
34974 2. Compile your code with `-fno-implicit-templates' to disable the
34975 implicit generation of template instances, and explicitly
34976 instantiate all the ones you use. This approach requires more
34977 knowledge of exactly which instances you need than do the others,
34978 but it's less mysterious and allows greater control. You can
34979 scatter the explicit instantiations throughout your program,
34980 perhaps putting them in the translation units where the instances
34981 are used or the translation units that define the templates
34982 themselves; you can put all of the explicit instantiations you
34983 need into one big file; or you can create small files like
34988 template class Foo<int>;
34989 template ostream& operator <<
34990 (ostream&, const Foo<int>&);
34992 for each of the instances you need, and create a template
34993 instantiation library from those.
34995 If you are using Cfront-model code, you can probably get away with
34996 not using `-fno-implicit-templates' when compiling files that don't
34997 `#include' the member template definitions.
34999 If you use one big file to do the instantiations, you may want to
35000 compile it without `-fno-implicit-templates' so you get all of the
35001 instances required by your explicit instantiations (but not by any
35002 other files) without having to specify them as well.
35004 G++ has extended the template instantiation syntax given in the ISO
35005 standard to allow forward declaration of explicit instantiations
35006 (with `extern'), instantiation of the compiler support data for a
35007 template class (i.e. the vtable) without instantiating any of its
35008 members (with `inline'), and instantiation of only the static data
35009 members of a template class, without the support data or member
35010 functions (with (`static'):
35012 extern template int max (int, int);
35013 inline template class Foo<int>;
35014 static template class Foo<int>;
35016 3. Do nothing. Pretend G++ does implement automatic instantiation
35017 management. Code written for the Borland model will work fine, but
35018 each translation unit will contain instances of each of the
35019 templates it uses. In a large program, this can lead to an
35020 unacceptable amount of code duplication.
35023 File: gcc.info, Node: Bound member functions, Next: C++ Attributes, Prev: Template Instantiation, Up: C++ Extensions
35025 6.6 Extracting the function pointer from a bound pointer to member function
35026 ===========================================================================
35028 In C++, pointer to member functions (PMFs) are implemented using a wide
35029 pointer of sorts to handle all the possible call mechanisms; the PMF
35030 needs to store information about how to adjust the `this' pointer, and
35031 if the function pointed to is virtual, where to find the vtable, and
35032 where in the vtable to look for the member function. If you are using
35033 PMFs in an inner loop, you should really reconsider that decision. If
35034 that is not an option, you can extract the pointer to the function that
35035 would be called for a given object/PMF pair and call it directly inside
35036 the inner loop, to save a bit of time.
35038 Note that you will still be paying the penalty for the call through a
35039 function pointer; on most modern architectures, such a call defeats the
35040 branch prediction features of the CPU. This is also true of normal
35041 virtual function calls.
35043 The syntax for this extension is
35046 extern int (A::*fp)();
35047 typedef int (*fptr)(A *);
35049 fptr p = (fptr)(a.*fp);
35051 For PMF constants (i.e. expressions of the form `&Klasse::Member'), no
35052 object is needed to obtain the address of the function. They can be
35053 converted to function pointers directly:
35055 fptr p1 = (fptr)(&A::foo);
35057 You must specify `-Wno-pmf-conversions' to use this extension.
35060 File: gcc.info, Node: C++ Attributes, Next: Namespace Association, Prev: Bound member functions, Up: C++ Extensions
35062 6.7 C++-Specific Variable, Function, and Type Attributes
35063 ========================================================
35065 Some attributes only make sense for C++ programs.
35067 `init_priority (PRIORITY)'
35068 In Standard C++, objects defined at namespace scope are guaranteed
35069 to be initialized in an order in strict accordance with that of
35070 their definitions _in a given translation unit_. No guarantee is
35071 made for initializations across translation units. However, GNU
35072 C++ allows users to control the order of initialization of objects
35073 defined at namespace scope with the `init_priority' attribute by
35074 specifying a relative PRIORITY, a constant integral expression
35075 currently bounded between 101 and 65535 inclusive. Lower numbers
35076 indicate a higher priority.
35078 In the following example, `A' would normally be created before
35079 `B', but the `init_priority' attribute has reversed that order:
35081 Some_Class A __attribute__ ((init_priority (2000)));
35082 Some_Class B __attribute__ ((init_priority (543)));
35084 Note that the particular values of PRIORITY do not matter; only
35085 their relative ordering.
35088 This type attribute informs C++ that the class is a Java
35089 interface. It may only be applied to classes declared within an
35090 `extern "Java"' block. Calls to methods declared in this
35091 interface will be dispatched using GCJ's interface table
35092 mechanism, instead of regular virtual table dispatch.
35095 See also *note Namespace Association::.
35098 File: gcc.info, Node: Namespace Association, Next: Type Traits, Prev: C++ Attributes, Up: C++ Extensions
35100 6.8 Namespace Association
35101 =========================
35103 *Caution:* The semantics of this extension are not fully defined.
35104 Users should refrain from using this extension as its semantics may
35105 change subtly over time. It is possible that this extension will be
35106 removed in future versions of G++.
35108 A using-directive with `__attribute ((strong))' is stronger than a
35109 normal using-directive in two ways:
35111 * Templates from the used namespace can be specialized and explicitly
35112 instantiated as though they were members of the using namespace.
35114 * The using namespace is considered an associated namespace of all
35115 templates in the used namespace for purposes of argument-dependent
35118 The used namespace must be nested within the using namespace so that
35119 normal unqualified lookup works properly.
35121 This is useful for composing a namespace transparently from
35122 implementation namespaces. For example:
35126 template <class T> struct A { };
35128 using namespace debug __attribute ((__strong__));
35129 template <> struct A<int> { }; // ok to specialize
35131 template <class T> void f (A<T>);
35136 f (std::A<float>()); // lookup finds std::f
35141 File: gcc.info, Node: Type Traits, Next: Java Exceptions, Prev: Namespace Association, Up: C++ Extensions
35146 The C++ front-end implements syntactic extensions that allow to
35147 determine at compile time various characteristics of a type (or of a
35150 `__has_nothrow_assign (type)'
35151 If `type' is const qualified or is a reference type then the trait
35152 is false. Otherwise if `__has_trivial_assign (type)' is true then
35153 the trait is true, else if `type' is a cv class or union type with
35154 copy assignment operators that are known not to throw an exception
35155 then the trait is true, else it is false. Requires: `type' shall
35156 be a complete type, an array type of unknown bound, or is a `void'
35159 `__has_nothrow_copy (type)'
35160 If `__has_trivial_copy (type)' is true then the trait is true,
35161 else if `type' is a cv class or union type with copy constructors
35162 that are known not to throw an exception then the trait is true,
35163 else it is false. Requires: `type' shall be a complete type, an
35164 array type of unknown bound, or is a `void' type.
35166 `__has_nothrow_constructor (type)'
35167 If `__has_trivial_constructor (type)' is true then the trait is
35168 true, else if `type' is a cv class or union type (or array
35169 thereof) with a default constructor that is known not to throw an
35170 exception then the trait is true, else it is false. Requires:
35171 `type' shall be a complete type, an array type of unknown bound,
35172 or is a `void' type.
35174 `__has_trivial_assign (type)'
35175 If `type' is const qualified or is a reference type then the trait
35176 is false. Otherwise if `__is_pod (type)' is true then the trait is
35177 true, else if `type' is a cv class or union type with a trivial
35178 copy assignment ([class.copy]) then the trait is true, else it is
35179 false. Requires: `type' shall be a complete type, an array type
35180 of unknown bound, or is a `void' type.
35182 `__has_trivial_copy (type)'
35183 If `__is_pod (type)' is true or `type' is a reference type then
35184 the trait is true, else if `type' is a cv class or union type with
35185 a trivial copy constructor ([class.copy]) then the trait is true,
35186 else it is false. Requires: `type' shall be a complete type, an
35187 array type of unknown bound, or is a `void' type.
35189 `__has_trivial_constructor (type)'
35190 If `__is_pod (type)' is true then the trait is true, else if
35191 `type' is a cv class or union type (or array thereof) with a
35192 trivial default constructor ([class.ctor]) then the trait is true,
35193 else it is false. Requires: `type' shall be a complete type, an
35194 array type of unknown bound, or is a `void' type.
35196 `__has_trivial_destructor (type)'
35197 If `__is_pod (type)' is true or `type' is a reference type then
35198 the trait is true, else if `type' is a cv class or union type (or
35199 array thereof) with a trivial destructor ([class.dtor]) then the
35200 trait is true, else it is false. Requires: `type' shall be a
35201 complete type, an array type of unknown bound, or is a `void' type.
35203 `__has_virtual_destructor (type)'
35204 If `type' is a class type with a virtual destructor ([class.dtor])
35205 then the trait is true, else it is false. Requires: `type' shall
35206 be a complete type, an array type of unknown bound, or is a `void'
35209 `__is_abstract (type)'
35210 If `type' is an abstract class ([class.abstract]) then the trait
35211 is true, else it is false. Requires: `type' shall be a complete
35212 type, an array type of unknown bound, or is a `void' type.
35214 `__is_base_of (base_type, derived_type)'
35215 If `base_type' is a base class of `derived_type' ([class.derived])
35216 then the trait is true, otherwise it is false. Top-level cv
35217 qualifications of `base_type' and `derived_type' are ignored. For
35218 the purposes of this trait, a class type is considered is own
35219 base. Requires: if `__is_class (base_type)' and `__is_class
35220 (derived_type)' are true and `base_type' and `derived_type' are
35221 not the same type (disregarding cv-qualifiers), `derived_type'
35222 shall be a complete type. Diagnostic is produced if this
35223 requirement is not met.
35225 `__is_class (type)'
35226 If `type' is a cv class type, and not a union type
35227 ([basic.compound]) the trait is true, else it is false.
35229 `__is_empty (type)'
35230 If `__is_class (type)' is false then the trait is false.
35231 Otherwise `type' is considered empty if and only if: `type' has no
35232 non-static data members, or all non-static data members, if any,
35233 are bit-fields of length 0, and `type' has no virtual members, and
35234 `type' has no virtual base classes, and `type' has no base classes
35235 `base_type' for which `__is_empty (base_type)' is false.
35236 Requires: `type' shall be a complete type, an array type of
35237 unknown bound, or is a `void' type.
35240 If `type' is a cv enumeration type ([basic.compound]) the trait is
35241 true, else it is false.
35244 If `type' is a cv POD type ([basic.types]) then the trait is true,
35245 else it is false. Requires: `type' shall be a complete type, an
35246 array type of unknown bound, or is a `void' type.
35248 `__is_polymorphic (type)'
35249 If `type' is a polymorphic class ([class.virtual]) then the trait
35250 is true, else it is false. Requires: `type' shall be a complete
35251 type, an array type of unknown bound, or is a `void' type.
35253 `__is_union (type)'
35254 If `type' is a cv union type ([basic.compound]) the trait is true,
35259 File: gcc.info, Node: Java Exceptions, Next: Deprecated Features, Prev: Type Traits, Up: C++ Extensions
35261 6.10 Java Exceptions
35262 ====================
35264 The Java language uses a slightly different exception handling model
35265 from C++. Normally, GNU C++ will automatically detect when you are
35266 writing C++ code that uses Java exceptions, and handle them
35267 appropriately. However, if C++ code only needs to execute destructors
35268 when Java exceptions are thrown through it, GCC will guess incorrectly.
35269 Sample problematic code is:
35271 struct S { ~S(); };
35272 extern void bar(); // is written in Java, and may throw exceptions
35279 The usual effect of an incorrect guess is a link failure, complaining of
35280 a missing routine called `__gxx_personality_v0'.
35282 You can inform the compiler that Java exceptions are to be used in a
35283 translation unit, irrespective of what it might think, by writing
35284 `#pragma GCC java_exceptions' at the head of the file. This `#pragma'
35285 must appear before any functions that throw or catch exceptions, or run
35286 destructors when exceptions are thrown through them.
35288 You cannot mix Java and C++ exceptions in the same translation unit.
35289 It is believed to be safe to throw a C++ exception from one file through
35290 another file compiled for the Java exception model, or vice versa, but
35291 there may be bugs in this area.
35294 File: gcc.info, Node: Deprecated Features, Next: Backwards Compatibility, Prev: Java Exceptions, Up: C++ Extensions
35296 6.11 Deprecated Features
35297 ========================
35299 In the past, the GNU C++ compiler was extended to experiment with new
35300 features, at a time when the C++ language was still evolving. Now that
35301 the C++ standard is complete, some of those features are superseded by
35302 superior alternatives. Using the old features might cause a warning in
35303 some cases that the feature will be dropped in the future. In other
35304 cases, the feature might be gone already.
35306 While the list below is not exhaustive, it documents some of the
35307 options that are now deprecated:
35309 `-fexternal-templates'
35310 `-falt-external-templates'
35311 These are two of the many ways for G++ to implement template
35312 instantiation. *Note Template Instantiation::. The C++ standard
35313 clearly defines how template definitions have to be organized
35314 across implementation units. G++ has an implicit instantiation
35315 mechanism that should work just fine for standard-conforming code.
35317 `-fstrict-prototype'
35318 `-fno-strict-prototype'
35319 Previously it was possible to use an empty prototype parameter
35320 list to indicate an unspecified number of parameters (like C),
35321 rather than no parameters, as C++ demands. This feature has been
35322 removed, except where it is required for backwards compatibility.
35323 *Note Backwards Compatibility::.
35325 G++ allows a virtual function returning `void *' to be overridden by
35326 one returning a different pointer type. This extension to the
35327 covariant return type rules is now deprecated and will be removed from a
35330 The G++ minimum and maximum operators (`<?' and `>?') and their
35331 compound forms (`<?=') and `>?=') have been deprecated and are now
35332 removed from G++. Code using these operators should be modified to use
35333 `std::min' and `std::max' instead.
35335 The named return value extension has been deprecated, and is now
35338 The use of initializer lists with new expressions has been deprecated,
35339 and is now removed from G++.
35341 Floating and complex non-type template parameters have been deprecated,
35342 and are now removed from G++.
35344 The implicit typename extension has been deprecated and is now removed
35347 The use of default arguments in function pointers, function typedefs
35348 and other places where they are not permitted by the standard is
35349 deprecated and will be removed from a future version of G++.
35351 G++ allows floating-point literals to appear in integral constant
35352 expressions, e.g. ` enum E { e = int(2.2 * 3.7) } ' This extension is
35353 deprecated and will be removed from a future version.
35355 G++ allows static data members of const floating-point type to be
35356 declared with an initializer in a class definition. The standard only
35357 allows initializers for static members of const integral types and const
35358 enumeration types so this extension has been deprecated and will be
35359 removed from a future version.
35362 File: gcc.info, Node: Backwards Compatibility, Prev: Deprecated Features, Up: C++ Extensions
35364 6.12 Backwards Compatibility
35365 ============================
35367 Now that there is a definitive ISO standard C++, G++ has a specification
35368 to adhere to. The C++ language evolved over time, and features that
35369 used to be acceptable in previous drafts of the standard, such as the
35370 ARM [Annotated C++ Reference Manual], are no longer accepted. In order
35371 to allow compilation of C++ written to such drafts, G++ contains some
35372 backwards compatibilities. _All such backwards compatibility features
35373 are liable to disappear in future versions of G++._ They should be
35374 considered deprecated. *Note Deprecated Features::.
35377 If a variable is declared at for scope, it used to remain in scope
35378 until the end of the scope which contained the for statement
35379 (rather than just within the for scope). G++ retains this, but
35380 issues a warning, if such a variable is accessed outside the for
35383 `Implicit C language'
35384 Old C system header files did not contain an `extern "C" {...}'
35385 scope to set the language. On such systems, all header files are
35386 implicitly scoped inside a C language scope. Also, an empty
35387 prototype `()' will be treated as an unspecified number of
35388 arguments, rather than no arguments, as C++ demands.
35391 File: gcc.info, Node: Objective-C, Next: Compatibility, Prev: C++ Extensions, Up: Top
35393 7 GNU Objective-C runtime features
35394 **********************************
35396 This document is meant to describe some of the GNU Objective-C runtime
35397 features. It is not intended to teach you Objective-C, there are
35398 several resources on the Internet that present the language. Questions
35399 and comments about this document to Ovidiu Predescu <ovidiu@cup.hp.com>.
35403 * Executing code before main::
35405 * Garbage Collection::
35406 * Constant string objects::
35407 * compatibility_alias::
35410 File: gcc.info, Node: Executing code before main, Next: Type encoding, Prev: Objective-C, Up: Objective-C
35412 7.1 `+load': Executing code before main
35413 =======================================
35415 The GNU Objective-C runtime provides a way that allows you to execute
35416 code before the execution of the program enters the `main' function.
35417 The code is executed on a per-class and a per-category basis, through a
35418 special class method `+load'.
35420 This facility is very useful if you want to initialize global variables
35421 which can be accessed by the program directly, without sending a message
35422 to the class first. The usual way to initialize global variables, in
35423 the `+initialize' method, might not be useful because `+initialize' is
35424 only called when the first message is sent to a class object, which in
35425 some cases could be too late.
35427 Suppose for example you have a `FileStream' class that declares
35428 `Stdin', `Stdout' and `Stderr' as global variables, like below:
35431 FileStream *Stdin = nil;
35432 FileStream *Stdout = nil;
35433 FileStream *Stderr = nil;
35435 @implementation FileStream
35439 Stdin = [[FileStream new] initWithFd:0];
35440 Stdout = [[FileStream new] initWithFd:1];
35441 Stderr = [[FileStream new] initWithFd:2];
35444 /* Other methods here */
35447 In this example, the initialization of `Stdin', `Stdout' and `Stderr'
35448 in `+initialize' occurs too late. The programmer can send a message to
35449 one of these objects before the variables are actually initialized,
35450 thus sending messages to the `nil' object. The `+initialize' method
35451 which actually initializes the global variables is not invoked until
35452 the first message is sent to the class object. The solution would
35453 require these variables to be initialized just before entering `main'.
35455 The correct solution of the above problem is to use the `+load' method
35456 instead of `+initialize':
35459 @implementation FileStream
35463 Stdin = [[FileStream new] initWithFd:0];
35464 Stdout = [[FileStream new] initWithFd:1];
35465 Stderr = [[FileStream new] initWithFd:2];
35468 /* Other methods here */
35471 The `+load' is a method that is not overridden by categories. If a
35472 class and a category of it both implement `+load', both methods are
35473 invoked. This allows some additional initializations to be performed in
35476 This mechanism is not intended to be a replacement for `+initialize'.
35477 You should be aware of its limitations when you decide to use it
35478 instead of `+initialize'.
35482 * What you can and what you cannot do in +load::
35485 File: gcc.info, Node: What you can and what you cannot do in +load, Prev: Executing code before main, Up: Executing code before main
35487 7.1.1 What you can and what you cannot do in `+load'
35488 ----------------------------------------------------
35490 The `+load' implementation in the GNU runtime guarantees you the
35493 * you can write whatever C code you like;
35495 * you can send messages to Objective-C constant strings (`@"this is a
35496 constant string"');
35498 * you can allocate and send messages to objects whose class is
35499 implemented in the same file;
35501 * the `+load' implementation of all super classes of a class are
35502 executed before the `+load' of that class is executed;
35504 * the `+load' implementation of a class is executed before the
35505 `+load' implementation of any category.
35508 In particular, the following things, even if they can work in a
35509 particular case, are not guaranteed:
35511 * allocation of or sending messages to arbitrary objects;
35513 * allocation of or sending messages to objects whose classes have a
35514 category implemented in the same file;
35517 You should make no assumptions about receiving `+load' in sibling
35518 classes when you write `+load' of a class. The order in which sibling
35519 classes receive `+load' is not guaranteed.
35521 The order in which `+load' and `+initialize' are called could be
35522 problematic if this matters. If you don't allocate objects inside
35523 `+load', it is guaranteed that `+load' is called before `+initialize'.
35524 If you create an object inside `+load' the `+initialize' method of
35525 object's class is invoked even if `+load' was not invoked. Note if you
35526 explicitly call `+load' on a class, `+initialize' will be called first.
35527 To avoid possible problems try to implement only one of these methods.
35529 The `+load' method is also invoked when a bundle is dynamically loaded
35530 into your running program. This happens automatically without any
35531 intervening operation from you. When you write bundles and you need to
35532 write `+load' you can safely create and send messages to objects whose
35533 classes already exist in the running program. The same restrictions as
35534 above apply to classes defined in bundle.
35537 File: gcc.info, Node: Type encoding, Next: Garbage Collection, Prev: Executing code before main, Up: Objective-C
35542 The Objective-C compiler generates type encodings for all the types.
35543 These type encodings are used at runtime to find out information about
35544 selectors and methods and about objects and classes.
35546 The types are encoded in the following way:
35550 `unsigned char' `C'
35552 `unsigned short' `S'
35556 `unsigned long' `L'
35568 Complex types `j' followed by the inner type. For example
35569 `_Complex double' is encoded as "jd".
35570 bit-fields `b' followed by the starting position of the
35571 bit-field, the type of the bit-field and the size of
35572 the bit-field (the bit-fields encoding was changed
35573 from the NeXT's compiler encoding, see below)
35575 The encoding of bit-fields has changed to allow bit-fields to be
35576 properly handled by the runtime functions that compute sizes and
35577 alignments of types that contain bit-fields. The previous encoding
35578 contained only the size of the bit-field. Using only this information
35579 it is not possible to reliably compute the size occupied by the
35580 bit-field. This is very important in the presence of the Boehm's
35581 garbage collector because the objects are allocated using the typed
35582 memory facility available in this collector. The typed memory
35583 allocation requires information about where the pointers are located
35586 The position in the bit-field is the position, counting in bits, of the
35587 bit closest to the beginning of the structure.
35589 The non-atomic types are encoded as follows:
35591 pointers `^' followed by the pointed type.
35592 arrays `[' followed by the number of elements in the array
35593 followed by the type of the elements followed by `]'
35594 structures `{' followed by the name of the structure (or `?' if the
35595 structure is unnamed), the `=' sign, the type of the
35597 unions `(' followed by the name of the structure (or `?' if the
35598 union is unnamed), the `=' sign, the type of the members
35601 Here are some types and their encodings, as they are generated by the
35602 compiler on an i386 machine:
35605 Objective-C type Compiler encoding
35607 struct { `{?=i[3f]b128i3b131i2c}'
35616 In addition to the types the compiler also encodes the type
35617 specifiers. The table below describes the encoding of the current
35618 Objective-C type specifiers:
35630 The type specifiers are encoded just before the type. Unlike types
35631 however, the type specifiers are only encoded when they appear in method
35635 File: gcc.info, Node: Garbage Collection, Next: Constant string objects, Prev: Type encoding, Up: Objective-C
35637 7.3 Garbage Collection
35638 ======================
35640 Support for a new memory management policy has been added by using a
35641 powerful conservative garbage collector, known as the
35642 Boehm-Demers-Weiser conservative garbage collector. It is available
35643 from `http://www.hpl.hp.com/personal/Hans_Boehm/gc/'.
35645 To enable the support for it you have to configure the compiler using
35646 an additional argument, `--enable-objc-gc'. You need to have garbage
35647 collector installed before building the compiler. This will build an
35648 additional runtime library which has several enhancements to support
35649 the garbage collector. The new library has a new name, `libobjc_gc.a'
35650 to not conflict with the non-garbage-collected library.
35652 When the garbage collector is used, the objects are allocated using the
35653 so-called typed memory allocation mechanism available in the
35654 Boehm-Demers-Weiser collector. This mode requires precise information
35655 on where pointers are located inside objects. This information is
35656 computed once per class, immediately after the class has been
35659 There is a new runtime function `class_ivar_set_gcinvisible()' which
35660 can be used to declare a so-called "weak pointer" reference. Such a
35661 pointer is basically hidden for the garbage collector; this can be
35662 useful in certain situations, especially when you want to keep track of
35663 the allocated objects, yet allow them to be collected. This kind of
35664 pointers can only be members of objects, you cannot declare a global
35665 pointer as a weak reference. Every type which is a pointer type can be
35666 declared a weak pointer, including `id', `Class' and `SEL'.
35668 Here is an example of how to use this feature. Suppose you want to
35669 implement a class whose instances hold a weak pointer reference; the
35670 following class does this:
35673 @interface WeakPointer : Object
35675 const void* weakPointer;
35678 - initWithPointer:(const void*)p;
35679 - (const void*)weakPointer;
35683 @implementation WeakPointer
35687 class_ivar_set_gcinvisible (self, "weakPointer", YES);
35690 - initWithPointer:(const void*)p
35696 - (const void*)weakPointer
35698 return weakPointer;
35703 Weak pointers are supported through a new type character specifier
35704 represented by the `!' character. The `class_ivar_set_gcinvisible()'
35705 function adds or removes this specifier to the string type description
35706 of the instance variable named as argument.
35709 File: gcc.info, Node: Constant string objects, Next: compatibility_alias, Prev: Garbage Collection, Up: Objective-C
35711 7.4 Constant string objects
35712 ===========================
35714 GNU Objective-C provides constant string objects that are generated
35715 directly by the compiler. You declare a constant string object by
35716 prefixing a C constant string with the character `@':
35718 id myString = @"this is a constant string object";
35720 The constant string objects are by default instances of the
35721 `NXConstantString' class which is provided by the GNU Objective-C
35722 runtime. To get the definition of this class you must include the
35723 `objc/NXConstStr.h' header file.
35725 User defined libraries may want to implement their own constant string
35726 class. To be able to support them, the GNU Objective-C compiler
35727 provides a new command line options
35728 `-fconstant-string-class=CLASS-NAME'. The provided class should adhere
35729 to a strict structure, the same as `NXConstantString''s structure:
35732 @interface MyConstantStringClass
35740 `NXConstantString' inherits from `Object'; user class libraries may
35741 choose to inherit the customized constant string class from a different
35742 class than `Object'. There is no requirement in the methods the
35743 constant string class has to implement, but the final ivar layout of
35744 the class must be the compatible with the given structure.
35746 When the compiler creates the statically allocated constant string
35747 object, the `c_string' field will be filled by the compiler with the
35748 string; the `length' field will be filled by the compiler with the
35749 string length; the `isa' pointer will be filled with `NULL' by the
35750 compiler, and it will later be fixed up automatically at runtime by the
35751 GNU Objective-C runtime library to point to the class which was set by
35752 the `-fconstant-string-class' option when the object file is loaded (if
35753 you wonder how it works behind the scenes, the name of the class to
35754 use, and the list of static objects to fixup, are stored by the
35755 compiler in the object file in a place where the GNU runtime library
35756 will find them at runtime).
35758 As a result, when a file is compiled with the
35759 `-fconstant-string-class' option, all the constant string objects will
35760 be instances of the class specified as argument to this option. It is
35761 possible to have multiple compilation units referring to different
35762 constant string classes, neither the compiler nor the linker impose any
35763 restrictions in doing this.
35766 File: gcc.info, Node: compatibility_alias, Prev: Constant string objects, Up: Objective-C
35768 7.5 compatibility_alias
35769 =======================
35771 This is a feature of the Objective-C compiler rather than of the
35772 runtime, anyway since it is documented nowhere and its existence was
35773 forgotten, we are documenting it here.
35775 The keyword `@compatibility_alias' allows you to define a class name
35776 as equivalent to another class name. For example:
35778 @compatibility_alias WOApplication GSWApplication;
35780 tells the compiler that each time it encounters `WOApplication' as a
35781 class name, it should replace it with `GSWApplication' (that is,
35782 `WOApplication' is just an alias for `GSWApplication').
35784 There are some constraints on how this can be used--
35786 * `WOApplication' (the alias) must not be an existing class;
35788 * `GSWApplication' (the real class) must be an existing class.
35792 File: gcc.info, Node: Compatibility, Next: Gcov, Prev: Objective-C, Up: Top
35794 8 Binary Compatibility
35795 **********************
35797 Binary compatibility encompasses several related concepts:
35799 "application binary interface (ABI)"
35800 The set of runtime conventions followed by all of the tools that
35801 deal with binary representations of a program, including
35802 compilers, assemblers, linkers, and language runtime support.
35803 Some ABIs are formal with a written specification, possibly
35804 designed by multiple interested parties. Others are simply the
35805 way things are actually done by a particular set of tools.
35808 A compiler conforms to an ABI if it generates code that follows
35809 all of the specifications enumerated by that ABI. A library
35810 conforms to an ABI if it is implemented according to that ABI. An
35811 application conforms to an ABI if it is built using tools that
35812 conform to that ABI and does not contain source code that
35813 specifically changes behavior specified by the ABI.
35815 "calling conventions"
35816 Calling conventions are a subset of an ABI that specify of how
35817 arguments are passed and function results are returned.
35820 Different sets of tools are interoperable if they generate files
35821 that can be used in the same program. The set of tools includes
35822 compilers, assemblers, linkers, libraries, header files, startup
35823 files, and debuggers. Binaries produced by different sets of
35824 tools are not interoperable unless they implement the same ABI.
35825 This applies to different versions of the same tools as well as
35826 tools from different vendors.
35829 Whether a function in a binary built by one set of tools can call a
35830 function in a binary built by a different set of tools is a subset
35831 of interoperability.
35833 "implementation-defined features"
35834 Language standards include lists of implementation-defined
35835 features whose behavior can vary from one implementation to
35836 another. Some of these features are normally covered by a
35837 platform's ABI and others are not. The features that are not
35838 covered by an ABI generally affect how a program behaves, but not
35842 Conformance to the same ABI and the same behavior of
35843 implementation-defined features are both relevant for
35846 The application binary interface implemented by a C or C++ compiler
35847 affects code generation and runtime support for:
35849 * size and alignment of data types
35851 * layout of structured types
35853 * calling conventions
35855 * register usage conventions
35857 * interfaces for runtime arithmetic support
35859 * object file formats
35861 In addition, the application binary interface implemented by a C++
35862 compiler affects code generation and runtime support for:
35865 * exception handling
35867 * invoking constructors and destructors
35869 * layout, alignment, and padding of classes
35871 * layout and alignment of virtual tables
35873 Some GCC compilation options cause the compiler to generate code that
35874 does not conform to the platform's default ABI. Other options cause
35875 different program behavior for implementation-defined features that are
35876 not covered by an ABI. These options are provided for consistency with
35877 other compilers that do not follow the platform's default ABI or the
35878 usual behavior of implementation-defined features for the platform. Be
35879 very careful about using such options.
35881 Most platforms have a well-defined ABI that covers C code, but ABIs
35882 that cover C++ functionality are not yet common.
35884 Starting with GCC 3.2, GCC binary conventions for C++ are based on a
35885 written, vendor-neutral C++ ABI that was designed to be specific to
35886 64-bit Itanium but also includes generic specifications that apply to
35887 any platform. This C++ ABI is also implemented by other compiler
35888 vendors on some platforms, notably GNU/Linux and BSD systems. We have
35889 tried hard to provide a stable ABI that will be compatible with future
35890 GCC releases, but it is possible that we will encounter problems that
35891 make this difficult. Such problems could include different
35892 interpretations of the C++ ABI by different vendors, bugs in the ABI, or
35893 bugs in the implementation of the ABI in different compilers. GCC's
35894 `-Wabi' switch warns when G++ generates code that is probably not
35895 compatible with the C++ ABI.
35897 The C++ library used with a C++ compiler includes the Standard C++
35898 Library, with functionality defined in the C++ Standard, plus language
35899 runtime support. The runtime support is included in a C++ ABI, but
35900 there is no formal ABI for the Standard C++ Library. Two
35901 implementations of that library are interoperable if one follows the
35902 de-facto ABI of the other and if they are both built with the same
35903 compiler, or with compilers that conform to the same ABI for C++
35904 compiler and runtime support.
35906 When G++ and another C++ compiler conform to the same C++ ABI, but the
35907 implementations of the Standard C++ Library that they normally use do
35908 not follow the same ABI for the Standard C++ Library, object files
35909 built with those compilers can be used in the same program only if they
35910 use the same C++ library. This requires specifying the location of the
35911 C++ library header files when invoking the compiler whose usual library
35912 is not being used. The location of GCC's C++ header files depends on
35913 how the GCC build was configured, but can be seen by using the G++ `-v'
35914 option. With default configuration options for G++ 3.3 the compile
35915 line for a different C++ compiler needs to include
35917 -IGCC_INSTALL_DIRECTORY/include/c++/3.3
35919 Similarly, compiling code with G++ that must use a C++ library other
35920 than the GNU C++ library requires specifying the location of the header
35921 files for that other library.
35923 The most straightforward way to link a program to use a particular C++
35924 library is to use a C++ driver that specifies that C++ library by
35925 default. The `g++' driver, for example, tells the linker where to find
35926 GCC's C++ library (`libstdc++') plus the other libraries and startup
35927 files it needs, in the proper order.
35929 If a program must use a different C++ library and it's not possible to
35930 do the final link using a C++ driver that uses that library by default,
35931 it is necessary to tell `g++' the location and name of that library.
35932 It might also be necessary to specify different startup files and other
35933 runtime support libraries, and to suppress the use of GCC's support
35934 libraries with one or more of the options `-nostdlib', `-nostartfiles',
35935 and `-nodefaultlibs'.
35938 File: gcc.info, Node: Gcov, Next: Trouble, Prev: Compatibility, Up: Top
35940 9 `gcov'--a Test Coverage Program
35941 *********************************
35943 `gcov' is a tool you can use in conjunction with GCC to test code
35944 coverage in your programs.
35948 * Gcov Intro:: Introduction to gcov.
35949 * Invoking Gcov:: How to use gcov.
35950 * Gcov and Optimization:: Using gcov with GCC optimization.
35951 * Gcov Data Files:: The files used by gcov.
35952 * Cross-profiling:: Data file relocation.
35955 File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov
35957 9.1 Introduction to `gcov'
35958 ==========================
35960 `gcov' is a test coverage program. Use it in concert with GCC to
35961 analyze your programs to help create more efficient, faster running
35962 code and to discover untested parts of your program. You can use
35963 `gcov' as a profiling tool to help discover where your optimization
35964 efforts will best affect your code. You can also use `gcov' along with
35965 the other profiling tool, `gprof', to assess which parts of your code
35966 use the greatest amount of computing time.
35968 Profiling tools help you analyze your code's performance. Using a
35969 profiler such as `gcov' or `gprof', you can find out some basic
35970 performance statistics, such as:
35972 * how often each line of code executes
35974 * what lines of code are actually executed
35976 * how much computing time each section of code uses
35978 Once you know these things about how your code works when compiled, you
35979 can look at each module to see which modules should be optimized.
35980 `gcov' helps you determine where to work on optimization.
35982 Software developers also use coverage testing in concert with
35983 testsuites, to make sure software is actually good enough for a release.
35984 Testsuites can verify that a program works as expected; a coverage
35985 program tests to see how much of the program is exercised by the
35986 testsuite. Developers can then determine what kinds of test cases need
35987 to be added to the testsuites to create both better testing and a better
35990 You should compile your code without optimization if you plan to use
35991 `gcov' because the optimization, by combining some lines of code into
35992 one function, may not give you as much information as you need to look
35993 for `hot spots' where the code is using a great deal of computer time.
35994 Likewise, because `gcov' accumulates statistics by line (at the lowest
35995 resolution), it works best with a programming style that places only
35996 one statement on each line. If you use complicated macros that expand
35997 to loops or to other control structures, the statistics are less
35998 helpful--they only report on the line where the macro call appears. If
35999 your complex macros behave like functions, you can replace them with
36000 inline functions to solve this problem.
36002 `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates how
36003 many times each line of a source file `SOURCEFILE.c' has executed. You
36004 can use these logfiles along with `gprof' to aid in fine-tuning the
36005 performance of your programs. `gprof' gives timing information you can
36006 use along with the information you get from `gcov'.
36008 `gcov' works only on code compiled with GCC. It is not compatible
36009 with any other profiling or test coverage mechanism.
36012 File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov
36014 9.2 Invoking `gcov'
36015 ===================
36017 gcov [OPTIONS] SOURCEFILES
36019 `gcov' accepts the following options:
36023 Display help about using `gcov' (on the standard output), and exit
36024 without doing any further processing.
36028 Display the `gcov' version number (on the standard output), and
36029 exit without doing any further processing.
36033 Write individual execution counts for every basic block. Normally
36034 gcov outputs execution counts only for the main blocks of a line.
36035 With this option you can determine if blocks within a single line
36036 are not being executed.
36039 `--branch-probabilities'
36040 Write branch frequencies to the output file, and write branch
36041 summary info to the standard output. This option allows you to
36042 see how often each branch in your program was taken.
36043 Unconditional branches will not be shown, unless the `-u' option
36048 Write branch frequencies as the number of branches taken, rather
36049 than the percentage of branches taken.
36053 Do not create the `gcov' output file.
36056 `--long-file-names'
36057 Create long file names for included source files. For example, if
36058 the header file `x.h' contains code, and was included in the file
36059 `a.c', then running `gcov' on the file `a.c' will produce an
36060 output file called `a.c##x.h.gcov' instead of `x.h.gcov'. This
36061 can be useful if `x.h' is included in multiple source files. If
36062 you use the `-p' option, both the including and included file
36063 names will be complete path names.
36067 Preserve complete path information in the names of generated
36068 `.gcov' files. Without this option, just the filename component is
36069 used. With this option, all directories are used, with `/'
36070 characters translated to `#' characters, `.' directory components
36071 removed and `..' components renamed to `^'. This is useful if
36072 sourcefiles are in several different directories. It also affects
36076 `--function-summaries'
36077 Output summaries for each function in addition to the file level
36080 `-o DIRECTORY|FILE'
36081 `--object-directory DIRECTORY'
36082 `--object-file FILE'
36083 Specify either the directory containing the gcov data files, or the
36084 object path name. The `.gcno', and `.gcda' data files are
36085 searched for using this option. If a directory is specified, the
36086 data files are in that directory and named after the source file
36087 name, without its extension. If a file is specified here, the
36088 data files are named after that file, without its extension. If
36089 this option is not supplied, it defaults to the current directory.
36092 `--unconditional-branches'
36093 When branch probabilities are given, include those of
36094 unconditional branches. Unconditional branches are normally not
36098 `--intermediate-format'
36099 Output gcov file in an intermediate text format that can be used by
36100 `lcov' or other applications. It will output a single *.gcov file
36101 per *gcda file. No source code required.
36104 `gcov' should be run with the current directory the same as that when
36105 you invoked the compiler. Otherwise it will not be able to locate the
36106 source files. `gcov' produces files called `MANGLEDNAME.gcov' in the
36107 current directory. These contain the coverage information of the
36108 source file they correspond to. One `.gcov' file is produced for each
36109 source file containing code, which was compiled to produce the data
36110 files. The MANGLEDNAME part of the output file name is usually simply
36111 the source file name, but can be something more complicated if the `-l'
36112 or `-p' options are given. Refer to those options for details.
36114 The `.gcov' files contain the `:' separated fields along with program
36115 source code. The format is
36117 EXECUTION_COUNT:LINE_NUMBER:SOURCE LINE TEXT
36119 Additional block information may succeed each line, when requested by
36120 command line option. The EXECUTION_COUNT is `-' for lines containing
36121 no code and `#####' for lines which were never executed. Some lines of
36122 information at the start have LINE_NUMBER of zero.
36124 The preamble lines are of the form
36128 The ordering and number of these preamble lines will be augmented as
36129 `gcov' development progresses -- do not rely on them remaining
36130 unchanged. Use TAG to locate a particular preamble line.
36132 The additional block information is of the form
36136 The INFORMATION is human readable, but designed to be simple enough
36137 for machine parsing too.
36139 When printing percentages, 0% and 100% are only printed when the values
36140 are _exactly_ 0% and 100% respectively. Other values which would
36141 conventionally be rounded to 0% or 100% are instead printed as the
36142 nearest non-boundary value.
36144 When using `gcov', you must first compile your program with two
36145 special GCC options: `-fprofile-arcs -ftest-coverage'. This tells the
36146 compiler to generate additional information needed by gcov (basically a
36147 flow graph of the program) and also includes additional code in the
36148 object files for generating the extra profiling information needed by
36149 gcov. These additional files are placed in the directory where the
36150 object file is located.
36152 Running the program will cause profile output to be generated. For
36153 each source file compiled with `-fprofile-arcs', an accompanying
36154 `.gcda' file will be placed in the object file directory.
36156 Running `gcov' with your program's source file names as arguments will
36157 now produce a listing of the code along with frequency of execution for
36158 each line. For example, if your program is called `tmp.c', this is
36159 what you see when you use the basic `gcov' facility:
36161 $ gcc -fprofile-arcs -ftest-coverage tmp.c
36164 90.00% of 10 source lines executed in file tmp.c
36165 Creating tmp.c.gcov.
36167 The file `tmp.c.gcov' contains output from `gcov'. Here is a sample:
36170 -: 0:Graph:tmp.gcno
36174 -: 1:#include <stdio.h>
36176 -: 3:int main (void)
36178 1: 5: int i, total;
36182 11: 9: for (i = 0; i < 10; i++)
36183 10: 10: total += i;
36185 1: 12: if (total != 45)
36186 #####: 13: printf ("Failure\n");
36188 1: 15: printf ("Success\n");
36192 When you use the `-a' option, you will get individual block counts,
36193 and the output looks like this:
36196 -: 0:Graph:tmp.gcno
36200 -: 1:#include <stdio.h>
36202 -: 3:int main (void)
36205 1: 5: int i, total;
36209 11: 9: for (i = 0; i < 10; i++)
36211 10: 10: total += i;
36214 1: 12: if (total != 45)
36216 #####: 13: printf ("Failure\n");
36219 1: 15: printf ("Success\n");
36225 In this mode, each basic block is only shown on one line - the last
36226 line of the block. A multi-line block will only contribute to the
36227 execution count of that last line, and other lines will not be shown to
36228 contain code, unless previous blocks end on those lines. The total
36229 execution count of a line is shown and subsequent lines show the
36230 execution counts for individual blocks that end on that line. After
36231 each block, the branch and call counts of the block will be shown, if
36232 the `-b' option is given.
36234 Because of the way GCC instruments calls, a call count can be shown
36235 after a line with no individual blocks. As you can see, line 13
36236 contains a basic block that was not executed.
36238 When you use the `-b' option, your output looks like this:
36241 90.00% of 10 source lines executed in file tmp.c
36242 80.00% of 5 branches executed in file tmp.c
36243 80.00% of 5 branches taken at least once in file tmp.c
36244 50.00% of 2 calls executed in file tmp.c
36245 Creating tmp.c.gcov.
36247 Here is a sample of a resulting `tmp.c.gcov' file:
36250 -: 0:Graph:tmp.gcno
36254 -: 1:#include <stdio.h>
36256 -: 3:int main (void)
36257 function main called 1 returned 1 blocks executed 75%
36259 1: 5: int i, total;
36263 11: 9: for (i = 0; i < 10; i++)
36264 branch 0 taken 91% (fallthrough)
36266 10: 10: total += i;
36268 1: 12: if (total != 45)
36269 branch 0 taken 0% (fallthrough)
36270 branch 1 taken 100%
36271 #####: 13: printf ("Failure\n");
36272 call 0 never executed
36274 1: 15: printf ("Success\n");
36275 call 0 called 1 returned 100%
36279 For each function, a line is printed showing how many times the
36280 function is called, how many times it returns and what percentage of the
36281 function's blocks were executed.
36283 For each basic block, a line is printed after the last line of the
36284 basic block describing the branch or call that ends the basic block.
36285 There can be multiple branches and calls listed for a single source
36286 line if there are multiple basic blocks that end on that line. In this
36287 case, the branches and calls are each given a number. There is no
36288 simple way to map these branches and calls back to source constructs.
36289 In general, though, the lowest numbered branch or call will correspond
36290 to the leftmost construct on the source line.
36292 For a branch, if it was executed at least once, then a percentage
36293 indicating the number of times the branch was taken divided by the
36294 number of times the branch was executed will be printed. Otherwise, the
36295 message "never executed" is printed.
36297 For a call, if it was executed at least once, then a percentage
36298 indicating the number of times the call returned divided by the number
36299 of times the call was executed will be printed. This will usually be
36300 100%, but may be less for functions that call `exit' or `longjmp', and
36301 thus may not return every time they are called.
36303 The execution counts are cumulative. If the example program were
36304 executed again without removing the `.gcda' file, the count for the
36305 number of times each line in the source was executed would be added to
36306 the results of the previous run(s). This is potentially useful in
36307 several ways. For example, it could be used to accumulate data over a
36308 number of program runs as part of a test verification suite, or to
36309 provide more accurate long-term information over a large number of
36312 The data in the `.gcda' files is saved immediately before the program
36313 exits. For each source file compiled with `-fprofile-arcs', the
36314 profiling code first attempts to read in an existing `.gcda' file; if
36315 the file doesn't match the executable (differing number of basic block
36316 counts) it will ignore the contents of the file. It then adds in the
36317 new execution counts and finally writes the data to the file.
36320 File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov
36322 9.3 Using `gcov' with GCC Optimization
36323 ======================================
36325 If you plan to use `gcov' to help optimize your code, you must first
36326 compile your program with two special GCC options: `-fprofile-arcs
36327 -ftest-coverage'. Aside from that, you can use any other GCC options;
36328 but if you want to prove that every single line in your program was
36329 executed, you should not compile with optimization at the same time.
36330 On some machines the optimizer can eliminate some simple code lines by
36331 combining them with other lines. For example, code like this:
36338 can be compiled into one instruction on some machines. In this case,
36339 there is no way for `gcov' to calculate separate execution counts for
36340 each line because there isn't separate code for each line. Hence the
36341 `gcov' output looks like this if you compiled the program with
36344 100: 12:if (a != b)
36349 The output shows that this block of code, combined by optimization,
36350 executed 100 times. In one sense this result is correct, because there
36351 was only one instruction representing all four of these lines. However,
36352 the output does not indicate how many times the result was 0 and how
36353 many times the result was 1.
36355 Inlineable functions can create unexpected line counts. Line counts
36356 are shown for the source code of the inlineable function, but what is
36357 shown depends on where the function is inlined, or if it is not inlined
36360 If the function is not inlined, the compiler must emit an out of line
36361 copy of the function, in any object file that needs it. If `fileA.o'
36362 and `fileB.o' both contain out of line bodies of a particular
36363 inlineable function, they will also both contain coverage counts for
36364 that function. When `fileA.o' and `fileB.o' are linked together, the
36365 linker will, on many systems, select one of those out of line bodies
36366 for all calls to that function, and remove or ignore the other.
36367 Unfortunately, it will not remove the coverage counters for the unused
36368 function body. Hence when instrumented, all but one use of that
36369 function will show zero counts.
36371 If the function is inlined in several places, the block structure in
36372 each location might not be the same. For instance, a condition might
36373 now be calculable at compile time in some instances. Because the
36374 coverage of all the uses of the inline function will be shown for the
36375 same source lines, the line counts themselves might seem inconsistent.
36378 File: gcc.info, Node: Gcov Data Files, Next: Cross-profiling, Prev: Gcov and Optimization, Up: Gcov
36380 9.4 Brief description of `gcov' data files
36381 ==========================================
36383 `gcov' uses two files for profiling. The names of these files are
36384 derived from the original _object_ file by substituting the file suffix
36385 with either `.gcno', or `.gcda'. All of these files are placed in the
36386 same directory as the object file, and contain data stored in a
36387 platform-independent format.
36389 The `.gcno' file is generated when the source file is compiled with
36390 the GCC `-ftest-coverage' option. It contains information to
36391 reconstruct the basic block graphs and assign source line numbers to
36394 The `.gcda' file is generated when a program containing object files
36395 built with the GCC `-fprofile-arcs' option is executed. A separate
36396 `.gcda' file is created for each object file compiled with this option.
36397 It contains arc transition counts, and some summary information.
36399 The full details of the file format is specified in `gcov-io.h', and
36400 functions provided in that header file should be used to access the
36404 File: gcc.info, Node: Cross-profiling, Prev: Gcov Data Files, Up: Gcov
36406 9.5 Data file relocation to support cross-profiling
36407 ===================================================
36409 Running the program will cause profile output to be generated. For each
36410 source file compiled with `-fprofile-arcs', an accompanying `.gcda'
36411 file will be placed in the object file directory. That implicitly
36412 requires running the program on the same system as it was built or
36413 having the same absolute directory structure on the target system. The
36414 program will try to create the needed directory structure, if it is not
36417 To support cross-profiling, a program compiled with `-fprofile-arcs'
36418 can relocate the data files based on two environment variables:
36420 * GCOV_PREFIX contains the prefix to add to the absolute paths in
36421 the object file. Prefix must be absolute as well, otherwise its
36422 value is ignored. The default is no prefix.
36424 * GCOV_PREFIX_STRIP indicates the how many initial directory names
36425 to strip off the hardwired absolute paths. Default value is 0.
36427 _Note:_ GCOV_PREFIX_STRIP has no effect if GCOV_PREFIX is
36428 undefined, empty or non-absolute.
36430 For example, if the object file `/user/build/foo.o' was built with
36431 `-fprofile-arcs', the final executable will try to create the data file
36432 `/user/build/foo.gcda' when running on the target system. This will
36433 fail if the corresponding directory does not exist and it is unable to
36434 create it. This can be overcome by, for example, setting the
36435 environment as `GCOV_PREFIX=/target/run' and `GCOV_PREFIX_STRIP=1'.
36436 Such a setting will name the data file `/target/run/build/foo.gcda'.
36438 You must move the data files to the expected directory tree in order to
36439 use them for profile directed optimizations (`--use-profile'), or to
36440 use the `gcov' tool.
36443 File: gcc.info, Node: Trouble, Next: Bugs, Prev: Gcov, Up: Top
36445 10 Known Causes of Trouble with GCC
36446 ***********************************
36448 This section describes known problems that affect users of GCC. Most
36449 of these are not GCC bugs per se--if they were, we would fix them. But
36450 the result for a user may be like the result of a bug.
36452 Some of these problems are due to bugs in other software, some are
36453 missing features that are too much work to add, and some are places
36454 where people's opinions differ as to what is best.
36458 * Actual Bugs:: Bugs we will fix later.
36459 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
36460 * Interoperation:: Problems using GCC with other compilers,
36461 and with certain linkers, assemblers and debuggers.
36462 * Incompatibilities:: GCC is incompatible with traditional C.
36463 * Fixed Headers:: GCC uses corrected versions of system header files.
36464 This is necessary, but doesn't always work smoothly.
36465 * Standard Libraries:: GCC uses the system C library, which might not be
36466 compliant with the ISO C standard.
36467 * Disappointments:: Regrettable things we can't change, but not quite bugs.
36468 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
36469 * Protoize Caveats:: Things to watch out for when using `protoize'.
36470 * Non-bugs:: Things we think are right, but some others disagree.
36471 * Warnings and Errors:: Which problems in your code get warnings,
36472 and which get errors.
36475 File: gcc.info, Node: Actual Bugs, Next: Cross-Compiler Problems, Up: Trouble
36477 10.1 Actual Bugs We Haven't Fixed Yet
36478 =====================================
36480 * The `fixincludes' script interacts badly with automounters; if the
36481 directory of system header files is automounted, it tends to be
36482 unmounted while `fixincludes' is running. This would seem to be a
36483 bug in the automounter. We don't know any good way to work around
36486 * The `fixproto' script will sometimes add prototypes for the
36487 `sigsetjmp' and `siglongjmp' functions that reference the
36488 `jmp_buf' type before that type is defined. To work around this,
36489 edit the offending file and place the typedef in front of the
36493 File: gcc.info, Node: Cross-Compiler Problems, Next: Interoperation, Prev: Actual Bugs, Up: Trouble
36495 10.2 Cross-Compiler Problems
36496 ============================
36498 You may run into problems with cross compilation on certain machines,
36499 for several reasons.
36501 * At present, the program `mips-tfile' which adds debug support to
36502 object files on MIPS systems does not work in a cross compile
36506 File: gcc.info, Node: Interoperation, Next: Incompatibilities, Prev: Cross-Compiler Problems, Up: Trouble
36508 10.3 Interoperation
36509 ===================
36511 This section lists various difficulties encountered in using GCC
36512 together with other compilers or with the assemblers, linkers,
36513 libraries and debuggers on certain systems.
36515 * On many platforms, GCC supports a different ABI for C++ than do
36516 other compilers, so the object files compiled by GCC cannot be
36517 used with object files generated by another C++ compiler.
36519 An area where the difference is most apparent is name mangling.
36520 The use of different name mangling is intentional, to protect you
36521 from more subtle problems. Compilers differ as to many internal
36522 details of C++ implementation, including: how class instances are
36523 laid out, how multiple inheritance is implemented, and how virtual
36524 function calls are handled. If the name encoding were made the
36525 same, your programs would link against libraries provided from
36526 other compilers--but the programs would then crash when run.
36527 Incompatible libraries are then detected at link time, rather than
36530 * On some BSD systems, including some versions of Ultrix, use of
36531 profiling causes static variable destructors (currently used only
36532 in C++) not to be run.
36534 * On some SGI systems, when you use `-lgl_s' as an option, it gets
36535 translated magically to `-lgl_s -lX11_s -lc_s'. Naturally, this
36536 does not happen when you use GCC. You must specify all three
36537 options explicitly.
36539 * On a SPARC, GCC aligns all values of type `double' on an 8-byte
36540 boundary, and it expects every `double' to be so aligned. The Sun
36541 compiler usually gives `double' values 8-byte alignment, with one
36542 exception: function arguments of type `double' may not be aligned.
36544 As a result, if a function compiled with Sun CC takes the address
36545 of an argument of type `double' and passes this pointer of type
36546 `double *' to a function compiled with GCC, dereferencing the
36547 pointer may cause a fatal signal.
36549 One way to solve this problem is to compile your entire program
36550 with GCC. Another solution is to modify the function that is
36551 compiled with Sun CC to copy the argument into a local variable;
36552 local variables are always properly aligned. A third solution is
36553 to modify the function that uses the pointer to dereference it via
36554 the following function `access_double' instead of directly with
36558 access_double (double *unaligned_ptr)
36560 union d2i { double d; int i[2]; };
36562 union d2i *p = (union d2i *) unaligned_ptr;
36571 Storing into the pointer can be done likewise with the same union.
36573 * On Solaris, the `malloc' function in the `libmalloc.a' library may
36574 allocate memory that is only 4 byte aligned. Since GCC on the
36575 SPARC assumes that doubles are 8 byte aligned, this may result in a
36576 fatal signal if doubles are stored in memory allocated by the
36577 `libmalloc.a' library.
36579 The solution is to not use the `libmalloc.a' library. Use instead
36580 `malloc' and related functions from `libc.a'; they do not have
36583 * On the HP PA machine, ADB sometimes fails to work on functions
36584 compiled with GCC. Specifically, it fails to work on functions
36585 that use `alloca' or variable-size arrays. This is because GCC
36586 doesn't generate HP-UX unwind descriptors for such functions. It
36587 may even be impossible to generate them.
36589 * Debugging (`-g') is not supported on the HP PA machine, unless you
36590 use the preliminary GNU tools.
36592 * Taking the address of a label may generate errors from the HP-UX
36593 PA assembler. GAS for the PA does not have this problem.
36595 * Using floating point parameters for indirect calls to static
36596 functions will not work when using the HP assembler. There simply
36597 is no way for GCC to specify what registers hold arguments for
36598 static functions when using the HP assembler. GAS for the PA does
36599 not have this problem.
36601 * In extremely rare cases involving some very large functions you may
36602 receive errors from the HP linker complaining about an out of
36603 bounds unconditional branch offset. This used to occur more often
36604 in previous versions of GCC, but is now exceptionally rare. If
36605 you should run into it, you can work around by making your
36608 * GCC compiled code sometimes emits warnings from the HP-UX
36609 assembler of the form:
36611 (warning) Use of GR3 when
36612 frame >= 8192 may cause conflict.
36614 These warnings are harmless and can be safely ignored.
36616 * In extremely rare cases involving some very large functions you may
36617 receive errors from the AIX Assembler complaining about a
36618 displacement that is too large. If you should run into it, you
36619 can work around by making your function smaller.
36621 * The `libstdc++.a' library in GCC relies on the SVR4 dynamic linker
36622 semantics which merges global symbols between libraries and
36623 applications, especially necessary for C++ streams functionality.
36624 This is not the default behavior of AIX shared libraries and
36625 dynamic linking. `libstdc++.a' is built on AIX with
36626 "runtime-linking" enabled so that symbol merging can occur. To
36627 utilize this feature, the application linked with `libstdc++.a'
36628 must include the `-Wl,-brtl' flag on the link line. G++ cannot
36629 impose this because this option may interfere with the semantics
36630 of the user program and users may not always use `g++' to link his
36631 or her application. Applications are not required to use the
36632 `-Wl,-brtl' flag on the link line--the rest of the `libstdc++.a'
36633 library which is not dependent on the symbol merging semantics
36634 will continue to function correctly.
36636 * An application can interpose its own definition of functions for
36637 functions invoked by `libstdc++.a' with "runtime-linking" enabled
36638 on AIX. To accomplish this the application must be linked with
36639 "runtime-linking" option and the functions explicitly must be
36640 exported by the application (`-Wl,-brtl,-bE:exportfile').
36642 * AIX on the RS/6000 provides support (NLS) for environments outside
36643 of the United States. Compilers and assemblers use NLS to support
36644 locale-specific representations of various objects including
36645 floating-point numbers (`.' vs `,' for separating decimal
36646 fractions). There have been problems reported where the library
36647 linked with GCC does not produce the same floating-point formats
36648 that the assembler accepts. If you have this problem, set the
36649 `LANG' environment variable to `C' or `En_US'.
36651 * Even if you specify `-fdollars-in-identifiers', you cannot
36652 successfully use `$' in identifiers on the RS/6000 due to a
36653 restriction in the IBM assembler. GAS supports these identifiers.
36657 File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: Interoperation, Up: Trouble
36659 10.4 Incompatibilities of GCC
36660 =============================
36662 There are several noteworthy incompatibilities between GNU C and K&R
36663 (non-ISO) versions of C.
36665 * GCC normally makes string constants read-only. If several
36666 identical-looking string constants are used, GCC stores only one
36667 copy of the string.
36669 One consequence is that you cannot call `mktemp' with a string
36670 constant argument. The function `mktemp' always alters the string
36671 its argument points to.
36673 Another consequence is that `sscanf' does not work on some very
36674 old systems when passed a string constant as its format control
36675 string or input. This is because `sscanf' incorrectly tries to
36676 write into the string constant. Likewise `fscanf' and `scanf'.
36678 The solution to these problems is to change the program to use
36679 `char'-array variables with initialization strings for these
36680 purposes instead of string constants.
36682 * `-2147483648' is positive.
36684 This is because 2147483648 cannot fit in the type `int', so
36685 (following the ISO C rules) its data type is `unsigned long int'.
36686 Negating this value yields 2147483648 again.
36688 * GCC does not substitute macro arguments when they appear inside of
36689 string constants. For example, the following macro in GCC
36693 will produce output `"a"' regardless of what the argument A is.
36695 * When you use `setjmp' and `longjmp', the only automatic variables
36696 guaranteed to remain valid are those declared `volatile'. This is
36697 a consequence of automatic register allocation. Consider this
36711 /* `longjmp (j)' may occur in `fun3'. */
36712 return a + fun3 ();
36715 Here `a' may or may not be restored to its first value when the
36716 `longjmp' occurs. If `a' is allocated in a register, then its
36717 first value is restored; otherwise, it keeps the last value stored
36720 If you use the `-W' option with the `-O' option, you will get a
36721 warning when GCC thinks such a problem might be possible.
36723 * Programs that use preprocessing directives in the middle of macro
36724 arguments do not work with GCC. For example, a program like this
36731 ISO C does not permit such a construct.
36733 * K&R compilers allow comments to cross over an inclusion boundary
36734 (i.e. started in an include file and ended in the including file).
36736 * Declarations of external variables and functions within a block
36737 apply only to the block containing the declaration. In other
36738 words, they have the same scope as any other declaration in the
36741 In some other C compilers, a `extern' declaration affects all the
36742 rest of the file even if it happens within a block.
36744 * In traditional C, you can combine `long', etc., with a typedef
36745 name, as shown here:
36748 typedef long foo bar;
36750 In ISO C, this is not allowed: `long' and other type modifiers
36751 require an explicit `int'.
36753 * PCC allows typedef names to be used as function parameters.
36755 * Traditional C allows the following erroneous pair of declarations
36756 to appear together in a given scope:
36761 * GCC treats all characters of identifiers as significant.
36762 According to K&R-1 (2.2), "No more than the first eight characters
36763 are significant, although more may be used.". Also according to
36764 K&R-1 (2.2), "An identifier is a sequence of letters and digits;
36765 the first character must be a letter. The underscore _ counts as
36766 a letter.", but GCC also allows dollar signs in identifiers.
36768 * PCC allows whitespace in the middle of compound assignment
36769 operators such as `+='. GCC, following the ISO standard, does not
36772 * GCC complains about unterminated character constants inside of
36773 preprocessing conditionals that fail. Some programs have English
36774 comments enclosed in conditionals that are guaranteed to fail; if
36775 these comments contain apostrophes, GCC will probably report an
36776 error. For example, this code would produce an error:
36779 You can't expect this to work.
36782 The best solution to such a problem is to put the text into an
36783 actual C comment delimited by `/*...*/'.
36785 * Many user programs contain the declaration `long time ();'. In the
36786 past, the system header files on many systems did not actually
36787 declare `time', so it did not matter what type your program
36788 declared it to return. But in systems with ISO C headers, `time'
36789 is declared to return `time_t', and if that is not the same as
36790 `long', then `long time ();' is erroneous.
36792 The solution is to change your program to use appropriate system
36793 headers (`<time.h>' on systems with ISO C headers) and not to
36794 declare `time' if the system header files declare it, or failing
36795 that to use `time_t' as the return type of `time'.
36797 * When compiling functions that return `float', PCC converts it to a
36798 double. GCC actually returns a `float'. If you are concerned
36799 with PCC compatibility, you should declare your functions to return
36800 `double'; you might as well say what you mean.
36802 * When compiling functions that return structures or unions, GCC
36803 output code normally uses a method different from that used on most
36804 versions of Unix. As a result, code compiled with GCC cannot call
36805 a structure-returning function compiled with PCC, and vice versa.
36807 The method used by GCC is as follows: a structure or union which is
36808 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or
36809 union with any other size is stored into an address supplied by
36810 the caller (usually in a special, fixed register, but on some
36811 machines it is passed on the stack). The target hook
36812 `TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address.
36814 By contrast, PCC on most target machines returns structures and
36815 unions of any size by copying the data into an area of static
36816 storage, and then returning the address of that storage as if it
36817 were a pointer value. The caller must copy the data from that
36818 memory area to the place where the value is wanted. GCC does not
36819 use this method because it is slower and nonreentrant.
36821 On some newer machines, PCC uses a reentrant convention for all
36822 structure and union returning. GCC on most of these machines uses
36823 a compatible convention when returning structures and unions in
36824 memory, but still returns small structures and unions in registers.
36826 You can tell GCC to use a compatible convention for all structure
36827 and union returning with the option `-fpcc-struct-return'.
36829 * GCC complains about program fragments such as `0x74ae-0x4000'
36830 which appear to be two hexadecimal constants separated by the minus
36831 operator. Actually, this string is a single "preprocessing token".
36832 Each such token must correspond to one token in C. Since this
36833 does not, GCC prints an error message. Although it may appear
36834 obvious that what is meant is an operator and two values, the ISO
36835 C standard specifically requires that this be treated as erroneous.
36837 A "preprocessing token" is a "preprocessing number" if it begins
36838 with a digit and is followed by letters, underscores, digits,
36839 periods and `e+', `e-', `E+', `E-', `p+', `p-', `P+', or `P-'
36840 character sequences. (In strict C89 mode, the sequences `p+',
36841 `p-', `P+' and `P-' cannot appear in preprocessing numbers.)
36843 To make the above program fragment valid, place whitespace in
36844 front of the minus sign. This whitespace will end the
36845 preprocessing number.
36848 File: gcc.info, Node: Fixed Headers, Next: Standard Libraries, Prev: Incompatibilities, Up: Trouble
36850 10.5 Fixed Header Files
36851 =======================
36853 GCC needs to install corrected versions of some system header files.
36854 This is because most target systems have some header files that won't
36855 work with GCC unless they are changed. Some have bugs, some are
36856 incompatible with ISO C, and some depend on special features of other
36859 Installing GCC automatically creates and installs the fixed header
36860 files, by running a program called `fixincludes'. Normally, you don't
36861 need to pay attention to this. But there are cases where it doesn't do
36862 the right thing automatically.
36864 * If you update the system's header files, such as by installing a
36865 new system version, the fixed header files of GCC are not
36866 automatically updated. They can be updated using the `mkheaders'
36867 script installed in `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
36869 * On some systems, header file directories contain machine-specific
36870 symbolic links in certain places. This makes it possible to share
36871 most of the header files among hosts running the same version of
36872 the system on different machine models.
36874 The programs that fix the header files do not understand this
36875 special way of using symbolic links; therefore, the directory of
36876 fixed header files is good only for the machine model used to
36879 It is possible to make separate sets of fixed header files for the
36880 different machine models, and arrange a structure of symbolic
36881 links so as to use the proper set, but you'll have to do this by
36885 File: gcc.info, Node: Standard Libraries, Next: Disappointments, Prev: Fixed Headers, Up: Trouble
36887 10.6 Standard Libraries
36888 =======================
36890 GCC by itself attempts to be a conforming freestanding implementation.
36891 *Note Language Standards Supported by GCC: Standards, for details of
36892 what this means. Beyond the library facilities required of such an
36893 implementation, the rest of the C library is supplied by the vendor of
36894 the operating system. If that C library doesn't conform to the C
36895 standards, then your programs might get warnings (especially when using
36896 `-Wall') that you don't expect.
36898 For example, the `sprintf' function on SunOS 4.1.3 returns `char *'
36899 while the C standard says that `sprintf' returns an `int'. The
36900 `fixincludes' program could make the prototype for this function match
36901 the Standard, but that would be wrong, since the function will still
36904 If you need a Standard compliant library, then you need to find one, as
36905 GCC does not provide one. The GNU C library (called `glibc') provides
36906 ISO C, POSIX, BSD, SystemV and X/Open compatibility for GNU/Linux and
36907 HURD-based GNU systems; no recent version of it supports other systems,
36908 though some very old versions did. Version 2.2 of the GNU C library
36909 includes nearly complete C99 support. You could also ask your
36910 operating system vendor if newer libraries are available.
36913 File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Standard Libraries, Up: Trouble
36915 10.7 Disappointments and Misunderstandings
36916 ==========================================
36918 These problems are perhaps regrettable, but we don't know any practical
36921 * Certain local variables aren't recognized by debuggers when you
36922 compile with optimization.
36924 This occurs because sometimes GCC optimizes the variable out of
36925 existence. There is no way to tell the debugger how to compute the
36926 value such a variable "would have had", and it is not clear that
36927 would be desirable anyway. So GCC simply does not mention the
36928 eliminated variable when it writes debugging information.
36930 You have to expect a certain amount of disagreement between the
36931 executable and your source code, when you use optimization.
36933 * Users often think it is a bug when GCC reports an error for code
36936 int foo (struct mumble *);
36938 struct mumble { ... };
36940 int foo (struct mumble *x)
36943 This code really is erroneous, because the scope of `struct
36944 mumble' in the prototype is limited to the argument list
36945 containing it. It does not refer to the `struct mumble' defined
36946 with file scope immediately below--they are two unrelated types
36947 with similar names in different scopes.
36949 But in the definition of `foo', the file-scope type is used
36950 because that is available to be inherited. Thus, the definition
36951 and the prototype do not match, and you get an error.
36953 This behavior may seem silly, but it's what the ISO standard
36954 specifies. It is easy enough for you to make your code work by
36955 moving the definition of `struct mumble' above the prototype.
36956 It's not worth being incompatible with ISO C just to avoid an
36957 error for the example shown above.
36959 * Accesses to bit-fields even in volatile objects works by accessing
36960 larger objects, such as a byte or a word. You cannot rely on what
36961 size of object is accessed in order to read or write the
36962 bit-field; it may even vary for a given bit-field according to the
36965 If you care about controlling the amount of memory that is
36966 accessed, use volatile but do not use bit-fields.
36968 * GCC comes with shell scripts to fix certain known problems in
36969 system header files. They install corrected copies of various
36970 header files in a special directory where only GCC will normally
36971 look for them. The scripts adapt to various systems by searching
36972 all the system header files for the problem cases that we know
36975 If new system header files are installed, nothing automatically
36976 arranges to update the corrected header files. They can be
36977 updated using the `mkheaders' script installed in
36978 `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
36980 * On 68000 and x86 systems, for instance, you can get paradoxical
36981 results if you test the precise values of floating point numbers.
36982 For example, you can find that a floating point value which is not
36983 a NaN is not equal to itself. This results from the fact that the
36984 floating point registers hold a few more bits of precision than
36985 fit in a `double' in memory. Compiled code moves values between
36986 memory and floating point registers at its convenience, and moving
36987 them into memory truncates them.
36989 You can partially avoid this problem by using the `-ffloat-store'
36990 option (*note Optimize Options::).
36992 * On AIX and other platforms without weak symbol support, templates
36993 need to be instantiated explicitly and symbols for static members
36994 of templates will not be generated.
36996 * On AIX, GCC scans object files and library archives for static
36997 constructors and destructors when linking an application before the
36998 linker prunes unreferenced symbols. This is necessary to prevent
36999 the AIX linker from mistakenly assuming that static constructor or
37000 destructor are unused and removing them before the scanning can
37001 occur. All static constructors and destructors found will be
37002 referenced even though the modules in which they occur may not be
37003 used by the program. This may lead to both increased executable
37004 size and unexpected symbol references.
37007 File: gcc.info, Node: C++ Misunderstandings, Next: Protoize Caveats, Prev: Disappointments, Up: Trouble
37009 10.8 Common Misunderstandings with GNU C++
37010 ==========================================
37012 C++ is a complex language and an evolving one, and its standard
37013 definition (the ISO C++ standard) was only recently completed. As a
37014 result, your C++ compiler may occasionally surprise you, even when its
37015 behavior is correct. This section discusses some areas that frequently
37016 give rise to questions of this sort.
37020 * Static Definitions:: Static member declarations are not definitions
37021 * Name lookup:: Name lookup, templates, and accessing members of base classes
37022 * Temporaries:: Temporaries may vanish before you expect
37023 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
37026 File: gcc.info, Node: Static Definitions, Next: Name lookup, Up: C++ Misunderstandings
37028 10.8.1 Declare _and_ Define Static Members
37029 ------------------------------------------
37031 When a class has static data members, it is not enough to _declare_ the
37032 static member; you must also _define_ it. For example:
37041 This declaration only establishes that the class `Foo' has an `int'
37042 named `Foo::bar', and a member function named `Foo::method'. But you
37043 still need to define _both_ `method' and `bar' elsewhere. According to
37044 the ISO standard, you must supply an initializer in one (and only one)
37045 source file, such as:
37049 Other C++ compilers may not correctly implement the standard behavior.
37050 As a result, when you switch to `g++' from one of these compilers, you
37051 may discover that a program that appeared to work correctly in fact
37052 does not conform to the standard: `g++' reports as undefined symbols
37053 any static data members that lack definitions.
37056 File: gcc.info, Node: Name lookup, Next: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings
37058 10.8.2 Name lookup, templates, and accessing members of base classes
37059 --------------------------------------------------------------------
37061 The C++ standard prescribes that all names that are not dependent on
37062 template parameters are bound to their present definitions when parsing
37063 a template function or class.(1) Only names that are dependent are
37064 looked up at the point of instantiation. For example, consider
37069 template <typename T>
37078 static const int N;
37081 Here, the names `foo' and `N' appear in a context that does not depend
37082 on the type of `T'. The compiler will thus require that they are
37083 defined in the context of use in the template, not only before the
37084 point of instantiation, and will here use `::foo(double)' and `A::N',
37085 respectively. In particular, it will convert the integer value to a
37086 `double' when passing it to `::foo(double)'.
37088 Conversely, `bar' and the call to `foo' in the fourth marked line are
37089 used in contexts that do depend on the type of `T', so they are only
37090 looked up at the point of instantiation, and you can provide
37091 declarations for them after declaring the template, but before
37092 instantiating it. In particular, if you instantiate `A::f<int>', the
37093 last line will call an overloaded `::foo(int)' if one was provided,
37094 even if after the declaration of `struct A'.
37096 This distinction between lookup of dependent and non-dependent names is
37097 called two-stage (or dependent) name lookup. G++ implements it since
37100 Two-stage name lookup sometimes leads to situations with behavior
37101 different from non-template codes. The most common is probably this:
37103 template <typename T> struct Base {
37107 template <typename T> struct Derived : public Base<T> {
37108 int get_i() { return i; }
37111 In `get_i()', `i' is not used in a dependent context, so the compiler
37112 will look for a name declared at the enclosing namespace scope (which
37113 is the global scope here). It will not look into the base class, since
37114 that is dependent and you may declare specializations of `Base' even
37115 after declaring `Derived', so the compiler can't really know what `i'
37116 would refer to. If there is no global variable `i', then you will get
37119 In order to make it clear that you want the member of the base class,
37120 you need to defer lookup until instantiation time, at which the base
37121 class is known. For this, you need to access `i' in a dependent
37122 context, by either using `this->i' (remember that `this' is of type
37123 `Derived<T>*', so is obviously dependent), or using `Base<T>::i'.
37124 Alternatively, `Base<T>::i' might be brought into scope by a
37125 `using'-declaration.
37127 Another, similar example involves calling member functions of a base
37130 template <typename T> struct Base {
37134 template <typename T> struct Derived : Base<T> {
37135 int g() { return f(); };
37138 Again, the call to `f()' is not dependent on template arguments (there
37139 are no arguments that depend on the type `T', and it is also not
37140 otherwise specified that the call should be in a dependent context).
37141 Thus a global declaration of such a function must be available, since
37142 the one in the base class is not visible until instantiation time. The
37143 compiler will consequently produce the following error message:
37145 x.cc: In member function `int Derived<T>::g()':
37146 x.cc:6: error: there are no arguments to `f' that depend on a template
37147 parameter, so a declaration of `f' must be available
37148 x.cc:6: error: (if you use `-fpermissive', G++ will accept your code, but
37149 allowing the use of an undeclared name is deprecated)
37151 To make the code valid either use `this->f()', or `Base<T>::f()'.
37152 Using the `-fpermissive' flag will also let the compiler accept the
37153 code, by marking all function calls for which no declaration is visible
37154 at the time of definition of the template for later lookup at
37155 instantiation time, as if it were a dependent call. We do not
37156 recommend using `-fpermissive' to work around invalid code, and it will
37157 also only catch cases where functions in base classes are called, not
37158 where variables in base classes are used (as in the example above).
37160 Note that some compilers (including G++ versions prior to 3.4) get
37161 these examples wrong and accept above code without an error. Those
37162 compilers do not implement two-stage name lookup correctly.
37164 ---------- Footnotes ----------
37166 (1) The C++ standard just uses the term "dependent" for names that
37167 depend on the type or value of template parameters. This shorter term
37168 will also be used in the rest of this section.
37171 File: gcc.info, Node: Temporaries, Next: Copy Assignment, Prev: Name lookup, Up: C++ Misunderstandings
37173 10.8.3 Temporaries May Vanish Before You Expect
37174 -----------------------------------------------
37176 It is dangerous to use pointers or references to _portions_ of a
37177 temporary object. The compiler may very well delete the object before
37178 you expect it to, leaving a pointer to garbage. The most common place
37179 where this problem crops up is in classes like string classes,
37180 especially ones that define a conversion function to type `char *' or
37181 `const char *'--which is one reason why the standard `string' class
37182 requires you to call the `c_str' member function. However, any class
37183 that returns a pointer to some internal structure is potentially
37184 subject to this problem.
37186 For example, a program may use a function `strfunc' that returns
37187 `string' objects, and another function `charfunc' that operates on
37188 pointers to `char':
37191 void charfunc (const char *);
37196 const char *p = strfunc().c_str();
37203 In this situation, it may seem reasonable to save a pointer to the C
37204 string returned by the `c_str' member function and use that rather than
37205 call `c_str' repeatedly. However, the temporary string created by the
37206 call to `strfunc' is destroyed after `p' is initialized, at which point
37207 `p' is left pointing to freed memory.
37209 Code like this may run successfully under some other compilers,
37210 particularly obsolete cfront-based compilers that delete temporaries
37211 along with normal local variables. However, the GNU C++ behavior is
37212 standard-conforming, so if your program depends on late destruction of
37213 temporaries it is not portable.
37215 The safe way to write such code is to give the temporary a name, which
37216 forces it to remain until the end of the scope of the name. For
37219 const string& tmp = strfunc ();
37220 charfunc (tmp.c_str ());
37223 File: gcc.info, Node: Copy Assignment, Prev: Temporaries, Up: C++ Misunderstandings
37225 10.8.4 Implicit Copy-Assignment for Virtual Bases
37226 -------------------------------------------------
37228 When a base class is virtual, only one subobject of the base class
37229 belongs to each full object. Also, the constructors and destructors are
37230 invoked only once, and called from the most-derived class. However,
37231 such objects behave unspecified when being assigned. For example:
37235 Base(char *n) : name(strdup(n)){}
37236 Base& operator= (const Base& other){
37238 name = strdup (other.name);
37242 struct A:virtual Base{
37247 struct B:virtual Base{
37252 struct Derived:public A, public B{
37253 Derived():Base("Derived"){}
37256 void func(Derived &d1, Derived &d2)
37261 The C++ standard specifies that `Base::Base' is only called once when
37262 constructing or copy-constructing a Derived object. It is unspecified
37263 whether `Base::operator=' is called more than once when the implicit
37264 copy-assignment for Derived objects is invoked (as it is inside `func'
37267 G++ implements the "intuitive" algorithm for copy-assignment: assign
37268 all direct bases, then assign all members. In that algorithm, the
37269 virtual base subobject can be encountered more than once. In the
37270 example, copying proceeds in the following order: `val', `name' (via
37271 `strdup'), `bval', and `name' again.
37273 If application code relies on copy-assignment, a user-defined
37274 copy-assignment operator removes any uncertainties. With such an
37275 operator, the application can define whether and how the virtual base
37276 subobject is assigned.
37279 File: gcc.info, Node: Protoize Caveats, Next: Non-bugs, Prev: C++ Misunderstandings, Up: Trouble
37281 10.9 Caveats of using `protoize'
37282 ================================
37284 The conversion programs `protoize' and `unprotoize' can sometimes
37285 change a source file in a way that won't work unless you rearrange it.
37287 * `protoize' can insert references to a type name or type tag before
37288 the definition, or in a file where they are not defined.
37290 If this happens, compiler error messages should show you where the
37291 new references are, so fixing the file by hand is straightforward.
37293 * There are some C constructs which `protoize' cannot figure out.
37294 For example, it can't determine argument types for declaring a
37295 pointer-to-function variable; this you must do by hand. `protoize'
37296 inserts a comment containing `???' each time it finds such a
37297 variable; so you can find all such variables by searching for this
37298 string. ISO C does not require declaring the argument types of
37299 pointer-to-function types.
37301 * Using `unprotoize' can easily introduce bugs. If the program
37302 relied on prototypes to bring about conversion of arguments, these
37303 conversions will not take place in the program without prototypes.
37304 One case in which you can be sure `unprotoize' is safe is when you
37305 are removing prototypes that were made with `protoize'; if the
37306 program worked before without any prototypes, it will work again
37309 You can find all the places where this problem might occur by
37310 compiling the program with the `-Wtraditional-conversion' option.
37311 It prints a warning whenever an argument is converted.
37313 * Both conversion programs can be confused if there are macro calls
37314 in and around the text to be converted. In other words, the
37315 standard syntax for a declaration or definition must not result
37316 from expanding a macro. This problem is inherent in the design of
37317 C and cannot be fixed. If only a few functions have confusing
37318 macro calls, you can easily convert them manually.
37320 * `protoize' cannot get the argument types for a function whose
37321 definition was not actually compiled due to preprocessing
37322 conditionals. When this happens, `protoize' changes nothing in
37323 regard to such a function. `protoize' tries to detect such
37324 instances and warn about them.
37326 You can generally work around this problem by using `protoize' step
37327 by step, each time specifying a different set of `-D' options for
37328 compilation, until all of the functions have been converted.
37329 There is no automatic way to verify that you have got them all,
37332 * Confusion may result if there is an occasion to convert a function
37333 declaration or definition in a region of source code where there
37334 is more than one formal parameter list present. Thus, attempts to
37335 convert code containing multiple (conditionally compiled) versions
37336 of a single function header (in the same vicinity) may not produce
37337 the desired (or expected) results.
37339 If you plan on converting source files which contain such code, it
37340 is recommended that you first make sure that each conditionally
37341 compiled region of source code which contains an alternative
37342 function header also contains at least one additional follower
37343 token (past the final right parenthesis of the function header).
37344 This should circumvent the problem.
37346 * `unprotoize' can become confused when trying to convert a function
37347 definition or declaration which contains a declaration for a
37348 pointer-to-function formal argument which has the same name as the
37349 function being defined or declared. We recommend you avoid such
37350 choices of formal parameter names.
37352 * You might also want to correct some of the indentation by hand and
37353 break long lines. (The conversion programs don't write lines
37354 longer than eighty characters in any case.)
37357 File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: Protoize Caveats, Up: Trouble
37359 10.10 Certain Changes We Don't Want to Make
37360 ===========================================
37362 This section lists changes that people frequently request, but which we
37363 do not make because we think GCC is better without them.
37365 * Checking the number and type of arguments to a function which has
37366 an old-fashioned definition and no prototype.
37368 Such a feature would work only occasionally--only for calls that
37369 appear in the same file as the called function, following the
37370 definition. The only way to check all calls reliably is to add a
37371 prototype for the function. But adding a prototype eliminates the
37372 motivation for this feature. So the feature is not worthwhile.
37374 * Warning about using an expression whose type is signed as a shift
37377 Shift count operands are probably signed more often than unsigned.
37378 Warning about this would cause far more annoyance than good.
37380 * Warning about assigning a signed value to an unsigned variable.
37382 Such assignments must be very common; warning about them would
37383 cause more annoyance than good.
37385 * Warning when a non-void function value is ignored.
37387 C contains many standard functions that return a value that most
37388 programs choose to ignore. One obvious example is `printf'.
37389 Warning about this practice only leads the defensive programmer to
37390 clutter programs with dozens of casts to `void'. Such casts are
37391 required so frequently that they become visual noise. Writing
37392 those casts becomes so automatic that they no longer convey useful
37393 information about the intentions of the programmer. For functions
37394 where the return value should never be ignored, use the
37395 `warn_unused_result' function attribute (*note Function
37398 * Making `-fshort-enums' the default.
37400 This would cause storage layout to be incompatible with most other
37401 C compilers. And it doesn't seem very important, given that you
37402 can get the same result in other ways. The case where it matters
37403 most is when the enumeration-valued object is inside a structure,
37404 and in that case you can specify a field width explicitly.
37406 * Making bit-fields unsigned by default on particular machines where
37407 "the ABI standard" says to do so.
37409 The ISO C standard leaves it up to the implementation whether a
37410 bit-field declared plain `int' is signed or not. This in effect
37411 creates two alternative dialects of C.
37413 The GNU C compiler supports both dialects; you can specify the
37414 signed dialect with `-fsigned-bitfields' and the unsigned dialect
37415 with `-funsigned-bitfields'. However, this leaves open the
37416 question of which dialect to use by default.
37418 Currently, the preferred dialect makes plain bit-fields signed,
37419 because this is simplest. Since `int' is the same as `signed int'
37420 in every other context, it is cleanest for them to be the same in
37421 bit-fields as well.
37423 Some computer manufacturers have published Application Binary
37424 Interface standards which specify that plain bit-fields should be
37425 unsigned. It is a mistake, however, to say anything about this
37426 issue in an ABI. This is because the handling of plain bit-fields
37427 distinguishes two dialects of C. Both dialects are meaningful on
37428 every type of machine. Whether a particular object file was
37429 compiled using signed bit-fields or unsigned is of no concern to
37430 other object files, even if they access the same bit-fields in the
37431 same data structures.
37433 A given program is written in one or the other of these two
37434 dialects. The program stands a chance to work on most any machine
37435 if it is compiled with the proper dialect. It is unlikely to work
37436 at all if compiled with the wrong dialect.
37438 Many users appreciate the GNU C compiler because it provides an
37439 environment that is uniform across machines. These users would be
37440 inconvenienced if the compiler treated plain bit-fields
37441 differently on certain machines.
37443 Occasionally users write programs intended only for a particular
37444 machine type. On these occasions, the users would benefit if the
37445 GNU C compiler were to support by default the same dialect as the
37446 other compilers on that machine. But such applications are rare.
37447 And users writing a program to run on more than one type of
37448 machine cannot possibly benefit from this kind of compatibility.
37450 This is why GCC does and will treat plain bit-fields in the same
37451 fashion on all types of machines (by default).
37453 There are some arguments for making bit-fields unsigned by default
37454 on all machines. If, for example, this becomes a universal de
37455 facto standard, it would make sense for GCC to go along with it.
37456 This is something to be considered in the future.
37458 (Of course, users strongly concerned about portability should
37459 indicate explicitly in each bit-field whether it is signed or not.
37460 In this way, they write programs which have the same meaning in
37463 * Undefining `__STDC__' when `-ansi' is not used.
37465 Currently, GCC defines `__STDC__' unconditionally. This provides
37466 good results in practice.
37468 Programmers normally use conditionals on `__STDC__' to ask whether
37469 it is safe to use certain features of ISO C, such as function
37470 prototypes or ISO token concatenation. Since plain `gcc' supports
37471 all the features of ISO C, the correct answer to these questions is
37474 Some users try to use `__STDC__' to check for the availability of
37475 certain library facilities. This is actually incorrect usage in
37476 an ISO C program, because the ISO C standard says that a conforming
37477 freestanding implementation should define `__STDC__' even though it
37478 does not have the library facilities. `gcc -ansi -pedantic' is a
37479 conforming freestanding implementation, and it is therefore
37480 required to define `__STDC__', even though it does not come with
37483 Sometimes people say that defining `__STDC__' in a compiler that
37484 does not completely conform to the ISO C standard somehow violates
37485 the standard. This is illogical. The standard is a standard for
37486 compilers that claim to support ISO C, such as `gcc -ansi'--not
37487 for other compilers such as plain `gcc'. Whatever the ISO C
37488 standard says is relevant to the design of plain `gcc' without
37489 `-ansi' only for pragmatic reasons, not as a requirement.
37491 GCC normally defines `__STDC__' to be 1, and in addition defines
37492 `__STRICT_ANSI__' if you specify the `-ansi' option, or a `-std'
37493 option for strict conformance to some version of ISO C. On some
37494 hosts, system include files use a different convention, where
37495 `__STDC__' is normally 0, but is 1 if the user specifies strict
37496 conformance to the C Standard. GCC follows the host convention
37497 when processing system include files, but when processing user
37498 files it follows the usual GNU C convention.
37500 * Undefining `__STDC__' in C++.
37502 Programs written to compile with C++-to-C translators get the
37503 value of `__STDC__' that goes with the C compiler that is
37504 subsequently used. These programs must test `__STDC__' to
37505 determine what kind of C preprocessor that compiler uses: whether
37506 they should concatenate tokens in the ISO C fashion or in the
37507 traditional fashion.
37509 These programs work properly with GNU C++ if `__STDC__' is defined.
37510 They would not work otherwise.
37512 In addition, many header files are written to provide prototypes
37513 in ISO C but not in traditional C. Many of these header files can
37514 work without change in C++ provided `__STDC__' is defined. If
37515 `__STDC__' is not defined, they will all fail, and will all need
37516 to be changed to test explicitly for C++ as well.
37518 * Deleting "empty" loops.
37520 Historically, GCC has not deleted "empty" loops under the
37521 assumption that the most likely reason you would put one in a
37522 program is to have a delay, so deleting them will not make real
37523 programs run any faster.
37525 However, the rationale here is that optimization of a nonempty loop
37526 cannot produce an empty one. This held for carefully written C
37527 compiled with less powerful optimizers but is not always the case
37528 for carefully written C++ or with more powerful optimizers. Thus
37529 GCC will remove operations from loops whenever it can determine
37530 those operations are not externally visible (apart from the time
37531 taken to execute them, of course). In case the loop can be proved
37532 to be finite, GCC will also remove the loop itself.
37534 Be aware of this when performing timing tests, for instance the
37535 following loop can be completely removed, provided
37536 `some_expression' can provably not change any global state.
37542 for (ix = 0; ix != 10000; ix++)
37543 sum += some_expression;
37546 Even though `sum' is accumulated in the loop, no use is made of
37547 that summation, so the accumulation can be removed.
37549 * Making side effects happen in the same order as in some other
37552 It is never safe to depend on the order of evaluation of side
37553 effects. For example, a function call like this may very well
37554 behave differently from one compiler to another:
37556 void func (int, int);
37561 There is no guarantee (in either the C or the C++ standard language
37562 definitions) that the increments will be evaluated in any
37563 particular order. Either increment might happen first. `func'
37564 might get the arguments `2, 3', or it might get `3, 2', or even
37567 * Making certain warnings into errors by default.
37569 Some ISO C testsuites report failure when the compiler does not
37570 produce an error message for a certain program.
37572 ISO C requires a "diagnostic" message for certain kinds of invalid
37573 programs, but a warning is defined by GCC to count as a
37574 diagnostic. If GCC produces a warning but not an error, that is
37575 correct ISO C support. If testsuites call this "failure", they
37576 should be run with the GCC option `-pedantic-errors', which will
37577 turn these warnings into errors.
37581 File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble
37583 10.11 Warning Messages and Error Messages
37584 =========================================
37586 The GNU compiler can produce two kinds of diagnostics: errors and
37587 warnings. Each kind has a different purpose:
37589 "Errors" report problems that make it impossible to compile your
37590 program. GCC reports errors with the source file name and line
37591 number where the problem is apparent.
37593 "Warnings" report other unusual conditions in your code that _may_
37594 indicate a problem, although compilation can (and does) proceed.
37595 Warning messages also report the source file name and line number,
37596 but include the text `warning:' to distinguish them from error
37599 Warnings may indicate danger points where you should check to make sure
37600 that your program really does what you intend; or the use of obsolete
37601 features; or the use of nonstandard features of GNU C or C++. Many
37602 warnings are issued only if you ask for them, with one of the `-W'
37603 options (for instance, `-Wall' requests a variety of useful warnings).
37605 GCC always tries to compile your program if possible; it never
37606 gratuitously rejects a program whose meaning is clear merely because
37607 (for instance) it fails to conform to a standard. In some cases,
37608 however, the C and C++ standards specify that certain extensions are
37609 forbidden, and a diagnostic _must_ be issued by a conforming compiler.
37610 The `-pedantic' option tells GCC to issue warnings in such cases;
37611 `-pedantic-errors' says to make them errors instead. This does not
37612 mean that _all_ non-ISO constructs get warnings or errors.
37614 *Note Options to Request or Suppress Warnings: Warning Options, for
37615 more detail on these and related command-line options.
37618 File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top
37623 Your bug reports play an essential role in making GCC reliable.
37625 When you encounter a problem, the first thing to do is to see if it is
37626 already known. *Note Trouble::. If it isn't known, then you should
37627 report the problem.
37631 * Criteria: Bug Criteria. Have you really found a bug?
37632 * Reporting: Bug Reporting. How to report a bug effectively.
37633 * Known: Trouble. Known problems.
37634 * Help: Service. Where to ask for help.
37637 File: gcc.info, Node: Bug Criteria, Next: Bug Reporting, Up: Bugs
37639 11.1 Have You Found a Bug?
37640 ==========================
37642 If you are not sure whether you have found a bug, here are some
37645 * If the compiler gets a fatal signal, for any input whatever, that
37646 is a compiler bug. Reliable compilers never crash.
37648 * If the compiler produces invalid assembly code, for any input
37649 whatever (except an `asm' statement), that is a compiler bug,
37650 unless the compiler reports errors (not just warnings) which would
37651 ordinarily prevent the assembler from being run.
37653 * If the compiler produces valid assembly code that does not
37654 correctly execute the input source code, that is a compiler bug.
37656 However, you must double-check to make sure, because you may have a
37657 program whose behavior is undefined, which happened by chance to
37658 give the desired results with another C or C++ compiler.
37660 For example, in many nonoptimizing compilers, you can write `x;'
37661 at the end of a function instead of `return x;', with the same
37662 results. But the value of the function is undefined if `return'
37663 is omitted; it is not a bug when GCC produces different results.
37665 Problems often result from expressions with two increment
37666 operators, as in `f (*p++, *p++)'. Your previous compiler might
37667 have interpreted that expression the way you intended; GCC might
37668 interpret it another way. Neither compiler is wrong. The bug is
37671 After you have localized the error to a single source line, it
37672 should be easy to check for these things. If your program is
37673 correct and well defined, you have found a compiler bug.
37675 * If the compiler produces an error message for valid input, that is
37678 * If the compiler does not produce an error message for invalid
37679 input, that is a compiler bug. However, you should note that your
37680 idea of "invalid input" might be someone else's idea of "an
37681 extension" or "support for traditional practice".
37683 * If you are an experienced user of one of the languages GCC
37684 supports, your suggestions for improvement of GCC are welcome in
37688 File: gcc.info, Node: Bug Reporting, Prev: Bug Criteria, Up: Bugs
37690 11.2 How and where to Report Bugs
37691 =================================
37693 Bugs should be reported to the bug database at
37694 `http://gcc.gnu.org/bugs.html'.
37697 File: gcc.info, Node: Service, Next: Contributing, Prev: Bugs, Up: Top
37699 12 How To Get Help with GCC
37700 ***************************
37702 If you need help installing, using or changing GCC, there are two ways
37705 * Send a message to a suitable network mailing list. First try
37706 <gcc-help@gcc.gnu.org> (for help installing or using GCC), and if
37707 that brings no response, try <gcc@gcc.gnu.org>. For help changing
37708 GCC, ask <gcc@gcc.gnu.org>. If you think you have found a bug in
37709 GCC, please report it following the instructions at *note Bug
37712 * Look in the service directory for someone who might help you for a
37713 fee. The service directory is found at
37714 `http://www.gnu.org/prep/service.html'.
37716 For further information, see `http://gcc.gnu.org/faq.html#support'.
37719 File: gcc.info, Node: Contributing, Next: Funding, Prev: Service, Up: Top
37721 13 Contributing to GCC Development
37722 **********************************
37724 If you would like to help pretest GCC releases to assure they work well,
37725 current development sources are available by SVN (see
37726 `http://gcc.gnu.org/svn.html'). Source and binary snapshots are also
37727 available for FTP; see `http://gcc.gnu.org/snapshots.html'.
37729 If you would like to work on improvements to GCC, please read the
37730 advice at these URLs:
37732 `http://gcc.gnu.org/contribute.html'
37733 `http://gcc.gnu.org/contributewhy.html'
37735 for information on how to make useful contributions and avoid
37736 duplication of effort. Suggested projects are listed at
37737 `http://gcc.gnu.org/projects/'.
37740 File: gcc.info, Node: Funding, Next: GNU Project, Prev: Contributing, Up: Top
37742 Funding Free Software
37743 *********************
37745 If you want to have more free software a few years from now, it makes
37746 sense for you to help encourage people to contribute funds for its
37747 development. The most effective approach known is to encourage
37748 commercial redistributors to donate.
37750 Users of free software systems can boost the pace of development by
37751 encouraging for-a-fee distributors to donate part of their selling price
37752 to free software developers--the Free Software Foundation, and others.
37754 The way to convince distributors to do this is to demand it and expect
37755 it from them. So when you compare distributors, judge them partly by
37756 how much they give to free software development. Show distributors
37757 they must compete to be the one who gives the most.
37759 To make this approach work, you must insist on numbers that you can
37760 compare, such as, "We will donate ten dollars to the Frobnitz project
37761 for each disk sold." Don't be satisfied with a vague promise, such as
37762 "A portion of the profits are donated," since it doesn't give a basis
37765 Even a precise fraction "of the profits from this disk" is not very
37766 meaningful, since creative accounting and unrelated business decisions
37767 can greatly alter what fraction of the sales price counts as profit.
37768 If the price you pay is $50, ten percent of the profit is probably less
37769 than a dollar; it might be a few cents, or nothing at all.
37771 Some redistributors do development work themselves. This is useful
37772 too; but to keep everyone honest, you need to inquire how much they do,
37773 and what kind. Some kinds of development make much more long-term
37774 difference than others. For example, maintaining a separate version of
37775 a program contributes very little; maintaining the standard version of a
37776 program for the whole community contributes much. Easy new ports
37777 contribute little, since someone else would surely do them; difficult
37778 ports such as adding a new CPU to the GNU Compiler Collection
37779 contribute more; major new features or packages contribute the most.
37781 By establishing the idea that supporting further development is "the
37782 proper thing to do" when distributing free software for a fee, we can
37783 assure a steady flow of resources into making more free software.
37785 Copyright (C) 1994 Free Software Foundation, Inc.
37786 Verbatim copying and redistribution of this section is permitted
37787 without royalty; alteration is not permitted.
37790 File: gcc.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top
37792 The GNU Project and GNU/Linux
37793 *****************************
37795 The GNU Project was launched in 1984 to develop a complete Unix-like
37796 operating system which is free software: the GNU system. (GNU is a
37797 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".)
37798 Variants of the GNU operating system, which use the kernel Linux, are
37799 now widely used; though these systems are often referred to as "Linux",
37800 they are more accurately called GNU/Linux systems.
37802 For more information, see:
37803 `http://www.gnu.org/'
37804 `http://www.gnu.org/gnu/linux-and-gnu.html'
37807 File: gcc.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top
37809 GNU General Public License
37810 **************************
37812 Version 3, 29 June 2007
37814 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/'
37816 Everyone is permitted to copy and distribute verbatim copies of this
37817 license document, but changing it is not allowed.
37822 The GNU General Public License is a free, copyleft license for software
37823 and other kinds of works.
37825 The licenses for most software and other practical works are designed
37826 to take away your freedom to share and change the works. By contrast,
37827 the GNU General Public License is intended to guarantee your freedom to
37828 share and change all versions of a program-to make sure it remains free
37829 software for all its users. We, the Free Software Foundation, use the
37830 GNU General Public License for most of our software; it applies also to
37831 any other work released this way by its authors. You can apply it to
37832 your programs, too.
37834 When we speak of free software, we are referring to freedom, not
37835 price. Our General Public Licenses are designed to make sure that you
37836 have the freedom to distribute copies of free software (and charge for
37837 them if you wish), that you receive source code or can get it if you
37838 want it, that you can change the software or use pieces of it in new
37839 free programs, and that you know you can do these things.
37841 To protect your rights, we need to prevent others from denying you
37842 these rights or asking you to surrender the rights. Therefore, you
37843 have certain responsibilities if you distribute copies of the software,
37844 or if you modify it: responsibilities to respect the freedom of others.
37846 For example, if you distribute copies of such a program, whether
37847 gratis or for a fee, you must pass on to the recipients the same
37848 freedoms that you received. You must make sure that they, too, receive
37849 or can get the source code. And you must show them these terms so they
37852 Developers that use the GNU GPL protect your rights with two steps:
37853 (1) assert copyright on the software, and (2) offer you this License
37854 giving you legal permission to copy, distribute and/or modify it.
37856 For the developers' and authors' protection, the GPL clearly explains
37857 that there is no warranty for this free software. For both users' and
37858 authors' sake, the GPL requires that modified versions be marked as
37859 changed, so that their problems will not be attributed erroneously to
37860 authors of previous versions.
37862 Some devices are designed to deny users access to install or run
37863 modified versions of the software inside them, although the
37864 manufacturer can do so. This is fundamentally incompatible with the
37865 aim of protecting users' freedom to change the software. The
37866 systematic pattern of such abuse occurs in the area of products for
37867 individuals to use, which is precisely where it is most unacceptable.
37868 Therefore, we have designed this version of the GPL to prohibit the
37869 practice for those products. If such problems arise substantially in
37870 other domains, we stand ready to extend this provision to those domains
37871 in future versions of the GPL, as needed to protect the freedom of
37874 Finally, every program is threatened constantly by software patents.
37875 States should not allow patents to restrict development and use of
37876 software on general-purpose computers, but in those that do, we wish to
37877 avoid the special danger that patents applied to a free program could
37878 make it effectively proprietary. To prevent this, the GPL assures that
37879 patents cannot be used to render the program non-free.
37881 The precise terms and conditions for copying, distribution and
37882 modification follow.
37884 TERMS AND CONDITIONS
37885 ====================
37889 "This License" refers to version 3 of the GNU General Public
37892 "Copyright" also means copyright-like laws that apply to other
37893 kinds of works, such as semiconductor masks.
37895 "The Program" refers to any copyrightable work licensed under this
37896 License. Each licensee is addressed as "you". "Licensees" and
37897 "recipients" may be individuals or organizations.
37899 To "modify" a work means to copy from or adapt all or part of the
37900 work in a fashion requiring copyright permission, other than the
37901 making of an exact copy. The resulting work is called a "modified
37902 version" of the earlier work or a work "based on" the earlier work.
37904 A "covered work" means either the unmodified Program or a work
37905 based on the Program.
37907 To "propagate" a work means to do anything with it that, without
37908 permission, would make you directly or secondarily liable for
37909 infringement under applicable copyright law, except executing it
37910 on a computer or modifying a private copy. Propagation includes
37911 copying, distribution (with or without modification), making
37912 available to the public, and in some countries other activities as
37915 To "convey" a work means any kind of propagation that enables other
37916 parties to make or receive copies. Mere interaction with a user
37917 through a computer network, with no transfer of a copy, is not
37920 An interactive user interface displays "Appropriate Legal Notices"
37921 to the extent that it includes a convenient and prominently visible
37922 feature that (1) displays an appropriate copyright notice, and (2)
37923 tells the user that there is no warranty for the work (except to
37924 the extent that warranties are provided), that licensees may
37925 convey the work under this License, and how to view a copy of this
37926 License. If the interface presents a list of user commands or
37927 options, such as a menu, a prominent item in the list meets this
37932 The "source code" for a work means the preferred form of the work
37933 for making modifications to it. "Object code" means any
37934 non-source form of a work.
37936 A "Standard Interface" means an interface that either is an
37937 official standard defined by a recognized standards body, or, in
37938 the case of interfaces specified for a particular programming
37939 language, one that is widely used among developers working in that
37942 The "System Libraries" of an executable work include anything,
37943 other than the work as a whole, that (a) is included in the normal
37944 form of packaging a Major Component, but which is not part of that
37945 Major Component, and (b) serves only to enable use of the work
37946 with that Major Component, or to implement a Standard Interface
37947 for which an implementation is available to the public in source
37948 code form. A "Major Component", in this context, means a major
37949 essential component (kernel, window system, and so on) of the
37950 specific operating system (if any) on which the executable work
37951 runs, or a compiler used to produce the work, or an object code
37952 interpreter used to run it.
37954 The "Corresponding Source" for a work in object code form means all
37955 the source code needed to generate, install, and (for an executable
37956 work) run the object code and to modify the work, including
37957 scripts to control those activities. However, it does not include
37958 the work's System Libraries, or general-purpose tools or generally
37959 available free programs which are used unmodified in performing
37960 those activities but which are not part of the work. For example,
37961 Corresponding Source includes interface definition files
37962 associated with source files for the work, and the source code for
37963 shared libraries and dynamically linked subprograms that the work
37964 is specifically designed to require, such as by intimate data
37965 communication or control flow between those subprograms and other
37968 The Corresponding Source need not include anything that users can
37969 regenerate automatically from other parts of the Corresponding
37972 The Corresponding Source for a work in source code form is that
37975 2. Basic Permissions.
37977 All rights granted under this License are granted for the term of
37978 copyright on the Program, and are irrevocable provided the stated
37979 conditions are met. This License explicitly affirms your unlimited
37980 permission to run the unmodified Program. The output from running
37981 a covered work is covered by this License only if the output,
37982 given its content, constitutes a covered work. This License
37983 acknowledges your rights of fair use or other equivalent, as
37984 provided by copyright law.
37986 You may make, run and propagate covered works that you do not
37987 convey, without conditions so long as your license otherwise
37988 remains in force. You may convey covered works to others for the
37989 sole purpose of having them make modifications exclusively for
37990 you, or provide you with facilities for running those works,
37991 provided that you comply with the terms of this License in
37992 conveying all material for which you do not control copyright.
37993 Those thus making or running the covered works for you must do so
37994 exclusively on your behalf, under your direction and control, on
37995 terms that prohibit them from making any copies of your
37996 copyrighted material outside their relationship with you.
37998 Conveying under any other circumstances is permitted solely under
37999 the conditions stated below. Sublicensing is not allowed; section
38000 10 makes it unnecessary.
38002 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
38004 No covered work shall be deemed part of an effective technological
38005 measure under any applicable law fulfilling obligations under
38006 article 11 of the WIPO copyright treaty adopted on 20 December
38007 1996, or similar laws prohibiting or restricting circumvention of
38010 When you convey a covered work, you waive any legal power to forbid
38011 circumvention of technological measures to the extent such
38012 circumvention is effected by exercising rights under this License
38013 with respect to the covered work, and you disclaim any intention
38014 to limit operation or modification of the work as a means of
38015 enforcing, against the work's users, your or third parties' legal
38016 rights to forbid circumvention of technological measures.
38018 4. Conveying Verbatim Copies.
38020 You may convey verbatim copies of the Program's source code as you
38021 receive it, in any medium, provided that you conspicuously and
38022 appropriately publish on each copy an appropriate copyright notice;
38023 keep intact all notices stating that this License and any
38024 non-permissive terms added in accord with section 7 apply to the
38025 code; keep intact all notices of the absence of any warranty; and
38026 give all recipients a copy of this License along with the Program.
38028 You may charge any price or no price for each copy that you convey,
38029 and you may offer support or warranty protection for a fee.
38031 5. Conveying Modified Source Versions.
38033 You may convey a work based on the Program, or the modifications to
38034 produce it from the Program, in the form of source code under the
38035 terms of section 4, provided that you also meet all of these
38038 a. The work must carry prominent notices stating that you
38039 modified it, and giving a relevant date.
38041 b. The work must carry prominent notices stating that it is
38042 released under this License and any conditions added under
38043 section 7. This requirement modifies the requirement in
38044 section 4 to "keep intact all notices".
38046 c. You must license the entire work, as a whole, under this
38047 License to anyone who comes into possession of a copy. This
38048 License will therefore apply, along with any applicable
38049 section 7 additional terms, to the whole of the work, and all
38050 its parts, regardless of how they are packaged. This License
38051 gives no permission to license the work in any other way, but
38052 it does not invalidate such permission if you have separately
38055 d. If the work has interactive user interfaces, each must display
38056 Appropriate Legal Notices; however, if the Program has
38057 interactive interfaces that do not display Appropriate Legal
38058 Notices, your work need not make them do so.
38060 A compilation of a covered work with other separate and independent
38061 works, which are not by their nature extensions of the covered
38062 work, and which are not combined with it such as to form a larger
38063 program, in or on a volume of a storage or distribution medium, is
38064 called an "aggregate" if the compilation and its resulting
38065 copyright are not used to limit the access or legal rights of the
38066 compilation's users beyond what the individual works permit.
38067 Inclusion of a covered work in an aggregate does not cause this
38068 License to apply to the other parts of the aggregate.
38070 6. Conveying Non-Source Forms.
38072 You may convey a covered work in object code form under the terms
38073 of sections 4 and 5, provided that you also convey the
38074 machine-readable Corresponding Source under the terms of this
38075 License, in one of these ways:
38077 a. Convey the object code in, or embodied in, a physical product
38078 (including a physical distribution medium), accompanied by the
38079 Corresponding Source fixed on a durable physical medium
38080 customarily used for software interchange.
38082 b. Convey the object code in, or embodied in, a physical product
38083 (including a physical distribution medium), accompanied by a
38084 written offer, valid for at least three years and valid for
38085 as long as you offer spare parts or customer support for that
38086 product model, to give anyone who possesses the object code
38087 either (1) a copy of the Corresponding Source for all the
38088 software in the product that is covered by this License, on a
38089 durable physical medium customarily used for software
38090 interchange, for a price no more than your reasonable cost of
38091 physically performing this conveying of source, or (2) access
38092 to copy the Corresponding Source from a network server at no
38095 c. Convey individual copies of the object code with a copy of
38096 the written offer to provide the Corresponding Source. This
38097 alternative is allowed only occasionally and noncommercially,
38098 and only if you received the object code with such an offer,
38099 in accord with subsection 6b.
38101 d. Convey the object code by offering access from a designated
38102 place (gratis or for a charge), and offer equivalent access
38103 to the Corresponding Source in the same way through the same
38104 place at no further charge. You need not require recipients
38105 to copy the Corresponding Source along with the object code.
38106 If the place to copy the object code is a network server, the
38107 Corresponding Source may be on a different server (operated
38108 by you or a third party) that supports equivalent copying
38109 facilities, provided you maintain clear directions next to
38110 the object code saying where to find the Corresponding Source.
38111 Regardless of what server hosts the Corresponding Source, you
38112 remain obligated to ensure that it is available for as long
38113 as needed to satisfy these requirements.
38115 e. Convey the object code using peer-to-peer transmission,
38116 provided you inform other peers where the object code and
38117 Corresponding Source of the work are being offered to the
38118 general public at no charge under subsection 6d.
38121 A separable portion of the object code, whose source code is
38122 excluded from the Corresponding Source as a System Library, need
38123 not be included in conveying the object code work.
38125 A "User Product" is either (1) a "consumer product", which means
38126 any tangible personal property which is normally used for personal,
38127 family, or household purposes, or (2) anything designed or sold for
38128 incorporation into a dwelling. In determining whether a product
38129 is a consumer product, doubtful cases shall be resolved in favor of
38130 coverage. For a particular product received by a particular user,
38131 "normally used" refers to a typical or common use of that class of
38132 product, regardless of the status of the particular user or of the
38133 way in which the particular user actually uses, or expects or is
38134 expected to use, the product. A product is a consumer product
38135 regardless of whether the product has substantial commercial,
38136 industrial or non-consumer uses, unless such uses represent the
38137 only significant mode of use of the product.
38139 "Installation Information" for a User Product means any methods,
38140 procedures, authorization keys, or other information required to
38141 install and execute modified versions of a covered work in that
38142 User Product from a modified version of its Corresponding Source.
38143 The information must suffice to ensure that the continued
38144 functioning of the modified object code is in no case prevented or
38145 interfered with solely because modification has been made.
38147 If you convey an object code work under this section in, or with,
38148 or specifically for use in, a User Product, and the conveying
38149 occurs as part of a transaction in which the right of possession
38150 and use of the User Product is transferred to the recipient in
38151 perpetuity or for a fixed term (regardless of how the transaction
38152 is characterized), the Corresponding Source conveyed under this
38153 section must be accompanied by the Installation Information. But
38154 this requirement does not apply if neither you nor any third party
38155 retains the ability to install modified object code on the User
38156 Product (for example, the work has been installed in ROM).
38158 The requirement to provide Installation Information does not
38159 include a requirement to continue to provide support service,
38160 warranty, or updates for a work that has been modified or
38161 installed by the recipient, or for the User Product in which it
38162 has been modified or installed. Access to a network may be denied
38163 when the modification itself materially and adversely affects the
38164 operation of the network or violates the rules and protocols for
38165 communication across the network.
38167 Corresponding Source conveyed, and Installation Information
38168 provided, in accord with this section must be in a format that is
38169 publicly documented (and with an implementation available to the
38170 public in source code form), and must require no special password
38171 or key for unpacking, reading or copying.
38173 7. Additional Terms.
38175 "Additional permissions" are terms that supplement the terms of
38176 this License by making exceptions from one or more of its
38177 conditions. Additional permissions that are applicable to the
38178 entire Program shall be treated as though they were included in
38179 this License, to the extent that they are valid under applicable
38180 law. If additional permissions apply only to part of the Program,
38181 that part may be used separately under those permissions, but the
38182 entire Program remains governed by this License without regard to
38183 the additional permissions.
38185 When you convey a copy of a covered work, you may at your option
38186 remove any additional permissions from that copy, or from any part
38187 of it. (Additional permissions may be written to require their own
38188 removal in certain cases when you modify the work.) You may place
38189 additional permissions on material, added by you to a covered work,
38190 for which you have or can give appropriate copyright permission.
38192 Notwithstanding any other provision of this License, for material
38193 you add to a covered work, you may (if authorized by the copyright
38194 holders of that material) supplement the terms of this License
38197 a. Disclaiming warranty or limiting liability differently from
38198 the terms of sections 15 and 16 of this License; or
38200 b. Requiring preservation of specified reasonable legal notices
38201 or author attributions in that material or in the Appropriate
38202 Legal Notices displayed by works containing it; or
38204 c. Prohibiting misrepresentation of the origin of that material,
38205 or requiring that modified versions of such material be
38206 marked in reasonable ways as different from the original
38209 d. Limiting the use for publicity purposes of names of licensors
38210 or authors of the material; or
38212 e. Declining to grant rights under trademark law for use of some
38213 trade names, trademarks, or service marks; or
38215 f. Requiring indemnification of licensors and authors of that
38216 material by anyone who conveys the material (or modified
38217 versions of it) with contractual assumptions of liability to
38218 the recipient, for any liability that these contractual
38219 assumptions directly impose on those licensors and authors.
38221 All other non-permissive additional terms are considered "further
38222 restrictions" within the meaning of section 10. If the Program as
38223 you received it, or any part of it, contains a notice stating that
38224 it is governed by this License along with a term that is a further
38225 restriction, you may remove that term. If a license document
38226 contains a further restriction but permits relicensing or
38227 conveying under this License, you may add to a covered work
38228 material governed by the terms of that license document, provided
38229 that the further restriction does not survive such relicensing or
38232 If you add terms to a covered work in accord with this section, you
38233 must place, in the relevant source files, a statement of the
38234 additional terms that apply to those files, or a notice indicating
38235 where to find the applicable terms.
38237 Additional terms, permissive or non-permissive, may be stated in
38238 the form of a separately written license, or stated as exceptions;
38239 the above requirements apply either way.
38243 You may not propagate or modify a covered work except as expressly
38244 provided under this License. Any attempt otherwise to propagate or
38245 modify it is void, and will automatically terminate your rights
38246 under this License (including any patent licenses granted under
38247 the third paragraph of section 11).
38249 However, if you cease all violation of this License, then your
38250 license from a particular copyright holder is reinstated (a)
38251 provisionally, unless and until the copyright holder explicitly
38252 and finally terminates your license, and (b) permanently, if the
38253 copyright holder fails to notify you of the violation by some
38254 reasonable means prior to 60 days after the cessation.
38256 Moreover, your license from a particular copyright holder is
38257 reinstated permanently if the copyright holder notifies you of the
38258 violation by some reasonable means, this is the first time you have
38259 received notice of violation of this License (for any work) from
38260 that copyright holder, and you cure the violation prior to 30 days
38261 after your receipt of the notice.
38263 Termination of your rights under this section does not terminate
38264 the licenses of parties who have received copies or rights from
38265 you under this License. If your rights have been terminated and
38266 not permanently reinstated, you do not qualify to receive new
38267 licenses for the same material under section 10.
38269 9. Acceptance Not Required for Having Copies.
38271 You are not required to accept this License in order to receive or
38272 run a copy of the Program. Ancillary propagation of a covered work
38273 occurring solely as a consequence of using peer-to-peer
38274 transmission to receive a copy likewise does not require
38275 acceptance. However, nothing other than this License grants you
38276 permission to propagate or modify any covered work. These actions
38277 infringe copyright if you do not accept this License. Therefore,
38278 by modifying or propagating a covered work, you indicate your
38279 acceptance of this License to do so.
38281 10. Automatic Licensing of Downstream Recipients.
38283 Each time you convey a covered work, the recipient automatically
38284 receives a license from the original licensors, to run, modify and
38285 propagate that work, subject to this License. You are not
38286 responsible for enforcing compliance by third parties with this
38289 An "entity transaction" is a transaction transferring control of an
38290 organization, or substantially all assets of one, or subdividing an
38291 organization, or merging organizations. If propagation of a
38292 covered work results from an entity transaction, each party to that
38293 transaction who receives a copy of the work also receives whatever
38294 licenses to the work the party's predecessor in interest had or
38295 could give under the previous paragraph, plus a right to
38296 possession of the Corresponding Source of the work from the
38297 predecessor in interest, if the predecessor has it or can get it
38298 with reasonable efforts.
38300 You may not impose any further restrictions on the exercise of the
38301 rights granted or affirmed under this License. For example, you
38302 may not impose a license fee, royalty, or other charge for
38303 exercise of rights granted under this License, and you may not
38304 initiate litigation (including a cross-claim or counterclaim in a
38305 lawsuit) alleging that any patent claim is infringed by making,
38306 using, selling, offering for sale, or importing the Program or any
38311 A "contributor" is a copyright holder who authorizes use under this
38312 License of the Program or a work on which the Program is based.
38313 The work thus licensed is called the contributor's "contributor
38316 A contributor's "essential patent claims" are all patent claims
38317 owned or controlled by the contributor, whether already acquired or
38318 hereafter acquired, that would be infringed by some manner,
38319 permitted by this License, of making, using, or selling its
38320 contributor version, but do not include claims that would be
38321 infringed only as a consequence of further modification of the
38322 contributor version. For purposes of this definition, "control"
38323 includes the right to grant patent sublicenses in a manner
38324 consistent with the requirements of this License.
38326 Each contributor grants you a non-exclusive, worldwide,
38327 royalty-free patent license under the contributor's essential
38328 patent claims, to make, use, sell, offer for sale, import and
38329 otherwise run, modify and propagate the contents of its
38330 contributor version.
38332 In the following three paragraphs, a "patent license" is any
38333 express agreement or commitment, however denominated, not to
38334 enforce a patent (such as an express permission to practice a
38335 patent or covenant not to sue for patent infringement). To
38336 "grant" such a patent license to a party means to make such an
38337 agreement or commitment not to enforce a patent against the party.
38339 If you convey a covered work, knowingly relying on a patent
38340 license, and the Corresponding Source of the work is not available
38341 for anyone to copy, free of charge and under the terms of this
38342 License, through a publicly available network server or other
38343 readily accessible means, then you must either (1) cause the
38344 Corresponding Source to be so available, or (2) arrange to deprive
38345 yourself of the benefit of the patent license for this particular
38346 work, or (3) arrange, in a manner consistent with the requirements
38347 of this License, to extend the patent license to downstream
38348 recipients. "Knowingly relying" means you have actual knowledge
38349 that, but for the patent license, your conveying the covered work
38350 in a country, or your recipient's use of the covered work in a
38351 country, would infringe one or more identifiable patents in that
38352 country that you have reason to believe are valid.
38354 If, pursuant to or in connection with a single transaction or
38355 arrangement, you convey, or propagate by procuring conveyance of, a
38356 covered work, and grant a patent license to some of the parties
38357 receiving the covered work authorizing them to use, propagate,
38358 modify or convey a specific copy of the covered work, then the
38359 patent license you grant is automatically extended to all
38360 recipients of the covered work and works based on it.
38362 A patent license is "discriminatory" if it does not include within
38363 the scope of its coverage, prohibits the exercise of, or is
38364 conditioned on the non-exercise of one or more of the rights that
38365 are specifically granted under this License. You may not convey a
38366 covered work if you are a party to an arrangement with a third
38367 party that is in the business of distributing software, under
38368 which you make payment to the third party based on the extent of
38369 your activity of conveying the work, and under which the third
38370 party grants, to any of the parties who would receive the covered
38371 work from you, a discriminatory patent license (a) in connection
38372 with copies of the covered work conveyed by you (or copies made
38373 from those copies), or (b) primarily for and in connection with
38374 specific products or compilations that contain the covered work,
38375 unless you entered into that arrangement, or that patent license
38376 was granted, prior to 28 March 2007.
38378 Nothing in this License shall be construed as excluding or limiting
38379 any implied license or other defenses to infringement that may
38380 otherwise be available to you under applicable patent law.
38382 12. No Surrender of Others' Freedom.
38384 If conditions are imposed on you (whether by court order,
38385 agreement or otherwise) that contradict the conditions of this
38386 License, they do not excuse you from the conditions of this
38387 License. If you cannot convey a covered work so as to satisfy
38388 simultaneously your obligations under this License and any other
38389 pertinent obligations, then as a consequence you may not convey it
38390 at all. For example, if you agree to terms that obligate you to
38391 collect a royalty for further conveying from those to whom you
38392 convey the Program, the only way you could satisfy both those
38393 terms and this License would be to refrain entirely from conveying
38396 13. Use with the GNU Affero General Public License.
38398 Notwithstanding any other provision of this License, you have
38399 permission to link or combine any covered work with a work licensed
38400 under version 3 of the GNU Affero General Public License into a
38401 single combined work, and to convey the resulting work. The terms
38402 of this License will continue to apply to the part which is the
38403 covered work, but the special requirements of the GNU Affero
38404 General Public License, section 13, concerning interaction through
38405 a network will apply to the combination as such.
38407 14. Revised Versions of this License.
38409 The Free Software Foundation may publish revised and/or new
38410 versions of the GNU General Public License from time to time.
38411 Such new versions will be similar in spirit to the present
38412 version, but may differ in detail to address new problems or
38415 Each version is given a distinguishing version number. If the
38416 Program specifies that a certain numbered version of the GNU
38417 General Public License "or any later version" applies to it, you
38418 have the option of following the terms and conditions either of
38419 that numbered version or of any later version published by the
38420 Free Software Foundation. If the Program does not specify a
38421 version number of the GNU General Public License, you may choose
38422 any version ever published by the Free Software Foundation.
38424 If the Program specifies that a proxy can decide which future
38425 versions of the GNU General Public License can be used, that
38426 proxy's public statement of acceptance of a version permanently
38427 authorizes you to choose that version for the Program.
38429 Later license versions may give you additional or different
38430 permissions. However, no additional obligations are imposed on any
38431 author or copyright holder as a result of your choosing to follow a
38434 15. Disclaimer of Warranty.
38436 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
38437 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE
38438 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS"
38439 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED,
38440 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
38441 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE
38442 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.
38443 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
38444 NECESSARY SERVICING, REPAIR OR CORRECTION.
38446 16. Limitation of Liability.
38448 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
38449 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES
38450 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU
38451 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
38452 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
38453 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA
38454 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
38455 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
38456 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF
38457 THE POSSIBILITY OF SUCH DAMAGES.
38459 17. Interpretation of Sections 15 and 16.
38461 If the disclaimer of warranty and limitation of liability provided
38462 above cannot be given local legal effect according to their terms,
38463 reviewing courts shall apply local law that most closely
38464 approximates an absolute waiver of all civil liability in
38465 connection with the Program, unless a warranty or assumption of
38466 liability accompanies a copy of the Program in return for a fee.
38469 END OF TERMS AND CONDITIONS
38470 ===========================
38472 How to Apply These Terms to Your New Programs
38473 =============================================
38475 If you develop a new program, and you want it to be of the greatest
38476 possible use to the public, the best way to achieve this is to make it
38477 free software which everyone can redistribute and change under these
38480 To do so, attach the following notices to the program. It is safest
38481 to attach them to the start of each source file to most effectively
38482 state the exclusion of warranty; and each file should have at least the
38483 "copyright" line and a pointer to where the full notice is found.
38485 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES.
38486 Copyright (C) YEAR NAME OF AUTHOR
38488 This program is free software: you can redistribute it and/or modify
38489 it under the terms of the GNU General Public License as published by
38490 the Free Software Foundation, either version 3 of the License, or (at
38491 your option) any later version.
38493 This program is distributed in the hope that it will be useful, but
38494 WITHOUT ANY WARRANTY; without even the implied warranty of
38495 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
38496 General Public License for more details.
38498 You should have received a copy of the GNU General Public License
38499 along with this program. If not, see `http://www.gnu.org/licenses/'.
38501 Also add information on how to contact you by electronic and paper
38504 If the program does terminal interaction, make it output a short
38505 notice like this when it starts in an interactive mode:
38507 PROGRAM Copyright (C) YEAR NAME OF AUTHOR
38508 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
38509 This is free software, and you are welcome to redistribute it
38510 under certain conditions; type `show c' for details.
38512 The hypothetical commands `show w' and `show c' should show the
38513 appropriate parts of the General Public License. Of course, your
38514 program's commands might be different; for a GUI interface, you would
38515 use an "about box".
38517 You should also get your employer (if you work as a programmer) or
38518 school, if any, to sign a "copyright disclaimer" for the program, if
38519 necessary. For more information on this, and how to apply and follow
38520 the GNU GPL, see `http://www.gnu.org/licenses/'.
38522 The GNU General Public License does not permit incorporating your
38523 program into proprietary programs. If your program is a subroutine
38524 library, you may consider it more useful to permit linking proprietary
38525 applications with the library. If this is what you want to do, use the
38526 GNU Lesser General Public License instead of this License. But first,
38527 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'.
38530 File: gcc.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top
38532 GNU Free Documentation License
38533 ******************************
38535 Version 1.2, November 2002
38537 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc.
38538 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
38540 Everyone is permitted to copy and distribute verbatim copies
38541 of this license document, but changing it is not allowed.
38545 The purpose of this License is to make a manual, textbook, or other
38546 functional and useful document "free" in the sense of freedom: to
38547 assure everyone the effective freedom to copy and redistribute it,
38548 with or without modifying it, either commercially or
38549 noncommercially. Secondarily, this License preserves for the
38550 author and publisher a way to get credit for their work, while not
38551 being considered responsible for modifications made by others.
38553 This License is a kind of "copyleft", which means that derivative
38554 works of the document must themselves be free in the same sense.
38555 It complements the GNU General Public License, which is a copyleft
38556 license designed for free software.
38558 We have designed this License in order to use it for manuals for
38559 free software, because free software needs free documentation: a
38560 free program should come with manuals providing the same freedoms
38561 that the software does. But this License is not limited to
38562 software manuals; it can be used for any textual work, regardless
38563 of subject matter or whether it is published as a printed book.
38564 We recommend this License principally for works whose purpose is
38565 instruction or reference.
38567 1. APPLICABILITY AND DEFINITIONS
38569 This License applies to any manual or other work, in any medium,
38570 that contains a notice placed by the copyright holder saying it
38571 can be distributed under the terms of this License. Such a notice
38572 grants a world-wide, royalty-free license, unlimited in duration,
38573 to use that work under the conditions stated herein. The
38574 "Document", below, refers to any such manual or work. Any member
38575 of the public is a licensee, and is addressed as "you". You
38576 accept the license if you copy, modify or distribute the work in a
38577 way requiring permission under copyright law.
38579 A "Modified Version" of the Document means any work containing the
38580 Document or a portion of it, either copied verbatim, or with
38581 modifications and/or translated into another language.
38583 A "Secondary Section" is a named appendix or a front-matter section
38584 of the Document that deals exclusively with the relationship of the
38585 publishers or authors of the Document to the Document's overall
38586 subject (or to related matters) and contains nothing that could
38587 fall directly within that overall subject. (Thus, if the Document
38588 is in part a textbook of mathematics, a Secondary Section may not
38589 explain any mathematics.) The relationship could be a matter of
38590 historical connection with the subject or with related matters, or
38591 of legal, commercial, philosophical, ethical or political position
38594 The "Invariant Sections" are certain Secondary Sections whose
38595 titles are designated, as being those of Invariant Sections, in
38596 the notice that says that the Document is released under this
38597 License. If a section does not fit the above definition of
38598 Secondary then it is not allowed to be designated as Invariant.
38599 The Document may contain zero Invariant Sections. If the Document
38600 does not identify any Invariant Sections then there are none.
38602 The "Cover Texts" are certain short passages of text that are
38603 listed, as Front-Cover Texts or Back-Cover Texts, in the notice
38604 that says that the Document is released under this License. A
38605 Front-Cover Text may be at most 5 words, and a Back-Cover Text may
38606 be at most 25 words.
38608 A "Transparent" copy of the Document means a machine-readable copy,
38609 represented in a format whose specification is available to the
38610 general public, that is suitable for revising the document
38611 straightforwardly with generic text editors or (for images
38612 composed of pixels) generic paint programs or (for drawings) some
38613 widely available drawing editor, and that is suitable for input to
38614 text formatters or for automatic translation to a variety of
38615 formats suitable for input to text formatters. A copy made in an
38616 otherwise Transparent file format whose markup, or absence of
38617 markup, has been arranged to thwart or discourage subsequent
38618 modification by readers is not Transparent. An image format is
38619 not Transparent if used for any substantial amount of text. A
38620 copy that is not "Transparent" is called "Opaque".
38622 Examples of suitable formats for Transparent copies include plain
38623 ASCII without markup, Texinfo input format, LaTeX input format,
38624 SGML or XML using a publicly available DTD, and
38625 standard-conforming simple HTML, PostScript or PDF designed for
38626 human modification. Examples of transparent image formats include
38627 PNG, XCF and JPG. Opaque formats include proprietary formats that
38628 can be read and edited only by proprietary word processors, SGML or
38629 XML for which the DTD and/or processing tools are not generally
38630 available, and the machine-generated HTML, PostScript or PDF
38631 produced by some word processors for output purposes only.
38633 The "Title Page" means, for a printed book, the title page itself,
38634 plus such following pages as are needed to hold, legibly, the
38635 material this License requires to appear in the title page. For
38636 works in formats which do not have any title page as such, "Title
38637 Page" means the text near the most prominent appearance of the
38638 work's title, preceding the beginning of the body of the text.
38640 A section "Entitled XYZ" means a named subunit of the Document
38641 whose title either is precisely XYZ or contains XYZ in parentheses
38642 following text that translates XYZ in another language. (Here XYZ
38643 stands for a specific section name mentioned below, such as
38644 "Acknowledgements", "Dedications", "Endorsements", or "History".)
38645 To "Preserve the Title" of such a section when you modify the
38646 Document means that it remains a section "Entitled XYZ" according
38647 to this definition.
38649 The Document may include Warranty Disclaimers next to the notice
38650 which states that this License applies to the Document. These
38651 Warranty Disclaimers are considered to be included by reference in
38652 this License, but only as regards disclaiming warranties: any other
38653 implication that these Warranty Disclaimers may have is void and
38654 has no effect on the meaning of this License.
38656 2. VERBATIM COPYING
38658 You may copy and distribute the Document in any medium, either
38659 commercially or noncommercially, provided that this License, the
38660 copyright notices, and the license notice saying this License
38661 applies to the Document are reproduced in all copies, and that you
38662 add no other conditions whatsoever to those of this License. You
38663 may not use technical measures to obstruct or control the reading
38664 or further copying of the copies you make or distribute. However,
38665 you may accept compensation in exchange for copies. If you
38666 distribute a large enough number of copies you must also follow
38667 the conditions in section 3.
38669 You may also lend copies, under the same conditions stated above,
38670 and you may publicly display copies.
38672 3. COPYING IN QUANTITY
38674 If you publish printed copies (or copies in media that commonly
38675 have printed covers) of the Document, numbering more than 100, and
38676 the Document's license notice requires Cover Texts, you must
38677 enclose the copies in covers that carry, clearly and legibly, all
38678 these Cover Texts: Front-Cover Texts on the front cover, and
38679 Back-Cover Texts on the back cover. Both covers must also clearly
38680 and legibly identify you as the publisher of these copies. The
38681 front cover must present the full title with all words of the
38682 title equally prominent and visible. You may add other material
38683 on the covers in addition. Copying with changes limited to the
38684 covers, as long as they preserve the title of the Document and
38685 satisfy these conditions, can be treated as verbatim copying in
38688 If the required texts for either cover are too voluminous to fit
38689 legibly, you should put the first ones listed (as many as fit
38690 reasonably) on the actual cover, and continue the rest onto
38693 If you publish or distribute Opaque copies of the Document
38694 numbering more than 100, you must either include a
38695 machine-readable Transparent copy along with each Opaque copy, or
38696 state in or with each Opaque copy a computer-network location from
38697 which the general network-using public has access to download
38698 using public-standard network protocols a complete Transparent
38699 copy of the Document, free of added material. If you use the
38700 latter option, you must take reasonably prudent steps, when you
38701 begin distribution of Opaque copies in quantity, to ensure that
38702 this Transparent copy will remain thus accessible at the stated
38703 location until at least one year after the last time you
38704 distribute an Opaque copy (directly or through your agents or
38705 retailers) of that edition to the public.
38707 It is requested, but not required, that you contact the authors of
38708 the Document well before redistributing any large number of
38709 copies, to give them a chance to provide you with an updated
38710 version of the Document.
38714 You may copy and distribute a Modified Version of the Document
38715 under the conditions of sections 2 and 3 above, provided that you
38716 release the Modified Version under precisely this License, with
38717 the Modified Version filling the role of the Document, thus
38718 licensing distribution and modification of the Modified Version to
38719 whoever possesses a copy of it. In addition, you must do these
38720 things in the Modified Version:
38722 A. Use in the Title Page (and on the covers, if any) a title
38723 distinct from that of the Document, and from those of
38724 previous versions (which should, if there were any, be listed
38725 in the History section of the Document). You may use the
38726 same title as a previous version if the original publisher of
38727 that version gives permission.
38729 B. List on the Title Page, as authors, one or more persons or
38730 entities responsible for authorship of the modifications in
38731 the Modified Version, together with at least five of the
38732 principal authors of the Document (all of its principal
38733 authors, if it has fewer than five), unless they release you
38734 from this requirement.
38736 C. State on the Title page the name of the publisher of the
38737 Modified Version, as the publisher.
38739 D. Preserve all the copyright notices of the Document.
38741 E. Add an appropriate copyright notice for your modifications
38742 adjacent to the other copyright notices.
38744 F. Include, immediately after the copyright notices, a license
38745 notice giving the public permission to use the Modified
38746 Version under the terms of this License, in the form shown in
38747 the Addendum below.
38749 G. Preserve in that license notice the full lists of Invariant
38750 Sections and required Cover Texts given in the Document's
38753 H. Include an unaltered copy of this License.
38755 I. Preserve the section Entitled "History", Preserve its Title,
38756 and add to it an item stating at least the title, year, new
38757 authors, and publisher of the Modified Version as given on
38758 the Title Page. If there is no section Entitled "History" in
38759 the Document, create one stating the title, year, authors,
38760 and publisher of the Document as given on its Title Page,
38761 then add an item describing the Modified Version as stated in
38762 the previous sentence.
38764 J. Preserve the network location, if any, given in the Document
38765 for public access to a Transparent copy of the Document, and
38766 likewise the network locations given in the Document for
38767 previous versions it was based on. These may be placed in
38768 the "History" section. You may omit a network location for a
38769 work that was published at least four years before the
38770 Document itself, or if the original publisher of the version
38771 it refers to gives permission.
38773 K. For any section Entitled "Acknowledgements" or "Dedications",
38774 Preserve the Title of the section, and preserve in the
38775 section all the substance and tone of each of the contributor
38776 acknowledgements and/or dedications given therein.
38778 L. Preserve all the Invariant Sections of the Document,
38779 unaltered in their text and in their titles. Section numbers
38780 or the equivalent are not considered part of the section
38783 M. Delete any section Entitled "Endorsements". Such a section
38784 may not be included in the Modified Version.
38786 N. Do not retitle any existing section to be Entitled
38787 "Endorsements" or to conflict in title with any Invariant
38790 O. Preserve any Warranty Disclaimers.
38792 If the Modified Version includes new front-matter sections or
38793 appendices that qualify as Secondary Sections and contain no
38794 material copied from the Document, you may at your option
38795 designate some or all of these sections as invariant. To do this,
38796 add their titles to the list of Invariant Sections in the Modified
38797 Version's license notice. These titles must be distinct from any
38798 other section titles.
38800 You may add a section Entitled "Endorsements", provided it contains
38801 nothing but endorsements of your Modified Version by various
38802 parties--for example, statements of peer review or that the text
38803 has been approved by an organization as the authoritative
38804 definition of a standard.
38806 You may add a passage of up to five words as a Front-Cover Text,
38807 and a passage of up to 25 words as a Back-Cover Text, to the end
38808 of the list of Cover Texts in the Modified Version. Only one
38809 passage of Front-Cover Text and one of Back-Cover Text may be
38810 added by (or through arrangements made by) any one entity. If the
38811 Document already includes a cover text for the same cover,
38812 previously added by you or by arrangement made by the same entity
38813 you are acting on behalf of, you may not add another; but you may
38814 replace the old one, on explicit permission from the previous
38815 publisher that added the old one.
38817 The author(s) and publisher(s) of the Document do not by this
38818 License give permission to use their names for publicity for or to
38819 assert or imply endorsement of any Modified Version.
38821 5. COMBINING DOCUMENTS
38823 You may combine the Document with other documents released under
38824 this License, under the terms defined in section 4 above for
38825 modified versions, provided that you include in the combination
38826 all of the Invariant Sections of all of the original documents,
38827 unmodified, and list them all as Invariant Sections of your
38828 combined work in its license notice, and that you preserve all
38829 their Warranty Disclaimers.
38831 The combined work need only contain one copy of this License, and
38832 multiple identical Invariant Sections may be replaced with a single
38833 copy. If there are multiple Invariant Sections with the same name
38834 but different contents, make the title of each such section unique
38835 by adding at the end of it, in parentheses, the name of the
38836 original author or publisher of that section if known, or else a
38837 unique number. Make the same adjustment to the section titles in
38838 the list of Invariant Sections in the license notice of the
38841 In the combination, you must combine any sections Entitled
38842 "History" in the various original documents, forming one section
38843 Entitled "History"; likewise combine any sections Entitled
38844 "Acknowledgements", and any sections Entitled "Dedications". You
38845 must delete all sections Entitled "Endorsements."
38847 6. COLLECTIONS OF DOCUMENTS
38849 You may make a collection consisting of the Document and other
38850 documents released under this License, and replace the individual
38851 copies of this License in the various documents with a single copy
38852 that is included in the collection, provided that you follow the
38853 rules of this License for verbatim copying of each of the
38854 documents in all other respects.
38856 You may extract a single document from such a collection, and
38857 distribute it individually under this License, provided you insert
38858 a copy of this License into the extracted document, and follow
38859 this License in all other respects regarding verbatim copying of
38862 7. AGGREGATION WITH INDEPENDENT WORKS
38864 A compilation of the Document or its derivatives with other
38865 separate and independent documents or works, in or on a volume of
38866 a storage or distribution medium, is called an "aggregate" if the
38867 copyright resulting from the compilation is not used to limit the
38868 legal rights of the compilation's users beyond what the individual
38869 works permit. When the Document is included in an aggregate, this
38870 License does not apply to the other works in the aggregate which
38871 are not themselves derivative works of the Document.
38873 If the Cover Text requirement of section 3 is applicable to these
38874 copies of the Document, then if the Document is less than one half
38875 of the entire aggregate, the Document's Cover Texts may be placed
38876 on covers that bracket the Document within the aggregate, or the
38877 electronic equivalent of covers if the Document is in electronic
38878 form. Otherwise they must appear on printed covers that bracket
38879 the whole aggregate.
38883 Translation is considered a kind of modification, so you may
38884 distribute translations of the Document under the terms of section
38885 4. Replacing Invariant Sections with translations requires special
38886 permission from their copyright holders, but you may include
38887 translations of some or all Invariant Sections in addition to the
38888 original versions of these Invariant Sections. You may include a
38889 translation of this License, and all the license notices in the
38890 Document, and any Warranty Disclaimers, provided that you also
38891 include the original English version of this License and the
38892 original versions of those notices and disclaimers. In case of a
38893 disagreement between the translation and the original version of
38894 this License or a notice or disclaimer, the original version will
38897 If a section in the Document is Entitled "Acknowledgements",
38898 "Dedications", or "History", the requirement (section 4) to
38899 Preserve its Title (section 1) will typically require changing the
38904 You may not copy, modify, sublicense, or distribute the Document
38905 except as expressly provided for under this License. Any other
38906 attempt to copy, modify, sublicense or distribute the Document is
38907 void, and will automatically terminate your rights under this
38908 License. However, parties who have received copies, or rights,
38909 from you under this License will not have their licenses
38910 terminated so long as such parties remain in full compliance.
38912 10. FUTURE REVISIONS OF THIS LICENSE
38914 The Free Software Foundation may publish new, revised versions of
38915 the GNU Free Documentation License from time to time. Such new
38916 versions will be similar in spirit to the present version, but may
38917 differ in detail to address new problems or concerns. See
38918 `http://www.gnu.org/copyleft/'.
38920 Each version of the License is given a distinguishing version
38921 number. If the Document specifies that a particular numbered
38922 version of this License "or any later version" applies to it, you
38923 have the option of following the terms and conditions either of
38924 that specified version or of any later version that has been
38925 published (not as a draft) by the Free Software Foundation. If
38926 the Document does not specify a version number of this License,
38927 you may choose any version ever published (not as a draft) by the
38928 Free Software Foundation.
38930 ADDENDUM: How to use this License for your documents
38931 ====================================================
38933 To use this License in a document you have written, include a copy of
38934 the License in the document and put the following copyright and license
38935 notices just after the title page:
38937 Copyright (C) YEAR YOUR NAME.
38938 Permission is granted to copy, distribute and/or modify this document
38939 under the terms of the GNU Free Documentation License, Version 1.2
38940 or any later version published by the Free Software Foundation;
38941 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
38942 Texts. A copy of the license is included in the section entitled ``GNU
38943 Free Documentation License''.
38945 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts,
38946 replace the "with...Texts." line with this:
38948 with the Invariant Sections being LIST THEIR TITLES, with
38949 the Front-Cover Texts being LIST, and with the Back-Cover Texts
38952 If you have Invariant Sections without Cover Texts, or some other
38953 combination of the three, merge those two alternatives to suit the
38956 If your document contains nontrivial examples of program code, we
38957 recommend releasing these examples in parallel under your choice of
38958 free software license, such as the GNU General Public License, to
38959 permit their use in free software.
38962 File: gcc.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top
38964 Contributors to GCC
38965 *******************
38967 The GCC project would like to thank its many contributors. Without
38968 them the project would not have been nearly as successful as it has
38969 been. Any omissions in this list are accidental. Feel free to contact
38970 <law@redhat.com> or <gerald@pfeifer.com> if you have been left out or
38971 some of your contributions are not listed. Please keep this list in
38972 alphabetical order.
38974 * Analog Devices helped implement the support for complex data types
38977 * John David Anglin for threading-related fixes and improvements to
38978 libstdc++-v3, and the HP-UX port.
38980 * James van Artsdalen wrote the code that makes efficient use of the
38981 Intel 80387 register stack.
38983 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta
38986 * Alasdair Baird for various bug fixes.
38988 * Giovanni Bajo for analyzing lots of complicated C++ problem
38991 * Peter Barada for his work to improve code generation for new
38994 * Gerald Baumgartner added the signature extension to the C++ front
38997 * Godmar Back for his Java improvements and encouragement.
38999 * Scott Bambrough for help porting the Java compiler.
39001 * Wolfgang Bangerth for processing tons of bug reports.
39003 * Jon Beniston for his Microsoft Windows port of Java.
39005 * Daniel Berlin for better DWARF2 support, faster/better
39006 optimizations, improved alias analysis, plus migrating GCC to
39009 * Geoff Berry for his Java object serialization work and various
39012 * Uros Bizjak for the implementation of x87 math built-in functions
39013 and for various middle end and i386 back end improvements and bug
39016 * Eric Blake for helping to make GCJ and libgcj conform to the
39019 * Janne Blomqvist for contributions to GNU Fortran.
39021 * Segher Boessenkool for various fixes.
39023 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and
39026 * Neil Booth for work on cpplib, lang hooks, debug hooks and other
39027 miscellaneous clean-ups.
39029 * Steven Bosscher for integrating the GNU Fortran front end into GCC
39030 and for contributing to the tree-ssa branch.
39032 * Eric Botcazou for fixing middle- and backend bugs left and right.
39034 * Per Bothner for his direction via the steering committee and
39035 various improvements to the infrastructure for supporting new
39036 languages. Chill front end implementation. Initial
39037 implementations of cpplib, fix-header, config.guess, libio, and
39038 past C++ library (libg++) maintainer. Dreaming up, designing and
39039 implementing much of GCJ.
39041 * Devon Bowen helped port GCC to the Tahoe.
39043 * Don Bowman for mips-vxworks contributions.
39045 * Dave Brolley for work on cpplib and Chill.
39047 * Paul Brook for work on the ARM architecture and maintaining GNU
39050 * Robert Brown implemented the support for Encore 32000 systems.
39052 * Christian Bruel for improvements to local store elimination.
39054 * Herman A.J. ten Brugge for various fixes.
39056 * Joerg Brunsmann for Java compiler hacking and help with the GCJ
39059 * Joe Buck for his direction via the steering committee.
39061 * Craig Burley for leadership of the G77 Fortran effort.
39063 * Stephan Buys for contributing Doxygen notes for libstdc++.
39065 * Paolo Carlini for libstdc++ work: lots of efficiency improvements
39066 to the C++ strings, streambufs and formatted I/O, hard detective
39067 work on the frustrating localization issues, and keeping up with
39068 the problem reports.
39070 * John Carr for his alias work, SPARC hacking, infrastructure
39071 improvements, previous contributions to the steering committee,
39072 loop optimizations, etc.
39074 * Stephane Carrez for 68HC11 and 68HC12 ports.
39076 * Steve Chamberlain for support for the Renesas SH and H8 processors
39077 and the PicoJava processor, and for GCJ config fixes.
39079 * Glenn Chambers for help with the GCJ FAQ.
39081 * John-Marc Chandonia for various libgcj patches.
39083 * Scott Christley for his Objective-C contributions.
39085 * Eric Christopher for his Java porting help and clean-ups.
39087 * Branko Cibej for more warning contributions.
39089 * The GNU Classpath project for all of their merged runtime code.
39091 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and
39092 other random hacking.
39094 * Michael Cook for libstdc++ cleanup patches to reduce warnings.
39096 * R. Kelley Cook for making GCC buildable from a read-only directory
39097 as well as other miscellaneous build process and documentation
39100 * Ralf Corsepius for SH testing and minor bug fixing.
39102 * Stan Cox for care and feeding of the x86 port and lots of behind
39103 the scenes hacking.
39105 * Alex Crain provided changes for the 3b1.
39107 * Ian Dall for major improvements to the NS32k port.
39109 * Paul Dale for his work to add uClinux platform support to the m68k
39112 * Dario Dariol contributed the four varieties of sample programs
39113 that print a copy of their source.
39115 * Russell Davidson for fstream and stringstream fixes in libstdc++.
39117 * Bud Davis for work on the G77 and GNU Fortran compilers.
39119 * Mo DeJong for GCJ and libgcj bug fixes.
39121 * DJ Delorie for the DJGPP port, build and libiberty maintenance,
39122 various bug fixes, and the M32C port.
39124 * Arnaud Desitter for helping to debug GNU Fortran.
39126 * Gabriel Dos Reis for contributions to G++, contributions and
39127 maintenance of GCC diagnostics infrastructure, libstdc++-v3,
39128 including `valarray<>', `complex<>', maintaining the numerics
39129 library (including that pesky `<limits>' :-) and keeping
39130 up-to-date anything to do with numbers.
39132 * Ulrich Drepper for his work on glibc, testing of GCC using glibc,
39133 ISO C99 support, CFG dumping support, etc., plus support of the
39134 C++ runtime libraries including for all kinds of C interface
39135 issues, contributing and maintaining `complex<>', sanity checking
39136 and disbursement, configuration architecture, libio maintenance,
39137 and early math work.
39139 * Zdenek Dvorak for a new loop unroller and various fixes.
39141 * Richard Earnshaw for his ongoing work with the ARM.
39143 * David Edelsohn for his direction via the steering committee,
39144 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa
39145 loop changes, doing the entire AIX port of libstdc++ with his bare
39146 hands, and for ensuring GCC properly keeps working on AIX.
39148 * Kevin Ediger for the floating point formatting of num_put::do_put
39151 * Phil Edwards for libstdc++ work including configuration hackery,
39152 documentation maintainer, chief breaker of the web pages, the
39153 occasional iostream bug fix, and work on shared library symbol
39156 * Paul Eggert for random hacking all over GCC.
39158 * Mark Elbrecht for various DJGPP improvements, and for libstdc++
39159 configuration support for locales and fstream-related fixes.
39161 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and
39164 * Christian Ehrhardt for dealing with bug reports.
39166 * Ben Elliston for his work to move the Objective-C runtime into its
39167 own subdirectory and for his work on autoconf.
39169 * Revital Eres for work on the PowerPC 750CL port.
39171 * Marc Espie for OpenBSD support.
39173 * Doug Evans for much of the global optimization framework, arc,
39174 m32r, and SPARC work.
39176 * Christopher Faylor for his work on the Cygwin port and for caring
39177 and feeding the gcc.gnu.org box and saving its users tons of spam.
39179 * Fred Fish for BeOS support and Ada fixes.
39181 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ.
39183 * Peter Gerwinski for various bug fixes and the Pascal front end.
39185 * Kaveh R. Ghazi for his direction via the steering committee,
39186 amazing work to make `-W -Wall -W* -Werror' useful, and
39187 continuously testing GCC on a plethora of platforms. Kaveh
39188 extends his gratitude to the CAIP Center at Rutgers University for
39189 providing him with computing resources to work on Free Software
39190 since the late 1980s.
39192 * John Gilmore for a donation to the FSF earmarked improving GNU
39195 * Judy Goldberg for c++ contributions.
39197 * Torbjorn Granlund for various fixes and the c-torture testsuite,
39198 multiply- and divide-by-constant optimization, improved long long
39199 support, improved leaf function register allocation, and his
39200 direction via the steering committee.
39202 * Anthony Green for his `-Os' contributions and Java front end work.
39204 * Stu Grossman for gdb hacking, allowing GCJ developers to debug
39207 * Michael K. Gschwind contributed the port to the PDP-11.
39209 * Ron Guilmette implemented the `protoize' and `unprotoize' tools,
39210 the support for Dwarf symbolic debugging information, and much of
39211 the support for System V Release 4. He has also worked heavily on
39212 the Intel 386 and 860 support.
39214 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload
39217 * Bruno Haible for improvements in the runtime overhead for EH, new
39218 warnings and assorted bug fixes.
39220 * Andrew Haley for his amazing Java compiler and library efforts.
39222 * Chris Hanson assisted in making GCC work on HP-UX for the 9000
39225 * Michael Hayes for various thankless work he's done trying to get
39226 the c30/c40 ports functional. Lots of loop and unroll
39227 improvements and fixes.
39229 * Dara Hazeghi for wading through myriads of target-specific bug
39232 * Kate Hedstrom for staking the G77 folks with an initial testsuite.
39234 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64
39235 work, loop opts, and generally fixing lots of old problems we've
39236 ignored for years, flow rewrite and lots of further stuff,
39237 including reviewing tons of patches.
39239 * Aldy Hernandez for working on the PowerPC port, SIMD support, and
39242 * Nobuyuki Hikichi of Software Research Associates, Tokyo,
39243 contributed the support for the Sony NEWS machine.
39245 * Kazu Hirata for caring and feeding the Renesas H8/300 port and
39248 * Katherine Holcomb for work on GNU Fortran.
39250 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots
39251 of testing and bug fixing, particularly of GCC configury code.
39253 * Steve Holmgren for MachTen patches.
39255 * Jan Hubicka for his x86 port improvements.
39257 * Falk Hueffner for working on C and optimization bug reports.
39259 * Bernardo Innocenti for his m68k work, including merging of
39260 ColdFire improvements and uClinux support.
39262 * Christian Iseli for various bug fixes.
39264 * Kamil Iskra for general m68k hacking.
39266 * Lee Iverson for random fixes and MIPS testing.
39268 * Andreas Jaeger for testing and benchmarking of GCC and various bug
39271 * Jakub Jelinek for his SPARC work and sibling call optimizations as
39272 well as lots of bug fixes and test cases, and for improving the
39275 * Janis Johnson for ia64 testing and fixes, her quality improvement
39276 sidetracks, and web page maintenance.
39278 * Kean Johnston for SCO OpenServer support and various fixes.
39280 * Tim Josling for the sample language treelang based originally on
39281 Richard Kenner's "toy" language.
39283 * Nicolai Josuttis for additional libstdc++ documentation.
39285 * Klaus Kaempf for his ongoing work to make alpha-vms a viable
39288 * Steven G. Kargl for work on GNU Fortran.
39290 * David Kashtan of SRI adapted GCC to VMS.
39292 * Ryszard Kabatek for many, many libstdc++ bug fixes and
39293 optimizations of strings, especially member functions, and for
39296 * Geoffrey Keating for his ongoing work to make the PPC work for
39297 GNU/Linux and his automatic regression tester.
39299 * Brendan Kehoe for his ongoing work with G++ and for a lot of early
39300 work in just about every part of libstdc++.
39302 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the
39305 * Richard Kenner of the New York University Ultracomputer Research
39306 Laboratory wrote the machine descriptions for the AMD 29000, the
39307 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the
39308 support for instruction attributes. He also made changes to
39309 better support RISC processors including changes to common
39310 subexpression elimination, strength reduction, function calling
39311 sequence handling, and condition code support, in addition to
39312 generalizing the code for frame pointer elimination and delay slot
39313 scheduling. Richard Kenner was also the head maintainer of GCC
39316 * Mumit Khan for various contributions to the Cygwin and Mingw32
39317 ports and maintaining binary releases for Microsoft Windows hosts,
39318 and for massive libstdc++ porting work to Cygwin/Mingw32.
39320 * Robin Kirkham for cpu32 support.
39322 * Mark Klein for PA improvements.
39324 * Thomas Koenig for various bug fixes.
39326 * Bruce Korb for the new and improved fixincludes code.
39328 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3
39331 * Charles LaBrec contributed the support for the Integrated Solutions
39334 * Asher Langton and Mike Kumbera for contributing Cray pointer
39335 support to GNU Fortran, and for other GNU Fortran improvements.
39337 * Jeff Law for his direction via the steering committee,
39338 coordinating the entire egcs project and GCC 2.95, rolling out
39339 snapshots and releases, handling merges from GCC2, reviewing tons
39340 of patches that might have fallen through the cracks else, and
39341 random but extensive hacking.
39343 * Marc Lehmann for his direction via the steering committee and
39344 helping with analysis and improvements of x86 performance.
39346 * Victor Leikehman for work on GNU Fortran.
39348 * Ted Lemon wrote parts of the RTL reader and printer.
39350 * Kriang Lerdsuwanakij for C++ improvements including template as
39351 template parameter support, and many C++ fixes.
39353 * Warren Levy for tremendous work on libgcj (Java Runtime Library)
39354 and random work on the Java front end.
39356 * Alain Lichnewsky ported GCC to the MIPS CPU.
39358 * Oskar Liljeblad for hacking on AWT and his many Java bug reports
39361 * Robert Lipe for OpenServer support, new testsuites, testing, etc.
39363 * Chen Liqin for various S+core related fixes/improvement, and for
39364 maintaining the S+core port.
39366 * Weiwen Liu for testing and various bug fixes.
39368 * Manuel Lo'pez-Iba'n~ez for improving `-Wconversion' and many other
39369 diagnostics fixes and improvements.
39371 * Dave Love for his ongoing work with the Fortran front end and
39374 * Martin von Lo"wis for internal consistency checking infrastructure,
39375 various C++ improvements including namespace support, and tons of
39376 assistance with libstdc++/compiler merges.
39378 * H.J. Lu for his previous contributions to the steering committee,
39379 many x86 bug reports, prototype patches, and keeping the GNU/Linux
39382 * Greg McGary for random fixes and (someday) bounded pointers.
39384 * Andrew MacLeod for his ongoing work in building a real EH system,
39385 various code generation improvements, work on the global
39388 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC
39389 hacking improvements to compile-time performance, overall
39390 knowledge and direction in the area of instruction scheduling, and
39391 design and implementation of the automaton based instruction
39394 * Bob Manson for his behind the scenes work on dejagnu.
39396 * Philip Martin for lots of libstdc++ string and vector iterator
39397 fixes and improvements, and string clean up and testsuites.
39399 * All of the Mauve project contributors, for Java test code.
39401 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements.
39403 * Adam Megacz for his work on the Microsoft Windows port of GCJ.
39405 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS,
39406 powerpc, haifa, ECOFF debug support, and other assorted hacking.
39408 * Jason Merrill for his direction via the steering committee and
39409 leading the G++ effort.
39411 * Martin Michlmayr for testing GCC on several architectures using the
39412 entire Debian archive.
39414 * David Miller for his direction via the steering committee, lots of
39415 SPARC work, improvements in jump.c and interfacing with the Linux
39418 * Gary Miller ported GCC to Charles River Data Systems machines.
39420 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning
39421 the entire libstdc++ testsuite namespace-compatible.
39423 * Mark Mitchell for his direction via the steering committee,
39424 mountains of C++ work, load/store hoisting out of loops, alias
39425 analysis improvements, ISO C `restrict' support, and serving as
39426 release manager for GCC 3.x.
39428 * Alan Modra for various GNU/Linux bits and testing.
39430 * Toon Moene for his direction via the steering committee, Fortran
39431 maintenance, and his ongoing work to make us make Fortran run fast.
39433 * Jason Molenda for major help in the care and feeding of all the
39434 services on the gcc.gnu.org (formerly egcs.cygnus.com)
39435 machine--mail, web services, ftp services, etc etc. Doing all
39436 this work on scrap paper and the backs of envelopes would have
39439 * Catherine Moore for fixing various ugly problems we have sent her
39440 way, including the haifa bug which was killing the Alpha & PowerPC
39443 * Mike Moreton for his various Java patches.
39445 * David Mosberger-Tang for various Alpha improvements, and for the
39446 initial IA-64 port.
39448 * Stephen Moshier contributed the floating point emulator that
39449 assists in cross-compilation and permits support for floating
39450 point numbers wider than 64 bits and for ISO C99 support.
39452 * Bill Moyer for his behind the scenes work on various issues.
39454 * Philippe De Muyter for his work on the m68k port.
39456 * Joseph S. Myers for his work on the PDP-11 port, format checking
39457 and ISO C99 support, and continuous emphasis on (and contributions
39460 * Nathan Myers for his work on libstdc++-v3: architecture and
39461 authorship through the first three snapshots, including
39462 implementation of locale infrastructure, string, shadow C headers,
39463 and the initial project documentation (DESIGN, CHECKLIST, and so
39464 forth). Later, more work on MT-safe string and shadow headers.
39466 * Felix Natter for documentation on porting libstdc++.
39468 * Nathanael Nerode for cleaning up the configuration/build process.
39470 * NeXT, Inc. donated the front end that supports the Objective-C
39473 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to
39474 the search engine setup, various documentation fixes and other
39477 * Geoff Noer for his work on getting cygwin native builds working.
39479 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance
39480 tracking web pages, GIMPLE tuples, and assorted fixes.
39482 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64,
39483 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and
39484 related infrastructure improvements.
39486 * Alexandre Oliva for various build infrastructure improvements,
39487 scripts and amazing testing work, including keeping libtool issues
39490 * Stefan Olsson for work on mt_alloc.
39492 * Melissa O'Neill for various NeXT fixes.
39494 * Rainer Orth for random MIPS work, including improvements to GCC's
39495 o32 ABI support, improvements to dejagnu's MIPS support, Java
39496 configuration clean-ups and porting work, etc.
39498 * Hartmut Penner for work on the s390 port.
39500 * Paul Petersen wrote the machine description for the Alliant FX/8.
39502 * Alexandre Petit-Bianco for implementing much of the Java compiler
39503 and continued Java maintainership.
39505 * Matthias Pfaller for major improvements to the NS32k port.
39507 * Gerald Pfeifer for his direction via the steering committee,
39508 pointing out lots of problems we need to solve, maintenance of the
39509 web pages, and taking care of documentation maintenance in general.
39511 * Andrew Pinski for processing bug reports by the dozen.
39513 * Ovidiu Predescu for his work on the Objective-C front end and
39516 * Jerry Quinn for major performance improvements in C++ formatted
39519 * Ken Raeburn for various improvements to checker, MIPS ports and
39520 various cleanups in the compiler.
39522 * Rolf W. Rasmussen for hacking on AWT.
39524 * David Reese of Sun Microsystems contributed to the Solaris on
39527 * Volker Reichelt for keeping up with the problem reports.
39529 * Joern Rennecke for maintaining the sh port, loop, regmove & reload
39532 * Loren J. Rittle for improvements to libstdc++-v3 including the
39533 FreeBSD port, threading fixes, thread-related configury changes,
39534 critical threading documentation, and solutions to really tricky
39535 I/O problems, as well as keeping GCC properly working on FreeBSD
39536 and continuous testing.
39538 * Craig Rodrigues for processing tons of bug reports.
39540 * Ola Ro"nnerup for work on mt_alloc.
39542 * Gavin Romig-Koch for lots of behind the scenes MIPS work.
39544 * David Ronis inspired and encouraged Craig to rewrite the G77
39545 documentation in texinfo format by contributing a first pass at a
39546 translation of the old `g77-0.5.16/f/DOC' file.
39548 * Ken Rose for fixes to GCC's delay slot filling code.
39550 * Paul Rubin wrote most of the preprocessor.
39552 * Pe'tur Runo'lfsson for major performance improvements in C++
39553 formatted I/O and large file support in C++ filebuf.
39555 * Chip Salzenberg for libstdc++ patches and improvements to locales,
39556 traits, Makefiles, libio, libtool hackery, and "long long" support.
39558 * Juha Sarlin for improvements to the H8 code generator.
39560 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series
39563 * Roger Sayle for improvements to constant folding and GCC's RTL
39564 optimizers as well as for fixing numerous bugs.
39566 * Bradley Schatz for his work on the GCJ FAQ.
39568 * Peter Schauer wrote the code to allow debugging to work on the
39571 * William Schelter did most of the work on the Intel 80386 support.
39573 * Tobias Schlu"ter for work on GNU Fortran.
39575 * Bernd Schmidt for various code generation improvements and major
39576 work in the reload pass as well a serving as release manager for
39579 * Peter Schmid for constant testing of libstdc++--especially
39580 application testing, going above and beyond what was requested for
39581 the release criteria--and libstdc++ header file tweaks.
39583 * Jason Schroeder for jcf-dump patches.
39585 * Andreas Schwab for his work on the m68k port.
39587 * Lars Segerlund for work on GNU Fortran.
39589 * Joel Sherrill for his direction via the steering committee, RTEMS
39590 contributions and RTEMS testing.
39592 * Nathan Sidwell for many C++ fixes/improvements.
39594 * Jeffrey Siegal for helping RMS with the original design of GCC,
39595 some code which handles the parse tree and RTL data structures,
39596 constant folding and help with the original VAX & m68k ports.
39598 * Kenny Simpson for prompting libstdc++ fixes due to defect reports
39599 from the LWG (thereby keeping GCC in line with updates from the
39602 * Franz Sirl for his ongoing work with making the PPC port stable
39605 * Andrey Slepuhin for assorted AIX hacking.
39607 * Trevor Smigiel for contributing the SPU port.
39609 * Christopher Smith did the port for Convex machines.
39611 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports.
39613 * Randy Smith finished the Sun FPA support.
39615 * Scott Snyder for queue, iterator, istream, and string fixes and
39616 libstdc++ testsuite entries. Also for providing the patch to G77
39617 to add rudimentary support for `INTEGER*1', `INTEGER*2', and
39620 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique.
39622 * Richard Stallman, for writing the original GCC and launching the
39625 * Jan Stein of the Chalmers Computer Society provided support for
39626 Genix, as well as part of the 32000 machine description.
39628 * Nigel Stephens for various mips16 related fixes/improvements.
39630 * Jonathan Stone wrote the machine description for the Pyramid
39633 * Graham Stott for various infrastructure improvements.
39635 * John Stracke for his Java HTTP protocol fixes.
39637 * Mike Stump for his Elxsi port, G++ contributions over the years
39638 and more recently his vxworks contributions
39640 * Jeff Sturm for Java porting help, bug fixes, and encouragement.
39642 * Shigeya Suzuki for this fixes for the bsdi platforms.
39644 * Ian Lance Taylor for his mips16 work, general configury hacking,
39647 * Holger Teutsch provided the support for the Clipper CPU.
39649 * Gary Thomas for his ongoing work to make the PPC work for
39652 * Philipp Thomas for random bug fixes throughout the compiler
39654 * Jason Thorpe for thread support in libstdc++ on NetBSD.
39656 * Kresten Krab Thorup wrote the run time support for the Objective-C
39657 language and the fantastic Java bytecode interpreter.
39659 * Michael Tiemann for random bug fixes, the first instruction
39660 scheduler, initial C++ support, function integration, NS32k, SPARC
39661 and M88k machine description work, delay slot scheduling.
39663 * Andreas Tobler for his work porting libgcj to Darwin.
39665 * Teemu Torma for thread safe exception handling support.
39667 * Leonard Tower wrote parts of the parser, RTL generator, and RTL
39668 definitions, and of the VAX machine description.
39670 * Daniel Towner and Hariharan Sandanagobalane contributed and
39671 maintain the picoChip port.
39673 * Tom Tromey for internationalization support and for his many Java
39674 contributions and libgcj maintainership.
39676 * Lassi Tuura for improvements to config.guess to determine HP
39679 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes.
39681 * Andy Vaught for the design and initial implementation of the GNU
39684 * Brent Verner for work with the libstdc++ cshadow files and their
39685 associated configure steps.
39687 * Todd Vierling for contributions for NetBSD ports.
39689 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML
39692 * Dean Wakerley for converting the install documentation from HTML
39693 to texinfo in time for GCC 3.0.
39695 * Krister Walfridsson for random bug fixes.
39697 * Feng Wang for contributions to GNU Fortran.
39699 * Stephen M. Webb for time and effort on making libstdc++ shadow
39700 files work with the tricky Solaris 8+ headers, and for pushing the
39701 build-time header tree.
39703 * John Wehle for various improvements for the x86 code generator,
39704 related infrastructure improvements to help x86 code generation,
39705 value range propagation and other work, WE32k port.
39707 * Ulrich Weigand for work on the s390 port.
39709 * Zack Weinberg for major work on cpplib and various other bug fixes.
39711 * Matt Welsh for help with Linux Threads support in GCJ.
39713 * Urban Widmark for help fixing java.io.
39715 * Mark Wielaard for new Java library code and his work integrating
39718 * Dale Wiles helped port GCC to the Tahoe.
39720 * Bob Wilson from Tensilica, Inc. for the Xtensa port.
39722 * Jim Wilson for his direction via the steering committee, tackling
39723 hard problems in various places that nobody else wanted to work
39724 on, strength reduction and other loop optimizations.
39726 * Paul Woegerer and Tal Agmon for the CRX port.
39728 * Carlo Wood for various fixes.
39730 * Tom Wood for work on the m88k port.
39732 * Canqun Yang for work on GNU Fortran.
39734 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine
39735 description for the Tron architecture (specifically, the Gmicro).
39737 * Kevin Zachmann helped port GCC to the Tahoe.
39739 * Ayal Zaks for Swing Modulo Scheduling (SMS).
39741 * Xiaoqiang Zhang for work on GNU Fortran.
39743 * Gilles Zunino for help porting Java to Irix.
39746 The following people are recognized for their contributions to GNAT,
39747 the Ada front end of GCC:
39750 * Romain Berrendonner
39800 * Hristian Kirtchev
39843 The following people are recognized for their contributions of new
39844 features, bug reports, testing and integration of classpath/libgcj for
39846 * Lillian Angel for `JTree' implementation and lots Free Swing
39847 additions and bug fixes.
39849 * Wolfgang Baer for `GapContent' bug fixes.
39851 * Anthony Balkissoon for `JList', Free Swing 1.5 updates and mouse
39852 event fixes, lots of Free Swing work including `JTable' editing.
39854 * Stuart Ballard for RMI constant fixes.
39856 * Goffredo Baroncelli for `HTTPURLConnection' fixes.
39858 * Gary Benson for `MessageFormat' fixes.
39860 * Daniel Bonniot for `Serialization' fixes.
39862 * Chris Burdess for lots of gnu.xml and http protocol fixes, `StAX'
39863 and `DOM xml:id' support.
39865 * Ka-Hing Cheung for `TreePath' and `TreeSelection' fixes.
39867 * Archie Cobbs for build fixes, VM interface updates,
39868 `URLClassLoader' updates.
39870 * Kelley Cook for build fixes.
39872 * Martin Cordova for Suggestions for better `SocketTimeoutException'.
39874 * David Daney for `BitSet' bug fixes, `HttpURLConnection' rewrite
39877 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo
39878 2D support. Lots of imageio framework additions, lots of AWT and
39879 Free Swing bug fixes.
39881 * Jeroen Frijters for `ClassLoader' and nio cleanups, serialization
39882 fixes, better `Proxy' support, bug fixes and IKVM integration.
39884 * Santiago Gala for `AccessControlContext' fixes.
39886 * Nicolas Geoffray for `VMClassLoader' and `AccessController'
39889 * David Gilbert for `basic' and `metal' icon and plaf support and
39890 lots of documenting, Lots of Free Swing and metal theme additions.
39891 `MetalIconFactory' implementation.
39893 * Anthony Green for `MIDI' framework, `ALSA' and `DSSI' providers.
39895 * Andrew Haley for `Serialization' and `URLClassLoader' fixes, gcj
39898 * Kim Ho for `JFileChooser' implementation.
39900 * Andrew John Hughes for `Locale' and net fixes, URI RFC2986
39901 updates, `Serialization' fixes, `Properties' XML support and
39902 generic branch work, VMIntegration guide update.
39904 * Bastiaan Huisman for `TimeZone' bug fixing.
39906 * Andreas Jaeger for mprec updates.
39908 * Paul Jenner for better `-Werror' support.
39910 * Ito Kazumitsu for `NetworkInterface' implementation and updates.
39912 * Roman Kennke for `BoxLayout', `GrayFilter' and `SplitPane', plus
39913 bug fixes all over. Lots of Free Swing work including styled text.
39915 * Simon Kitching for `String' cleanups and optimization suggestions.
39917 * Michael Koch for configuration fixes, `Locale' updates, bug and
39920 * Guilhem Lavaux for configuration, thread and channel fixes and
39921 Kaffe integration. JCL native `Pointer' updates. Logger bug fixes.
39923 * David Lichteblau for JCL support library global/local reference
39926 * Aaron Luchko for JDWP updates and documentation fixes.
39928 * Ziga Mahkovec for `Graphics2D' upgraded to Cairo 0.5 and new regex
39931 * Sven de Marothy for BMP imageio support, CSS and `TextLayout'
39932 fixes. `GtkImage' rewrite, 2D, awt, free swing and date/time fixes
39933 and implementing the Qt4 peers.
39935 * Casey Marshall for crypto algorithm fixes, `FileChannel' lock,
39936 `SystemLogger' and `FileHandler' rotate implementations, NIO
39937 `FileChannel.map' support, security and policy updates.
39939 * Bryce McKinlay for RMI work.
39941 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus
39942 testing and documenting.
39944 * Kalle Olavi Niemitalo for build fixes.
39946 * Rainer Orth for build fixes.
39948 * Andrew Overholt for `File' locking fixes.
39950 * Ingo Proetel for `Image', `Logger' and `URLClassLoader' updates.
39952 * Olga Rodimina for `MenuSelectionManager' implementation.
39954 * Jan Roehrich for `BasicTreeUI' and `JTree' fixes.
39956 * Julian Scheid for documentation updates and gjdoc support.
39958 * Christian Schlichtherle for zip fixes and cleanups.
39960 * Robert Schuster for documentation updates and beans fixes,
39961 `TreeNode' enumerations and `ActionCommand' and various fixes, XML
39962 and URL, AWT and Free Swing bug fixes.
39964 * Keith Seitz for lots of JDWP work.
39966 * Christian Thalinger for 64-bit cleanups, Configuration and VM
39967 interface fixes and `CACAO' integration, `fdlibm' updates.
39969 * Gael Thomas for `VMClassLoader' boot packages support suggestions.
39971 * Andreas Tobler for Darwin and Solaris testing and fixing, `Qt4'
39972 support for Darwin/OS X, `Graphics2D' support, `gtk+' updates.
39974 * Dalibor Topic for better `DEBUG' support, build cleanups and Kaffe
39975 integration. `Qt4' build infrastructure, `SHA1PRNG' and
39976 `GdkPixbugDecoder' updates.
39978 * Tom Tromey for Eclipse integration, generics work, lots of bug
39979 fixes and gcj integration including coordinating The Big Merge.
39981 * Mark Wielaard for bug fixes, packaging and release management,
39982 `Clipboard' implementation, system call interrupts and network
39983 timeouts and `GdkPixpufDecoder' fixes.
39986 In addition to the above, all of which also contributed time and
39987 energy in testing GCC, we would like to thank the following for their
39988 contributions to testing:
39990 * Michael Abd-El-Malek
40000 * David Billinghurst
40004 * Stephane Bortzmeyer
40014 * Bradford Castalia
40036 * Charles-Antoine Gauthier
40058 * Kevin B. Hendricks
40062 * Christian Joensson
40070 * Anand Krishnaswamy
40072 * A. O. V. Le Blanc
40136 * Pedro A. M. Vazquez
40146 And finally we'd like to thank everyone who uses the compiler, provides
40147 feedback and generally reminds us why we're doing this work in the first
40151 File: gcc.info, Node: Option Index, Next: Keyword Index, Prev: Contributors, Up: Top
40156 GCC's command line options are indexed here without any initial `-' or
40157 `--'. Where an option has both positive and negative forms (such as
40158 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are
40159 indexed under the most appropriate form; it may sometimes be useful to
40160 look up both forms.
40165 * ###: Overall Options. (line 204)
40166 * -fdump-statistics: Debugging Options. (line 623)
40167 * A: Preprocessor Options.
40169 * all_load: Darwin Options. (line 112)
40170 * allowable_client: Darwin Options. (line 199)
40171 * ansi <1>: Non-bugs. (line 107)
40172 * ansi <2>: Other Builtins. (line 22)
40173 * ansi <3>: Preprocessor Options.
40175 * ansi <4>: C Dialect Options. (line 11)
40176 * ansi: Standards. (line 16)
40177 * arch_errors_fatal: Darwin Options. (line 116)
40178 * aux-info: C Dialect Options. (line 140)
40179 * b: Target Options. (line 13)
40180 * B: Directory Options. (line 41)
40181 * bcopy-builtin: PDP-11 Options. (line 32)
40182 * Bdynamic: VxWorks Options. (line 22)
40183 * bind_at_load: Darwin Options. (line 120)
40184 * Bstatic: VxWorks Options. (line 22)
40185 * bundle: Darwin Options. (line 125)
40186 * bundle_loader: Darwin Options. (line 129)
40187 * c: Link Options. (line 20)
40188 * C: Preprocessor Options.
40190 * c: Overall Options. (line 159)
40191 * client_name: Darwin Options. (line 199)
40192 * combine: Overall Options. (line 215)
40193 * compatibility_version: Darwin Options. (line 199)
40194 * coverage: Debugging Options. (line 272)
40195 * current_version: Darwin Options. (line 199)
40196 * D: Preprocessor Options.
40198 * d: Debugging Options. (line 336)
40199 * dA: Debugging Options. (line 539)
40200 * dD <1>: Preprocessor Options.
40202 * dD: Debugging Options. (line 543)
40203 * dead_strip: Darwin Options. (line 199)
40204 * dependency-file: Darwin Options. (line 199)
40205 * dH: Debugging Options. (line 547)
40206 * dI: Preprocessor Options.
40208 * dM: Preprocessor Options.
40210 * dm: Debugging Options. (line 550)
40211 * dN: Preprocessor Options.
40213 * dP: Debugging Options. (line 559)
40214 * dp: Debugging Options. (line 554)
40215 * dU: Preprocessor Options.
40217 * dumpmachine: Debugging Options. (line 952)
40218 * dumpspecs: Debugging Options. (line 960)
40219 * dumpversion: Debugging Options. (line 956)
40220 * dv: Debugging Options. (line 563)
40221 * dx: Debugging Options. (line 568)
40222 * dy: Debugging Options. (line 572)
40223 * dylib_file: Darwin Options. (line 199)
40224 * dylinker_install_name: Darwin Options. (line 199)
40225 * dynamic: Darwin Options. (line 199)
40226 * dynamiclib: Darwin Options. (line 133)
40227 * E <1>: Link Options. (line 20)
40228 * E: Overall Options. (line 180)
40229 * EB <1>: MIPS Options. (line 7)
40230 * EB: ARC Options. (line 12)
40231 * EL <1>: MIPS Options. (line 10)
40232 * EL: ARC Options. (line 9)
40233 * exported_symbols_list: Darwin Options. (line 199)
40234 * F: Darwin Options. (line 32)
40235 * fabi-version: C++ Dialect Options.
40237 * falign-functions: Optimize Options. (line 1184)
40238 * falign-jumps: Optimize Options. (line 1234)
40239 * falign-labels: Optimize Options. (line 1202)
40240 * falign-loops: Optimize Options. (line 1220)
40241 * fargument-alias: Code Gen Options. (line 413)
40242 * fargument-noalias: Code Gen Options. (line 413)
40243 * fargument-noalias-anything: Code Gen Options. (line 413)
40244 * fargument-noalias-global: Code Gen Options. (line 413)
40245 * fassociative-math: Optimize Options. (line 1411)
40246 * fasynchronous-unwind-tables: Code Gen Options. (line 64)
40247 * fauto-inc-dec: Optimize Options. (line 455)
40248 * fbounds-check: Code Gen Options. (line 15)
40249 * fbranch-probabilities: Optimize Options. (line 1544)
40250 * fbranch-target-load-optimize: Optimize Options. (line 1652)
40251 * fbranch-target-load-optimize2: Optimize Options. (line 1658)
40252 * fbtr-bb-exclusive: Optimize Options. (line 1662)
40253 * fcall-saved: Code Gen Options. (line 262)
40254 * fcall-used: Code Gen Options. (line 248)
40255 * fcaller-saves: Optimize Options. (line 676)
40256 * fcheck-data-deps: Optimize Options. (line 897)
40257 * fcheck-new: C++ Dialect Options.
40259 * fcommon: Variable Attributes.
40261 * fcond-mismatch: C Dialect Options. (line 258)
40262 * fconserve-space: C++ Dialect Options.
40264 * fconserve-stack: Optimize Options. (line 689)
40265 * fconstant-string-class: Objective-C and Objective-C++ Dialect Options.
40267 * fcprop-registers: Optimize Options. (line 1292)
40268 * fcrossjumping: Optimize Options. (line 448)
40269 * fcse-follow-jumps: Optimize Options. (line 376)
40270 * fcse-skip-blocks: Optimize Options. (line 385)
40271 * fcx-fortran-rules: Optimize Options. (line 1530)
40272 * fcx-limited-range: Optimize Options. (line 1518)
40273 * fdata-sections: Optimize Options. (line 1633)
40274 * fdbg-cnt: Debugging Options. (line 325)
40275 * fdbg-cnt-list: Debugging Options. (line 322)
40276 * fdce: Optimize Options. (line 461)
40277 * fdebug-prefix-map: Debugging Options. (line 219)
40278 * fdelayed-branch: Optimize Options. (line 557)
40279 * fdelete-null-pointer-checks: Optimize Options. (line 484)
40280 * fdiagnostics-show-location: Language Independent Options.
40282 * fdiagnostics-show-option: Language Independent Options.
40284 * fdirectives-only: Preprocessor Options.
40286 * fdollars-in-identifiers <1>: Interoperation. (line 146)
40287 * fdollars-in-identifiers: Preprocessor Options.
40289 * fdse: Optimize Options. (line 465)
40290 * fdump-class-hierarchy: Debugging Options. (line 597)
40291 * fdump-ipa: Debugging Options. (line 605)
40292 * fdump-noaddr: Debugging Options. (line 575)
40293 * fdump-rtl-alignments: Debugging Options. (line 351)
40294 * fdump-rtl-all: Debugging Options. (line 536)
40295 * fdump-rtl-asmcons: Debugging Options. (line 354)
40296 * fdump-rtl-auto_inc_dec: Debugging Options. (line 358)
40297 * fdump-rtl-barriers: Debugging Options. (line 362)
40298 * fdump-rtl-bbpart: Debugging Options. (line 365)
40299 * fdump-rtl-bbro: Debugging Options. (line 368)
40300 * fdump-rtl-btl2: Debugging Options. (line 372)
40301 * fdump-rtl-bypass: Debugging Options. (line 376)
40302 * fdump-rtl-ce1: Debugging Options. (line 387)
40303 * fdump-rtl-ce2: Debugging Options. (line 387)
40304 * fdump-rtl-ce3: Debugging Options. (line 387)
40305 * fdump-rtl-combine: Debugging Options. (line 379)
40306 * fdump-rtl-compgotos: Debugging Options. (line 382)
40307 * fdump-rtl-cprop_hardreg: Debugging Options. (line 391)
40308 * fdump-rtl-csa: Debugging Options. (line 394)
40309 * fdump-rtl-cse1: Debugging Options. (line 398)
40310 * fdump-rtl-cse2: Debugging Options. (line 398)
40311 * fdump-rtl-dbr: Debugging Options. (line 405)
40312 * fdump-rtl-dce: Debugging Options. (line 402)
40313 * fdump-rtl-dce1: Debugging Options. (line 409)
40314 * fdump-rtl-dce2: Debugging Options. (line 409)
40315 * fdump-rtl-dfinish: Debugging Options. (line 533)
40316 * fdump-rtl-dfinit: Debugging Options. (line 533)
40317 * fdump-rtl-eh: Debugging Options. (line 413)
40318 * fdump-rtl-eh_ranges: Debugging Options. (line 416)
40319 * fdump-rtl-expand: Debugging Options. (line 419)
40320 * fdump-rtl-fwprop1: Debugging Options. (line 423)
40321 * fdump-rtl-fwprop2: Debugging Options. (line 423)
40322 * fdump-rtl-gcse1: Debugging Options. (line 428)
40323 * fdump-rtl-gcse2: Debugging Options. (line 428)
40324 * fdump-rtl-init-regs: Debugging Options. (line 432)
40325 * fdump-rtl-initvals: Debugging Options. (line 435)
40326 * fdump-rtl-into_cfglayout: Debugging Options. (line 438)
40327 * fdump-rtl-ira: Debugging Options. (line 441)
40328 * fdump-rtl-jump: Debugging Options. (line 444)
40329 * fdump-rtl-loop2: Debugging Options. (line 447)
40330 * fdump-rtl-mach: Debugging Options. (line 451)
40331 * fdump-rtl-mode_sw: Debugging Options. (line 455)
40332 * fdump-rtl-outof_cfglayout: Debugging Options. (line 461)
40333 * fdump-rtl-peephole2: Debugging Options. (line 464)
40334 * fdump-rtl-postreload: Debugging Options. (line 467)
40335 * fdump-rtl-pro_and_epilogue: Debugging Options. (line 470)
40336 * fdump-rtl-regclass: Debugging Options. (line 533)
40337 * fdump-rtl-regmove: Debugging Options. (line 473)
40338 * fdump-rtl-rnreg: Debugging Options. (line 458)
40339 * fdump-rtl-sched1: Debugging Options. (line 477)
40340 * fdump-rtl-sched2: Debugging Options. (line 477)
40341 * fdump-rtl-see: Debugging Options. (line 481)
40342 * fdump-rtl-seqabstr: Debugging Options. (line 484)
40343 * fdump-rtl-shorten: Debugging Options. (line 487)
40344 * fdump-rtl-sibling: Debugging Options. (line 490)
40345 * fdump-rtl-sms: Debugging Options. (line 503)
40346 * fdump-rtl-split1: Debugging Options. (line 497)
40347 * fdump-rtl-split2: Debugging Options. (line 497)
40348 * fdump-rtl-split3: Debugging Options. (line 497)
40349 * fdump-rtl-split4: Debugging Options. (line 497)
40350 * fdump-rtl-split5: Debugging Options. (line 497)
40351 * fdump-rtl-stack: Debugging Options. (line 507)
40352 * fdump-rtl-subreg1: Debugging Options. (line 513)
40353 * fdump-rtl-subreg2: Debugging Options. (line 513)
40354 * fdump-rtl-subregs_of_mode_finish: Debugging Options. (line 533)
40355 * fdump-rtl-subregs_of_mode_init: Debugging Options. (line 533)
40356 * fdump-rtl-unshare: Debugging Options. (line 517)
40357 * fdump-rtl-vartrack: Debugging Options. (line 520)
40358 * fdump-rtl-vregs: Debugging Options. (line 523)
40359 * fdump-rtl-web: Debugging Options. (line 526)
40360 * fdump-translation-unit: Debugging Options. (line 588)
40361 * fdump-tree: Debugging Options. (line 634)
40362 * fdump-tree-alias: Debugging Options. (line 719)
40363 * fdump-tree-all: Debugging Options. (line 804)
40364 * fdump-tree-ccp: Debugging Options. (line 723)
40365 * fdump-tree-cfg: Debugging Options. (line 699)
40366 * fdump-tree-ch: Debugging Options. (line 711)
40367 * fdump-tree-copyprop: Debugging Options. (line 739)
40368 * fdump-tree-copyrename: Debugging Options. (line 785)
40369 * fdump-tree-dce: Debugging Options. (line 747)
40370 * fdump-tree-dom: Debugging Options. (line 765)
40371 * fdump-tree-dse: Debugging Options. (line 770)
40372 * fdump-tree-forwprop: Debugging Options. (line 780)
40373 * fdump-tree-fre: Debugging Options. (line 735)
40374 * fdump-tree-gimple: Debugging Options. (line 694)
40375 * fdump-tree-mudflap: Debugging Options. (line 751)
40376 * fdump-tree-nrv: Debugging Options. (line 790)
40377 * fdump-tree-phiopt: Debugging Options. (line 775)
40378 * fdump-tree-pre: Debugging Options. (line 731)
40379 * fdump-tree-sink: Debugging Options. (line 761)
40380 * fdump-tree-sra: Debugging Options. (line 756)
40381 * fdump-tree-ssa: Debugging Options. (line 715)
40382 * fdump-tree-store_copyprop: Debugging Options. (line 743)
40383 * fdump-tree-storeccp: Debugging Options. (line 727)
40384 * fdump-tree-vcg: Debugging Options. (line 703)
40385 * fdump-tree-vect: Debugging Options. (line 795)
40386 * fdump-tree-vrp: Debugging Options. (line 800)
40387 * fdump-unnumbered: Debugging Options. (line 581)
40388 * fdwarf2-cfi-asm: Debugging Options. (line 223)
40389 * fearly-inlining: Optimize Options. (line 220)
40390 * feliminate-dwarf2-dups: Debugging Options. (line 136)
40391 * feliminate-unused-debug-symbols: Debugging Options. (line 52)
40392 * feliminate-unused-debug-types: Debugging Options. (line 964)
40393 * fexceptions: Code Gen Options. (line 34)
40394 * fexec-charset: Preprocessor Options.
40396 * fexpensive-optimizations: Optimize Options. (line 497)
40397 * fextended-identifiers: Preprocessor Options.
40399 * ffast-math: Optimize Options. (line 1362)
40400 * ffinite-math-only: Optimize Options. (line 1435)
40401 * ffix-and-continue: Darwin Options. (line 106)
40402 * ffixed: Code Gen Options. (line 236)
40403 * ffloat-store <1>: Disappointments. (line 77)
40404 * ffloat-store: Optimize Options. (line 1348)
40405 * ffor-scope: C++ Dialect Options.
40407 * fforward-propagate: Optimize Options. (line 149)
40408 * ffreestanding <1>: Function Attributes.
40410 * ffreestanding <2>: Warning Options. (line 194)
40411 * ffreestanding <3>: C Dialect Options. (line 211)
40412 * ffreestanding: Standards. (line 84)
40413 * ffriend-injection: C++ Dialect Options.
40415 * ffunction-sections: Optimize Options. (line 1633)
40416 * fgcse: Optimize Options. (line 399)
40417 * fgcse-after-reload: Optimize Options. (line 435)
40418 * fgcse-las: Optimize Options. (line 428)
40419 * fgcse-lm: Optimize Options. (line 410)
40420 * fgcse-sm: Optimize Options. (line 419)
40421 * fgnu-runtime: Objective-C and Objective-C++ Dialect Options.
40423 * fgnu89-inline: C Dialect Options. (line 120)
40424 * fhosted: C Dialect Options. (line 204)
40425 * fif-conversion: Optimize Options. (line 469)
40426 * fif-conversion2: Optimize Options. (line 478)
40427 * filelist: Darwin Options. (line 199)
40428 * findirect-data: Darwin Options. (line 106)
40429 * findirect-inlining: Optimize Options. (line 193)
40430 * finhibit-size-directive: Code Gen Options. (line 158)
40431 * finline-functions: Optimize Options. (line 201)
40432 * finline-functions-called-once: Optimize Options. (line 212)
40433 * finline-limit: Optimize Options. (line 230)
40434 * finline-small-functions: Optimize Options. (line 185)
40435 * finput-charset: Preprocessor Options.
40437 * finstrument-functions <1>: Function Attributes.
40439 * finstrument-functions: Code Gen Options. (line 292)
40440 * finstrument-functions-exclude-file-list: Code Gen Options. (line 329)
40441 * finstrument-functions-exclude-function-list: Code Gen Options.
40443 * fipa-cp: Optimize Options. (line 742)
40444 * fipa-cp-clone: Optimize Options. (line 750)
40445 * fipa-matrix-reorg: Optimize Options. (line 760)
40446 * fipa-pta: Optimize Options. (line 738)
40447 * fipa-pure-const: Optimize Options. (line 715)
40448 * fipa-reference: Optimize Options. (line 719)
40449 * fipa-struct-reorg: Optimize Options. (line 723)
40450 * fira-coalesce: Optimize Options. (line 536)
40451 * fira-verbose: Optimize Options. (line 552)
40452 * fivopts: Optimize Options. (line 933)
40453 * fkeep-inline-functions <1>: Inline. (line 51)
40454 * fkeep-inline-functions: Optimize Options. (line 256)
40455 * fkeep-static-consts: Optimize Options. (line 263)
40456 * flat_namespace: Darwin Options. (line 199)
40457 * flax-vector-conversions: C Dialect Options. (line 263)
40458 * fleading-underscore: Code Gen Options. (line 430)
40459 * fmem-report: Debugging Options. (line 247)
40460 * fmerge-all-constants: Optimize Options. (line 282)
40461 * fmerge-constants: Optimize Options. (line 272)
40462 * fmerge-debug-strings: Debugging Options. (line 211)
40463 * fmessage-length: Language Independent Options.
40465 * fmodulo-sched: Optimize Options. (line 293)
40466 * fmodulo-sched-allow-regmoves: Optimize Options. (line 298)
40467 * fmove-loop-invariants: Optimize Options. (line 1623)
40468 * fms-extensions <1>: Unnamed Fields. (line 37)
40469 * fms-extensions <2>: C++ Dialect Options.
40471 * fms-extensions: C Dialect Options. (line 229)
40472 * fmudflap: Optimize Options. (line 338)
40473 * fmudflapir: Optimize Options. (line 338)
40474 * fmudflapth: Optimize Options. (line 338)
40475 * fnext-runtime: Objective-C and Objective-C++ Dialect Options.
40477 * fno-access-control: C++ Dialect Options.
40479 * fno-asm: C Dialect Options. (line 156)
40480 * fno-branch-count-reg: Optimize Options. (line 305)
40481 * fno-builtin <1>: Other Builtins. (line 14)
40482 * fno-builtin <2>: Function Attributes.
40484 * fno-builtin <3>: Warning Options. (line 194)
40485 * fno-builtin: C Dialect Options. (line 170)
40486 * fno-common <1>: Variable Attributes.
40488 * fno-common: Code Gen Options. (line 135)
40489 * fno-default-inline <1>: Inline. (line 71)
40490 * fno-default-inline <2>: Optimize Options. (line 134)
40491 * fno-default-inline: C++ Dialect Options.
40493 * fno-defer-pop: Optimize Options. (line 141)
40494 * fno-dwarf2-cfi-asm: Debugging Options. (line 223)
40495 * fno-elide-constructors: C++ Dialect Options.
40497 * fno-enforce-eh-specs: C++ Dialect Options.
40499 * fno-for-scope: C++ Dialect Options.
40501 * fno-function-cse: Optimize Options. (line 315)
40502 * fno-gnu-keywords: C++ Dialect Options.
40504 * fno-guess-branch-probability: Optimize Options. (line 1056)
40505 * fno-ident: Code Gen Options. (line 155)
40506 * fno-implement-inlines <1>: C++ Interface. (line 75)
40507 * fno-implement-inlines: C++ Dialect Options.
40509 * fno-implicit-inline-templates: C++ Dialect Options.
40511 * fno-implicit-templates <1>: Template Instantiation.
40513 * fno-implicit-templates: C++ Dialect Options.
40515 * fno-inline: Optimize Options. (line 179)
40516 * fno-ira-share-save-slots: Optimize Options. (line 540)
40517 * fno-ira-share-spill-slots: Optimize Options. (line 546)
40518 * fno-jump-tables: Code Gen Options. (line 228)
40519 * fno-math-errno: Optimize Options. (line 1376)
40520 * fno-merge-debug-strings: Debugging Options. (line 211)
40521 * fno-nil-receivers: Objective-C and Objective-C++ Dialect Options.
40523 * fno-nonansi-builtins: C++ Dialect Options.
40525 * fno-operator-names: C++ Dialect Options.
40527 * fno-optional-diags: C++ Dialect Options.
40529 * fno-peephole: Optimize Options. (line 1047)
40530 * fno-peephole2: Optimize Options. (line 1047)
40531 * fno-rtti: C++ Dialect Options.
40533 * fno-sched-interblock: Optimize Options. (line 583)
40534 * fno-sched-spec: Optimize Options. (line 588)
40535 * fno-show-column: Preprocessor Options.
40537 * fno-signed-bitfields: C Dialect Options. (line 296)
40538 * fno-signed-zeros: Optimize Options. (line 1447)
40539 * fno-stack-limit: Code Gen Options. (line 396)
40540 * fno-threadsafe-statics: C++ Dialect Options.
40542 * fno-toplevel-reorder: Optimize Options. (line 1254)
40543 * fno-trapping-math: Optimize Options. (line 1457)
40544 * fno-unsigned-bitfields: C Dialect Options. (line 296)
40545 * fno-use-cxa-get-exception-ptr: C++ Dialect Options.
40547 * fno-weak: C++ Dialect Options.
40549 * fno-working-directory: Preprocessor Options.
40551 * fno-zero-initialized-in-bss: Optimize Options. (line 326)
40552 * fnon-call-exceptions: Code Gen Options. (line 48)
40553 * fobjc-call-cxx-cdtors: Objective-C and Objective-C++ Dialect Options.
40555 * fobjc-direct-dispatch: Objective-C and Objective-C++ Dialect Options.
40557 * fobjc-exceptions: Objective-C and Objective-C++ Dialect Options.
40559 * fobjc-gc: Objective-C and Objective-C++ Dialect Options.
40561 * fomit-frame-pointer: Optimize Options. (line 158)
40562 * fopenmp: C Dialect Options. (line 221)
40563 * foptimize-register-move: Optimize Options. (line 504)
40564 * foptimize-sibling-calls: Optimize Options. (line 174)
40565 * force_cpusubtype_ALL: Darwin Options. (line 138)
40566 * force_flat_namespace: Darwin Options. (line 199)
40567 * fpack-struct: Code Gen Options. (line 279)
40568 * fpcc-struct-return <1>: Incompatibilities. (line 170)
40569 * fpcc-struct-return: Code Gen Options. (line 70)
40570 * fpch-deps: Preprocessor Options.
40572 * fpch-preprocess: Preprocessor Options.
40574 * fpeel-loops: Optimize Options. (line 1615)
40575 * fpermissive: C++ Dialect Options.
40577 * fPIC: Code Gen Options. (line 205)
40578 * fpic: Code Gen Options. (line 184)
40579 * fPIE: Code Gen Options. (line 218)
40580 * fpie: Code Gen Options. (line 218)
40581 * fpost-ipa-mem-report: Debugging Options. (line 253)
40582 * fpre-ipa-mem-report: Debugging Options. (line 251)
40583 * fpredictive-commoning: Optimize Options. (line 1029)
40584 * fprefetch-loop-arrays: Optimize Options. (line 1036)
40585 * fpreprocessed: Preprocessor Options.
40587 * fprofile-arcs <1>: Other Builtins. (line 242)
40588 * fprofile-arcs: Debugging Options. (line 257)
40589 * fprofile-correction: Optimize Options. (line 1299)
40590 * fprofile-dir: Optimize Options. (line 1306)
40591 * fprofile-generate: Optimize Options. (line 1316)
40592 * fprofile-use: Optimize Options. (line 1329)
40593 * fprofile-values: Optimize Options. (line 1563)
40594 * frandom-string: Debugging Options. (line 833)
40595 * freciprocal-math: Optimize Options. (line 1426)
40596 * frecord-gcc-switches: Code Gen Options. (line 174)
40597 * freg-struct-return: Code Gen Options. (line 88)
40598 * fregmove: Optimize Options. (line 504)
40599 * frename-registers: Optimize Options. (line 1582)
40600 * freorder-blocks: Optimize Options. (line 1073)
40601 * freorder-blocks-and-partition: Optimize Options. (line 1079)
40602 * freorder-functions: Optimize Options. (line 1090)
40603 * freplace-objc-classes: Objective-C and Objective-C++ Dialect Options.
40605 * frepo <1>: Template Instantiation.
40607 * frepo: C++ Dialect Options.
40609 * frerun-cse-after-loop: Optimize Options. (line 393)
40610 * freschedule-modulo-scheduled-loops: Optimize Options. (line 652)
40611 * frounding-math: Optimize Options. (line 1472)
40612 * frtl-abstract-sequences: Optimize Options. (line 1492)
40613 * fsched-spec-load: Optimize Options. (line 593)
40614 * fsched-spec-load-dangerous: Optimize Options. (line 598)
40615 * fsched-stalled-insns: Optimize Options. (line 604)
40616 * fsched-stalled-insns-dep: Optimize Options. (line 614)
40617 * fsched-verbose: Debugging Options. (line 843)
40618 * fsched2-use-superblocks: Optimize Options. (line 624)
40619 * fsched2-use-traces: Optimize Options. (line 635)
40620 * fschedule-insns: Optimize Options. (line 564)
40621 * fschedule-insns2: Optimize Options. (line 574)
40622 * fsection-anchors: Optimize Options. (line 1678)
40623 * fsee: Optimize Options. (line 647)
40624 * fsel-sched-pipelining: Optimize Options. (line 666)
40625 * fsel-sched-pipelining-outer-loops: Optimize Options. (line 671)
40626 * fselective-scheduling: Optimize Options. (line 658)
40627 * fselective-scheduling2: Optimize Options. (line 662)
40628 * fshort-double: Code Gen Options. (line 117)
40629 * fshort-enums <1>: Non-bugs. (line 42)
40630 * fshort-enums <2>: Type Attributes. (line 113)
40631 * fshort-enums <3>: Structures unions enumerations and bit-fields implementation.
40633 * fshort-enums: Code Gen Options. (line 106)
40634 * fshort-wchar: Code Gen Options. (line 125)
40635 * fsignaling-nans: Optimize Options. (line 1499)
40636 * fsigned-bitfields <1>: Non-bugs. (line 57)
40637 * fsigned-bitfields: C Dialect Options. (line 296)
40638 * fsigned-char <1>: Characters implementation.
40640 * fsigned-char: C Dialect Options. (line 286)
40641 * fsingle-precision-constant: Optimize Options. (line 1514)
40642 * fsplit-ivs-in-unroller: Optimize Options. (line 1010)
40643 * fsplit-wide-types: Optimize Options. (line 368)
40644 * fstack-check: Code Gen Options. (line 357)
40645 * fstack-limit-register: Code Gen Options. (line 396)
40646 * fstack-limit-symbol: Code Gen Options. (line 396)
40647 * fstack-protector: Optimize Options. (line 1666)
40648 * fstack-protector-all: Optimize Options. (line 1675)
40649 * fstats: C++ Dialect Options.
40651 * fstrict-aliasing: Optimize Options. (line 1103)
40652 * fstrict-overflow: Optimize Options. (line 1149)
40653 * fsyntax-only: Warning Options. (line 14)
40654 * ftabstop: Preprocessor Options.
40656 * ftemplate-depth: C++ Dialect Options.
40658 * ftest-coverage: Debugging Options. (line 313)
40659 * fthread-jumps: Optimize Options. (line 359)
40660 * ftime-report: Debugging Options. (line 243)
40661 * ftls-model: Code Gen Options. (line 441)
40662 * ftracer: Optimize Options. (line 993)
40663 * ftrapv: Code Gen Options. (line 22)
40664 * ftree-builtin-call-dce: Optimize Options. (line 788)
40665 * ftree-ccp: Optimize Options. (line 774)
40666 * ftree-ch: Optimize Options. (line 808)
40667 * ftree-copy-prop: Optimize Options. (line 710)
40668 * ftree-copyrename: Optimize Options. (line 953)
40669 * ftree-dce: Optimize Options. (line 784)
40670 * ftree-dominator-opts: Optimize Options. (line 794)
40671 * ftree-dse: Optimize Options. (line 801)
40672 * ftree-fre: Optimize Options. (line 703)
40673 * ftree-loop-im: Optimize Options. (line 918)
40674 * ftree-loop-ivcanon: Optimize Options. (line 927)
40675 * ftree-loop-linear: Optimize Options. (line 819)
40676 * ftree-loop-optimize: Optimize Options. (line 815)
40677 * ftree-parallelize-loops: Optimize Options. (line 938)
40678 * ftree-pre: Optimize Options. (line 699)
40679 * ftree-reassoc: Optimize Options. (line 695)
40680 * ftree-sink: Optimize Options. (line 770)
40681 * ftree-sra: Optimize Options. (line 947)
40682 * ftree-ter: Optimize Options. (line 960)
40683 * ftree-vect-loop-version: Optimize Options. (line 972)
40684 * ftree-vectorize: Optimize Options. (line 968)
40685 * ftree-vectorizer-verbose: Debugging Options. (line 808)
40686 * ftree-vrp: Optimize Options. (line 984)
40687 * funit-at-a-time: Optimize Options. (line 1247)
40688 * funroll-all-loops: Optimize Options. (line 1004)
40689 * funroll-loops: Optimize Options. (line 998)
40690 * funsafe-loop-optimizations: Optimize Options. (line 440)
40691 * funsafe-math-optimizations: Optimize Options. (line 1394)
40692 * funsigned-bitfields <1>: Non-bugs. (line 57)
40693 * funsigned-bitfields <2>: Structures unions enumerations and bit-fields implementation.
40695 * funsigned-bitfields: C Dialect Options. (line 296)
40696 * funsigned-char <1>: Characters implementation.
40698 * funsigned-char: C Dialect Options. (line 268)
40699 * funswitch-loops: Optimize Options. (line 1627)
40700 * funwind-tables: Code Gen Options. (line 57)
40701 * fuse-cxa-atexit: C++ Dialect Options.
40703 * fvar-tracking: Debugging Options. (line 888)
40704 * fvariable-expansion-in-unroller: Optimize Options. (line 1024)
40705 * fvect-cost-model: Optimize Options. (line 981)
40706 * fverbose-asm: Code Gen Options. (line 165)
40707 * fvisibility: Code Gen Options. (line 449)
40708 * fvisibility-inlines-hidden: C++ Dialect Options.
40710 * fvisibility-ms-compat: C++ Dialect Options.
40712 * fvpt: Optimize Options. (line 1573)
40713 * fweb: Optimize Options. (line 1266)
40714 * fwhole-program: Optimize Options. (line 1277)
40715 * fwide-exec-charset: Preprocessor Options.
40717 * fworking-directory: Preprocessor Options.
40719 * fwrapv: Code Gen Options. (line 26)
40720 * fzero-link: Objective-C and Objective-C++ Dialect Options.
40722 * G <1>: System V Options. (line 10)
40723 * G <2>: RS/6000 and PowerPC Options.
40725 * G <3>: MIPS Options. (line 314)
40726 * G: M32R/D Options. (line 57)
40727 * g: Debugging Options. (line 10)
40728 * gcoff: Debugging Options. (line 70)
40729 * gdwarf-2: Debugging Options. (line 88)
40730 * gdwarf-4: Debugging Options. (line 95)
40731 * gen-decls: Objective-C and Objective-C++ Dialect Options.
40733 * gfull: Darwin Options. (line 71)
40734 * ggdb: Debugging Options. (line 38)
40735 * gnu-ld: HPPA Options. (line 111)
40736 * gstabs: Debugging Options. (line 44)
40737 * gstabs+: Debugging Options. (line 64)
40738 * gused: Darwin Options. (line 66)
40739 * gvms: Debugging Options. (line 103)
40740 * gxcoff: Debugging Options. (line 75)
40741 * gxcoff+: Debugging Options. (line 80)
40742 * H: Preprocessor Options.
40744 * headerpad_max_install_names: Darwin Options. (line 199)
40745 * help <1>: Preprocessor Options.
40747 * help: Overall Options. (line 231)
40748 * hp-ld: HPPA Options. (line 123)
40749 * I <1>: Directory Options. (line 10)
40750 * I: Preprocessor Options.
40752 * I- <1>: Directory Options. (line 107)
40753 * I-: Preprocessor Options.
40755 * idirafter: Preprocessor Options.
40757 * iframework: Darwin Options. (line 59)
40758 * imacros: Preprocessor Options.
40760 * image_base: Darwin Options. (line 199)
40761 * imultilib: Preprocessor Options.
40763 * include: Preprocessor Options.
40765 * init: Darwin Options. (line 199)
40766 * install_name: Darwin Options. (line 199)
40767 * iprefix: Preprocessor Options.
40769 * iquote <1>: Directory Options. (line 31)
40770 * iquote: Preprocessor Options.
40772 * isysroot: Preprocessor Options.
40774 * isystem: Preprocessor Options.
40776 * iwithprefix: Preprocessor Options.
40778 * iwithprefixbefore: Preprocessor Options.
40780 * keep_private_externs: Darwin Options. (line 199)
40781 * L: Directory Options. (line 37)
40782 * l: Link Options. (line 26)
40783 * lobjc: Link Options. (line 53)
40784 * M: Preprocessor Options.
40786 * m1: SH Options. (line 9)
40787 * m10: PDP-11 Options. (line 29)
40788 * m128bit-long-double: i386 and x86-64 Options.
40790 * m16-bit: CRIS Options. (line 64)
40791 * m2: SH Options. (line 12)
40792 * m210: MCore Options. (line 43)
40793 * m3: SH Options. (line 18)
40794 * m31: S/390 and zSeries Options.
40796 * m32 <1>: SPARC Options. (line 191)
40797 * m32 <2>: RS/6000 and PowerPC Options.
40799 * m32: i386 and x86-64 Options.
40801 * m32-bit: CRIS Options. (line 64)
40802 * m32r: M32R/D Options. (line 15)
40803 * m32r2: M32R/D Options. (line 9)
40804 * m32rx: M32R/D Options. (line 12)
40805 * m340: MCore Options. (line 43)
40806 * m3dnow: i386 and x86-64 Options.
40808 * m3e: SH Options. (line 21)
40809 * m4: SH Options. (line 35)
40810 * m4-nofpu: SH Options. (line 24)
40811 * m4-single: SH Options. (line 31)
40812 * m4-single-only: SH Options. (line 27)
40813 * m40: PDP-11 Options. (line 23)
40814 * m45: PDP-11 Options. (line 26)
40815 * m4a: SH Options. (line 50)
40816 * m4a-nofpu: SH Options. (line 38)
40817 * m4a-single: SH Options. (line 46)
40818 * m4a-single-only: SH Options. (line 42)
40819 * m4al: SH Options. (line 53)
40820 * m4byte-functions: MCore Options. (line 27)
40821 * m5200: M680x0 Options. (line 143)
40822 * m5206e: M680x0 Options. (line 152)
40823 * m528x: M680x0 Options. (line 156)
40824 * m5307: M680x0 Options. (line 160)
40825 * m5407: M680x0 Options. (line 164)
40826 * m64 <1>: SPARC Options. (line 191)
40827 * m64 <2>: S/390 and zSeries Options.
40829 * m64 <3>: RS/6000 and PowerPC Options.
40831 * m64: i386 and x86-64 Options.
40833 * m68000: M680x0 Options. (line 91)
40834 * m68010: M680x0 Options. (line 99)
40835 * m68020: M680x0 Options. (line 105)
40836 * m68020-40: M680x0 Options. (line 174)
40837 * m68020-60: M680x0 Options. (line 183)
40838 * m68030: M680x0 Options. (line 110)
40839 * m68040: M680x0 Options. (line 115)
40840 * m68060: M680x0 Options. (line 124)
40841 * m6811: M68hc1x Options. (line 13)
40842 * m6812: M68hc1x Options. (line 18)
40843 * m68881: M680x0 Options. (line 193)
40844 * m68hc11: M68hc1x Options. (line 13)
40845 * m68hc12: M68hc1x Options. (line 18)
40846 * m68hcs12: M68hc1x Options. (line 23)
40847 * m68S12: M68hc1x Options. (line 23)
40848 * m8-bit: CRIS Options. (line 64)
40849 * m96bit-long-double: i386 and x86-64 Options.
40851 * mabi <1>: RS/6000 and PowerPC Options.
40853 * mabi: ARM Options. (line 10)
40854 * mabi-mmixware: MMIX Options. (line 20)
40855 * mabi=32: MIPS Options. (line 129)
40856 * mabi=64: MIPS Options. (line 129)
40857 * mabi=eabi: MIPS Options. (line 129)
40858 * mabi=gnu: MMIX Options. (line 20)
40859 * mabi=ibmlongdouble: RS/6000 and PowerPC Options.
40861 * mabi=ieeelongdouble: RS/6000 and PowerPC Options.
40863 * mabi=n32: MIPS Options. (line 129)
40864 * mabi=no-spe: RS/6000 and PowerPC Options.
40866 * mabi=o64: MIPS Options. (line 129)
40867 * mabi=spe: RS/6000 and PowerPC Options.
40869 * mabicalls: MIPS Options. (line 153)
40870 * mabort-on-noreturn: ARM Options. (line 149)
40871 * mabshi: PDP-11 Options. (line 55)
40872 * mac0: PDP-11 Options. (line 16)
40873 * macc-4: FRV Options. (line 113)
40874 * macc-8: FRV Options. (line 116)
40875 * maccumulate-outgoing-args: i386 and x86-64 Options.
40877 * madjust-unroll: SH Options. (line 196)
40878 * mads: RS/6000 and PowerPC Options.
40880 * maix-struct-return: RS/6000 and PowerPC Options.
40882 * maix32: RS/6000 and PowerPC Options.
40884 * maix64: RS/6000 and PowerPC Options.
40886 * malign-300: H8/300 Options. (line 31)
40887 * malign-double: i386 and x86-64 Options.
40889 * malign-int: M680x0 Options. (line 263)
40890 * malign-labels: FRV Options. (line 104)
40891 * malign-loops: M32R/D Options. (line 73)
40892 * malign-natural: RS/6000 and PowerPC Options.
40894 * malign-power: RS/6000 and PowerPC Options.
40896 * malloc-cc: FRV Options. (line 25)
40897 * malpha-as: DEC Alpha Options. (line 159)
40898 * maltivec: RS/6000 and PowerPC Options.
40900 * mam33: MN10300 Options. (line 17)
40901 * mandroid: ARM Options. (line 264)
40902 * mapcs: ARM Options. (line 22)
40903 * mapcs-frame: ARM Options. (line 14)
40904 * mapp-regs <1>: V850 Options. (line 57)
40905 * mapp-regs: SPARC Options. (line 10)
40906 * march <1>: S/390 and zSeries Options.
40908 * march <2>: MIPS Options. (line 14)
40909 * march <3>: M680x0 Options. (line 12)
40910 * march <4>: i386 and x86-64 Options.
40912 * march <5>: HPPA Options. (line 9)
40913 * march <6>: CRIS Options. (line 10)
40914 * march: ARM Options. (line 112)
40915 * masm=DIALECT: i386 and x86-64 Options.
40917 * mauto-incdec: M68hc1x Options. (line 26)
40918 * mauto-pic: IA-64 Options. (line 50)
40919 * mavoid-indexed-addresses: RS/6000 and PowerPC Options.
40921 * mb: SH Options. (line 58)
40922 * mbackchain: S/390 and zSeries Options.
40924 * mbase-addresses: MMIX Options. (line 54)
40925 * mbcopy: PDP-11 Options. (line 36)
40926 * mbig: RS/6000 and PowerPC Options.
40928 * mbig-endian <1>: RS/6000 and PowerPC Options.
40930 * mbig-endian <2>: MCore Options. (line 39)
40931 * mbig-endian <3>: IA-64 Options. (line 9)
40932 * mbig-endian: ARM Options. (line 72)
40933 * mbig-switch <1>: V850 Options. (line 52)
40934 * mbig-switch: HPPA Options. (line 23)
40935 * mbigtable: SH Options. (line 74)
40936 * mbit-align: RS/6000 and PowerPC Options.
40938 * mbitfield: M680x0 Options. (line 231)
40939 * mbitops: SH Options. (line 78)
40940 * mbranch-cheap: PDP-11 Options. (line 65)
40941 * mbranch-cost: MIPS Options. (line 610)
40942 * mbranch-cost=NUMBER: M32R/D Options. (line 82)
40943 * mbranch-expensive: PDP-11 Options. (line 61)
40944 * mbranch-hints: SPU Options. (line 27)
40945 * mbranch-likely: MIPS Options. (line 617)
40946 * mbranch-predict: MMIX Options. (line 49)
40947 * mbss-plt: RS/6000 and PowerPC Options.
40949 * mbuild-constants: DEC Alpha Options. (line 142)
40950 * mbwx: DEC Alpha Options. (line 171)
40951 * mc68000: M680x0 Options. (line 91)
40952 * mc68020: M680x0 Options. (line 105)
40953 * mcall-gnu: RS/6000 and PowerPC Options.
40955 * mcall-linux: RS/6000 and PowerPC Options.
40957 * mcall-netbsd: RS/6000 and PowerPC Options.
40959 * mcall-prologues: AVR Options. (line 43)
40960 * mcall-solaris: RS/6000 and PowerPC Options.
40962 * mcall-sysv: RS/6000 and PowerPC Options.
40964 * mcall-sysv-eabi: RS/6000 and PowerPC Options.
40966 * mcall-sysv-noeabi: RS/6000 and PowerPC Options.
40968 * mcallee-super-interworking: ARM Options. (line 238)
40969 * mcaller-super-interworking: ARM Options. (line 244)
40970 * mcallgraph-data: MCore Options. (line 31)
40971 * mcc-init: CRIS Options. (line 41)
40972 * mcfv4e: M680x0 Options. (line 168)
40973 * mcheck-zero-division: MIPS Options. (line 425)
40974 * mcirrus-fix-invalid-insns: ARM Options. (line 189)
40975 * mcix: DEC Alpha Options. (line 171)
40976 * mcld: i386 and x86-64 Options.
40978 * mcmodel=embmedany: SPARC Options. (line 213)
40979 * mcmodel=kernel: i386 and x86-64 Options.
40981 * mcmodel=large: i386 and x86-64 Options.
40983 * mcmodel=medany: SPARC Options. (line 207)
40984 * mcmodel=medium: i386 and x86-64 Options.
40986 * mcmodel=medlow: SPARC Options. (line 196)
40987 * mcmodel=medmid: SPARC Options. (line 201)
40988 * mcmodel=small: i386 and x86-64 Options.
40990 * mcmpb: RS/6000 and PowerPC Options.
40992 * mcode-readable: MIPS Options. (line 385)
40993 * mcond-exec: FRV Options. (line 152)
40994 * mcond-move: FRV Options. (line 128)
40995 * mconsole: i386 and x86-64 Windows Options.
40997 * mconst-align: CRIS Options. (line 55)
40998 * mconst16: Xtensa Options. (line 10)
40999 * mconstant-gp: IA-64 Options. (line 46)
41000 * mcorea: Blackfin Options. (line 149)
41001 * mcoreb: Blackfin Options. (line 155)
41002 * mcpu <1>: SPARC Options. (line 96)
41003 * mcpu <2>: RS/6000 and PowerPC Options.
41005 * mcpu <3>: picoChip Options. (line 9)
41006 * mcpu <4>: M680x0 Options. (line 28)
41007 * mcpu <5>: i386 and x86-64 Options.
41009 * mcpu <6>: FRV Options. (line 212)
41010 * mcpu <7>: DEC Alpha Options. (line 223)
41011 * mcpu <8>: CRIS Options. (line 10)
41012 * mcpu <9>: ARM Options. (line 84)
41013 * mcpu: ARC Options. (line 23)
41014 * mcpu32: M680x0 Options. (line 134)
41015 * mcpu= <1>: M32C Options. (line 7)
41016 * mcpu=: Blackfin Options. (line 7)
41017 * mcsync-anomaly: Blackfin Options. (line 55)
41018 * mcx16: i386 and x86-64 Options.
41020 * mcygwin: i386 and x86-64 Windows Options.
41022 * MD: Preprocessor Options.
41024 * mdalign: SH Options. (line 64)
41025 * mdata: ARC Options. (line 30)
41026 * mdata-align: CRIS Options. (line 55)
41027 * mdebug <1>: S/390 and zSeries Options.
41029 * mdebug: M32R/D Options. (line 69)
41030 * mdec-asm: PDP-11 Options. (line 78)
41031 * mdisable-callt: V850 Options. (line 80)
41032 * mdisable-fpregs: HPPA Options. (line 33)
41033 * mdisable-indexing: HPPA Options. (line 40)
41034 * mdiv <1>: MCore Options. (line 15)
41035 * mdiv: M680x0 Options. (line 205)
41036 * mdiv=STRATEGY: SH Options. (line 141)
41037 * mdivide-breaks: MIPS Options. (line 431)
41038 * mdivide-traps: MIPS Options. (line 431)
41039 * mdivsi3_libfunc=NAME: SH Options. (line 182)
41040 * mdll: i386 and x86-64 Windows Options.
41042 * mdlmzb: RS/6000 and PowerPC Options.
41044 * mdmx: MIPS Options. (line 278)
41045 * mdouble: FRV Options. (line 38)
41046 * mdouble-float <1>: RS/6000 and PowerPC Options.
41048 * mdouble-float: MIPS Options. (line 236)
41049 * mdsp: MIPS Options. (line 255)
41050 * mdspr2: MIPS Options. (line 261)
41051 * mdual-nops: SPU Options. (line 55)
41052 * mdwarf2-asm: IA-64 Options. (line 79)
41053 * mdword: FRV Options. (line 32)
41054 * mdynamic-no-pic: RS/6000 and PowerPC Options.
41056 * meabi: RS/6000 and PowerPC Options.
41058 * mearly-stop-bits: IA-64 Options. (line 85)
41059 * meb: Score Options. (line 9)
41060 * mel: Score Options. (line 12)
41061 * melf <1>: MMIX Options. (line 44)
41062 * melf: CRIS Options. (line 87)
41063 * memb: RS/6000 and PowerPC Options.
41065 * membedded-data: MIPS Options. (line 372)
41066 * memregs=: M32C Options. (line 21)
41067 * mep: V850 Options. (line 16)
41068 * mepsilon: MMIX Options. (line 15)
41069 * merror-reloc: SPU Options. (line 10)
41070 * mesa: S/390 and zSeries Options.
41072 * metrax100: CRIS Options. (line 26)
41073 * metrax4: CRIS Options. (line 26)
41074 * mexplicit-relocs <1>: MIPS Options. (line 416)
41075 * mexplicit-relocs: DEC Alpha Options. (line 184)
41076 * mextern-sdata: MIPS Options. (line 334)
41077 * MF: Preprocessor Options.
41079 * mfast-fp: Blackfin Options. (line 128)
41080 * mfast-indirect-calls: HPPA Options. (line 52)
41081 * mfaster-structs: SPARC Options. (line 71)
41082 * mfdpic: FRV Options. (line 56)
41083 * mfix: DEC Alpha Options. (line 171)
41084 * mfix-and-continue: Darwin Options. (line 106)
41085 * mfix-cortex-m3-ldrd: ARC Options. (line 36)
41086 * mfix-r10000: MIPS Options. (line 502)
41087 * mfix-r4000: MIPS Options. (line 481)
41088 * mfix-r4400: MIPS Options. (line 495)
41089 * mfix-sb1: MIPS Options. (line 534)
41090 * mfix-vr4120: MIPS Options. (line 513)
41091 * mfix-vr4130: MIPS Options. (line 527)
41092 * mfixed-cc: FRV Options. (line 28)
41093 * mfixed-range <1>: SPU Options. (line 47)
41094 * mfixed-range <2>: SH Options. (line 189)
41095 * mfixed-range <3>: IA-64 Options. (line 90)
41096 * mfixed-range: HPPA Options. (line 59)
41097 * mflip-mips16: MIPS Options. (line 109)
41098 * mfloat-abi: ARM Options. (line 41)
41099 * mfloat-gprs: RS/6000 and PowerPC Options.
41101 * mfloat-ieee: DEC Alpha Options. (line 179)
41102 * mfloat-vax: DEC Alpha Options. (line 179)
41103 * mfloat32: PDP-11 Options. (line 52)
41104 * mfloat64: PDP-11 Options. (line 48)
41105 * mflush-func: MIPS Options. (line 601)
41106 * mflush-func=NAME: M32R/D Options. (line 94)
41107 * mflush-trap=NUMBER: M32R/D Options. (line 87)
41108 * mfmovd: SH Options. (line 81)
41109 * mfp: ARM Options. (line 124)
41110 * mfp-exceptions: MIPS Options. (line 628)
41111 * mfp-reg: DEC Alpha Options. (line 25)
41112 * mfp-rounding-mode: DEC Alpha Options. (line 85)
41113 * mfp-trap-mode: DEC Alpha Options. (line 63)
41114 * mfp32: MIPS Options. (line 219)
41115 * mfp64: MIPS Options. (line 222)
41116 * mfpe: ARM Options. (line 124)
41117 * mfpr-32: FRV Options. (line 13)
41118 * mfpr-64: FRV Options. (line 16)
41119 * mfprnd: RS/6000 and PowerPC Options.
41121 * mfpu <1>: SPARC Options. (line 20)
41122 * mfpu <2>: RS/6000 and PowerPC Options.
41124 * mfpu <3>: PDP-11 Options. (line 9)
41125 * mfpu: ARM Options. (line 124)
41126 * mfull-toc: RS/6000 and PowerPC Options.
41128 * mfused-madd <1>: Xtensa Options. (line 19)
41129 * mfused-madd <2>: S/390 and zSeries Options.
41131 * mfused-madd <3>: RS/6000 and PowerPC Options.
41133 * mfused-madd <4>: MIPS Options. (line 466)
41134 * mfused-madd: i386 and x86-64 Options.
41136 * mg: VAX Options. (line 17)
41137 * MG: Preprocessor Options.
41139 * mgas <1>: HPPA Options. (line 75)
41140 * mgas: DEC Alpha Options. (line 159)
41141 * mgen-cell-microcode: RS/6000 and PowerPC Options.
41143 * mgettrcost=NUMBER: SH Options. (line 211)
41144 * mglibc: GNU/Linux Options. (line 9)
41145 * mgnu: VAX Options. (line 13)
41146 * mgnu-as: IA-64 Options. (line 18)
41147 * mgnu-ld: IA-64 Options. (line 23)
41148 * mgotplt: CRIS Options. (line 81)
41149 * mgp32: MIPS Options. (line 213)
41150 * mgp64: MIPS Options. (line 216)
41151 * mgpopt: MIPS Options. (line 357)
41152 * mgpr-32: FRV Options. (line 7)
41153 * mgpr-64: FRV Options. (line 10)
41154 * mgprel-ro: FRV Options. (line 79)
41155 * mh: H8/300 Options. (line 14)
41156 * mhard-dfp <1>: S/390 and zSeries Options.
41158 * mhard-dfp: RS/6000 and PowerPC Options.
41160 * mhard-float <1>: SPARC Options. (line 20)
41161 * mhard-float <2>: S/390 and zSeries Options.
41163 * mhard-float <3>: RS/6000 and PowerPC Options.
41165 * mhard-float <4>: MIPS Options. (line 225)
41166 * mhard-float <5>: M680x0 Options. (line 193)
41167 * mhard-float <6>: FRV Options. (line 19)
41168 * mhard-float: ARM Options. (line 62)
41169 * mhard-quad-float: SPARC Options. (line 41)
41170 * mhardlit: MCore Options. (line 10)
41171 * mhint-max-distance: SPU Options. (line 67)
41172 * mhint-max-nops: SPU Options. (line 61)
41173 * mhitachi: SH Options. (line 84)
41174 * micplb: Blackfin Options. (line 168)
41175 * mid-shared-library: Blackfin Options. (line 76)
41176 * mieee <1>: SH Options. (line 99)
41177 * mieee: DEC Alpha Options. (line 39)
41178 * mieee-conformant: DEC Alpha Options. (line 134)
41179 * mieee-fp: i386 and x86-64 Options.
41181 * mieee-with-inexact: DEC Alpha Options. (line 52)
41182 * milp32: IA-64 Options. (line 114)
41183 * mimpure-text: SPARC Options. (line 81)
41184 * mincoming-stack-boundary: i386 and x86-64 Options.
41186 * mindexed-addressing: SH Options. (line 201)
41187 * minit-stack: AVR Options. (line 35)
41188 * minline-all-stringops: i386 and x86-64 Options.
41190 * minline-float-divide-max-throughput: IA-64 Options. (line 58)
41191 * minline-float-divide-min-latency: IA-64 Options. (line 54)
41192 * minline-ic_invalidate: SH Options. (line 106)
41193 * minline-int-divide-max-throughput: IA-64 Options. (line 66)
41194 * minline-int-divide-min-latency: IA-64 Options. (line 62)
41195 * minline-plt <1>: FRV Options. (line 64)
41196 * minline-plt: Blackfin Options. (line 133)
41197 * minline-sqrt-max-throughput: IA-64 Options. (line 74)
41198 * minline-sqrt-min-latency: IA-64 Options. (line 70)
41199 * minline-stringops-dynamically: i386 and x86-64 Options.
41201 * minmax: M68hc1x Options. (line 31)
41202 * minsert-sched-nops: RS/6000 and PowerPC Options.
41204 * mint16: PDP-11 Options. (line 40)
41205 * mint32 <1>: PDP-11 Options. (line 44)
41206 * mint32: H8/300 Options. (line 28)
41207 * mint8: AVR Options. (line 55)
41208 * minterlink-mips16: MIPS Options. (line 116)
41209 * minvalid-symbols: SH Options. (line 234)
41210 * mips1: MIPS Options. (line 76)
41211 * mips16: MIPS Options. (line 101)
41212 * mips2: MIPS Options. (line 79)
41213 * mips3: MIPS Options. (line 82)
41214 * mips32: MIPS Options. (line 88)
41215 * mips32r2: MIPS Options. (line 91)
41216 * mips3d: MIPS Options. (line 284)
41217 * mips4: MIPS Options. (line 85)
41218 * mips64: MIPS Options. (line 94)
41219 * mips64r2: MIPS Options. (line 97)
41220 * misel: RS/6000 and PowerPC Options.
41222 * misize: SH Options. (line 118)
41223 * missue-rate=NUMBER: M32R/D Options. (line 79)
41224 * mjump-in-delay: HPPA Options. (line 28)
41225 * mkernel: Darwin Options. (line 84)
41226 * mknuthdiv: MMIX Options. (line 33)
41227 * ml: SH Options. (line 61)
41228 * mlarge-data: DEC Alpha Options. (line 195)
41229 * mlarge-data-threshold=NUMBER: i386 and x86-64 Options.
41231 * mlarge-mem: SPU Options. (line 35)
41232 * mlarge-text: DEC Alpha Options. (line 213)
41233 * mleaf-id-shared-library: Blackfin Options. (line 87)
41234 * mlibfuncs: MMIX Options. (line 10)
41235 * mlibrary-pic: FRV Options. (line 110)
41236 * mlinked-fp: FRV Options. (line 94)
41237 * mlinker-opt: HPPA Options. (line 85)
41238 * mlinux: CRIS Options. (line 91)
41239 * mlittle: RS/6000 and PowerPC Options.
41241 * mlittle-endian <1>: SPARC Options. (line 185)
41242 * mlittle-endian <2>: RS/6000 and PowerPC Options.
41244 * mlittle-endian <3>: MCore Options. (line 39)
41245 * mlittle-endian <4>: IA-64 Options. (line 13)
41246 * mlittle-endian: ARM Options. (line 68)
41247 * mllsc: MIPS Options. (line 241)
41248 * mlocal-sdata: MIPS Options. (line 322)
41249 * mlong-calls <1>: V850 Options. (line 10)
41250 * mlong-calls <2>: MIPS Options. (line 452)
41251 * mlong-calls <3>: M68hc1x Options. (line 35)
41252 * mlong-calls <4>: FRV Options. (line 99)
41253 * mlong-calls <5>: Blackfin Options. (line 116)
41254 * mlong-calls: ARM Options. (line 154)
41255 * mlong-double-128: S/390 and zSeries Options.
41257 * mlong-double-64: S/390 and zSeries Options.
41259 * mlong-load-store: HPPA Options. (line 66)
41260 * mlong32: MIPS Options. (line 297)
41261 * mlong64: MIPS Options. (line 292)
41262 * mlongcall: RS/6000 and PowerPC Options.
41264 * mlongcalls: Xtensa Options. (line 67)
41265 * mlow-64k: Blackfin Options. (line 65)
41266 * mlp64: IA-64 Options. (line 114)
41267 * MM: Preprocessor Options.
41269 * mmac <1>: Score Options. (line 21)
41270 * mmac: CRX Options. (line 9)
41271 * mmad: MIPS Options. (line 461)
41272 * mmangle-cpu: ARC Options. (line 15)
41273 * mmax: DEC Alpha Options. (line 171)
41274 * mmax-stack-frame: CRIS Options. (line 22)
41275 * mmcu: AVR Options. (line 9)
41276 * MMD: Preprocessor Options.
41278 * mmedia: FRV Options. (line 44)
41279 * mmemcpy: MIPS Options. (line 446)
41280 * mmemory-latency: DEC Alpha Options. (line 276)
41281 * mmfcrf: RS/6000 and PowerPC Options.
41283 * mmfpgpr: RS/6000 and PowerPC Options.
41285 * mminimal-toc: RS/6000 and PowerPC Options.
41287 * mmmx: i386 and x86-64 Options.
41289 * mmodel=large: M32R/D Options. (line 33)
41290 * mmodel=medium: M32R/D Options. (line 27)
41291 * mmodel=small: M32R/D Options. (line 18)
41292 * mmt: MIPS Options. (line 289)
41293 * mmul-bug-workaround: CRIS Options. (line 31)
41294 * mmuladd: FRV Options. (line 50)
41295 * mmulhw: RS/6000 and PowerPC Options.
41297 * mmult-bug: MN10300 Options. (line 9)
41298 * mmulti-cond-exec: FRV Options. (line 176)
41299 * mmulticore: Blackfin Options. (line 137)
41300 * mmultiple: RS/6000 and PowerPC Options.
41302 * mmvcle: S/390 and zSeries Options.
41304 * mmvme: RS/6000 and PowerPC Options.
41306 * mn: H8/300 Options. (line 20)
41307 * mnested-cond-exec: FRV Options. (line 189)
41308 * mnew-mnemonics: RS/6000 and PowerPC Options.
41310 * mnhwloop: Score Options. (line 15)
41311 * mno-3dnow: i386 and x86-64 Options.
41313 * mno-4byte-functions: MCore Options. (line 27)
41314 * mno-abicalls: MIPS Options. (line 153)
41315 * mno-abshi: PDP-11 Options. (line 58)
41316 * mno-ac0: PDP-11 Options. (line 20)
41317 * mno-align-double: i386 and x86-64 Options.
41319 * mno-align-int: M680x0 Options. (line 263)
41320 * mno-align-loops: M32R/D Options. (line 76)
41321 * mno-align-stringops: i386 and x86-64 Options.
41323 * mno-altivec: RS/6000 and PowerPC Options.
41325 * mno-am33: MN10300 Options. (line 20)
41326 * mno-app-regs <1>: V850 Options. (line 61)
41327 * mno-app-regs: SPARC Options. (line 10)
41328 * mno-avoid-indexed-addresses: RS/6000 and PowerPC Options.
41330 * mno-backchain: S/390 and zSeries Options.
41332 * mno-base-addresses: MMIX Options. (line 54)
41333 * mno-bit-align: RS/6000 and PowerPC Options.
41335 * mno-bitfield: M680x0 Options. (line 227)
41336 * mno-branch-likely: MIPS Options. (line 617)
41337 * mno-branch-predict: MMIX Options. (line 49)
41338 * mno-bwx: DEC Alpha Options. (line 171)
41339 * mno-callgraph-data: MCore Options. (line 31)
41340 * mno-check-zero-division: MIPS Options. (line 425)
41341 * mno-cirrus-fix-invalid-insns: ARM Options. (line 189)
41342 * mno-cix: DEC Alpha Options. (line 171)
41343 * mno-cmpb: RS/6000 and PowerPC Options.
41345 * mno-cond-exec: FRV Options. (line 158)
41346 * mno-cond-move: FRV Options. (line 134)
41347 * mno-const-align: CRIS Options. (line 55)
41348 * mno-const16: Xtensa Options. (line 10)
41349 * mno-crt0: MN10300 Options. (line 31)
41350 * mno-csync-anomaly: Blackfin Options. (line 61)
41351 * mno-cygwin: i386 and x86-64 Windows Options.
41353 * mno-data-align: CRIS Options. (line 55)
41354 * mno-debug: S/390 and zSeries Options.
41356 * mno-div <1>: MCore Options. (line 15)
41357 * mno-div: M680x0 Options. (line 205)
41358 * mno-dlmzb: RS/6000 and PowerPC Options.
41360 * mno-double: FRV Options. (line 41)
41361 * mno-dsp: MIPS Options. (line 255)
41362 * mno-dspr2: MIPS Options. (line 261)
41363 * mno-dwarf2-asm: IA-64 Options. (line 79)
41364 * mno-dword: FRV Options. (line 35)
41365 * mno-eabi: RS/6000 and PowerPC Options.
41367 * mno-early-stop-bits: IA-64 Options. (line 85)
41368 * mno-eflags: FRV Options. (line 125)
41369 * mno-embedded-data: MIPS Options. (line 372)
41370 * mno-ep: V850 Options. (line 16)
41371 * mno-epsilon: MMIX Options. (line 15)
41372 * mno-explicit-relocs <1>: MIPS Options. (line 416)
41373 * mno-explicit-relocs: DEC Alpha Options. (line 184)
41374 * mno-extern-sdata: MIPS Options. (line 334)
41375 * mno-fancy-math-387: i386 and x86-64 Options.
41377 * mno-faster-structs: SPARC Options. (line 71)
41378 * mno-fix: DEC Alpha Options. (line 171)
41379 * mno-fix-r10000: MIPS Options. (line 502)
41380 * mno-fix-r4000: MIPS Options. (line 481)
41381 * mno-fix-r4400: MIPS Options. (line 495)
41382 * mno-float32: PDP-11 Options. (line 48)
41383 * mno-float64: PDP-11 Options. (line 52)
41384 * mno-flush-func: M32R/D Options. (line 99)
41385 * mno-flush-trap: M32R/D Options. (line 91)
41386 * mno-fp-in-toc: RS/6000 and PowerPC Options.
41388 * mno-fp-regs: DEC Alpha Options. (line 25)
41389 * mno-fp-ret-in-387: i386 and x86-64 Options.
41391 * mno-fprnd: RS/6000 and PowerPC Options.
41393 * mno-fpu: SPARC Options. (line 25)
41394 * mno-fused-madd <1>: Xtensa Options. (line 19)
41395 * mno-fused-madd <2>: S/390 and zSeries Options.
41397 * mno-fused-madd <3>: RS/6000 and PowerPC Options.
41399 * mno-fused-madd: MIPS Options. (line 466)
41400 * mno-gnu-as: IA-64 Options. (line 18)
41401 * mno-gnu-ld: IA-64 Options. (line 23)
41402 * mno-gotplt: CRIS Options. (line 81)
41403 * mno-gpopt: MIPS Options. (line 357)
41404 * mno-hard-dfp <1>: S/390 and zSeries Options.
41406 * mno-hard-dfp: RS/6000 and PowerPC Options.
41408 * mno-hardlit: MCore Options. (line 10)
41409 * mno-id-shared-library: Blackfin Options. (line 83)
41410 * mno-ieee-fp: i386 and x86-64 Options.
41412 * mno-int16: PDP-11 Options. (line 44)
41413 * mno-int32: PDP-11 Options. (line 40)
41414 * mno-interlink-mips16: MIPS Options. (line 116)
41415 * mno-interrupts: AVR Options. (line 39)
41416 * mno-isel: RS/6000 and PowerPC Options.
41418 * mno-knuthdiv: MMIX Options. (line 33)
41419 * mno-leaf-id-shared-library: Blackfin Options. (line 93)
41420 * mno-libfuncs: MMIX Options. (line 10)
41421 * mno-llsc: MIPS Options. (line 241)
41422 * mno-local-sdata: MIPS Options. (line 322)
41423 * mno-long-calls <1>: V850 Options. (line 10)
41424 * mno-long-calls <2>: MIPS Options. (line 452)
41425 * mno-long-calls <3>: M68hc1x Options. (line 35)
41426 * mno-long-calls <4>: HPPA Options. (line 136)
41427 * mno-long-calls <5>: Blackfin Options. (line 116)
41428 * mno-long-calls: ARM Options. (line 154)
41429 * mno-longcall: RS/6000 and PowerPC Options.
41431 * mno-longcalls: Xtensa Options. (line 67)
41432 * mno-low-64k: Blackfin Options. (line 69)
41433 * mno-lsim: FR30 Options. (line 14)
41434 * mno-mad: MIPS Options. (line 461)
41435 * mno-max: DEC Alpha Options. (line 171)
41436 * mno-mdmx: MIPS Options. (line 278)
41437 * mno-media: FRV Options. (line 47)
41438 * mno-memcpy: MIPS Options. (line 446)
41439 * mno-mfcrf: RS/6000 and PowerPC Options.
41441 * mno-mfpgpr: RS/6000 and PowerPC Options.
41443 * mno-mips16: MIPS Options. (line 101)
41444 * mno-mips3d: MIPS Options. (line 284)
41445 * mno-mmx: i386 and x86-64 Options.
41447 * mno-mt: MIPS Options. (line 289)
41448 * mno-mul-bug-workaround: CRIS Options. (line 31)
41449 * mno-muladd: FRV Options. (line 53)
41450 * mno-mulhw: RS/6000 and PowerPC Options.
41452 * mno-mult-bug: MN10300 Options. (line 13)
41453 * mno-multi-cond-exec: FRV Options. (line 183)
41454 * mno-multiple: RS/6000 and PowerPC Options.
41456 * mno-mvcle: S/390 and zSeries Options.
41458 * mno-nested-cond-exec: FRV Options. (line 195)
41459 * mno-optimize-membar: FRV Options. (line 205)
41460 * mno-pack: FRV Options. (line 122)
41461 * mno-packed-stack: S/390 and zSeries Options.
41463 * mno-paired: RS/6000 and PowerPC Options.
41465 * mno-paired-single: MIPS Options. (line 272)
41466 * mno-pic: IA-64 Options. (line 26)
41467 * mno-plt: MIPS Options. (line 180)
41468 * mno-popcntb: RS/6000 and PowerPC Options.
41470 * mno-power: RS/6000 and PowerPC Options.
41472 * mno-power2: RS/6000 and PowerPC Options.
41474 * mno-powerpc: RS/6000 and PowerPC Options.
41476 * mno-powerpc-gfxopt: RS/6000 and PowerPC Options.
41478 * mno-powerpc-gpopt: RS/6000 and PowerPC Options.
41480 * mno-powerpc64: RS/6000 and PowerPC Options.
41482 * mno-prolog-function: V850 Options. (line 23)
41483 * mno-prologue-epilogue: CRIS Options. (line 71)
41484 * mno-prototype: RS/6000 and PowerPC Options.
41486 * mno-push-args: i386 and x86-64 Options.
41488 * mno-register-names: IA-64 Options. (line 37)
41489 * mno-regnames: RS/6000 and PowerPC Options.
41491 * mno-relax-immediate: MCore Options. (line 19)
41492 * mno-relocatable: RS/6000 and PowerPC Options.
41494 * mno-relocatable-lib: RS/6000 and PowerPC Options.
41496 * mno-rtd: M680x0 Options. (line 258)
41497 * mno-scc: FRV Options. (line 146)
41498 * mno-sched-ar-data-spec: IA-64 Options. (line 128)
41499 * mno-sched-ar-in-data-spec: IA-64 Options. (line 149)
41500 * mno-sched-br-data-spec: IA-64 Options. (line 121)
41501 * mno-sched-br-in-data-spec: IA-64 Options. (line 142)
41502 * mno-sched-control-ldc: IA-64 Options. (line 168)
41503 * mno-sched-control-spec: IA-64 Options. (line 135)
41504 * mno-sched-count-spec-in-critical-path: IA-64 Options. (line 194)
41505 * mno-sched-in-control-spec: IA-64 Options. (line 156)
41506 * mno-sched-ldc: IA-64 Options. (line 162)
41507 * mno-sched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
41508 * mno-sched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
41509 * mno-sched-prolog: ARM Options. (line 32)
41510 * mno-sched-spec-verbose: IA-64 Options. (line 176)
41511 * mno-sdata <1>: RS/6000 and PowerPC Options.
41513 * mno-sdata: IA-64 Options. (line 42)
41514 * mno-sep-data: Blackfin Options. (line 111)
41515 * mno-serialize-volatile: Xtensa Options. (line 35)
41516 * mno-short: M680x0 Options. (line 222)
41517 * mno-side-effects: CRIS Options. (line 46)
41518 * mno-single-exit: MMIX Options. (line 66)
41519 * mno-slow-bytes: MCore Options. (line 35)
41520 * mno-small-exec: S/390 and zSeries Options.
41522 * mno-smartmips: MIPS Options. (line 268)
41523 * mno-soft-float: DEC Alpha Options. (line 10)
41524 * mno-space-regs: HPPA Options. (line 45)
41525 * mno-spe: RS/6000 and PowerPC Options.
41527 * mno-specld-anomaly: Blackfin Options. (line 51)
41528 * mno-split: PDP-11 Options. (line 71)
41529 * mno-split-addresses: MIPS Options. (line 410)
41530 * mno-sse: i386 and x86-64 Options.
41532 * mno-stack-align: CRIS Options. (line 55)
41533 * mno-stack-bias: SPARC Options. (line 222)
41534 * mno-strict-align <1>: RS/6000 and PowerPC Options.
41536 * mno-strict-align: M680x0 Options. (line 283)
41537 * mno-string: RS/6000 and PowerPC Options.
41539 * mno-sum-in-toc: RS/6000 and PowerPC Options.
41541 * mno-swdiv: RS/6000 and PowerPC Options.
41543 * mno-sym32: MIPS Options. (line 307)
41544 * mno-tablejump: AVR Options. (line 47)
41545 * mno-target-align: Xtensa Options. (line 54)
41546 * mno-text-section-literals: Xtensa Options. (line 42)
41547 * mno-toc: RS/6000 and PowerPC Options.
41549 * mno-toplevel-symbols: MMIX Options. (line 40)
41550 * mno-tpf-trace: S/390 and zSeries Options.
41552 * mno-unaligned-doubles: SPARC Options. (line 59)
41553 * mno-uninit-const-in-rodata: MIPS Options. (line 380)
41554 * mno-update: RS/6000 and PowerPC Options.
41556 * mno-v8plus: SPARC Options. (line 170)
41557 * mno-vis: SPARC Options. (line 177)
41558 * mno-vliw-branch: FRV Options. (line 170)
41559 * mno-volatile-asm-stop: IA-64 Options. (line 32)
41560 * mno-vrsave: RS/6000 and PowerPC Options.
41562 * mno-wide-bitfields: MCore Options. (line 23)
41563 * mno-xgot <1>: MIPS Options. (line 190)
41564 * mno-xgot: M680x0 Options. (line 315)
41565 * mno-xl-compat: RS/6000 and PowerPC Options.
41567 * mno-zero-extend: MMIX Options. (line 27)
41568 * mnobitfield: M680x0 Options. (line 227)
41569 * mnomacsave: SH Options. (line 95)
41570 * mnominmax: M68hc1x Options. (line 31)
41571 * mnop-fun-dllimport: i386 and x86-64 Windows Options.
41573 * mold-mnemonics: RS/6000 and PowerPC Options.
41575 * momit-leaf-frame-pointer <1>: i386 and x86-64 Options.
41577 * momit-leaf-frame-pointer: Blackfin Options. (line 39)
41578 * mone-byte-bool: Darwin Options. (line 92)
41579 * moptimize-membar: FRV Options. (line 201)
41580 * MP: Preprocessor Options.
41582 * mpa-risc-1-0: HPPA Options. (line 19)
41583 * mpa-risc-1-1: HPPA Options. (line 19)
41584 * mpa-risc-2-0: HPPA Options. (line 19)
41585 * mpack: FRV Options. (line 119)
41586 * mpacked-stack: S/390 and zSeries Options.
41588 * mpadstruct: SH Options. (line 121)
41589 * mpaired: RS/6000 and PowerPC Options.
41591 * mpaired-single: MIPS Options. (line 272)
41592 * mpc32: i386 and x86-64 Options.
41594 * mpc64: i386 and x86-64 Options.
41596 * mpc80: i386 and x86-64 Options.
41598 * mpcrel: M680x0 Options. (line 275)
41599 * mpdebug: CRIS Options. (line 35)
41600 * mpe: RS/6000 and PowerPC Options.
41602 * mpic-register: ARM Options. (line 185)
41603 * mplt: MIPS Options. (line 180)
41604 * mpoke-function-name: ARM Options. (line 199)
41605 * mpopcntb: RS/6000 and PowerPC Options.
41607 * mportable-runtime: HPPA Options. (line 71)
41608 * mpower: RS/6000 and PowerPC Options.
41610 * mpower2: RS/6000 and PowerPC Options.
41612 * mpowerpc: RS/6000 and PowerPC Options.
41614 * mpowerpc-gfxopt: RS/6000 and PowerPC Options.
41616 * mpowerpc-gpopt: RS/6000 and PowerPC Options.
41618 * mpowerpc64: RS/6000 and PowerPC Options.
41620 * mprefergot: SH Options. (line 128)
41621 * mpreferred-stack-boundary: i386 and x86-64 Options.
41623 * mprioritize-restricted-insns: RS/6000 and PowerPC Options.
41625 * mprolog-function: V850 Options. (line 23)
41626 * mprologue-epilogue: CRIS Options. (line 71)
41627 * mprototype: RS/6000 and PowerPC Options.
41629 * mpt-fixed: SH Options. (line 215)
41630 * mpush-args <1>: i386 and x86-64 Options.
41632 * mpush-args: CRX Options. (line 13)
41633 * MQ: Preprocessor Options.
41635 * mr10k-cache-barrier: MIPS Options. (line 539)
41636 * mrecip: i386 and x86-64 Options.
41638 * mregister-names: IA-64 Options. (line 37)
41639 * mregnames: RS/6000 and PowerPC Options.
41641 * mregparm: i386 and x86-64 Options.
41643 * mrelax <1>: SH Options. (line 70)
41644 * mrelax <2>: MN10300 Options. (line 34)
41645 * mrelax: H8/300 Options. (line 9)
41646 * mrelax-immediate: MCore Options. (line 19)
41647 * mrelocatable: RS/6000 and PowerPC Options.
41649 * mrelocatable-lib: RS/6000 and PowerPC Options.
41651 * mreturn-pointer-on-d0: MN10300 Options. (line 24)
41652 * mrodata: ARC Options. (line 30)
41653 * mrtd <1>: Function Attributes.
41655 * mrtd <2>: M680x0 Options. (line 236)
41656 * mrtd: i386 and x86-64 Options.
41658 * mrtp: VxWorks Options. (line 11)
41659 * ms: H8/300 Options. (line 17)
41660 * ms2600: H8/300 Options. (line 24)
41661 * msafe-dma: SPU Options. (line 17)
41662 * msafe-hints: SPU Options. (line 72)
41663 * msahf: i386 and x86-64 Options.
41665 * mscc: FRV Options. (line 140)
41666 * msched-ar-data-spec: IA-64 Options. (line 128)
41667 * msched-ar-in-data-spec: IA-64 Options. (line 149)
41668 * msched-br-data-spec: IA-64 Options. (line 121)
41669 * msched-br-in-data-spec: IA-64 Options. (line 142)
41670 * msched-control-ldc: IA-64 Options. (line 168)
41671 * msched-control-spec: IA-64 Options. (line 135)
41672 * msched-costly-dep: RS/6000 and PowerPC Options.
41674 * msched-count-spec-in-critical-path: IA-64 Options. (line 194)
41675 * msched-in-control-spec: IA-64 Options. (line 156)
41676 * msched-ldc: IA-64 Options. (line 162)
41677 * msched-prefer-non-control-spec-insns: IA-64 Options. (line 187)
41678 * msched-prefer-non-data-spec-insns: IA-64 Options. (line 180)
41679 * msched-spec-verbose: IA-64 Options. (line 176)
41680 * mschedule: HPPA Options. (line 78)
41681 * mscore5: Score Options. (line 25)
41682 * mscore5u: Score Options. (line 28)
41683 * mscore7: Score Options. (line 31)
41684 * mscore7d: Score Options. (line 34)
41685 * msda: V850 Options. (line 40)
41686 * msdata <1>: RS/6000 and PowerPC Options.
41688 * msdata: IA-64 Options. (line 42)
41689 * msdata=data: RS/6000 and PowerPC Options.
41691 * msdata=default: RS/6000 and PowerPC Options.
41693 * msdata=eabi: RS/6000 and PowerPC Options.
41695 * msdata=none <1>: RS/6000 and PowerPC Options.
41697 * msdata=none: M32R/D Options. (line 40)
41698 * msdata=sdata: M32R/D Options. (line 49)
41699 * msdata=sysv: RS/6000 and PowerPC Options.
41701 * msdata=use: M32R/D Options. (line 53)
41702 * msdram: Blackfin Options. (line 162)
41703 * msecure-plt: RS/6000 and PowerPC Options.
41705 * msep-data: Blackfin Options. (line 105)
41706 * mserialize-volatile: Xtensa Options. (line 35)
41707 * mshared-library-id: Blackfin Options. (line 98)
41708 * mshort <1>: M68hc1x Options. (line 40)
41709 * mshort: M680x0 Options. (line 216)
41710 * msim <1>: Xstormy16 Options. (line 9)
41711 * msim <2>: RS/6000 and PowerPC Options.
41713 * msim <3>: M32C Options. (line 13)
41714 * msim: Blackfin Options. (line 32)
41715 * msimple-fpu: RS/6000 and PowerPC Options.
41717 * msingle-exit: MMIX Options. (line 66)
41718 * msingle-float <1>: RS/6000 and PowerPC Options.
41720 * msingle-float: MIPS Options. (line 232)
41721 * msingle-pic-base: ARM Options. (line 179)
41722 * msio: HPPA Options. (line 105)
41723 * msize: AVR Options. (line 32)
41724 * mslow-bytes: MCore Options. (line 35)
41725 * msmall-data: DEC Alpha Options. (line 195)
41726 * msmall-exec: S/390 and zSeries Options.
41728 * msmall-mem: SPU Options. (line 35)
41729 * msmall-model: FR30 Options. (line 9)
41730 * msmall-text: DEC Alpha Options. (line 213)
41731 * msmartmips: MIPS Options. (line 268)
41732 * msoft-float <1>: SPARC Options. (line 25)
41733 * msoft-float <2>: S/390 and zSeries Options.
41735 * msoft-float <3>: RS/6000 and PowerPC Options.
41737 * msoft-float <4>: PDP-11 Options. (line 13)
41738 * msoft-float <5>: MIPS Options. (line 228)
41739 * msoft-float <6>: M680x0 Options. (line 199)
41740 * msoft-float <7>: i386 and x86-64 Options.
41742 * msoft-float <8>: HPPA Options. (line 91)
41743 * msoft-float <9>: FRV Options. (line 22)
41744 * msoft-float <10>: DEC Alpha Options. (line 10)
41745 * msoft-float: ARM Options. (line 65)
41746 * msoft-quad-float: SPARC Options. (line 45)
41747 * msoft-reg-count: M68hc1x Options. (line 43)
41748 * mspace <1>: V850 Options. (line 30)
41749 * mspace: SH Options. (line 125)
41750 * mspe: RS/6000 and PowerPC Options.
41752 * mspecld-anomaly: Blackfin Options. (line 46)
41753 * msplit: PDP-11 Options. (line 68)
41754 * msplit-addresses: MIPS Options. (line 410)
41755 * msse: i386 and x86-64 Options.
41757 * msse2avx: i386 and x86-64 Options.
41759 * msseregparm: i386 and x86-64 Options.
41761 * mstack-align: CRIS Options. (line 55)
41762 * mstack-bias: SPARC Options. (line 222)
41763 * mstack-check-l1: Blackfin Options. (line 72)
41764 * mstack-guard: S/390 and zSeries Options.
41766 * mstack-increment: MCore Options. (line 50)
41767 * mstack-size: S/390 and zSeries Options.
41769 * mstackrealign: i386 and x86-64 Options.
41771 * mstdmain: SPU Options. (line 40)
41772 * mstrict-align <1>: RS/6000 and PowerPC Options.
41774 * mstrict-align: M680x0 Options. (line 283)
41775 * mstring: RS/6000 and PowerPC Options.
41777 * mstringop-strategy=ALG: i386 and x86-64 Options.
41779 * mstructure-size-boundary: ARM Options. (line 134)
41780 * msvr4-struct-return: RS/6000 and PowerPC Options.
41782 * mswdiv: RS/6000 and PowerPC Options.
41784 * msym32: MIPS Options. (line 307)
41785 * mt: IA-64 Options. (line 106)
41786 * MT: Preprocessor Options.
41788 * mtarget-align: Xtensa Options. (line 54)
41789 * mtda: V850 Options. (line 34)
41790 * mtext: ARC Options. (line 30)
41791 * mtext-section-literals: Xtensa Options. (line 42)
41792 * mthread: i386 and x86-64 Windows Options.
41794 * mthreads: i386 and x86-64 Options.
41796 * mthumb: ARM Options. (line 220)
41797 * mthumb-interwork: ARM Options. (line 25)
41798 * mtiny-stack: AVR Options. (line 52)
41799 * mtls-direct-seg-refs: i386 and x86-64 Options.
41801 * mtls-size: IA-64 Options. (line 97)
41802 * mtoc: RS/6000 and PowerPC Options.
41804 * mtomcat-stats: FRV Options. (line 209)
41805 * mtoplevel-symbols: MMIX Options. (line 40)
41806 * mtp: ARM Options. (line 250)
41807 * mtpcs-frame: ARM Options. (line 226)
41808 * mtpcs-leaf-frame: ARM Options. (line 232)
41809 * mtpf-trace: S/390 and zSeries Options.
41811 * mtrap-precision: DEC Alpha Options. (line 109)
41812 * mtune <1>: SPARC Options. (line 158)
41813 * mtune <2>: S/390 and zSeries Options.
41815 * mtune <3>: RS/6000 and PowerPC Options.
41817 * mtune <4>: MIPS Options. (line 61)
41818 * mtune <5>: M680x0 Options. (line 66)
41819 * mtune <6>: IA-64 Options. (line 101)
41820 * mtune <7>: i386 and x86-64 Options.
41822 * mtune <8>: DEC Alpha Options. (line 267)
41823 * mtune <9>: CRIS Options. (line 16)
41824 * mtune: ARM Options. (line 102)
41825 * muclibc: GNU/Linux Options. (line 13)
41826 * muls: Score Options. (line 18)
41827 * multcost=NUMBER: SH Options. (line 138)
41828 * multi_module: Darwin Options. (line 199)
41829 * multilib-library-pic: FRV Options. (line 89)
41830 * multiply_defined: Darwin Options. (line 199)
41831 * multiply_defined_unused: Darwin Options. (line 199)
41832 * munaligned-doubles: SPARC Options. (line 59)
41833 * muninit-const-in-rodata: MIPS Options. (line 380)
41834 * munix: VAX Options. (line 9)
41835 * munix-asm: PDP-11 Options. (line 74)
41836 * munsafe-dma: SPU Options. (line 17)
41837 * mupdate: RS/6000 and PowerPC Options.
41839 * musermode: SH Options. (line 133)
41840 * mv850: V850 Options. (line 49)
41841 * mv850e: V850 Options. (line 69)
41842 * mv850e1: V850 Options. (line 64)
41843 * mv8plus: SPARC Options. (line 170)
41844 * mveclibabi: i386 and x86-64 Options.
41846 * mvis: SPARC Options. (line 177)
41847 * mvliw-branch: FRV Options. (line 164)
41848 * mvms-return-codes: DEC Alpha/VMS Options.
41850 * mvolatile-asm-stop: IA-64 Options. (line 32)
41851 * mvr4130-align: MIPS Options. (line 638)
41852 * mvrsave: RS/6000 and PowerPC Options.
41854 * mvxworks: RS/6000 and PowerPC Options.
41856 * mwarn-cell-microcode: RS/6000 and PowerPC Options.
41858 * mwarn-dynamicstack: S/390 and zSeries Options.
41860 * mwarn-framesize: S/390 and zSeries Options.
41862 * mwarn-reloc: SPU Options. (line 10)
41863 * mwide-bitfields: MCore Options. (line 23)
41864 * mwin32: i386 and x86-64 Windows Options.
41866 * mwindows: i386 and x86-64 Windows Options.
41868 * mword-relocations: ARM Options. (line 258)
41869 * mwords-little-endian: ARM Options. (line 76)
41870 * mxgot <1>: MIPS Options. (line 190)
41871 * mxgot: M680x0 Options. (line 315)
41872 * mxilinx-fpu: RS/6000 and PowerPC Options.
41874 * mxl-compat: RS/6000 and PowerPC Options.
41876 * myellowknife: RS/6000 and PowerPC Options.
41878 * mzarch: S/390 and zSeries Options.
41880 * mzda: V850 Options. (line 45)
41881 * mzero-extend: MMIX Options. (line 27)
41882 * no-integrated-cpp: C Dialect Options. (line 240)
41883 * no-lsim: MCore Options. (line 46)
41884 * no-red-zone: i386 and x86-64 Options.
41886 * no_dead_strip_inits_and_terms: Darwin Options. (line 199)
41887 * noall_load: Darwin Options. (line 199)
41888 * nocpp: MIPS Options. (line 476)
41889 * nodefaultlibs: Link Options. (line 62)
41890 * nofixprebinding: Darwin Options. (line 199)
41891 * nolibdld: HPPA Options. (line 188)
41892 * nomultidefs: Darwin Options. (line 199)
41893 * non-static: VxWorks Options. (line 16)
41894 * noprebind: Darwin Options. (line 199)
41895 * noseglinkedit: Darwin Options. (line 199)
41896 * nostartfiles: Link Options. (line 57)
41897 * nostdinc: Preprocessor Options.
41899 * nostdinc++ <1>: Preprocessor Options.
41901 * nostdinc++: C++ Dialect Options.
41903 * nostdlib: Link Options. (line 71)
41904 * o: Preprocessor Options.
41906 * O: Optimize Options. (line 29)
41907 * o: Overall Options. (line 187)
41908 * O0: Optimize Options. (line 106)
41909 * O1: Optimize Options. (line 29)
41910 * O2: Optimize Options. (line 67)
41911 * O3: Optimize Options. (line 100)
41912 * Os: Optimize Options. (line 110)
41913 * P: Preprocessor Options.
41915 * p: Debugging Options. (line 227)
41916 * pagezero_size: Darwin Options. (line 199)
41917 * param: Optimize Options. (line 1702)
41918 * pass-exit-codes: Overall Options. (line 145)
41919 * pedantic <1>: Warnings and Errors.
41921 * pedantic <2>: Alternate Keywords. (line 29)
41922 * pedantic <3>: C Extensions. (line 6)
41923 * pedantic <4>: Preprocessor Options.
41925 * pedantic <5>: Warning Options. (line 53)
41926 * pedantic: Standards. (line 16)
41927 * pedantic-errors <1>: Warnings and Errors.
41929 * pedantic-errors <2>: Non-bugs. (line 216)
41930 * pedantic-errors <3>: Preprocessor Options.
41932 * pedantic-errors <4>: Warning Options. (line 95)
41933 * pedantic-errors: Standards. (line 16)
41934 * pg: Debugging Options. (line 233)
41935 * pie: Link Options. (line 92)
41936 * pipe: Overall Options. (line 209)
41937 * prebind: Darwin Options. (line 199)
41938 * prebind_all_twolevel_modules: Darwin Options. (line 199)
41939 * preprocessor: Preprocessor Options.
41941 * print-file-name: Debugging Options. (line 898)
41942 * print-libgcc-file-name: Debugging Options. (line 919)
41943 * print-multi-directory: Debugging Options. (line 904)
41944 * print-multi-lib: Debugging Options. (line 909)
41945 * print-objc-runtime-info: Objective-C and Objective-C++ Dialect Options.
41947 * print-prog-name: Debugging Options. (line 916)
41948 * print-search-dirs: Debugging Options. (line 927)
41949 * print-sysroot: Debugging Options. (line 940)
41950 * print-sysroot-headers-suffix: Debugging Options. (line 947)
41951 * private_bundle: Darwin Options. (line 199)
41952 * pthread <1>: SPARC Options. (line 242)
41953 * pthread <2>: RS/6000 and PowerPC Options.
41955 * pthread: IA-64 Options. (line 106)
41956 * pthreads: SPARC Options. (line 236)
41957 * Q: Debugging Options. (line 239)
41958 * Qn: System V Options. (line 18)
41959 * Qy: System V Options. (line 14)
41960 * rdynamic: Link Options. (line 98)
41961 * read_only_relocs: Darwin Options. (line 199)
41962 * remap: Preprocessor Options.
41964 * s: Link Options. (line 105)
41965 * S <1>: Link Options. (line 20)
41966 * S: Overall Options. (line 170)
41967 * save-temps: Debugging Options. (line 860)
41968 * sectalign: Darwin Options. (line 199)
41969 * sectcreate: Darwin Options. (line 199)
41970 * sectobjectsymbols: Darwin Options. (line 199)
41971 * sectorder: Darwin Options. (line 199)
41972 * seg1addr: Darwin Options. (line 199)
41973 * seg_addr_table: Darwin Options. (line 199)
41974 * seg_addr_table_filename: Darwin Options. (line 199)
41975 * segaddr: Darwin Options. (line 199)
41976 * seglinkedit: Darwin Options. (line 199)
41977 * segprot: Darwin Options. (line 199)
41978 * segs_read_only_addr: Darwin Options. (line 199)
41979 * segs_read_write_addr: Darwin Options. (line 199)
41980 * shared: Link Options. (line 114)
41981 * shared-libgcc: Link Options. (line 122)
41982 * sim: CRIS Options. (line 95)
41983 * sim2: CRIS Options. (line 101)
41984 * single_module: Darwin Options. (line 199)
41985 * specs: Directory Options. (line 84)
41986 * static <1>: HPPA Options. (line 192)
41987 * static <2>: Darwin Options. (line 199)
41988 * static: Link Options. (line 109)
41989 * static-libgcc: Link Options. (line 122)
41990 * std <1>: Non-bugs. (line 107)
41991 * std <2>: Other Builtins. (line 22)
41992 * std <3>: C Dialect Options. (line 47)
41993 * std: Standards. (line 16)
41994 * std=: Preprocessor Options.
41996 * sub_library: Darwin Options. (line 199)
41997 * sub_umbrella: Darwin Options. (line 199)
41998 * symbolic: Link Options. (line 157)
41999 * sysroot: Directory Options. (line 92)
42000 * T: Link Options. (line 163)
42001 * target-help <1>: Preprocessor Options.
42003 * target-help: Overall Options. (line 240)
42004 * threads <1>: SPARC Options. (line 230)
42005 * threads: HPPA Options. (line 205)
42006 * time: Debugging Options. (line 874)
42007 * tls: FRV Options. (line 75)
42008 * TLS: FRV Options. (line 72)
42009 * traditional <1>: Incompatibilities. (line 6)
42010 * traditional: C Dialect Options. (line 252)
42011 * traditional-cpp <1>: Preprocessor Options.
42013 * traditional-cpp: C Dialect Options. (line 252)
42014 * trigraphs <1>: Preprocessor Options.
42016 * trigraphs: C Dialect Options. (line 236)
42017 * twolevel_namespace: Darwin Options. (line 199)
42018 * u: Link Options. (line 196)
42019 * U: Preprocessor Options.
42021 * umbrella: Darwin Options. (line 199)
42022 * undef: Preprocessor Options.
42024 * undefined: Darwin Options. (line 199)
42025 * unexported_symbols_list: Darwin Options. (line 199)
42026 * V: Target Options. (line 25)
42027 * v <1>: Preprocessor Options.
42029 * v: Overall Options. (line 198)
42030 * version <1>: Preprocessor Options.
42032 * version: Overall Options. (line 348)
42033 * W: Incompatibilities. (line 64)
42034 * w: Preprocessor Options.
42036 * W: Warning Options. (line 146)
42037 * w: Warning Options. (line 18)
42038 * Wa: Assembler Options. (line 9)
42039 * Wabi: C++ Dialect Options.
42041 * Waddress: Warning Options. (line 953)
42042 * Waggregate-return: Warning Options. (line 971)
42043 * Wall <1>: Standard Libraries. (line 6)
42044 * Wall <2>: Preprocessor Options.
42046 * Wall: Warning Options. (line 99)
42047 * Warray-bounds: Warning Options. (line 691)
42048 * Wassign-intercept: Objective-C and Objective-C++ Dialect Options.
42050 * Wattributes: Warning Options. (line 976)
42051 * Wbad-function-cast: Warning Options. (line 869)
42052 * Wbuiltin-macro-redefined: Warning Options. (line 982)
42053 * Wcast-align: Warning Options. (line 889)
42054 * Wcast-qual: Warning Options. (line 884)
42055 * Wchar-subscripts: Warning Options. (line 184)
42056 * Wclobbered: Warning Options. (line 909)
42057 * Wcomment <1>: Preprocessor Options.
42059 * Wcomment: Warning Options. (line 189)
42060 * Wcomments: Preprocessor Options.
42062 * Wconversion: Warning Options. (line 913)
42063 * Wcoverage-mismatch: Language Independent Options.
42065 * Wctor-dtor-privacy: C++ Dialect Options.
42067 * Wdeclaration-after-statement: Warning Options. (line 812)
42068 * Wdeprecated: Warning Options. (line 1119)
42069 * Wdeprecated-declarations: Warning Options. (line 1123)
42070 * Wdisabled-optimization: Warning Options. (line 1272)
42071 * Wdiv-by-zero: Warning Options. (line 696)
42072 * weak_reference_mismatches: Darwin Options. (line 199)
42073 * Weffc++: C++ Dialect Options.
42075 * Wempty-body: Warning Options. (line 932)
42076 * Wendif-labels <1>: Preprocessor Options.
42078 * Wendif-labels: Warning Options. (line 822)
42079 * Wenum-compare: Warning Options. (line 936)
42080 * Werror <1>: Preprocessor Options.
42082 * Werror: Warning Options. (line 21)
42083 * Werror=: Warning Options. (line 24)
42084 * Wextra: Warning Options. (line 146)
42085 * Wfatal-errors: Warning Options. (line 38)
42086 * Wfloat-equal: Warning Options. (line 712)
42087 * Wformat <1>: Function Attributes.
42089 * Wformat: Warning Options. (line 194)
42090 * Wformat-contains-nul: Warning Options. (line 233)
42091 * Wformat-extra-args: Warning Options. (line 237)
42092 * Wformat-nonliteral <1>: Function Attributes.
42094 * Wformat-nonliteral: Warning Options. (line 255)
42095 * Wformat-security: Warning Options. (line 260)
42096 * Wformat-y2k: Warning Options. (line 229)
42097 * Wformat-zero-length: Warning Options. (line 251)
42098 * Wformat=2: Warning Options. (line 271)
42099 * Wframe-larger-than: Warning Options. (line 834)
42100 * whatsloaded: Darwin Options. (line 199)
42101 * whyload: Darwin Options. (line 199)
42102 * Wignored-qualifiers: Warning Options. (line 310)
42103 * Wimplicit: Warning Options. (line 306)
42104 * Wimplicit-function-declaration: Warning Options. (line 300)
42105 * Wimplicit-int: Warning Options. (line 296)
42106 * Winit-self: Warning Options. (line 283)
42107 * Winline <1>: Inline. (line 63)
42108 * Winline: Warning Options. (line 1211)
42109 * Wint-to-pointer-cast: Warning Options. (line 1238)
42110 * Winvalid-offsetof: Warning Options. (line 1224)
42111 * Winvalid-pch: Warning Options. (line 1246)
42112 * Wl: Link Options. (line 188)
42113 * Wlarger-than-LEN: Warning Options. (line 831)
42114 * Wlarger-than=LEN: Warning Options. (line 831)
42115 * Wlogical-op: Warning Options. (line 966)
42116 * Wlong-long: Warning Options. (line 1250)
42117 * Wmain: Warning Options. (line 321)
42118 * Wmissing-braces: Warning Options. (line 328)
42119 * Wmissing-declarations: Warning Options. (line 1017)
42120 * Wmissing-field-initializers: Warning Options. (line 1025)
42121 * Wmissing-format-attribute: Warning Options. (line 1051)
42122 * Wmissing-include-dirs: Warning Options. (line 338)
42123 * Wmissing-noreturn: Warning Options. (line 1043)
42124 * Wmissing-parameter-type: Warning Options. (line 1003)
42125 * Wmissing-prototypes: Warning Options. (line 1011)
42126 * Wmultichar: Warning Options. (line 1070)
42127 * Wnested-externs: Warning Options. (line 1186)
42128 * Wno-abi: C++ Dialect Options.
42130 * Wno-address: Warning Options. (line 953)
42131 * Wno-aggregate-return: Warning Options. (line 971)
42132 * Wno-all: Warning Options. (line 99)
42133 * Wno-array-bounds: Warning Options. (line 691)
42134 * Wno-assign-intercept: Objective-C and Objective-C++ Dialect Options.
42136 * Wno-attributes: Warning Options. (line 976)
42137 * Wno-bad-function-cast: Warning Options. (line 869)
42138 * Wno-builtin-macro-redefined: Warning Options. (line 982)
42139 * Wno-cast-align: Warning Options. (line 889)
42140 * Wno-cast-qual: Warning Options. (line 884)
42141 * Wno-char-subscripts: Warning Options. (line 184)
42142 * Wno-clobbered: Warning Options. (line 909)
42143 * Wno-comment: Warning Options. (line 189)
42144 * Wno-conversion: Warning Options. (line 913)
42145 * Wno-ctor-dtor-privacy: C++ Dialect Options.
42147 * Wno-declaration-after-statement: Warning Options. (line 812)
42148 * Wno-deprecated: Warning Options. (line 1119)
42149 * Wno-deprecated-declarations: Warning Options. (line 1123)
42150 * Wno-disabled-optimization: Warning Options. (line 1272)
42151 * Wno-div-by-zero: Warning Options. (line 696)
42152 * Wno-effc++: C++ Dialect Options.
42154 * Wno-empty-body: Warning Options. (line 932)
42155 * Wno-endif-labels: Warning Options. (line 822)
42156 * Wno-enum-compare: Warning Options. (line 936)
42157 * Wno-error: Warning Options. (line 21)
42158 * Wno-error=: Warning Options. (line 24)
42159 * Wno-extra: Warning Options. (line 146)
42160 * Wno-fatal-errors: Warning Options. (line 38)
42161 * Wno-float-equal: Warning Options. (line 712)
42162 * Wno-format: Warning Options. (line 194)
42163 * Wno-format-contains-nul: Warning Options. (line 233)
42164 * Wno-format-extra-args: Warning Options. (line 237)
42165 * Wno-format-nonliteral: Warning Options. (line 255)
42166 * Wno-format-security: Warning Options. (line 260)
42167 * Wno-format-y2k: Warning Options. (line 229)
42168 * Wno-format-zero-length: Warning Options. (line 251)
42169 * Wno-format=2: Warning Options. (line 271)
42170 * Wno-ignored-qualifiers: Warning Options. (line 310)
42171 * Wno-implicit: Warning Options. (line 306)
42172 * Wno-implicit-function-declaration: Warning Options. (line 300)
42173 * Wno-implicit-int: Warning Options. (line 296)
42174 * Wno-init-self: Warning Options. (line 283)
42175 * Wno-inline: Warning Options. (line 1211)
42176 * Wno-int-to-pointer-cast: Warning Options. (line 1238)
42177 * Wno-invalid-offsetof: Warning Options. (line 1224)
42178 * Wno-invalid-pch: Warning Options. (line 1246)
42179 * Wno-logical-op: Warning Options. (line 966)
42180 * Wno-long-long: Warning Options. (line 1250)
42181 * Wno-main: Warning Options. (line 321)
42182 * Wno-missing-braces: Warning Options. (line 328)
42183 * Wno-missing-declarations: Warning Options. (line 1017)
42184 * Wno-missing-field-initializers: Warning Options. (line 1025)
42185 * Wno-missing-format-attribute: Warning Options. (line 1051)
42186 * Wno-missing-include-dirs: Warning Options. (line 338)
42187 * Wno-missing-noreturn: Warning Options. (line 1043)
42188 * Wno-missing-parameter-type: Warning Options. (line 1003)
42189 * Wno-missing-prototypes: Warning Options. (line 1011)
42190 * Wno-mudflap: Warning Options. (line 1292)
42191 * Wno-multichar: Warning Options. (line 1070)
42192 * Wno-nested-externs: Warning Options. (line 1186)
42193 * Wno-non-template-friend: C++ Dialect Options.
42195 * Wno-non-virtual-dtor: C++ Dialect Options.
42197 * Wno-nonnull: Warning Options. (line 276)
42198 * Wno-old-style-cast: C++ Dialect Options.
42200 * Wno-old-style-declaration: Warning Options. (line 993)
42201 * Wno-old-style-definition: Warning Options. (line 999)
42202 * Wno-overflow: Warning Options. (line 1129)
42203 * Wno-overlength-strings: Warning Options. (line 1296)
42204 * Wno-overloaded-virtual: C++ Dialect Options.
42206 * Wno-override-init: Warning Options. (line 1132)
42207 * Wno-packed: Warning Options. (line 1140)
42208 * Wno-packed-bitfield-compat: Warning Options. (line 1157)
42209 * Wno-padded: Warning Options. (line 1174)
42210 * Wno-parentheses: Warning Options. (line 341)
42211 * Wno-pedantic-ms-format: Warning Options. (line 849)
42212 * Wno-pmf-conversions <1>: Bound member functions.
42214 * Wno-pmf-conversions: C++ Dialect Options.
42216 * Wno-pointer-arith: Warning Options. (line 855)
42217 * Wno-pointer-sign: Warning Options. (line 1281)
42218 * Wno-pointer-to-int-cast: Warning Options. (line 1242)
42219 * Wno-pragmas: Warning Options. (line 594)
42220 * Wno-protocol: Objective-C and Objective-C++ Dialect Options.
42222 * Wno-redundant-decls: Warning Options. (line 1181)
42223 * Wno-reorder: C++ Dialect Options.
42225 * Wno-return-type: Warning Options. (line 431)
42226 * Wno-selector: Objective-C and Objective-C++ Dialect Options.
42228 * Wno-sequence-point: Warning Options. (line 385)
42229 * Wno-shadow: Warning Options. (line 826)
42230 * Wno-sign-compare: Warning Options. (line 940)
42231 * Wno-sign-conversion: Warning Options. (line 947)
42232 * Wno-sign-promo: C++ Dialect Options.
42234 * Wno-stack-protector: Warning Options. (line 1287)
42235 * Wno-strict-aliasing: Warning Options. (line 599)
42236 * Wno-strict-aliasing=n: Warning Options. (line 607)
42237 * Wno-strict-null-sentinel: C++ Dialect Options.
42239 * Wno-strict-overflow: Warning Options. (line 640)
42240 * Wno-strict-prototypes: Warning Options. (line 987)
42241 * Wno-strict-selector-match: Objective-C and Objective-C++ Dialect Options.
42243 * Wno-switch: Warning Options. (line 446)
42244 * Wno-switch-default: Warning Options. (line 454)
42245 * Wno-switch-enum: Warning Options. (line 457)
42246 * Wno-sync-nand: Warning Options. (line 463)
42247 * Wno-system-headers: Warning Options. (line 701)
42248 * Wno-traditional: Warning Options. (line 727)
42249 * Wno-traditional-conversion: Warning Options. (line 804)
42250 * Wno-trigraphs: Warning Options. (line 468)
42251 * Wno-type-limits: Warning Options. (line 862)
42252 * Wno-undeclared-selector: Objective-C and Objective-C++ Dialect Options.
42254 * Wno-undef: Warning Options. (line 819)
42255 * Wno-uninitialized: Warning Options. (line 517)
42256 * Wno-unknown-pragmas: Warning Options. (line 587)
42257 * Wno-unreachable-code: Warning Options. (line 1189)
42258 * Wno-unsafe-loop-optimizations: Warning Options. (line 843)
42259 * Wno-unused: Warning Options. (line 510)
42260 * Wno-unused-function: Warning Options. (line 473)
42261 * Wno-unused-label: Warning Options. (line 478)
42262 * Wno-unused-parameter: Warning Options. (line 485)
42263 * Wno-unused-value: Warning Options. (line 500)
42264 * Wno-unused-variable: Warning Options. (line 492)
42265 * Wno-variadic-macros: Warning Options. (line 1256)
42266 * Wno-vla: Warning Options. (line 1262)
42267 * Wno-volatile-register-var: Warning Options. (line 1266)
42268 * Wno-write-strings: Warning Options. (line 895)
42269 * Wnon-template-friend: C++ Dialect Options.
42271 * Wnon-virtual-dtor: C++ Dialect Options.
42273 * Wnonnull: Warning Options. (line 276)
42274 * Wnormalized=: Warning Options. (line 1076)
42275 * Wold-style-cast: C++ Dialect Options.
42277 * Wold-style-declaration: Warning Options. (line 993)
42278 * Wold-style-definition: Warning Options. (line 999)
42279 * Woverflow: Warning Options. (line 1129)
42280 * Woverlength-strings: Warning Options. (line 1296)
42281 * Woverloaded-virtual: C++ Dialect Options.
42283 * Woverride-init: Warning Options. (line 1132)
42284 * Wp: Preprocessor Options.
42286 * Wpacked: Warning Options. (line 1140)
42287 * Wpacked-bitfield-compat: Warning Options. (line 1157)
42288 * Wpadded: Warning Options. (line 1174)
42289 * Wparentheses: Warning Options. (line 341)
42290 * Wpedantic-ms-format: Warning Options. (line 849)
42291 * Wpmf-conversions: C++ Dialect Options.
42293 * Wpointer-arith <1>: Pointer Arith. (line 13)
42294 * Wpointer-arith: Warning Options. (line 855)
42295 * Wpointer-sign: Warning Options. (line 1281)
42296 * Wpointer-to-int-cast: Warning Options. (line 1242)
42297 * Wpragmas: Warning Options. (line 594)
42298 * Wprotocol: Objective-C and Objective-C++ Dialect Options.
42300 * wrapper: Overall Options. (line 351)
42301 * Wredundant-decls: Warning Options. (line 1181)
42302 * Wreorder: C++ Dialect Options.
42304 * Wreturn-type: Warning Options. (line 431)
42305 * Wselector: Objective-C and Objective-C++ Dialect Options.
42307 * Wsequence-point: Warning Options. (line 385)
42308 * Wshadow: Warning Options. (line 826)
42309 * Wsign-compare: Warning Options. (line 940)
42310 * Wsign-conversion: Warning Options. (line 947)
42311 * Wsign-promo: C++ Dialect Options.
42313 * Wstack-protector: Warning Options. (line 1287)
42314 * Wstrict-aliasing: Warning Options. (line 599)
42315 * Wstrict-aliasing=n: Warning Options. (line 607)
42316 * Wstrict-null-sentinel: C++ Dialect Options.
42318 * Wstrict-overflow: Warning Options. (line 640)
42319 * Wstrict-prototypes: Warning Options. (line 987)
42320 * Wstrict-selector-match: Objective-C and Objective-C++ Dialect Options.
42322 * Wswitch: Warning Options. (line 446)
42323 * Wswitch-default: Warning Options. (line 454)
42324 * Wswitch-enum: Warning Options. (line 457)
42325 * Wsync-nand: Warning Options. (line 463)
42326 * Wsystem-headers <1>: Preprocessor Options.
42328 * Wsystem-headers: Warning Options. (line 701)
42329 * Wtraditional <1>: Preprocessor Options.
42331 * Wtraditional: Warning Options. (line 727)
42332 * Wtraditional-conversion <1>: Protoize Caveats. (line 31)
42333 * Wtraditional-conversion: Warning Options. (line 804)
42334 * Wtrigraphs <1>: Preprocessor Options.
42336 * Wtrigraphs: Warning Options. (line 468)
42337 * Wtype-limits: Warning Options. (line 862)
42338 * Wundeclared-selector: Objective-C and Objective-C++ Dialect Options.
42340 * Wundef <1>: Preprocessor Options.
42342 * Wundef: Warning Options. (line 819)
42343 * Wuninitialized: Warning Options. (line 517)
42344 * Wunknown-pragmas: Warning Options. (line 587)
42345 * Wunreachable-code: Warning Options. (line 1189)
42346 * Wunsafe-loop-optimizations: Warning Options. (line 843)
42347 * Wunused: Warning Options. (line 510)
42348 * Wunused-function: Warning Options. (line 473)
42349 * Wunused-label: Warning Options. (line 478)
42350 * Wunused-macros: Preprocessor Options.
42352 * Wunused-parameter: Warning Options. (line 485)
42353 * Wunused-value: Warning Options. (line 500)
42354 * Wunused-variable: Warning Options. (line 492)
42355 * Wvariadic-macros: Warning Options. (line 1256)
42356 * Wvla: Warning Options. (line 1262)
42357 * Wvolatile-register-var: Warning Options. (line 1266)
42358 * Wwrite-strings: Warning Options. (line 895)
42359 * x <1>: Preprocessor Options.
42361 * x: Overall Options. (line 122)
42362 * Xassembler: Assembler Options. (line 13)
42363 * Xbind-lazy: VxWorks Options. (line 26)
42364 * Xbind-now: VxWorks Options. (line 30)
42365 * Xlinker: Link Options. (line 169)
42366 * Ym: System V Options. (line 26)
42367 * YP: System V Options. (line 22)
42370 File: gcc.info, Node: Keyword Index, Prev: Option Index, Up: Top
42378 * ! in constraint: Multi-Alternative. (line 33)
42379 * # in constraint: Modifiers. (line 57)
42380 * #pragma: Pragmas. (line 6)
42381 * #pragma implementation: C++ Interface. (line 39)
42382 * #pragma implementation, implied: C++ Interface. (line 46)
42383 * #pragma interface: C++ Interface. (line 20)
42384 * #pragma, reason for not using: Function Attributes.
42386 * $: Dollar Signs. (line 6)
42387 * % in constraint: Modifiers. (line 45)
42388 * %include: Spec Files. (line 27)
42389 * %include_noerr: Spec Files. (line 31)
42390 * %rename: Spec Files. (line 35)
42391 * & in constraint: Modifiers. (line 25)
42392 * ': Incompatibilities. (line 116)
42393 * (: Constructing Calls. (line 53)
42394 * * in constraint: Modifiers. (line 62)
42395 * + in constraint: Modifiers. (line 12)
42396 * -lgcc, use with -nodefaultlibs: Link Options. (line 79)
42397 * -lgcc, use with -nostdlib: Link Options. (line 79)
42398 * -nodefaultlibs and unresolved references: Link Options. (line 79)
42399 * -nostdlib and unresolved references: Link Options. (line 79)
42400 * .sdata/.sdata2 references (PowerPC): RS/6000 and PowerPC Options.
42402 * //: C++ Comments. (line 6)
42403 * 0 in constraint: Simple Constraints. (line 117)
42404 * < in constraint: Simple Constraints. (line 48)
42405 * = in constraint: Modifiers. (line 8)
42406 * > in constraint: Simple Constraints. (line 52)
42407 * ? in constraint: Multi-Alternative. (line 27)
42408 * ?: extensions: Conditionals. (line 6)
42409 * ?: side effect: Conditionals. (line 20)
42410 * _ in variables in macros: Typeof. (line 42)
42411 * __builtin___clear_cache: Other Builtins. (line 274)
42412 * __builtin___fprintf_chk: Object Size Checking.
42414 * __builtin___memcpy_chk: Object Size Checking.
42416 * __builtin___memmove_chk: Object Size Checking.
42418 * __builtin___mempcpy_chk: Object Size Checking.
42420 * __builtin___memset_chk: Object Size Checking.
42422 * __builtin___printf_chk: Object Size Checking.
42424 * __builtin___snprintf_chk: Object Size Checking.
42426 * __builtin___sprintf_chk: Object Size Checking.
42428 * __builtin___stpcpy_chk: Object Size Checking.
42430 * __builtin___strcat_chk: Object Size Checking.
42432 * __builtin___strcpy_chk: Object Size Checking.
42434 * __builtin___strncat_chk: Object Size Checking.
42436 * __builtin___strncpy_chk: Object Size Checking.
42438 * __builtin___vfprintf_chk: Object Size Checking.
42440 * __builtin___vprintf_chk: Object Size Checking.
42442 * __builtin___vsnprintf_chk: Object Size Checking.
42444 * __builtin___vsprintf_chk: Object Size Checking.
42446 * __builtin_apply: Constructing Calls. (line 31)
42447 * __builtin_apply_args: Constructing Calls. (line 20)
42448 * __builtin_bswap32: Other Builtins. (line 493)
42449 * __builtin_bswap64: Other Builtins. (line 498)
42450 * __builtin_choose_expr: Other Builtins. (line 156)
42451 * __builtin_clz: Other Builtins. (line 426)
42452 * __builtin_clzl: Other Builtins. (line 444)
42453 * __builtin_clzll: Other Builtins. (line 464)
42454 * __builtin_constant_p: Other Builtins. (line 196)
42455 * __builtin_ctz: Other Builtins. (line 430)
42456 * __builtin_ctzl: Other Builtins. (line 448)
42457 * __builtin_ctzll: Other Builtins. (line 468)
42458 * __builtin_expect: Other Builtins. (line 242)
42459 * __builtin_ffs: Other Builtins. (line 422)
42460 * __builtin_ffsl: Other Builtins. (line 440)
42461 * __builtin_ffsll: Other Builtins. (line 460)
42462 * __builtin_fpclassify: Other Builtins. (line 6)
42463 * __builtin_frame_address: Return Address. (line 34)
42464 * __builtin_huge_val: Other Builtins. (line 325)
42465 * __builtin_huge_valf: Other Builtins. (line 330)
42466 * __builtin_huge_vall: Other Builtins. (line 333)
42467 * __builtin_inf: Other Builtins. (line 348)
42468 * __builtin_infd128: Other Builtins. (line 358)
42469 * __builtin_infd32: Other Builtins. (line 352)
42470 * __builtin_infd64: Other Builtins. (line 355)
42471 * __builtin_inff: Other Builtins. (line 362)
42472 * __builtin_infl: Other Builtins. (line 367)
42473 * __builtin_isfinite: Other Builtins. (line 6)
42474 * __builtin_isgreater: Other Builtins. (line 6)
42475 * __builtin_isgreaterequal: Other Builtins. (line 6)
42476 * __builtin_isinf_sign: Other Builtins. (line 6)
42477 * __builtin_isless: Other Builtins. (line 6)
42478 * __builtin_islessequal: Other Builtins. (line 6)
42479 * __builtin_islessgreater: Other Builtins. (line 6)
42480 * __builtin_isnormal: Other Builtins. (line 6)
42481 * __builtin_isunordered: Other Builtins. (line 6)
42482 * __builtin_nan: Other Builtins. (line 378)
42483 * __builtin_nand128: Other Builtins. (line 400)
42484 * __builtin_nand32: Other Builtins. (line 394)
42485 * __builtin_nand64: Other Builtins. (line 397)
42486 * __builtin_nanf: Other Builtins. (line 404)
42487 * __builtin_nanl: Other Builtins. (line 407)
42488 * __builtin_nans: Other Builtins. (line 411)
42489 * __builtin_nansf: Other Builtins. (line 415)
42490 * __builtin_nansl: Other Builtins. (line 418)
42491 * __builtin_object_size: Object Size Checking.
42493 * __builtin_offsetof: Offsetof. (line 6)
42494 * __builtin_parity: Other Builtins. (line 437)
42495 * __builtin_parityl: Other Builtins. (line 456)
42496 * __builtin_parityll: Other Builtins. (line 476)
42497 * __builtin_popcount: Other Builtins. (line 434)
42498 * __builtin_popcountl: Other Builtins. (line 452)
42499 * __builtin_popcountll: Other Builtins. (line 472)
42500 * __builtin_powi: Other Builtins. (line 6)
42501 * __builtin_powif: Other Builtins. (line 6)
42502 * __builtin_powil: Other Builtins. (line 6)
42503 * __builtin_prefetch: Other Builtins. (line 286)
42504 * __builtin_return: Constructing Calls. (line 48)
42505 * __builtin_return_address: Return Address. (line 11)
42506 * __builtin_trap: Other Builtins. (line 266)
42507 * __builtin_types_compatible_p: Other Builtins. (line 110)
42508 * __complex__ keyword: Complex. (line 6)
42509 * __declspec(dllexport): Function Attributes.
42511 * __declspec(dllimport): Function Attributes.
42513 * __extension__: Alternate Keywords. (line 29)
42514 * __float128 data type: Floating Types. (line 6)
42515 * __float80 data type: Floating Types. (line 6)
42516 * __func__ identifier: Function Names. (line 6)
42517 * __FUNCTION__ identifier: Function Names. (line 6)
42518 * __imag__ keyword: Complex. (line 27)
42519 * __PRETTY_FUNCTION__ identifier: Function Names. (line 6)
42520 * __real__ keyword: Complex. (line 27)
42521 * __STDC_HOSTED__: Standards. (line 13)
42522 * __sync_add_and_fetch: Atomic Builtins. (line 61)
42523 * __sync_and_and_fetch: Atomic Builtins. (line 61)
42524 * __sync_bool_compare_and_swap: Atomic Builtins. (line 73)
42525 * __sync_fetch_and_add: Atomic Builtins. (line 45)
42526 * __sync_fetch_and_and: Atomic Builtins. (line 45)
42527 * __sync_fetch_and_nand: Atomic Builtins. (line 45)
42528 * __sync_fetch_and_or: Atomic Builtins. (line 45)
42529 * __sync_fetch_and_sub: Atomic Builtins. (line 45)
42530 * __sync_fetch_and_xor: Atomic Builtins. (line 45)
42531 * __sync_lock_release: Atomic Builtins. (line 103)
42532 * __sync_lock_test_and_set: Atomic Builtins. (line 85)
42533 * __sync_nand_and_fetch: Atomic Builtins. (line 61)
42534 * __sync_or_and_fetch: Atomic Builtins. (line 61)
42535 * __sync_sub_and_fetch: Atomic Builtins. (line 61)
42536 * __sync_synchronize: Atomic Builtins. (line 82)
42537 * __sync_val_compare_and_swap: Atomic Builtins. (line 73)
42538 * __sync_xor_and_fetch: Atomic Builtins. (line 61)
42539 * __thread: Thread-Local. (line 6)
42540 * _Accum data type: Fixed-Point. (line 6)
42541 * _Complex keyword: Complex. (line 6)
42542 * _Decimal128 data type: Decimal Float. (line 6)
42543 * _Decimal32 data type: Decimal Float. (line 6)
42544 * _Decimal64 data type: Decimal Float. (line 6)
42545 * _exit: Other Builtins. (line 6)
42546 * _Exit: Other Builtins. (line 6)
42547 * _Fract data type: Fixed-Point. (line 6)
42548 * _Sat data type: Fixed-Point. (line 6)
42549 * ABI: Compatibility. (line 6)
42550 * abort: Other Builtins. (line 6)
42551 * abs: Other Builtins. (line 6)
42552 * accessing volatiles: Volatiles. (line 6)
42553 * acos: Other Builtins. (line 6)
42554 * acosf: Other Builtins. (line 6)
42555 * acosh: Other Builtins. (line 6)
42556 * acoshf: Other Builtins. (line 6)
42557 * acoshl: Other Builtins. (line 6)
42558 * acosl: Other Builtins. (line 6)
42559 * Ada: G++ and GCC. (line 6)
42560 * additional floating types: Floating Types. (line 6)
42561 * address constraints: Simple Constraints. (line 144)
42562 * address of a label: Labels as Values. (line 6)
42563 * address_operand: Simple Constraints. (line 148)
42564 * alias attribute: Function Attributes.
42566 * aliasing of parameters: Code Gen Options. (line 409)
42567 * aligned attribute <1>: Type Attributes. (line 31)
42568 * aligned attribute <2>: Variable Attributes.
42570 * aligned attribute: Function Attributes.
42572 * alignment: Alignment. (line 6)
42573 * alloc_size attribute: Function Attributes.
42575 * alloca: Other Builtins. (line 6)
42576 * alloca vs variable-length arrays: Variable Length. (line 27)
42577 * Allow nesting in an interrupt handler on the Blackfin processor.: Function Attributes.
42579 * alternate keywords: Alternate Keywords. (line 6)
42580 * always_inline function attribute: Function Attributes.
42582 * AMD x86-64 Options: i386 and x86-64 Options.
42584 * AMD1: Standards. (line 13)
42585 * ANSI C: Standards. (line 13)
42586 * ANSI C standard: Standards. (line 13)
42587 * ANSI C89: Standards. (line 13)
42588 * ANSI support: C Dialect Options. (line 10)
42589 * ANSI X3.159-1989: Standards. (line 13)
42590 * apostrophes: Incompatibilities. (line 116)
42591 * application binary interface: Compatibility. (line 6)
42592 * ARC Options: ARC Options. (line 6)
42593 * ARM [Annotated C++ Reference Manual]: Backwards Compatibility.
42595 * ARM options: ARM Options. (line 6)
42596 * arrays of length zero: Zero Length. (line 6)
42597 * arrays of variable length: Variable Length. (line 6)
42598 * arrays, non-lvalue: Subscripting. (line 6)
42599 * artificial function attribute: Function Attributes.
42601 * asin: Other Builtins. (line 6)
42602 * asinf: Other Builtins. (line 6)
42603 * asinh: Other Builtins. (line 6)
42604 * asinhf: Other Builtins. (line 6)
42605 * asinhl: Other Builtins. (line 6)
42606 * asinl: Other Builtins. (line 6)
42607 * asm constraints: Constraints. (line 6)
42608 * asm expressions: Extended Asm. (line 6)
42609 * assembler instructions: Extended Asm. (line 6)
42610 * assembler names for identifiers: Asm Labels. (line 6)
42611 * assembly code, invalid: Bug Criteria. (line 12)
42612 * atan: Other Builtins. (line 6)
42613 * atan2: Other Builtins. (line 6)
42614 * atan2f: Other Builtins. (line 6)
42615 * atan2l: Other Builtins. (line 6)
42616 * atanf: Other Builtins. (line 6)
42617 * atanh: Other Builtins. (line 6)
42618 * atanhf: Other Builtins. (line 6)
42619 * atanhl: Other Builtins. (line 6)
42620 * atanl: Other Builtins. (line 6)
42621 * attribute of types: Type Attributes. (line 6)
42622 * attribute of variables: Variable Attributes.
42624 * attribute syntax: Attribute Syntax. (line 6)
42625 * autoincrement/decrement addressing: Simple Constraints. (line 30)
42626 * automatic inline for C++ member fns: Inline. (line 71)
42627 * AVR Options: AVR Options. (line 6)
42628 * Backwards Compatibility: Backwards Compatibility.
42630 * base class members: Name lookup. (line 6)
42631 * bcmp: Other Builtins. (line 6)
42632 * below100 attribute: Variable Attributes.
42634 * binary compatibility: Compatibility. (line 6)
42635 * Binary constants using the 0b prefix: Binary constants. (line 6)
42636 * Blackfin Options: Blackfin Options. (line 6)
42637 * bound pointer to member function: Bound member functions.
42639 * bounds checking: Optimize Options. (line 338)
42640 * bug criteria: Bug Criteria. (line 6)
42641 * bugs: Bugs. (line 6)
42642 * bugs, known: Trouble. (line 6)
42643 * built-in functions <1>: Other Builtins. (line 6)
42644 * built-in functions: C Dialect Options. (line 170)
42645 * bzero: Other Builtins. (line 6)
42646 * C compilation options: Invoking GCC. (line 17)
42647 * C intermediate output, nonexistent: G++ and GCC. (line 35)
42648 * C language extensions: C Extensions. (line 6)
42649 * C language, traditional: C Dialect Options. (line 250)
42650 * C standard: Standards. (line 13)
42651 * C standards: Standards. (line 13)
42652 * c++: Invoking G++. (line 14)
42653 * C++: G++ and GCC. (line 30)
42654 * C++ comments: C++ Comments. (line 6)
42655 * C++ compilation options: Invoking GCC. (line 23)
42656 * C++ interface and implementation headers: C++ Interface. (line 6)
42657 * C++ language extensions: C++ Extensions. (line 6)
42658 * C++ member fns, automatically inline: Inline. (line 71)
42659 * C++ misunderstandings: C++ Misunderstandings.
42661 * C++ options, command line: C++ Dialect Options.
42663 * C++ pragmas, effect on inlining: C++ Interface. (line 66)
42664 * C++ source file suffixes: Invoking G++. (line 6)
42665 * C++ static data, declaring and defining: Static Definitions.
42667 * C89: Standards. (line 13)
42668 * C90: Standards. (line 13)
42669 * C94: Standards. (line 13)
42670 * C95: Standards. (line 13)
42671 * C99: Standards. (line 13)
42672 * C9X: Standards. (line 13)
42673 * C_INCLUDE_PATH: Environment Variables.
42675 * cabs: Other Builtins. (line 6)
42676 * cabsf: Other Builtins. (line 6)
42677 * cabsl: Other Builtins. (line 6)
42678 * cacos: Other Builtins. (line 6)
42679 * cacosf: Other Builtins. (line 6)
42680 * cacosh: Other Builtins. (line 6)
42681 * cacoshf: Other Builtins. (line 6)
42682 * cacoshl: Other Builtins. (line 6)
42683 * cacosl: Other Builtins. (line 6)
42684 * calling functions through the function vector on H8/300, M16C, M32C and SH2A processors: Function Attributes.
42686 * calloc: Other Builtins. (line 6)
42687 * carg: Other Builtins. (line 6)
42688 * cargf: Other Builtins. (line 6)
42689 * cargl: Other Builtins. (line 6)
42690 * case labels in initializers: Designated Inits. (line 6)
42691 * case ranges: Case Ranges. (line 6)
42692 * casin: Other Builtins. (line 6)
42693 * casinf: Other Builtins. (line 6)
42694 * casinh: Other Builtins. (line 6)
42695 * casinhf: Other Builtins. (line 6)
42696 * casinhl: Other Builtins. (line 6)
42697 * casinl: Other Builtins. (line 6)
42698 * cast to a union: Cast to Union. (line 6)
42699 * catan: Other Builtins. (line 6)
42700 * catanf: Other Builtins. (line 6)
42701 * catanh: Other Builtins. (line 6)
42702 * catanhf: Other Builtins. (line 6)
42703 * catanhl: Other Builtins. (line 6)
42704 * catanl: Other Builtins. (line 6)
42705 * cbrt: Other Builtins. (line 6)
42706 * cbrtf: Other Builtins. (line 6)
42707 * cbrtl: Other Builtins. (line 6)
42708 * ccos: Other Builtins. (line 6)
42709 * ccosf: Other Builtins. (line 6)
42710 * ccosh: Other Builtins. (line 6)
42711 * ccoshf: Other Builtins. (line 6)
42712 * ccoshl: Other Builtins. (line 6)
42713 * ccosl: Other Builtins. (line 6)
42714 * ceil: Other Builtins. (line 6)
42715 * ceilf: Other Builtins. (line 6)
42716 * ceill: Other Builtins. (line 6)
42717 * cexp: Other Builtins. (line 6)
42718 * cexpf: Other Builtins. (line 6)
42719 * cexpl: Other Builtins. (line 6)
42720 * character set, execution: Preprocessor Options.
42722 * character set, input: Preprocessor Options.
42724 * character set, input normalization: Warning Options. (line 1076)
42725 * character set, wide execution: Preprocessor Options.
42727 * cimag: Other Builtins. (line 6)
42728 * cimagf: Other Builtins. (line 6)
42729 * cimagl: Other Builtins. (line 6)
42730 * cleanup attribute: Variable Attributes.
42732 * clog: Other Builtins. (line 6)
42733 * clogf: Other Builtins. (line 6)
42734 * clogl: Other Builtins. (line 6)
42735 * COBOL: G++ and GCC. (line 23)
42736 * code generation conventions: Code Gen Options. (line 6)
42737 * code, mixed with declarations: Mixed Declarations. (line 6)
42738 * cold function attribute: Function Attributes.
42740 * command options: Invoking GCC. (line 6)
42741 * comments, C++ style: C++ Comments. (line 6)
42742 * common attribute: Variable Attributes.
42744 * comparison of signed and unsigned values, warning: Warning Options.
42746 * compiler bugs, reporting: Bug Reporting. (line 6)
42747 * compiler compared to C++ preprocessor: G++ and GCC. (line 35)
42748 * compiler options, C++: C++ Dialect Options.
42750 * compiler options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
42752 * compiler version, specifying: Target Options. (line 6)
42753 * COMPILER_PATH: Environment Variables.
42755 * complex conjugation: Complex. (line 34)
42756 * complex numbers: Complex. (line 6)
42757 * compound literals: Compound Literals. (line 6)
42758 * computed gotos: Labels as Values. (line 6)
42759 * conditional expressions, extensions: Conditionals. (line 6)
42760 * conflicting types: Disappointments. (line 21)
42761 * conj: Other Builtins. (line 6)
42762 * conjf: Other Builtins. (line 6)
42763 * conjl: Other Builtins. (line 6)
42764 * const applied to function: Function Attributes.
42766 * const function attribute: Function Attributes.
42768 * constants in constraints: Simple Constraints. (line 60)
42769 * constraint modifier characters: Modifiers. (line 6)
42770 * constraint, matching: Simple Constraints. (line 129)
42771 * constraints, asm: Constraints. (line 6)
42772 * constraints, machine specific: Machine Constraints.
42774 * constructing calls: Constructing Calls. (line 6)
42775 * constructor expressions: Compound Literals. (line 6)
42776 * constructor function attribute: Function Attributes.
42778 * contributors: Contributors. (line 6)
42779 * copysign: Other Builtins. (line 6)
42780 * copysignf: Other Builtins. (line 6)
42781 * copysignl: Other Builtins. (line 6)
42782 * core dump: Bug Criteria. (line 9)
42783 * cos: Other Builtins. (line 6)
42784 * cosf: Other Builtins. (line 6)
42785 * cosh: Other Builtins. (line 6)
42786 * coshf: Other Builtins. (line 6)
42787 * coshl: Other Builtins. (line 6)
42788 * cosl: Other Builtins. (line 6)
42789 * CPATH: Environment Variables.
42791 * CPLUS_INCLUDE_PATH: Environment Variables.
42793 * cpow: Other Builtins. (line 6)
42794 * cpowf: Other Builtins. (line 6)
42795 * cpowl: Other Builtins. (line 6)
42796 * cproj: Other Builtins. (line 6)
42797 * cprojf: Other Builtins. (line 6)
42798 * cprojl: Other Builtins. (line 6)
42799 * creal: Other Builtins. (line 6)
42800 * crealf: Other Builtins. (line 6)
42801 * creall: Other Builtins. (line 6)
42802 * CRIS Options: CRIS Options. (line 6)
42803 * cross compiling: Target Options. (line 6)
42804 * CRX Options: CRX Options. (line 6)
42805 * csin: Other Builtins. (line 6)
42806 * csinf: Other Builtins. (line 6)
42807 * csinh: Other Builtins. (line 6)
42808 * csinhf: Other Builtins. (line 6)
42809 * csinhl: Other Builtins. (line 6)
42810 * csinl: Other Builtins. (line 6)
42811 * csqrt: Other Builtins. (line 6)
42812 * csqrtf: Other Builtins. (line 6)
42813 * csqrtl: Other Builtins. (line 6)
42814 * ctan: Other Builtins. (line 6)
42815 * ctanf: Other Builtins. (line 6)
42816 * ctanh: Other Builtins. (line 6)
42817 * ctanhf: Other Builtins. (line 6)
42818 * ctanhl: Other Builtins. (line 6)
42819 * ctanl: Other Builtins. (line 6)
42820 * Darwin options: Darwin Options. (line 6)
42821 * dcgettext: Other Builtins. (line 6)
42822 * DD integer suffix: Decimal Float. (line 6)
42823 * dd integer suffix: Decimal Float. (line 6)
42824 * deallocating variable length arrays: Variable Length. (line 23)
42825 * debugging information options: Debugging Options. (line 6)
42826 * decimal floating types: Decimal Float. (line 6)
42827 * declaration scope: Incompatibilities. (line 80)
42828 * declarations inside expressions: Statement Exprs. (line 6)
42829 * declarations, mixed with code: Mixed Declarations. (line 6)
42830 * declaring attributes of functions: Function Attributes.
42832 * declaring static data in C++: Static Definitions. (line 6)
42833 * defining static data in C++: Static Definitions. (line 6)
42834 * dependencies for make as output: Environment Variables.
42836 * dependencies, make: Preprocessor Options.
42838 * DEPENDENCIES_OUTPUT: Environment Variables.
42840 * dependent name lookup: Name lookup. (line 6)
42841 * deprecated attribute: Variable Attributes.
42843 * deprecated attribute.: Function Attributes.
42845 * designated initializers: Designated Inits. (line 6)
42846 * designator lists: Designated Inits. (line 94)
42847 * designators: Designated Inits. (line 61)
42848 * destructor function attribute: Function Attributes.
42850 * DF integer suffix: Decimal Float. (line 6)
42851 * df integer suffix: Decimal Float. (line 6)
42852 * dgettext: Other Builtins. (line 6)
42853 * diagnostic messages: Language Independent Options.
42855 * dialect options: C Dialect Options. (line 6)
42856 * digits in constraint: Simple Constraints. (line 117)
42857 * directory options: Directory Options. (line 6)
42858 * DL integer suffix: Decimal Float. (line 6)
42859 * dl integer suffix: Decimal Float. (line 6)
42860 * dollar signs in identifier names: Dollar Signs. (line 6)
42861 * double-word arithmetic: Long Long. (line 6)
42862 * downward funargs: Nested Functions. (line 6)
42863 * drem: Other Builtins. (line 6)
42864 * dremf: Other Builtins. (line 6)
42865 * dreml: Other Builtins. (line 6)
42866 * E in constraint: Simple Constraints. (line 79)
42867 * earlyclobber operand: Modifiers. (line 25)
42868 * eight bit data on the H8/300, H8/300H, and H8S: Function Attributes.
42870 * empty structures: Empty Structures. (line 6)
42871 * environment variables: Environment Variables.
42873 * erf: Other Builtins. (line 6)
42874 * erfc: Other Builtins. (line 6)
42875 * erfcf: Other Builtins. (line 6)
42876 * erfcl: Other Builtins. (line 6)
42877 * erff: Other Builtins. (line 6)
42878 * erfl: Other Builtins. (line 6)
42879 * error function attribute: Function Attributes.
42881 * error messages: Warnings and Errors.
42883 * escaped newlines: Escaped Newlines. (line 6)
42884 * exception handler functions on the Blackfin processor: Function Attributes.
42886 * exclamation point: Multi-Alternative. (line 33)
42887 * exit: Other Builtins. (line 6)
42888 * exp: Other Builtins. (line 6)
42889 * exp10: Other Builtins. (line 6)
42890 * exp10f: Other Builtins. (line 6)
42891 * exp10l: Other Builtins. (line 6)
42892 * exp2: Other Builtins. (line 6)
42893 * exp2f: Other Builtins. (line 6)
42894 * exp2l: Other Builtins. (line 6)
42895 * expf: Other Builtins. (line 6)
42896 * expl: Other Builtins. (line 6)
42897 * explicit register variables: Explicit Reg Vars. (line 6)
42898 * expm1: Other Builtins. (line 6)
42899 * expm1f: Other Builtins. (line 6)
42900 * expm1l: Other Builtins. (line 6)
42901 * expressions containing statements: Statement Exprs. (line 6)
42902 * expressions, constructor: Compound Literals. (line 6)
42903 * extended asm: Extended Asm. (line 6)
42904 * extensible constraints: Simple Constraints. (line 153)
42905 * extensions, ?:: Conditionals. (line 6)
42906 * extensions, C language: C Extensions. (line 6)
42907 * extensions, C++ language: C++ Extensions. (line 6)
42908 * external declaration scope: Incompatibilities. (line 80)
42909 * externally_visible attribute.: Function Attributes.
42911 * F in constraint: Simple Constraints. (line 84)
42912 * fabs: Other Builtins. (line 6)
42913 * fabsf: Other Builtins. (line 6)
42914 * fabsl: Other Builtins. (line 6)
42915 * fatal signal: Bug Criteria. (line 9)
42916 * fdim: Other Builtins. (line 6)
42917 * fdimf: Other Builtins. (line 6)
42918 * fdiml: Other Builtins. (line 6)
42919 * FDL, GNU Free Documentation License: GNU Free Documentation License.
42921 * ffs: Other Builtins. (line 6)
42922 * file name suffix: Overall Options. (line 14)
42923 * file names: Link Options. (line 10)
42924 * fixed-point types: Fixed-Point. (line 6)
42925 * flatten function attribute: Function Attributes.
42927 * flexible array members: Zero Length. (line 6)
42928 * float as function value type: Incompatibilities. (line 141)
42929 * floating point precision <1>: Disappointments. (line 68)
42930 * floating point precision: Optimize Options. (line 1352)
42931 * floor: Other Builtins. (line 6)
42932 * floorf: Other Builtins. (line 6)
42933 * floorl: Other Builtins. (line 6)
42934 * fma: Other Builtins. (line 6)
42935 * fmaf: Other Builtins. (line 6)
42936 * fmal: Other Builtins. (line 6)
42937 * fmax: Other Builtins. (line 6)
42938 * fmaxf: Other Builtins. (line 6)
42939 * fmaxl: Other Builtins. (line 6)
42940 * fmin: Other Builtins. (line 6)
42941 * fminf: Other Builtins. (line 6)
42942 * fminl: Other Builtins. (line 6)
42943 * fmod: Other Builtins. (line 6)
42944 * fmodf: Other Builtins. (line 6)
42945 * fmodl: Other Builtins. (line 6)
42946 * force_align_arg_pointer attribute: Function Attributes.
42948 * format function attribute: Function Attributes.
42950 * format_arg function attribute: Function Attributes.
42952 * Fortran: G++ and GCC. (line 6)
42953 * forwarding calls: Constructing Calls. (line 6)
42954 * fprintf: Other Builtins. (line 6)
42955 * fprintf_unlocked: Other Builtins. (line 6)
42956 * fputs: Other Builtins. (line 6)
42957 * fputs_unlocked: Other Builtins. (line 6)
42958 * FR30 Options: FR30 Options. (line 6)
42959 * freestanding environment: Standards. (line 13)
42960 * freestanding implementation: Standards. (line 13)
42961 * frexp: Other Builtins. (line 6)
42962 * frexpf: Other Builtins. (line 6)
42963 * frexpl: Other Builtins. (line 6)
42964 * FRV Options: FRV Options. (line 6)
42965 * fscanf: Other Builtins. (line 6)
42966 * fscanf, and constant strings: Incompatibilities. (line 17)
42967 * function addressability on the M32R/D: Function Attributes.
42969 * function attributes: Function Attributes.
42971 * function pointers, arithmetic: Pointer Arith. (line 6)
42972 * function prototype declarations: Function Prototypes.
42974 * function without a prologue/epilogue code: Function Attributes.
42976 * function, size of pointer to: Pointer Arith. (line 6)
42977 * functions called via pointer on the RS/6000 and PowerPC: Function Attributes.
42979 * functions in arbitrary sections: Function Attributes.
42981 * functions that are passed arguments in registers on the 386: Function Attributes.
42983 * functions that behave like malloc: Function Attributes.
42985 * functions that do not pop the argument stack on the 386: Function Attributes.
42987 * functions that do pop the argument stack on the 386: Function Attributes.
42989 * functions that have different compilation options on the 386: Function Attributes.
42991 * functions that have different optimization options: Function Attributes.
42993 * functions that have no side effects: Function Attributes.
42995 * functions that never return: Function Attributes.
42997 * functions that pop the argument stack on the 386: Function Attributes.
42999 * functions that return more than once: Function Attributes.
43001 * functions which do not handle memory bank switching on 68HC11/68HC12: Function Attributes.
43003 * functions which handle memory bank switching: Function Attributes.
43005 * functions with non-null pointer arguments: Function Attributes.
43007 * functions with printf, scanf, strftime or strfmon style arguments: Function Attributes.
43009 * g in constraint: Simple Constraints. (line 110)
43010 * G in constraint: Simple Constraints. (line 88)
43011 * g++: Invoking G++. (line 14)
43012 * G++: G++ and GCC. (line 30)
43013 * gamma: Other Builtins. (line 6)
43014 * gamma_r: Other Builtins. (line 6)
43015 * gammaf: Other Builtins. (line 6)
43016 * gammaf_r: Other Builtins. (line 6)
43017 * gammal: Other Builtins. (line 6)
43018 * gammal_r: Other Builtins. (line 6)
43019 * GCC: G++ and GCC. (line 6)
43020 * GCC command options: Invoking GCC. (line 6)
43021 * GCC_EXEC_PREFIX: Environment Variables.
43023 * gcc_struct: Type Attributes. (line 309)
43024 * gcc_struct attribute: Variable Attributes.
43026 * gcov: Debugging Options. (line 271)
43027 * gettext: Other Builtins. (line 6)
43028 * global offset table: Code Gen Options. (line 184)
43029 * global register after longjmp: Global Reg Vars. (line 66)
43030 * global register variables: Global Reg Vars. (line 6)
43031 * GNAT: G++ and GCC. (line 30)
43032 * GNU C Compiler: G++ and GCC. (line 6)
43033 * GNU Compiler Collection: G++ and GCC. (line 6)
43034 * gnu_inline function attribute: Function Attributes.
43036 * goto with computed label: Labels as Values. (line 6)
43037 * gprof: Debugging Options. (line 232)
43038 * grouping options: Invoking GCC. (line 26)
43039 * H in constraint: Simple Constraints. (line 88)
43040 * hardware models and configurations, specifying: Submodel Options.
43042 * hex floats: Hex Floats. (line 6)
43043 * HK fixed-suffix: Fixed-Point. (line 6)
43044 * hk fixed-suffix: Fixed-Point. (line 6)
43045 * hosted environment <1>: C Dialect Options. (line 204)
43046 * hosted environment: Standards. (line 13)
43047 * hosted implementation: Standards. (line 13)
43048 * hot function attribute: Function Attributes.
43050 * HPPA Options: HPPA Options. (line 6)
43051 * HR fixed-suffix: Fixed-Point. (line 6)
43052 * hr fixed-suffix: Fixed-Point. (line 6)
43053 * hypot: Other Builtins. (line 6)
43054 * hypotf: Other Builtins. (line 6)
43055 * hypotl: Other Builtins. (line 6)
43056 * I in constraint: Simple Constraints. (line 71)
43057 * i in constraint: Simple Constraints. (line 60)
43058 * i386 and x86-64 Windows Options: i386 and x86-64 Windows Options.
43060 * i386 Options: i386 and x86-64 Options.
43062 * IA-64 Options: IA-64 Options. (line 6)
43063 * IBM RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
43065 * identifier names, dollar signs in: Dollar Signs. (line 6)
43066 * identifiers, names in assembler code: Asm Labels. (line 6)
43067 * ilogb: Other Builtins. (line 6)
43068 * ilogbf: Other Builtins. (line 6)
43069 * ilogbl: Other Builtins. (line 6)
43070 * imaxabs: Other Builtins. (line 6)
43071 * implementation-defined behavior, C language: C Implementation.
43073 * implied #pragma implementation: C++ Interface. (line 46)
43074 * incompatibilities of GCC: Incompatibilities. (line 6)
43075 * increment operators: Bug Criteria. (line 17)
43076 * index: Other Builtins. (line 6)
43077 * indirect calls on ARM: Function Attributes.
43079 * indirect calls on MIPS: Function Attributes.
43081 * init_priority attribute: C++ Attributes. (line 9)
43082 * initializations in expressions: Compound Literals. (line 6)
43083 * initializers with labeled elements: Designated Inits. (line 6)
43084 * initializers, non-constant: Initializers. (line 6)
43085 * inline automatic for C++ member fns: Inline. (line 71)
43086 * inline functions: Inline. (line 6)
43087 * inline functions, omission of: Inline. (line 51)
43088 * inlining and C++ pragmas: C++ Interface. (line 66)
43089 * installation trouble: Trouble. (line 6)
43090 * integrating function code: Inline. (line 6)
43091 * Intel 386 Options: i386 and x86-64 Options.
43093 * interface and implementation headers, C++: C++ Interface. (line 6)
43094 * intermediate C version, nonexistent: G++ and GCC. (line 35)
43095 * interrupt handler functions: Function Attributes.
43097 * interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors: Function Attributes.
43099 * interrupt service routines on ARM: Function Attributes.
43101 * interrupt thread functions on fido: Function Attributes.
43103 * introduction: Top. (line 6)
43104 * invalid assembly code: Bug Criteria. (line 12)
43105 * invalid input: Bug Criteria. (line 42)
43106 * invoking g++: Invoking G++. (line 22)
43107 * isalnum: Other Builtins. (line 6)
43108 * isalpha: Other Builtins. (line 6)
43109 * isascii: Other Builtins. (line 6)
43110 * isblank: Other Builtins. (line 6)
43111 * iscntrl: Other Builtins. (line 6)
43112 * isdigit: Other Builtins. (line 6)
43113 * isgraph: Other Builtins. (line 6)
43114 * islower: Other Builtins. (line 6)
43115 * ISO 9899: Standards. (line 13)
43116 * ISO C: Standards. (line 13)
43117 * ISO C standard: Standards. (line 13)
43118 * ISO C90: Standards. (line 13)
43119 * ISO C94: Standards. (line 13)
43120 * ISO C95: Standards. (line 13)
43121 * ISO C99: Standards. (line 13)
43122 * ISO C9X: Standards. (line 13)
43123 * ISO support: C Dialect Options. (line 10)
43124 * ISO/IEC 9899: Standards. (line 13)
43125 * isprint: Other Builtins. (line 6)
43126 * ispunct: Other Builtins. (line 6)
43127 * isspace: Other Builtins. (line 6)
43128 * isupper: Other Builtins. (line 6)
43129 * iswalnum: Other Builtins. (line 6)
43130 * iswalpha: Other Builtins. (line 6)
43131 * iswblank: Other Builtins. (line 6)
43132 * iswcntrl: Other Builtins. (line 6)
43133 * iswdigit: Other Builtins. (line 6)
43134 * iswgraph: Other Builtins. (line 6)
43135 * iswlower: Other Builtins. (line 6)
43136 * iswprint: Other Builtins. (line 6)
43137 * iswpunct: Other Builtins. (line 6)
43138 * iswspace: Other Builtins. (line 6)
43139 * iswupper: Other Builtins. (line 6)
43140 * iswxdigit: Other Builtins. (line 6)
43141 * isxdigit: Other Builtins. (line 6)
43142 * j0: Other Builtins. (line 6)
43143 * j0f: Other Builtins. (line 6)
43144 * j0l: Other Builtins. (line 6)
43145 * j1: Other Builtins. (line 6)
43146 * j1f: Other Builtins. (line 6)
43147 * j1l: Other Builtins. (line 6)
43148 * Java: G++ and GCC. (line 6)
43149 * java_interface attribute: C++ Attributes. (line 29)
43150 * jn: Other Builtins. (line 6)
43151 * jnf: Other Builtins. (line 6)
43152 * jnl: Other Builtins. (line 6)
43153 * K fixed-suffix: Fixed-Point. (line 6)
43154 * k fixed-suffix: Fixed-Point. (line 6)
43155 * keywords, alternate: Alternate Keywords. (line 6)
43156 * known causes of trouble: Trouble. (line 6)
43157 * l1_data variable attribute: Variable Attributes.
43159 * l1_data_A variable attribute: Variable Attributes.
43161 * l1_data_B variable attribute: Variable Attributes.
43163 * l1_text function attribute: Function Attributes.
43165 * labeled elements in initializers: Designated Inits. (line 6)
43166 * labels as values: Labels as Values. (line 6)
43167 * labs: Other Builtins. (line 6)
43168 * LANG: Environment Variables.
43170 * language dialect options: C Dialect Options. (line 6)
43171 * LC_ALL: Environment Variables.
43173 * LC_CTYPE: Environment Variables.
43175 * LC_MESSAGES: Environment Variables.
43177 * ldexp: Other Builtins. (line 6)
43178 * ldexpf: Other Builtins. (line 6)
43179 * ldexpl: Other Builtins. (line 6)
43180 * length-zero arrays: Zero Length. (line 6)
43181 * lgamma: Other Builtins. (line 6)
43182 * lgamma_r: Other Builtins. (line 6)
43183 * lgammaf: Other Builtins. (line 6)
43184 * lgammaf_r: Other Builtins. (line 6)
43185 * lgammal: Other Builtins. (line 6)
43186 * lgammal_r: Other Builtins. (line 6)
43187 * Libraries: Link Options. (line 24)
43188 * LIBRARY_PATH: Environment Variables.
43190 * link options: Link Options. (line 6)
43191 * linker script: Link Options. (line 163)
43192 * LK fixed-suffix: Fixed-Point. (line 6)
43193 * lk fixed-suffix: Fixed-Point. (line 6)
43194 * LL integer suffix: Long Long. (line 6)
43195 * llabs: Other Builtins. (line 6)
43196 * LLK fixed-suffix: Fixed-Point. (line 6)
43197 * llk fixed-suffix: Fixed-Point. (line 6)
43198 * LLR fixed-suffix: Fixed-Point. (line 6)
43199 * llr fixed-suffix: Fixed-Point. (line 6)
43200 * llrint: Other Builtins. (line 6)
43201 * llrintf: Other Builtins. (line 6)
43202 * llrintl: Other Builtins. (line 6)
43203 * llround: Other Builtins. (line 6)
43204 * llroundf: Other Builtins. (line 6)
43205 * llroundl: Other Builtins. (line 6)
43206 * load address instruction: Simple Constraints. (line 144)
43207 * local labels: Local Labels. (line 6)
43208 * local variables in macros: Typeof. (line 42)
43209 * local variables, specifying registers: Local Reg Vars. (line 6)
43210 * locale: Environment Variables.
43212 * locale definition: Environment Variables.
43214 * log: Other Builtins. (line 6)
43215 * log10: Other Builtins. (line 6)
43216 * log10f: Other Builtins. (line 6)
43217 * log10l: Other Builtins. (line 6)
43218 * log1p: Other Builtins. (line 6)
43219 * log1pf: Other Builtins. (line 6)
43220 * log1pl: Other Builtins. (line 6)
43221 * log2: Other Builtins. (line 6)
43222 * log2f: Other Builtins. (line 6)
43223 * log2l: Other Builtins. (line 6)
43224 * logb: Other Builtins. (line 6)
43225 * logbf: Other Builtins. (line 6)
43226 * logbl: Other Builtins. (line 6)
43227 * logf: Other Builtins. (line 6)
43228 * logl: Other Builtins. (line 6)
43229 * long long data types: Long Long. (line 6)
43230 * longjmp: Global Reg Vars. (line 66)
43231 * longjmp incompatibilities: Incompatibilities. (line 39)
43232 * longjmp warnings: Warning Options. (line 570)
43233 * LR fixed-suffix: Fixed-Point. (line 6)
43234 * lr fixed-suffix: Fixed-Point. (line 6)
43235 * lrint: Other Builtins. (line 6)
43236 * lrintf: Other Builtins. (line 6)
43237 * lrintl: Other Builtins. (line 6)
43238 * lround: Other Builtins. (line 6)
43239 * lroundf: Other Builtins. (line 6)
43240 * lroundl: Other Builtins. (line 6)
43241 * m in constraint: Simple Constraints. (line 17)
43242 * M32C options: M32C Options. (line 6)
43243 * M32R/D options: M32R/D Options. (line 6)
43244 * M680x0 options: M680x0 Options. (line 6)
43245 * M68hc1x options: M68hc1x Options. (line 6)
43246 * machine dependent options: Submodel Options. (line 6)
43247 * machine specific constraints: Machine Constraints.
43249 * macro with variable arguments: Variadic Macros. (line 6)
43250 * macros containing asm: Extended Asm. (line 241)
43251 * macros, inline alternative: Inline. (line 6)
43252 * macros, local labels: Local Labels. (line 6)
43253 * macros, local variables in: Typeof. (line 42)
43254 * macros, statements in expressions: Statement Exprs. (line 6)
43255 * macros, types of arguments: Typeof. (line 6)
43256 * make: Preprocessor Options.
43258 * malloc: Other Builtins. (line 6)
43259 * malloc attribute: Function Attributes.
43261 * matching constraint: Simple Constraints. (line 129)
43262 * MCore options: MCore Options. (line 6)
43263 * member fns, automatically inline: Inline. (line 71)
43264 * memchr: Other Builtins. (line 6)
43265 * memcmp: Other Builtins. (line 6)
43266 * memcpy: Other Builtins. (line 6)
43267 * memory references in constraints: Simple Constraints. (line 17)
43268 * mempcpy: Other Builtins. (line 6)
43269 * memset: Other Builtins. (line 6)
43270 * Mercury: G++ and GCC. (line 23)
43271 * message formatting: Language Independent Options.
43273 * messages, warning: Warning Options. (line 6)
43274 * messages, warning and error: Warnings and Errors.
43276 * middle-operands, omitted: Conditionals. (line 6)
43277 * MIPS options: MIPS Options. (line 6)
43278 * mips16 attribute: Function Attributes.
43280 * misunderstandings in C++: C++ Misunderstandings.
43282 * mixed declarations and code: Mixed Declarations. (line 6)
43283 * mktemp, and constant strings: Incompatibilities. (line 13)
43284 * MMIX Options: MMIX Options. (line 6)
43285 * MN10300 options: MN10300 Options. (line 6)
43286 * mode attribute: Variable Attributes.
43288 * modf: Other Builtins. (line 6)
43289 * modff: Other Builtins. (line 6)
43290 * modfl: Other Builtins. (line 6)
43291 * modifiers in constraints: Modifiers. (line 6)
43292 * ms_abi attribute: Function Attributes.
43294 * ms_struct: Type Attributes. (line 309)
43295 * ms_struct attribute: Variable Attributes.
43297 * mudflap: Optimize Options. (line 338)
43298 * multiple alternative constraints: Multi-Alternative. (line 6)
43299 * multiprecision arithmetic: Long Long. (line 6)
43300 * n in constraint: Simple Constraints. (line 65)
43301 * names used in assembler code: Asm Labels. (line 6)
43302 * naming convention, implementation headers: C++ Interface. (line 46)
43303 * nearbyint: Other Builtins. (line 6)
43304 * nearbyintf: Other Builtins. (line 6)
43305 * nearbyintl: Other Builtins. (line 6)
43306 * nested functions: Nested Functions. (line 6)
43307 * newlines (escaped): Escaped Newlines. (line 6)
43308 * nextafter: Other Builtins. (line 6)
43309 * nextafterf: Other Builtins. (line 6)
43310 * nextafterl: Other Builtins. (line 6)
43311 * nexttoward: Other Builtins. (line 6)
43312 * nexttowardf: Other Builtins. (line 6)
43313 * nexttowardl: Other Builtins. (line 6)
43314 * NFC: Warning Options. (line 1076)
43315 * NFKC: Warning Options. (line 1076)
43316 * NMI handler functions on the Blackfin processor: Function Attributes.
43318 * no_instrument_function function attribute: Function Attributes.
43320 * nocommon attribute: Variable Attributes.
43322 * noinline function attribute: Function Attributes.
43324 * nomips16 attribute: Function Attributes.
43326 * non-constant initializers: Initializers. (line 6)
43327 * non-static inline function: Inline. (line 85)
43328 * nonnull function attribute: Function Attributes.
43330 * noreturn function attribute: Function Attributes.
43332 * nothrow function attribute: Function Attributes.
43334 * o in constraint: Simple Constraints. (line 23)
43335 * OBJC_INCLUDE_PATH: Environment Variables.
43337 * Objective-C <1>: Standards. (line 153)
43338 * Objective-C: G++ and GCC. (line 6)
43339 * Objective-C and Objective-C++ options, command line: Objective-C and Objective-C++ Dialect Options.
43341 * Objective-C++ <1>: Standards. (line 153)
43342 * Objective-C++: G++ and GCC. (line 6)
43343 * offsettable address: Simple Constraints. (line 23)
43344 * old-style function definitions: Function Prototypes.
43346 * omitted middle-operands: Conditionals. (line 6)
43347 * open coding: Inline. (line 6)
43348 * openmp parallel: C Dialect Options. (line 221)
43349 * operand constraints, asm: Constraints. (line 6)
43350 * optimize function attribute: Function Attributes.
43352 * optimize options: Optimize Options. (line 6)
43353 * options to control diagnostics formatting: Language Independent Options.
43355 * options to control warnings: Warning Options. (line 6)
43356 * options, C++: C++ Dialect Options.
43358 * options, code generation: Code Gen Options. (line 6)
43359 * options, debugging: Debugging Options. (line 6)
43360 * options, dialect: C Dialect Options. (line 6)
43361 * options, directory search: Directory Options. (line 6)
43362 * options, GCC command: Invoking GCC. (line 6)
43363 * options, grouping: Invoking GCC. (line 26)
43364 * options, linking: Link Options. (line 6)
43365 * options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
43367 * options, optimization: Optimize Options. (line 6)
43368 * options, order: Invoking GCC. (line 30)
43369 * options, preprocessor: Preprocessor Options.
43371 * order of evaluation, side effects: Non-bugs. (line 196)
43372 * order of options: Invoking GCC. (line 30)
43373 * other register constraints: Simple Constraints. (line 153)
43374 * output file option: Overall Options. (line 186)
43375 * overloaded virtual fn, warning: C++ Dialect Options.
43377 * p in constraint: Simple Constraints. (line 144)
43378 * packed attribute: Variable Attributes.
43380 * parameter forward declaration: Variable Length. (line 60)
43381 * parameters, aliased: Code Gen Options. (line 409)
43382 * Pascal: G++ and GCC. (line 23)
43383 * PDP-11 Options: PDP-11 Options. (line 6)
43384 * PIC: Code Gen Options. (line 184)
43385 * picoChip options: picoChip Options. (line 6)
43386 * pmf: Bound member functions.
43388 * pointer arguments: Function Attributes.
43390 * pointer to member function: Bound member functions.
43392 * portions of temporary objects, pointers to: Temporaries. (line 6)
43393 * pow: Other Builtins. (line 6)
43394 * pow10: Other Builtins. (line 6)
43395 * pow10f: Other Builtins. (line 6)
43396 * pow10l: Other Builtins. (line 6)
43397 * PowerPC options: PowerPC Options. (line 6)
43398 * powf: Other Builtins. (line 6)
43399 * powl: Other Builtins. (line 6)
43400 * pragma GCC optimize: Function Specific Option Pragmas.
43402 * pragma GCC pop_options: Function Specific Option Pragmas.
43404 * pragma GCC push_options: Function Specific Option Pragmas.
43406 * pragma GCC reset_options: Function Specific Option Pragmas.
43408 * pragma GCC target: Function Specific Option Pragmas.
43410 * pragma, align: Solaris Pragmas. (line 11)
43411 * pragma, diagnostic: Diagnostic Pragmas. (line 14)
43412 * pragma, extern_prefix: Symbol-Renaming Pragmas.
43414 * pragma, fini: Solaris Pragmas. (line 19)
43415 * pragma, init: Solaris Pragmas. (line 24)
43416 * pragma, long_calls: ARM Pragmas. (line 11)
43417 * pragma, long_calls_off: ARM Pragmas. (line 17)
43418 * pragma, longcall: RS/6000 and PowerPC Pragmas.
43420 * pragma, mark: Darwin Pragmas. (line 11)
43421 * pragma, memregs: M32C Pragmas. (line 7)
43422 * pragma, no_long_calls: ARM Pragmas. (line 14)
43423 * pragma, options align: Darwin Pragmas. (line 14)
43424 * pragma, pop_macro: Push/Pop Macro Pragmas.
43426 * pragma, push_macro: Push/Pop Macro Pragmas.
43428 * pragma, reason for not using: Function Attributes.
43430 * pragma, redefine_extname: Symbol-Renaming Pragmas.
43432 * pragma, segment: Darwin Pragmas. (line 21)
43433 * pragma, unused: Darwin Pragmas. (line 24)
43434 * pragma, visibility: Visibility Pragmas. (line 8)
43435 * pragma, weak: Weak Pragmas. (line 10)
43436 * pragmas: Pragmas. (line 6)
43437 * pragmas in C++, effect on inlining: C++ Interface. (line 66)
43438 * pragmas, interface and implementation: C++ Interface. (line 6)
43439 * pragmas, warning of unknown: Warning Options. (line 587)
43440 * precompiled headers: Precompiled Headers.
43442 * preprocessing numbers: Incompatibilities. (line 173)
43443 * preprocessing tokens: Incompatibilities. (line 173)
43444 * preprocessor options: Preprocessor Options.
43446 * printf: Other Builtins. (line 6)
43447 * printf_unlocked: Other Builtins. (line 6)
43448 * prof: Debugging Options. (line 226)
43449 * progmem variable attribute: Variable Attributes.
43451 * promotion of formal parameters: Function Prototypes.
43453 * pure function attribute: Function Attributes.
43455 * push address instruction: Simple Constraints. (line 144)
43456 * putchar: Other Builtins. (line 6)
43457 * puts: Other Builtins. (line 6)
43458 * Q floating point suffix: Floating Types. (line 6)
43459 * q floating point suffix: Floating Types. (line 6)
43460 * qsort, and global register variables: Global Reg Vars. (line 42)
43461 * question mark: Multi-Alternative. (line 27)
43462 * R fixed-suffix: Fixed-Point. (line 6)
43463 * r fixed-suffix: Fixed-Point. (line 6)
43464 * r in constraint: Simple Constraints. (line 56)
43465 * ranges in case statements: Case Ranges. (line 6)
43466 * read-only strings: Incompatibilities. (line 9)
43467 * register variable after longjmp: Global Reg Vars. (line 66)
43468 * registers: Extended Asm. (line 6)
43469 * registers for local variables: Local Reg Vars. (line 6)
43470 * registers in constraints: Simple Constraints. (line 56)
43471 * registers, global allocation: Explicit Reg Vars. (line 6)
43472 * registers, global variables in: Global Reg Vars. (line 6)
43473 * regparm attribute: Function Attributes.
43475 * relocation truncated to fit (ColdFire): M680x0 Options. (line 325)
43476 * relocation truncated to fit (MIPS): MIPS Options. (line 198)
43477 * remainder: Other Builtins. (line 6)
43478 * remainderf: Other Builtins. (line 6)
43479 * remainderl: Other Builtins. (line 6)
43480 * remquo: Other Builtins. (line 6)
43481 * remquof: Other Builtins. (line 6)
43482 * remquol: Other Builtins. (line 6)
43483 * reordering, warning: C++ Dialect Options.
43485 * reporting bugs: Bugs. (line 6)
43486 * resbank attribute: Function Attributes.
43488 * rest argument (in macro): Variadic Macros. (line 6)
43489 * restricted pointers: Restricted Pointers.
43491 * restricted references: Restricted Pointers.
43493 * restricted this pointer: Restricted Pointers.
43495 * returns_twice attribute: Function Attributes.
43497 * rindex: Other Builtins. (line 6)
43498 * rint: Other Builtins. (line 6)
43499 * rintf: Other Builtins. (line 6)
43500 * rintl: Other Builtins. (line 6)
43501 * round: Other Builtins. (line 6)
43502 * roundf: Other Builtins. (line 6)
43503 * roundl: Other Builtins. (line 6)
43504 * RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
43506 * RTTI: Vague Linkage. (line 43)
43507 * run-time options: Code Gen Options. (line 6)
43508 * s in constraint: Simple Constraints. (line 92)
43509 * S/390 and zSeries Options: S/390 and zSeries Options.
43511 * save all registers on the Blackfin, H8/300, H8/300H, and H8S: Function Attributes.
43513 * scalb: Other Builtins. (line 6)
43514 * scalbf: Other Builtins. (line 6)
43515 * scalbl: Other Builtins. (line 6)
43516 * scalbln: Other Builtins. (line 6)
43517 * scalblnf: Other Builtins. (line 6)
43518 * scalbn: Other Builtins. (line 6)
43519 * scalbnf: Other Builtins. (line 6)
43520 * scanf, and constant strings: Incompatibilities. (line 17)
43521 * scanfnl: Other Builtins. (line 6)
43522 * scope of a variable length array: Variable Length. (line 23)
43523 * scope of declaration: Disappointments. (line 21)
43524 * scope of external declarations: Incompatibilities. (line 80)
43525 * Score Options: Score Options. (line 6)
43526 * search path: Directory Options. (line 6)
43527 * section function attribute: Function Attributes.
43529 * section variable attribute: Variable Attributes.
43531 * sentinel function attribute: Function Attributes.
43533 * setjmp: Global Reg Vars. (line 66)
43534 * setjmp incompatibilities: Incompatibilities. (line 39)
43535 * shared strings: Incompatibilities. (line 9)
43536 * shared variable attribute: Variable Attributes.
43538 * side effect in ?:: Conditionals. (line 20)
43539 * side effects, macro argument: Statement Exprs. (line 35)
43540 * side effects, order of evaluation: Non-bugs. (line 196)
43541 * signal handler functions on the AVR processors: Function Attributes.
43543 * signbit: Other Builtins. (line 6)
43544 * signbitd128: Other Builtins. (line 6)
43545 * signbitd32: Other Builtins. (line 6)
43546 * signbitd64: Other Builtins. (line 6)
43547 * signbitf: Other Builtins. (line 6)
43548 * signbitl: Other Builtins. (line 6)
43549 * signed and unsigned values, comparison warning: Warning Options.
43551 * significand: Other Builtins. (line 6)
43552 * significandf: Other Builtins. (line 6)
43553 * significandl: Other Builtins. (line 6)
43554 * simple constraints: Simple Constraints. (line 6)
43555 * sin: Other Builtins. (line 6)
43556 * sincos: Other Builtins. (line 6)
43557 * sincosf: Other Builtins. (line 6)
43558 * sincosl: Other Builtins. (line 6)
43559 * sinf: Other Builtins. (line 6)
43560 * sinh: Other Builtins. (line 6)
43561 * sinhf: Other Builtins. (line 6)
43562 * sinhl: Other Builtins. (line 6)
43563 * sinl: Other Builtins. (line 6)
43564 * sizeof: Typeof. (line 6)
43565 * smaller data references: M32R/D Options. (line 57)
43566 * smaller data references (PowerPC): RS/6000 and PowerPC Options.
43568 * snprintf: Other Builtins. (line 6)
43569 * SPARC options: SPARC Options. (line 6)
43570 * Spec Files: Spec Files. (line 6)
43571 * specified registers: Explicit Reg Vars. (line 6)
43572 * specifying compiler version and target machine: Target Options.
43574 * specifying hardware config: Submodel Options. (line 6)
43575 * specifying machine version: Target Options. (line 6)
43576 * specifying registers for local variables: Local Reg Vars. (line 6)
43577 * speed of compilation: Precompiled Headers.
43579 * sprintf: Other Builtins. (line 6)
43580 * SPU options: SPU Options. (line 6)
43581 * sqrt: Other Builtins. (line 6)
43582 * sqrtf: Other Builtins. (line 6)
43583 * sqrtl: Other Builtins. (line 6)
43584 * sscanf: Other Builtins. (line 6)
43585 * sscanf, and constant strings: Incompatibilities. (line 17)
43586 * sseregparm attribute: Function Attributes.
43588 * statements inside expressions: Statement Exprs. (line 6)
43589 * static data in C++, declaring and defining: Static Definitions.
43591 * stpcpy: Other Builtins. (line 6)
43592 * stpncpy: Other Builtins. (line 6)
43593 * strcasecmp: Other Builtins. (line 6)
43594 * strcat: Other Builtins. (line 6)
43595 * strchr: Other Builtins. (line 6)
43596 * strcmp: Other Builtins. (line 6)
43597 * strcpy: Other Builtins. (line 6)
43598 * strcspn: Other Builtins. (line 6)
43599 * strdup: Other Builtins. (line 6)
43600 * strfmon: Other Builtins. (line 6)
43601 * strftime: Other Builtins. (line 6)
43602 * string constants: Incompatibilities. (line 9)
43603 * strlen: Other Builtins. (line 6)
43604 * strncasecmp: Other Builtins. (line 6)
43605 * strncat: Other Builtins. (line 6)
43606 * strncmp: Other Builtins. (line 6)
43607 * strncpy: Other Builtins. (line 6)
43608 * strndup: Other Builtins. (line 6)
43609 * strpbrk: Other Builtins. (line 6)
43610 * strrchr: Other Builtins. (line 6)
43611 * strspn: Other Builtins. (line 6)
43612 * strstr: Other Builtins. (line 6)
43613 * struct: Unnamed Fields. (line 6)
43614 * structures: Incompatibilities. (line 146)
43615 * structures, constructor expression: Compound Literals. (line 6)
43616 * submodel options: Submodel Options. (line 6)
43617 * subscripting: Subscripting. (line 6)
43618 * subscripting and function values: Subscripting. (line 6)
43619 * suffixes for C++ source: Invoking G++. (line 6)
43620 * SUNPRO_DEPENDENCIES: Environment Variables.
43622 * suppressing warnings: Warning Options. (line 6)
43623 * surprises in C++: C++ Misunderstandings.
43625 * syntax checking: Warning Options. (line 13)
43626 * syscall_linkage attribute: Function Attributes.
43628 * system headers, warnings from: Warning Options. (line 701)
43629 * sysv_abi attribute: Function Attributes.
43631 * tan: Other Builtins. (line 6)
43632 * tanf: Other Builtins. (line 6)
43633 * tanh: Other Builtins. (line 6)
43634 * tanhf: Other Builtins. (line 6)
43635 * tanhl: Other Builtins. (line 6)
43636 * tanl: Other Builtins. (line 6)
43637 * target function attribute: Function Attributes.
43639 * target machine, specifying: Target Options. (line 6)
43640 * target options: Target Options. (line 6)
43641 * target("abm") attribute: Function Attributes.
43643 * target("aes") attribute: Function Attributes.
43645 * target("align-stringops") attribute: Function Attributes.
43647 * target("arch=ARCH") attribute: Function Attributes.
43649 * target("cld") attribute: Function Attributes.
43651 * target("fancy-math-387") attribute: Function Attributes.
43653 * target("fpmath=FPMATH") attribute: Function Attributes.
43655 * target("fused-madd") attribute: Function Attributes.
43657 * target("ieee-fp") attribute: Function Attributes.
43659 * target("inline-all-stringops") attribute: Function Attributes.
43661 * target("inline-stringops-dynamically") attribute: Function Attributes.
43663 * target("mmx") attribute: Function Attributes.
43665 * target("pclmul") attribute: Function Attributes.
43667 * target("popcnt") attribute: Function Attributes.
43669 * target("recip") attribute: Function Attributes.
43671 * target("sse") attribute: Function Attributes.
43673 * target("sse2") attribute: Function Attributes.
43675 * target("sse3") attribute: Function Attributes.
43677 * target("sse4") attribute: Function Attributes.
43679 * target("sse4.1") attribute: Function Attributes.
43681 * target("sse4.2") attribute: Function Attributes.
43683 * target("sse4a") attribute: Function Attributes.
43685 * target("sse5") attribute: Function Attributes.
43687 * target("ssse3") attribute: Function Attributes.
43689 * target("tune=TUNE") attribute: Function Attributes.
43691 * TC1: Standards. (line 13)
43692 * TC2: Standards. (line 13)
43693 * TC3: Standards. (line 13)
43694 * Technical Corrigenda: Standards. (line 13)
43695 * Technical Corrigendum 1: Standards. (line 13)
43696 * Technical Corrigendum 2: Standards. (line 13)
43697 * Technical Corrigendum 3: Standards. (line 13)
43698 * template instantiation: Template Instantiation.
43700 * temporaries, lifetime of: Temporaries. (line 6)
43701 * tgamma: Other Builtins. (line 6)
43702 * tgammaf: Other Builtins. (line 6)
43703 * tgammal: Other Builtins. (line 6)
43704 * Thread-Local Storage: Thread-Local. (line 6)
43705 * thunks: Nested Functions. (line 6)
43706 * tiny data section on the H8/300H and H8S: Function Attributes.
43708 * TLS: Thread-Local. (line 6)
43709 * tls_model attribute: Variable Attributes.
43711 * TMPDIR: Environment Variables.
43713 * toascii: Other Builtins. (line 6)
43714 * tolower: Other Builtins. (line 6)
43715 * toupper: Other Builtins. (line 6)
43716 * towlower: Other Builtins. (line 6)
43717 * towupper: Other Builtins. (line 6)
43718 * traditional C language: C Dialect Options. (line 250)
43719 * trunc: Other Builtins. (line 6)
43720 * truncf: Other Builtins. (line 6)
43721 * truncl: Other Builtins. (line 6)
43722 * two-stage name lookup: Name lookup. (line 6)
43723 * type alignment: Alignment. (line 6)
43724 * type attributes: Type Attributes. (line 6)
43725 * type_info: Vague Linkage. (line 43)
43726 * typedef names as function parameters: Incompatibilities. (line 97)
43727 * typeof: Typeof. (line 6)
43728 * UHK fixed-suffix: Fixed-Point. (line 6)
43729 * uhk fixed-suffix: Fixed-Point. (line 6)
43730 * UHR fixed-suffix: Fixed-Point. (line 6)
43731 * uhr fixed-suffix: Fixed-Point. (line 6)
43732 * UK fixed-suffix: Fixed-Point. (line 6)
43733 * uk fixed-suffix: Fixed-Point. (line 6)
43734 * ULK fixed-suffix: Fixed-Point. (line 6)
43735 * ulk fixed-suffix: Fixed-Point. (line 6)
43736 * ULL integer suffix: Long Long. (line 6)
43737 * ULLK fixed-suffix: Fixed-Point. (line 6)
43738 * ullk fixed-suffix: Fixed-Point. (line 6)
43739 * ULLR fixed-suffix: Fixed-Point. (line 6)
43740 * ullr fixed-suffix: Fixed-Point. (line 6)
43741 * ULR fixed-suffix: Fixed-Point. (line 6)
43742 * ulr fixed-suffix: Fixed-Point. (line 6)
43743 * undefined behavior: Bug Criteria. (line 17)
43744 * undefined function value: Bug Criteria. (line 17)
43745 * underscores in variables in macros: Typeof. (line 42)
43746 * union: Unnamed Fields. (line 6)
43747 * union, casting to a: Cast to Union. (line 6)
43748 * unions: Incompatibilities. (line 146)
43749 * unknown pragmas, warning: Warning Options. (line 587)
43750 * unresolved references and -nodefaultlibs: Link Options. (line 79)
43751 * unresolved references and -nostdlib: Link Options. (line 79)
43752 * unused attribute.: Function Attributes.
43754 * UR fixed-suffix: Fixed-Point. (line 6)
43755 * ur fixed-suffix: Fixed-Point. (line 6)
43756 * used attribute.: Function Attributes.
43758 * User stack pointer in interrupts on the Blackfin: Function Attributes.
43760 * V in constraint: Simple Constraints. (line 43)
43761 * V850 Options: V850 Options. (line 6)
43762 * vague linkage: Vague Linkage. (line 6)
43763 * value after longjmp: Global Reg Vars. (line 66)
43764 * variable addressability on the IA-64: Function Attributes.
43766 * variable addressability on the M32R/D: Variable Attributes.
43768 * variable alignment: Alignment. (line 6)
43769 * variable attributes: Variable Attributes.
43771 * variable number of arguments: Variadic Macros. (line 6)
43772 * variable-length array scope: Variable Length. (line 23)
43773 * variable-length arrays: Variable Length. (line 6)
43774 * variables in specified registers: Explicit Reg Vars. (line 6)
43775 * variables, local, in macros: Typeof. (line 42)
43776 * variadic macros: Variadic Macros. (line 6)
43777 * VAX options: VAX Options. (line 6)
43778 * version_id attribute: Function Attributes.
43780 * vfprintf: Other Builtins. (line 6)
43781 * vfscanf: Other Builtins. (line 6)
43782 * visibility attribute: Function Attributes.
43784 * VLAs: Variable Length. (line 6)
43785 * void pointers, arithmetic: Pointer Arith. (line 6)
43786 * void, size of pointer to: Pointer Arith. (line 6)
43787 * volatile access: Volatiles. (line 6)
43788 * volatile applied to function: Function Attributes.
43790 * volatile read: Volatiles. (line 6)
43791 * volatile write: Volatiles. (line 6)
43792 * vprintf: Other Builtins. (line 6)
43793 * vscanf: Other Builtins. (line 6)
43794 * vsnprintf: Other Builtins. (line 6)
43795 * vsprintf: Other Builtins. (line 6)
43796 * vsscanf: Other Builtins. (line 6)
43797 * vtable: Vague Linkage. (line 28)
43798 * VxWorks Options: VxWorks Options. (line 6)
43799 * W floating point suffix: Floating Types. (line 6)
43800 * w floating point suffix: Floating Types. (line 6)
43801 * warn_unused_result attribute: Function Attributes.
43803 * warning for comparison of signed and unsigned values: Warning Options.
43805 * warning for overloaded virtual fn: C++ Dialect Options.
43807 * warning for reordering of member initializers: C++ Dialect Options.
43809 * warning for unknown pragmas: Warning Options. (line 587)
43810 * warning function attribute: Function Attributes.
43812 * warning messages: Warning Options. (line 6)
43813 * warnings from system headers: Warning Options. (line 701)
43814 * warnings vs errors: Warnings and Errors.
43816 * weak attribute: Function Attributes.
43818 * weakref attribute: Function Attributes.
43820 * whitespace: Incompatibilities. (line 112)
43821 * X in constraint: Simple Constraints. (line 114)
43822 * X3.159-1989: Standards. (line 13)
43823 * x86-64 options: x86-64 Options. (line 6)
43824 * x86-64 Options: i386 and x86-64 Options.
43826 * Xstormy16 Options: Xstormy16 Options. (line 6)
43827 * Xtensa Options: Xtensa Options. (line 6)
43828 * y0: Other Builtins. (line 6)
43829 * y0f: Other Builtins. (line 6)
43830 * y0l: Other Builtins. (line 6)
43831 * y1: Other Builtins. (line 6)
43832 * y1f: Other Builtins. (line 6)
43833 * y1l: Other Builtins. (line 6)
43834 * yn: Other Builtins. (line 6)
43835 * ynf: Other Builtins. (line 6)
43836 * ynl: Other Builtins. (line 6)
43837 * zero-length arrays: Zero Length. (line 6)
43838 * zero-size structures: Empty Structures. (line 6)
43839 * zSeries options: zSeries Options. (line 6)
43845 Node: G++ and GCC
\7f3758
43846 Node: Standards
\7f5823
43847 Node: Invoking GCC
\7f14798
43848 Node: Option Summary
\7f18627
43849 Node: Overall Options
\7f51342
43850 Node: Invoking G++
\7f65679
43851 Node: C Dialect Options
\7f67202
43852 Node: C++ Dialect Options
\7f81093
43853 Node: Objective-C and Objective-C++ Dialect Options
\7f102083
43854 Node: Language Independent Options
\7f113864
43855 Node: Warning Options
\7f116634
43856 Node: Debugging Options
\7f174981
43857 Node: Optimize Options
\7f214603
43858 Ref: Type-punning
\7f261405
43859 Node: Preprocessor Options
\7f318749
43860 Ref: Wtrigraphs
\7f322834
43861 Ref: dashMF
\7f327582
43862 Ref: fdollars-in-identifiers
\7f338101
43863 Node: Assembler Options
\7f346662
43864 Node: Link Options
\7f347367
43865 Ref: Link Options-Footnote-1
\7f356837
43866 Node: Directory Options
\7f357171
43867 Node: Spec Files
\7f363233
43868 Node: Target Options
\7f383572
43869 Node: Submodel Options
\7f385090
43870 Node: ARC Options
\7f386789
43871 Node: ARM Options
\7f388276
43872 Node: AVR Options
\7f401852
43873 Node: Blackfin Options
\7f404070
43874 Node: CRIS Options
\7f411962
43875 Node: CRX Options
\7f415703
43876 Node: Darwin Options
\7f416128
43877 Node: DEC Alpha Options
\7f423621
43878 Node: DEC Alpha/VMS Options
\7f435537
43879 Node: FR30 Options
\7f435923
43880 Node: FRV Options
\7f436498
43881 Node: GNU/Linux Options
\7f443215
43882 Node: H8/300 Options
\7f443673
43883 Node: HPPA Options
\7f444740
43884 Node: i386 and x86-64 Options
\7f454240
43885 Node: IA-64 Options
\7f482225
43886 Node: M32C Options
\7f489550
43887 Node: M32R/D Options
\7f490841
43888 Node: M680x0 Options
\7f494428
43889 Node: M68hc1x Options
\7f508248
43890 Node: MCore Options
\7f509816
43891 Node: MIPS Options
\7f511324
43892 Node: MMIX Options
\7f537359
43893 Node: MN10300 Options
\7f539841
43894 Node: PDP-11 Options
\7f541263
43895 Node: picoChip Options
\7f543103
43896 Node: PowerPC Options
\7f545302
43897 Node: RS/6000 and PowerPC Options
\7f545538
43898 Node: S/390 and zSeries Options
\7f576285
43899 Node: Score Options
\7f584233
43900 Node: SH Options
\7f585061
43901 Node: SPARC Options
\7f595339
43902 Node: SPU Options
\7f606312
43903 Node: System V Options
\7f609600
43904 Node: V850 Options
\7f610423
43905 Node: VAX Options
\7f613563
43906 Node: VxWorks Options
\7f614111
43907 Node: x86-64 Options
\7f615266
43908 Node: i386 and x86-64 Windows Options
\7f615484
43909 Node: Xstormy16 Options
\7f617803
43910 Node: Xtensa Options
\7f618092
43911 Node: zSeries Options
\7f622239
43912 Node: Code Gen Options
\7f622435
43913 Node: Environment Variables
\7f647014
43914 Node: Precompiled Headers
\7f654910
43915 Node: Running Protoize
\7f661136
43916 Node: C Implementation
\7f667473
43917 Node: Translation implementation
\7f669136
43918 Node: Environment implementation
\7f669710
43919 Node: Identifiers implementation
\7f670260
43920 Node: Characters implementation
\7f671314
43921 Node: Integers implementation
\7f674120
43922 Node: Floating point implementation
\7f675945
43923 Node: Arrays and pointers implementation
\7f678874
43924 Ref: Arrays and pointers implementation-Footnote-1
\7f680309
43925 Node: Hints implementation
\7f680433
43926 Node: Structures unions enumerations and bit-fields implementation
\7f681899
43927 Node: Qualifiers implementation
\7f683885
43928 Node: Declarators implementation
\7f685657
43929 Node: Statements implementation
\7f685999
43930 Node: Preprocessing directives implementation
\7f686326
43931 Node: Library functions implementation
\7f688431
43932 Node: Architecture implementation
\7f689071
43933 Node: Locale-specific behavior implementation
\7f689774
43934 Node: C Extensions
\7f690079
43935 Node: Statement Exprs
\7f694690
43936 Node: Local Labels
\7f699203
43937 Node: Labels as Values
\7f702182
43938 Ref: Labels as Values-Footnote-1
\7f704555
43939 Node: Nested Functions
\7f704738
43940 Node: Constructing Calls
\7f708632
43941 Node: Typeof
\7f713355
43942 Node: Conditionals
\7f716521
43943 Node: Long Long
\7f717412
43944 Node: Complex
\7f718913
43945 Node: Floating Types
\7f721483
43946 Node: Decimal Float
\7f722602
43947 Node: Hex Floats
\7f724591
43948 Node: Fixed-Point
\7f725632
43949 Node: Zero Length
\7f728917
43950 Node: Empty Structures
\7f732195
43951 Node: Variable Length
\7f732611
43952 Node: Variadic Macros
\7f735378
43953 Node: Escaped Newlines
\7f737760
43954 Node: Subscripting
\7f738599
43955 Node: Pointer Arith
\7f739322
43956 Node: Initializers
\7f739890
43957 Node: Compound Literals
\7f740386
43958 Node: Designated Inits
\7f742561
43959 Node: Case Ranges
\7f746216
43960 Node: Cast to Union
\7f746899
43961 Node: Mixed Declarations
\7f747995
43962 Node: Function Attributes
\7f748501
43963 Node: Attribute Syntax
\7f811117
43964 Node: Function Prototypes
\7f821387
43965 Node: C++ Comments
\7f823168
43966 Node: Dollar Signs
\7f823687
43967 Node: Character Escapes
\7f824152
43968 Node: Alignment
\7f824446
43969 Node: Variable Attributes
\7f825820
43970 Ref: i386 Variable Attributes
\7f840410
43971 Node: Type Attributes
\7f846395
43972 Ref: i386 Type Attributes
\7f860016
43973 Ref: PowerPC Type Attributes
\7f860856
43974 Ref: SPU Type Attributes
\7f861718
43975 Node: Inline
\7f862009
43976 Node: Extended Asm
\7f866956
43977 Ref: Example of asm with clobbered asm reg
\7f873042
43978 Node: Constraints
\7f887261
43979 Node: Simple Constraints
\7f888111
43980 Node: Multi-Alternative
\7f894782
43981 Node: Modifiers
\7f896499
43982 Node: Machine Constraints
\7f899393
43983 Node: Asm Labels
\7f931606
43984 Node: Explicit Reg Vars
\7f933282
43985 Node: Global Reg Vars
\7f934890
43986 Node: Local Reg Vars
\7f939440
43987 Node: Alternate Keywords
\7f941881
43988 Node: Incomplete Enums
\7f943309
43989 Node: Function Names
\7f944066
43990 Node: Return Address
\7f946228
43991 Node: Vector Extensions
\7f949025
43992 Node: Offsetof
\7f952527
43993 Node: Atomic Builtins
\7f953341
43994 Node: Object Size Checking
\7f958719
43995 Node: Other Builtins
\7f964147
43996 Node: Target Builtins
\7f988955
43997 Node: Alpha Built-in Functions
\7f989849
43998 Node: ARM iWMMXt Built-in Functions
\7f992848
43999 Node: ARM NEON Intrinsics
\7f999567
44000 Node: Blackfin Built-in Functions
\7f1207405
44001 Node: FR-V Built-in Functions
\7f1208019
44002 Node: Argument Types
\7f1208878
44003 Node: Directly-mapped Integer Functions
\7f1210634
44004 Node: Directly-mapped Media Functions
\7f1211716
44005 Node: Raw read/write Functions
\7f1218748
44006 Node: Other Built-in Functions
\7f1219660
44007 Node: X86 Built-in Functions
\7f1220849
44008 Node: MIPS DSP Built-in Functions
\7f1265240
44009 Node: MIPS Paired-Single Support
\7f1277687
44010 Node: MIPS Loongson Built-in Functions
\7f1279188
44011 Node: Paired-Single Arithmetic
\7f1285706
44012 Node: Paired-Single Built-in Functions
\7f1286652
44013 Node: MIPS-3D Built-in Functions
\7f1289322
44014 Node: picoChip Built-in Functions
\7f1294697
44015 Node: Other MIPS Built-in Functions
\7f1296059
44016 Node: PowerPC AltiVec Built-in Functions
\7f1296583
44017 Node: SPARC VIS Built-in Functions
\7f1398007
44018 Node: SPU Built-in Functions
\7f1399699
44019 Node: Target Format Checks
\7f1401481
44020 Node: Solaris Format Checks
\7f1401888
44021 Node: Pragmas
\7f1402285
44022 Node: ARM Pragmas
\7f1402979
44023 Node: M32C Pragmas
\7f1403582
44024 Node: RS/6000 and PowerPC Pragmas
\7f1404158
44025 Node: Darwin Pragmas
\7f1404900
44026 Node: Solaris Pragmas
\7f1405967
44027 Node: Symbol-Renaming Pragmas
\7f1407128
44028 Node: Structure-Packing Pragmas
\7f1409750
44029 Node: Weak Pragmas
\7f1411402
44030 Node: Diagnostic Pragmas
\7f1412204
44031 Node: Visibility Pragmas
\7f1414838
44032 Node: Push/Pop Macro Pragmas
\7f1415590
44033 Node: Function Specific Option Pragmas
\7f1416563
44034 Node: Unnamed Fields
\7f1418778
44035 Node: Thread-Local
\7f1420288
44036 Node: C99 Thread-Local Edits
\7f1422397
44037 Node: C++98 Thread-Local Edits
\7f1424409
44038 Node: Binary constants
\7f1427854
44039 Node: C++ Extensions
\7f1428525
44040 Node: Volatiles
\7f1430167
44041 Node: Restricted Pointers
\7f1432843
44042 Node: Vague Linkage
\7f1434437
44043 Node: C++ Interface
\7f1438093
44044 Ref: C++ Interface-Footnote-1
\7f1442390
44045 Node: Template Instantiation
\7f1442527
44046 Node: Bound member functions
\7f1449539
44047 Node: C++ Attributes
\7f1451082
44048 Node: Namespace Association
\7f1452740
44049 Node: Type Traits
\7f1454154
44050 Node: Java Exceptions
\7f1459701
44051 Node: Deprecated Features
\7f1461098
44052 Node: Backwards Compatibility
\7f1464063
44053 Node: Objective-C
\7f1465421
44054 Node: Executing code before main
\7f1466002
44055 Node: What you can and what you cannot do in +load
\7f1468608
44056 Node: Type encoding
\7f1470775
44057 Node: Garbage Collection
\7f1474162
44058 Node: Constant string objects
\7f1476786
44059 Node: compatibility_alias
\7f1479294
44060 Node: Compatibility
\7f1480172
44061 Node: Gcov
\7f1486739
44062 Node: Gcov Intro
\7f1487270
44063 Node: Invoking Gcov
\7f1489986
44064 Node: Gcov and Optimization
\7f1502067
44065 Node: Gcov Data Files
\7f1504720
44066 Node: Cross-profiling
\7f1505858
44067 Node: Trouble
\7f1507684
44068 Node: Actual Bugs
\7f1509240
44069 Node: Cross-Compiler Problems
\7f1509980
44070 Node: Interoperation
\7f1510394
44071 Node: Incompatibilities
\7f1517531
44072 Node: Fixed Headers
\7f1525681
44073 Node: Standard Libraries
\7f1527344
44074 Node: Disappointments
\7f1528716
44075 Node: C++ Misunderstandings
\7f1533074
44076 Node: Static Definitions
\7f1533893
44077 Node: Name lookup
\7f1534946
44078 Ref: Name lookup-Footnote-1
\7f1539724
44079 Node: Temporaries
\7f1539911
44080 Node: Copy Assignment
\7f1541887
44081 Node: Protoize Caveats
\7f1543694
44082 Node: Non-bugs
\7f1547667
44083 Node: Warnings and Errors
\7f1558171
44084 Node: Bugs
\7f1559935
44085 Node: Bug Criteria
\7f1560499
44086 Node: Bug Reporting
\7f1562709
44087 Node: Service
\7f1562930
44088 Node: Contributing
\7f1563749
44089 Node: Funding
\7f1564489
44090 Node: GNU Project
\7f1566978
44091 Node: Copying
\7f1567624
44092 Node: GNU Free Documentation License
\7f1605152
44093 Node: Contributors
\7f1627558
44094 Node: Option Index
\7f1663885
44095 Node: Keyword Index
\7f1823137