1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-call-cxx-cdtors @gol
212 -fobjc-direct-dispatch @gol
213 -fobjc-exceptions @gol
215 -freplace-objc-classes @gol
218 -Wassign-intercept @gol
219 -Wno-protocol -Wselector @gol
220 -Wstrict-selector-match @gol
221 -Wundeclared-selector}
223 @item Language Independent Options
224 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
225 @gccoptlist{-fmessage-length=@var{n} @gol
226 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
227 -fdiagnostics-show-option}
229 @item Warning Options
230 @xref{Warning Options,,Options to Request or Suppress Warnings}.
231 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
232 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
233 -Wno-attributes -Wno-builtin-macro-redefined @gol
234 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
235 -Wchar-subscripts -Wclobbered -Wcomment @gol
236 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
237 -Wno-deprecated-declarations -Wdisabled-optimization @gol
238 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
239 -Wno-endif-labels -Werror -Werror=* @gol
240 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
241 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
242 -Wformat-security -Wformat-y2k @gol
243 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
244 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
245 -Winit-self -Winline @gol
246 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
247 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
248 -Wlogical-op -Wlong-long @gol
249 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
250 -Wmissing-format-attribute -Wmissing-include-dirs @gol
252 -Wno-multichar -Wnonnull -Wno-overflow @gol
253 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
254 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
255 -Wpointer-arith -Wno-pointer-to-int-cast @gol
256 -Wredundant-decls @gol
257 -Wreturn-type -Wsequence-point -Wshadow @gol
258 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
259 -Wstrict-aliasing -Wstrict-aliasing=n @gol
260 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
261 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
262 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
263 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
264 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
265 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
266 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
267 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
268 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
270 @item C and Objective-C-only Warning Options
271 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
272 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
273 -Wold-style-declaration -Wold-style-definition @gol
274 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
275 -Wdeclaration-after-statement -Wpointer-sign}
277 @item Debugging Options
278 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
279 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
280 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
281 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
282 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
283 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
284 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
285 -fdump-statistics @gol
287 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
291 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-nrv -fdump-tree-vect @gol
301 -fdump-tree-sink @gol
302 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
306 -ftree-vectorizer-verbose=@var{n} @gol
307 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
308 -fdump-final-insns=@var{file} @gol
309 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
310 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
311 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
312 -fenable-icf-debug @gol
313 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
314 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
315 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
316 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
317 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
318 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
319 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
320 -gvms -gxcoff -gxcoff+ @gol
321 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
322 -fdebug-prefix-map=@var{old}=@var{new} @gol
323 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
324 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
325 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
326 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
327 -print-prog-name=@var{program} -print-search-dirs -Q @gol
328 -print-sysroot -print-sysroot-headers-suffix @gol
329 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
331 @item Optimization Options
332 @xref{Optimize Options,,Options that Control Optimization}.
333 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
334 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
335 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
336 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
337 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
338 -fcprop-registers -fcrossjumping @gol
339 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
340 -fcx-limited-range @gol
341 -fdata-sections -fdce -fdce @gol
342 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
343 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
344 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
345 -fforward-propagate -ffunction-sections @gol
346 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
347 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
348 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
349 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
350 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
351 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
352 -fira-region=@var{region} @gol
353 -fira-loop-pressure -fno-ira-share-save-slots @gol
354 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
355 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
356 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
357 -floop-parallelize-all -flto -flto-compression-level -flto-report @gol
358 -fltrans -fltrans-output-list -fmerge-all-constants -fmerge-constants @gol
359 -fmodulo-sched -fmodulo-sched-allow-regmoves -fmove-loop-invariants @gol
360 -fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
361 -fno-default-inline @gol
362 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
363 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
364 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
365 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
366 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
367 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
368 -fprefetch-loop-arrays @gol
369 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
370 -fprofile-generate=@var{path} @gol
371 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
372 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
373 -freorder-blocks-and-partition -freorder-functions @gol
374 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
375 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
376 -fsched-spec-load -fsched-spec-load-dangerous @gol
377 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
378 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
379 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
380 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
381 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
382 -fselective-scheduling -fselective-scheduling2 @gol
383 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
384 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
385 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
386 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
388 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
389 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
390 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
391 -ftree-loop-if-convert-memory-writes -ftree-loop-im @gol
392 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
393 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
394 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
395 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
396 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
397 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
398 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
399 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
400 -fwhole-program -fwhopr[=@var{n}] -fwpa -fuse-linker-plugin @gol
401 --param @var{name}=@var{value}
402 -O -O0 -O1 -O2 -O3 -Os -Ofast}
404 @item Preprocessor Options
405 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
406 @gccoptlist{-A@var{question}=@var{answer} @gol
407 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
408 -C -dD -dI -dM -dN @gol
409 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
410 -idirafter @var{dir} @gol
411 -include @var{file} -imacros @var{file} @gol
412 -iprefix @var{file} -iwithprefix @var{dir} @gol
413 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
414 -imultilib @var{dir} -isysroot @var{dir} @gol
415 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
416 -P -fworking-directory -remap @gol
417 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
418 -Xpreprocessor @var{option}}
420 @item Assembler Option
421 @xref{Assembler Options,,Passing Options to the Assembler}.
422 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
425 @xref{Link Options,,Options for Linking}.
426 @gccoptlist{@var{object-file-name} -l@var{library} @gol
427 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
428 -s -static -static-libgcc -static-libstdc++ -shared @gol
429 -shared-libgcc -symbolic @gol
430 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
433 @item Directory Options
434 @xref{Directory Options,,Options for Directory Search}.
435 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
436 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
439 @item Machine Dependent Options
440 @xref{Submodel Options,,Hardware Models and Configurations}.
441 @c This list is ordered alphanumerically by subsection name.
442 @c Try and put the significant identifier (CPU or system) first,
443 @c so users have a clue at guessing where the ones they want will be.
446 @gccoptlist{-EB -EL @gol
447 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
448 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
451 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
452 -mabi=@var{name} @gol
453 -mapcs-stack-check -mno-apcs-stack-check @gol
454 -mapcs-float -mno-apcs-float @gol
455 -mapcs-reentrant -mno-apcs-reentrant @gol
456 -msched-prolog -mno-sched-prolog @gol
457 -mlittle-endian -mbig-endian -mwords-little-endian @gol
458 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
459 -mfp16-format=@var{name}
460 -mthumb-interwork -mno-thumb-interwork @gol
461 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
462 -mstructure-size-boundary=@var{n} @gol
463 -mabort-on-noreturn @gol
464 -mlong-calls -mno-long-calls @gol
465 -msingle-pic-base -mno-single-pic-base @gol
466 -mpic-register=@var{reg} @gol
467 -mnop-fun-dllimport @gol
468 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
469 -mpoke-function-name @gol
471 -mtpcs-frame -mtpcs-leaf-frame @gol
472 -mcaller-super-interworking -mcallee-super-interworking @gol
474 -mword-relocations @gol
475 -mfix-cortex-m3-ldrd}
478 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
479 -mcall-prologues -mtiny-stack -mint8}
481 @emph{Blackfin Options}
482 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
483 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
484 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
485 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
486 -mno-id-shared-library -mshared-library-id=@var{n} @gol
487 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
488 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
489 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
493 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
494 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
495 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
496 -mstack-align -mdata-align -mconst-align @gol
497 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
498 -melf -maout -melinux -mlinux -sim -sim2 @gol
499 -mmul-bug-workaround -mno-mul-bug-workaround}
502 @gccoptlist{-mmac -mpush-args}
504 @emph{Darwin Options}
505 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
506 -arch_only -bind_at_load -bundle -bundle_loader @gol
507 -client_name -compatibility_version -current_version @gol
509 -dependency-file -dylib_file -dylinker_install_name @gol
510 -dynamic -dynamiclib -exported_symbols_list @gol
511 -filelist -flat_namespace -force_cpusubtype_ALL @gol
512 -force_flat_namespace -headerpad_max_install_names @gol
514 -image_base -init -install_name -keep_private_externs @gol
515 -multi_module -multiply_defined -multiply_defined_unused @gol
516 -noall_load -no_dead_strip_inits_and_terms @gol
517 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
518 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
519 -private_bundle -read_only_relocs -sectalign @gol
520 -sectobjectsymbols -whyload -seg1addr @gol
521 -sectcreate -sectobjectsymbols -sectorder @gol
522 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
523 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
524 -segprot -segs_read_only_addr -segs_read_write_addr @gol
525 -single_module -static -sub_library -sub_umbrella @gol
526 -twolevel_namespace -umbrella -undefined @gol
527 -unexported_symbols_list -weak_reference_mismatches @gol
528 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
529 -mkernel -mone-byte-bool}
531 @emph{DEC Alpha Options}
532 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
533 -mieee -mieee-with-inexact -mieee-conformant @gol
534 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
535 -mtrap-precision=@var{mode} -mbuild-constants @gol
536 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
537 -mbwx -mmax -mfix -mcix @gol
538 -mfloat-vax -mfloat-ieee @gol
539 -mexplicit-relocs -msmall-data -mlarge-data @gol
540 -msmall-text -mlarge-text @gol
541 -mmemory-latency=@var{time}}
543 @emph{DEC Alpha/VMS Options}
544 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
547 @gccoptlist{-msmall-model -mno-lsim}
550 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
551 -mhard-float -msoft-float @gol
552 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
553 -mdouble -mno-double @gol
554 -mmedia -mno-media -mmuladd -mno-muladd @gol
555 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
556 -mlinked-fp -mlong-calls -malign-labels @gol
557 -mlibrary-pic -macc-4 -macc-8 @gol
558 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
559 -moptimize-membar -mno-optimize-membar @gol
560 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
561 -mvliw-branch -mno-vliw-branch @gol
562 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
563 -mno-nested-cond-exec -mtomcat-stats @gol
567 @emph{GNU/Linux Options}
568 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
569 -tno-android-cc -tno-android-ld}
571 @emph{H8/300 Options}
572 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
575 @gccoptlist{-march=@var{architecture-type} @gol
576 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
577 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
578 -mfixed-range=@var{register-range} @gol
579 -mjump-in-delay -mlinker-opt -mlong-calls @gol
580 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
581 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
582 -mno-jump-in-delay -mno-long-load-store @gol
583 -mno-portable-runtime -mno-soft-float @gol
584 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
585 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
586 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
587 -munix=@var{unix-std} -nolibdld -static -threads}
589 @emph{i386 and x86-64 Options}
590 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
591 -mfpmath=@var{unit} @gol
592 -masm=@var{dialect} -mno-fancy-math-387 @gol
593 -mno-fp-ret-in-387 -msoft-float @gol
594 -mno-wide-multiply -mrtd -malign-double @gol
595 -mpreferred-stack-boundary=@var{num}
596 -mincoming-stack-boundary=@var{num} @gol
597 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
598 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
599 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
600 -msse4a -m3dnow -mpopcnt -mabm -mfma4 -mxop -mlwp @gol
601 -mthreads -mno-align-stringops -minline-all-stringops @gol
602 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
603 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
604 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
605 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
606 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
607 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
608 -mcmodel=@var{code-model} -mabi=@var{name} @gol
609 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
610 -msse2avx -mfentry -m8bit-idiv}
613 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
614 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
615 -mconstant-gp -mauto-pic -mfused-madd @gol
616 -minline-float-divide-min-latency @gol
617 -minline-float-divide-max-throughput @gol
618 -mno-inline-float-divide @gol
619 -minline-int-divide-min-latency @gol
620 -minline-int-divide-max-throughput @gol
621 -mno-inline-int-divide @gol
622 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
623 -mno-inline-sqrt @gol
624 -mdwarf2-asm -mearly-stop-bits @gol
625 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
626 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
627 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
628 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
629 -msched-spec-ldc -msched-spec-control-ldc @gol
630 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
631 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
632 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
633 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
635 @emph{IA-64/VMS Options}
636 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
639 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
640 -msign-extend-enabled -muser-enabled}
642 @emph{M32R/D Options}
643 @gccoptlist{-m32r2 -m32rx -m32r @gol
645 -malign-loops -mno-align-loops @gol
646 -missue-rate=@var{number} @gol
647 -mbranch-cost=@var{number} @gol
648 -mmodel=@var{code-size-model-type} @gol
649 -msdata=@var{sdata-type} @gol
650 -mno-flush-func -mflush-func=@var{name} @gol
651 -mno-flush-trap -mflush-trap=@var{number} @gol
655 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
657 @emph{M680x0 Options}
658 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
659 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
660 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
661 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
662 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
663 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
664 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
665 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
668 @emph{M68hc1x Options}
669 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
670 -mauto-incdec -minmax -mlong-calls -mshort @gol
671 -msoft-reg-count=@var{count}}
674 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
675 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
676 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
677 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
678 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
681 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
682 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
683 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
684 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
687 @emph{MicroBlaze Options}
688 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
689 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
690 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
691 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
692 -mxl-mode-@var{app-model}}
695 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
696 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
697 -mips64 -mips64r2 @gol
698 -mips16 -mno-mips16 -mflip-mips16 @gol
699 -minterlink-mips16 -mno-interlink-mips16 @gol
700 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
701 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
702 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
703 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
704 -mfpu=@var{fpu-type} @gol
705 -msmartmips -mno-smartmips @gol
706 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
707 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
708 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
709 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
710 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
711 -membedded-data -mno-embedded-data @gol
712 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
713 -mcode-readable=@var{setting} @gol
714 -msplit-addresses -mno-split-addresses @gol
715 -mexplicit-relocs -mno-explicit-relocs @gol
716 -mcheck-zero-division -mno-check-zero-division @gol
717 -mdivide-traps -mdivide-breaks @gol
718 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
719 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
720 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
721 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
722 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
723 -mflush-func=@var{func} -mno-flush-func @gol
724 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
725 -mfp-exceptions -mno-fp-exceptions @gol
726 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
727 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
730 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
731 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
732 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
733 -mno-base-addresses -msingle-exit -mno-single-exit}
735 @emph{MN10300 Options}
736 @gccoptlist{-mmult-bug -mno-mult-bug @gol
737 -mam33 -mno-am33 @gol
738 -mam33-2 -mno-am33-2 @gol
739 -mreturn-pointer-on-d0 @gol
742 @emph{PDP-11 Options}
743 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
744 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
745 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
746 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
747 -mbranch-expensive -mbranch-cheap @gol
748 -msplit -mno-split -munix-asm -mdec-asm}
750 @emph{picoChip Options}
751 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
752 -msymbol-as-address -mno-inefficient-warnings}
754 @emph{PowerPC Options}
755 See RS/6000 and PowerPC Options.
757 @emph{RS/6000 and PowerPC Options}
758 @gccoptlist{-mcpu=@var{cpu-type} @gol
759 -mtune=@var{cpu-type} @gol
760 -mcmodel=@var{code-model} @gol
761 -mpower -mno-power -mpower2 -mno-power2 @gol
762 -mpowerpc -mpowerpc64 -mno-powerpc @gol
763 -maltivec -mno-altivec @gol
764 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
765 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
766 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
767 -mfprnd -mno-fprnd @gol
768 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
769 -mnew-mnemonics -mold-mnemonics @gol
770 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
771 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
772 -malign-power -malign-natural @gol
773 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
774 -msingle-float -mdouble-float -msimple-fpu @gol
775 -mstring -mno-string -mupdate -mno-update @gol
776 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
777 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
778 -mstrict-align -mno-strict-align -mrelocatable @gol
779 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
780 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
781 -mdynamic-no-pic -maltivec -mswdiv @gol
782 -mprioritize-restricted-insns=@var{priority} @gol
783 -msched-costly-dep=@var{dependence_type} @gol
784 -minsert-sched-nops=@var{scheme} @gol
785 -mcall-sysv -mcall-netbsd @gol
786 -maix-struct-return -msvr4-struct-return @gol
787 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
788 -mblock-move-inline-limit=@var{num} @gol
789 -misel -mno-isel @gol
790 -misel=yes -misel=no @gol
792 -mspe=yes -mspe=no @gol
794 -mgen-cell-microcode -mwarn-cell-microcode @gol
795 -mvrsave -mno-vrsave @gol
796 -mmulhw -mno-mulhw @gol
797 -mdlmzb -mno-dlmzb @gol
798 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
799 -mprototype -mno-prototype @gol
800 -msim -mmvme -mads -myellowknife -memb -msdata @gol
801 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
802 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision
803 -mno-recip-precision @gol
804 -mveclibabi=@var{type} -mfriz -mno-friz}
807 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
809 -mbig-endian-data -mlittle-endian-data @gol
812 -mas100-syntax -mno-as100-syntax@gol
814 -mmax-constant-size=@gol
816 -msave-acc-in-interrupts}
818 @emph{S/390 and zSeries Options}
819 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
820 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
821 -mlong-double-64 -mlong-double-128 @gol
822 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
823 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
824 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
825 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
826 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
829 @gccoptlist{-meb -mel @gol
833 -mscore5 -mscore5u -mscore7 -mscore7d}
836 @gccoptlist{-m1 -m2 -m2e @gol
837 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
839 -m4-nofpu -m4-single-only -m4-single -m4 @gol
840 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
841 -m5-64media -m5-64media-nofpu @gol
842 -m5-32media -m5-32media-nofpu @gol
843 -m5-compact -m5-compact-nofpu @gol
844 -mb -ml -mdalign -mrelax @gol
845 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
846 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
847 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
848 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
849 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
850 -maccumulate-outgoing-args -minvalid-symbols}
852 @emph{Solaris 2 Options}
853 @gccoptlist{-mimpure-text -mno-impure-text @gol
854 -threads -pthreads -pthread}
857 @gccoptlist{-mcpu=@var{cpu-type} @gol
858 -mtune=@var{cpu-type} @gol
859 -mcmodel=@var{code-model} @gol
860 -m32 -m64 -mapp-regs -mno-app-regs @gol
861 -mfaster-structs -mno-faster-structs @gol
862 -mfpu -mno-fpu -mhard-float -msoft-float @gol
863 -mhard-quad-float -msoft-quad-float @gol
865 -mstack-bias -mno-stack-bias @gol
866 -munaligned-doubles -mno-unaligned-doubles @gol
867 -mv8plus -mno-v8plus -mvis -mno-vis}
870 @gccoptlist{-mwarn-reloc -merror-reloc @gol
871 -msafe-dma -munsafe-dma @gol
873 -msmall-mem -mlarge-mem -mstdmain @gol
874 -mfixed-range=@var{register-range} @gol
876 -maddress-space-conversion -mno-address-space-conversion @gol
877 -mcache-size=@var{cache-size} @gol
878 -matomic-updates -mno-atomic-updates}
880 @emph{System V Options}
881 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
884 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
885 -mprolog-function -mno-prolog-function -mspace @gol
886 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
887 -mapp-regs -mno-app-regs @gol
888 -mdisable-callt -mno-disable-callt @gol
896 @gccoptlist{-mg -mgnu -munix}
898 @emph{VxWorks Options}
899 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
900 -Xbind-lazy -Xbind-now}
902 @emph{x86-64 Options}
903 See i386 and x86-64 Options.
905 @emph{i386 and x86-64 Windows Options}
906 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
907 -mnop-fun-dllimport -mthread @gol
908 -municode -mwin32 -mwindows -fno-set-stack-executable}
910 @emph{Xstormy16 Options}
913 @emph{Xtensa Options}
914 @gccoptlist{-mconst16 -mno-const16 @gol
915 -mfused-madd -mno-fused-madd @gol
917 -mserialize-volatile -mno-serialize-volatile @gol
918 -mtext-section-literals -mno-text-section-literals @gol
919 -mtarget-align -mno-target-align @gol
920 -mlongcalls -mno-longcalls}
922 @emph{zSeries Options}
923 See S/390 and zSeries Options.
925 @item Code Generation Options
926 @xref{Code Gen Options,,Options for Code Generation Conventions}.
927 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
928 -ffixed-@var{reg} -fexceptions @gol
929 -fnon-call-exceptions -funwind-tables @gol
930 -fasynchronous-unwind-tables @gol
931 -finhibit-size-directive -finstrument-functions @gol
932 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
933 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
934 -fno-common -fno-ident @gol
935 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
936 -fno-jump-tables @gol
937 -frecord-gcc-switches @gol
938 -freg-struct-return -fshort-enums @gol
939 -fshort-double -fshort-wchar @gol
940 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
941 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
942 -fno-stack-limit -fsplit-stack @gol
943 -fleading-underscore -ftls-model=@var{model} @gol
944 -ftrapv -fwrapv -fbounds-check @gol
949 * Overall Options:: Controlling the kind of output:
950 an executable, object files, assembler files,
951 or preprocessed source.
952 * C Dialect Options:: Controlling the variant of C language compiled.
953 * C++ Dialect Options:: Variations on C++.
954 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
956 * Language Independent Options:: Controlling how diagnostics should be
958 * Warning Options:: How picky should the compiler be?
959 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
960 * Optimize Options:: How much optimization?
961 * Preprocessor Options:: Controlling header files and macro definitions.
962 Also, getting dependency information for Make.
963 * Assembler Options:: Passing options to the assembler.
964 * Link Options:: Specifying libraries and so on.
965 * Directory Options:: Where to find header files and libraries.
966 Where to find the compiler executable files.
967 * Spec Files:: How to pass switches to sub-processes.
968 * Target Options:: Running a cross-compiler, or an old version of GCC.
971 @node Overall Options
972 @section Options Controlling the Kind of Output
974 Compilation can involve up to four stages: preprocessing, compilation
975 proper, assembly and linking, always in that order. GCC is capable of
976 preprocessing and compiling several files either into several
977 assembler input files, or into one assembler input file; then each
978 assembler input file produces an object file, and linking combines all
979 the object files (those newly compiled, and those specified as input)
980 into an executable file.
982 @cindex file name suffix
983 For any given input file, the file name suffix determines what kind of
988 C source code which must be preprocessed.
991 C source code which should not be preprocessed.
994 C++ source code which should not be preprocessed.
997 Objective-C source code. Note that you must link with the @file{libobjc}
998 library to make an Objective-C program work.
1001 Objective-C source code which should not be preprocessed.
1005 Objective-C++ source code. Note that you must link with the @file{libobjc}
1006 library to make an Objective-C++ program work. Note that @samp{.M} refers
1007 to a literal capital M@.
1009 @item @var{file}.mii
1010 Objective-C++ source code which should not be preprocessed.
1013 C, C++, Objective-C or Objective-C++ header file to be turned into a
1014 precompiled header (default), or C, C++ header file to be turned into an
1015 Ada spec (via the @option{-fdump-ada-spec} switch).
1018 @itemx @var{file}.cp
1019 @itemx @var{file}.cxx
1020 @itemx @var{file}.cpp
1021 @itemx @var{file}.CPP
1022 @itemx @var{file}.c++
1024 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1025 the last two letters must both be literally @samp{x}. Likewise,
1026 @samp{.C} refers to a literal capital C@.
1030 Objective-C++ source code which must be preprocessed.
1032 @item @var{file}.mii
1033 Objective-C++ source code which should not be preprocessed.
1037 @itemx @var{file}.hp
1038 @itemx @var{file}.hxx
1039 @itemx @var{file}.hpp
1040 @itemx @var{file}.HPP
1041 @itemx @var{file}.h++
1042 @itemx @var{file}.tcc
1043 C++ header file to be turned into a precompiled header or Ada spec.
1046 @itemx @var{file}.for
1047 @itemx @var{file}.ftn
1048 Fixed form Fortran source code which should not be preprocessed.
1051 @itemx @var{file}.FOR
1052 @itemx @var{file}.fpp
1053 @itemx @var{file}.FPP
1054 @itemx @var{file}.FTN
1055 Fixed form Fortran source code which must be preprocessed (with the traditional
1058 @item @var{file}.f90
1059 @itemx @var{file}.f95
1060 @itemx @var{file}.f03
1061 @itemx @var{file}.f08
1062 Free form Fortran source code which should not be preprocessed.
1064 @item @var{file}.F90
1065 @itemx @var{file}.F95
1066 @itemx @var{file}.F03
1067 @itemx @var{file}.F08
1068 Free form Fortran source code which must be preprocessed (with the
1069 traditional preprocessor).
1071 @c FIXME: Descriptions of Java file types.
1077 @item @var{file}.ads
1078 Ada source code file which contains a library unit declaration (a
1079 declaration of a package, subprogram, or generic, or a generic
1080 instantiation), or a library unit renaming declaration (a package,
1081 generic, or subprogram renaming declaration). Such files are also
1084 @item @var{file}.adb
1085 Ada source code file containing a library unit body (a subprogram or
1086 package body). Such files are also called @dfn{bodies}.
1088 @c GCC also knows about some suffixes for languages not yet included:
1099 @itemx @var{file}.sx
1100 Assembler code which must be preprocessed.
1103 An object file to be fed straight into linking.
1104 Any file name with no recognized suffix is treated this way.
1108 You can specify the input language explicitly with the @option{-x} option:
1111 @item -x @var{language}
1112 Specify explicitly the @var{language} for the following input files
1113 (rather than letting the compiler choose a default based on the file
1114 name suffix). This option applies to all following input files until
1115 the next @option{-x} option. Possible values for @var{language} are:
1117 c c-header cpp-output
1118 c++ c++-header c++-cpp-output
1119 objective-c objective-c-header objective-c-cpp-output
1120 objective-c++ objective-c++-header objective-c++-cpp-output
1121 assembler assembler-with-cpp
1123 f77 f77-cpp-input f95 f95-cpp-input
1128 Turn off any specification of a language, so that subsequent files are
1129 handled according to their file name suffixes (as they are if @option{-x}
1130 has not been used at all).
1132 @item -pass-exit-codes
1133 @opindex pass-exit-codes
1134 Normally the @command{gcc} program will exit with the code of 1 if any
1135 phase of the compiler returns a non-success return code. If you specify
1136 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1137 numerically highest error produced by any phase that returned an error
1138 indication. The C, C++, and Fortran frontends return 4, if an internal
1139 compiler error is encountered.
1142 If you only want some of the stages of compilation, you can use
1143 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1144 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1145 @command{gcc} is to stop. Note that some combinations (for example,
1146 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1151 Compile or assemble the source files, but do not link. The linking
1152 stage simply is not done. The ultimate output is in the form of an
1153 object file for each source file.
1155 By default, the object file name for a source file is made by replacing
1156 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1158 Unrecognized input files, not requiring compilation or assembly, are
1163 Stop after the stage of compilation proper; do not assemble. The output
1164 is in the form of an assembler code file for each non-assembler input
1167 By default, the assembler file name for a source file is made by
1168 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1170 Input files that don't require compilation are ignored.
1174 Stop after the preprocessing stage; do not run the compiler proper. The
1175 output is in the form of preprocessed source code, which is sent to the
1178 Input files which don't require preprocessing are ignored.
1180 @cindex output file option
1183 Place output in file @var{file}. This applies regardless to whatever
1184 sort of output is being produced, whether it be an executable file,
1185 an object file, an assembler file or preprocessed C code.
1187 If @option{-o} is not specified, the default is to put an executable
1188 file in @file{a.out}, the object file for
1189 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1190 assembler file in @file{@var{source}.s}, a precompiled header file in
1191 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1196 Print (on standard error output) the commands executed to run the stages
1197 of compilation. Also print the version number of the compiler driver
1198 program and of the preprocessor and the compiler proper.
1202 Like @option{-v} except the commands are not executed and arguments
1203 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1204 This is useful for shell scripts to capture the driver-generated command lines.
1208 Use pipes rather than temporary files for communication between the
1209 various stages of compilation. This fails to work on some systems where
1210 the assembler is unable to read from a pipe; but the GNU assembler has
1215 Print (on the standard output) a description of the command line options
1216 understood by @command{gcc}. If the @option{-v} option is also specified
1217 then @option{--help} will also be passed on to the various processes
1218 invoked by @command{gcc}, so that they can display the command line options
1219 they accept. If the @option{-Wextra} option has also been specified
1220 (prior to the @option{--help} option), then command line options which
1221 have no documentation associated with them will also be displayed.
1224 @opindex target-help
1225 Print (on the standard output) a description of target-specific command
1226 line options for each tool. For some targets extra target-specific
1227 information may also be printed.
1229 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1230 Print (on the standard output) a description of the command line
1231 options understood by the compiler that fit into all specified classes
1232 and qualifiers. These are the supported classes:
1235 @item @samp{optimizers}
1236 This will display all of the optimization options supported by the
1239 @item @samp{warnings}
1240 This will display all of the options controlling warning messages
1241 produced by the compiler.
1244 This will display target-specific options. Unlike the
1245 @option{--target-help} option however, target-specific options of the
1246 linker and assembler will not be displayed. This is because those
1247 tools do not currently support the extended @option{--help=} syntax.
1250 This will display the values recognized by the @option{--param}
1253 @item @var{language}
1254 This will display the options supported for @var{language}, where
1255 @var{language} is the name of one of the languages supported in this
1259 This will display the options that are common to all languages.
1262 These are the supported qualifiers:
1265 @item @samp{undocumented}
1266 Display only those options which are undocumented.
1269 Display options which take an argument that appears after an equal
1270 sign in the same continuous piece of text, such as:
1271 @samp{--help=target}.
1273 @item @samp{separate}
1274 Display options which take an argument that appears as a separate word
1275 following the original option, such as: @samp{-o output-file}.
1278 Thus for example to display all the undocumented target-specific
1279 switches supported by the compiler the following can be used:
1282 --help=target,undocumented
1285 The sense of a qualifier can be inverted by prefixing it with the
1286 @samp{^} character, so for example to display all binary warning
1287 options (i.e., ones that are either on or off and that do not take an
1288 argument), which have a description the following can be used:
1291 --help=warnings,^joined,^undocumented
1294 The argument to @option{--help=} should not consist solely of inverted
1297 Combining several classes is possible, although this usually
1298 restricts the output by so much that there is nothing to display. One
1299 case where it does work however is when one of the classes is
1300 @var{target}. So for example to display all the target-specific
1301 optimization options the following can be used:
1304 --help=target,optimizers
1307 The @option{--help=} option can be repeated on the command line. Each
1308 successive use will display its requested class of options, skipping
1309 those that have already been displayed.
1311 If the @option{-Q} option appears on the command line before the
1312 @option{--help=} option, then the descriptive text displayed by
1313 @option{--help=} is changed. Instead of describing the displayed
1314 options, an indication is given as to whether the option is enabled,
1315 disabled or set to a specific value (assuming that the compiler
1316 knows this at the point where the @option{--help=} option is used).
1318 Here is a truncated example from the ARM port of @command{gcc}:
1321 % gcc -Q -mabi=2 --help=target -c
1322 The following options are target specific:
1324 -mabort-on-noreturn [disabled]
1328 The output is sensitive to the effects of previous command line
1329 options, so for example it is possible to find out which optimizations
1330 are enabled at @option{-O2} by using:
1333 -Q -O2 --help=optimizers
1336 Alternatively you can discover which binary optimizations are enabled
1337 by @option{-O3} by using:
1340 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1341 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1342 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1345 @item -no-canonical-prefixes
1346 @opindex no-canonical-prefixes
1347 Do not expand any symbolic links, resolve references to @samp{/../}
1348 or @samp{/./}, or make the path absolute when generating a relative
1353 Display the version number and copyrights of the invoked GCC@.
1357 Invoke all subcommands under a wrapper program. It takes a single
1358 comma separated list as an argument, which will be used to invoke
1362 gcc -c t.c -wrapper gdb,--args
1365 This will invoke all subprograms of gcc under "gdb --args",
1366 thus cc1 invocation will be "gdb --args cc1 ...".
1368 @item -fplugin=@var{name}.so
1369 Load the plugin code in file @var{name}.so, assumed to be a
1370 shared object to be dlopen'd by the compiler. The base name of
1371 the shared object file is used to identify the plugin for the
1372 purposes of argument parsing (See
1373 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1374 Each plugin should define the callback functions specified in the
1377 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1378 Define an argument called @var{key} with a value of @var{value}
1379 for the plugin called @var{name}.
1381 @item -fdump-ada-spec@r{[}-slim@r{]}
1382 For C and C++ source and include files, generate corresponding Ada
1383 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1384 GNAT User's Guide}, which provides detailed documentation on this feature.
1386 @include @value{srcdir}/../libiberty/at-file.texi
1390 @section Compiling C++ Programs
1392 @cindex suffixes for C++ source
1393 @cindex C++ source file suffixes
1394 C++ source files conventionally use one of the suffixes @samp{.C},
1395 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1396 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1397 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1398 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1399 files with these names and compiles them as C++ programs even if you
1400 call the compiler the same way as for compiling C programs (usually
1401 with the name @command{gcc}).
1405 However, the use of @command{gcc} does not add the C++ library.
1406 @command{g++} is a program that calls GCC and treats @samp{.c},
1407 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1408 files unless @option{-x} is used, and automatically specifies linking
1409 against the C++ library. This program is also useful when
1410 precompiling a C header file with a @samp{.h} extension for use in C++
1411 compilations. On many systems, @command{g++} is also installed with
1412 the name @command{c++}.
1414 @cindex invoking @command{g++}
1415 When you compile C++ programs, you may specify many of the same
1416 command-line options that you use for compiling programs in any
1417 language; or command-line options meaningful for C and related
1418 languages; or options that are meaningful only for C++ programs.
1419 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1420 explanations of options for languages related to C@.
1421 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1422 explanations of options that are meaningful only for C++ programs.
1424 @node C Dialect Options
1425 @section Options Controlling C Dialect
1426 @cindex dialect options
1427 @cindex language dialect options
1428 @cindex options, dialect
1430 The following options control the dialect of C (or languages derived
1431 from C, such as C++, Objective-C and Objective-C++) that the compiler
1435 @cindex ANSI support
1439 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1440 equivalent to @samp{-std=c++98}.
1442 This turns off certain features of GCC that are incompatible with ISO
1443 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1444 such as the @code{asm} and @code{typeof} keywords, and
1445 predefined macros such as @code{unix} and @code{vax} that identify the
1446 type of system you are using. It also enables the undesirable and
1447 rarely used ISO trigraph feature. For the C compiler,
1448 it disables recognition of C++ style @samp{//} comments as well as
1449 the @code{inline} keyword.
1451 The alternate keywords @code{__asm__}, @code{__extension__},
1452 @code{__inline__} and @code{__typeof__} continue to work despite
1453 @option{-ansi}. You would not want to use them in an ISO C program, of
1454 course, but it is useful to put them in header files that might be included
1455 in compilations done with @option{-ansi}. Alternate predefined macros
1456 such as @code{__unix__} and @code{__vax__} are also available, with or
1457 without @option{-ansi}.
1459 The @option{-ansi} option does not cause non-ISO programs to be
1460 rejected gratuitously. For that, @option{-pedantic} is required in
1461 addition to @option{-ansi}. @xref{Warning Options}.
1463 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1464 option is used. Some header files may notice this macro and refrain
1465 from declaring certain functions or defining certain macros that the
1466 ISO standard doesn't call for; this is to avoid interfering with any
1467 programs that might use these names for other things.
1469 Functions that would normally be built in but do not have semantics
1470 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1471 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1472 built-in functions provided by GCC}, for details of the functions
1477 Determine the language standard. @xref{Standards,,Language Standards
1478 Supported by GCC}, for details of these standard versions. This option
1479 is currently only supported when compiling C or C++.
1481 The compiler can accept several base standards, such as @samp{c90} or
1482 @samp{c++98}, and GNU dialects of those standards, such as
1483 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1484 compiler will accept all programs following that standard and those
1485 using GNU extensions that do not contradict it. For example,
1486 @samp{-std=c90} turns off certain features of GCC that are
1487 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1488 keywords, but not other GNU extensions that do not have a meaning in
1489 ISO C90, such as omitting the middle term of a @code{?:}
1490 expression. On the other hand, by specifying a GNU dialect of a
1491 standard, all features the compiler support are enabled, even when
1492 those features change the meaning of the base standard and some
1493 strict-conforming programs may be rejected. The particular standard
1494 is used by @option{-pedantic} to identify which features are GNU
1495 extensions given that version of the standard. For example
1496 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1497 comments, while @samp{-std=gnu99 -pedantic} would not.
1499 A value for this option must be provided; possible values are
1505 Support all ISO C90 programs (certain GNU extensions that conflict
1506 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1508 @item iso9899:199409
1509 ISO C90 as modified in amendment 1.
1515 ISO C99. Note that this standard is not yet fully supported; see
1516 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1517 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1520 ISO C1X, the draft of the next revision of the ISO C standard.
1521 Support is limited and experimental and features enabled by this
1522 option may be changed or removed if changed in or removed from the
1527 GNU dialect of ISO C90 (including some C99 features). This
1528 is the default for C code.
1532 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1533 this will become the default. The name @samp{gnu9x} is deprecated.
1536 GNU dialect of ISO C1X. Support is limited and experimental and
1537 features enabled by this option may be changed or removed if changed
1538 in or removed from the standard draft.
1541 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1545 GNU dialect of @option{-std=c++98}. This is the default for
1549 The working draft of the upcoming ISO C++0x standard. This option
1550 enables experimental features that are likely to be included in
1551 C++0x. The working draft is constantly changing, and any feature that is
1552 enabled by this flag may be removed from future versions of GCC if it is
1553 not part of the C++0x standard.
1556 GNU dialect of @option{-std=c++0x}. This option enables
1557 experimental features that may be removed in future versions of GCC.
1560 @item -fgnu89-inline
1561 @opindex fgnu89-inline
1562 The option @option{-fgnu89-inline} tells GCC to use the traditional
1563 GNU semantics for @code{inline} functions when in C99 mode.
1564 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1565 is accepted and ignored by GCC versions 4.1.3 up to but not including
1566 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1567 C99 mode. Using this option is roughly equivalent to adding the
1568 @code{gnu_inline} function attribute to all inline functions
1569 (@pxref{Function Attributes}).
1571 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1572 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1573 specifies the default behavior). This option was first supported in
1574 GCC 4.3. This option is not supported in @option{-std=c90} or
1575 @option{-std=gnu90} mode.
1577 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1578 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1579 in effect for @code{inline} functions. @xref{Common Predefined
1580 Macros,,,cpp,The C Preprocessor}.
1582 @item -aux-info @var{filename}
1584 Output to the given filename prototyped declarations for all functions
1585 declared and/or defined in a translation unit, including those in header
1586 files. This option is silently ignored in any language other than C@.
1588 Besides declarations, the file indicates, in comments, the origin of
1589 each declaration (source file and line), whether the declaration was
1590 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1591 @samp{O} for old, respectively, in the first character after the line
1592 number and the colon), and whether it came from a declaration or a
1593 definition (@samp{C} or @samp{F}, respectively, in the following
1594 character). In the case of function definitions, a K&R-style list of
1595 arguments followed by their declarations is also provided, inside
1596 comments, after the declaration.
1600 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1601 keyword, so that code can use these words as identifiers. You can use
1602 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1603 instead. @option{-ansi} implies @option{-fno-asm}.
1605 In C++, this switch only affects the @code{typeof} keyword, since
1606 @code{asm} and @code{inline} are standard keywords. You may want to
1607 use the @option{-fno-gnu-keywords} flag instead, which has the same
1608 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1609 switch only affects the @code{asm} and @code{typeof} keywords, since
1610 @code{inline} is a standard keyword in ISO C99.
1613 @itemx -fno-builtin-@var{function}
1614 @opindex fno-builtin
1615 @cindex built-in functions
1616 Don't recognize built-in functions that do not begin with
1617 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1618 functions provided by GCC}, for details of the functions affected,
1619 including those which are not built-in functions when @option{-ansi} or
1620 @option{-std} options for strict ISO C conformance are used because they
1621 do not have an ISO standard meaning.
1623 GCC normally generates special code to handle certain built-in functions
1624 more efficiently; for instance, calls to @code{alloca} may become single
1625 instructions that adjust the stack directly, and calls to @code{memcpy}
1626 may become inline copy loops. The resulting code is often both smaller
1627 and faster, but since the function calls no longer appear as such, you
1628 cannot set a breakpoint on those calls, nor can you change the behavior
1629 of the functions by linking with a different library. In addition,
1630 when a function is recognized as a built-in function, GCC may use
1631 information about that function to warn about problems with calls to
1632 that function, or to generate more efficient code, even if the
1633 resulting code still contains calls to that function. For example,
1634 warnings are given with @option{-Wformat} for bad calls to
1635 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1636 known not to modify global memory.
1638 With the @option{-fno-builtin-@var{function}} option
1639 only the built-in function @var{function} is
1640 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1641 function is named that is not built-in in this version of GCC, this
1642 option is ignored. There is no corresponding
1643 @option{-fbuiltin-@var{function}} option; if you wish to enable
1644 built-in functions selectively when using @option{-fno-builtin} or
1645 @option{-ffreestanding}, you may define macros such as:
1648 #define abs(n) __builtin_abs ((n))
1649 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1654 @cindex hosted environment
1656 Assert that compilation takes place in a hosted environment. This implies
1657 @option{-fbuiltin}. A hosted environment is one in which the
1658 entire standard library is available, and in which @code{main} has a return
1659 type of @code{int}. Examples are nearly everything except a kernel.
1660 This is equivalent to @option{-fno-freestanding}.
1662 @item -ffreestanding
1663 @opindex ffreestanding
1664 @cindex hosted environment
1666 Assert that compilation takes place in a freestanding environment. This
1667 implies @option{-fno-builtin}. A freestanding environment
1668 is one in which the standard library may not exist, and program startup may
1669 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1670 This is equivalent to @option{-fno-hosted}.
1672 @xref{Standards,,Language Standards Supported by GCC}, for details of
1673 freestanding and hosted environments.
1677 @cindex OpenMP parallel
1678 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1679 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1680 compiler generates parallel code according to the OpenMP Application
1681 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1682 implies @option{-pthread}, and thus is only supported on targets that
1683 have support for @option{-pthread}.
1685 @item -fms-extensions
1686 @opindex fms-extensions
1687 Accept some non-standard constructs used in Microsoft header files.
1689 It allows for c++ that member-names in structures can be similiar
1690 to previous types declarations.
1699 Some cases of unnamed fields in structures and unions are only
1700 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1701 fields within structs/unions}, for details.
1703 @item -fplan9-extensions
1704 Accept some non-standard constructs used in Plan 9 code.
1706 This enables @option{-fms-extensions}, permits passing pointers to
1707 structures with anonymous fields to functions which expect pointers to
1708 elements of the type of the field, and permits referring to anonymous
1709 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1710 struct/union fields within structs/unions}, for details. This is only
1711 supported for C, not C++.
1715 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1716 options for strict ISO C conformance) implies @option{-trigraphs}.
1718 @item -no-integrated-cpp
1719 @opindex no-integrated-cpp
1720 Performs a compilation in two passes: preprocessing and compiling. This
1721 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1722 @option{-B} option. The user supplied compilation step can then add in
1723 an additional preprocessing step after normal preprocessing but before
1724 compiling. The default is to use the integrated cpp (internal cpp)
1726 The semantics of this option will change if "cc1", "cc1plus", and
1727 "cc1obj" are merged.
1729 @cindex traditional C language
1730 @cindex C language, traditional
1732 @itemx -traditional-cpp
1733 @opindex traditional-cpp
1734 @opindex traditional
1735 Formerly, these options caused GCC to attempt to emulate a pre-standard
1736 C compiler. They are now only supported with the @option{-E} switch.
1737 The preprocessor continues to support a pre-standard mode. See the GNU
1738 CPP manual for details.
1740 @item -fcond-mismatch
1741 @opindex fcond-mismatch
1742 Allow conditional expressions with mismatched types in the second and
1743 third arguments. The value of such an expression is void. This option
1744 is not supported for C++.
1746 @item -flax-vector-conversions
1747 @opindex flax-vector-conversions
1748 Allow implicit conversions between vectors with differing numbers of
1749 elements and/or incompatible element types. This option should not be
1752 @item -funsigned-char
1753 @opindex funsigned-char
1754 Let the type @code{char} be unsigned, like @code{unsigned char}.
1756 Each kind of machine has a default for what @code{char} should
1757 be. It is either like @code{unsigned char} by default or like
1758 @code{signed char} by default.
1760 Ideally, a portable program should always use @code{signed char} or
1761 @code{unsigned char} when it depends on the signedness of an object.
1762 But many programs have been written to use plain @code{char} and
1763 expect it to be signed, or expect it to be unsigned, depending on the
1764 machines they were written for. This option, and its inverse, let you
1765 make such a program work with the opposite default.
1767 The type @code{char} is always a distinct type from each of
1768 @code{signed char} or @code{unsigned char}, even though its behavior
1769 is always just like one of those two.
1772 @opindex fsigned-char
1773 Let the type @code{char} be signed, like @code{signed char}.
1775 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1776 the negative form of @option{-funsigned-char}. Likewise, the option
1777 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1779 @item -fsigned-bitfields
1780 @itemx -funsigned-bitfields
1781 @itemx -fno-signed-bitfields
1782 @itemx -fno-unsigned-bitfields
1783 @opindex fsigned-bitfields
1784 @opindex funsigned-bitfields
1785 @opindex fno-signed-bitfields
1786 @opindex fno-unsigned-bitfields
1787 These options control whether a bit-field is signed or unsigned, when the
1788 declaration does not use either @code{signed} or @code{unsigned}. By
1789 default, such a bit-field is signed, because this is consistent: the
1790 basic integer types such as @code{int} are signed types.
1793 @node C++ Dialect Options
1794 @section Options Controlling C++ Dialect
1796 @cindex compiler options, C++
1797 @cindex C++ options, command line
1798 @cindex options, C++
1799 This section describes the command-line options that are only meaningful
1800 for C++ programs; but you can also use most of the GNU compiler options
1801 regardless of what language your program is in. For example, you
1802 might compile a file @code{firstClass.C} like this:
1805 g++ -g -frepo -O -c firstClass.C
1809 In this example, only @option{-frepo} is an option meant
1810 only for C++ programs; you can use the other options with any
1811 language supported by GCC@.
1813 Here is a list of options that are @emph{only} for compiling C++ programs:
1817 @item -fabi-version=@var{n}
1818 @opindex fabi-version
1819 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1820 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1821 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1822 the version that conforms most closely to the C++ ABI specification.
1823 Therefore, the ABI obtained using version 0 will change as ABI bugs
1826 The default is version 2.
1828 Version 3 corrects an error in mangling a constant address as a
1831 Version 4 implements a standard mangling for vector types.
1833 See also @option{-Wabi}.
1835 @item -fno-access-control
1836 @opindex fno-access-control
1837 Turn off all access checking. This switch is mainly useful for working
1838 around bugs in the access control code.
1842 Check that the pointer returned by @code{operator new} is non-null
1843 before attempting to modify the storage allocated. This check is
1844 normally unnecessary because the C++ standard specifies that
1845 @code{operator new} will only return @code{0} if it is declared
1846 @samp{throw()}, in which case the compiler will always check the
1847 return value even without this option. In all other cases, when
1848 @code{operator new} has a non-empty exception specification, memory
1849 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1850 @samp{new (nothrow)}.
1852 @item -fconserve-space
1853 @opindex fconserve-space
1854 Put uninitialized or runtime-initialized global variables into the
1855 common segment, as C does. This saves space in the executable at the
1856 cost of not diagnosing duplicate definitions. If you compile with this
1857 flag and your program mysteriously crashes after @code{main()} has
1858 completed, you may have an object that is being destroyed twice because
1859 two definitions were merged.
1861 This option is no longer useful on most targets, now that support has
1862 been added for putting variables into BSS without making them common.
1864 @item -fno-deduce-init-list
1865 @opindex fno-deduce-init-list
1866 Disable deduction of a template type parameter as
1867 std::initializer_list from a brace-enclosed initializer list, i.e.
1870 template <class T> auto forward(T t) -> decltype (realfn (t))
1877 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1881 This option is present because this deduction is an extension to the
1882 current specification in the C++0x working draft, and there was
1883 some concern about potential overload resolution problems.
1885 @item -ffriend-injection
1886 @opindex ffriend-injection
1887 Inject friend functions into the enclosing namespace, so that they are
1888 visible outside the scope of the class in which they are declared.
1889 Friend functions were documented to work this way in the old Annotated
1890 C++ Reference Manual, and versions of G++ before 4.1 always worked
1891 that way. However, in ISO C++ a friend function which is not declared
1892 in an enclosing scope can only be found using argument dependent
1893 lookup. This option causes friends to be injected as they were in
1896 This option is for compatibility, and may be removed in a future
1899 @item -fno-elide-constructors
1900 @opindex fno-elide-constructors
1901 The C++ standard allows an implementation to omit creating a temporary
1902 which is only used to initialize another object of the same type.
1903 Specifying this option disables that optimization, and forces G++ to
1904 call the copy constructor in all cases.
1906 @item -fno-enforce-eh-specs
1907 @opindex fno-enforce-eh-specs
1908 Don't generate code to check for violation of exception specifications
1909 at runtime. This option violates the C++ standard, but may be useful
1910 for reducing code size in production builds, much like defining
1911 @samp{NDEBUG}. This does not give user code permission to throw
1912 exceptions in violation of the exception specifications; the compiler
1913 will still optimize based on the specifications, so throwing an
1914 unexpected exception will result in undefined behavior.
1917 @itemx -fno-for-scope
1919 @opindex fno-for-scope
1920 If @option{-ffor-scope} is specified, the scope of variables declared in
1921 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1922 as specified by the C++ standard.
1923 If @option{-fno-for-scope} is specified, the scope of variables declared in
1924 a @i{for-init-statement} extends to the end of the enclosing scope,
1925 as was the case in old versions of G++, and other (traditional)
1926 implementations of C++.
1928 The default if neither flag is given to follow the standard,
1929 but to allow and give a warning for old-style code that would
1930 otherwise be invalid, or have different behavior.
1932 @item -fno-gnu-keywords
1933 @opindex fno-gnu-keywords
1934 Do not recognize @code{typeof} as a keyword, so that code can use this
1935 word as an identifier. You can use the keyword @code{__typeof__} instead.
1936 @option{-ansi} implies @option{-fno-gnu-keywords}.
1938 @item -fno-implicit-templates
1939 @opindex fno-implicit-templates
1940 Never emit code for non-inline templates which are instantiated
1941 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1942 @xref{Template Instantiation}, for more information.
1944 @item -fno-implicit-inline-templates
1945 @opindex fno-implicit-inline-templates
1946 Don't emit code for implicit instantiations of inline templates, either.
1947 The default is to handle inlines differently so that compiles with and
1948 without optimization will need the same set of explicit instantiations.
1950 @item -fno-implement-inlines
1951 @opindex fno-implement-inlines
1952 To save space, do not emit out-of-line copies of inline functions
1953 controlled by @samp{#pragma implementation}. This will cause linker
1954 errors if these functions are not inlined everywhere they are called.
1956 @item -fms-extensions
1957 @opindex fms-extensions
1958 Disable pedantic warnings about constructs used in MFC, such as implicit
1959 int and getting a pointer to member function via non-standard syntax.
1961 @item -fno-nonansi-builtins
1962 @opindex fno-nonansi-builtins
1963 Disable built-in declarations of functions that are not mandated by
1964 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1965 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1968 @opindex fnothrow-opt
1969 Treat a @code{throw()} exception specification as though it were a
1970 @code{noexcept} specification to reduce or eliminate the text size
1971 overhead relative to a function with no exception specification. If
1972 the function has local variables of types with non-trivial
1973 destructors, the exception specification will actually make the
1974 function smaller because the EH cleanups for those variables can be
1975 optimized away. The semantic effect is that an exception thrown out of
1976 a function with such an exception specification will result in a call
1977 to @code{terminate} rather than @code{unexpected}.
1979 @item -fno-operator-names
1980 @opindex fno-operator-names
1981 Do not treat the operator name keywords @code{and}, @code{bitand},
1982 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1983 synonyms as keywords.
1985 @item -fno-optional-diags
1986 @opindex fno-optional-diags
1987 Disable diagnostics that the standard says a compiler does not need to
1988 issue. Currently, the only such diagnostic issued by G++ is the one for
1989 a name having multiple meanings within a class.
1992 @opindex fpermissive
1993 Downgrade some diagnostics about nonconformant code from errors to
1994 warnings. Thus, using @option{-fpermissive} will allow some
1995 nonconforming code to compile.
1997 @item -fno-pretty-templates
1998 @opindex fno-pretty-templates
1999 When an error message refers to a specialization of a function
2000 template, the compiler will normally print the signature of the
2001 template followed by the template arguments and any typedefs or
2002 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2003 rather than @code{void f(int)}) so that it's clear which template is
2004 involved. When an error message refers to a specialization of a class
2005 template, the compiler will omit any template arguments which match
2006 the default template arguments for that template. If either of these
2007 behaviors make it harder to understand the error message rather than
2008 easier, using @option{-fno-pretty-templates} will disable them.
2012 Enable automatic template instantiation at link time. This option also
2013 implies @option{-fno-implicit-templates}. @xref{Template
2014 Instantiation}, for more information.
2018 Disable generation of information about every class with virtual
2019 functions for use by the C++ runtime type identification features
2020 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2021 of the language, you can save some space by using this flag. Note that
2022 exception handling uses the same information, but it will generate it as
2023 needed. The @samp{dynamic_cast} operator can still be used for casts that
2024 do not require runtime type information, i.e.@: casts to @code{void *} or to
2025 unambiguous base classes.
2029 Emit statistics about front-end processing at the end of the compilation.
2030 This information is generally only useful to the G++ development team.
2032 @item -fstrict-enums
2033 @opindex fstrict-enums
2034 Allow the compiler to optimize using the assumption that a value of
2035 enumeration type can only be one of the values of the enumeration (as
2036 defined in the C++ standard; basically, a value which can be
2037 represented in the minimum number of bits needed to represent all the
2038 enumerators). This assumption may not be valid if the program uses a
2039 cast to convert an arbitrary integer value to the enumeration type.
2041 @item -ftemplate-depth=@var{n}
2042 @opindex ftemplate-depth
2043 Set the maximum instantiation depth for template classes to @var{n}.
2044 A limit on the template instantiation depth is needed to detect
2045 endless recursions during template class instantiation. ANSI/ISO C++
2046 conforming programs must not rely on a maximum depth greater than 17
2047 (changed to 1024 in C++0x).
2049 @item -fno-threadsafe-statics
2050 @opindex fno-threadsafe-statics
2051 Do not emit the extra code to use the routines specified in the C++
2052 ABI for thread-safe initialization of local statics. You can use this
2053 option to reduce code size slightly in code that doesn't need to be
2056 @item -fuse-cxa-atexit
2057 @opindex fuse-cxa-atexit
2058 Register destructors for objects with static storage duration with the
2059 @code{__cxa_atexit} function rather than the @code{atexit} function.
2060 This option is required for fully standards-compliant handling of static
2061 destructors, but will only work if your C library supports
2062 @code{__cxa_atexit}.
2064 @item -fno-use-cxa-get-exception-ptr
2065 @opindex fno-use-cxa-get-exception-ptr
2066 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2067 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2068 if the runtime routine is not available.
2070 @item -fvisibility-inlines-hidden
2071 @opindex fvisibility-inlines-hidden
2072 This switch declares that the user does not attempt to compare
2073 pointers to inline methods where the addresses of the two functions
2074 were taken in different shared objects.
2076 The effect of this is that GCC may, effectively, mark inline methods with
2077 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2078 appear in the export table of a DSO and do not require a PLT indirection
2079 when used within the DSO@. Enabling this option can have a dramatic effect
2080 on load and link times of a DSO as it massively reduces the size of the
2081 dynamic export table when the library makes heavy use of templates.
2083 The behavior of this switch is not quite the same as marking the
2084 methods as hidden directly, because it does not affect static variables
2085 local to the function or cause the compiler to deduce that
2086 the function is defined in only one shared object.
2088 You may mark a method as having a visibility explicitly to negate the
2089 effect of the switch for that method. For example, if you do want to
2090 compare pointers to a particular inline method, you might mark it as
2091 having default visibility. Marking the enclosing class with explicit
2092 visibility will have no effect.
2094 Explicitly instantiated inline methods are unaffected by this option
2095 as their linkage might otherwise cross a shared library boundary.
2096 @xref{Template Instantiation}.
2098 @item -fvisibility-ms-compat
2099 @opindex fvisibility-ms-compat
2100 This flag attempts to use visibility settings to make GCC's C++
2101 linkage model compatible with that of Microsoft Visual Studio.
2103 The flag makes these changes to GCC's linkage model:
2107 It sets the default visibility to @code{hidden}, like
2108 @option{-fvisibility=hidden}.
2111 Types, but not their members, are not hidden by default.
2114 The One Definition Rule is relaxed for types without explicit
2115 visibility specifications which are defined in more than one different
2116 shared object: those declarations are permitted if they would have
2117 been permitted when this option was not used.
2120 In new code it is better to use @option{-fvisibility=hidden} and
2121 export those classes which are intended to be externally visible.
2122 Unfortunately it is possible for code to rely, perhaps accidentally,
2123 on the Visual Studio behavior.
2125 Among the consequences of these changes are that static data members
2126 of the same type with the same name but defined in different shared
2127 objects will be different, so changing one will not change the other;
2128 and that pointers to function members defined in different shared
2129 objects may not compare equal. When this flag is given, it is a
2130 violation of the ODR to define types with the same name differently.
2134 Do not use weak symbol support, even if it is provided by the linker.
2135 By default, G++ will use weak symbols if they are available. This
2136 option exists only for testing, and should not be used by end-users;
2137 it will result in inferior code and has no benefits. This option may
2138 be removed in a future release of G++.
2142 Do not search for header files in the standard directories specific to
2143 C++, but do still search the other standard directories. (This option
2144 is used when building the C++ library.)
2147 In addition, these optimization, warning, and code generation options
2148 have meanings only for C++ programs:
2151 @item -fno-default-inline
2152 @opindex fno-default-inline
2153 Do not assume @samp{inline} for functions defined inside a class scope.
2154 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2155 functions will have linkage like inline functions; they just won't be
2158 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2161 Warn when G++ generates code that is probably not compatible with the
2162 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2163 all such cases, there are probably some cases that are not warned about,
2164 even though G++ is generating incompatible code. There may also be
2165 cases where warnings are emitted even though the code that is generated
2168 You should rewrite your code to avoid these warnings if you are
2169 concerned about the fact that code generated by G++ may not be binary
2170 compatible with code generated by other compilers.
2172 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2177 A template with a non-type template parameter of reference type is
2178 mangled incorrectly:
2181 template <int &> struct S @{@};
2185 This is fixed in @option{-fabi-version=3}.
2188 SIMD vector types declared using @code{__attribute ((vector_size))} are
2189 mangled in a non-standard way that does not allow for overloading of
2190 functions taking vectors of different sizes.
2192 The mangling is changed in @option{-fabi-version=4}.
2195 The known incompatibilities in @option{-fabi-version=1} include:
2200 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2201 pack data into the same byte as a base class. For example:
2204 struct A @{ virtual void f(); int f1 : 1; @};
2205 struct B : public A @{ int f2 : 1; @};
2209 In this case, G++ will place @code{B::f2} into the same byte
2210 as@code{A::f1}; other compilers will not. You can avoid this problem
2211 by explicitly padding @code{A} so that its size is a multiple of the
2212 byte size on your platform; that will cause G++ and other compilers to
2213 layout @code{B} identically.
2216 Incorrect handling of tail-padding for virtual bases. G++ does not use
2217 tail padding when laying out virtual bases. For example:
2220 struct A @{ virtual void f(); char c1; @};
2221 struct B @{ B(); char c2; @};
2222 struct C : public A, public virtual B @{@};
2226 In this case, G++ will not place @code{B} into the tail-padding for
2227 @code{A}; other compilers will. You can avoid this problem by
2228 explicitly padding @code{A} so that its size is a multiple of its
2229 alignment (ignoring virtual base classes); that will cause G++ and other
2230 compilers to layout @code{C} identically.
2233 Incorrect handling of bit-fields with declared widths greater than that
2234 of their underlying types, when the bit-fields appear in a union. For
2238 union U @{ int i : 4096; @};
2242 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2243 union too small by the number of bits in an @code{int}.
2246 Empty classes can be placed at incorrect offsets. For example:
2256 struct C : public B, public A @{@};
2260 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2261 it should be placed at offset zero. G++ mistakenly believes that the
2262 @code{A} data member of @code{B} is already at offset zero.
2265 Names of template functions whose types involve @code{typename} or
2266 template template parameters can be mangled incorrectly.
2269 template <typename Q>
2270 void f(typename Q::X) @{@}
2272 template <template <typename> class Q>
2273 void f(typename Q<int>::X) @{@}
2277 Instantiations of these templates may be mangled incorrectly.
2281 It also warns psABI related changes. The known psABI changes at this
2287 For SYSV/x86-64, when passing union with long double, it is changed to
2288 pass in memory as specified in psABI. For example:
2298 @code{union U} will always be passed in memory.
2302 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2303 @opindex Wctor-dtor-privacy
2304 @opindex Wno-ctor-dtor-privacy
2305 Warn when a class seems unusable because all the constructors or
2306 destructors in that class are private, and it has neither friends nor
2307 public static member functions.
2309 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2311 @opindex Wno-noexcept
2312 Warn when a noexcept-expression evaluates to false because of a call
2313 to a function that does not have a non-throwing exception
2314 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2315 the compiler to never throw an exception.
2317 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2318 @opindex Wnon-virtual-dtor
2319 @opindex Wno-non-virtual-dtor
2320 Warn when a class has virtual functions and accessible non-virtual
2321 destructor, in which case it would be possible but unsafe to delete
2322 an instance of a derived class through a pointer to the base class.
2323 This warning is also enabled if -Weffc++ is specified.
2325 @item -Wreorder @r{(C++ and Objective-C++ only)}
2327 @opindex Wno-reorder
2328 @cindex reordering, warning
2329 @cindex warning for reordering of member initializers
2330 Warn when the order of member initializers given in the code does not
2331 match the order in which they must be executed. For instance:
2337 A(): j (0), i (1) @{ @}
2341 The compiler will rearrange the member initializers for @samp{i}
2342 and @samp{j} to match the declaration order of the members, emitting
2343 a warning to that effect. This warning is enabled by @option{-Wall}.
2346 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2349 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2352 Warn about violations of the following style guidelines from Scott Meyers'
2353 @cite{Effective C++} book:
2357 Item 11: Define a copy constructor and an assignment operator for classes
2358 with dynamically allocated memory.
2361 Item 12: Prefer initialization to assignment in constructors.
2364 Item 14: Make destructors virtual in base classes.
2367 Item 15: Have @code{operator=} return a reference to @code{*this}.
2370 Item 23: Don't try to return a reference when you must return an object.
2374 Also warn about violations of the following style guidelines from
2375 Scott Meyers' @cite{More Effective C++} book:
2379 Item 6: Distinguish between prefix and postfix forms of increment and
2380 decrement operators.
2383 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2387 When selecting this option, be aware that the standard library
2388 headers do not obey all of these guidelines; use @samp{grep -v}
2389 to filter out those warnings.
2391 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2392 @opindex Wstrict-null-sentinel
2393 @opindex Wno-strict-null-sentinel
2394 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2395 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2396 to @code{__null}. Although it is a null pointer constant not a null pointer,
2397 it is guaranteed to be of the same size as a pointer. But this use is
2398 not portable across different compilers.
2400 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2401 @opindex Wno-non-template-friend
2402 @opindex Wnon-template-friend
2403 Disable warnings when non-templatized friend functions are declared
2404 within a template. Since the advent of explicit template specification
2405 support in G++, if the name of the friend is an unqualified-id (i.e.,
2406 @samp{friend foo(int)}), the C++ language specification demands that the
2407 friend declare or define an ordinary, nontemplate function. (Section
2408 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2409 could be interpreted as a particular specialization of a templatized
2410 function. Because this non-conforming behavior is no longer the default
2411 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2412 check existing code for potential trouble spots and is on by default.
2413 This new compiler behavior can be turned off with
2414 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2415 but disables the helpful warning.
2417 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2418 @opindex Wold-style-cast
2419 @opindex Wno-old-style-cast
2420 Warn if an old-style (C-style) cast to a non-void type is used within
2421 a C++ program. The new-style casts (@samp{dynamic_cast},
2422 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2423 less vulnerable to unintended effects and much easier to search for.
2425 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2426 @opindex Woverloaded-virtual
2427 @opindex Wno-overloaded-virtual
2428 @cindex overloaded virtual function, warning
2429 @cindex warning for overloaded virtual function
2430 Warn when a function declaration hides virtual functions from a
2431 base class. For example, in:
2438 struct B: public A @{
2443 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2451 will fail to compile.
2453 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2454 @opindex Wno-pmf-conversions
2455 @opindex Wpmf-conversions
2456 Disable the diagnostic for converting a bound pointer to member function
2459 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2460 @opindex Wsign-promo
2461 @opindex Wno-sign-promo
2462 Warn when overload resolution chooses a promotion from unsigned or
2463 enumerated type to a signed type, over a conversion to an unsigned type of
2464 the same size. Previous versions of G++ would try to preserve
2465 unsignedness, but the standard mandates the current behavior.
2470 A& operator = (int);
2480 In this example, G++ will synthesize a default @samp{A& operator =
2481 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2484 @node Objective-C and Objective-C++ Dialect Options
2485 @section Options Controlling Objective-C and Objective-C++ Dialects
2487 @cindex compiler options, Objective-C and Objective-C++
2488 @cindex Objective-C and Objective-C++ options, command line
2489 @cindex options, Objective-C and Objective-C++
2490 (NOTE: This manual does not describe the Objective-C and Objective-C++
2491 languages themselves. See @xref{Standards,,Language Standards
2492 Supported by GCC}, for references.)
2494 This section describes the command-line options that are only meaningful
2495 for Objective-C and Objective-C++ programs, but you can also use most of
2496 the language-independent GNU compiler options.
2497 For example, you might compile a file @code{some_class.m} like this:
2500 gcc -g -fgnu-runtime -O -c some_class.m
2504 In this example, @option{-fgnu-runtime} is an option meant only for
2505 Objective-C and Objective-C++ programs; you can use the other options with
2506 any language supported by GCC@.
2508 Note that since Objective-C is an extension of the C language, Objective-C
2509 compilations may also use options specific to the C front-end (e.g.,
2510 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2511 C++-specific options (e.g., @option{-Wabi}).
2513 Here is a list of options that are @emph{only} for compiling Objective-C
2514 and Objective-C++ programs:
2517 @item -fconstant-string-class=@var{class-name}
2518 @opindex fconstant-string-class
2519 Use @var{class-name} as the name of the class to instantiate for each
2520 literal string specified with the syntax @code{@@"@dots{}"}. The default
2521 class name is @code{NXConstantString} if the GNU runtime is being used, and
2522 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2523 @option{-fconstant-cfstrings} option, if also present, will override the
2524 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2525 to be laid out as constant CoreFoundation strings.
2528 @opindex fgnu-runtime
2529 Generate object code compatible with the standard GNU Objective-C
2530 runtime. This is the default for most types of systems.
2532 @item -fnext-runtime
2533 @opindex fnext-runtime
2534 Generate output compatible with the NeXT runtime. This is the default
2535 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2536 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2539 @item -fno-nil-receivers
2540 @opindex fno-nil-receivers
2541 Assume that all Objective-C message dispatches (@code{[receiver
2542 message:arg]}) in this translation unit ensure that the receiver is
2543 not @code{nil}. This allows for more efficient entry points in the
2544 runtime to be used. Currently, this option is only available in
2545 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2547 @item -fobjc-call-cxx-cdtors
2548 @opindex fobjc-call-cxx-cdtors
2549 For each Objective-C class, check if any of its instance variables is a
2550 C++ object with a non-trivial default constructor. If so, synthesize a
2551 special @code{- (id) .cxx_construct} instance method that will run
2552 non-trivial default constructors on any such instance variables, in order,
2553 and then return @code{self}. Similarly, check if any instance variable
2554 is a C++ object with a non-trivial destructor, and if so, synthesize a
2555 special @code{- (void) .cxx_destruct} method that will run
2556 all such default destructors, in reverse order.
2558 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2559 methods thusly generated will only operate on instance variables
2560 declared in the current Objective-C class, and not those inherited
2561 from superclasses. It is the responsibility of the Objective-C
2562 runtime to invoke all such methods in an object's inheritance
2563 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2564 by the runtime immediately after a new object instance is allocated;
2565 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2566 before the runtime deallocates an object instance.
2568 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2569 support for invoking the @code{- (id) .cxx_construct} and
2570 @code{- (void) .cxx_destruct} methods.
2572 @item -fobjc-direct-dispatch
2573 @opindex fobjc-direct-dispatch
2574 Allow fast jumps to the message dispatcher. On Darwin this is
2575 accomplished via the comm page.
2577 @item -fobjc-exceptions
2578 @opindex fobjc-exceptions
2579 Enable syntactic support for structured exception handling in
2580 Objective-C, similar to what is offered by C++ and Java. This option
2581 is required to use the Objective-C keywords @code{@@try},
2582 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2583 @code{@@synchronized}. This option is available with both the GNU
2584 runtime and the NeXT runtime (but not available in conjunction with
2585 the NeXT runtime on Mac OS X 10.2 and earlier).
2589 Enable garbage collection (GC) in Objective-C and Objective-C++
2590 programs. This option is only available with the NeXT runtime; the
2591 GNU runtime has a different garbage collection implementation that
2592 does not require special compiler flags.
2594 @item -freplace-objc-classes
2595 @opindex freplace-objc-classes
2596 Emit a special marker instructing @command{ld(1)} not to statically link in
2597 the resulting object file, and allow @command{dyld(1)} to load it in at
2598 run time instead. This is used in conjunction with the Fix-and-Continue
2599 debugging mode, where the object file in question may be recompiled and
2600 dynamically reloaded in the course of program execution, without the need
2601 to restart the program itself. Currently, Fix-and-Continue functionality
2602 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2607 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2608 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2609 compile time) with static class references that get initialized at load time,
2610 which improves run-time performance. Specifying the @option{-fzero-link} flag
2611 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2612 to be retained. This is useful in Zero-Link debugging mode, since it allows
2613 for individual class implementations to be modified during program execution.
2614 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2615 regardless of command line options.
2619 Dump interface declarations for all classes seen in the source file to a
2620 file named @file{@var{sourcename}.decl}.
2622 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2623 @opindex Wassign-intercept
2624 @opindex Wno-assign-intercept
2625 Warn whenever an Objective-C assignment is being intercepted by the
2628 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2629 @opindex Wno-protocol
2631 If a class is declared to implement a protocol, a warning is issued for
2632 every method in the protocol that is not implemented by the class. The
2633 default behavior is to issue a warning for every method not explicitly
2634 implemented in the class, even if a method implementation is inherited
2635 from the superclass. If you use the @option{-Wno-protocol} option, then
2636 methods inherited from the superclass are considered to be implemented,
2637 and no warning is issued for them.
2639 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2641 @opindex Wno-selector
2642 Warn if multiple methods of different types for the same selector are
2643 found during compilation. The check is performed on the list of methods
2644 in the final stage of compilation. Additionally, a check is performed
2645 for each selector appearing in a @code{@@selector(@dots{})}
2646 expression, and a corresponding method for that selector has been found
2647 during compilation. Because these checks scan the method table only at
2648 the end of compilation, these warnings are not produced if the final
2649 stage of compilation is not reached, for example because an error is
2650 found during compilation, or because the @option{-fsyntax-only} option is
2653 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2654 @opindex Wstrict-selector-match
2655 @opindex Wno-strict-selector-match
2656 Warn if multiple methods with differing argument and/or return types are
2657 found for a given selector when attempting to send a message using this
2658 selector to a receiver of type @code{id} or @code{Class}. When this flag
2659 is off (which is the default behavior), the compiler will omit such warnings
2660 if any differences found are confined to types which share the same size
2663 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2664 @opindex Wundeclared-selector
2665 @opindex Wno-undeclared-selector
2666 Warn if a @code{@@selector(@dots{})} expression referring to an
2667 undeclared selector is found. A selector is considered undeclared if no
2668 method with that name has been declared before the
2669 @code{@@selector(@dots{})} expression, either explicitly in an
2670 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2671 an @code{@@implementation} section. This option always performs its
2672 checks as soon as a @code{@@selector(@dots{})} expression is found,
2673 while @option{-Wselector} only performs its checks in the final stage of
2674 compilation. This also enforces the coding style convention
2675 that methods and selectors must be declared before being used.
2677 @item -print-objc-runtime-info
2678 @opindex print-objc-runtime-info
2679 Generate C header describing the largest structure that is passed by
2684 @node Language Independent Options
2685 @section Options to Control Diagnostic Messages Formatting
2686 @cindex options to control diagnostics formatting
2687 @cindex diagnostic messages
2688 @cindex message formatting
2690 Traditionally, diagnostic messages have been formatted irrespective of
2691 the output device's aspect (e.g.@: its width, @dots{}). The options described
2692 below can be used to control the diagnostic messages formatting
2693 algorithm, e.g.@: how many characters per line, how often source location
2694 information should be reported. Right now, only the C++ front end can
2695 honor these options. However it is expected, in the near future, that
2696 the remaining front ends would be able to digest them correctly.
2699 @item -fmessage-length=@var{n}
2700 @opindex fmessage-length
2701 Try to format error messages so that they fit on lines of about @var{n}
2702 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2703 the front ends supported by GCC@. If @var{n} is zero, then no
2704 line-wrapping will be done; each error message will appear on a single
2707 @opindex fdiagnostics-show-location
2708 @item -fdiagnostics-show-location=once
2709 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2710 reporter to emit @emph{once} source location information; that is, in
2711 case the message is too long to fit on a single physical line and has to
2712 be wrapped, the source location won't be emitted (as prefix) again,
2713 over and over, in subsequent continuation lines. This is the default
2716 @item -fdiagnostics-show-location=every-line
2717 Only meaningful in line-wrapping mode. Instructs the diagnostic
2718 messages reporter to emit the same source location information (as
2719 prefix) for physical lines that result from the process of breaking
2720 a message which is too long to fit on a single line.
2722 @item -fdiagnostics-show-option
2723 @opindex fdiagnostics-show-option
2724 This option instructs the diagnostic machinery to add text to each
2725 diagnostic emitted, which indicates which command line option directly
2726 controls that diagnostic, when such an option is known to the
2727 diagnostic machinery.
2729 @item -Wcoverage-mismatch
2730 @opindex Wcoverage-mismatch
2731 Warn if feedback profiles do not match when using the
2732 @option{-fprofile-use} option.
2733 If a source file was changed between @option{-fprofile-gen} and
2734 @option{-fprofile-use}, the files with the profile feedback can fail
2735 to match the source file and GCC can not use the profile feedback
2736 information. By default, this warning is enabled and is treated as an
2737 error. @option{-Wno-coverage-mismatch} can be used to disable the
2738 warning or @option{-Wno-error=coverage-mismatch} can be used to
2739 disable the error. Disable the error for this warning can result in
2740 poorly optimized code, so disabling the error is useful only in the
2741 case of very minor changes such as bug fixes to an existing code-base.
2742 Completely disabling the warning is not recommended.
2746 @node Warning Options
2747 @section Options to Request or Suppress Warnings
2748 @cindex options to control warnings
2749 @cindex warning messages
2750 @cindex messages, warning
2751 @cindex suppressing warnings
2753 Warnings are diagnostic messages that report constructions which
2754 are not inherently erroneous but which are risky or suggest there
2755 may have been an error.
2757 The following language-independent options do not enable specific
2758 warnings but control the kinds of diagnostics produced by GCC.
2761 @cindex syntax checking
2763 @opindex fsyntax-only
2764 Check the code for syntax errors, but don't do anything beyond that.
2768 Inhibit all warning messages.
2773 Make all warnings into errors.
2778 Make the specified warning into an error. The specifier for a warning
2779 is appended, for example @option{-Werror=switch} turns the warnings
2780 controlled by @option{-Wswitch} into errors. This switch takes a
2781 negative form, to be used to negate @option{-Werror} for specific
2782 warnings, for example @option{-Wno-error=switch} makes
2783 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2784 is in effect. You can use the @option{-fdiagnostics-show-option}
2785 option to have each controllable warning amended with the option which
2786 controls it, to determine what to use with this option.
2788 Note that specifying @option{-Werror=}@var{foo} automatically implies
2789 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2792 @item -Wfatal-errors
2793 @opindex Wfatal-errors
2794 @opindex Wno-fatal-errors
2795 This option causes the compiler to abort compilation on the first error
2796 occurred rather than trying to keep going and printing further error
2801 You can request many specific warnings with options beginning
2802 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2803 implicit declarations. Each of these specific warning options also
2804 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2805 example, @option{-Wno-implicit}. This manual lists only one of the
2806 two forms, whichever is not the default. For further,
2807 language-specific options also refer to @ref{C++ Dialect Options} and
2808 @ref{Objective-C and Objective-C++ Dialect Options}.
2810 When an unrecognized warning option is requested (e.g.,
2811 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2812 that the option is not recognized. However, if the @option{-Wno-} form
2813 is used, the behavior is slightly different: No diagnostic will be
2814 produced for @option{-Wno-unknown-warning} unless other diagnostics
2815 are being produced. This allows the use of new @option{-Wno-} options
2816 with old compilers, but if something goes wrong, the compiler will
2817 warn that an unrecognized option was used.
2822 Issue all the warnings demanded by strict ISO C and ISO C++;
2823 reject all programs that use forbidden extensions, and some other
2824 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2825 version of the ISO C standard specified by any @option{-std} option used.
2827 Valid ISO C and ISO C++ programs should compile properly with or without
2828 this option (though a rare few will require @option{-ansi} or a
2829 @option{-std} option specifying the required version of ISO C)@. However,
2830 without this option, certain GNU extensions and traditional C and C++
2831 features are supported as well. With this option, they are rejected.
2833 @option{-pedantic} does not cause warning messages for use of the
2834 alternate keywords whose names begin and end with @samp{__}. Pedantic
2835 warnings are also disabled in the expression that follows
2836 @code{__extension__}. However, only system header files should use
2837 these escape routes; application programs should avoid them.
2838 @xref{Alternate Keywords}.
2840 Some users try to use @option{-pedantic} to check programs for strict ISO
2841 C conformance. They soon find that it does not do quite what they want:
2842 it finds some non-ISO practices, but not all---only those for which
2843 ISO C @emph{requires} a diagnostic, and some others for which
2844 diagnostics have been added.
2846 A feature to report any failure to conform to ISO C might be useful in
2847 some instances, but would require considerable additional work and would
2848 be quite different from @option{-pedantic}. We don't have plans to
2849 support such a feature in the near future.
2851 Where the standard specified with @option{-std} represents a GNU
2852 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2853 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2854 extended dialect is based. Warnings from @option{-pedantic} are given
2855 where they are required by the base standard. (It would not make sense
2856 for such warnings to be given only for features not in the specified GNU
2857 C dialect, since by definition the GNU dialects of C include all
2858 features the compiler supports with the given option, and there would be
2859 nothing to warn about.)
2861 @item -pedantic-errors
2862 @opindex pedantic-errors
2863 Like @option{-pedantic}, except that errors are produced rather than
2869 This enables all the warnings about constructions that some users
2870 consider questionable, and that are easy to avoid (or modify to
2871 prevent the warning), even in conjunction with macros. This also
2872 enables some language-specific warnings described in @ref{C++ Dialect
2873 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2875 @option{-Wall} turns on the following warning flags:
2877 @gccoptlist{-Waddress @gol
2878 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2880 -Wchar-subscripts @gol
2881 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2882 -Wimplicit-int @r{(C and Objective-C only)} @gol
2883 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2886 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2887 -Wmissing-braces @gol
2893 -Wsequence-point @gol
2894 -Wsign-compare @r{(only in C++)} @gol
2895 -Wstrict-aliasing @gol
2896 -Wstrict-overflow=1 @gol
2899 -Wuninitialized @gol
2900 -Wunknown-pragmas @gol
2901 -Wunused-function @gol
2904 -Wunused-variable @gol
2905 -Wvolatile-register-var @gol
2908 Note that some warning flags are not implied by @option{-Wall}. Some of
2909 them warn about constructions that users generally do not consider
2910 questionable, but which occasionally you might wish to check for;
2911 others warn about constructions that are necessary or hard to avoid in
2912 some cases, and there is no simple way to modify the code to suppress
2913 the warning. Some of them are enabled by @option{-Wextra} but many of
2914 them must be enabled individually.
2920 This enables some extra warning flags that are not enabled by
2921 @option{-Wall}. (This option used to be called @option{-W}. The older
2922 name is still supported, but the newer name is more descriptive.)
2924 @gccoptlist{-Wclobbered @gol
2926 -Wignored-qualifiers @gol
2927 -Wmissing-field-initializers @gol
2928 -Wmissing-parameter-type @r{(C only)} @gol
2929 -Wold-style-declaration @r{(C only)} @gol
2930 -Woverride-init @gol
2933 -Wuninitialized @gol
2934 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2935 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2938 The option @option{-Wextra} also prints warning messages for the
2944 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2945 @samp{>}, or @samp{>=}.
2948 (C++ only) An enumerator and a non-enumerator both appear in a
2949 conditional expression.
2952 (C++ only) Ambiguous virtual bases.
2955 (C++ only) Subscripting an array which has been declared @samp{register}.
2958 (C++ only) Taking the address of a variable which has been declared
2962 (C++ only) A base class is not initialized in a derived class' copy
2967 @item -Wchar-subscripts
2968 @opindex Wchar-subscripts
2969 @opindex Wno-char-subscripts
2970 Warn if an array subscript has type @code{char}. This is a common cause
2971 of error, as programmers often forget that this type is signed on some
2973 This warning is enabled by @option{-Wall}.
2977 @opindex Wno-comment
2978 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2979 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2980 This warning is enabled by @option{-Wall}.
2983 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
2985 Suppress warning messages emitted by @code{#warning} directives.
2987 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
2988 @opindex Wdouble-promotion
2989 @opindex Wno-double-promotion
2990 Give a warning when a value of type @code{float} is implicitly
2991 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
2992 floating-point unit implement @code{float} in hardware, but emulate
2993 @code{double} in software. On such a machine, doing computations
2994 using @code{double} values is much more expensive because of the
2995 overhead required for software emulation.
2997 It is easy to accidentally do computations with @code{double} because
2998 floating-point literals are implicitly of type @code{double}. For
3002 float area(float radius)
3004 return 3.14159 * radius * radius;
3008 the compiler will perform the entire computation with @code{double}
3009 because the floating-point literal is a @code{double}.
3014 @opindex ffreestanding
3015 @opindex fno-builtin
3016 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3017 the arguments supplied have types appropriate to the format string
3018 specified, and that the conversions specified in the format string make
3019 sense. This includes standard functions, and others specified by format
3020 attributes (@pxref{Function Attributes}), in the @code{printf},
3021 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3022 not in the C standard) families (or other target-specific families).
3023 Which functions are checked without format attributes having been
3024 specified depends on the standard version selected, and such checks of
3025 functions without the attribute specified are disabled by
3026 @option{-ffreestanding} or @option{-fno-builtin}.
3028 The formats are checked against the format features supported by GNU
3029 libc version 2.2. These include all ISO C90 and C99 features, as well
3030 as features from the Single Unix Specification and some BSD and GNU
3031 extensions. Other library implementations may not support all these
3032 features; GCC does not support warning about features that go beyond a
3033 particular library's limitations. However, if @option{-pedantic} is used
3034 with @option{-Wformat}, warnings will be given about format features not
3035 in the selected standard version (but not for @code{strfmon} formats,
3036 since those are not in any version of the C standard). @xref{C Dialect
3037 Options,,Options Controlling C Dialect}.
3039 Since @option{-Wformat} also checks for null format arguments for
3040 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3042 @option{-Wformat} is included in @option{-Wall}. For more control over some
3043 aspects of format checking, the options @option{-Wformat-y2k},
3044 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3045 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3046 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3049 @opindex Wformat-y2k
3050 @opindex Wno-format-y2k
3051 If @option{-Wformat} is specified, also warn about @code{strftime}
3052 formats which may yield only a two-digit year.
3054 @item -Wno-format-contains-nul
3055 @opindex Wno-format-contains-nul
3056 @opindex Wformat-contains-nul
3057 If @option{-Wformat} is specified, do not warn about format strings that
3060 @item -Wno-format-extra-args
3061 @opindex Wno-format-extra-args
3062 @opindex Wformat-extra-args
3063 If @option{-Wformat} is specified, do not warn about excess arguments to a
3064 @code{printf} or @code{scanf} format function. The C standard specifies
3065 that such arguments are ignored.
3067 Where the unused arguments lie between used arguments that are
3068 specified with @samp{$} operand number specifications, normally
3069 warnings are still given, since the implementation could not know what
3070 type to pass to @code{va_arg} to skip the unused arguments. However,
3071 in the case of @code{scanf} formats, this option will suppress the
3072 warning if the unused arguments are all pointers, since the Single
3073 Unix Specification says that such unused arguments are allowed.
3075 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3076 @opindex Wno-format-zero-length
3077 @opindex Wformat-zero-length
3078 If @option{-Wformat} is specified, do not warn about zero-length formats.
3079 The C standard specifies that zero-length formats are allowed.
3081 @item -Wformat-nonliteral
3082 @opindex Wformat-nonliteral
3083 @opindex Wno-format-nonliteral
3084 If @option{-Wformat} is specified, also warn if the format string is not a
3085 string literal and so cannot be checked, unless the format function
3086 takes its format arguments as a @code{va_list}.
3088 @item -Wformat-security
3089 @opindex Wformat-security
3090 @opindex Wno-format-security
3091 If @option{-Wformat} is specified, also warn about uses of format
3092 functions that represent possible security problems. At present, this
3093 warns about calls to @code{printf} and @code{scanf} functions where the
3094 format string is not a string literal and there are no format arguments,
3095 as in @code{printf (foo);}. This may be a security hole if the format
3096 string came from untrusted input and contains @samp{%n}. (This is
3097 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3098 in future warnings may be added to @option{-Wformat-security} that are not
3099 included in @option{-Wformat-nonliteral}.)
3103 @opindex Wno-format=2
3104 Enable @option{-Wformat} plus format checks not included in
3105 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3106 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3108 @item -Wnonnull @r{(C and Objective-C only)}
3110 @opindex Wno-nonnull
3111 Warn about passing a null pointer for arguments marked as
3112 requiring a non-null value by the @code{nonnull} function attribute.
3114 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3115 can be disabled with the @option{-Wno-nonnull} option.
3117 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3119 @opindex Wno-init-self
3120 Warn about uninitialized variables which are initialized with themselves.
3121 Note this option can only be used with the @option{-Wuninitialized} option.
3123 For example, GCC will warn about @code{i} being uninitialized in the
3124 following snippet only when @option{-Winit-self} has been specified:
3135 @item -Wimplicit-int @r{(C and Objective-C only)}
3136 @opindex Wimplicit-int
3137 @opindex Wno-implicit-int
3138 Warn when a declaration does not specify a type.
3139 This warning is enabled by @option{-Wall}.
3141 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3142 @opindex Wimplicit-function-declaration
3143 @opindex Wno-implicit-function-declaration
3144 Give a warning whenever a function is used before being declared. In
3145 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3146 enabled by default and it is made into an error by
3147 @option{-pedantic-errors}. This warning is also enabled by
3150 @item -Wimplicit @r{(C and Objective-C only)}
3152 @opindex Wno-implicit
3153 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3154 This warning is enabled by @option{-Wall}.
3156 @item -Wignored-qualifiers @r{(C and C++ only)}
3157 @opindex Wignored-qualifiers
3158 @opindex Wno-ignored-qualifiers
3159 Warn if the return type of a function has a type qualifier
3160 such as @code{const}. For ISO C such a type qualifier has no effect,
3161 since the value returned by a function is not an lvalue.
3162 For C++, the warning is only emitted for scalar types or @code{void}.
3163 ISO C prohibits qualified @code{void} return types on function
3164 definitions, so such return types always receive a warning
3165 even without this option.
3167 This warning is also enabled by @option{-Wextra}.
3172 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3173 a function with external linkage, returning int, taking either zero
3174 arguments, two, or three arguments of appropriate types. This warning
3175 is enabled by default in C++ and is enabled by either @option{-Wall}
3176 or @option{-pedantic}.
3178 @item -Wmissing-braces
3179 @opindex Wmissing-braces
3180 @opindex Wno-missing-braces
3181 Warn if an aggregate or union initializer is not fully bracketed. In
3182 the following example, the initializer for @samp{a} is not fully
3183 bracketed, but that for @samp{b} is fully bracketed.
3186 int a[2][2] = @{ 0, 1, 2, 3 @};
3187 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3190 This warning is enabled by @option{-Wall}.
3192 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3193 @opindex Wmissing-include-dirs
3194 @opindex Wno-missing-include-dirs
3195 Warn if a user-supplied include directory does not exist.
3198 @opindex Wparentheses
3199 @opindex Wno-parentheses
3200 Warn if parentheses are omitted in certain contexts, such
3201 as when there is an assignment in a context where a truth value
3202 is expected, or when operators are nested whose precedence people
3203 often get confused about.
3205 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3206 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3207 interpretation from that of ordinary mathematical notation.
3209 Also warn about constructions where there may be confusion to which
3210 @code{if} statement an @code{else} branch belongs. Here is an example of
3225 In C/C++, every @code{else} branch belongs to the innermost possible
3226 @code{if} statement, which in this example is @code{if (b)}. This is
3227 often not what the programmer expected, as illustrated in the above
3228 example by indentation the programmer chose. When there is the
3229 potential for this confusion, GCC will issue a warning when this flag
3230 is specified. To eliminate the warning, add explicit braces around
3231 the innermost @code{if} statement so there is no way the @code{else}
3232 could belong to the enclosing @code{if}. The resulting code would
3249 Also warn for dangerous uses of the
3250 ?: with omitted middle operand GNU extension. When the condition
3251 in the ?: operator is a boolean expression the omitted value will
3252 be always 1. Often the user expects it to be a value computed
3253 inside the conditional expression instead.
3255 This warning is enabled by @option{-Wall}.
3257 @item -Wsequence-point
3258 @opindex Wsequence-point
3259 @opindex Wno-sequence-point
3260 Warn about code that may have undefined semantics because of violations
3261 of sequence point rules in the C and C++ standards.
3263 The C and C++ standards defines the order in which expressions in a C/C++
3264 program are evaluated in terms of @dfn{sequence points}, which represent
3265 a partial ordering between the execution of parts of the program: those
3266 executed before the sequence point, and those executed after it. These
3267 occur after the evaluation of a full expression (one which is not part
3268 of a larger expression), after the evaluation of the first operand of a
3269 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3270 function is called (but after the evaluation of its arguments and the
3271 expression denoting the called function), and in certain other places.
3272 Other than as expressed by the sequence point rules, the order of
3273 evaluation of subexpressions of an expression is not specified. All
3274 these rules describe only a partial order rather than a total order,
3275 since, for example, if two functions are called within one expression
3276 with no sequence point between them, the order in which the functions
3277 are called is not specified. However, the standards committee have
3278 ruled that function calls do not overlap.
3280 It is not specified when between sequence points modifications to the
3281 values of objects take effect. Programs whose behavior depends on this
3282 have undefined behavior; the C and C++ standards specify that ``Between
3283 the previous and next sequence point an object shall have its stored
3284 value modified at most once by the evaluation of an expression.
3285 Furthermore, the prior value shall be read only to determine the value
3286 to be stored.''. If a program breaks these rules, the results on any
3287 particular implementation are entirely unpredictable.
3289 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3290 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3291 diagnosed by this option, and it may give an occasional false positive
3292 result, but in general it has been found fairly effective at detecting
3293 this sort of problem in programs.
3295 The standard is worded confusingly, therefore there is some debate
3296 over the precise meaning of the sequence point rules in subtle cases.
3297 Links to discussions of the problem, including proposed formal
3298 definitions, may be found on the GCC readings page, at
3299 @uref{http://gcc.gnu.org/@/readings.html}.
3301 This warning is enabled by @option{-Wall} for C and C++.
3304 @opindex Wreturn-type
3305 @opindex Wno-return-type
3306 Warn whenever a function is defined with a return-type that defaults
3307 to @code{int}. Also warn about any @code{return} statement with no
3308 return-value in a function whose return-type is not @code{void}
3309 (falling off the end of the function body is considered returning
3310 without a value), and about a @code{return} statement with an
3311 expression in a function whose return-type is @code{void}.
3313 For C++, a function without return type always produces a diagnostic
3314 message, even when @option{-Wno-return-type} is specified. The only
3315 exceptions are @samp{main} and functions defined in system headers.
3317 This warning is enabled by @option{-Wall}.
3322 Warn whenever a @code{switch} statement has an index of enumerated type
3323 and lacks a @code{case} for one or more of the named codes of that
3324 enumeration. (The presence of a @code{default} label prevents this
3325 warning.) @code{case} labels outside the enumeration range also
3326 provoke warnings when this option is used (even if there is a
3327 @code{default} label).
3328 This warning is enabled by @option{-Wall}.
3330 @item -Wswitch-default
3331 @opindex Wswitch-default
3332 @opindex Wno-switch-default
3333 Warn whenever a @code{switch} statement does not have a @code{default}
3337 @opindex Wswitch-enum
3338 @opindex Wno-switch-enum
3339 Warn whenever a @code{switch} statement has an index of enumerated type
3340 and lacks a @code{case} for one or more of the named codes of that
3341 enumeration. @code{case} labels outside the enumeration range also
3342 provoke warnings when this option is used. The only difference
3343 between @option{-Wswitch} and this option is that this option gives a
3344 warning about an omitted enumeration code even if there is a
3345 @code{default} label.
3347 @item -Wsync-nand @r{(C and C++ only)}
3349 @opindex Wno-sync-nand
3350 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3351 built-in functions are used. These functions changed semantics in GCC 4.4.
3355 @opindex Wno-trigraphs
3356 Warn if any trigraphs are encountered that might change the meaning of
3357 the program (trigraphs within comments are not warned about).
3358 This warning is enabled by @option{-Wall}.
3360 @item -Wunused-but-set-parameter
3361 @opindex Wunused-but-set-parameter
3362 @opindex Wno-unused-but-set-parameter
3363 Warn whenever a function parameter is assigned to, but otherwise unused
3364 (aside from its declaration).
3366 To suppress this warning use the @samp{unused} attribute
3367 (@pxref{Variable Attributes}).
3369 This warning is also enabled by @option{-Wunused} together with
3372 @item -Wunused-but-set-variable
3373 @opindex Wunused-but-set-variable
3374 @opindex Wno-unused-but-set-variable
3375 Warn whenever a local variable is assigned to, but otherwise unused
3376 (aside from its declaration).
3377 This warning is enabled by @option{-Wall}.
3379 To suppress this warning use the @samp{unused} attribute
3380 (@pxref{Variable Attributes}).
3382 This warning is also enabled by @option{-Wunused}, which is enabled
3385 @item -Wunused-function
3386 @opindex Wunused-function
3387 @opindex Wno-unused-function
3388 Warn whenever a static function is declared but not defined or a
3389 non-inline static function is unused.
3390 This warning is enabled by @option{-Wall}.
3392 @item -Wunused-label
3393 @opindex Wunused-label
3394 @opindex Wno-unused-label
3395 Warn whenever a label is declared but not used.
3396 This warning is enabled by @option{-Wall}.
3398 To suppress this warning use the @samp{unused} attribute
3399 (@pxref{Variable Attributes}).
3401 @item -Wunused-parameter
3402 @opindex Wunused-parameter
3403 @opindex Wno-unused-parameter
3404 Warn whenever a function parameter is unused aside from its declaration.
3406 To suppress this warning use the @samp{unused} attribute
3407 (@pxref{Variable Attributes}).
3409 @item -Wno-unused-result
3410 @opindex Wunused-result
3411 @opindex Wno-unused-result
3412 Do not warn if a caller of a function marked with attribute
3413 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3414 its return value. The default is @option{-Wunused-result}.
3416 @item -Wunused-variable
3417 @opindex Wunused-variable
3418 @opindex Wno-unused-variable
3419 Warn whenever a local variable or non-constant static variable is unused
3420 aside from its declaration.
3421 This warning is enabled by @option{-Wall}.
3423 To suppress this warning use the @samp{unused} attribute
3424 (@pxref{Variable Attributes}).
3426 @item -Wunused-value
3427 @opindex Wunused-value
3428 @opindex Wno-unused-value
3429 Warn whenever a statement computes a result that is explicitly not
3430 used. To suppress this warning cast the unused expression to
3431 @samp{void}. This includes an expression-statement or the left-hand
3432 side of a comma expression that contains no side effects. For example,
3433 an expression such as @samp{x[i,j]} will cause a warning, while
3434 @samp{x[(void)i,j]} will not.
3436 This warning is enabled by @option{-Wall}.
3441 All the above @option{-Wunused} options combined.
3443 In order to get a warning about an unused function parameter, you must
3444 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3445 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3447 @item -Wuninitialized
3448 @opindex Wuninitialized
3449 @opindex Wno-uninitialized
3450 Warn if an automatic variable is used without first being initialized
3451 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3452 warn if a non-static reference or non-static @samp{const} member
3453 appears in a class without constructors.
3455 If you want to warn about code which uses the uninitialized value of the
3456 variable in its own initializer, use the @option{-Winit-self} option.
3458 These warnings occur for individual uninitialized or clobbered
3459 elements of structure, union or array variables as well as for
3460 variables which are uninitialized or clobbered as a whole. They do
3461 not occur for variables or elements declared @code{volatile}. Because
3462 these warnings depend on optimization, the exact variables or elements
3463 for which there are warnings will depend on the precise optimization
3464 options and version of GCC used.
3466 Note that there may be no warning about a variable that is used only
3467 to compute a value that itself is never used, because such
3468 computations may be deleted by data flow analysis before the warnings
3471 These warnings are made optional because GCC is not smart
3472 enough to see all the reasons why the code might be correct
3473 despite appearing to have an error. Here is one example of how
3494 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3495 always initialized, but GCC doesn't know this. Here is
3496 another common case:
3501 if (change_y) save_y = y, y = new_y;
3503 if (change_y) y = save_y;
3508 This has no bug because @code{save_y} is used only if it is set.
3510 @cindex @code{longjmp} warnings
3511 This option also warns when a non-volatile automatic variable might be
3512 changed by a call to @code{longjmp}. These warnings as well are possible
3513 only in optimizing compilation.
3515 The compiler sees only the calls to @code{setjmp}. It cannot know
3516 where @code{longjmp} will be called; in fact, a signal handler could
3517 call it at any point in the code. As a result, you may get a warning
3518 even when there is in fact no problem because @code{longjmp} cannot
3519 in fact be called at the place which would cause a problem.
3521 Some spurious warnings can be avoided if you declare all the functions
3522 you use that never return as @code{noreturn}. @xref{Function
3525 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3527 @item -Wunknown-pragmas
3528 @opindex Wunknown-pragmas
3529 @opindex Wno-unknown-pragmas
3530 @cindex warning for unknown pragmas
3531 @cindex unknown pragmas, warning
3532 @cindex pragmas, warning of unknown
3533 Warn when a #pragma directive is encountered which is not understood by
3534 GCC@. If this command line option is used, warnings will even be issued
3535 for unknown pragmas in system header files. This is not the case if
3536 the warnings were only enabled by the @option{-Wall} command line option.
3539 @opindex Wno-pragmas
3541 Do not warn about misuses of pragmas, such as incorrect parameters,
3542 invalid syntax, or conflicts between pragmas. See also
3543 @samp{-Wunknown-pragmas}.
3545 @item -Wstrict-aliasing
3546 @opindex Wstrict-aliasing
3547 @opindex Wno-strict-aliasing
3548 This option is only active when @option{-fstrict-aliasing} is active.
3549 It warns about code which might break the strict aliasing rules that the
3550 compiler is using for optimization. The warning does not catch all
3551 cases, but does attempt to catch the more common pitfalls. It is
3552 included in @option{-Wall}.
3553 It is equivalent to @option{-Wstrict-aliasing=3}
3555 @item -Wstrict-aliasing=n
3556 @opindex Wstrict-aliasing=n
3557 @opindex Wno-strict-aliasing=n
3558 This option is only active when @option{-fstrict-aliasing} is active.
3559 It warns about code which might break the strict aliasing rules that the
3560 compiler is using for optimization.
3561 Higher levels correspond to higher accuracy (fewer false positives).
3562 Higher levels also correspond to more effort, similar to the way -O works.
3563 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3566 Level 1: Most aggressive, quick, least accurate.
3567 Possibly useful when higher levels
3568 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3569 false negatives. However, it has many false positives.
3570 Warns for all pointer conversions between possibly incompatible types,
3571 even if never dereferenced. Runs in the frontend only.
3573 Level 2: Aggressive, quick, not too precise.
3574 May still have many false positives (not as many as level 1 though),
3575 and few false negatives (but possibly more than level 1).
3576 Unlike level 1, it only warns when an address is taken. Warns about
3577 incomplete types. Runs in the frontend only.
3579 Level 3 (default for @option{-Wstrict-aliasing}):
3580 Should have very few false positives and few false
3581 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3582 Takes care of the common pun+dereference pattern in the frontend:
3583 @code{*(int*)&some_float}.
3584 If optimization is enabled, it also runs in the backend, where it deals
3585 with multiple statement cases using flow-sensitive points-to information.
3586 Only warns when the converted pointer is dereferenced.
3587 Does not warn about incomplete types.
3589 @item -Wstrict-overflow
3590 @itemx -Wstrict-overflow=@var{n}
3591 @opindex Wstrict-overflow
3592 @opindex Wno-strict-overflow
3593 This option is only active when @option{-fstrict-overflow} is active.
3594 It warns about cases where the compiler optimizes based on the
3595 assumption that signed overflow does not occur. Note that it does not
3596 warn about all cases where the code might overflow: it only warns
3597 about cases where the compiler implements some optimization. Thus
3598 this warning depends on the optimization level.
3600 An optimization which assumes that signed overflow does not occur is
3601 perfectly safe if the values of the variables involved are such that
3602 overflow never does, in fact, occur. Therefore this warning can
3603 easily give a false positive: a warning about code which is not
3604 actually a problem. To help focus on important issues, several
3605 warning levels are defined. No warnings are issued for the use of
3606 undefined signed overflow when estimating how many iterations a loop
3607 will require, in particular when determining whether a loop will be
3611 @item -Wstrict-overflow=1
3612 Warn about cases which are both questionable and easy to avoid. For
3613 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3614 compiler will simplify this to @code{1}. This level of
3615 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3616 are not, and must be explicitly requested.
3618 @item -Wstrict-overflow=2
3619 Also warn about other cases where a comparison is simplified to a
3620 constant. For example: @code{abs (x) >= 0}. This can only be
3621 simplified when @option{-fstrict-overflow} is in effect, because
3622 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3623 zero. @option{-Wstrict-overflow} (with no level) is the same as
3624 @option{-Wstrict-overflow=2}.
3626 @item -Wstrict-overflow=3
3627 Also warn about other cases where a comparison is simplified. For
3628 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3630 @item -Wstrict-overflow=4
3631 Also warn about other simplifications not covered by the above cases.
3632 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3634 @item -Wstrict-overflow=5
3635 Also warn about cases where the compiler reduces the magnitude of a
3636 constant involved in a comparison. For example: @code{x + 2 > y} will
3637 be simplified to @code{x + 1 >= y}. This is reported only at the
3638 highest warning level because this simplification applies to many
3639 comparisons, so this warning level will give a very large number of
3643 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3644 @opindex Wsuggest-attribute=
3645 @opindex Wno-suggest-attribute=
3646 Warn for cases where adding an attribute may be beneficial. The
3647 attributes currently supported are listed below.
3650 @item -Wsuggest-attribute=pure
3651 @itemx -Wsuggest-attribute=const
3652 @itemx -Wsuggest-attribute=noreturn
3653 @opindex Wsuggest-attribute=pure
3654 @opindex Wno-suggest-attribute=pure
3655 @opindex Wsuggest-attribute=const
3656 @opindex Wno-suggest-attribute=const
3657 @opindex Wsuggest-attribute=noreturn
3658 @opindex Wno-suggest-attribute=noreturn
3660 Warn about functions which might be candidates for attributes
3661 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3662 functions visible in other compilation units or (in the case of @code{pure} and
3663 @code{const}) if it cannot prove that the function returns normally. A function
3664 returns normally if it doesn't contain an infinite loop nor returns abnormally
3665 by throwing, calling @code{abort()} or trapping. This analysis requires option
3666 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3667 higher. Higher optimization levels improve the accuracy of the analysis.
3670 @item -Warray-bounds
3671 @opindex Wno-array-bounds
3672 @opindex Warray-bounds
3673 This option is only active when @option{-ftree-vrp} is active
3674 (default for @option{-O2} and above). It warns about subscripts to arrays
3675 that are always out of bounds. This warning is enabled by @option{-Wall}.
3677 @item -Wno-div-by-zero
3678 @opindex Wno-div-by-zero
3679 @opindex Wdiv-by-zero
3680 Do not warn about compile-time integer division by zero. Floating point
3681 division by zero is not warned about, as it can be a legitimate way of
3682 obtaining infinities and NaNs.
3684 @item -Wsystem-headers
3685 @opindex Wsystem-headers
3686 @opindex Wno-system-headers
3687 @cindex warnings from system headers
3688 @cindex system headers, warnings from
3689 Print warning messages for constructs found in system header files.
3690 Warnings from system headers are normally suppressed, on the assumption
3691 that they usually do not indicate real problems and would only make the
3692 compiler output harder to read. Using this command line option tells
3693 GCC to emit warnings from system headers as if they occurred in user
3694 code. However, note that using @option{-Wall} in conjunction with this
3695 option will @emph{not} warn about unknown pragmas in system
3696 headers---for that, @option{-Wunknown-pragmas} must also be used.
3699 @opindex Wtrampolines
3700 @opindex Wno-trampolines
3701 Warn about trampolines generated for pointers to nested functions.
3703 A trampoline is a small piece of data or code that is created at run
3704 time on the stack when the address of a nested function is taken, and
3705 is used to call the nested function indirectly. For some targets, it
3706 is made up of data only and thus requires no special treatment. But,
3707 for most targets, it is made up of code and thus requires the stack
3708 to be made executable in order for the program to work properly.
3711 @opindex Wfloat-equal
3712 @opindex Wno-float-equal
3713 Warn if floating point values are used in equality comparisons.
3715 The idea behind this is that sometimes it is convenient (for the
3716 programmer) to consider floating-point values as approximations to
3717 infinitely precise real numbers. If you are doing this, then you need
3718 to compute (by analyzing the code, or in some other way) the maximum or
3719 likely maximum error that the computation introduces, and allow for it
3720 when performing comparisons (and when producing output, but that's a
3721 different problem). In particular, instead of testing for equality, you
3722 would check to see whether the two values have ranges that overlap; and
3723 this is done with the relational operators, so equality comparisons are
3726 @item -Wtraditional @r{(C and Objective-C only)}
3727 @opindex Wtraditional
3728 @opindex Wno-traditional
3729 Warn about certain constructs that behave differently in traditional and
3730 ISO C@. Also warn about ISO C constructs that have no traditional C
3731 equivalent, and/or problematic constructs which should be avoided.
3735 Macro parameters that appear within string literals in the macro body.
3736 In traditional C macro replacement takes place within string literals,
3737 but does not in ISO C@.
3740 In traditional C, some preprocessor directives did not exist.
3741 Traditional preprocessors would only consider a line to be a directive
3742 if the @samp{#} appeared in column 1 on the line. Therefore
3743 @option{-Wtraditional} warns about directives that traditional C
3744 understands but would ignore because the @samp{#} does not appear as the
3745 first character on the line. It also suggests you hide directives like
3746 @samp{#pragma} not understood by traditional C by indenting them. Some
3747 traditional implementations would not recognize @samp{#elif}, so it
3748 suggests avoiding it altogether.
3751 A function-like macro that appears without arguments.
3754 The unary plus operator.
3757 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3758 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3759 constants.) Note, these suffixes appear in macros defined in the system
3760 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3761 Use of these macros in user code might normally lead to spurious
3762 warnings, however GCC's integrated preprocessor has enough context to
3763 avoid warning in these cases.
3766 A function declared external in one block and then used after the end of
3770 A @code{switch} statement has an operand of type @code{long}.
3773 A non-@code{static} function declaration follows a @code{static} one.
3774 This construct is not accepted by some traditional C compilers.
3777 The ISO type of an integer constant has a different width or
3778 signedness from its traditional type. This warning is only issued if
3779 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3780 typically represent bit patterns, are not warned about.
3783 Usage of ISO string concatenation is detected.
3786 Initialization of automatic aggregates.
3789 Identifier conflicts with labels. Traditional C lacks a separate
3790 namespace for labels.
3793 Initialization of unions. If the initializer is zero, the warning is
3794 omitted. This is done under the assumption that the zero initializer in
3795 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3796 initializer warnings and relies on default initialization to zero in the
3800 Conversions by prototypes between fixed/floating point values and vice
3801 versa. The absence of these prototypes when compiling with traditional
3802 C would cause serious problems. This is a subset of the possible
3803 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3806 Use of ISO C style function definitions. This warning intentionally is
3807 @emph{not} issued for prototype declarations or variadic functions
3808 because these ISO C features will appear in your code when using
3809 libiberty's traditional C compatibility macros, @code{PARAMS} and
3810 @code{VPARAMS}. This warning is also bypassed for nested functions
3811 because that feature is already a GCC extension and thus not relevant to
3812 traditional C compatibility.
3815 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3816 @opindex Wtraditional-conversion
3817 @opindex Wno-traditional-conversion
3818 Warn if a prototype causes a type conversion that is different from what
3819 would happen to the same argument in the absence of a prototype. This
3820 includes conversions of fixed point to floating and vice versa, and
3821 conversions changing the width or signedness of a fixed point argument
3822 except when the same as the default promotion.
3824 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3825 @opindex Wdeclaration-after-statement
3826 @opindex Wno-declaration-after-statement
3827 Warn when a declaration is found after a statement in a block. This
3828 construct, known from C++, was introduced with ISO C99 and is by default
3829 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3830 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3835 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3837 @item -Wno-endif-labels
3838 @opindex Wno-endif-labels
3839 @opindex Wendif-labels
3840 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3845 Warn whenever a local variable or type declaration shadows another variable,
3846 parameter, type, or class member (in C++), or whenever a built-in function
3847 is shadowed. Note that in C++, the compiler will not warn if a local variable
3848 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3850 @item -Wlarger-than=@var{len}
3851 @opindex Wlarger-than=@var{len}
3852 @opindex Wlarger-than-@var{len}
3853 Warn whenever an object of larger than @var{len} bytes is defined.
3855 @item -Wframe-larger-than=@var{len}
3856 @opindex Wframe-larger-than
3857 Warn if the size of a function frame is larger than @var{len} bytes.
3858 The computation done to determine the stack frame size is approximate
3859 and not conservative.
3860 The actual requirements may be somewhat greater than @var{len}
3861 even if you do not get a warning. In addition, any space allocated
3862 via @code{alloca}, variable-length arrays, or related constructs
3863 is not included by the compiler when determining
3864 whether or not to issue a warning.
3866 @item -Wunsafe-loop-optimizations
3867 @opindex Wunsafe-loop-optimizations
3868 @opindex Wno-unsafe-loop-optimizations
3869 Warn if the loop cannot be optimized because the compiler could not
3870 assume anything on the bounds of the loop indices. With
3871 @option{-funsafe-loop-optimizations} warn if the compiler made
3874 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3875 @opindex Wno-pedantic-ms-format
3876 @opindex Wpedantic-ms-format
3877 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3878 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3879 depending on the MS runtime, when you are using the options @option{-Wformat}
3880 and @option{-pedantic} without gnu-extensions.
3882 @item -Wpointer-arith
3883 @opindex Wpointer-arith
3884 @opindex Wno-pointer-arith
3885 Warn about anything that depends on the ``size of'' a function type or
3886 of @code{void}. GNU C assigns these types a size of 1, for
3887 convenience in calculations with @code{void *} pointers and pointers
3888 to functions. In C++, warn also when an arithmetic operation involves
3889 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3892 @opindex Wtype-limits
3893 @opindex Wno-type-limits
3894 Warn if a comparison is always true or always false due to the limited
3895 range of the data type, but do not warn for constant expressions. For
3896 example, warn if an unsigned variable is compared against zero with
3897 @samp{<} or @samp{>=}. This warning is also enabled by
3900 @item -Wbad-function-cast @r{(C and Objective-C only)}
3901 @opindex Wbad-function-cast
3902 @opindex Wno-bad-function-cast
3903 Warn whenever a function call is cast to a non-matching type.
3904 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3906 @item -Wc++-compat @r{(C and Objective-C only)}
3907 Warn about ISO C constructs that are outside of the common subset of
3908 ISO C and ISO C++, e.g.@: request for implicit conversion from
3909 @code{void *} to a pointer to non-@code{void} type.
3911 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3912 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3913 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3914 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3918 @opindex Wno-cast-qual
3919 Warn whenever a pointer is cast so as to remove a type qualifier from
3920 the target type. For example, warn if a @code{const char *} is cast
3921 to an ordinary @code{char *}.
3923 Also warn when making a cast which introduces a type qualifier in an
3924 unsafe way. For example, casting @code{char **} to @code{const char **}
3925 is unsafe, as in this example:
3928 /* p is char ** value. */
3929 const char **q = (const char **) p;
3930 /* Assignment of readonly string to const char * is OK. */
3932 /* Now char** pointer points to read-only memory. */
3937 @opindex Wcast-align
3938 @opindex Wno-cast-align
3939 Warn whenever a pointer is cast such that the required alignment of the
3940 target is increased. For example, warn if a @code{char *} is cast to
3941 an @code{int *} on machines where integers can only be accessed at
3942 two- or four-byte boundaries.
3944 @item -Wwrite-strings
3945 @opindex Wwrite-strings
3946 @opindex Wno-write-strings
3947 When compiling C, give string constants the type @code{const
3948 char[@var{length}]} so that copying the address of one into a
3949 non-@code{const} @code{char *} pointer will get a warning. These
3950 warnings will help you find at compile time code that can try to write
3951 into a string constant, but only if you have been very careful about
3952 using @code{const} in declarations and prototypes. Otherwise, it will
3953 just be a nuisance. This is why we did not make @option{-Wall} request
3956 When compiling C++, warn about the deprecated conversion from string
3957 literals to @code{char *}. This warning is enabled by default for C++
3962 @opindex Wno-clobbered
3963 Warn for variables that might be changed by @samp{longjmp} or
3964 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3967 @opindex Wconversion
3968 @opindex Wno-conversion
3969 Warn for implicit conversions that may alter a value. This includes
3970 conversions between real and integer, like @code{abs (x)} when
3971 @code{x} is @code{double}; conversions between signed and unsigned,
3972 like @code{unsigned ui = -1}; and conversions to smaller types, like
3973 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3974 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3975 changed by the conversion like in @code{abs (2.0)}. Warnings about
3976 conversions between signed and unsigned integers can be disabled by
3977 using @option{-Wno-sign-conversion}.
3979 For C++, also warn for confusing overload resolution for user-defined
3980 conversions; and conversions that will never use a type conversion
3981 operator: conversions to @code{void}, the same type, a base class or a
3982 reference to them. Warnings about conversions between signed and
3983 unsigned integers are disabled by default in C++ unless
3984 @option{-Wsign-conversion} is explicitly enabled.
3986 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
3987 @opindex Wconversion-null
3988 @opindex Wno-conversion-null
3989 Do not warn for conversions between @code{NULL} and non-pointer
3990 types. @option{-Wconversion-null} is enabled by default.
3993 @opindex Wempty-body
3994 @opindex Wno-empty-body
3995 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3996 while} statement. This warning is also enabled by @option{-Wextra}.
3998 @item -Wenum-compare
3999 @opindex Wenum-compare
4000 @opindex Wno-enum-compare
4001 Warn about a comparison between values of different enum types. In C++
4002 this warning is enabled by default. In C this warning is enabled by
4005 @item -Wjump-misses-init @r{(C, Objective-C only)}
4006 @opindex Wjump-misses-init
4007 @opindex Wno-jump-misses-init
4008 Warn if a @code{goto} statement or a @code{switch} statement jumps
4009 forward across the initialization of a variable, or jumps backward to a
4010 label after the variable has been initialized. This only warns about
4011 variables which are initialized when they are declared. This warning is
4012 only supported for C and Objective C; in C++ this sort of branch is an
4015 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4016 can be disabled with the @option{-Wno-jump-misses-init} option.
4018 @item -Wsign-compare
4019 @opindex Wsign-compare
4020 @opindex Wno-sign-compare
4021 @cindex warning for comparison of signed and unsigned values
4022 @cindex comparison of signed and unsigned values, warning
4023 @cindex signed and unsigned values, comparison warning
4024 Warn when a comparison between signed and unsigned values could produce
4025 an incorrect result when the signed value is converted to unsigned.
4026 This warning is also enabled by @option{-Wextra}; to get the other warnings
4027 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4029 @item -Wsign-conversion
4030 @opindex Wsign-conversion
4031 @opindex Wno-sign-conversion
4032 Warn for implicit conversions that may change the sign of an integer
4033 value, like assigning a signed integer expression to an unsigned
4034 integer variable. An explicit cast silences the warning. In C, this
4035 option is enabled also by @option{-Wconversion}.
4039 @opindex Wno-address
4040 Warn about suspicious uses of memory addresses. These include using
4041 the address of a function in a conditional expression, such as
4042 @code{void func(void); if (func)}, and comparisons against the memory
4043 address of a string literal, such as @code{if (x == "abc")}. Such
4044 uses typically indicate a programmer error: the address of a function
4045 always evaluates to true, so their use in a conditional usually
4046 indicate that the programmer forgot the parentheses in a function
4047 call; and comparisons against string literals result in unspecified
4048 behavior and are not portable in C, so they usually indicate that the
4049 programmer intended to use @code{strcmp}. This warning is enabled by
4053 @opindex Wlogical-op
4054 @opindex Wno-logical-op
4055 Warn about suspicious uses of logical operators in expressions.
4056 This includes using logical operators in contexts where a
4057 bit-wise operator is likely to be expected.
4059 @item -Waggregate-return
4060 @opindex Waggregate-return
4061 @opindex Wno-aggregate-return
4062 Warn if any functions that return structures or unions are defined or
4063 called. (In languages where you can return an array, this also elicits
4066 @item -Wno-attributes
4067 @opindex Wno-attributes
4068 @opindex Wattributes
4069 Do not warn if an unexpected @code{__attribute__} is used, such as
4070 unrecognized attributes, function attributes applied to variables,
4071 etc. This will not stop errors for incorrect use of supported
4074 @item -Wno-builtin-macro-redefined
4075 @opindex Wno-builtin-macro-redefined
4076 @opindex Wbuiltin-macro-redefined
4077 Do not warn if certain built-in macros are redefined. This suppresses
4078 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4079 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4081 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4082 @opindex Wstrict-prototypes
4083 @opindex Wno-strict-prototypes
4084 Warn if a function is declared or defined without specifying the
4085 argument types. (An old-style function definition is permitted without
4086 a warning if preceded by a declaration which specifies the argument
4089 @item -Wold-style-declaration @r{(C and Objective-C only)}
4090 @opindex Wold-style-declaration
4091 @opindex Wno-old-style-declaration
4092 Warn for obsolescent usages, according to the C Standard, in a
4093 declaration. For example, warn if storage-class specifiers like
4094 @code{static} are not the first things in a declaration. This warning
4095 is also enabled by @option{-Wextra}.
4097 @item -Wold-style-definition @r{(C and Objective-C only)}
4098 @opindex Wold-style-definition
4099 @opindex Wno-old-style-definition
4100 Warn if an old-style function definition is used. A warning is given
4101 even if there is a previous prototype.
4103 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4104 @opindex Wmissing-parameter-type
4105 @opindex Wno-missing-parameter-type
4106 A function parameter is declared without a type specifier in K&R-style
4113 This warning is also enabled by @option{-Wextra}.
4115 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4116 @opindex Wmissing-prototypes
4117 @opindex Wno-missing-prototypes
4118 Warn if a global function is defined without a previous prototype
4119 declaration. This warning is issued even if the definition itself
4120 provides a prototype. The aim is to detect global functions that fail
4121 to be declared in header files.
4123 @item -Wmissing-declarations
4124 @opindex Wmissing-declarations
4125 @opindex Wno-missing-declarations
4126 Warn if a global function is defined without a previous declaration.
4127 Do so even if the definition itself provides a prototype.
4128 Use this option to detect global functions that are not declared in
4129 header files. In C++, no warnings are issued for function templates,
4130 or for inline functions, or for functions in anonymous namespaces.
4132 @item -Wmissing-field-initializers
4133 @opindex Wmissing-field-initializers
4134 @opindex Wno-missing-field-initializers
4138 Warn if a structure's initializer has some fields missing. For
4139 example, the following code would cause such a warning, because
4140 @code{x.h} is implicitly zero:
4143 struct s @{ int f, g, h; @};
4144 struct s x = @{ 3, 4 @};
4147 This option does not warn about designated initializers, so the following
4148 modification would not trigger a warning:
4151 struct s @{ int f, g, h; @};
4152 struct s x = @{ .f = 3, .g = 4 @};
4155 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4156 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4158 @item -Wmissing-format-attribute
4159 @opindex Wmissing-format-attribute
4160 @opindex Wno-missing-format-attribute
4163 Warn about function pointers which might be candidates for @code{format}
4164 attributes. Note these are only possible candidates, not absolute ones.
4165 GCC will guess that function pointers with @code{format} attributes that
4166 are used in assignment, initialization, parameter passing or return
4167 statements should have a corresponding @code{format} attribute in the
4168 resulting type. I.e.@: the left-hand side of the assignment or
4169 initialization, the type of the parameter variable, or the return type
4170 of the containing function respectively should also have a @code{format}
4171 attribute to avoid the warning.
4173 GCC will also warn about function definitions which might be
4174 candidates for @code{format} attributes. Again, these are only
4175 possible candidates. GCC will guess that @code{format} attributes
4176 might be appropriate for any function that calls a function like
4177 @code{vprintf} or @code{vscanf}, but this might not always be the
4178 case, and some functions for which @code{format} attributes are
4179 appropriate may not be detected.
4181 @item -Wno-multichar
4182 @opindex Wno-multichar
4184 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4185 Usually they indicate a typo in the user's code, as they have
4186 implementation-defined values, and should not be used in portable code.
4188 @item -Wnormalized=<none|id|nfc|nfkc>
4189 @opindex Wnormalized=
4192 @cindex character set, input normalization
4193 In ISO C and ISO C++, two identifiers are different if they are
4194 different sequences of characters. However, sometimes when characters
4195 outside the basic ASCII character set are used, you can have two
4196 different character sequences that look the same. To avoid confusion,
4197 the ISO 10646 standard sets out some @dfn{normalization rules} which
4198 when applied ensure that two sequences that look the same are turned into
4199 the same sequence. GCC can warn you if you are using identifiers which
4200 have not been normalized; this option controls that warning.
4202 There are four levels of warning that GCC supports. The default is
4203 @option{-Wnormalized=nfc}, which warns about any identifier which is
4204 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4205 recommended form for most uses.
4207 Unfortunately, there are some characters which ISO C and ISO C++ allow
4208 in identifiers that when turned into NFC aren't allowable as
4209 identifiers. That is, there's no way to use these symbols in portable
4210 ISO C or C++ and have all your identifiers in NFC@.
4211 @option{-Wnormalized=id} suppresses the warning for these characters.
4212 It is hoped that future versions of the standards involved will correct
4213 this, which is why this option is not the default.
4215 You can switch the warning off for all characters by writing
4216 @option{-Wnormalized=none}. You would only want to do this if you
4217 were using some other normalization scheme (like ``D''), because
4218 otherwise you can easily create bugs that are literally impossible to see.
4220 Some characters in ISO 10646 have distinct meanings but look identical
4221 in some fonts or display methodologies, especially once formatting has
4222 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4223 LETTER N'', will display just like a regular @code{n} which has been
4224 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4225 normalization scheme to convert all these into a standard form as
4226 well, and GCC will warn if your code is not in NFKC if you use
4227 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4228 about every identifier that contains the letter O because it might be
4229 confused with the digit 0, and so is not the default, but may be
4230 useful as a local coding convention if the programming environment is
4231 unable to be fixed to display these characters distinctly.
4233 @item -Wno-deprecated
4234 @opindex Wno-deprecated
4235 @opindex Wdeprecated
4236 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4238 @item -Wno-deprecated-declarations
4239 @opindex Wno-deprecated-declarations
4240 @opindex Wdeprecated-declarations
4241 Do not warn about uses of functions (@pxref{Function Attributes}),
4242 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4243 Attributes}) marked as deprecated by using the @code{deprecated}
4247 @opindex Wno-overflow
4249 Do not warn about compile-time overflow in constant expressions.
4251 @item -Woverride-init @r{(C and Objective-C only)}
4252 @opindex Woverride-init
4253 @opindex Wno-override-init
4257 Warn if an initialized field without side effects is overridden when
4258 using designated initializers (@pxref{Designated Inits, , Designated
4261 This warning is included in @option{-Wextra}. To get other
4262 @option{-Wextra} warnings without this one, use @samp{-Wextra
4263 -Wno-override-init}.
4268 Warn if a structure is given the packed attribute, but the packed
4269 attribute has no effect on the layout or size of the structure.
4270 Such structures may be mis-aligned for little benefit. For
4271 instance, in this code, the variable @code{f.x} in @code{struct bar}
4272 will be misaligned even though @code{struct bar} does not itself
4273 have the packed attribute:
4280 @} __attribute__((packed));
4288 @item -Wpacked-bitfield-compat
4289 @opindex Wpacked-bitfield-compat
4290 @opindex Wno-packed-bitfield-compat
4291 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4292 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4293 the change can lead to differences in the structure layout. GCC
4294 informs you when the offset of such a field has changed in GCC 4.4.
4295 For example there is no longer a 4-bit padding between field @code{a}
4296 and @code{b} in this structure:
4303 @} __attribute__ ((packed));
4306 This warning is enabled by default. Use
4307 @option{-Wno-packed-bitfield-compat} to disable this warning.
4312 Warn if padding is included in a structure, either to align an element
4313 of the structure or to align the whole structure. Sometimes when this
4314 happens it is possible to rearrange the fields of the structure to
4315 reduce the padding and so make the structure smaller.
4317 @item -Wredundant-decls
4318 @opindex Wredundant-decls
4319 @opindex Wno-redundant-decls
4320 Warn if anything is declared more than once in the same scope, even in
4321 cases where multiple declaration is valid and changes nothing.
4323 @item -Wnested-externs @r{(C and Objective-C only)}
4324 @opindex Wnested-externs
4325 @opindex Wno-nested-externs
4326 Warn if an @code{extern} declaration is encountered within a function.
4331 Warn if a function can not be inlined and it was declared as inline.
4332 Even with this option, the compiler will not warn about failures to
4333 inline functions declared in system headers.
4335 The compiler uses a variety of heuristics to determine whether or not
4336 to inline a function. For example, the compiler takes into account
4337 the size of the function being inlined and the amount of inlining
4338 that has already been done in the current function. Therefore,
4339 seemingly insignificant changes in the source program can cause the
4340 warnings produced by @option{-Winline} to appear or disappear.
4342 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4343 @opindex Wno-invalid-offsetof
4344 @opindex Winvalid-offsetof
4345 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4346 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4347 to a non-POD type is undefined. In existing C++ implementations,
4348 however, @samp{offsetof} typically gives meaningful results even when
4349 applied to certain kinds of non-POD types. (Such as a simple
4350 @samp{struct} that fails to be a POD type only by virtue of having a
4351 constructor.) This flag is for users who are aware that they are
4352 writing nonportable code and who have deliberately chosen to ignore the
4355 The restrictions on @samp{offsetof} may be relaxed in a future version
4356 of the C++ standard.
4358 @item -Wno-int-to-pointer-cast
4359 @opindex Wno-int-to-pointer-cast
4360 @opindex Wint-to-pointer-cast
4361 Suppress warnings from casts to pointer type of an integer of a
4362 different size. In C++, casting to a pointer type of smaller size is
4363 an error. @option{Wint-to-pointer-cast} is enabled by default.
4366 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4367 @opindex Wno-pointer-to-int-cast
4368 @opindex Wpointer-to-int-cast
4369 Suppress warnings from casts from a pointer to an integer type of a
4373 @opindex Winvalid-pch
4374 @opindex Wno-invalid-pch
4375 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4376 the search path but can't be used.
4380 @opindex Wno-long-long
4381 Warn if @samp{long long} type is used. This is enabled by either
4382 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4383 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4385 @item -Wvariadic-macros
4386 @opindex Wvariadic-macros
4387 @opindex Wno-variadic-macros
4388 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4389 alternate syntax when in pedantic ISO C99 mode. This is default.
4390 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4395 Warn if variable length array is used in the code.
4396 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4397 the variable length array.
4399 @item -Wvolatile-register-var
4400 @opindex Wvolatile-register-var
4401 @opindex Wno-volatile-register-var
4402 Warn if a register variable is declared volatile. The volatile
4403 modifier does not inhibit all optimizations that may eliminate reads
4404 and/or writes to register variables. This warning is enabled by
4407 @item -Wdisabled-optimization
4408 @opindex Wdisabled-optimization
4409 @opindex Wno-disabled-optimization
4410 Warn if a requested optimization pass is disabled. This warning does
4411 not generally indicate that there is anything wrong with your code; it
4412 merely indicates that GCC's optimizers were unable to handle the code
4413 effectively. Often, the problem is that your code is too big or too
4414 complex; GCC will refuse to optimize programs when the optimization
4415 itself is likely to take inordinate amounts of time.
4417 @item -Wpointer-sign @r{(C and Objective-C only)}
4418 @opindex Wpointer-sign
4419 @opindex Wno-pointer-sign
4420 Warn for pointer argument passing or assignment with different signedness.
4421 This option is only supported for C and Objective-C@. It is implied by
4422 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4423 @option{-Wno-pointer-sign}.
4425 @item -Wstack-protector
4426 @opindex Wstack-protector
4427 @opindex Wno-stack-protector
4428 This option is only active when @option{-fstack-protector} is active. It
4429 warns about functions that will not be protected against stack smashing.
4432 @opindex Wno-mudflap
4433 Suppress warnings about constructs that cannot be instrumented by
4436 @item -Woverlength-strings
4437 @opindex Woverlength-strings
4438 @opindex Wno-overlength-strings
4439 Warn about string constants which are longer than the ``minimum
4440 maximum'' length specified in the C standard. Modern compilers
4441 generally allow string constants which are much longer than the
4442 standard's minimum limit, but very portable programs should avoid
4443 using longer strings.
4445 The limit applies @emph{after} string constant concatenation, and does
4446 not count the trailing NUL@. In C90, the limit was 509 characters; in
4447 C99, it was raised to 4095. C++98 does not specify a normative
4448 minimum maximum, so we do not diagnose overlength strings in C++@.
4450 This option is implied by @option{-pedantic}, and can be disabled with
4451 @option{-Wno-overlength-strings}.
4453 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4454 @opindex Wunsuffixed-float-constants
4456 GCC will issue a warning for any floating constant that does not have
4457 a suffix. When used together with @option{-Wsystem-headers} it will
4458 warn about such constants in system header files. This can be useful
4459 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4460 from the decimal floating-point extension to C99.
4463 @node Debugging Options
4464 @section Options for Debugging Your Program or GCC
4465 @cindex options, debugging
4466 @cindex debugging information options
4468 GCC has various special options that are used for debugging
4469 either your program or GCC:
4474 Produce debugging information in the operating system's native format
4475 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4478 On most systems that use stabs format, @option{-g} enables use of extra
4479 debugging information that only GDB can use; this extra information
4480 makes debugging work better in GDB but will probably make other debuggers
4482 refuse to read the program. If you want to control for certain whether
4483 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4484 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4486 GCC allows you to use @option{-g} with
4487 @option{-O}. The shortcuts taken by optimized code may occasionally
4488 produce surprising results: some variables you declared may not exist
4489 at all; flow of control may briefly move where you did not expect it;
4490 some statements may not be executed because they compute constant
4491 results or their values were already at hand; some statements may
4492 execute in different places because they were moved out of loops.
4494 Nevertheless it proves possible to debug optimized output. This makes
4495 it reasonable to use the optimizer for programs that might have bugs.
4497 The following options are useful when GCC is generated with the
4498 capability for more than one debugging format.
4502 Produce debugging information for use by GDB@. This means to use the
4503 most expressive format available (DWARF 2, stabs, or the native format
4504 if neither of those are supported), including GDB extensions if at all
4509 Produce debugging information in stabs format (if that is supported),
4510 without GDB extensions. This is the format used by DBX on most BSD
4511 systems. On MIPS, Alpha and System V Release 4 systems this option
4512 produces stabs debugging output which is not understood by DBX or SDB@.
4513 On System V Release 4 systems this option requires the GNU assembler.
4515 @item -feliminate-unused-debug-symbols
4516 @opindex feliminate-unused-debug-symbols
4517 Produce debugging information in stabs format (if that is supported),
4518 for only symbols that are actually used.
4520 @item -femit-class-debug-always
4521 Instead of emitting debugging information for a C++ class in only one
4522 object file, emit it in all object files using the class. This option
4523 should be used only with debuggers that are unable to handle the way GCC
4524 normally emits debugging information for classes because using this
4525 option will increase the size of debugging information by as much as a
4530 Produce debugging information in stabs format (if that is supported),
4531 using GNU extensions understood only by the GNU debugger (GDB)@. The
4532 use of these extensions is likely to make other debuggers crash or
4533 refuse to read the program.
4537 Produce debugging information in COFF format (if that is supported).
4538 This is the format used by SDB on most System V systems prior to
4543 Produce debugging information in XCOFF format (if that is supported).
4544 This is the format used by the DBX debugger on IBM RS/6000 systems.
4548 Produce debugging information in XCOFF format (if that is supported),
4549 using GNU extensions understood only by the GNU debugger (GDB)@. The
4550 use of these extensions is likely to make other debuggers crash or
4551 refuse to read the program, and may cause assemblers other than the GNU
4552 assembler (GAS) to fail with an error.
4554 @item -gdwarf-@var{version}
4555 @opindex gdwarf-@var{version}
4556 Produce debugging information in DWARF format (if that is
4557 supported). This is the format used by DBX on IRIX 6. The value
4558 of @var{version} may be either 2, 3 or 4; the default version is 2.
4560 Note that with DWARF version 2 some ports require, and will always
4561 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4563 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4564 for maximum benefit.
4566 @item -gstrict-dwarf
4567 @opindex gstrict-dwarf
4568 Disallow using extensions of later DWARF standard version than selected
4569 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4570 DWARF extensions from later standard versions is allowed.
4572 @item -gno-strict-dwarf
4573 @opindex gno-strict-dwarf
4574 Allow using extensions of later DWARF standard version than selected with
4575 @option{-gdwarf-@var{version}}.
4579 Produce debugging information in VMS debug format (if that is
4580 supported). This is the format used by DEBUG on VMS systems.
4583 @itemx -ggdb@var{level}
4584 @itemx -gstabs@var{level}
4585 @itemx -gcoff@var{level}
4586 @itemx -gxcoff@var{level}
4587 @itemx -gvms@var{level}
4588 Request debugging information and also use @var{level} to specify how
4589 much information. The default level is 2.
4591 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4594 Level 1 produces minimal information, enough for making backtraces in
4595 parts of the program that you don't plan to debug. This includes
4596 descriptions of functions and external variables, but no information
4597 about local variables and no line numbers.
4599 Level 3 includes extra information, such as all the macro definitions
4600 present in the program. Some debuggers support macro expansion when
4601 you use @option{-g3}.
4603 @option{-gdwarf-2} does not accept a concatenated debug level, because
4604 GCC used to support an option @option{-gdwarf} that meant to generate
4605 debug information in version 1 of the DWARF format (which is very
4606 different from version 2), and it would have been too confusing. That
4607 debug format is long obsolete, but the option cannot be changed now.
4608 Instead use an additional @option{-g@var{level}} option to change the
4609 debug level for DWARF.
4613 Turn off generation of debug info, if leaving out this option would have
4614 generated it, or turn it on at level 2 otherwise. The position of this
4615 argument in the command line does not matter, it takes effect after all
4616 other options are processed, and it does so only once, no matter how
4617 many times it is given. This is mainly intended to be used with
4618 @option{-fcompare-debug}.
4620 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4621 @opindex fdump-final-insns
4622 Dump the final internal representation (RTL) to @var{file}. If the
4623 optional argument is omitted (or if @var{file} is @code{.}), the name
4624 of the dump file will be determined by appending @code{.gkd} to the
4625 compilation output file name.
4627 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4628 @opindex fcompare-debug
4629 @opindex fno-compare-debug
4630 If no error occurs during compilation, run the compiler a second time,
4631 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4632 passed to the second compilation. Dump the final internal
4633 representation in both compilations, and print an error if they differ.
4635 If the equal sign is omitted, the default @option{-gtoggle} is used.
4637 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4638 and nonzero, implicitly enables @option{-fcompare-debug}. If
4639 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4640 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4643 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4644 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4645 of the final representation and the second compilation, preventing even
4646 @env{GCC_COMPARE_DEBUG} from taking effect.
4648 To verify full coverage during @option{-fcompare-debug} testing, set
4649 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4650 which GCC will reject as an invalid option in any actual compilation
4651 (rather than preprocessing, assembly or linking). To get just a
4652 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4653 not overridden} will do.
4655 @item -fcompare-debug-second
4656 @opindex fcompare-debug-second
4657 This option is implicitly passed to the compiler for the second
4658 compilation requested by @option{-fcompare-debug}, along with options to
4659 silence warnings, and omitting other options that would cause
4660 side-effect compiler outputs to files or to the standard output. Dump
4661 files and preserved temporary files are renamed so as to contain the
4662 @code{.gk} additional extension during the second compilation, to avoid
4663 overwriting those generated by the first.
4665 When this option is passed to the compiler driver, it causes the
4666 @emph{first} compilation to be skipped, which makes it useful for little
4667 other than debugging the compiler proper.
4669 @item -feliminate-dwarf2-dups
4670 @opindex feliminate-dwarf2-dups
4671 Compress DWARF2 debugging information by eliminating duplicated
4672 information about each symbol. This option only makes sense when
4673 generating DWARF2 debugging information with @option{-gdwarf-2}.
4675 @item -femit-struct-debug-baseonly
4676 Emit debug information for struct-like types
4677 only when the base name of the compilation source file
4678 matches the base name of file in which the struct was defined.
4680 This option substantially reduces the size of debugging information,
4681 but at significant potential loss in type information to the debugger.
4682 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4683 See @option{-femit-struct-debug-detailed} for more detailed control.
4685 This option works only with DWARF 2.
4687 @item -femit-struct-debug-reduced
4688 Emit debug information for struct-like types
4689 only when the base name of the compilation source file
4690 matches the base name of file in which the type was defined,
4691 unless the struct is a template or defined in a system header.
4693 This option significantly reduces the size of debugging information,
4694 with some potential loss in type information to the debugger.
4695 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4696 See @option{-femit-struct-debug-detailed} for more detailed control.
4698 This option works only with DWARF 2.
4700 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4701 Specify the struct-like types
4702 for which the compiler will generate debug information.
4703 The intent is to reduce duplicate struct debug information
4704 between different object files within the same program.
4706 This option is a detailed version of
4707 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4708 which will serve for most needs.
4710 A specification has the syntax@*
4711 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4713 The optional first word limits the specification to
4714 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4715 A struct type is used directly when it is the type of a variable, member.
4716 Indirect uses arise through pointers to structs.
4717 That is, when use of an incomplete struct would be legal, the use is indirect.
4719 @samp{struct one direct; struct two * indirect;}.
4721 The optional second word limits the specification to
4722 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4723 Generic structs are a bit complicated to explain.
4724 For C++, these are non-explicit specializations of template classes,
4725 or non-template classes within the above.
4726 Other programming languages have generics,
4727 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4729 The third word specifies the source files for those
4730 structs for which the compiler will emit debug information.
4731 The values @samp{none} and @samp{any} have the normal meaning.
4732 The value @samp{base} means that
4733 the base of name of the file in which the type declaration appears
4734 must match the base of the name of the main compilation file.
4735 In practice, this means that
4736 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4737 but types declared in other header will not.
4738 The value @samp{sys} means those types satisfying @samp{base}
4739 or declared in system or compiler headers.
4741 You may need to experiment to determine the best settings for your application.
4743 The default is @samp{-femit-struct-debug-detailed=all}.
4745 This option works only with DWARF 2.
4747 @item -fenable-icf-debug
4748 @opindex fenable-icf-debug
4749 Generate additional debug information to support identical code folding (ICF).
4750 This option only works with DWARF version 2 or higher.
4752 @item -fno-merge-debug-strings
4753 @opindex fmerge-debug-strings
4754 @opindex fno-merge-debug-strings
4755 Direct the linker to not merge together strings in the debugging
4756 information which are identical in different object files. Merging is
4757 not supported by all assemblers or linkers. Merging decreases the size
4758 of the debug information in the output file at the cost of increasing
4759 link processing time. Merging is enabled by default.
4761 @item -fdebug-prefix-map=@var{old}=@var{new}
4762 @opindex fdebug-prefix-map
4763 When compiling files in directory @file{@var{old}}, record debugging
4764 information describing them as in @file{@var{new}} instead.
4766 @item -fno-dwarf2-cfi-asm
4767 @opindex fdwarf2-cfi-asm
4768 @opindex fno-dwarf2-cfi-asm
4769 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4770 instead of using GAS @code{.cfi_*} directives.
4772 @cindex @command{prof}
4775 Generate extra code to write profile information suitable for the
4776 analysis program @command{prof}. You must use this option when compiling
4777 the source files you want data about, and you must also use it when
4780 @cindex @command{gprof}
4783 Generate extra code to write profile information suitable for the
4784 analysis program @command{gprof}. You must use this option when compiling
4785 the source files you want data about, and you must also use it when
4790 Makes the compiler print out each function name as it is compiled, and
4791 print some statistics about each pass when it finishes.
4794 @opindex ftime-report
4795 Makes the compiler print some statistics about the time consumed by each
4796 pass when it finishes.
4799 @opindex fmem-report
4800 Makes the compiler print some statistics about permanent memory
4801 allocation when it finishes.
4803 @item -fpre-ipa-mem-report
4804 @opindex fpre-ipa-mem-report
4805 @item -fpost-ipa-mem-report
4806 @opindex fpost-ipa-mem-report
4807 Makes the compiler print some statistics about permanent memory
4808 allocation before or after interprocedural optimization.
4811 @opindex fstack-usage
4812 Makes the compiler output stack usage information for the program, on a
4813 per-function basis. The filename for the dump is made by appending
4814 @file{.su} to the AUXNAME. AUXNAME is generated from the name of
4815 the output file, if explicitly specified and it is not an executable,
4816 otherwise it is the basename of the source file. An entry is made up
4821 The name of the function.
4825 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4828 The qualifier @code{static} means that the function manipulates the stack
4829 statically: a fixed number of bytes are allocated for the frame on function
4830 entry and released on function exit; no stack adjustments are otherwise made
4831 in the function. The second field is this fixed number of bytes.
4833 The qualifier @code{dynamic} means that the function manipulates the stack
4834 dynamically: in addition to the static allocation described above, stack
4835 adjustments are made in the body of the function, for example to push/pop
4836 arguments around function calls. If the qualifier @code{bounded} is also
4837 present, the amount of these adjustments is bounded at compile-time and
4838 the second field is an upper bound of the total amount of stack used by
4839 the function. If it is not present, the amount of these adjustments is
4840 not bounded at compile-time and the second field only represents the
4843 @item -fprofile-arcs
4844 @opindex fprofile-arcs
4845 Add code so that program flow @dfn{arcs} are instrumented. During
4846 execution the program records how many times each branch and call is
4847 executed and how many times it is taken or returns. When the compiled
4848 program exits it saves this data to a file called
4849 @file{@var{auxname}.gcda} for each source file. The data may be used for
4850 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4851 test coverage analysis (@option{-ftest-coverage}). Each object file's
4852 @var{auxname} is generated from the name of the output file, if
4853 explicitly specified and it is not the final executable, otherwise it is
4854 the basename of the source file. In both cases any suffix is removed
4855 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4856 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4857 @xref{Cross-profiling}.
4859 @cindex @command{gcov}
4863 This option is used to compile and link code instrumented for coverage
4864 analysis. The option is a synonym for @option{-fprofile-arcs}
4865 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4866 linking). See the documentation for those options for more details.
4871 Compile the source files with @option{-fprofile-arcs} plus optimization
4872 and code generation options. For test coverage analysis, use the
4873 additional @option{-ftest-coverage} option. You do not need to profile
4874 every source file in a program.
4877 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4878 (the latter implies the former).
4881 Run the program on a representative workload to generate the arc profile
4882 information. This may be repeated any number of times. You can run
4883 concurrent instances of your program, and provided that the file system
4884 supports locking, the data files will be correctly updated. Also
4885 @code{fork} calls are detected and correctly handled (double counting
4889 For profile-directed optimizations, compile the source files again with
4890 the same optimization and code generation options plus
4891 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4892 Control Optimization}).
4895 For test coverage analysis, use @command{gcov} to produce human readable
4896 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4897 @command{gcov} documentation for further information.
4901 With @option{-fprofile-arcs}, for each function of your program GCC
4902 creates a program flow graph, then finds a spanning tree for the graph.
4903 Only arcs that are not on the spanning tree have to be instrumented: the
4904 compiler adds code to count the number of times that these arcs are
4905 executed. When an arc is the only exit or only entrance to a block, the
4906 instrumentation code can be added to the block; otherwise, a new basic
4907 block must be created to hold the instrumentation code.
4910 @item -ftest-coverage
4911 @opindex ftest-coverage
4912 Produce a notes file that the @command{gcov} code-coverage utility
4913 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4914 show program coverage. Each source file's note file is called
4915 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4916 above for a description of @var{auxname} and instructions on how to
4917 generate test coverage data. Coverage data will match the source files
4918 more closely, if you do not optimize.
4920 @item -fdbg-cnt-list
4921 @opindex fdbg-cnt-list
4922 Print the name and the counter upperbound for all debug counters.
4924 @item -fdbg-cnt=@var{counter-value-list}
4926 Set the internal debug counter upperbound. @var{counter-value-list}
4927 is a comma-separated list of @var{name}:@var{value} pairs
4928 which sets the upperbound of each debug counter @var{name} to @var{value}.
4929 All debug counters have the initial upperbound of @var{UINT_MAX},
4930 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4931 e.g. With -fdbg-cnt=dce:10,tail_call:0
4932 dbg_cnt(dce) will return true only for first 10 invocations
4933 and dbg_cnt(tail_call) will return false always.
4935 @item -d@var{letters}
4936 @itemx -fdump-rtl-@var{pass}
4938 Says to make debugging dumps during compilation at times specified by
4939 @var{letters}. This is used for debugging the RTL-based passes of the
4940 compiler. The file names for most of the dumps are made by appending
4941 a pass number and a word to the @var{dumpname}, and the files are
4942 created in the directory of the output file. @var{dumpname} is
4943 generated from the name of the output file, if explicitly specified
4944 and it is not an executable, otherwise it is the basename of the
4945 source file. These switches may have different effects when
4946 @option{-E} is used for preprocessing.
4948 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4949 @option{-d} option @var{letters}. Here are the possible
4950 letters for use in @var{pass} and @var{letters}, and their meanings:
4954 @item -fdump-rtl-alignments
4955 @opindex fdump-rtl-alignments
4956 Dump after branch alignments have been computed.
4958 @item -fdump-rtl-asmcons
4959 @opindex fdump-rtl-asmcons
4960 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4962 @item -fdump-rtl-auto_inc_dec
4963 @opindex fdump-rtl-auto_inc_dec
4964 Dump after auto-inc-dec discovery. This pass is only run on
4965 architectures that have auto inc or auto dec instructions.
4967 @item -fdump-rtl-barriers
4968 @opindex fdump-rtl-barriers
4969 Dump after cleaning up the barrier instructions.
4971 @item -fdump-rtl-bbpart
4972 @opindex fdump-rtl-bbpart
4973 Dump after partitioning hot and cold basic blocks.
4975 @item -fdump-rtl-bbro
4976 @opindex fdump-rtl-bbro
4977 Dump after block reordering.
4979 @item -fdump-rtl-btl1
4980 @itemx -fdump-rtl-btl2
4981 @opindex fdump-rtl-btl2
4982 @opindex fdump-rtl-btl2
4983 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4984 after the two branch
4985 target load optimization passes.
4987 @item -fdump-rtl-bypass
4988 @opindex fdump-rtl-bypass
4989 Dump after jump bypassing and control flow optimizations.
4991 @item -fdump-rtl-combine
4992 @opindex fdump-rtl-combine
4993 Dump after the RTL instruction combination pass.
4995 @item -fdump-rtl-compgotos
4996 @opindex fdump-rtl-compgotos
4997 Dump after duplicating the computed gotos.
4999 @item -fdump-rtl-ce1
5000 @itemx -fdump-rtl-ce2
5001 @itemx -fdump-rtl-ce3
5002 @opindex fdump-rtl-ce1
5003 @opindex fdump-rtl-ce2
5004 @opindex fdump-rtl-ce3
5005 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5006 @option{-fdump-rtl-ce3} enable dumping after the three
5007 if conversion passes.
5009 @itemx -fdump-rtl-cprop_hardreg
5010 @opindex fdump-rtl-cprop_hardreg
5011 Dump after hard register copy propagation.
5013 @itemx -fdump-rtl-csa
5014 @opindex fdump-rtl-csa
5015 Dump after combining stack adjustments.
5017 @item -fdump-rtl-cse1
5018 @itemx -fdump-rtl-cse2
5019 @opindex fdump-rtl-cse1
5020 @opindex fdump-rtl-cse2
5021 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5022 the two common sub-expression elimination passes.
5024 @itemx -fdump-rtl-dce
5025 @opindex fdump-rtl-dce
5026 Dump after the standalone dead code elimination passes.
5028 @itemx -fdump-rtl-dbr
5029 @opindex fdump-rtl-dbr
5030 Dump after delayed branch scheduling.
5032 @item -fdump-rtl-dce1
5033 @itemx -fdump-rtl-dce2
5034 @opindex fdump-rtl-dce1
5035 @opindex fdump-rtl-dce2
5036 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5037 the two dead store elimination passes.
5040 @opindex fdump-rtl-eh
5041 Dump after finalization of EH handling code.
5043 @item -fdump-rtl-eh_ranges
5044 @opindex fdump-rtl-eh_ranges
5045 Dump after conversion of EH handling range regions.
5047 @item -fdump-rtl-expand
5048 @opindex fdump-rtl-expand
5049 Dump after RTL generation.
5051 @item -fdump-rtl-fwprop1
5052 @itemx -fdump-rtl-fwprop2
5053 @opindex fdump-rtl-fwprop1
5054 @opindex fdump-rtl-fwprop2
5055 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5056 dumping after the two forward propagation passes.
5058 @item -fdump-rtl-gcse1
5059 @itemx -fdump-rtl-gcse2
5060 @opindex fdump-rtl-gcse1
5061 @opindex fdump-rtl-gcse2
5062 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5063 after global common subexpression elimination.
5065 @item -fdump-rtl-init-regs
5066 @opindex fdump-rtl-init-regs
5067 Dump after the initialization of the registers.
5069 @item -fdump-rtl-initvals
5070 @opindex fdump-rtl-initvals
5071 Dump after the computation of the initial value sets.
5073 @itemx -fdump-rtl-into_cfglayout
5074 @opindex fdump-rtl-into_cfglayout
5075 Dump after converting to cfglayout mode.
5077 @item -fdump-rtl-ira
5078 @opindex fdump-rtl-ira
5079 Dump after iterated register allocation.
5081 @item -fdump-rtl-jump
5082 @opindex fdump-rtl-jump
5083 Dump after the second jump optimization.
5085 @item -fdump-rtl-loop2
5086 @opindex fdump-rtl-loop2
5087 @option{-fdump-rtl-loop2} enables dumping after the rtl
5088 loop optimization passes.
5090 @item -fdump-rtl-mach
5091 @opindex fdump-rtl-mach
5092 Dump after performing the machine dependent reorganization pass, if that
5095 @item -fdump-rtl-mode_sw
5096 @opindex fdump-rtl-mode_sw
5097 Dump after removing redundant mode switches.
5099 @item -fdump-rtl-rnreg
5100 @opindex fdump-rtl-rnreg
5101 Dump after register renumbering.
5103 @itemx -fdump-rtl-outof_cfglayout
5104 @opindex fdump-rtl-outof_cfglayout
5105 Dump after converting from cfglayout mode.
5107 @item -fdump-rtl-peephole2
5108 @opindex fdump-rtl-peephole2
5109 Dump after the peephole pass.
5111 @item -fdump-rtl-postreload
5112 @opindex fdump-rtl-postreload
5113 Dump after post-reload optimizations.
5115 @itemx -fdump-rtl-pro_and_epilogue
5116 @opindex fdump-rtl-pro_and_epilogue
5117 Dump after generating the function pro and epilogues.
5119 @item -fdump-rtl-regmove
5120 @opindex fdump-rtl-regmove
5121 Dump after the register move pass.
5123 @item -fdump-rtl-sched1
5124 @itemx -fdump-rtl-sched2
5125 @opindex fdump-rtl-sched1
5126 @opindex fdump-rtl-sched2
5127 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5128 after the basic block scheduling passes.
5130 @item -fdump-rtl-see
5131 @opindex fdump-rtl-see
5132 Dump after sign extension elimination.
5134 @item -fdump-rtl-seqabstr
5135 @opindex fdump-rtl-seqabstr
5136 Dump after common sequence discovery.
5138 @item -fdump-rtl-shorten
5139 @opindex fdump-rtl-shorten
5140 Dump after shortening branches.
5142 @item -fdump-rtl-sibling
5143 @opindex fdump-rtl-sibling
5144 Dump after sibling call optimizations.
5146 @item -fdump-rtl-split1
5147 @itemx -fdump-rtl-split2
5148 @itemx -fdump-rtl-split3
5149 @itemx -fdump-rtl-split4
5150 @itemx -fdump-rtl-split5
5151 @opindex fdump-rtl-split1
5152 @opindex fdump-rtl-split2
5153 @opindex fdump-rtl-split3
5154 @opindex fdump-rtl-split4
5155 @opindex fdump-rtl-split5
5156 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5157 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5158 @option{-fdump-rtl-split5} enable dumping after five rounds of
5159 instruction splitting.
5161 @item -fdump-rtl-sms
5162 @opindex fdump-rtl-sms
5163 Dump after modulo scheduling. This pass is only run on some
5166 @item -fdump-rtl-stack
5167 @opindex fdump-rtl-stack
5168 Dump after conversion from GCC's "flat register file" registers to the
5169 x87's stack-like registers. This pass is only run on x86 variants.
5171 @item -fdump-rtl-subreg1
5172 @itemx -fdump-rtl-subreg2
5173 @opindex fdump-rtl-subreg1
5174 @opindex fdump-rtl-subreg2
5175 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5176 the two subreg expansion passes.
5178 @item -fdump-rtl-unshare
5179 @opindex fdump-rtl-unshare
5180 Dump after all rtl has been unshared.
5182 @item -fdump-rtl-vartrack
5183 @opindex fdump-rtl-vartrack
5184 Dump after variable tracking.
5186 @item -fdump-rtl-vregs
5187 @opindex fdump-rtl-vregs
5188 Dump after converting virtual registers to hard registers.
5190 @item -fdump-rtl-web
5191 @opindex fdump-rtl-web
5192 Dump after live range splitting.
5194 @item -fdump-rtl-regclass
5195 @itemx -fdump-rtl-subregs_of_mode_init
5196 @itemx -fdump-rtl-subregs_of_mode_finish
5197 @itemx -fdump-rtl-dfinit
5198 @itemx -fdump-rtl-dfinish
5199 @opindex fdump-rtl-regclass
5200 @opindex fdump-rtl-subregs_of_mode_init
5201 @opindex fdump-rtl-subregs_of_mode_finish
5202 @opindex fdump-rtl-dfinit
5203 @opindex fdump-rtl-dfinish
5204 These dumps are defined but always produce empty files.
5206 @item -fdump-rtl-all
5207 @opindex fdump-rtl-all
5208 Produce all the dumps listed above.
5212 Annotate the assembler output with miscellaneous debugging information.
5216 Dump all macro definitions, at the end of preprocessing, in addition to
5221 Produce a core dump whenever an error occurs.
5225 Print statistics on memory usage, at the end of the run, to
5230 Annotate the assembler output with a comment indicating which
5231 pattern and alternative was used. The length of each instruction is
5236 Dump the RTL in the assembler output as a comment before each instruction.
5237 Also turns on @option{-dp} annotation.
5241 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5242 dump a representation of the control flow graph suitable for viewing with VCG
5243 to @file{@var{file}.@var{pass}.vcg}.
5247 Just generate RTL for a function instead of compiling it. Usually used
5248 with @option{-fdump-rtl-expand}.
5252 Dump debugging information during parsing, to standard error.
5256 @opindex fdump-noaddr
5257 When doing debugging dumps, suppress address output. This makes it more
5258 feasible to use diff on debugging dumps for compiler invocations with
5259 different compiler binaries and/or different
5260 text / bss / data / heap / stack / dso start locations.
5262 @item -fdump-unnumbered
5263 @opindex fdump-unnumbered
5264 When doing debugging dumps, suppress instruction numbers and address output.
5265 This makes it more feasible to use diff on debugging dumps for compiler
5266 invocations with different options, in particular with and without
5269 @item -fdump-unnumbered-links
5270 @opindex fdump-unnumbered-links
5271 When doing debugging dumps (see @option{-d} option above), suppress
5272 instruction numbers for the links to the previous and next instructions
5275 @item -fdump-translation-unit @r{(C++ only)}
5276 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5277 @opindex fdump-translation-unit
5278 Dump a representation of the tree structure for the entire translation
5279 unit to a file. The file name is made by appending @file{.tu} to the
5280 source file name, and the file is created in the same directory as the
5281 output file. If the @samp{-@var{options}} form is used, @var{options}
5282 controls the details of the dump as described for the
5283 @option{-fdump-tree} options.
5285 @item -fdump-class-hierarchy @r{(C++ only)}
5286 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5287 @opindex fdump-class-hierarchy
5288 Dump a representation of each class's hierarchy and virtual function
5289 table layout to a file. The file name is made by appending
5290 @file{.class} to the source file name, and the file is created in the
5291 same directory as the output file. If the @samp{-@var{options}} form
5292 is used, @var{options} controls the details of the dump as described
5293 for the @option{-fdump-tree} options.
5295 @item -fdump-ipa-@var{switch}
5297 Control the dumping at various stages of inter-procedural analysis
5298 language tree to a file. The file name is generated by appending a
5299 switch specific suffix to the source file name, and the file is created
5300 in the same directory as the output file. The following dumps are
5305 Enables all inter-procedural analysis dumps.
5308 Dumps information about call-graph optimization, unused function removal,
5309 and inlining decisions.
5312 Dump after function inlining.
5316 @item -fdump-statistics-@var{option}
5317 @opindex fdump-statistics
5318 Enable and control dumping of pass statistics in a separate file. The
5319 file name is generated by appending a suffix ending in
5320 @samp{.statistics} to the source file name, and the file is created in
5321 the same directory as the output file. If the @samp{-@var{option}}
5322 form is used, @samp{-stats} will cause counters to be summed over the
5323 whole compilation unit while @samp{-details} will dump every event as
5324 the passes generate them. The default with no option is to sum
5325 counters for each function compiled.
5327 @item -fdump-tree-@var{switch}
5328 @itemx -fdump-tree-@var{switch}-@var{options}
5330 Control the dumping at various stages of processing the intermediate
5331 language tree to a file. The file name is generated by appending a
5332 switch specific suffix to the source file name, and the file is
5333 created in the same directory as the output file. If the
5334 @samp{-@var{options}} form is used, @var{options} is a list of
5335 @samp{-} separated options that control the details of the dump. Not
5336 all options are applicable to all dumps, those which are not
5337 meaningful will be ignored. The following options are available
5341 Print the address of each node. Usually this is not meaningful as it
5342 changes according to the environment and source file. Its primary use
5343 is for tying up a dump file with a debug environment.
5345 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5346 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5347 use working backward from mangled names in the assembly file.
5349 Inhibit dumping of members of a scope or body of a function merely
5350 because that scope has been reached. Only dump such items when they
5351 are directly reachable by some other path. When dumping pretty-printed
5352 trees, this option inhibits dumping the bodies of control structures.
5354 Print a raw representation of the tree. By default, trees are
5355 pretty-printed into a C-like representation.
5357 Enable more detailed dumps (not honored by every dump option).
5359 Enable dumping various statistics about the pass (not honored by every dump
5362 Enable showing basic block boundaries (disabled in raw dumps).
5364 Enable showing virtual operands for every statement.
5366 Enable showing line numbers for statements.
5368 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5370 Enable showing the tree dump for each statement.
5372 Enable showing the EH region number holding each statement.
5374 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5375 and @option{lineno}.
5378 The following tree dumps are possible:
5382 @opindex fdump-tree-original
5383 Dump before any tree based optimization, to @file{@var{file}.original}.
5386 @opindex fdump-tree-optimized
5387 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5390 @opindex fdump-tree-gimple
5391 Dump each function before and after the gimplification pass to a file. The
5392 file name is made by appending @file{.gimple} to the source file name.
5395 @opindex fdump-tree-cfg
5396 Dump the control flow graph of each function to a file. The file name is
5397 made by appending @file{.cfg} to the source file name.
5400 @opindex fdump-tree-vcg
5401 Dump the control flow graph of each function to a file in VCG format. The
5402 file name is made by appending @file{.vcg} to the source file name. Note
5403 that if the file contains more than one function, the generated file cannot
5404 be used directly by VCG@. You will need to cut and paste each function's
5405 graph into its own separate file first.
5408 @opindex fdump-tree-ch
5409 Dump each function after copying loop headers. The file name is made by
5410 appending @file{.ch} to the source file name.
5413 @opindex fdump-tree-ssa
5414 Dump SSA related information to a file. The file name is made by appending
5415 @file{.ssa} to the source file name.
5418 @opindex fdump-tree-alias
5419 Dump aliasing information for each function. The file name is made by
5420 appending @file{.alias} to the source file name.
5423 @opindex fdump-tree-ccp
5424 Dump each function after CCP@. The file name is made by appending
5425 @file{.ccp} to the source file name.
5428 @opindex fdump-tree-storeccp
5429 Dump each function after STORE-CCP@. The file name is made by appending
5430 @file{.storeccp} to the source file name.
5433 @opindex fdump-tree-pre
5434 Dump trees after partial redundancy elimination. The file name is made
5435 by appending @file{.pre} to the source file name.
5438 @opindex fdump-tree-fre
5439 Dump trees after full redundancy elimination. The file name is made
5440 by appending @file{.fre} to the source file name.
5443 @opindex fdump-tree-copyprop
5444 Dump trees after copy propagation. The file name is made
5445 by appending @file{.copyprop} to the source file name.
5447 @item store_copyprop
5448 @opindex fdump-tree-store_copyprop
5449 Dump trees after store copy-propagation. The file name is made
5450 by appending @file{.store_copyprop} to the source file name.
5453 @opindex fdump-tree-dce
5454 Dump each function after dead code elimination. The file name is made by
5455 appending @file{.dce} to the source file name.
5458 @opindex fdump-tree-mudflap
5459 Dump each function after adding mudflap instrumentation. The file name is
5460 made by appending @file{.mudflap} to the source file name.
5463 @opindex fdump-tree-sra
5464 Dump each function after performing scalar replacement of aggregates. The
5465 file name is made by appending @file{.sra} to the source file name.
5468 @opindex fdump-tree-sink
5469 Dump each function after performing code sinking. The file name is made
5470 by appending @file{.sink} to the source file name.
5473 @opindex fdump-tree-dom
5474 Dump each function after applying dominator tree optimizations. The file
5475 name is made by appending @file{.dom} to the source file name.
5478 @opindex fdump-tree-dse
5479 Dump each function after applying dead store elimination. The file
5480 name is made by appending @file{.dse} to the source file name.
5483 @opindex fdump-tree-phiopt
5484 Dump each function after optimizing PHI nodes into straightline code. The file
5485 name is made by appending @file{.phiopt} to the source file name.
5488 @opindex fdump-tree-forwprop
5489 Dump each function after forward propagating single use variables. The file
5490 name is made by appending @file{.forwprop} to the source file name.
5493 @opindex fdump-tree-copyrename
5494 Dump each function after applying the copy rename optimization. The file
5495 name is made by appending @file{.copyrename} to the source file name.
5498 @opindex fdump-tree-nrv
5499 Dump each function after applying the named return value optimization on
5500 generic trees. The file name is made by appending @file{.nrv} to the source
5504 @opindex fdump-tree-vect
5505 Dump each function after applying vectorization of loops. The file name is
5506 made by appending @file{.vect} to the source file name.
5509 @opindex fdump-tree-slp
5510 Dump each function after applying vectorization of basic blocks. The file name
5511 is made by appending @file{.slp} to the source file name.
5514 @opindex fdump-tree-vrp
5515 Dump each function after Value Range Propagation (VRP). The file name
5516 is made by appending @file{.vrp} to the source file name.
5519 @opindex fdump-tree-all
5520 Enable all the available tree dumps with the flags provided in this option.
5523 @item -ftree-vectorizer-verbose=@var{n}
5524 @opindex ftree-vectorizer-verbose
5525 This option controls the amount of debugging output the vectorizer prints.
5526 This information is written to standard error, unless
5527 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5528 in which case it is output to the usual dump listing file, @file{.vect}.
5529 For @var{n}=0 no diagnostic information is reported.
5530 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5531 and the total number of loops that got vectorized.
5532 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5533 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5534 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5535 level that @option{-fdump-tree-vect-stats} uses.
5536 Higher verbosity levels mean either more information dumped for each
5537 reported loop, or same amount of information reported for more loops:
5538 if @var{n}=3, vectorizer cost model information is reported.
5539 If @var{n}=4, alignment related information is added to the reports.
5540 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5541 memory access-patterns) is added to the reports.
5542 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5543 that did not pass the first analysis phase (i.e., may not be countable, or
5544 may have complicated control-flow).
5545 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5546 If @var{n}=8, SLP related information is added to the reports.
5547 For @var{n}=9, all the information the vectorizer generates during its
5548 analysis and transformation is reported. This is the same verbosity level
5549 that @option{-fdump-tree-vect-details} uses.
5551 @item -frandom-seed=@var{string}
5552 @opindex frandom-seed
5553 This option provides a seed that GCC uses when it would otherwise use
5554 random numbers. It is used to generate certain symbol names
5555 that have to be different in every compiled file. It is also used to
5556 place unique stamps in coverage data files and the object files that
5557 produce them. You can use the @option{-frandom-seed} option to produce
5558 reproducibly identical object files.
5560 The @var{string} should be different for every file you compile.
5562 @item -fsched-verbose=@var{n}
5563 @opindex fsched-verbose
5564 On targets that use instruction scheduling, this option controls the
5565 amount of debugging output the scheduler prints. This information is
5566 written to standard error, unless @option{-fdump-rtl-sched1} or
5567 @option{-fdump-rtl-sched2} is specified, in which case it is output
5568 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5569 respectively. However for @var{n} greater than nine, the output is
5570 always printed to standard error.
5572 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5573 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5574 For @var{n} greater than one, it also output basic block probabilities,
5575 detailed ready list information and unit/insn info. For @var{n} greater
5576 than two, it includes RTL at abort point, control-flow and regions info.
5577 And for @var{n} over four, @option{-fsched-verbose} also includes
5581 @itemx -save-temps=cwd
5583 Store the usual ``temporary'' intermediate files permanently; place them
5584 in the current directory and name them based on the source file. Thus,
5585 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5586 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5587 preprocessed @file{foo.i} output file even though the compiler now
5588 normally uses an integrated preprocessor.
5590 When used in combination with the @option{-x} command line option,
5591 @option{-save-temps} is sensible enough to avoid over writing an
5592 input source file with the same extension as an intermediate file.
5593 The corresponding intermediate file may be obtained by renaming the
5594 source file before using @option{-save-temps}.
5596 If you invoke GCC in parallel, compiling several different source
5597 files that share a common base name in different subdirectories or the
5598 same source file compiled for multiple output destinations, it is
5599 likely that the different parallel compilers will interfere with each
5600 other, and overwrite the temporary files. For instance:
5603 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5604 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5607 may result in @file{foo.i} and @file{foo.o} being written to
5608 simultaneously by both compilers.
5610 @item -save-temps=obj
5611 @opindex save-temps=obj
5612 Store the usual ``temporary'' intermediate files permanently. If the
5613 @option{-o} option is used, the temporary files are based on the
5614 object file. If the @option{-o} option is not used, the
5615 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5620 gcc -save-temps=obj -c foo.c
5621 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5622 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5625 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5626 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5627 @file{dir2/yfoobar.o}.
5629 @item -time@r{[}=@var{file}@r{]}
5631 Report the CPU time taken by each subprocess in the compilation
5632 sequence. For C source files, this is the compiler proper and assembler
5633 (plus the linker if linking is done).
5635 Without the specification of an output file, the output looks like this:
5642 The first number on each line is the ``user time'', that is time spent
5643 executing the program itself. The second number is ``system time'',
5644 time spent executing operating system routines on behalf of the program.
5645 Both numbers are in seconds.
5647 With the specification of an output file, the output is appended to the
5648 named file, and it looks like this:
5651 0.12 0.01 cc1 @var{options}
5652 0.00 0.01 as @var{options}
5655 The ``user time'' and the ``system time'' are moved before the program
5656 name, and the options passed to the program are displayed, so that one
5657 can later tell what file was being compiled, and with which options.
5659 @item -fvar-tracking
5660 @opindex fvar-tracking
5661 Run variable tracking pass. It computes where variables are stored at each
5662 position in code. Better debugging information is then generated
5663 (if the debugging information format supports this information).
5665 It is enabled by default when compiling with optimization (@option{-Os},
5666 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5667 the debug info format supports it.
5669 @item -fvar-tracking-assignments
5670 @opindex fvar-tracking-assignments
5671 @opindex fno-var-tracking-assignments
5672 Annotate assignments to user variables early in the compilation and
5673 attempt to carry the annotations over throughout the compilation all the
5674 way to the end, in an attempt to improve debug information while
5675 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5677 It can be enabled even if var-tracking is disabled, in which case
5678 annotations will be created and maintained, but discarded at the end.
5680 @item -fvar-tracking-assignments-toggle
5681 @opindex fvar-tracking-assignments-toggle
5682 @opindex fno-var-tracking-assignments-toggle
5683 Toggle @option{-fvar-tracking-assignments}, in the same way that
5684 @option{-gtoggle} toggles @option{-g}.
5686 @item -print-file-name=@var{library}
5687 @opindex print-file-name
5688 Print the full absolute name of the library file @var{library} that
5689 would be used when linking---and don't do anything else. With this
5690 option, GCC does not compile or link anything; it just prints the
5693 @item -print-multi-directory
5694 @opindex print-multi-directory
5695 Print the directory name corresponding to the multilib selected by any
5696 other switches present in the command line. This directory is supposed
5697 to exist in @env{GCC_EXEC_PREFIX}.
5699 @item -print-multi-lib
5700 @opindex print-multi-lib
5701 Print the mapping from multilib directory names to compiler switches
5702 that enable them. The directory name is separated from the switches by
5703 @samp{;}, and each switch starts with an @samp{@@} instead of the
5704 @samp{-}, without spaces between multiple switches. This is supposed to
5705 ease shell-processing.
5707 @item -print-multi-os-directory
5708 @opindex print-multi-os-directory
5709 Print the path to OS libraries for the selected
5710 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5711 present in the @file{lib} subdirectory and no multilibs are used, this is
5712 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5713 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5714 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5715 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5717 @item -print-prog-name=@var{program}
5718 @opindex print-prog-name
5719 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5721 @item -print-libgcc-file-name
5722 @opindex print-libgcc-file-name
5723 Same as @option{-print-file-name=libgcc.a}.
5725 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5726 but you do want to link with @file{libgcc.a}. You can do
5729 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5732 @item -print-search-dirs
5733 @opindex print-search-dirs
5734 Print the name of the configured installation directory and a list of
5735 program and library directories @command{gcc} will search---and don't do anything else.
5737 This is useful when @command{gcc} prints the error message
5738 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5739 To resolve this you either need to put @file{cpp0} and the other compiler
5740 components where @command{gcc} expects to find them, or you can set the environment
5741 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5742 Don't forget the trailing @samp{/}.
5743 @xref{Environment Variables}.
5745 @item -print-sysroot
5746 @opindex print-sysroot
5747 Print the target sysroot directory that will be used during
5748 compilation. This is the target sysroot specified either at configure
5749 time or using the @option{--sysroot} option, possibly with an extra
5750 suffix that depends on compilation options. If no target sysroot is
5751 specified, the option prints nothing.
5753 @item -print-sysroot-headers-suffix
5754 @opindex print-sysroot-headers-suffix
5755 Print the suffix added to the target sysroot when searching for
5756 headers, or give an error if the compiler is not configured with such
5757 a suffix---and don't do anything else.
5760 @opindex dumpmachine
5761 Print the compiler's target machine (for example,
5762 @samp{i686-pc-linux-gnu})---and don't do anything else.
5765 @opindex dumpversion
5766 Print the compiler version (for example, @samp{3.0})---and don't do
5771 Print the compiler's built-in specs---and don't do anything else. (This
5772 is used when GCC itself is being built.) @xref{Spec Files}.
5774 @item -feliminate-unused-debug-types
5775 @opindex feliminate-unused-debug-types
5776 Normally, when producing DWARF2 output, GCC will emit debugging
5777 information for all types declared in a compilation
5778 unit, regardless of whether or not they are actually used
5779 in that compilation unit. Sometimes this is useful, such as
5780 if, in the debugger, you want to cast a value to a type that is
5781 not actually used in your program (but is declared). More often,
5782 however, this results in a significant amount of wasted space.
5783 With this option, GCC will avoid producing debug symbol output
5784 for types that are nowhere used in the source file being compiled.
5787 @node Optimize Options
5788 @section Options That Control Optimization
5789 @cindex optimize options
5790 @cindex options, optimization
5792 These options control various sorts of optimizations.
5794 Without any optimization option, the compiler's goal is to reduce the
5795 cost of compilation and to make debugging produce the expected
5796 results. Statements are independent: if you stop the program with a
5797 breakpoint between statements, you can then assign a new value to any
5798 variable or change the program counter to any other statement in the
5799 function and get exactly the results you would expect from the source
5802 Turning on optimization flags makes the compiler attempt to improve
5803 the performance and/or code size at the expense of compilation time
5804 and possibly the ability to debug the program.
5806 The compiler performs optimization based on the knowledge it has of the
5807 program. Compiling multiple files at once to a single output file mode allows
5808 the compiler to use information gained from all of the files when compiling
5811 Not all optimizations are controlled directly by a flag. Only
5812 optimizations that have a flag are listed in this section.
5814 Most optimizations are only enabled if an @option{-O} level is set on
5815 the command line. Otherwise they are disabled, even if individual
5816 optimization flags are specified.
5818 Depending on the target and how GCC was configured, a slightly different
5819 set of optimizations may be enabled at each @option{-O} level than
5820 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5821 to find out the exact set of optimizations that are enabled at each level.
5822 @xref{Overall Options}, for examples.
5829 Optimize. Optimizing compilation takes somewhat more time, and a lot
5830 more memory for a large function.
5832 With @option{-O}, the compiler tries to reduce code size and execution
5833 time, without performing any optimizations that take a great deal of
5836 @option{-O} turns on the following optimization flags:
5839 -fcprop-registers @gol
5842 -fdelayed-branch @gol
5844 -fguess-branch-probability @gol
5845 -fif-conversion2 @gol
5846 -fif-conversion @gol
5847 -fipa-pure-const @gol
5849 -fipa-reference @gol
5851 -fsplit-wide-types @gol
5853 -ftree-builtin-call-dce @gol
5856 -ftree-copyrename @gol
5858 -ftree-dominator-opts @gol
5860 -ftree-forwprop @gol
5868 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5869 where doing so does not interfere with debugging.
5873 Optimize even more. GCC performs nearly all supported optimizations
5874 that do not involve a space-speed tradeoff.
5875 As compared to @option{-O}, this option increases both compilation time
5876 and the performance of the generated code.
5878 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5879 also turns on the following optimization flags:
5880 @gccoptlist{-fthread-jumps @gol
5881 -falign-functions -falign-jumps @gol
5882 -falign-loops -falign-labels @gol
5885 -fcse-follow-jumps -fcse-skip-blocks @gol
5886 -fdelete-null-pointer-checks @gol
5887 -fexpensive-optimizations @gol
5888 -fgcse -fgcse-lm @gol
5889 -finline-small-functions @gol
5890 -findirect-inlining @gol
5892 -foptimize-sibling-calls @gol
5893 -fpartial-inlining @gol
5896 -freorder-blocks -freorder-functions @gol
5897 -frerun-cse-after-loop @gol
5898 -fsched-interblock -fsched-spec @gol
5899 -fschedule-insns -fschedule-insns2 @gol
5900 -fstrict-aliasing -fstrict-overflow @gol
5901 -ftree-switch-conversion @gol
5905 Please note the warning under @option{-fgcse} about
5906 invoking @option{-O2} on programs that use computed gotos.
5910 Optimize yet more. @option{-O3} turns on all optimizations specified
5911 by @option{-O2} and also turns on the @option{-finline-functions},
5912 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5913 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5914 @option{-fipa-cp-clone} options.
5918 Reduce compilation time and make debugging produce the expected
5919 results. This is the default.
5923 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5924 do not typically increase code size. It also performs further
5925 optimizations designed to reduce code size.
5927 @option{-Os} disables the following optimization flags:
5928 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5929 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5930 -fprefetch-loop-arrays -ftree-vect-loop-version}
5934 Disregard strict standards compliance. @option{-Ofast} enables all
5935 @option{-O3} optimizations. It also enables optimizations that are not
5936 valid for all standard compliant programs.
5937 It turns on @option{-ffast-math}.
5939 If you use multiple @option{-O} options, with or without level numbers,
5940 the last such option is the one that is effective.
5943 Options of the form @option{-f@var{flag}} specify machine-independent
5944 flags. Most flags have both positive and negative forms; the negative
5945 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5946 below, only one of the forms is listed---the one you typically will
5947 use. You can figure out the other form by either removing @samp{no-}
5950 The following options control specific optimizations. They are either
5951 activated by @option{-O} options or are related to ones that are. You
5952 can use the following flags in the rare cases when ``fine-tuning'' of
5953 optimizations to be performed is desired.
5956 @item -fno-default-inline
5957 @opindex fno-default-inline
5958 Do not make member functions inline by default merely because they are
5959 defined inside the class scope (C++ only). Otherwise, when you specify
5960 @w{@option{-O}}, member functions defined inside class scope are compiled
5961 inline by default; i.e., you don't need to add @samp{inline} in front of
5962 the member function name.
5964 @item -fno-defer-pop
5965 @opindex fno-defer-pop
5966 Always pop the arguments to each function call as soon as that function
5967 returns. For machines which must pop arguments after a function call,
5968 the compiler normally lets arguments accumulate on the stack for several
5969 function calls and pops them all at once.
5971 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5973 @item -fforward-propagate
5974 @opindex fforward-propagate
5975 Perform a forward propagation pass on RTL@. The pass tries to combine two
5976 instructions and checks if the result can be simplified. If loop unrolling
5977 is active, two passes are performed and the second is scheduled after
5980 This option is enabled by default at optimization levels @option{-O},
5981 @option{-O2}, @option{-O3}, @option{-Os}.
5983 @item -fomit-frame-pointer
5984 @opindex fomit-frame-pointer
5985 Don't keep the frame pointer in a register for functions that
5986 don't need one. This avoids the instructions to save, set up and
5987 restore frame pointers; it also makes an extra register available
5988 in many functions. @strong{It also makes debugging impossible on
5991 On some machines, such as the VAX, this flag has no effect, because
5992 the standard calling sequence automatically handles the frame pointer
5993 and nothing is saved by pretending it doesn't exist. The
5994 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5995 whether a target machine supports this flag. @xref{Registers,,Register
5996 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5998 Starting with GCC version 4.6, the default setting (when not optimizing for
5999 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6000 @option{-fomit-frame-pointer}. The default can be reverted to
6001 @option{-fno-omit-frame-pointer} by configuring GCC with the
6002 @option{--enable-frame-pointer} configure option.
6004 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6006 @item -foptimize-sibling-calls
6007 @opindex foptimize-sibling-calls
6008 Optimize sibling and tail recursive calls.
6010 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6014 Don't pay attention to the @code{inline} keyword. Normally this option
6015 is used to keep the compiler from expanding any functions inline.
6016 Note that if you are not optimizing, no functions can be expanded inline.
6018 @item -finline-small-functions
6019 @opindex finline-small-functions
6020 Integrate functions into their callers when their body is smaller than expected
6021 function call code (so overall size of program gets smaller). The compiler
6022 heuristically decides which functions are simple enough to be worth integrating
6025 Enabled at level @option{-O2}.
6027 @item -findirect-inlining
6028 @opindex findirect-inlining
6029 Inline also indirect calls that are discovered to be known at compile
6030 time thanks to previous inlining. This option has any effect only
6031 when inlining itself is turned on by the @option{-finline-functions}
6032 or @option{-finline-small-functions} options.
6034 Enabled at level @option{-O2}.
6036 @item -finline-functions
6037 @opindex finline-functions
6038 Integrate all simple functions into their callers. The compiler
6039 heuristically decides which functions are simple enough to be worth
6040 integrating in this way.
6042 If all calls to a given function are integrated, and the function is
6043 declared @code{static}, then the function is normally not output as
6044 assembler code in its own right.
6046 Enabled at level @option{-O3}.
6048 @item -finline-functions-called-once
6049 @opindex finline-functions-called-once
6050 Consider all @code{static} functions called once for inlining into their
6051 caller even if they are not marked @code{inline}. If a call to a given
6052 function is integrated, then the function is not output as assembler code
6055 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6057 @item -fearly-inlining
6058 @opindex fearly-inlining
6059 Inline functions marked by @code{always_inline} and functions whose body seems
6060 smaller than the function call overhead early before doing
6061 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6062 makes profiling significantly cheaper and usually inlining faster on programs
6063 having large chains of nested wrapper functions.
6069 Perform interprocedural scalar replacement of aggregates, removal of
6070 unused parameters and replacement of parameters passed by reference
6071 by parameters passed by value.
6073 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6075 @item -finline-limit=@var{n}
6076 @opindex finline-limit
6077 By default, GCC limits the size of functions that can be inlined. This flag
6078 allows coarse control of this limit. @var{n} is the size of functions that
6079 can be inlined in number of pseudo instructions.
6081 Inlining is actually controlled by a number of parameters, which may be
6082 specified individually by using @option{--param @var{name}=@var{value}}.
6083 The @option{-finline-limit=@var{n}} option sets some of these parameters
6087 @item max-inline-insns-single
6088 is set to @var{n}/2.
6089 @item max-inline-insns-auto
6090 is set to @var{n}/2.
6093 See below for a documentation of the individual
6094 parameters controlling inlining and for the defaults of these parameters.
6096 @emph{Note:} there may be no value to @option{-finline-limit} that results
6097 in default behavior.
6099 @emph{Note:} pseudo instruction represents, in this particular context, an
6100 abstract measurement of function's size. In no way does it represent a count
6101 of assembly instructions and as such its exact meaning might change from one
6102 release to an another.
6104 @item -fkeep-inline-functions
6105 @opindex fkeep-inline-functions
6106 In C, emit @code{static} functions that are declared @code{inline}
6107 into the object file, even if the function has been inlined into all
6108 of its callers. This switch does not affect functions using the
6109 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6110 inline functions into the object file.
6112 @item -fkeep-static-consts
6113 @opindex fkeep-static-consts
6114 Emit variables declared @code{static const} when optimization isn't turned
6115 on, even if the variables aren't referenced.
6117 GCC enables this option by default. If you want to force the compiler to
6118 check if the variable was referenced, regardless of whether or not
6119 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6121 @item -fmerge-constants
6122 @opindex fmerge-constants
6123 Attempt to merge identical constants (string constants and floating point
6124 constants) across compilation units.
6126 This option is the default for optimized compilation if the assembler and
6127 linker support it. Use @option{-fno-merge-constants} to inhibit this
6130 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6132 @item -fmerge-all-constants
6133 @opindex fmerge-all-constants
6134 Attempt to merge identical constants and identical variables.
6136 This option implies @option{-fmerge-constants}. In addition to
6137 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6138 arrays or initialized constant variables with integral or floating point
6139 types. Languages like C or C++ require each variable, including multiple
6140 instances of the same variable in recursive calls, to have distinct locations,
6141 so using this option will result in non-conforming
6144 @item -fmodulo-sched
6145 @opindex fmodulo-sched
6146 Perform swing modulo scheduling immediately before the first scheduling
6147 pass. This pass looks at innermost loops and reorders their
6148 instructions by overlapping different iterations.
6150 @item -fmodulo-sched-allow-regmoves
6151 @opindex fmodulo-sched-allow-regmoves
6152 Perform more aggressive SMS based modulo scheduling with register moves
6153 allowed. By setting this flag certain anti-dependences edges will be
6154 deleted which will trigger the generation of reg-moves based on the
6155 life-range analysis. This option is effective only with
6156 @option{-fmodulo-sched} enabled.
6158 @item -fno-branch-count-reg
6159 @opindex fno-branch-count-reg
6160 Do not use ``decrement and branch'' instructions on a count register,
6161 but instead generate a sequence of instructions that decrement a
6162 register, compare it against zero, then branch based upon the result.
6163 This option is only meaningful on architectures that support such
6164 instructions, which include x86, PowerPC, IA-64 and S/390.
6166 The default is @option{-fbranch-count-reg}.
6168 @item -fno-function-cse
6169 @opindex fno-function-cse
6170 Do not put function addresses in registers; make each instruction that
6171 calls a constant function contain the function's address explicitly.
6173 This option results in less efficient code, but some strange hacks
6174 that alter the assembler output may be confused by the optimizations
6175 performed when this option is not used.
6177 The default is @option{-ffunction-cse}
6179 @item -fno-zero-initialized-in-bss
6180 @opindex fno-zero-initialized-in-bss
6181 If the target supports a BSS section, GCC by default puts variables that
6182 are initialized to zero into BSS@. This can save space in the resulting
6185 This option turns off this behavior because some programs explicitly
6186 rely on variables going to the data section. E.g., so that the
6187 resulting executable can find the beginning of that section and/or make
6188 assumptions based on that.
6190 The default is @option{-fzero-initialized-in-bss}.
6192 @item -fmudflap -fmudflapth -fmudflapir
6196 @cindex bounds checking
6198 For front-ends that support it (C and C++), instrument all risky
6199 pointer/array dereferencing operations, some standard library
6200 string/heap functions, and some other associated constructs with
6201 range/validity tests. Modules so instrumented should be immune to
6202 buffer overflows, invalid heap use, and some other classes of C/C++
6203 programming errors. The instrumentation relies on a separate runtime
6204 library (@file{libmudflap}), which will be linked into a program if
6205 @option{-fmudflap} is given at link time. Run-time behavior of the
6206 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6207 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6210 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6211 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6212 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6213 instrumentation should ignore pointer reads. This produces less
6214 instrumentation (and therefore faster execution) and still provides
6215 some protection against outright memory corrupting writes, but allows
6216 erroneously read data to propagate within a program.
6218 @item -fthread-jumps
6219 @opindex fthread-jumps
6220 Perform optimizations where we check to see if a jump branches to a
6221 location where another comparison subsumed by the first is found. If
6222 so, the first branch is redirected to either the destination of the
6223 second branch or a point immediately following it, depending on whether
6224 the condition is known to be true or false.
6226 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6228 @item -fsplit-wide-types
6229 @opindex fsplit-wide-types
6230 When using a type that occupies multiple registers, such as @code{long
6231 long} on a 32-bit system, split the registers apart and allocate them
6232 independently. This normally generates better code for those types,
6233 but may make debugging more difficult.
6235 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6238 @item -fcse-follow-jumps
6239 @opindex fcse-follow-jumps
6240 In common subexpression elimination (CSE), scan through jump instructions
6241 when the target of the jump is not reached by any other path. For
6242 example, when CSE encounters an @code{if} statement with an
6243 @code{else} clause, CSE will follow the jump when the condition
6246 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6248 @item -fcse-skip-blocks
6249 @opindex fcse-skip-blocks
6250 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6251 follow jumps which conditionally skip over blocks. When CSE
6252 encounters a simple @code{if} statement with no else clause,
6253 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6254 body of the @code{if}.
6256 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6258 @item -frerun-cse-after-loop
6259 @opindex frerun-cse-after-loop
6260 Re-run common subexpression elimination after loop optimizations has been
6263 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6267 Perform a global common subexpression elimination pass.
6268 This pass also performs global constant and copy propagation.
6270 @emph{Note:} When compiling a program using computed gotos, a GCC
6271 extension, you may get better runtime performance if you disable
6272 the global common subexpression elimination pass by adding
6273 @option{-fno-gcse} to the command line.
6275 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6279 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6280 attempt to move loads which are only killed by stores into themselves. This
6281 allows a loop containing a load/store sequence to be changed to a load outside
6282 the loop, and a copy/store within the loop.
6284 Enabled by default when gcse is enabled.
6288 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6289 global common subexpression elimination. This pass will attempt to move
6290 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6291 loops containing a load/store sequence can be changed to a load before
6292 the loop and a store after the loop.
6294 Not enabled at any optimization level.
6298 When @option{-fgcse-las} is enabled, the global common subexpression
6299 elimination pass eliminates redundant loads that come after stores to the
6300 same memory location (both partial and full redundancies).
6302 Not enabled at any optimization level.
6304 @item -fgcse-after-reload
6305 @opindex fgcse-after-reload
6306 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6307 pass is performed after reload. The purpose of this pass is to cleanup
6310 @item -funsafe-loop-optimizations
6311 @opindex funsafe-loop-optimizations
6312 If given, the loop optimizer will assume that loop indices do not
6313 overflow, and that the loops with nontrivial exit condition are not
6314 infinite. This enables a wider range of loop optimizations even if
6315 the loop optimizer itself cannot prove that these assumptions are valid.
6316 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6317 if it finds this kind of loop.
6319 @item -fcrossjumping
6320 @opindex fcrossjumping
6321 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6322 resulting code may or may not perform better than without cross-jumping.
6324 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6326 @item -fauto-inc-dec
6327 @opindex fauto-inc-dec
6328 Combine increments or decrements of addresses with memory accesses.
6329 This pass is always skipped on architectures that do not have
6330 instructions to support this. Enabled by default at @option{-O} and
6331 higher on architectures that support this.
6335 Perform dead code elimination (DCE) on RTL@.
6336 Enabled by default at @option{-O} and higher.
6340 Perform dead store elimination (DSE) on RTL@.
6341 Enabled by default at @option{-O} and higher.
6343 @item -fif-conversion
6344 @opindex fif-conversion
6345 Attempt to transform conditional jumps into branch-less equivalents. This
6346 include use of conditional moves, min, max, set flags and abs instructions, and
6347 some tricks doable by standard arithmetics. The use of conditional execution
6348 on chips where it is available is controlled by @code{if-conversion2}.
6350 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6352 @item -fif-conversion2
6353 @opindex fif-conversion2
6354 Use conditional execution (where available) to transform conditional jumps into
6355 branch-less equivalents.
6357 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6359 @item -fdelete-null-pointer-checks
6360 @opindex fdelete-null-pointer-checks
6361 Assume that programs cannot safely dereference null pointers, and that
6362 no code or data element resides there. This enables simple constant
6363 folding optimizations at all optimization levels. In addition, other
6364 optimization passes in GCC use this flag to control global dataflow
6365 analyses that eliminate useless checks for null pointers; these assume
6366 that if a pointer is checked after it has already been dereferenced,
6369 Note however that in some environments this assumption is not true.
6370 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6371 for programs which depend on that behavior.
6373 Some targets, especially embedded ones, disable this option at all levels.
6374 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6375 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6376 are enabled independently at different optimization levels.
6378 @item -fexpensive-optimizations
6379 @opindex fexpensive-optimizations
6380 Perform a number of minor optimizations that are relatively expensive.
6382 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6384 @item -foptimize-register-move
6386 @opindex foptimize-register-move
6388 Attempt to reassign register numbers in move instructions and as
6389 operands of other simple instructions in order to maximize the amount of
6390 register tying. This is especially helpful on machines with two-operand
6393 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6396 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6398 @item -fira-algorithm=@var{algorithm}
6399 Use specified coloring algorithm for the integrated register
6400 allocator. The @var{algorithm} argument should be @code{priority} or
6401 @code{CB}. The first algorithm specifies Chow's priority coloring,
6402 the second one specifies Chaitin-Briggs coloring. The second
6403 algorithm can be unimplemented for some architectures. If it is
6404 implemented, it is the default because Chaitin-Briggs coloring as a
6405 rule generates a better code.
6407 @item -fira-region=@var{region}
6408 Use specified regions for the integrated register allocator. The
6409 @var{region} argument should be one of @code{all}, @code{mixed}, or
6410 @code{one}. The first value means using all loops as register
6411 allocation regions, the second value which is the default means using
6412 all loops except for loops with small register pressure as the
6413 regions, and third one means using all function as a single region.
6414 The first value can give best result for machines with small size and
6415 irregular register set, the third one results in faster and generates
6416 decent code and the smallest size code, and the default value usually
6417 give the best results in most cases and for most architectures.
6419 @item -fira-loop-pressure
6420 @opindex fira-loop-pressure
6421 Use IRA to evaluate register pressure in loops for decision to move
6422 loop invariants. Usage of this option usually results in generation
6423 of faster and smaller code on machines with big register files (>= 32
6424 registers) but it can slow compiler down.
6426 This option is enabled at level @option{-O3} for some targets.
6428 @item -fno-ira-share-save-slots
6429 @opindex fno-ira-share-save-slots
6430 Switch off sharing stack slots used for saving call used hard
6431 registers living through a call. Each hard register will get a
6432 separate stack slot and as a result function stack frame will be
6435 @item -fno-ira-share-spill-slots
6436 @opindex fno-ira-share-spill-slots
6437 Switch off sharing stack slots allocated for pseudo-registers. Each
6438 pseudo-register which did not get a hard register will get a separate
6439 stack slot and as a result function stack frame will be bigger.
6441 @item -fira-verbose=@var{n}
6442 @opindex fira-verbose
6443 Set up how verbose dump file for the integrated register allocator
6444 will be. Default value is 5. If the value is greater or equal to 10,
6445 the dump file will be stderr as if the value were @var{n} minus 10.
6447 @item -fdelayed-branch
6448 @opindex fdelayed-branch
6449 If supported for the target machine, attempt to reorder instructions
6450 to exploit instruction slots available after delayed branch
6453 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6455 @item -fschedule-insns
6456 @opindex fschedule-insns
6457 If supported for the target machine, attempt to reorder instructions to
6458 eliminate execution stalls due to required data being unavailable. This
6459 helps machines that have slow floating point or memory load instructions
6460 by allowing other instructions to be issued until the result of the load
6461 or floating point instruction is required.
6463 Enabled at levels @option{-O2}, @option{-O3}.
6465 @item -fschedule-insns2
6466 @opindex fschedule-insns2
6467 Similar to @option{-fschedule-insns}, but requests an additional pass of
6468 instruction scheduling after register allocation has been done. This is
6469 especially useful on machines with a relatively small number of
6470 registers and where memory load instructions take more than one cycle.
6472 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6474 @item -fno-sched-interblock
6475 @opindex fno-sched-interblock
6476 Don't schedule instructions across basic blocks. This is normally
6477 enabled by default when scheduling before register allocation, i.e.@:
6478 with @option{-fschedule-insns} or at @option{-O2} or higher.
6480 @item -fno-sched-spec
6481 @opindex fno-sched-spec
6482 Don't allow speculative motion of non-load instructions. This is normally
6483 enabled by default when scheduling before register allocation, i.e.@:
6484 with @option{-fschedule-insns} or at @option{-O2} or higher.
6486 @item -fsched-pressure
6487 @opindex fsched-pressure
6488 Enable register pressure sensitive insn scheduling before the register
6489 allocation. This only makes sense when scheduling before register
6490 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6491 @option{-O2} or higher. Usage of this option can improve the
6492 generated code and decrease its size by preventing register pressure
6493 increase above the number of available hard registers and as a
6494 consequence register spills in the register allocation.
6496 @item -fsched-spec-load
6497 @opindex fsched-spec-load
6498 Allow speculative motion of some load instructions. This only makes
6499 sense when scheduling before register allocation, i.e.@: with
6500 @option{-fschedule-insns} or at @option{-O2} or higher.
6502 @item -fsched-spec-load-dangerous
6503 @opindex fsched-spec-load-dangerous
6504 Allow speculative motion of more load instructions. This only makes
6505 sense when scheduling before register allocation, i.e.@: with
6506 @option{-fschedule-insns} or at @option{-O2} or higher.
6508 @item -fsched-stalled-insns
6509 @itemx -fsched-stalled-insns=@var{n}
6510 @opindex fsched-stalled-insns
6511 Define how many insns (if any) can be moved prematurely from the queue
6512 of stalled insns into the ready list, during the second scheduling pass.
6513 @option{-fno-sched-stalled-insns} means that no insns will be moved
6514 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6515 on how many queued insns can be moved prematurely.
6516 @option{-fsched-stalled-insns} without a value is equivalent to
6517 @option{-fsched-stalled-insns=1}.
6519 @item -fsched-stalled-insns-dep
6520 @itemx -fsched-stalled-insns-dep=@var{n}
6521 @opindex fsched-stalled-insns-dep
6522 Define how many insn groups (cycles) will be examined for a dependency
6523 on a stalled insn that is candidate for premature removal from the queue
6524 of stalled insns. This has an effect only during the second scheduling pass,
6525 and only if @option{-fsched-stalled-insns} is used.
6526 @option{-fno-sched-stalled-insns-dep} is equivalent to
6527 @option{-fsched-stalled-insns-dep=0}.
6528 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6529 @option{-fsched-stalled-insns-dep=1}.
6531 @item -fsched2-use-superblocks
6532 @opindex fsched2-use-superblocks
6533 When scheduling after register allocation, do use superblock scheduling
6534 algorithm. Superblock scheduling allows motion across basic block boundaries
6535 resulting on faster schedules. This option is experimental, as not all machine
6536 descriptions used by GCC model the CPU closely enough to avoid unreliable
6537 results from the algorithm.
6539 This only makes sense when scheduling after register allocation, i.e.@: with
6540 @option{-fschedule-insns2} or at @option{-O2} or higher.
6542 @item -fsched-group-heuristic
6543 @opindex fsched-group-heuristic
6544 Enable the group heuristic in the scheduler. This heuristic favors
6545 the instruction that belongs to a schedule group. This is enabled
6546 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6547 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6549 @item -fsched-critical-path-heuristic
6550 @opindex fsched-critical-path-heuristic
6551 Enable the critical-path heuristic in the scheduler. This heuristic favors
6552 instructions on the critical path. This is enabled by default when
6553 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6554 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6556 @item -fsched-spec-insn-heuristic
6557 @opindex fsched-spec-insn-heuristic
6558 Enable the speculative instruction heuristic in the scheduler. This
6559 heuristic favors speculative instructions with greater dependency weakness.
6560 This is enabled by default when scheduling is enabled, i.e.@:
6561 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6562 or at @option{-O2} or higher.
6564 @item -fsched-rank-heuristic
6565 @opindex fsched-rank-heuristic
6566 Enable the rank heuristic in the scheduler. This heuristic favors
6567 the instruction belonging to a basic block with greater size or frequency.
6568 This is enabled by default when scheduling is enabled, i.e.@:
6569 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6570 at @option{-O2} or higher.
6572 @item -fsched-last-insn-heuristic
6573 @opindex fsched-last-insn-heuristic
6574 Enable the last-instruction heuristic in the scheduler. This heuristic
6575 favors the instruction that is less dependent on the last instruction
6576 scheduled. This is enabled by default when scheduling is enabled,
6577 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6578 at @option{-O2} or higher.
6580 @item -fsched-dep-count-heuristic
6581 @opindex fsched-dep-count-heuristic
6582 Enable the dependent-count heuristic in the scheduler. This heuristic
6583 favors the instruction that has more instructions depending on it.
6584 This is enabled by default when scheduling is enabled, i.e.@:
6585 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6586 at @option{-O2} or higher.
6588 @item -freschedule-modulo-scheduled-loops
6589 @opindex freschedule-modulo-scheduled-loops
6590 The modulo scheduling comes before the traditional scheduling, if a loop
6591 was modulo scheduled we may want to prevent the later scheduling passes
6592 from changing its schedule, we use this option to control that.
6594 @item -fselective-scheduling
6595 @opindex fselective-scheduling
6596 Schedule instructions using selective scheduling algorithm. Selective
6597 scheduling runs instead of the first scheduler pass.
6599 @item -fselective-scheduling2
6600 @opindex fselective-scheduling2
6601 Schedule instructions using selective scheduling algorithm. Selective
6602 scheduling runs instead of the second scheduler pass.
6604 @item -fsel-sched-pipelining
6605 @opindex fsel-sched-pipelining
6606 Enable software pipelining of innermost loops during selective scheduling.
6607 This option has no effect until one of @option{-fselective-scheduling} or
6608 @option{-fselective-scheduling2} is turned on.
6610 @item -fsel-sched-pipelining-outer-loops
6611 @opindex fsel-sched-pipelining-outer-loops
6612 When pipelining loops during selective scheduling, also pipeline outer loops.
6613 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6615 @item -fcaller-saves
6616 @opindex fcaller-saves
6617 Enable values to be allocated in registers that will be clobbered by
6618 function calls, by emitting extra instructions to save and restore the
6619 registers around such calls. Such allocation is done only when it
6620 seems to result in better code than would otherwise be produced.
6622 This option is always enabled by default on certain machines, usually
6623 those which have no call-preserved registers to use instead.
6625 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6627 @item -fcombine-stack-adjustments
6628 @opindex fcombine-stack-adjustments
6629 Tracks stack adjustments (pushes and pops) and stack memory references
6630 and then tries to find ways to combine them.
6632 Enabled by default at @option{-O1} and higher.
6634 @item -fconserve-stack
6635 @opindex fconserve-stack
6636 Attempt to minimize stack usage. The compiler will attempt to use less
6637 stack space, even if that makes the program slower. This option
6638 implies setting the @option{large-stack-frame} parameter to 100
6639 and the @option{large-stack-frame-growth} parameter to 400.
6641 @item -ftree-reassoc
6642 @opindex ftree-reassoc
6643 Perform reassociation on trees. This flag is enabled by default
6644 at @option{-O} and higher.
6648 Perform partial redundancy elimination (PRE) on trees. This flag is
6649 enabled by default at @option{-O2} and @option{-O3}.
6651 @item -ftree-forwprop
6652 @opindex ftree-forwprop
6653 Perform forward propagation on trees. This flag is enabled by default
6654 at @option{-O} and higher.
6658 Perform full redundancy elimination (FRE) on trees. The difference
6659 between FRE and PRE is that FRE only considers expressions
6660 that are computed on all paths leading to the redundant computation.
6661 This analysis is faster than PRE, though it exposes fewer redundancies.
6662 This flag is enabled by default at @option{-O} and higher.
6664 @item -ftree-phiprop
6665 @opindex ftree-phiprop
6666 Perform hoisting of loads from conditional pointers on trees. This
6667 pass is enabled by default at @option{-O} and higher.
6669 @item -ftree-copy-prop
6670 @opindex ftree-copy-prop
6671 Perform copy propagation on trees. This pass eliminates unnecessary
6672 copy operations. This flag is enabled by default at @option{-O} and
6675 @item -fipa-pure-const
6676 @opindex fipa-pure-const
6677 Discover which functions are pure or constant.
6678 Enabled by default at @option{-O} and higher.
6680 @item -fipa-reference
6681 @opindex fipa-reference
6682 Discover which static variables do not escape cannot escape the
6684 Enabled by default at @option{-O} and higher.
6686 @item -fipa-struct-reorg
6687 @opindex fipa-struct-reorg
6688 Perform structure reorganization optimization, that change C-like structures
6689 layout in order to better utilize spatial locality. This transformation is
6690 affective for programs containing arrays of structures. Available in two
6691 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6692 or static (which uses built-in heuristics). It works only in whole program
6693 mode, so it requires @option{-fwhole-program} to be
6694 enabled. Structures considered @samp{cold} by this transformation are not
6695 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6697 With this flag, the program debug info reflects a new structure layout.
6701 Perform interprocedural pointer analysis and interprocedural modification
6702 and reference analysis. This option can cause excessive memory and
6703 compile-time usage on large compilation units. It is not enabled by
6704 default at any optimization level.
6707 @opindex fipa-profile
6708 Perform interprocedural profile propagation. The functions called only from
6709 cold functions are marked as cold. Also functions executed once (such as
6710 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6711 functions and loop less parts of functions executed once are then optimized for
6713 Enabled by default at @option{-O} and higher.
6717 Perform interprocedural constant propagation.
6718 This optimization analyzes the program to determine when values passed
6719 to functions are constants and then optimizes accordingly.
6720 This optimization can substantially increase performance
6721 if the application has constants passed to functions.
6722 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6724 @item -fipa-cp-clone
6725 @opindex fipa-cp-clone
6726 Perform function cloning to make interprocedural constant propagation stronger.
6727 When enabled, interprocedural constant propagation will perform function cloning
6728 when externally visible function can be called with constant arguments.
6729 Because this optimization can create multiple copies of functions,
6730 it may significantly increase code size
6731 (see @option{--param ipcp-unit-growth=@var{value}}).
6732 This flag is enabled by default at @option{-O3}.
6734 @item -fipa-matrix-reorg
6735 @opindex fipa-matrix-reorg
6736 Perform matrix flattening and transposing.
6737 Matrix flattening tries to replace an @math{m}-dimensional matrix
6738 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6739 This reduces the level of indirection needed for accessing the elements
6740 of the matrix. The second optimization is matrix transposing that
6741 attempts to change the order of the matrix's dimensions in order to
6742 improve cache locality.
6743 Both optimizations need the @option{-fwhole-program} flag.
6744 Transposing is enabled only if profiling information is available.
6748 Perform forward store motion on trees. This flag is
6749 enabled by default at @option{-O} and higher.
6751 @item -ftree-bit-ccp
6752 @opindex ftree-bit-ccp
6753 Perform sparse conditional bit constant propagation on trees and propagate
6754 pointer alignment information.
6755 This pass only operates on local scalar variables and is enabled by default
6756 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6760 Perform sparse conditional constant propagation (CCP) on trees. This
6761 pass only operates on local scalar variables and is enabled by default
6762 at @option{-O} and higher.
6764 @item -ftree-switch-conversion
6765 Perform conversion of simple initializations in a switch to
6766 initializations from a scalar array. This flag is enabled by default
6767 at @option{-O2} and higher.
6771 Perform dead code elimination (DCE) on trees. This flag is enabled by
6772 default at @option{-O} and higher.
6774 @item -ftree-builtin-call-dce
6775 @opindex ftree-builtin-call-dce
6776 Perform conditional dead code elimination (DCE) for calls to builtin functions
6777 that may set @code{errno} but are otherwise side-effect free. This flag is
6778 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6781 @item -ftree-dominator-opts
6782 @opindex ftree-dominator-opts
6783 Perform a variety of simple scalar cleanups (constant/copy
6784 propagation, redundancy elimination, range propagation and expression
6785 simplification) based on a dominator tree traversal. This also
6786 performs jump threading (to reduce jumps to jumps). This flag is
6787 enabled by default at @option{-O} and higher.
6791 Perform dead store elimination (DSE) on trees. A dead store is a store into
6792 a memory location which will later be overwritten by another store without
6793 any intervening loads. In this case the earlier store can be deleted. This
6794 flag is enabled by default at @option{-O} and higher.
6798 Perform loop header copying on trees. This is beneficial since it increases
6799 effectiveness of code motion optimizations. It also saves one jump. This flag
6800 is enabled by default at @option{-O} and higher. It is not enabled
6801 for @option{-Os}, since it usually increases code size.
6803 @item -ftree-loop-optimize
6804 @opindex ftree-loop-optimize
6805 Perform loop optimizations on trees. This flag is enabled by default
6806 at @option{-O} and higher.
6808 @item -ftree-loop-linear
6809 @opindex ftree-loop-linear
6810 Perform linear loop transformations on tree. This flag can improve cache
6811 performance and allow further loop optimizations to take place.
6813 @item -floop-interchange
6814 @opindex floop-interchange
6815 Perform loop interchange transformations on loops. Interchanging two
6816 nested loops switches the inner and outer loops. For example, given a
6821 A(J, I) = A(J, I) * C
6825 loop interchange will transform the loop as if the user had written:
6829 A(J, I) = A(J, I) * C
6833 which can be beneficial when @code{N} is larger than the caches,
6834 because in Fortran, the elements of an array are stored in memory
6835 contiguously by column, and the original loop iterates over rows,
6836 potentially creating at each access a cache miss. This optimization
6837 applies to all the languages supported by GCC and is not limited to
6838 Fortran. To use this code transformation, GCC has to be configured
6839 with @option{--with-ppl} and @option{--with-cloog} to enable the
6840 Graphite loop transformation infrastructure.
6842 @item -floop-strip-mine
6843 @opindex floop-strip-mine
6844 Perform loop strip mining transformations on loops. Strip mining
6845 splits a loop into two nested loops. The outer loop has strides
6846 equal to the strip size and the inner loop has strides of the
6847 original loop within a strip. The strip length can be changed
6848 using the @option{loop-block-tile-size} parameter. For example,
6855 loop strip mining will transform the loop as if the user had written:
6858 DO I = II, min (II + 50, N)
6863 This optimization applies to all the languages supported by GCC and is
6864 not limited to Fortran. To use this code transformation, GCC has to
6865 be configured with @option{--with-ppl} and @option{--with-cloog} to
6866 enable the Graphite loop transformation infrastructure.
6869 @opindex floop-block
6870 Perform loop blocking transformations on loops. Blocking strip mines
6871 each loop in the loop nest such that the memory accesses of the
6872 element loops fit inside caches. The strip length can be changed
6873 using the @option{loop-block-tile-size} parameter. For example, given
6878 A(J, I) = B(I) + C(J)
6882 loop blocking will transform the loop as if the user had written:
6886 DO I = II, min (II + 50, N)
6887 DO J = JJ, min (JJ + 50, M)
6888 A(J, I) = B(I) + C(J)
6894 which can be beneficial when @code{M} is larger than the caches,
6895 because the innermost loop will iterate over a smaller amount of data
6896 that can be kept in the caches. This optimization applies to all the
6897 languages supported by GCC and is not limited to Fortran. To use this
6898 code transformation, GCC has to be configured with @option{--with-ppl}
6899 and @option{--with-cloog} to enable the Graphite loop transformation
6902 @item -fgraphite-identity
6903 @opindex fgraphite-identity
6904 Enable the identity transformation for graphite. For every SCoP we generate
6905 the polyhedral representation and transform it back to gimple. Using
6906 @option{-fgraphite-identity} we can check the costs or benefits of the
6907 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6908 are also performed by the code generator CLooG, like index splitting and
6909 dead code elimination in loops.
6911 @item -floop-flatten
6912 @opindex floop-flatten
6913 Removes the loop nesting structure: transforms the loop nest into a
6914 single loop. This transformation can be useful to vectorize all the
6915 levels of the loop nest.
6917 @item -floop-parallelize-all
6918 @opindex floop-parallelize-all
6919 Use the Graphite data dependence analysis to identify loops that can
6920 be parallelized. Parallelize all the loops that can be analyzed to
6921 not contain loop carried dependences without checking that it is
6922 profitable to parallelize the loops.
6924 @item -fcheck-data-deps
6925 @opindex fcheck-data-deps
6926 Compare the results of several data dependence analyzers. This option
6927 is used for debugging the data dependence analyzers.
6929 @item -ftree-loop-if-convert
6930 Attempt to transform conditional jumps in the innermost loops to
6931 branch-less equivalents. The intent is to remove control-flow from
6932 the innermost loops in order to improve the ability of the
6933 vectorization pass to handle these loops. This is enabled by default
6934 if vectorization is enabled.
6936 @item -ftree-loop-if-convert-stores
6937 Attempt to also if-convert conditional jumps containing memory writes.
6938 This transformation can be unsafe for multi-threaded programs as it
6939 transforms conditional memory writes into unconditional memory writes.
6942 for (i = 0; i < N; i++)
6946 would be transformed to
6948 for (i = 0; i < N; i++)
6949 A[i] = cond ? expr : A[i];
6951 potentially producing data races.
6953 @item -ftree-loop-distribution
6954 Perform loop distribution. This flag can improve cache performance on
6955 big loop bodies and allow further loop optimizations, like
6956 parallelization or vectorization, to take place. For example, the loop
6973 @item -ftree-loop-distribute-patterns
6974 Perform loop distribution of patterns that can be code generated with
6975 calls to a library. This flag is enabled by default at @option{-O3}.
6977 This pass distributes the initialization loops and generates a call to
6978 memset zero. For example, the loop
6994 and the initialization loop is transformed into a call to memset zero.
6996 @item -ftree-loop-im
6997 @opindex ftree-loop-im
6998 Perform loop invariant motion on trees. This pass moves only invariants that
6999 would be hard to handle at RTL level (function calls, operations that expand to
7000 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7001 operands of conditions that are invariant out of the loop, so that we can use
7002 just trivial invariantness analysis in loop unswitching. The pass also includes
7005 @item -ftree-loop-ivcanon
7006 @opindex ftree-loop-ivcanon
7007 Create a canonical counter for number of iterations in the loop for that
7008 determining number of iterations requires complicated analysis. Later
7009 optimizations then may determine the number easily. Useful especially
7010 in connection with unrolling.
7014 Perform induction variable optimizations (strength reduction, induction
7015 variable merging and induction variable elimination) on trees.
7017 @item -ftree-parallelize-loops=n
7018 @opindex ftree-parallelize-loops
7019 Parallelize loops, i.e., split their iteration space to run in n threads.
7020 This is only possible for loops whose iterations are independent
7021 and can be arbitrarily reordered. The optimization is only
7022 profitable on multiprocessor machines, for loops that are CPU-intensive,
7023 rather than constrained e.g.@: by memory bandwidth. This option
7024 implies @option{-pthread}, and thus is only supported on targets
7025 that have support for @option{-pthread}.
7029 Perform function-local points-to analysis on trees. This flag is
7030 enabled by default at @option{-O} and higher.
7034 Perform scalar replacement of aggregates. This pass replaces structure
7035 references with scalars to prevent committing structures to memory too
7036 early. This flag is enabled by default at @option{-O} and higher.
7038 @item -ftree-copyrename
7039 @opindex ftree-copyrename
7040 Perform copy renaming on trees. This pass attempts to rename compiler
7041 temporaries to other variables at copy locations, usually resulting in
7042 variable names which more closely resemble the original variables. This flag
7043 is enabled by default at @option{-O} and higher.
7047 Perform temporary expression replacement during the SSA->normal phase. Single
7048 use/single def temporaries are replaced at their use location with their
7049 defining expression. This results in non-GIMPLE code, but gives the expanders
7050 much more complex trees to work on resulting in better RTL generation. This is
7051 enabled by default at @option{-O} and higher.
7053 @item -ftree-vectorize
7054 @opindex ftree-vectorize
7055 Perform loop vectorization on trees. This flag is enabled by default at
7058 @item -ftree-slp-vectorize
7059 @opindex ftree-slp-vectorize
7060 Perform basic block vectorization on trees. This flag is enabled by default at
7061 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7063 @item -ftree-vect-loop-version
7064 @opindex ftree-vect-loop-version
7065 Perform loop versioning when doing loop vectorization on trees. When a loop
7066 appears to be vectorizable except that data alignment or data dependence cannot
7067 be determined at compile time then vectorized and non-vectorized versions of
7068 the loop are generated along with runtime checks for alignment or dependence
7069 to control which version is executed. This option is enabled by default
7070 except at level @option{-Os} where it is disabled.
7072 @item -fvect-cost-model
7073 @opindex fvect-cost-model
7074 Enable cost model for vectorization.
7078 Perform Value Range Propagation on trees. This is similar to the
7079 constant propagation pass, but instead of values, ranges of values are
7080 propagated. This allows the optimizers to remove unnecessary range
7081 checks like array bound checks and null pointer checks. This is
7082 enabled by default at @option{-O2} and higher. Null pointer check
7083 elimination is only done if @option{-fdelete-null-pointer-checks} is
7088 Perform tail duplication to enlarge superblock size. This transformation
7089 simplifies the control flow of the function allowing other optimizations to do
7092 @item -funroll-loops
7093 @opindex funroll-loops
7094 Unroll loops whose number of iterations can be determined at compile
7095 time or upon entry to the loop. @option{-funroll-loops} implies
7096 @option{-frerun-cse-after-loop}. This option makes code larger,
7097 and may or may not make it run faster.
7099 @item -funroll-all-loops
7100 @opindex funroll-all-loops
7101 Unroll all loops, even if their number of iterations is uncertain when
7102 the loop is entered. This usually makes programs run more slowly.
7103 @option{-funroll-all-loops} implies the same options as
7104 @option{-funroll-loops},
7106 @item -fsplit-ivs-in-unroller
7107 @opindex fsplit-ivs-in-unroller
7108 Enables expressing of values of induction variables in later iterations
7109 of the unrolled loop using the value in the first iteration. This breaks
7110 long dependency chains, thus improving efficiency of the scheduling passes.
7112 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7113 same effect. However in cases the loop body is more complicated than
7114 a single basic block, this is not reliable. It also does not work at all
7115 on some of the architectures due to restrictions in the CSE pass.
7117 This optimization is enabled by default.
7119 @item -fvariable-expansion-in-unroller
7120 @opindex fvariable-expansion-in-unroller
7121 With this option, the compiler will create multiple copies of some
7122 local variables when unrolling a loop which can result in superior code.
7124 @item -fpartial-inlining
7125 @opindex fpartial-inlining
7126 Inline parts of functions. This option has any effect only
7127 when inlining itself is turned on by the @option{-finline-functions}
7128 or @option{-finline-small-functions} options.
7130 Enabled at level @option{-O2}.
7132 @item -fpredictive-commoning
7133 @opindex fpredictive-commoning
7134 Perform predictive commoning optimization, i.e., reusing computations
7135 (especially memory loads and stores) performed in previous
7136 iterations of loops.
7138 This option is enabled at level @option{-O3}.
7140 @item -fprefetch-loop-arrays
7141 @opindex fprefetch-loop-arrays
7142 If supported by the target machine, generate instructions to prefetch
7143 memory to improve the performance of loops that access large arrays.
7145 This option may generate better or worse code; results are highly
7146 dependent on the structure of loops within the source code.
7148 Disabled at level @option{-Os}.
7151 @itemx -fno-peephole2
7152 @opindex fno-peephole
7153 @opindex fno-peephole2
7154 Disable any machine-specific peephole optimizations. The difference
7155 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7156 are implemented in the compiler; some targets use one, some use the
7157 other, a few use both.
7159 @option{-fpeephole} is enabled by default.
7160 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7162 @item -fno-guess-branch-probability
7163 @opindex fno-guess-branch-probability
7164 Do not guess branch probabilities using heuristics.
7166 GCC will use heuristics to guess branch probabilities if they are
7167 not provided by profiling feedback (@option{-fprofile-arcs}). These
7168 heuristics are based on the control flow graph. If some branch probabilities
7169 are specified by @samp{__builtin_expect}, then the heuristics will be
7170 used to guess branch probabilities for the rest of the control flow graph,
7171 taking the @samp{__builtin_expect} info into account. The interactions
7172 between the heuristics and @samp{__builtin_expect} can be complex, and in
7173 some cases, it may be useful to disable the heuristics so that the effects
7174 of @samp{__builtin_expect} are easier to understand.
7176 The default is @option{-fguess-branch-probability} at levels
7177 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7179 @item -freorder-blocks
7180 @opindex freorder-blocks
7181 Reorder basic blocks in the compiled function in order to reduce number of
7182 taken branches and improve code locality.
7184 Enabled at levels @option{-O2}, @option{-O3}.
7186 @item -freorder-blocks-and-partition
7187 @opindex freorder-blocks-and-partition
7188 In addition to reordering basic blocks in the compiled function, in order
7189 to reduce number of taken branches, partitions hot and cold basic blocks
7190 into separate sections of the assembly and .o files, to improve
7191 paging and cache locality performance.
7193 This optimization is automatically turned off in the presence of
7194 exception handling, for linkonce sections, for functions with a user-defined
7195 section attribute and on any architecture that does not support named
7198 @item -freorder-functions
7199 @opindex freorder-functions
7200 Reorder functions in the object file in order to
7201 improve code locality. This is implemented by using special
7202 subsections @code{.text.hot} for most frequently executed functions and
7203 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7204 the linker so object file format must support named sections and linker must
7205 place them in a reasonable way.
7207 Also profile feedback must be available in to make this option effective. See
7208 @option{-fprofile-arcs} for details.
7210 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7212 @item -fstrict-aliasing
7213 @opindex fstrict-aliasing
7214 Allow the compiler to assume the strictest aliasing rules applicable to
7215 the language being compiled. For C (and C++), this activates
7216 optimizations based on the type of expressions. In particular, an
7217 object of one type is assumed never to reside at the same address as an
7218 object of a different type, unless the types are almost the same. For
7219 example, an @code{unsigned int} can alias an @code{int}, but not a
7220 @code{void*} or a @code{double}. A character type may alias any other
7223 @anchor{Type-punning}Pay special attention to code like this:
7236 The practice of reading from a different union member than the one most
7237 recently written to (called ``type-punning'') is common. Even with
7238 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7239 is accessed through the union type. So, the code above will work as
7240 expected. @xref{Structures unions enumerations and bit-fields
7241 implementation}. However, this code might not:
7252 Similarly, access by taking the address, casting the resulting pointer
7253 and dereferencing the result has undefined behavior, even if the cast
7254 uses a union type, e.g.:
7258 return ((union a_union *) &d)->i;
7262 The @option{-fstrict-aliasing} option is enabled at levels
7263 @option{-O2}, @option{-O3}, @option{-Os}.
7265 @item -fstrict-overflow
7266 @opindex fstrict-overflow
7267 Allow the compiler to assume strict signed overflow rules, depending
7268 on the language being compiled. For C (and C++) this means that
7269 overflow when doing arithmetic with signed numbers is undefined, which
7270 means that the compiler may assume that it will not happen. This
7271 permits various optimizations. For example, the compiler will assume
7272 that an expression like @code{i + 10 > i} will always be true for
7273 signed @code{i}. This assumption is only valid if signed overflow is
7274 undefined, as the expression is false if @code{i + 10} overflows when
7275 using twos complement arithmetic. When this option is in effect any
7276 attempt to determine whether an operation on signed numbers will
7277 overflow must be written carefully to not actually involve overflow.
7279 This option also allows the compiler to assume strict pointer
7280 semantics: given a pointer to an object, if adding an offset to that
7281 pointer does not produce a pointer to the same object, the addition is
7282 undefined. This permits the compiler to conclude that @code{p + u >
7283 p} is always true for a pointer @code{p} and unsigned integer
7284 @code{u}. This assumption is only valid because pointer wraparound is
7285 undefined, as the expression is false if @code{p + u} overflows using
7286 twos complement arithmetic.
7288 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7289 that integer signed overflow is fully defined: it wraps. When
7290 @option{-fwrapv} is used, there is no difference between
7291 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7292 integers. With @option{-fwrapv} certain types of overflow are
7293 permitted. For example, if the compiler gets an overflow when doing
7294 arithmetic on constants, the overflowed value can still be used with
7295 @option{-fwrapv}, but not otherwise.
7297 The @option{-fstrict-overflow} option is enabled at levels
7298 @option{-O2}, @option{-O3}, @option{-Os}.
7300 @item -falign-functions
7301 @itemx -falign-functions=@var{n}
7302 @opindex falign-functions
7303 Align the start of functions to the next power-of-two greater than
7304 @var{n}, skipping up to @var{n} bytes. For instance,
7305 @option{-falign-functions=32} aligns functions to the next 32-byte
7306 boundary, but @option{-falign-functions=24} would align to the next
7307 32-byte boundary only if this can be done by skipping 23 bytes or less.
7309 @option{-fno-align-functions} and @option{-falign-functions=1} are
7310 equivalent and mean that functions will not be aligned.
7312 Some assemblers only support this flag when @var{n} is a power of two;
7313 in that case, it is rounded up.
7315 If @var{n} is not specified or is zero, use a machine-dependent default.
7317 Enabled at levels @option{-O2}, @option{-O3}.
7319 @item -falign-labels
7320 @itemx -falign-labels=@var{n}
7321 @opindex falign-labels
7322 Align all branch targets to a power-of-two boundary, skipping up to
7323 @var{n} bytes like @option{-falign-functions}. This option can easily
7324 make code slower, because it must insert dummy operations for when the
7325 branch target is reached in the usual flow of the code.
7327 @option{-fno-align-labels} and @option{-falign-labels=1} are
7328 equivalent and mean that labels will not be aligned.
7330 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7331 are greater than this value, then their values are used instead.
7333 If @var{n} is not specified or is zero, use a machine-dependent default
7334 which is very likely to be @samp{1}, meaning no alignment.
7336 Enabled at levels @option{-O2}, @option{-O3}.
7339 @itemx -falign-loops=@var{n}
7340 @opindex falign-loops
7341 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7342 like @option{-falign-functions}. The hope is that the loop will be
7343 executed many times, which will make up for any execution of the dummy
7346 @option{-fno-align-loops} and @option{-falign-loops=1} are
7347 equivalent and mean that loops will not be aligned.
7349 If @var{n} is not specified or is zero, use a machine-dependent default.
7351 Enabled at levels @option{-O2}, @option{-O3}.
7354 @itemx -falign-jumps=@var{n}
7355 @opindex falign-jumps
7356 Align branch targets to a power-of-two boundary, for branch targets
7357 where the targets can only be reached by jumping, skipping up to @var{n}
7358 bytes like @option{-falign-functions}. In this case, no dummy operations
7361 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7362 equivalent and mean that loops will not be aligned.
7364 If @var{n} is not specified or is zero, use a machine-dependent default.
7366 Enabled at levels @option{-O2}, @option{-O3}.
7368 @item -funit-at-a-time
7369 @opindex funit-at-a-time
7370 This option is left for compatibility reasons. @option{-funit-at-a-time}
7371 has no effect, while @option{-fno-unit-at-a-time} implies
7372 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7376 @item -fno-toplevel-reorder
7377 @opindex fno-toplevel-reorder
7378 Do not reorder top-level functions, variables, and @code{asm}
7379 statements. Output them in the same order that they appear in the
7380 input file. When this option is used, unreferenced static variables
7381 will not be removed. This option is intended to support existing code
7382 which relies on a particular ordering. For new code, it is better to
7385 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7386 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7391 Constructs webs as commonly used for register allocation purposes and assign
7392 each web individual pseudo register. This allows the register allocation pass
7393 to operate on pseudos directly, but also strengthens several other optimization
7394 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7395 however, make debugging impossible, since variables will no longer stay in a
7398 Enabled by default with @option{-funroll-loops}.
7400 @item -fwhole-program
7401 @opindex fwhole-program
7402 Assume that the current compilation unit represents the whole program being
7403 compiled. All public functions and variables with the exception of @code{main}
7404 and those merged by attribute @code{externally_visible} become static functions
7405 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7406 While this option is equivalent to proper use of the @code{static} keyword for
7407 programs consisting of a single file, in combination with option
7408 @option{-flto} or @option{-fwhopr} this flag can be used to
7409 compile many smaller scale programs since the functions and variables become
7410 local for the whole combined compilation unit, not for the single source file
7413 This option implies @option{-fwhole-file} for Fortran programs.
7417 This option runs the standard link-time optimizer. When invoked
7418 with source code, it generates GIMPLE (one of GCC's internal
7419 representations) and writes it to special ELF sections in the object
7420 file. When the object files are linked together, all the function
7421 bodies are read from these ELF sections and instantiated as if they
7422 had been part of the same translation unit.
7424 To use the link-timer optimizer, @option{-flto} needs to be specified at
7425 compile time and during the final link. For example,
7428 gcc -c -O2 -flto foo.c
7429 gcc -c -O2 -flto bar.c
7430 gcc -o myprog -flto -O2 foo.o bar.o
7433 The first two invocations to GCC will save a bytecode representation
7434 of GIMPLE into special ELF sections inside @file{foo.o} and
7435 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7436 @file{foo.o} and @file{bar.o}, merge the two files into a single
7437 internal image, and compile the result as usual. Since both
7438 @file{foo.o} and @file{bar.o} are merged into a single image, this
7439 causes all the inter-procedural analyses and optimizations in GCC to
7440 work across the two files as if they were a single one. This means,
7441 for example, that the inliner will be able to inline functions in
7442 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7444 Another (simpler) way to enable link-time optimization is,
7447 gcc -o myprog -flto -O2 foo.c bar.c
7450 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7451 merge them together into a single GIMPLE representation and optimize
7452 them as usual to produce @file{myprog}.
7454 The only important thing to keep in mind is that to enable link-time
7455 optimizations the @option{-flto} flag needs to be passed to both the
7456 compile and the link commands.
7458 Note that when a file is compiled with @option{-flto}, the generated
7459 object file will be larger than a regular object file because it will
7460 contain GIMPLE bytecodes and the usual final code. This means that
7461 object files with LTO information can be linked as a normal object
7462 file. So, in the previous example, if the final link is done with
7465 gcc -o myprog foo.o bar.o
7468 The only difference will be that no inter-procedural optimizations
7469 will be applied to produce @file{myprog}. The two object files
7470 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7473 Additionally, the optimization flags used to compile individual files
7474 are not necessarily related to those used at link-time. For instance,
7477 gcc -c -O0 -flto foo.c
7478 gcc -c -O0 -flto bar.c
7479 gcc -o myprog -flto -O3 foo.o bar.o
7482 This will produce individual object files with unoptimized assembler
7483 code, but the resulting binary @file{myprog} will be optimized at
7484 @option{-O3}. Now, if the final binary is generated without
7485 @option{-flto}, then @file{myprog} will not be optimized.
7487 When producing the final binary with @option{-flto}, GCC will only
7488 apply link-time optimizations to those files that contain bytecode.
7489 Therefore, you can mix and match object files and libraries with
7490 GIMPLE bytecodes and final object code. GCC will automatically select
7491 which files to optimize in LTO mode and which files to link without
7494 There are some code generation flags that GCC will preserve when
7495 generating bytecodes, as they need to be used during the final link
7496 stage. Currently, the following options are saved into the GIMPLE
7497 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7498 @option{-m} target flags.
7500 At link time, these options are read-in and reapplied. Note that the
7501 current implementation makes no attempt at recognizing conflicting
7502 values for these options. If two or more files have a conflicting
7503 value (e.g., one file is compiled with @option{-fPIC} and another
7504 isn't), the compiler will simply use the last value read from the
7505 bytecode files. It is recommended, then, that all the files
7506 participating in the same link be compiled with the same options.
7508 Another feature of LTO is that it is possible to apply interprocedural
7509 optimizations on files written in different languages. This requires
7510 some support in the language front end. Currently, the C, C++ and
7511 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7512 something like this should work
7517 gfortran -c -flto baz.f90
7518 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7521 Notice that the final link is done with @command{g++} to get the C++
7522 runtime libraries and @option{-lgfortran} is added to get the Fortran
7523 runtime libraries. In general, when mixing languages in LTO mode, you
7524 should use the same link command used when mixing languages in a
7525 regular (non-LTO) compilation. This means that if your build process
7526 was mixing languages before, all you need to add is @option{-flto} to
7527 all the compile and link commands.
7529 If LTO encounters objects with C linkage declared with incompatible
7530 types in separate translation units to be linked together (undefined
7531 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7532 issued. The behavior is still undefined at runtime.
7534 If object files containing GIMPLE bytecode are stored in a library
7535 archive, say @file{libfoo.a}, it is possible to extract and use them
7536 in an LTO link if you are using @command{gold} as the linker (which,
7537 in turn requires GCC to be configured with @option{--enable-gold}).
7538 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7542 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7545 With the linker plugin enabled, @command{gold} will extract the needed
7546 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7547 to make them part of the aggregated GIMPLE image to be optimized.
7549 If you are not using @command{gold} and/or do not specify
7550 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7551 will be extracted and linked as usual, but they will not participate
7552 in the LTO optimization process.
7554 Link time optimizations do not require the presence of the whole
7555 program to operate. If the program does not require any symbols to
7556 be exported, it is possible to combine @option{-flto} and
7557 @option{-fwhopr} with @option{-fwhole-program} to allow the
7558 interprocedural optimizers to use more aggressive assumptions which
7559 may lead to improved optimization opportunities.
7561 Regarding portability: the current implementation of LTO makes no
7562 attempt at generating bytecode that can be ported between different
7563 types of hosts. The bytecode files are versioned and there is a
7564 strict version check, so bytecode files generated in one version of
7565 GCC will not work with an older/newer version of GCC.
7567 Link time optimization does not play well with generating debugging
7568 information. Combining @option{-flto} or @option{-fwhopr} with
7569 @option{-g} is experimental.
7571 This option is disabled by default.
7573 @item -fwhopr[=@var{n}]
7575 This option is identical in functionality to @option{-flto} but it
7576 differs in how the final link stage is executed. Instead of loading
7577 all the function bodies in memory, the callgraph is analyzed and
7578 optimization decisions are made (whole program analysis or WPA). Once
7579 optimization decisions are made, the callgraph is partitioned and the
7580 different sections are compiled separately (local transformations or
7581 LTRANS)@. This process allows optimizations on very large programs
7582 that otherwise would not fit in memory. This option enables
7583 @option{-fwpa} and @option{-fltrans} automatically.
7585 If you specify the optional @var{n} the link stage is executed in
7586 parallel using @var{n} parallel jobs by utilizing an installed
7587 @command{make} program. The environment variable @env{MAKE} may be
7588 used to override the program used.
7590 You can also specify @option{-fwhopr=jobserver} to use GNU make's
7591 job server mode to determine the number of parallel jobs. This
7592 is useful when the Makefile calling GCC is already parallel.
7593 The parent Makefile will need a @samp{+} prepended to the command recipe
7594 for this to work. This will likely only work if @env{MAKE} is
7597 Disabled by default.
7601 This is an internal option used by GCC when compiling with
7602 @option{-fwhopr}. You should never need to use it.
7604 This option runs the link-time optimizer in the whole-program-analysis
7605 (WPA) mode, which reads in summary information from all inputs and
7606 performs a whole-program analysis based on summary information only.
7607 It generates object files for subsequent runs of the link-time
7608 optimizer where individual object files are optimized using both
7609 summary information from the WPA mode and the actual function bodies.
7610 It then drives the LTRANS phase.
7612 Disabled by default.
7616 This is an internal option used by GCC when compiling with
7617 @option{-fwhopr}. You should never need to use it.
7619 This option runs the link-time optimizer in the local-transformation (LTRANS)
7620 mode, which reads in output from a previous run of the LTO in WPA mode.
7621 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7623 Disabled by default.
7625 @item -fltrans-output-list=@var{file}
7626 @opindex fltrans-output-list
7627 This is an internal option used by GCC when compiling with
7628 @option{-fwhopr}. You should never need to use it.
7630 This option specifies a file to which the names of LTRANS output files are
7631 written. This option is only meaningful in conjunction with @option{-fwpa}.
7633 Disabled by default.
7635 @item -flto-compression-level=@var{n}
7636 This option specifies the level of compression used for intermediate
7637 language written to LTO object files, and is only meaningful in
7638 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7639 values are 0 (no compression) to 9 (maximum compression). Values
7640 outside this range are clamped to either 0 or 9. If the option is not
7641 given, a default balanced compression setting is used.
7644 Prints a report with internal details on the workings of the link-time
7645 optimizer. The contents of this report vary from version to version,
7646 it is meant to be useful to GCC developers when processing object
7647 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7649 Disabled by default.
7651 @item -fuse-linker-plugin
7652 Enables the extraction of objects with GIMPLE bytecode information
7653 from library archives. This option relies on features available only
7654 in @command{gold}, so to use this you must configure GCC with
7655 @option{--enable-gold}. See @option{-flto} for a description on the
7656 effect of this flag and how to use it.
7658 Disabled by default.
7660 @item -fcprop-registers
7661 @opindex fcprop-registers
7662 After register allocation and post-register allocation instruction splitting,
7663 we perform a copy-propagation pass to try to reduce scheduling dependencies
7664 and occasionally eliminate the copy.
7666 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7668 @item -fprofile-correction
7669 @opindex fprofile-correction
7670 Profiles collected using an instrumented binary for multi-threaded programs may
7671 be inconsistent due to missed counter updates. When this option is specified,
7672 GCC will use heuristics to correct or smooth out such inconsistencies. By
7673 default, GCC will emit an error message when an inconsistent profile is detected.
7675 @item -fprofile-dir=@var{path}
7676 @opindex fprofile-dir
7678 Set the directory to search the profile data files in to @var{path}.
7679 This option affects only the profile data generated by
7680 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7681 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7682 and its related options.
7683 By default, GCC will use the current directory as @var{path}
7684 thus the profile data file will appear in the same directory as the object file.
7686 @item -fprofile-generate
7687 @itemx -fprofile-generate=@var{path}
7688 @opindex fprofile-generate
7690 Enable options usually used for instrumenting application to produce
7691 profile useful for later recompilation with profile feedback based
7692 optimization. You must use @option{-fprofile-generate} both when
7693 compiling and when linking your program.
7695 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7697 If @var{path} is specified, GCC will look at the @var{path} to find
7698 the profile feedback data files. See @option{-fprofile-dir}.
7701 @itemx -fprofile-use=@var{path}
7702 @opindex fprofile-use
7703 Enable profile feedback directed optimizations, and optimizations
7704 generally profitable only with profile feedback available.
7706 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7707 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7709 By default, GCC emits an error message if the feedback profiles do not
7710 match the source code. This error can be turned into a warning by using
7711 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7714 If @var{path} is specified, GCC will look at the @var{path} to find
7715 the profile feedback data files. See @option{-fprofile-dir}.
7718 The following options control compiler behavior regarding floating
7719 point arithmetic. These options trade off between speed and
7720 correctness. All must be specifically enabled.
7724 @opindex ffloat-store
7725 Do not store floating point variables in registers, and inhibit other
7726 options that might change whether a floating point value is taken from a
7729 @cindex floating point precision
7730 This option prevents undesirable excess precision on machines such as
7731 the 68000 where the floating registers (of the 68881) keep more
7732 precision than a @code{double} is supposed to have. Similarly for the
7733 x86 architecture. For most programs, the excess precision does only
7734 good, but a few programs rely on the precise definition of IEEE floating
7735 point. Use @option{-ffloat-store} for such programs, after modifying
7736 them to store all pertinent intermediate computations into variables.
7738 @item -fexcess-precision=@var{style}
7739 @opindex fexcess-precision
7740 This option allows further control over excess precision on machines
7741 where floating-point registers have more precision than the IEEE
7742 @code{float} and @code{double} types and the processor does not
7743 support operations rounding to those types. By default,
7744 @option{-fexcess-precision=fast} is in effect; this means that
7745 operations are carried out in the precision of the registers and that
7746 it is unpredictable when rounding to the types specified in the source
7747 code takes place. When compiling C, if
7748 @option{-fexcess-precision=standard} is specified then excess
7749 precision will follow the rules specified in ISO C99; in particular,
7750 both casts and assignments cause values to be rounded to their
7751 semantic types (whereas @option{-ffloat-store} only affects
7752 assignments). This option is enabled by default for C if a strict
7753 conformance option such as @option{-std=c99} is used.
7756 @option{-fexcess-precision=standard} is not implemented for languages
7757 other than C, and has no effect if
7758 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7759 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7760 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7761 semantics apply without excess precision, and in the latter, rounding
7766 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7767 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7768 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7770 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7772 This option is not turned on by any @option{-O} option since
7773 it can result in incorrect output for programs which depend on
7774 an exact implementation of IEEE or ISO rules/specifications for
7775 math functions. It may, however, yield faster code for programs
7776 that do not require the guarantees of these specifications.
7778 @item -fno-math-errno
7779 @opindex fno-math-errno
7780 Do not set ERRNO after calling math functions that are executed
7781 with a single instruction, e.g., sqrt. A program that relies on
7782 IEEE exceptions for math error handling may want to use this flag
7783 for speed while maintaining IEEE arithmetic compatibility.
7785 This option is not turned on by any @option{-O} option since
7786 it can result in incorrect output for programs which depend on
7787 an exact implementation of IEEE or ISO rules/specifications for
7788 math functions. It may, however, yield faster code for programs
7789 that do not require the guarantees of these specifications.
7791 The default is @option{-fmath-errno}.
7793 On Darwin systems, the math library never sets @code{errno}. There is
7794 therefore no reason for the compiler to consider the possibility that
7795 it might, and @option{-fno-math-errno} is the default.
7797 @item -funsafe-math-optimizations
7798 @opindex funsafe-math-optimizations
7800 Allow optimizations for floating-point arithmetic that (a) assume
7801 that arguments and results are valid and (b) may violate IEEE or
7802 ANSI standards. When used at link-time, it may include libraries
7803 or startup files that change the default FPU control word or other
7804 similar optimizations.
7806 This option is not turned on by any @option{-O} option since
7807 it can result in incorrect output for programs which depend on
7808 an exact implementation of IEEE or ISO rules/specifications for
7809 math functions. It may, however, yield faster code for programs
7810 that do not require the guarantees of these specifications.
7811 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7812 @option{-fassociative-math} and @option{-freciprocal-math}.
7814 The default is @option{-fno-unsafe-math-optimizations}.
7816 @item -fassociative-math
7817 @opindex fassociative-math
7819 Allow re-association of operands in series of floating-point operations.
7820 This violates the ISO C and C++ language standard by possibly changing
7821 computation result. NOTE: re-ordering may change the sign of zero as
7822 well as ignore NaNs and inhibit or create underflow or overflow (and
7823 thus cannot be used on a code which relies on rounding behavior like
7824 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7825 and thus may not be used when ordered comparisons are required.
7826 This option requires that both @option{-fno-signed-zeros} and
7827 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7828 much sense with @option{-frounding-math}. For Fortran the option
7829 is automatically enabled when both @option{-fno-signed-zeros} and
7830 @option{-fno-trapping-math} are in effect.
7832 The default is @option{-fno-associative-math}.
7834 @item -freciprocal-math
7835 @opindex freciprocal-math
7837 Allow the reciprocal of a value to be used instead of dividing by
7838 the value if this enables optimizations. For example @code{x / y}
7839 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7840 is subject to common subexpression elimination. Note that this loses
7841 precision and increases the number of flops operating on the value.
7843 The default is @option{-fno-reciprocal-math}.
7845 @item -ffinite-math-only
7846 @opindex ffinite-math-only
7847 Allow optimizations for floating-point arithmetic that assume
7848 that arguments and results are not NaNs or +-Infs.
7850 This option is not turned on by any @option{-O} option since
7851 it can result in incorrect output for programs which depend on
7852 an exact implementation of IEEE or ISO rules/specifications for
7853 math functions. It may, however, yield faster code for programs
7854 that do not require the guarantees of these specifications.
7856 The default is @option{-fno-finite-math-only}.
7858 @item -fno-signed-zeros
7859 @opindex fno-signed-zeros
7860 Allow optimizations for floating point arithmetic that ignore the
7861 signedness of zero. IEEE arithmetic specifies the behavior of
7862 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7863 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7864 This option implies that the sign of a zero result isn't significant.
7866 The default is @option{-fsigned-zeros}.
7868 @item -fno-trapping-math
7869 @opindex fno-trapping-math
7870 Compile code assuming that floating-point operations cannot generate
7871 user-visible traps. These traps include division by zero, overflow,
7872 underflow, inexact result and invalid operation. This option requires
7873 that @option{-fno-signaling-nans} be in effect. Setting this option may
7874 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7876 This option should never be turned on by any @option{-O} option since
7877 it can result in incorrect output for programs which depend on
7878 an exact implementation of IEEE or ISO rules/specifications for
7881 The default is @option{-ftrapping-math}.
7883 @item -frounding-math
7884 @opindex frounding-math
7885 Disable transformations and optimizations that assume default floating
7886 point rounding behavior. This is round-to-zero for all floating point
7887 to integer conversions, and round-to-nearest for all other arithmetic
7888 truncations. This option should be specified for programs that change
7889 the FP rounding mode dynamically, or that may be executed with a
7890 non-default rounding mode. This option disables constant folding of
7891 floating point expressions at compile-time (which may be affected by
7892 rounding mode) and arithmetic transformations that are unsafe in the
7893 presence of sign-dependent rounding modes.
7895 The default is @option{-fno-rounding-math}.
7897 This option is experimental and does not currently guarantee to
7898 disable all GCC optimizations that are affected by rounding mode.
7899 Future versions of GCC may provide finer control of this setting
7900 using C99's @code{FENV_ACCESS} pragma. This command line option
7901 will be used to specify the default state for @code{FENV_ACCESS}.
7903 @item -fsignaling-nans
7904 @opindex fsignaling-nans
7905 Compile code assuming that IEEE signaling NaNs may generate user-visible
7906 traps during floating-point operations. Setting this option disables
7907 optimizations that may change the number of exceptions visible with
7908 signaling NaNs. This option implies @option{-ftrapping-math}.
7910 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7913 The default is @option{-fno-signaling-nans}.
7915 This option is experimental and does not currently guarantee to
7916 disable all GCC optimizations that affect signaling NaN behavior.
7918 @item -fsingle-precision-constant
7919 @opindex fsingle-precision-constant
7920 Treat floating point constant as single precision constant instead of
7921 implicitly converting it to double precision constant.
7923 @item -fcx-limited-range
7924 @opindex fcx-limited-range
7925 When enabled, this option states that a range reduction step is not
7926 needed when performing complex division. Also, there is no checking
7927 whether the result of a complex multiplication or division is @code{NaN
7928 + I*NaN}, with an attempt to rescue the situation in that case. The
7929 default is @option{-fno-cx-limited-range}, but is enabled by
7930 @option{-ffast-math}.
7932 This option controls the default setting of the ISO C99
7933 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7936 @item -fcx-fortran-rules
7937 @opindex fcx-fortran-rules
7938 Complex multiplication and division follow Fortran rules. Range
7939 reduction is done as part of complex division, but there is no checking
7940 whether the result of a complex multiplication or division is @code{NaN
7941 + I*NaN}, with an attempt to rescue the situation in that case.
7943 The default is @option{-fno-cx-fortran-rules}.
7947 The following options control optimizations that may improve
7948 performance, but are not enabled by any @option{-O} options. This
7949 section includes experimental options that may produce broken code.
7952 @item -fbranch-probabilities
7953 @opindex fbranch-probabilities
7954 After running a program compiled with @option{-fprofile-arcs}
7955 (@pxref{Debugging Options,, Options for Debugging Your Program or
7956 @command{gcc}}), you can compile it a second time using
7957 @option{-fbranch-probabilities}, to improve optimizations based on
7958 the number of times each branch was taken. When the program
7959 compiled with @option{-fprofile-arcs} exits it saves arc execution
7960 counts to a file called @file{@var{sourcename}.gcda} for each source
7961 file. The information in this data file is very dependent on the
7962 structure of the generated code, so you must use the same source code
7963 and the same optimization options for both compilations.
7965 With @option{-fbranch-probabilities}, GCC puts a
7966 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7967 These can be used to improve optimization. Currently, they are only
7968 used in one place: in @file{reorg.c}, instead of guessing which path a
7969 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7970 exactly determine which path is taken more often.
7972 @item -fprofile-values
7973 @opindex fprofile-values
7974 If combined with @option{-fprofile-arcs}, it adds code so that some
7975 data about values of expressions in the program is gathered.
7977 With @option{-fbranch-probabilities}, it reads back the data gathered
7978 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7979 notes to instructions for their later usage in optimizations.
7981 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7985 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7986 a code to gather information about values of expressions.
7988 With @option{-fbranch-probabilities}, it reads back the data gathered
7989 and actually performs the optimizations based on them.
7990 Currently the optimizations include specialization of division operation
7991 using the knowledge about the value of the denominator.
7993 @item -frename-registers
7994 @opindex frename-registers
7995 Attempt to avoid false dependencies in scheduled code by making use
7996 of registers left over after register allocation. This optimization
7997 will most benefit processors with lots of registers. Depending on the
7998 debug information format adopted by the target, however, it can
7999 make debugging impossible, since variables will no longer stay in
8000 a ``home register''.
8002 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8006 Perform tail duplication to enlarge superblock size. This transformation
8007 simplifies the control flow of the function allowing other optimizations to do
8010 Enabled with @option{-fprofile-use}.
8012 @item -funroll-loops
8013 @opindex funroll-loops
8014 Unroll loops whose number of iterations can be determined at compile time or
8015 upon entry to the loop. @option{-funroll-loops} implies
8016 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8017 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8018 small constant number of iterations). This option makes code larger, and may
8019 or may not make it run faster.
8021 Enabled with @option{-fprofile-use}.
8023 @item -funroll-all-loops
8024 @opindex funroll-all-loops
8025 Unroll all loops, even if their number of iterations is uncertain when
8026 the loop is entered. This usually makes programs run more slowly.
8027 @option{-funroll-all-loops} implies the same options as
8028 @option{-funroll-loops}.
8031 @opindex fpeel-loops
8032 Peels the loops for that there is enough information that they do not
8033 roll much (from profile feedback). It also turns on complete loop peeling
8034 (i.e.@: complete removal of loops with small constant number of iterations).
8036 Enabled with @option{-fprofile-use}.
8038 @item -fmove-loop-invariants
8039 @opindex fmove-loop-invariants
8040 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8041 at level @option{-O1}
8043 @item -funswitch-loops
8044 @opindex funswitch-loops
8045 Move branches with loop invariant conditions out of the loop, with duplicates
8046 of the loop on both branches (modified according to result of the condition).
8048 @item -ffunction-sections
8049 @itemx -fdata-sections
8050 @opindex ffunction-sections
8051 @opindex fdata-sections
8052 Place each function or data item into its own section in the output
8053 file if the target supports arbitrary sections. The name of the
8054 function or the name of the data item determines the section's name
8057 Use these options on systems where the linker can perform optimizations
8058 to improve locality of reference in the instruction space. Most systems
8059 using the ELF object format and SPARC processors running Solaris 2 have
8060 linkers with such optimizations. AIX may have these optimizations in
8063 Only use these options when there are significant benefits from doing
8064 so. When you specify these options, the assembler and linker will
8065 create larger object and executable files and will also be slower.
8066 You will not be able to use @code{gprof} on all systems if you
8067 specify this option and you may have problems with debugging if
8068 you specify both this option and @option{-g}.
8070 @item -fbranch-target-load-optimize
8071 @opindex fbranch-target-load-optimize
8072 Perform branch target register load optimization before prologue / epilogue
8074 The use of target registers can typically be exposed only during reload,
8075 thus hoisting loads out of loops and doing inter-block scheduling needs
8076 a separate optimization pass.
8078 @item -fbranch-target-load-optimize2
8079 @opindex fbranch-target-load-optimize2
8080 Perform branch target register load optimization after prologue / epilogue
8083 @item -fbtr-bb-exclusive
8084 @opindex fbtr-bb-exclusive
8085 When performing branch target register load optimization, don't reuse
8086 branch target registers in within any basic block.
8088 @item -fstack-protector
8089 @opindex fstack-protector
8090 Emit extra code to check for buffer overflows, such as stack smashing
8091 attacks. This is done by adding a guard variable to functions with
8092 vulnerable objects. This includes functions that call alloca, and
8093 functions with buffers larger than 8 bytes. The guards are initialized
8094 when a function is entered and then checked when the function exits.
8095 If a guard check fails, an error message is printed and the program exits.
8097 @item -fstack-protector-all
8098 @opindex fstack-protector-all
8099 Like @option{-fstack-protector} except that all functions are protected.
8101 @item -fsection-anchors
8102 @opindex fsection-anchors
8103 Try to reduce the number of symbolic address calculations by using
8104 shared ``anchor'' symbols to address nearby objects. This transformation
8105 can help to reduce the number of GOT entries and GOT accesses on some
8108 For example, the implementation of the following function @code{foo}:
8112 int foo (void) @{ return a + b + c; @}
8115 would usually calculate the addresses of all three variables, but if you
8116 compile it with @option{-fsection-anchors}, it will access the variables
8117 from a common anchor point instead. The effect is similar to the
8118 following pseudocode (which isn't valid C):
8123 register int *xr = &x;
8124 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8128 Not all targets support this option.
8130 @item --param @var{name}=@var{value}
8132 In some places, GCC uses various constants to control the amount of
8133 optimization that is done. For example, GCC will not inline functions
8134 that contain more that a certain number of instructions. You can
8135 control some of these constants on the command-line using the
8136 @option{--param} option.
8138 The names of specific parameters, and the meaning of the values, are
8139 tied to the internals of the compiler, and are subject to change
8140 without notice in future releases.
8142 In each case, the @var{value} is an integer. The allowable choices for
8143 @var{name} are given in the following table:
8146 @item struct-reorg-cold-struct-ratio
8147 The threshold ratio (as a percentage) between a structure frequency
8148 and the frequency of the hottest structure in the program. This parameter
8149 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8150 We say that if the ratio of a structure frequency, calculated by profiling,
8151 to the hottest structure frequency in the program is less than this
8152 parameter, then structure reorganization is not applied to this structure.
8155 @item predictable-branch-outcome
8156 When branch is predicted to be taken with probability lower than this threshold
8157 (in percent), then it is considered well predictable. The default is 10.
8159 @item max-crossjump-edges
8160 The maximum number of incoming edges to consider for crossjumping.
8161 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8162 the number of edges incoming to each block. Increasing values mean
8163 more aggressive optimization, making the compile time increase with
8164 probably small improvement in executable size.
8166 @item min-crossjump-insns
8167 The minimum number of instructions which must be matched at the end
8168 of two blocks before crossjumping will be performed on them. This
8169 value is ignored in the case where all instructions in the block being
8170 crossjumped from are matched. The default value is 5.
8172 @item max-grow-copy-bb-insns
8173 The maximum code size expansion factor when copying basic blocks
8174 instead of jumping. The expansion is relative to a jump instruction.
8175 The default value is 8.
8177 @item max-goto-duplication-insns
8178 The maximum number of instructions to duplicate to a block that jumps
8179 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8180 passes, GCC factors computed gotos early in the compilation process,
8181 and unfactors them as late as possible. Only computed jumps at the
8182 end of a basic blocks with no more than max-goto-duplication-insns are
8183 unfactored. The default value is 8.
8185 @item max-delay-slot-insn-search
8186 The maximum number of instructions to consider when looking for an
8187 instruction to fill a delay slot. If more than this arbitrary number of
8188 instructions is searched, the time savings from filling the delay slot
8189 will be minimal so stop searching. Increasing values mean more
8190 aggressive optimization, making the compile time increase with probably
8191 small improvement in executable run time.
8193 @item max-delay-slot-live-search
8194 When trying to fill delay slots, the maximum number of instructions to
8195 consider when searching for a block with valid live register
8196 information. Increasing this arbitrarily chosen value means more
8197 aggressive optimization, increasing the compile time. This parameter
8198 should be removed when the delay slot code is rewritten to maintain the
8201 @item max-gcse-memory
8202 The approximate maximum amount of memory that will be allocated in
8203 order to perform the global common subexpression elimination
8204 optimization. If more memory than specified is required, the
8205 optimization will not be done.
8207 @item max-pending-list-length
8208 The maximum number of pending dependencies scheduling will allow
8209 before flushing the current state and starting over. Large functions
8210 with few branches or calls can create excessively large lists which
8211 needlessly consume memory and resources.
8213 @item max-inline-insns-single
8214 Several parameters control the tree inliner used in gcc.
8215 This number sets the maximum number of instructions (counted in GCC's
8216 internal representation) in a single function that the tree inliner
8217 will consider for inlining. This only affects functions declared
8218 inline and methods implemented in a class declaration (C++).
8219 The default value is 300.
8221 @item max-inline-insns-auto
8222 When you use @option{-finline-functions} (included in @option{-O3}),
8223 a lot of functions that would otherwise not be considered for inlining
8224 by the compiler will be investigated. To those functions, a different
8225 (more restrictive) limit compared to functions declared inline can
8227 The default value is 40.
8229 @item large-function-insns
8230 The limit specifying really large functions. For functions larger than this
8231 limit after inlining, inlining is constrained by
8232 @option{--param large-function-growth}. This parameter is useful primarily
8233 to avoid extreme compilation time caused by non-linear algorithms used by the
8235 The default value is 2700.
8237 @item large-function-growth
8238 Specifies maximal growth of large function caused by inlining in percents.
8239 The default value is 100 which limits large function growth to 2.0 times
8242 @item large-unit-insns
8243 The limit specifying large translation unit. Growth caused by inlining of
8244 units larger than this limit is limited by @option{--param inline-unit-growth}.
8245 For small units this might be too tight (consider unit consisting of function A
8246 that is inline and B that just calls A three time. If B is small relative to
8247 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8248 large units consisting of small inlineable functions however the overall unit
8249 growth limit is needed to avoid exponential explosion of code size. Thus for
8250 smaller units, the size is increased to @option{--param large-unit-insns}
8251 before applying @option{--param inline-unit-growth}. The default is 10000
8253 @item inline-unit-growth
8254 Specifies maximal overall growth of the compilation unit caused by inlining.
8255 The default value is 30 which limits unit growth to 1.3 times the original
8258 @item ipcp-unit-growth
8259 Specifies maximal overall growth of the compilation unit caused by
8260 interprocedural constant propagation. The default value is 10 which limits
8261 unit growth to 1.1 times the original size.
8263 @item large-stack-frame
8264 The limit specifying large stack frames. While inlining the algorithm is trying
8265 to not grow past this limit too much. Default value is 256 bytes.
8267 @item large-stack-frame-growth
8268 Specifies maximal growth of large stack frames caused by inlining in percents.
8269 The default value is 1000 which limits large stack frame growth to 11 times
8272 @item max-inline-insns-recursive
8273 @itemx max-inline-insns-recursive-auto
8274 Specifies maximum number of instructions out-of-line copy of self recursive inline
8275 function can grow into by performing recursive inlining.
8277 For functions declared inline @option{--param max-inline-insns-recursive} is
8278 taken into account. For function not declared inline, recursive inlining
8279 happens only when @option{-finline-functions} (included in @option{-O3}) is
8280 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8281 default value is 450.
8283 @item max-inline-recursive-depth
8284 @itemx max-inline-recursive-depth-auto
8285 Specifies maximum recursion depth used by the recursive inlining.
8287 For functions declared inline @option{--param max-inline-recursive-depth} is
8288 taken into account. For function not declared inline, recursive inlining
8289 happens only when @option{-finline-functions} (included in @option{-O3}) is
8290 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8293 @item min-inline-recursive-probability
8294 Recursive inlining is profitable only for function having deep recursion
8295 in average and can hurt for function having little recursion depth by
8296 increasing the prologue size or complexity of function body to other
8299 When profile feedback is available (see @option{-fprofile-generate}) the actual
8300 recursion depth can be guessed from probability that function will recurse via
8301 given call expression. This parameter limits inlining only to call expression
8302 whose probability exceeds given threshold (in percents). The default value is
8305 @item early-inlining-insns
8306 Specify growth that early inliner can make. In effect it increases amount of
8307 inlining for code having large abstraction penalty. The default value is 8.
8309 @item max-early-inliner-iterations
8310 @itemx max-early-inliner-iterations
8311 Limit of iterations of early inliner. This basically bounds number of nested
8312 indirect calls early inliner can resolve. Deeper chains are still handled by
8315 @item min-vect-loop-bound
8316 The minimum number of iterations under which a loop will not get vectorized
8317 when @option{-ftree-vectorize} is used. The number of iterations after
8318 vectorization needs to be greater than the value specified by this option
8319 to allow vectorization. The default value is 0.
8321 @item gcse-cost-distance-ratio
8322 Scaling factor in calculation of maximum distance an expression
8323 can be moved by GCSE optimizations. This is currently supported only in
8324 code hoisting pass. The bigger the ratio, the more agressive code hoisting
8325 will be with simple expressions, i.e., the expressions which have cost
8326 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8327 hoisting of simple expressions. The default value is 10.
8329 @item gcse-unrestricted-cost
8330 Cost, roughly measured as the cost of a single typical machine
8331 instruction, at which GCSE optimizations will not constrain
8332 the distance an expression can travel. This is currently
8333 supported only in code hoisting pass. The lesser the cost,
8334 the more aggressive code hoisting will be. Specifying 0 will
8335 allow all expressions to travel unrestricted distances.
8336 The default value is 3.
8338 @item max-hoist-depth
8339 The depth of search in the dominator tree for expressions to hoist.
8340 This is used to avoid quadratic behavior in hoisting algorithm.
8341 The value of 0 will avoid limiting the search, but may slow down compilation
8342 of huge functions. The default value is 30.
8344 @item max-unrolled-insns
8345 The maximum number of instructions that a loop should have if that loop
8346 is unrolled, and if the loop is unrolled, it determines how many times
8347 the loop code is unrolled.
8349 @item max-average-unrolled-insns
8350 The maximum number of instructions biased by probabilities of their execution
8351 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8352 it determines how many times the loop code is unrolled.
8354 @item max-unroll-times
8355 The maximum number of unrollings of a single loop.
8357 @item max-peeled-insns
8358 The maximum number of instructions that a loop should have if that loop
8359 is peeled, and if the loop is peeled, it determines how many times
8360 the loop code is peeled.
8362 @item max-peel-times
8363 The maximum number of peelings of a single loop.
8365 @item max-completely-peeled-insns
8366 The maximum number of insns of a completely peeled loop.
8368 @item max-completely-peel-times
8369 The maximum number of iterations of a loop to be suitable for complete peeling.
8371 @item max-completely-peel-loop-nest-depth
8372 The maximum depth of a loop nest suitable for complete peeling.
8374 @item max-unswitch-insns
8375 The maximum number of insns of an unswitched loop.
8377 @item max-unswitch-level
8378 The maximum number of branches unswitched in a single loop.
8381 The minimum cost of an expensive expression in the loop invariant motion.
8383 @item iv-consider-all-candidates-bound
8384 Bound on number of candidates for induction variables below that
8385 all candidates are considered for each use in induction variable
8386 optimizations. Only the most relevant candidates are considered
8387 if there are more candidates, to avoid quadratic time complexity.
8389 @item iv-max-considered-uses
8390 The induction variable optimizations give up on loops that contain more
8391 induction variable uses.
8393 @item iv-always-prune-cand-set-bound
8394 If number of candidates in the set is smaller than this value,
8395 we always try to remove unnecessary ivs from the set during its
8396 optimization when a new iv is added to the set.
8398 @item scev-max-expr-size
8399 Bound on size of expressions used in the scalar evolutions analyzer.
8400 Large expressions slow the analyzer.
8402 @item omega-max-vars
8403 The maximum number of variables in an Omega constraint system.
8404 The default value is 128.
8406 @item omega-max-geqs
8407 The maximum number of inequalities in an Omega constraint system.
8408 The default value is 256.
8411 The maximum number of equalities in an Omega constraint system.
8412 The default value is 128.
8414 @item omega-max-wild-cards
8415 The maximum number of wildcard variables that the Omega solver will
8416 be able to insert. The default value is 18.
8418 @item omega-hash-table-size
8419 The size of the hash table in the Omega solver. The default value is
8422 @item omega-max-keys
8423 The maximal number of keys used by the Omega solver. The default
8426 @item omega-eliminate-redundant-constraints
8427 When set to 1, use expensive methods to eliminate all redundant
8428 constraints. The default value is 0.
8430 @item vect-max-version-for-alignment-checks
8431 The maximum number of runtime checks that can be performed when
8432 doing loop versioning for alignment in the vectorizer. See option
8433 ftree-vect-loop-version for more information.
8435 @item vect-max-version-for-alias-checks
8436 The maximum number of runtime checks that can be performed when
8437 doing loop versioning for alias in the vectorizer. See option
8438 ftree-vect-loop-version for more information.
8440 @item max-iterations-to-track
8442 The maximum number of iterations of a loop the brute force algorithm
8443 for analysis of # of iterations of the loop tries to evaluate.
8445 @item hot-bb-count-fraction
8446 Select fraction of the maximal count of repetitions of basic block in program
8447 given basic block needs to have to be considered hot.
8449 @item hot-bb-frequency-fraction
8450 Select fraction of the maximal frequency of executions of basic block in
8451 function given basic block needs to have to be considered hot
8453 @item max-predicted-iterations
8454 The maximum number of loop iterations we predict statically. This is useful
8455 in cases where function contain single loop with known bound and other loop
8456 with unknown. We predict the known number of iterations correctly, while
8457 the unknown number of iterations average to roughly 10. This means that the
8458 loop without bounds would appear artificially cold relative to the other one.
8460 @item align-threshold
8462 Select fraction of the maximal frequency of executions of basic block in
8463 function given basic block will get aligned.
8465 @item align-loop-iterations
8467 A loop expected to iterate at lest the selected number of iterations will get
8470 @item tracer-dynamic-coverage
8471 @itemx tracer-dynamic-coverage-feedback
8473 This value is used to limit superblock formation once the given percentage of
8474 executed instructions is covered. This limits unnecessary code size
8477 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8478 feedback is available. The real profiles (as opposed to statically estimated
8479 ones) are much less balanced allowing the threshold to be larger value.
8481 @item tracer-max-code-growth
8482 Stop tail duplication once code growth has reached given percentage. This is
8483 rather hokey argument, as most of the duplicates will be eliminated later in
8484 cross jumping, so it may be set to much higher values than is the desired code
8487 @item tracer-min-branch-ratio
8489 Stop reverse growth when the reverse probability of best edge is less than this
8490 threshold (in percent).
8492 @item tracer-min-branch-ratio
8493 @itemx tracer-min-branch-ratio-feedback
8495 Stop forward growth if the best edge do have probability lower than this
8498 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8499 compilation for profile feedback and one for compilation without. The value
8500 for compilation with profile feedback needs to be more conservative (higher) in
8501 order to make tracer effective.
8503 @item max-cse-path-length
8505 Maximum number of basic blocks on path that cse considers. The default is 10.
8508 The maximum instructions CSE process before flushing. The default is 1000.
8510 @item ggc-min-expand
8512 GCC uses a garbage collector to manage its own memory allocation. This
8513 parameter specifies the minimum percentage by which the garbage
8514 collector's heap should be allowed to expand between collections.
8515 Tuning this may improve compilation speed; it has no effect on code
8518 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8519 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8520 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8521 GCC is not able to calculate RAM on a particular platform, the lower
8522 bound of 30% is used. Setting this parameter and
8523 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8524 every opportunity. This is extremely slow, but can be useful for
8527 @item ggc-min-heapsize
8529 Minimum size of the garbage collector's heap before it begins bothering
8530 to collect garbage. The first collection occurs after the heap expands
8531 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8532 tuning this may improve compilation speed, and has no effect on code
8535 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8536 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8537 with a lower bound of 4096 (four megabytes) and an upper bound of
8538 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8539 particular platform, the lower bound is used. Setting this parameter
8540 very large effectively disables garbage collection. Setting this
8541 parameter and @option{ggc-min-expand} to zero causes a full collection
8542 to occur at every opportunity.
8544 @item max-reload-search-insns
8545 The maximum number of instruction reload should look backward for equivalent
8546 register. Increasing values mean more aggressive optimization, making the
8547 compile time increase with probably slightly better performance. The default
8550 @item max-cselib-memory-locations
8551 The maximum number of memory locations cselib should take into account.
8552 Increasing values mean more aggressive optimization, making the compile time
8553 increase with probably slightly better performance. The default value is 500.
8555 @item reorder-blocks-duplicate
8556 @itemx reorder-blocks-duplicate-feedback
8558 Used by basic block reordering pass to decide whether to use unconditional
8559 branch or duplicate the code on its destination. Code is duplicated when its
8560 estimated size is smaller than this value multiplied by the estimated size of
8561 unconditional jump in the hot spots of the program.
8563 The @option{reorder-block-duplicate-feedback} is used only when profile
8564 feedback is available and may be set to higher values than
8565 @option{reorder-block-duplicate} since information about the hot spots is more
8568 @item max-sched-ready-insns
8569 The maximum number of instructions ready to be issued the scheduler should
8570 consider at any given time during the first scheduling pass. Increasing
8571 values mean more thorough searches, making the compilation time increase
8572 with probably little benefit. The default value is 100.
8574 @item max-sched-region-blocks
8575 The maximum number of blocks in a region to be considered for
8576 interblock scheduling. The default value is 10.
8578 @item max-pipeline-region-blocks
8579 The maximum number of blocks in a region to be considered for
8580 pipelining in the selective scheduler. The default value is 15.
8582 @item max-sched-region-insns
8583 The maximum number of insns in a region to be considered for
8584 interblock scheduling. The default value is 100.
8586 @item max-pipeline-region-insns
8587 The maximum number of insns in a region to be considered for
8588 pipelining in the selective scheduler. The default value is 200.
8591 The minimum probability (in percents) of reaching a source block
8592 for interblock speculative scheduling. The default value is 40.
8594 @item max-sched-extend-regions-iters
8595 The maximum number of iterations through CFG to extend regions.
8596 0 - disable region extension,
8597 N - do at most N iterations.
8598 The default value is 0.
8600 @item max-sched-insn-conflict-delay
8601 The maximum conflict delay for an insn to be considered for speculative motion.
8602 The default value is 3.
8604 @item sched-spec-prob-cutoff
8605 The minimal probability of speculation success (in percents), so that
8606 speculative insn will be scheduled.
8607 The default value is 40.
8609 @item sched-mem-true-dep-cost
8610 Minimal distance (in CPU cycles) between store and load targeting same
8611 memory locations. The default value is 1.
8613 @item selsched-max-lookahead
8614 The maximum size of the lookahead window of selective scheduling. It is a
8615 depth of search for available instructions.
8616 The default value is 50.
8618 @item selsched-max-sched-times
8619 The maximum number of times that an instruction will be scheduled during
8620 selective scheduling. This is the limit on the number of iterations
8621 through which the instruction may be pipelined. The default value is 2.
8623 @item selsched-max-insns-to-rename
8624 The maximum number of best instructions in the ready list that are considered
8625 for renaming in the selective scheduler. The default value is 2.
8627 @item max-last-value-rtl
8628 The maximum size measured as number of RTLs that can be recorded in an expression
8629 in combiner for a pseudo register as last known value of that register. The default
8632 @item integer-share-limit
8633 Small integer constants can use a shared data structure, reducing the
8634 compiler's memory usage and increasing its speed. This sets the maximum
8635 value of a shared integer constant. The default value is 256.
8637 @item min-virtual-mappings
8638 Specifies the minimum number of virtual mappings in the incremental
8639 SSA updater that should be registered to trigger the virtual mappings
8640 heuristic defined by virtual-mappings-ratio. The default value is
8643 @item virtual-mappings-ratio
8644 If the number of virtual mappings is virtual-mappings-ratio bigger
8645 than the number of virtual symbols to be updated, then the incremental
8646 SSA updater switches to a full update for those symbols. The default
8649 @item ssp-buffer-size
8650 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8651 protection when @option{-fstack-protection} is used.
8653 @item max-jump-thread-duplication-stmts
8654 Maximum number of statements allowed in a block that needs to be
8655 duplicated when threading jumps.
8657 @item max-fields-for-field-sensitive
8658 Maximum number of fields in a structure we will treat in
8659 a field sensitive manner during pointer analysis. The default is zero
8660 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8662 @item prefetch-latency
8663 Estimate on average number of instructions that are executed before
8664 prefetch finishes. The distance we prefetch ahead is proportional
8665 to this constant. Increasing this number may also lead to less
8666 streams being prefetched (see @option{simultaneous-prefetches}).
8668 @item simultaneous-prefetches
8669 Maximum number of prefetches that can run at the same time.
8671 @item l1-cache-line-size
8672 The size of cache line in L1 cache, in bytes.
8675 The size of L1 cache, in kilobytes.
8678 The size of L2 cache, in kilobytes.
8680 @item min-insn-to-prefetch-ratio
8681 The minimum ratio between the number of instructions and the
8682 number of prefetches to enable prefetching in a loop.
8684 @item prefetch-min-insn-to-mem-ratio
8685 The minimum ratio between the number of instructions and the
8686 number of memory references to enable prefetching in a loop.
8688 @item use-canonical-types
8689 Whether the compiler should use the ``canonical'' type system. By
8690 default, this should always be 1, which uses a more efficient internal
8691 mechanism for comparing types in C++ and Objective-C++. However, if
8692 bugs in the canonical type system are causing compilation failures,
8693 set this value to 0 to disable canonical types.
8695 @item switch-conversion-max-branch-ratio
8696 Switch initialization conversion will refuse to create arrays that are
8697 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8698 branches in the switch.
8700 @item max-partial-antic-length
8701 Maximum length of the partial antic set computed during the tree
8702 partial redundancy elimination optimization (@option{-ftree-pre}) when
8703 optimizing at @option{-O3} and above. For some sorts of source code
8704 the enhanced partial redundancy elimination optimization can run away,
8705 consuming all of the memory available on the host machine. This
8706 parameter sets a limit on the length of the sets that are computed,
8707 which prevents the runaway behavior. Setting a value of 0 for
8708 this parameter will allow an unlimited set length.
8710 @item sccvn-max-scc-size
8711 Maximum size of a strongly connected component (SCC) during SCCVN
8712 processing. If this limit is hit, SCCVN processing for the whole
8713 function will not be done and optimizations depending on it will
8714 be disabled. The default maximum SCC size is 10000.
8716 @item ira-max-loops-num
8717 IRA uses a regional register allocation by default. If a function
8718 contains loops more than number given by the parameter, only at most
8719 given number of the most frequently executed loops will form regions
8720 for the regional register allocation. The default value of the
8723 @item ira-max-conflict-table-size
8724 Although IRA uses a sophisticated algorithm of compression conflict
8725 table, the table can be still big for huge functions. If the conflict
8726 table for a function could be more than size in MB given by the
8727 parameter, the conflict table is not built and faster, simpler, and
8728 lower quality register allocation algorithm will be used. The
8729 algorithm do not use pseudo-register conflicts. The default value of
8730 the parameter is 2000.
8732 @item ira-loop-reserved-regs
8733 IRA can be used to evaluate more accurate register pressure in loops
8734 for decision to move loop invariants (see @option{-O3}). The number
8735 of available registers reserved for some other purposes is described
8736 by this parameter. The default value of the parameter is 2 which is
8737 minimal number of registers needed for execution of typical
8738 instruction. This value is the best found from numerous experiments.
8740 @item loop-invariant-max-bbs-in-loop
8741 Loop invariant motion can be very expensive, both in compile time and
8742 in amount of needed compile time memory, with very large loops. Loops
8743 with more basic blocks than this parameter won't have loop invariant
8744 motion optimization performed on them. The default value of the
8745 parameter is 1000 for -O1 and 10000 for -O2 and above.
8747 @item max-vartrack-size
8748 Sets a maximum number of hash table slots to use during variable
8749 tracking dataflow analysis of any function. If this limit is exceeded
8750 with variable tracking at assignments enabled, analysis for that
8751 function is retried without it, after removing all debug insns from
8752 the function. If the limit is exceeded even without debug insns, var
8753 tracking analysis is completely disabled for the function. Setting
8754 the parameter to zero makes it unlimited.
8756 @item min-nondebug-insn-uid
8757 Use uids starting at this parameter for nondebug insns. The range below
8758 the parameter is reserved exclusively for debug insns created by
8759 @option{-fvar-tracking-assignments}, but debug insns may get
8760 (non-overlapping) uids above it if the reserved range is exhausted.
8762 @item ipa-sra-ptr-growth-factor
8763 IPA-SRA will replace a pointer to an aggregate with one or more new
8764 parameters only when their cumulative size is less or equal to
8765 @option{ipa-sra-ptr-growth-factor} times the size of the original
8768 @item graphite-max-nb-scop-params
8769 To avoid exponential effects in the Graphite loop transforms, the
8770 number of parameters in a Static Control Part (SCoP) is bounded. The
8771 default value is 10 parameters. A variable whose value is unknown at
8772 compile time and defined outside a SCoP is a parameter of the SCoP.
8774 @item graphite-max-bbs-per-function
8775 To avoid exponential effects in the detection of SCoPs, the size of
8776 the functions analyzed by Graphite is bounded. The default value is
8779 @item loop-block-tile-size
8780 Loop blocking or strip mining transforms, enabled with
8781 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8782 loop in the loop nest by a given number of iterations. The strip
8783 length can be changed using the @option{loop-block-tile-size}
8784 parameter. The default value is 51 iterations.
8786 @item devirt-type-list-size
8787 IPA-CP attempts to track all possible types passed to a function's
8788 parameter in order to perform devirtualization.
8789 @option{devirt-type-list-size} is the maximum number of types it
8790 stores per a single formal parameter of a function.
8795 @node Preprocessor Options
8796 @section Options Controlling the Preprocessor
8797 @cindex preprocessor options
8798 @cindex options, preprocessor
8800 These options control the C preprocessor, which is run on each C source
8801 file before actual compilation.
8803 If you use the @option{-E} option, nothing is done except preprocessing.
8804 Some of these options make sense only together with @option{-E} because
8805 they cause the preprocessor output to be unsuitable for actual
8809 @item -Wp,@var{option}
8811 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8812 and pass @var{option} directly through to the preprocessor. If
8813 @var{option} contains commas, it is split into multiple options at the
8814 commas. However, many options are modified, translated or interpreted
8815 by the compiler driver before being passed to the preprocessor, and
8816 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8817 interface is undocumented and subject to change, so whenever possible
8818 you should avoid using @option{-Wp} and let the driver handle the
8821 @item -Xpreprocessor @var{option}
8822 @opindex Xpreprocessor
8823 Pass @var{option} as an option to the preprocessor. You can use this to
8824 supply system-specific preprocessor options which GCC does not know how to
8827 If you want to pass an option that takes an argument, you must use
8828 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8831 @include cppopts.texi
8833 @node Assembler Options
8834 @section Passing Options to the Assembler
8836 @c prevent bad page break with this line
8837 You can pass options to the assembler.
8840 @item -Wa,@var{option}
8842 Pass @var{option} as an option to the assembler. If @var{option}
8843 contains commas, it is split into multiple options at the commas.
8845 @item -Xassembler @var{option}
8847 Pass @var{option} as an option to the assembler. You can use this to
8848 supply system-specific assembler options which GCC does not know how to
8851 If you want to pass an option that takes an argument, you must use
8852 @option{-Xassembler} twice, once for the option and once for the argument.
8857 @section Options for Linking
8858 @cindex link options
8859 @cindex options, linking
8861 These options come into play when the compiler links object files into
8862 an executable output file. They are meaningless if the compiler is
8863 not doing a link step.
8867 @item @var{object-file-name}
8868 A file name that does not end in a special recognized suffix is
8869 considered to name an object file or library. (Object files are
8870 distinguished from libraries by the linker according to the file
8871 contents.) If linking is done, these object files are used as input
8880 If any of these options is used, then the linker is not run, and
8881 object file names should not be used as arguments. @xref{Overall
8885 @item -l@var{library}
8886 @itemx -l @var{library}
8888 Search the library named @var{library} when linking. (The second
8889 alternative with the library as a separate argument is only for
8890 POSIX compliance and is not recommended.)
8892 It makes a difference where in the command you write this option; the
8893 linker searches and processes libraries and object files in the order they
8894 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8895 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8896 to functions in @samp{z}, those functions may not be loaded.
8898 The linker searches a standard list of directories for the library,
8899 which is actually a file named @file{lib@var{library}.a}. The linker
8900 then uses this file as if it had been specified precisely by name.
8902 The directories searched include several standard system directories
8903 plus any that you specify with @option{-L}.
8905 Normally the files found this way are library files---archive files
8906 whose members are object files. The linker handles an archive file by
8907 scanning through it for members which define symbols that have so far
8908 been referenced but not defined. But if the file that is found is an
8909 ordinary object file, it is linked in the usual fashion. The only
8910 difference between using an @option{-l} option and specifying a file name
8911 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8912 and searches several directories.
8916 You need this special case of the @option{-l} option in order to
8917 link an Objective-C or Objective-C++ program.
8920 @opindex nostartfiles
8921 Do not use the standard system startup files when linking.
8922 The standard system libraries are used normally, unless @option{-nostdlib}
8923 or @option{-nodefaultlibs} is used.
8925 @item -nodefaultlibs
8926 @opindex nodefaultlibs
8927 Do not use the standard system libraries when linking.
8928 Only the libraries you specify will be passed to the linker, options
8929 specifying linkage of the system libraries, such as @code{-static-libgcc}
8930 or @code{-shared-libgcc}, will be ignored.
8931 The standard startup files are used normally, unless @option{-nostartfiles}
8932 is used. The compiler may generate calls to @code{memcmp},
8933 @code{memset}, @code{memcpy} and @code{memmove}.
8934 These entries are usually resolved by entries in
8935 libc. These entry points should be supplied through some other
8936 mechanism when this option is specified.
8940 Do not use the standard system startup files or libraries when linking.
8941 No startup files and only the libraries you specify will be passed to
8942 the linker, options specifying linkage of the system libraries, such as
8943 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8944 The compiler may generate calls to @code{memcmp}, @code{memset},
8945 @code{memcpy} and @code{memmove}.
8946 These entries are usually resolved by entries in
8947 libc. These entry points should be supplied through some other
8948 mechanism when this option is specified.
8950 @cindex @option{-lgcc}, use with @option{-nostdlib}
8951 @cindex @option{-nostdlib} and unresolved references
8952 @cindex unresolved references and @option{-nostdlib}
8953 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8954 @cindex @option{-nodefaultlibs} and unresolved references
8955 @cindex unresolved references and @option{-nodefaultlibs}
8956 One of the standard libraries bypassed by @option{-nostdlib} and
8957 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8958 that GCC uses to overcome shortcomings of particular machines, or special
8959 needs for some languages.
8960 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8961 Collection (GCC) Internals},
8962 for more discussion of @file{libgcc.a}.)
8963 In most cases, you need @file{libgcc.a} even when you want to avoid
8964 other standard libraries. In other words, when you specify @option{-nostdlib}
8965 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8966 This ensures that you have no unresolved references to internal GCC
8967 library subroutines. (For example, @samp{__main}, used to ensure C++
8968 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8969 GNU Compiler Collection (GCC) Internals}.)
8973 Produce a position independent executable on targets which support it.
8974 For predictable results, you must also specify the same set of options
8975 that were used to generate code (@option{-fpie}, @option{-fPIE},
8976 or model suboptions) when you specify this option.
8980 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8981 that support it. This instructs the linker to add all symbols, not
8982 only used ones, to the dynamic symbol table. This option is needed
8983 for some uses of @code{dlopen} or to allow obtaining backtraces
8984 from within a program.
8988 Remove all symbol table and relocation information from the executable.
8992 On systems that support dynamic linking, this prevents linking with the shared
8993 libraries. On other systems, this option has no effect.
8997 Produce a shared object which can then be linked with other objects to
8998 form an executable. Not all systems support this option. For predictable
8999 results, you must also specify the same set of options that were used to
9000 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9001 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9002 needs to build supplementary stub code for constructors to work. On
9003 multi-libbed systems, @samp{gcc -shared} must select the correct support
9004 libraries to link against. Failing to supply the correct flags may lead
9005 to subtle defects. Supplying them in cases where they are not necessary
9008 @item -shared-libgcc
9009 @itemx -static-libgcc
9010 @opindex shared-libgcc
9011 @opindex static-libgcc
9012 On systems that provide @file{libgcc} as a shared library, these options
9013 force the use of either the shared or static version respectively.
9014 If no shared version of @file{libgcc} was built when the compiler was
9015 configured, these options have no effect.
9017 There are several situations in which an application should use the
9018 shared @file{libgcc} instead of the static version. The most common
9019 of these is when the application wishes to throw and catch exceptions
9020 across different shared libraries. In that case, each of the libraries
9021 as well as the application itself should use the shared @file{libgcc}.
9023 Therefore, the G++ and GCJ drivers automatically add
9024 @option{-shared-libgcc} whenever you build a shared library or a main
9025 executable, because C++ and Java programs typically use exceptions, so
9026 this is the right thing to do.
9028 If, instead, you use the GCC driver to create shared libraries, you may
9029 find that they will not always be linked with the shared @file{libgcc}.
9030 If GCC finds, at its configuration time, that you have a non-GNU linker
9031 or a GNU linker that does not support option @option{--eh-frame-hdr},
9032 it will link the shared version of @file{libgcc} into shared libraries
9033 by default. Otherwise, it will take advantage of the linker and optimize
9034 away the linking with the shared version of @file{libgcc}, linking with
9035 the static version of libgcc by default. This allows exceptions to
9036 propagate through such shared libraries, without incurring relocation
9037 costs at library load time.
9039 However, if a library or main executable is supposed to throw or catch
9040 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9041 for the languages used in the program, or using the option
9042 @option{-shared-libgcc}, such that it is linked with the shared
9045 @item -static-libstdc++
9046 When the @command{g++} program is used to link a C++ program, it will
9047 normally automatically link against @option{libstdc++}. If
9048 @file{libstdc++} is available as a shared library, and the
9049 @option{-static} option is not used, then this will link against the
9050 shared version of @file{libstdc++}. That is normally fine. However, it
9051 is sometimes useful to freeze the version of @file{libstdc++} used by
9052 the program without going all the way to a fully static link. The
9053 @option{-static-libstdc++} option directs the @command{g++} driver to
9054 link @file{libstdc++} statically, without necessarily linking other
9055 libraries statically.
9059 Bind references to global symbols when building a shared object. Warn
9060 about any unresolved references (unless overridden by the link editor
9061 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9064 @item -T @var{script}
9066 @cindex linker script
9067 Use @var{script} as the linker script. This option is supported by most
9068 systems using the GNU linker. On some targets, such as bare-board
9069 targets without an operating system, the @option{-T} option may be required
9070 when linking to avoid references to undefined symbols.
9072 @item -Xlinker @var{option}
9074 Pass @var{option} as an option to the linker. You can use this to
9075 supply system-specific linker options which GCC does not know how to
9078 If you want to pass an option that takes a separate argument, you must use
9079 @option{-Xlinker} twice, once for the option and once for the argument.
9080 For example, to pass @option{-assert definitions}, you must write
9081 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9082 @option{-Xlinker "-assert definitions"}, because this passes the entire
9083 string as a single argument, which is not what the linker expects.
9085 When using the GNU linker, it is usually more convenient to pass
9086 arguments to linker options using the @option{@var{option}=@var{value}}
9087 syntax than as separate arguments. For example, you can specify
9088 @samp{-Xlinker -Map=output.map} rather than
9089 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9090 this syntax for command-line options.
9092 @item -Wl,@var{option}
9094 Pass @var{option} as an option to the linker. If @var{option} contains
9095 commas, it is split into multiple options at the commas. You can use this
9096 syntax to pass an argument to the option.
9097 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9098 linker. When using the GNU linker, you can also get the same effect with
9099 @samp{-Wl,-Map=output.map}.
9101 @item -u @var{symbol}
9103 Pretend the symbol @var{symbol} is undefined, to force linking of
9104 library modules to define it. You can use @option{-u} multiple times with
9105 different symbols to force loading of additional library modules.
9108 @node Directory Options
9109 @section Options for Directory Search
9110 @cindex directory options
9111 @cindex options, directory search
9114 These options specify directories to search for header files, for
9115 libraries and for parts of the compiler:
9120 Add the directory @var{dir} to the head of the list of directories to be
9121 searched for header files. This can be used to override a system header
9122 file, substituting your own version, since these directories are
9123 searched before the system header file directories. However, you should
9124 not use this option to add directories that contain vendor-supplied
9125 system header files (use @option{-isystem} for that). If you use more than
9126 one @option{-I} option, the directories are scanned in left-to-right
9127 order; the standard system directories come after.
9129 If a standard system include directory, or a directory specified with
9130 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9131 option will be ignored. The directory will still be searched but as a
9132 system directory at its normal position in the system include chain.
9133 This is to ensure that GCC's procedure to fix buggy system headers and
9134 the ordering for the include_next directive are not inadvertently changed.
9135 If you really need to change the search order for system directories,
9136 use the @option{-nostdinc} and/or @option{-isystem} options.
9138 @item -iplugindir=@var{dir}
9139 Set the directory to search for plugins which are passed
9140 by @option{-fplugin=@var{name}} instead of
9141 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9142 to be used by the user, but only passed by the driver.
9144 @item -iquote@var{dir}
9146 Add the directory @var{dir} to the head of the list of directories to
9147 be searched for header files only for the case of @samp{#include
9148 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9149 otherwise just like @option{-I}.
9153 Add directory @var{dir} to the list of directories to be searched
9156 @item -B@var{prefix}
9158 This option specifies where to find the executables, libraries,
9159 include files, and data files of the compiler itself.
9161 The compiler driver program runs one or more of the subprograms
9162 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9163 @var{prefix} as a prefix for each program it tries to run, both with and
9164 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9166 For each subprogram to be run, the compiler driver first tries the
9167 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9168 was not specified, the driver tries two standard prefixes, which are
9169 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9170 those results in a file name that is found, the unmodified program
9171 name is searched for using the directories specified in your
9172 @env{PATH} environment variable.
9174 The compiler will check to see if the path provided by the @option{-B}
9175 refers to a directory, and if necessary it will add a directory
9176 separator character at the end of the path.
9178 @option{-B} prefixes that effectively specify directory names also apply
9179 to libraries in the linker, because the compiler translates these
9180 options into @option{-L} options for the linker. They also apply to
9181 includes files in the preprocessor, because the compiler translates these
9182 options into @option{-isystem} options for the preprocessor. In this case,
9183 the compiler appends @samp{include} to the prefix.
9185 The run-time support file @file{libgcc.a} can also be searched for using
9186 the @option{-B} prefix, if needed. If it is not found there, the two
9187 standard prefixes above are tried, and that is all. The file is left
9188 out of the link if it is not found by those means.
9190 Another way to specify a prefix much like the @option{-B} prefix is to use
9191 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9194 As a special kludge, if the path provided by @option{-B} is
9195 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9196 9, then it will be replaced by @file{[dir/]include}. This is to help
9197 with boot-strapping the compiler.
9199 @item -specs=@var{file}
9201 Process @var{file} after the compiler reads in the standard @file{specs}
9202 file, in order to override the defaults that the @file{gcc} driver
9203 program uses when determining what switches to pass to @file{cc1},
9204 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9205 @option{-specs=@var{file}} can be specified on the command line, and they
9206 are processed in order, from left to right.
9208 @item --sysroot=@var{dir}
9210 Use @var{dir} as the logical root directory for headers and libraries.
9211 For example, if the compiler would normally search for headers in
9212 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9213 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9215 If you use both this option and the @option{-isysroot} option, then
9216 the @option{--sysroot} option will apply to libraries, but the
9217 @option{-isysroot} option will apply to header files.
9219 The GNU linker (beginning with version 2.16) has the necessary support
9220 for this option. If your linker does not support this option, the
9221 header file aspect of @option{--sysroot} will still work, but the
9222 library aspect will not.
9226 This option has been deprecated. Please use @option{-iquote} instead for
9227 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9228 Any directories you specify with @option{-I} options before the @option{-I-}
9229 option are searched only for the case of @samp{#include "@var{file}"};
9230 they are not searched for @samp{#include <@var{file}>}.
9232 If additional directories are specified with @option{-I} options after
9233 the @option{-I-}, these directories are searched for all @samp{#include}
9234 directives. (Ordinarily @emph{all} @option{-I} directories are used
9237 In addition, the @option{-I-} option inhibits the use of the current
9238 directory (where the current input file came from) as the first search
9239 directory for @samp{#include "@var{file}"}. There is no way to
9240 override this effect of @option{-I-}. With @option{-I.} you can specify
9241 searching the directory which was current when the compiler was
9242 invoked. That is not exactly the same as what the preprocessor does
9243 by default, but it is often satisfactory.
9245 @option{-I-} does not inhibit the use of the standard system directories
9246 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9253 @section Specifying subprocesses and the switches to pass to them
9256 @command{gcc} is a driver program. It performs its job by invoking a
9257 sequence of other programs to do the work of compiling, assembling and
9258 linking. GCC interprets its command-line parameters and uses these to
9259 deduce which programs it should invoke, and which command-line options
9260 it ought to place on their command lines. This behavior is controlled
9261 by @dfn{spec strings}. In most cases there is one spec string for each
9262 program that GCC can invoke, but a few programs have multiple spec
9263 strings to control their behavior. The spec strings built into GCC can
9264 be overridden by using the @option{-specs=} command-line switch to specify
9267 @dfn{Spec files} are plaintext files that are used to construct spec
9268 strings. They consist of a sequence of directives separated by blank
9269 lines. The type of directive is determined by the first non-whitespace
9270 character on the line and it can be one of the following:
9273 @item %@var{command}
9274 Issues a @var{command} to the spec file processor. The commands that can
9278 @item %include <@var{file}>
9279 @cindex @code{%include}
9280 Search for @var{file} and insert its text at the current point in the
9283 @item %include_noerr <@var{file}>
9284 @cindex @code{%include_noerr}
9285 Just like @samp{%include}, but do not generate an error message if the include
9286 file cannot be found.
9288 @item %rename @var{old_name} @var{new_name}
9289 @cindex @code{%rename}
9290 Rename the spec string @var{old_name} to @var{new_name}.
9294 @item *[@var{spec_name}]:
9295 This tells the compiler to create, override or delete the named spec
9296 string. All lines after this directive up to the next directive or
9297 blank line are considered to be the text for the spec string. If this
9298 results in an empty string then the spec will be deleted. (Or, if the
9299 spec did not exist, then nothing will happened.) Otherwise, if the spec
9300 does not currently exist a new spec will be created. If the spec does
9301 exist then its contents will be overridden by the text of this
9302 directive, unless the first character of that text is the @samp{+}
9303 character, in which case the text will be appended to the spec.
9305 @item [@var{suffix}]:
9306 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9307 and up to the next directive or blank line are considered to make up the
9308 spec string for the indicated suffix. When the compiler encounters an
9309 input file with the named suffix, it will processes the spec string in
9310 order to work out how to compile that file. For example:
9317 This says that any input file whose name ends in @samp{.ZZ} should be
9318 passed to the program @samp{z-compile}, which should be invoked with the
9319 command-line switch @option{-input} and with the result of performing the
9320 @samp{%i} substitution. (See below.)
9322 As an alternative to providing a spec string, the text that follows a
9323 suffix directive can be one of the following:
9326 @item @@@var{language}
9327 This says that the suffix is an alias for a known @var{language}. This is
9328 similar to using the @option{-x} command-line switch to GCC to specify a
9329 language explicitly. For example:
9336 Says that .ZZ files are, in fact, C++ source files.
9339 This causes an error messages saying:
9342 @var{name} compiler not installed on this system.
9346 GCC already has an extensive list of suffixes built into it.
9347 This directive will add an entry to the end of the list of suffixes, but
9348 since the list is searched from the end backwards, it is effectively
9349 possible to override earlier entries using this technique.
9353 GCC has the following spec strings built into it. Spec files can
9354 override these strings or create their own. Note that individual
9355 targets can also add their own spec strings to this list.
9358 asm Options to pass to the assembler
9359 asm_final Options to pass to the assembler post-processor
9360 cpp Options to pass to the C preprocessor
9361 cc1 Options to pass to the C compiler
9362 cc1plus Options to pass to the C++ compiler
9363 endfile Object files to include at the end of the link
9364 link Options to pass to the linker
9365 lib Libraries to include on the command line to the linker
9366 libgcc Decides which GCC support library to pass to the linker
9367 linker Sets the name of the linker
9368 predefines Defines to be passed to the C preprocessor
9369 signed_char Defines to pass to CPP to say whether @code{char} is signed
9371 startfile Object files to include at the start of the link
9374 Here is a small example of a spec file:
9380 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9383 This example renames the spec called @samp{lib} to @samp{old_lib} and
9384 then overrides the previous definition of @samp{lib} with a new one.
9385 The new definition adds in some extra command-line options before
9386 including the text of the old definition.
9388 @dfn{Spec strings} are a list of command-line options to be passed to their
9389 corresponding program. In addition, the spec strings can contain
9390 @samp{%}-prefixed sequences to substitute variable text or to
9391 conditionally insert text into the command line. Using these constructs
9392 it is possible to generate quite complex command lines.
9394 Here is a table of all defined @samp{%}-sequences for spec
9395 strings. Note that spaces are not generated automatically around the
9396 results of expanding these sequences. Therefore you can concatenate them
9397 together or combine them with constant text in a single argument.
9401 Substitute one @samp{%} into the program name or argument.
9404 Substitute the name of the input file being processed.
9407 Substitute the basename of the input file being processed.
9408 This is the substring up to (and not including) the last period
9409 and not including the directory.
9412 This is the same as @samp{%b}, but include the file suffix (text after
9416 Marks the argument containing or following the @samp{%d} as a
9417 temporary file name, so that that file will be deleted if GCC exits
9418 successfully. Unlike @samp{%g}, this contributes no text to the
9421 @item %g@var{suffix}
9422 Substitute a file name that has suffix @var{suffix} and is chosen
9423 once per compilation, and mark the argument in the same way as
9424 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9425 name is now chosen in a way that is hard to predict even when previously
9426 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9427 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9428 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9429 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9430 was simply substituted with a file name chosen once per compilation,
9431 without regard to any appended suffix (which was therefore treated
9432 just like ordinary text), making such attacks more likely to succeed.
9434 @item %u@var{suffix}
9435 Like @samp{%g}, but generates a new temporary file name even if
9436 @samp{%u@var{suffix}} was already seen.
9438 @item %U@var{suffix}
9439 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9440 new one if there is no such last file name. In the absence of any
9441 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9442 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9443 would involve the generation of two distinct file names, one
9444 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9445 simply substituted with a file name chosen for the previous @samp{%u},
9446 without regard to any appended suffix.
9448 @item %j@var{suffix}
9449 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9450 writable, and if save-temps is off; otherwise, substitute the name
9451 of a temporary file, just like @samp{%u}. This temporary file is not
9452 meant for communication between processes, but rather as a junk
9455 @item %|@var{suffix}
9456 @itemx %m@var{suffix}
9457 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9458 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9459 all. These are the two most common ways to instruct a program that it
9460 should read from standard input or write to standard output. If you
9461 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9462 construct: see for example @file{f/lang-specs.h}.
9464 @item %.@var{SUFFIX}
9465 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9466 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9467 terminated by the next space or %.
9470 Marks the argument containing or following the @samp{%w} as the
9471 designated output file of this compilation. This puts the argument
9472 into the sequence of arguments that @samp{%o} will substitute later.
9475 Substitutes the names of all the output files, with spaces
9476 automatically placed around them. You should write spaces
9477 around the @samp{%o} as well or the results are undefined.
9478 @samp{%o} is for use in the specs for running the linker.
9479 Input files whose names have no recognized suffix are not compiled
9480 at all, but they are included among the output files, so they will
9484 Substitutes the suffix for object files. Note that this is
9485 handled specially when it immediately follows @samp{%g, %u, or %U},
9486 because of the need for those to form complete file names. The
9487 handling is such that @samp{%O} is treated exactly as if it had already
9488 been substituted, except that @samp{%g, %u, and %U} do not currently
9489 support additional @var{suffix} characters following @samp{%O} as they would
9490 following, for example, @samp{.o}.
9493 Substitutes the standard macro predefinitions for the
9494 current target machine. Use this when running @code{cpp}.
9497 Like @samp{%p}, but puts @samp{__} before and after the name of each
9498 predefined macro, except for macros that start with @samp{__} or with
9499 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9503 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9504 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9505 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9506 and @option{-imultilib} as necessary.
9509 Current argument is the name of a library or startup file of some sort.
9510 Search for that file in a standard list of directories and substitute
9511 the full name found. The current working directory is included in the
9512 list of directories scanned.
9515 Current argument is the name of a linker script. Search for that file
9516 in the current list of directories to scan for libraries. If the file
9517 is located insert a @option{--script} option into the command line
9518 followed by the full path name found. If the file is not found then
9519 generate an error message. Note: the current working directory is not
9523 Print @var{str} as an error message. @var{str} is terminated by a newline.
9524 Use this when inconsistent options are detected.
9527 Substitute the contents of spec string @var{name} at this point.
9530 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9532 @item %x@{@var{option}@}
9533 Accumulate an option for @samp{%X}.
9536 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9540 Output the accumulated assembler options specified by @option{-Wa}.
9543 Output the accumulated preprocessor options specified by @option{-Wp}.
9546 Process the @code{asm} spec. This is used to compute the
9547 switches to be passed to the assembler.
9550 Process the @code{asm_final} spec. This is a spec string for
9551 passing switches to an assembler post-processor, if such a program is
9555 Process the @code{link} spec. This is the spec for computing the
9556 command line passed to the linker. Typically it will make use of the
9557 @samp{%L %G %S %D and %E} sequences.
9560 Dump out a @option{-L} option for each directory that GCC believes might
9561 contain startup files. If the target supports multilibs then the
9562 current multilib directory will be prepended to each of these paths.
9565 Process the @code{lib} spec. This is a spec string for deciding which
9566 libraries should be included on the command line to the linker.
9569 Process the @code{libgcc} spec. This is a spec string for deciding
9570 which GCC support library should be included on the command line to the linker.
9573 Process the @code{startfile} spec. This is a spec for deciding which
9574 object files should be the first ones passed to the linker. Typically
9575 this might be a file named @file{crt0.o}.
9578 Process the @code{endfile} spec. This is a spec string that specifies
9579 the last object files that will be passed to the linker.
9582 Process the @code{cpp} spec. This is used to construct the arguments
9583 to be passed to the C preprocessor.
9586 Process the @code{cc1} spec. This is used to construct the options to be
9587 passed to the actual C compiler (@samp{cc1}).
9590 Process the @code{cc1plus} spec. This is used to construct the options to be
9591 passed to the actual C++ compiler (@samp{cc1plus}).
9594 Substitute the variable part of a matched option. See below.
9595 Note that each comma in the substituted string is replaced by
9599 Remove all occurrences of @code{-S} from the command line. Note---this
9600 command is position dependent. @samp{%} commands in the spec string
9601 before this one will see @code{-S}, @samp{%} commands in the spec string
9602 after this one will not.
9604 @item %:@var{function}(@var{args})
9605 Call the named function @var{function}, passing it @var{args}.
9606 @var{args} is first processed as a nested spec string, then split
9607 into an argument vector in the usual fashion. The function returns
9608 a string which is processed as if it had appeared literally as part
9609 of the current spec.
9611 The following built-in spec functions are provided:
9615 The @code{getenv} spec function takes two arguments: an environment
9616 variable name and a string. If the environment variable is not
9617 defined, a fatal error is issued. Otherwise, the return value is the
9618 value of the environment variable concatenated with the string. For
9619 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9622 %:getenv(TOPDIR /include)
9625 expands to @file{/path/to/top/include}.
9627 @item @code{if-exists}
9628 The @code{if-exists} spec function takes one argument, an absolute
9629 pathname to a file. If the file exists, @code{if-exists} returns the
9630 pathname. Here is a small example of its usage:
9634 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9637 @item @code{if-exists-else}
9638 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9639 spec function, except that it takes two arguments. The first argument is
9640 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9641 returns the pathname. If it does not exist, it returns the second argument.
9642 This way, @code{if-exists-else} can be used to select one file or another,
9643 based on the existence of the first. Here is a small example of its usage:
9647 crt0%O%s %:if-exists(crti%O%s) \
9648 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9651 @item @code{replace-outfile}
9652 The @code{replace-outfile} spec function takes two arguments. It looks for the
9653 first argument in the outfiles array and replaces it with the second argument. Here
9654 is a small example of its usage:
9657 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9660 @item @code{remove-outfile}
9661 The @code{remove-outfile} spec function takes one argument. It looks for the
9662 first argument in the outfiles array and removes it. Here is a small example
9666 %:remove-outfile(-lm)
9669 @item @code{print-asm-header}
9670 The @code{print-asm-header} function takes no arguments and simply
9671 prints a banner like:
9677 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9680 It is used to separate compiler options from assembler options
9681 in the @option{--target-help} output.
9685 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9686 If that switch was not specified, this substitutes nothing. Note that
9687 the leading dash is omitted when specifying this option, and it is
9688 automatically inserted if the substitution is performed. Thus the spec
9689 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9690 and would output the command line option @option{-foo}.
9692 @item %W@{@code{S}@}
9693 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9696 @item %@{@code{S}*@}
9697 Substitutes all the switches specified to GCC whose names start
9698 with @code{-S}, but which also take an argument. This is used for
9699 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9700 GCC considers @option{-o foo} as being
9701 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9702 text, including the space. Thus two arguments would be generated.
9704 @item %@{@code{S}*&@code{T}*@}
9705 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9706 (the order of @code{S} and @code{T} in the spec is not significant).
9707 There can be any number of ampersand-separated variables; for each the
9708 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9710 @item %@{@code{S}:@code{X}@}
9711 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9713 @item %@{!@code{S}:@code{X}@}
9714 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9716 @item %@{@code{S}*:@code{X}@}
9717 Substitutes @code{X} if one or more switches whose names start with
9718 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9719 once, no matter how many such switches appeared. However, if @code{%*}
9720 appears somewhere in @code{X}, then @code{X} will be substituted once
9721 for each matching switch, with the @code{%*} replaced by the part of
9722 that switch that matched the @code{*}.
9724 @item %@{.@code{S}:@code{X}@}
9725 Substitutes @code{X}, if processing a file with suffix @code{S}.
9727 @item %@{!.@code{S}:@code{X}@}
9728 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9730 @item %@{,@code{S}:@code{X}@}
9731 Substitutes @code{X}, if processing a file for language @code{S}.
9733 @item %@{!,@code{S}:@code{X}@}
9734 Substitutes @code{X}, if not processing a file for language @code{S}.
9736 @item %@{@code{S}|@code{P}:@code{X}@}
9737 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9738 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9739 @code{*} sequences as well, although they have a stronger binding than
9740 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9741 alternatives must be starred, and only the first matching alternative
9744 For example, a spec string like this:
9747 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9750 will output the following command-line options from the following input
9751 command-line options:
9756 -d fred.c -foo -baz -boggle
9757 -d jim.d -bar -baz -boggle
9760 @item %@{S:X; T:Y; :D@}
9762 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9763 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9764 be as many clauses as you need. This may be combined with @code{.},
9765 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9770 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9771 construct may contain other nested @samp{%} constructs or spaces, or
9772 even newlines. They are processed as usual, as described above.
9773 Trailing white space in @code{X} is ignored. White space may also
9774 appear anywhere on the left side of the colon in these constructs,
9775 except between @code{.} or @code{*} and the corresponding word.
9777 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9778 handled specifically in these constructs. If another value of
9779 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9780 @option{-W} switch is found later in the command line, the earlier
9781 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9782 just one letter, which passes all matching options.
9784 The character @samp{|} at the beginning of the predicate text is used to
9785 indicate that a command should be piped to the following command, but
9786 only if @option{-pipe} is specified.
9788 It is built into GCC which switches take arguments and which do not.
9789 (You might think it would be useful to generalize this to allow each
9790 compiler's spec to say which switches take arguments. But this cannot
9791 be done in a consistent fashion. GCC cannot even decide which input
9792 files have been specified without knowing which switches take arguments,
9793 and it must know which input files to compile in order to tell which
9796 GCC also knows implicitly that arguments starting in @option{-l} are to be
9797 treated as compiler output files, and passed to the linker in their
9798 proper position among the other output files.
9800 @c man begin OPTIONS
9802 @node Target Options
9803 @section Specifying Target Machine and Compiler Version
9804 @cindex target options
9805 @cindex cross compiling
9806 @cindex specifying machine version
9807 @cindex specifying compiler version and target machine
9808 @cindex compiler version, specifying
9809 @cindex target machine, specifying
9811 The usual way to run GCC is to run the executable called @file{gcc}, or
9812 @file{<machine>-gcc} when cross-compiling, or
9813 @file{<machine>-gcc-<version>} to run a version other than the one that
9816 @node Submodel Options
9817 @section Hardware Models and Configurations
9818 @cindex submodel options
9819 @cindex specifying hardware config
9820 @cindex hardware models and configurations, specifying
9821 @cindex machine dependent options
9823 Each target machine types can have its own
9824 special options, starting with @samp{-m}, to choose among various
9825 hardware models or configurations---for example, 68010 vs 68020,
9826 floating coprocessor or none. A single installed version of the
9827 compiler can compile for any model or configuration, according to the
9830 Some configurations of the compiler also support additional special
9831 options, usually for compatibility with other compilers on the same
9834 @c This list is ordered alphanumerically by subsection name.
9835 @c It should be the same order and spelling as these options are listed
9836 @c in Machine Dependent Options
9842 * Blackfin Options::
9846 * DEC Alpha Options::
9847 * DEC Alpha/VMS Options::
9850 * GNU/Linux Options::
9853 * i386 and x86-64 Options::
9854 * i386 and x86-64 Windows Options::
9856 * IA-64/VMS Options::
9864 * MicroBlaze Options::
9869 * picoChip Options::
9871 * RS/6000 and PowerPC Options::
9873 * S/390 and zSeries Options::
9876 * Solaris 2 Options::
9879 * System V Options::
9884 * Xstormy16 Options::
9890 @subsection ARC Options
9893 These options are defined for ARC implementations:
9898 Compile code for little endian mode. This is the default.
9902 Compile code for big endian mode.
9905 @opindex mmangle-cpu
9906 Prepend the name of the cpu to all public symbol names.
9907 In multiple-processor systems, there are many ARC variants with different
9908 instruction and register set characteristics. This flag prevents code
9909 compiled for one cpu to be linked with code compiled for another.
9910 No facility exists for handling variants that are ``almost identical''.
9911 This is an all or nothing option.
9913 @item -mcpu=@var{cpu}
9915 Compile code for ARC variant @var{cpu}.
9916 Which variants are supported depend on the configuration.
9917 All variants support @option{-mcpu=base}, this is the default.
9919 @item -mtext=@var{text-section}
9920 @itemx -mdata=@var{data-section}
9921 @itemx -mrodata=@var{readonly-data-section}
9925 Put functions, data, and readonly data in @var{text-section},
9926 @var{data-section}, and @var{readonly-data-section} respectively
9927 by default. This can be overridden with the @code{section} attribute.
9928 @xref{Variable Attributes}.
9933 @subsection ARM Options
9936 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9940 @item -mabi=@var{name}
9942 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9943 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9946 @opindex mapcs-frame
9947 Generate a stack frame that is compliant with the ARM Procedure Call
9948 Standard for all functions, even if this is not strictly necessary for
9949 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9950 with this option will cause the stack frames not to be generated for
9951 leaf functions. The default is @option{-mno-apcs-frame}.
9955 This is a synonym for @option{-mapcs-frame}.
9958 @c not currently implemented
9959 @item -mapcs-stack-check
9960 @opindex mapcs-stack-check
9961 Generate code to check the amount of stack space available upon entry to
9962 every function (that actually uses some stack space). If there is
9963 insufficient space available then either the function
9964 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9965 called, depending upon the amount of stack space required. The run time
9966 system is required to provide these functions. The default is
9967 @option{-mno-apcs-stack-check}, since this produces smaller code.
9969 @c not currently implemented
9971 @opindex mapcs-float
9972 Pass floating point arguments using the float point registers. This is
9973 one of the variants of the APCS@. This option is recommended if the
9974 target hardware has a floating point unit or if a lot of floating point
9975 arithmetic is going to be performed by the code. The default is
9976 @option{-mno-apcs-float}, since integer only code is slightly increased in
9977 size if @option{-mapcs-float} is used.
9979 @c not currently implemented
9980 @item -mapcs-reentrant
9981 @opindex mapcs-reentrant
9982 Generate reentrant, position independent code. The default is
9983 @option{-mno-apcs-reentrant}.
9986 @item -mthumb-interwork
9987 @opindex mthumb-interwork
9988 Generate code which supports calling between the ARM and Thumb
9989 instruction sets. Without this option the two instruction sets cannot
9990 be reliably used inside one program. The default is
9991 @option{-mno-thumb-interwork}, since slightly larger code is generated
9992 when @option{-mthumb-interwork} is specified.
9994 @item -mno-sched-prolog
9995 @opindex mno-sched-prolog
9996 Prevent the reordering of instructions in the function prolog, or the
9997 merging of those instruction with the instructions in the function's
9998 body. This means that all functions will start with a recognizable set
9999 of instructions (or in fact one of a choice from a small set of
10000 different function prologues), and this information can be used to
10001 locate the start if functions inside an executable piece of code. The
10002 default is @option{-msched-prolog}.
10004 @item -mfloat-abi=@var{name}
10005 @opindex mfloat-abi
10006 Specifies which floating-point ABI to use. Permissible values
10007 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10009 Specifying @samp{soft} causes GCC to generate output containing
10010 library calls for floating-point operations.
10011 @samp{softfp} allows the generation of code using hardware floating-point
10012 instructions, but still uses the soft-float calling conventions.
10013 @samp{hard} allows generation of floating-point instructions
10014 and uses FPU-specific calling conventions.
10016 The default depends on the specific target configuration. Note that
10017 the hard-float and soft-float ABIs are not link-compatible; you must
10018 compile your entire program with the same ABI, and link with a
10019 compatible set of libraries.
10022 @opindex mhard-float
10023 Equivalent to @option{-mfloat-abi=hard}.
10026 @opindex msoft-float
10027 Equivalent to @option{-mfloat-abi=soft}.
10029 @item -mlittle-endian
10030 @opindex mlittle-endian
10031 Generate code for a processor running in little-endian mode. This is
10032 the default for all standard configurations.
10035 @opindex mbig-endian
10036 Generate code for a processor running in big-endian mode; the default is
10037 to compile code for a little-endian processor.
10039 @item -mwords-little-endian
10040 @opindex mwords-little-endian
10041 This option only applies when generating code for big-endian processors.
10042 Generate code for a little-endian word order but a big-endian byte
10043 order. That is, a byte order of the form @samp{32107654}. Note: this
10044 option should only be used if you require compatibility with code for
10045 big-endian ARM processors generated by versions of the compiler prior to
10048 @item -mcpu=@var{name}
10050 This specifies the name of the target ARM processor. GCC uses this name
10051 to determine what kind of instructions it can emit when generating
10052 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10053 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10054 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10055 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10056 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10058 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10059 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10060 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10061 @samp{strongarm1110},
10062 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10063 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10064 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10065 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10066 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10067 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10068 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10069 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10070 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10073 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10075 @item -mtune=@var{name}
10077 This option is very similar to the @option{-mcpu=} option, except that
10078 instead of specifying the actual target processor type, and hence
10079 restricting which instructions can be used, it specifies that GCC should
10080 tune the performance of the code as if the target were of the type
10081 specified in this option, but still choosing the instructions that it
10082 will generate based on the cpu specified by a @option{-mcpu=} option.
10083 For some ARM implementations better performance can be obtained by using
10086 @item -march=@var{name}
10088 This specifies the name of the target ARM architecture. GCC uses this
10089 name to determine what kind of instructions it can emit when generating
10090 assembly code. This option can be used in conjunction with or instead
10091 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10092 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10093 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10094 @samp{armv6}, @samp{armv6j},
10095 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10096 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10097 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10099 @item -mfpu=@var{name}
10100 @itemx -mfpe=@var{number}
10101 @itemx -mfp=@var{number}
10105 This specifies what floating point hardware (or hardware emulation) is
10106 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10107 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10108 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10109 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10110 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10111 @option{-mfp} and @option{-mfpe} are synonyms for
10112 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10115 If @option{-msoft-float} is specified this specifies the format of
10116 floating point values.
10118 If the selected floating-point hardware includes the NEON extension
10119 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10120 operations will not be used by GCC's auto-vectorization pass unless
10121 @option{-funsafe-math-optimizations} is also specified. This is
10122 because NEON hardware does not fully implement the IEEE 754 standard for
10123 floating-point arithmetic (in particular denormal values are treated as
10124 zero), so the use of NEON instructions may lead to a loss of precision.
10126 @item -mfp16-format=@var{name}
10127 @opindex mfp16-format
10128 Specify the format of the @code{__fp16} half-precision floating-point type.
10129 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10130 the default is @samp{none}, in which case the @code{__fp16} type is not
10131 defined. @xref{Half-Precision}, for more information.
10133 @item -mstructure-size-boundary=@var{n}
10134 @opindex mstructure-size-boundary
10135 The size of all structures and unions will be rounded up to a multiple
10136 of the number of bits set by this option. Permissible values are 8, 32
10137 and 64. The default value varies for different toolchains. For the COFF
10138 targeted toolchain the default value is 8. A value of 64 is only allowed
10139 if the underlying ABI supports it.
10141 Specifying the larger number can produce faster, more efficient code, but
10142 can also increase the size of the program. Different values are potentially
10143 incompatible. Code compiled with one value cannot necessarily expect to
10144 work with code or libraries compiled with another value, if they exchange
10145 information using structures or unions.
10147 @item -mabort-on-noreturn
10148 @opindex mabort-on-noreturn
10149 Generate a call to the function @code{abort} at the end of a
10150 @code{noreturn} function. It will be executed if the function tries to
10154 @itemx -mno-long-calls
10155 @opindex mlong-calls
10156 @opindex mno-long-calls
10157 Tells the compiler to perform function calls by first loading the
10158 address of the function into a register and then performing a subroutine
10159 call on this register. This switch is needed if the target function
10160 will lie outside of the 64 megabyte addressing range of the offset based
10161 version of subroutine call instruction.
10163 Even if this switch is enabled, not all function calls will be turned
10164 into long calls. The heuristic is that static functions, functions
10165 which have the @samp{short-call} attribute, functions that are inside
10166 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10167 definitions have already been compiled within the current compilation
10168 unit, will not be turned into long calls. The exception to this rule is
10169 that weak function definitions, functions with the @samp{long-call}
10170 attribute or the @samp{section} attribute, and functions that are within
10171 the scope of a @samp{#pragma long_calls} directive, will always be
10172 turned into long calls.
10174 This feature is not enabled by default. Specifying
10175 @option{-mno-long-calls} will restore the default behavior, as will
10176 placing the function calls within the scope of a @samp{#pragma
10177 long_calls_off} directive. Note these switches have no effect on how
10178 the compiler generates code to handle function calls via function
10181 @item -msingle-pic-base
10182 @opindex msingle-pic-base
10183 Treat the register used for PIC addressing as read-only, rather than
10184 loading it in the prologue for each function. The run-time system is
10185 responsible for initializing this register with an appropriate value
10186 before execution begins.
10188 @item -mpic-register=@var{reg}
10189 @opindex mpic-register
10190 Specify the register to be used for PIC addressing. The default is R10
10191 unless stack-checking is enabled, when R9 is used.
10193 @item -mcirrus-fix-invalid-insns
10194 @opindex mcirrus-fix-invalid-insns
10195 @opindex mno-cirrus-fix-invalid-insns
10196 Insert NOPs into the instruction stream to in order to work around
10197 problems with invalid Maverick instruction combinations. This option
10198 is only valid if the @option{-mcpu=ep9312} option has been used to
10199 enable generation of instructions for the Cirrus Maverick floating
10200 point co-processor. This option is not enabled by default, since the
10201 problem is only present in older Maverick implementations. The default
10202 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10205 @item -mpoke-function-name
10206 @opindex mpoke-function-name
10207 Write the name of each function into the text section, directly
10208 preceding the function prologue. The generated code is similar to this:
10212 .ascii "arm_poke_function_name", 0
10215 .word 0xff000000 + (t1 - t0)
10216 arm_poke_function_name
10218 stmfd sp!, @{fp, ip, lr, pc@}
10222 When performing a stack backtrace, code can inspect the value of
10223 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10224 location @code{pc - 12} and the top 8 bits are set, then we know that
10225 there is a function name embedded immediately preceding this location
10226 and has length @code{((pc[-3]) & 0xff000000)}.
10230 Generate code for the Thumb instruction set. The default is to
10231 use the 32-bit ARM instruction set.
10232 This option automatically enables either 16-bit Thumb-1 or
10233 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10234 and @option{-march=@var{name}} options. This option is not passed to the
10235 assembler. If you want to force assembler files to be interpreted as Thumb code,
10236 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10237 option directly to the assembler by prefixing it with @option{-Wa}.
10240 @opindex mtpcs-frame
10241 Generate a stack frame that is compliant with the Thumb Procedure Call
10242 Standard for all non-leaf functions. (A leaf function is one that does
10243 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10245 @item -mtpcs-leaf-frame
10246 @opindex mtpcs-leaf-frame
10247 Generate a stack frame that is compliant with the Thumb Procedure Call
10248 Standard for all leaf functions. (A leaf function is one that does
10249 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10251 @item -mcallee-super-interworking
10252 @opindex mcallee-super-interworking
10253 Gives all externally visible functions in the file being compiled an ARM
10254 instruction set header which switches to Thumb mode before executing the
10255 rest of the function. This allows these functions to be called from
10256 non-interworking code. This option is not valid in AAPCS configurations
10257 because interworking is enabled by default.
10259 @item -mcaller-super-interworking
10260 @opindex mcaller-super-interworking
10261 Allows calls via function pointers (including virtual functions) to
10262 execute correctly regardless of whether the target code has been
10263 compiled for interworking or not. There is a small overhead in the cost
10264 of executing a function pointer if this option is enabled. This option
10265 is not valid in AAPCS configurations because interworking is enabled
10268 @item -mtp=@var{name}
10270 Specify the access model for the thread local storage pointer. The valid
10271 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10272 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10273 (supported in the arm6k architecture), and @option{auto}, which uses the
10274 best available method for the selected processor. The default setting is
10277 @item -mword-relocations
10278 @opindex mword-relocations
10279 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10280 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10281 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10284 @item -mfix-cortex-m3-ldrd
10285 @opindex mfix-cortex-m3-ldrd
10286 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10287 with overlapping destination and base registers are used. This option avoids
10288 generating these instructions. This option is enabled by default when
10289 @option{-mcpu=cortex-m3} is specified.
10294 @subsection AVR Options
10295 @cindex AVR Options
10297 These options are defined for AVR implementations:
10300 @item -mmcu=@var{mcu}
10302 Specify ATMEL AVR instruction set or MCU type.
10304 Instruction set avr1 is for the minimal AVR core, not supported by the C
10305 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10306 attiny11, attiny12, attiny15, attiny28).
10308 Instruction set avr2 (default) is for the classic AVR core with up to
10309 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10310 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10311 at90c8534, at90s8535).
10313 Instruction set avr3 is for the classic AVR core with up to 128K program
10314 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10316 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10317 memory space (MCU types: atmega8, atmega83, atmega85).
10319 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10320 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10321 atmega64, atmega128, at43usb355, at94k).
10323 @item -mno-interrupts
10324 @opindex mno-interrupts
10325 Generated code is not compatible with hardware interrupts.
10326 Code size will be smaller.
10328 @item -mcall-prologues
10329 @opindex mcall-prologues
10330 Functions prologues/epilogues expanded as call to appropriate
10331 subroutines. Code size will be smaller.
10334 @opindex mtiny-stack
10335 Change only the low 8 bits of the stack pointer.
10339 Assume int to be 8 bit integer. This affects the sizes of all types: A
10340 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10341 and long long will be 4 bytes. Please note that this option does not
10342 comply to the C standards, but it will provide you with smaller code
10346 @node Blackfin Options
10347 @subsection Blackfin Options
10348 @cindex Blackfin Options
10351 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10353 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10354 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10355 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10356 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10357 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10358 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10359 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10361 The optional @var{sirevision} specifies the silicon revision of the target
10362 Blackfin processor. Any workarounds available for the targeted silicon revision
10363 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10364 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10365 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10366 hexadecimal digits representing the major and minor numbers in the silicon
10367 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10368 is not defined. If @var{sirevision} is @samp{any}, the
10369 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10370 If this optional @var{sirevision} is not used, GCC assumes the latest known
10371 silicon revision of the targeted Blackfin processor.
10373 Support for @samp{bf561} is incomplete. For @samp{bf561},
10374 Only the processor macro is defined.
10375 Without this option, @samp{bf532} is used as the processor by default.
10376 The corresponding predefined processor macros for @var{cpu} is to
10377 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10378 provided by libgloss to be linked in if @option{-msim} is not given.
10382 Specifies that the program will be run on the simulator. This causes
10383 the simulator BSP provided by libgloss to be linked in. This option
10384 has effect only for @samp{bfin-elf} toolchain.
10385 Certain other options, such as @option{-mid-shared-library} and
10386 @option{-mfdpic}, imply @option{-msim}.
10388 @item -momit-leaf-frame-pointer
10389 @opindex momit-leaf-frame-pointer
10390 Don't keep the frame pointer in a register for leaf functions. This
10391 avoids the instructions to save, set up and restore frame pointers and
10392 makes an extra register available in leaf functions. The option
10393 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10394 which might make debugging harder.
10396 @item -mspecld-anomaly
10397 @opindex mspecld-anomaly
10398 When enabled, the compiler will ensure that the generated code does not
10399 contain speculative loads after jump instructions. If this option is used,
10400 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10402 @item -mno-specld-anomaly
10403 @opindex mno-specld-anomaly
10404 Don't generate extra code to prevent speculative loads from occurring.
10406 @item -mcsync-anomaly
10407 @opindex mcsync-anomaly
10408 When enabled, the compiler will ensure that the generated code does not
10409 contain CSYNC or SSYNC instructions too soon after conditional branches.
10410 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10412 @item -mno-csync-anomaly
10413 @opindex mno-csync-anomaly
10414 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10415 occurring too soon after a conditional branch.
10419 When enabled, the compiler is free to take advantage of the knowledge that
10420 the entire program fits into the low 64k of memory.
10423 @opindex mno-low-64k
10424 Assume that the program is arbitrarily large. This is the default.
10426 @item -mstack-check-l1
10427 @opindex mstack-check-l1
10428 Do stack checking using information placed into L1 scratchpad memory by the
10431 @item -mid-shared-library
10432 @opindex mid-shared-library
10433 Generate code that supports shared libraries via the library ID method.
10434 This allows for execute in place and shared libraries in an environment
10435 without virtual memory management. This option implies @option{-fPIC}.
10436 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10438 @item -mno-id-shared-library
10439 @opindex mno-id-shared-library
10440 Generate code that doesn't assume ID based shared libraries are being used.
10441 This is the default.
10443 @item -mleaf-id-shared-library
10444 @opindex mleaf-id-shared-library
10445 Generate code that supports shared libraries via the library ID method,
10446 but assumes that this library or executable won't link against any other
10447 ID shared libraries. That allows the compiler to use faster code for jumps
10450 @item -mno-leaf-id-shared-library
10451 @opindex mno-leaf-id-shared-library
10452 Do not assume that the code being compiled won't link against any ID shared
10453 libraries. Slower code will be generated for jump and call insns.
10455 @item -mshared-library-id=n
10456 @opindex mshared-library-id
10457 Specified the identification number of the ID based shared library being
10458 compiled. Specifying a value of 0 will generate more compact code, specifying
10459 other values will force the allocation of that number to the current
10460 library but is no more space or time efficient than omitting this option.
10464 Generate code that allows the data segment to be located in a different
10465 area of memory from the text segment. This allows for execute in place in
10466 an environment without virtual memory management by eliminating relocations
10467 against the text section.
10469 @item -mno-sep-data
10470 @opindex mno-sep-data
10471 Generate code that assumes that the data segment follows the text segment.
10472 This is the default.
10475 @itemx -mno-long-calls
10476 @opindex mlong-calls
10477 @opindex mno-long-calls
10478 Tells the compiler to perform function calls by first loading the
10479 address of the function into a register and then performing a subroutine
10480 call on this register. This switch is needed if the target function
10481 will lie outside of the 24 bit addressing range of the offset based
10482 version of subroutine call instruction.
10484 This feature is not enabled by default. Specifying
10485 @option{-mno-long-calls} will restore the default behavior. Note these
10486 switches have no effect on how the compiler generates code to handle
10487 function calls via function pointers.
10491 Link with the fast floating-point library. This library relaxes some of
10492 the IEEE floating-point standard's rules for checking inputs against
10493 Not-a-Number (NAN), in the interest of performance.
10496 @opindex minline-plt
10497 Enable inlining of PLT entries in function calls to functions that are
10498 not known to bind locally. It has no effect without @option{-mfdpic}.
10501 @opindex mmulticore
10502 Build standalone application for multicore Blackfin processor. Proper
10503 start files and link scripts will be used to support multicore.
10504 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10505 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10506 @option{-mcorea} or @option{-mcoreb}. If it's used without
10507 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10508 programming model is used. In this model, the main function of Core B
10509 should be named as coreb_main. If it's used with @option{-mcorea} or
10510 @option{-mcoreb}, one application per core programming model is used.
10511 If this option is not used, single core application programming
10516 Build standalone application for Core A of BF561 when using
10517 one application per core programming model. Proper start files
10518 and link scripts will be used to support Core A. This option
10519 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10523 Build standalone application for Core B of BF561 when using
10524 one application per core programming model. Proper start files
10525 and link scripts will be used to support Core B. This option
10526 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10527 should be used instead of main. It must be used with
10528 @option{-mmulticore}.
10532 Build standalone application for SDRAM. Proper start files and
10533 link scripts will be used to put the application into SDRAM.
10534 Loader should initialize SDRAM before loading the application
10535 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10539 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10540 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10541 are enabled; for standalone applications the default is off.
10545 @subsection CRIS Options
10546 @cindex CRIS Options
10548 These options are defined specifically for the CRIS ports.
10551 @item -march=@var{architecture-type}
10552 @itemx -mcpu=@var{architecture-type}
10555 Generate code for the specified architecture. The choices for
10556 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10557 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10558 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10561 @item -mtune=@var{architecture-type}
10563 Tune to @var{architecture-type} everything applicable about the generated
10564 code, except for the ABI and the set of available instructions. The
10565 choices for @var{architecture-type} are the same as for
10566 @option{-march=@var{architecture-type}}.
10568 @item -mmax-stack-frame=@var{n}
10569 @opindex mmax-stack-frame
10570 Warn when the stack frame of a function exceeds @var{n} bytes.
10576 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10577 @option{-march=v3} and @option{-march=v8} respectively.
10579 @item -mmul-bug-workaround
10580 @itemx -mno-mul-bug-workaround
10581 @opindex mmul-bug-workaround
10582 @opindex mno-mul-bug-workaround
10583 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10584 models where it applies. This option is active by default.
10588 Enable CRIS-specific verbose debug-related information in the assembly
10589 code. This option also has the effect to turn off the @samp{#NO_APP}
10590 formatted-code indicator to the assembler at the beginning of the
10595 Do not use condition-code results from previous instruction; always emit
10596 compare and test instructions before use of condition codes.
10598 @item -mno-side-effects
10599 @opindex mno-side-effects
10600 Do not emit instructions with side-effects in addressing modes other than
10603 @item -mstack-align
10604 @itemx -mno-stack-align
10605 @itemx -mdata-align
10606 @itemx -mno-data-align
10607 @itemx -mconst-align
10608 @itemx -mno-const-align
10609 @opindex mstack-align
10610 @opindex mno-stack-align
10611 @opindex mdata-align
10612 @opindex mno-data-align
10613 @opindex mconst-align
10614 @opindex mno-const-align
10615 These options (no-options) arranges (eliminate arrangements) for the
10616 stack-frame, individual data and constants to be aligned for the maximum
10617 single data access size for the chosen CPU model. The default is to
10618 arrange for 32-bit alignment. ABI details such as structure layout are
10619 not affected by these options.
10627 Similar to the stack- data- and const-align options above, these options
10628 arrange for stack-frame, writable data and constants to all be 32-bit,
10629 16-bit or 8-bit aligned. The default is 32-bit alignment.
10631 @item -mno-prologue-epilogue
10632 @itemx -mprologue-epilogue
10633 @opindex mno-prologue-epilogue
10634 @opindex mprologue-epilogue
10635 With @option{-mno-prologue-epilogue}, the normal function prologue and
10636 epilogue that sets up the stack-frame are omitted and no return
10637 instructions or return sequences are generated in the code. Use this
10638 option only together with visual inspection of the compiled code: no
10639 warnings or errors are generated when call-saved registers must be saved,
10640 or storage for local variable needs to be allocated.
10644 @opindex mno-gotplt
10646 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10647 instruction sequences that load addresses for functions from the PLT part
10648 of the GOT rather than (traditional on other architectures) calls to the
10649 PLT@. The default is @option{-mgotplt}.
10653 Legacy no-op option only recognized with the cris-axis-elf and
10654 cris-axis-linux-gnu targets.
10658 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10662 This option, recognized for the cris-axis-elf arranges
10663 to link with input-output functions from a simulator library. Code,
10664 initialized data and zero-initialized data are allocated consecutively.
10668 Like @option{-sim}, but pass linker options to locate initialized data at
10669 0x40000000 and zero-initialized data at 0x80000000.
10673 @subsection CRX Options
10674 @cindex CRX Options
10676 These options are defined specifically for the CRX ports.
10682 Enable the use of multiply-accumulate instructions. Disabled by default.
10685 @opindex mpush-args
10686 Push instructions will be used to pass outgoing arguments when functions
10687 are called. Enabled by default.
10690 @node Darwin Options
10691 @subsection Darwin Options
10692 @cindex Darwin options
10694 These options are defined for all architectures running the Darwin operating
10697 FSF GCC on Darwin does not create ``fat'' object files; it will create
10698 an object file for the single architecture that it was built to
10699 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10700 @option{-arch} options are used; it does so by running the compiler or
10701 linker multiple times and joining the results together with
10704 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10705 @samp{i686}) is determined by the flags that specify the ISA
10706 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10707 @option{-force_cpusubtype_ALL} option can be used to override this.
10709 The Darwin tools vary in their behavior when presented with an ISA
10710 mismatch. The assembler, @file{as}, will only permit instructions to
10711 be used that are valid for the subtype of the file it is generating,
10712 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10713 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10714 and print an error if asked to create a shared library with a less
10715 restrictive subtype than its input files (for instance, trying to put
10716 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10717 for executables, @file{ld}, will quietly give the executable the most
10718 restrictive subtype of any of its input files.
10723 Add the framework directory @var{dir} to the head of the list of
10724 directories to be searched for header files. These directories are
10725 interleaved with those specified by @option{-I} options and are
10726 scanned in a left-to-right order.
10728 A framework directory is a directory with frameworks in it. A
10729 framework is a directory with a @samp{"Headers"} and/or
10730 @samp{"PrivateHeaders"} directory contained directly in it that ends
10731 in @samp{".framework"}. The name of a framework is the name of this
10732 directory excluding the @samp{".framework"}. Headers associated with
10733 the framework are found in one of those two directories, with
10734 @samp{"Headers"} being searched first. A subframework is a framework
10735 directory that is in a framework's @samp{"Frameworks"} directory.
10736 Includes of subframework headers can only appear in a header of a
10737 framework that contains the subframework, or in a sibling subframework
10738 header. Two subframeworks are siblings if they occur in the same
10739 framework. A subframework should not have the same name as a
10740 framework, a warning will be issued if this is violated. Currently a
10741 subframework cannot have subframeworks, in the future, the mechanism
10742 may be extended to support this. The standard frameworks can be found
10743 in @samp{"/System/Library/Frameworks"} and
10744 @samp{"/Library/Frameworks"}. An example include looks like
10745 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10746 the name of the framework and header.h is found in the
10747 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10749 @item -iframework@var{dir}
10750 @opindex iframework
10751 Like @option{-F} except the directory is a treated as a system
10752 directory. The main difference between this @option{-iframework} and
10753 @option{-F} is that with @option{-iframework} the compiler does not
10754 warn about constructs contained within header files found via
10755 @var{dir}. This option is valid only for the C family of languages.
10759 Emit debugging information for symbols that are used. For STABS
10760 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10761 This is by default ON@.
10765 Emit debugging information for all symbols and types.
10767 @item -mmacosx-version-min=@var{version}
10768 The earliest version of MacOS X that this executable will run on
10769 is @var{version}. Typical values of @var{version} include @code{10.1},
10770 @code{10.2}, and @code{10.3.9}.
10772 If the compiler was built to use the system's headers by default,
10773 then the default for this option is the system version on which the
10774 compiler is running, otherwise the default is to make choices which
10775 are compatible with as many systems and code bases as possible.
10779 Enable kernel development mode. The @option{-mkernel} option sets
10780 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10781 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10782 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10783 applicable. This mode also sets @option{-mno-altivec},
10784 @option{-msoft-float}, @option{-fno-builtin} and
10785 @option{-mlong-branch} for PowerPC targets.
10787 @item -mone-byte-bool
10788 @opindex mone-byte-bool
10789 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10790 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10791 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10792 option has no effect on x86.
10794 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10795 to generate code that is not binary compatible with code generated
10796 without that switch. Using this switch may require recompiling all
10797 other modules in a program, including system libraries. Use this
10798 switch to conform to a non-default data model.
10800 @item -mfix-and-continue
10801 @itemx -ffix-and-continue
10802 @itemx -findirect-data
10803 @opindex mfix-and-continue
10804 @opindex ffix-and-continue
10805 @opindex findirect-data
10806 Generate code suitable for fast turn around development. Needed to
10807 enable gdb to dynamically load @code{.o} files into already running
10808 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10809 are provided for backwards compatibility.
10813 Loads all members of static archive libraries.
10814 See man ld(1) for more information.
10816 @item -arch_errors_fatal
10817 @opindex arch_errors_fatal
10818 Cause the errors having to do with files that have the wrong architecture
10821 @item -bind_at_load
10822 @opindex bind_at_load
10823 Causes the output file to be marked such that the dynamic linker will
10824 bind all undefined references when the file is loaded or launched.
10828 Produce a Mach-o bundle format file.
10829 See man ld(1) for more information.
10831 @item -bundle_loader @var{executable}
10832 @opindex bundle_loader
10833 This option specifies the @var{executable} that will be loading the build
10834 output file being linked. See man ld(1) for more information.
10837 @opindex dynamiclib
10838 When passed this option, GCC will produce a dynamic library instead of
10839 an executable when linking, using the Darwin @file{libtool} command.
10841 @item -force_cpusubtype_ALL
10842 @opindex force_cpusubtype_ALL
10843 This causes GCC's output file to have the @var{ALL} subtype, instead of
10844 one controlled by the @option{-mcpu} or @option{-march} option.
10846 @item -allowable_client @var{client_name}
10847 @itemx -client_name
10848 @itemx -compatibility_version
10849 @itemx -current_version
10851 @itemx -dependency-file
10853 @itemx -dylinker_install_name
10855 @itemx -exported_symbols_list
10858 @itemx -flat_namespace
10859 @itemx -force_flat_namespace
10860 @itemx -headerpad_max_install_names
10863 @itemx -install_name
10864 @itemx -keep_private_externs
10865 @itemx -multi_module
10866 @itemx -multiply_defined
10867 @itemx -multiply_defined_unused
10870 @itemx -no_dead_strip_inits_and_terms
10871 @itemx -nofixprebinding
10872 @itemx -nomultidefs
10874 @itemx -noseglinkedit
10875 @itemx -pagezero_size
10877 @itemx -prebind_all_twolevel_modules
10878 @itemx -private_bundle
10880 @itemx -read_only_relocs
10882 @itemx -sectobjectsymbols
10886 @itemx -sectobjectsymbols
10889 @itemx -segs_read_only_addr
10891 @itemx -segs_read_write_addr
10892 @itemx -seg_addr_table
10893 @itemx -seg_addr_table_filename
10894 @itemx -seglinkedit
10896 @itemx -segs_read_only_addr
10897 @itemx -segs_read_write_addr
10898 @itemx -single_module
10900 @itemx -sub_library
10902 @itemx -sub_umbrella
10903 @itemx -twolevel_namespace
10906 @itemx -unexported_symbols_list
10907 @itemx -weak_reference_mismatches
10908 @itemx -whatsloaded
10909 @opindex allowable_client
10910 @opindex client_name
10911 @opindex compatibility_version
10912 @opindex current_version
10913 @opindex dead_strip
10914 @opindex dependency-file
10915 @opindex dylib_file
10916 @opindex dylinker_install_name
10918 @opindex exported_symbols_list
10920 @opindex flat_namespace
10921 @opindex force_flat_namespace
10922 @opindex headerpad_max_install_names
10923 @opindex image_base
10925 @opindex install_name
10926 @opindex keep_private_externs
10927 @opindex multi_module
10928 @opindex multiply_defined
10929 @opindex multiply_defined_unused
10930 @opindex noall_load
10931 @opindex no_dead_strip_inits_and_terms
10932 @opindex nofixprebinding
10933 @opindex nomultidefs
10935 @opindex noseglinkedit
10936 @opindex pagezero_size
10938 @opindex prebind_all_twolevel_modules
10939 @opindex private_bundle
10940 @opindex read_only_relocs
10942 @opindex sectobjectsymbols
10945 @opindex sectcreate
10946 @opindex sectobjectsymbols
10949 @opindex segs_read_only_addr
10950 @opindex segs_read_write_addr
10951 @opindex seg_addr_table
10952 @opindex seg_addr_table_filename
10953 @opindex seglinkedit
10955 @opindex segs_read_only_addr
10956 @opindex segs_read_write_addr
10957 @opindex single_module
10959 @opindex sub_library
10960 @opindex sub_umbrella
10961 @opindex twolevel_namespace
10964 @opindex unexported_symbols_list
10965 @opindex weak_reference_mismatches
10966 @opindex whatsloaded
10967 These options are passed to the Darwin linker. The Darwin linker man page
10968 describes them in detail.
10971 @node DEC Alpha Options
10972 @subsection DEC Alpha Options
10974 These @samp{-m} options are defined for the DEC Alpha implementations:
10977 @item -mno-soft-float
10978 @itemx -msoft-float
10979 @opindex mno-soft-float
10980 @opindex msoft-float
10981 Use (do not use) the hardware floating-point instructions for
10982 floating-point operations. When @option{-msoft-float} is specified,
10983 functions in @file{libgcc.a} will be used to perform floating-point
10984 operations. Unless they are replaced by routines that emulate the
10985 floating-point operations, or compiled in such a way as to call such
10986 emulations routines, these routines will issue floating-point
10987 operations. If you are compiling for an Alpha without floating-point
10988 operations, you must ensure that the library is built so as not to call
10991 Note that Alpha implementations without floating-point operations are
10992 required to have floating-point registers.
10995 @itemx -mno-fp-regs
10997 @opindex mno-fp-regs
10998 Generate code that uses (does not use) the floating-point register set.
10999 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11000 register set is not used, floating point operands are passed in integer
11001 registers as if they were integers and floating-point results are passed
11002 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11003 so any function with a floating-point argument or return value called by code
11004 compiled with @option{-mno-fp-regs} must also be compiled with that
11007 A typical use of this option is building a kernel that does not use,
11008 and hence need not save and restore, any floating-point registers.
11012 The Alpha architecture implements floating-point hardware optimized for
11013 maximum performance. It is mostly compliant with the IEEE floating
11014 point standard. However, for full compliance, software assistance is
11015 required. This option generates code fully IEEE compliant code
11016 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11017 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11018 defined during compilation. The resulting code is less efficient but is
11019 able to correctly support denormalized numbers and exceptional IEEE
11020 values such as not-a-number and plus/minus infinity. Other Alpha
11021 compilers call this option @option{-ieee_with_no_inexact}.
11023 @item -mieee-with-inexact
11024 @opindex mieee-with-inexact
11025 This is like @option{-mieee} except the generated code also maintains
11026 the IEEE @var{inexact-flag}. Turning on this option causes the
11027 generated code to implement fully-compliant IEEE math. In addition to
11028 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11029 macro. On some Alpha implementations the resulting code may execute
11030 significantly slower than the code generated by default. Since there is
11031 very little code that depends on the @var{inexact-flag}, you should
11032 normally not specify this option. Other Alpha compilers call this
11033 option @option{-ieee_with_inexact}.
11035 @item -mfp-trap-mode=@var{trap-mode}
11036 @opindex mfp-trap-mode
11037 This option controls what floating-point related traps are enabled.
11038 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11039 The trap mode can be set to one of four values:
11043 This is the default (normal) setting. The only traps that are enabled
11044 are the ones that cannot be disabled in software (e.g., division by zero
11048 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11052 Like @samp{u}, but the instructions are marked to be safe for software
11053 completion (see Alpha architecture manual for details).
11056 Like @samp{su}, but inexact traps are enabled as well.
11059 @item -mfp-rounding-mode=@var{rounding-mode}
11060 @opindex mfp-rounding-mode
11061 Selects the IEEE rounding mode. Other Alpha compilers call this option
11062 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11067 Normal IEEE rounding mode. Floating point numbers are rounded towards
11068 the nearest machine number or towards the even machine number in case
11072 Round towards minus infinity.
11075 Chopped rounding mode. Floating point numbers are rounded towards zero.
11078 Dynamic rounding mode. A field in the floating point control register
11079 (@var{fpcr}, see Alpha architecture reference manual) controls the
11080 rounding mode in effect. The C library initializes this register for
11081 rounding towards plus infinity. Thus, unless your program modifies the
11082 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11085 @item -mtrap-precision=@var{trap-precision}
11086 @opindex mtrap-precision
11087 In the Alpha architecture, floating point traps are imprecise. This
11088 means without software assistance it is impossible to recover from a
11089 floating trap and program execution normally needs to be terminated.
11090 GCC can generate code that can assist operating system trap handlers
11091 in determining the exact location that caused a floating point trap.
11092 Depending on the requirements of an application, different levels of
11093 precisions can be selected:
11097 Program precision. This option is the default and means a trap handler
11098 can only identify which program caused a floating point exception.
11101 Function precision. The trap handler can determine the function that
11102 caused a floating point exception.
11105 Instruction precision. The trap handler can determine the exact
11106 instruction that caused a floating point exception.
11109 Other Alpha compilers provide the equivalent options called
11110 @option{-scope_safe} and @option{-resumption_safe}.
11112 @item -mieee-conformant
11113 @opindex mieee-conformant
11114 This option marks the generated code as IEEE conformant. You must not
11115 use this option unless you also specify @option{-mtrap-precision=i} and either
11116 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11117 is to emit the line @samp{.eflag 48} in the function prologue of the
11118 generated assembly file. Under DEC Unix, this has the effect that
11119 IEEE-conformant math library routines will be linked in.
11121 @item -mbuild-constants
11122 @opindex mbuild-constants
11123 Normally GCC examines a 32- or 64-bit integer constant to
11124 see if it can construct it from smaller constants in two or three
11125 instructions. If it cannot, it will output the constant as a literal and
11126 generate code to load it from the data segment at runtime.
11128 Use this option to require GCC to construct @emph{all} integer constants
11129 using code, even if it takes more instructions (the maximum is six).
11131 You would typically use this option to build a shared library dynamic
11132 loader. Itself a shared library, it must relocate itself in memory
11133 before it can find the variables and constants in its own data segment.
11139 Select whether to generate code to be assembled by the vendor-supplied
11140 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11158 Indicate whether GCC should generate code to use the optional BWX,
11159 CIX, FIX and MAX instruction sets. The default is to use the instruction
11160 sets supported by the CPU type specified via @option{-mcpu=} option or that
11161 of the CPU on which GCC was built if none was specified.
11164 @itemx -mfloat-ieee
11165 @opindex mfloat-vax
11166 @opindex mfloat-ieee
11167 Generate code that uses (does not use) VAX F and G floating point
11168 arithmetic instead of IEEE single and double precision.
11170 @item -mexplicit-relocs
11171 @itemx -mno-explicit-relocs
11172 @opindex mexplicit-relocs
11173 @opindex mno-explicit-relocs
11174 Older Alpha assemblers provided no way to generate symbol relocations
11175 except via assembler macros. Use of these macros does not allow
11176 optimal instruction scheduling. GNU binutils as of version 2.12
11177 supports a new syntax that allows the compiler to explicitly mark
11178 which relocations should apply to which instructions. This option
11179 is mostly useful for debugging, as GCC detects the capabilities of
11180 the assembler when it is built and sets the default accordingly.
11183 @itemx -mlarge-data
11184 @opindex msmall-data
11185 @opindex mlarge-data
11186 When @option{-mexplicit-relocs} is in effect, static data is
11187 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11188 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11189 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11190 16-bit relocations off of the @code{$gp} register. This limits the
11191 size of the small data area to 64KB, but allows the variables to be
11192 directly accessed via a single instruction.
11194 The default is @option{-mlarge-data}. With this option the data area
11195 is limited to just below 2GB@. Programs that require more than 2GB of
11196 data must use @code{malloc} or @code{mmap} to allocate the data in the
11197 heap instead of in the program's data segment.
11199 When generating code for shared libraries, @option{-fpic} implies
11200 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11203 @itemx -mlarge-text
11204 @opindex msmall-text
11205 @opindex mlarge-text
11206 When @option{-msmall-text} is used, the compiler assumes that the
11207 code of the entire program (or shared library) fits in 4MB, and is
11208 thus reachable with a branch instruction. When @option{-msmall-data}
11209 is used, the compiler can assume that all local symbols share the
11210 same @code{$gp} value, and thus reduce the number of instructions
11211 required for a function call from 4 to 1.
11213 The default is @option{-mlarge-text}.
11215 @item -mcpu=@var{cpu_type}
11217 Set the instruction set and instruction scheduling parameters for
11218 machine type @var{cpu_type}. You can specify either the @samp{EV}
11219 style name or the corresponding chip number. GCC supports scheduling
11220 parameters for the EV4, EV5 and EV6 family of processors and will
11221 choose the default values for the instruction set from the processor
11222 you specify. If you do not specify a processor type, GCC will default
11223 to the processor on which the compiler was built.
11225 Supported values for @var{cpu_type} are
11231 Schedules as an EV4 and has no instruction set extensions.
11235 Schedules as an EV5 and has no instruction set extensions.
11239 Schedules as an EV5 and supports the BWX extension.
11244 Schedules as an EV5 and supports the BWX and MAX extensions.
11248 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11252 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11255 Native Linux/GNU toolchains also support the value @samp{native},
11256 which selects the best architecture option for the host processor.
11257 @option{-mcpu=native} has no effect if GCC does not recognize
11260 @item -mtune=@var{cpu_type}
11262 Set only the instruction scheduling parameters for machine type
11263 @var{cpu_type}. The instruction set is not changed.
11265 Native Linux/GNU toolchains also support the value @samp{native},
11266 which selects the best architecture option for the host processor.
11267 @option{-mtune=native} has no effect if GCC does not recognize
11270 @item -mmemory-latency=@var{time}
11271 @opindex mmemory-latency
11272 Sets the latency the scheduler should assume for typical memory
11273 references as seen by the application. This number is highly
11274 dependent on the memory access patterns used by the application
11275 and the size of the external cache on the machine.
11277 Valid options for @var{time} are
11281 A decimal number representing clock cycles.
11287 The compiler contains estimates of the number of clock cycles for
11288 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11289 (also called Dcache, Scache, and Bcache), as well as to main memory.
11290 Note that L3 is only valid for EV5.
11295 @node DEC Alpha/VMS Options
11296 @subsection DEC Alpha/VMS Options
11298 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11301 @item -mvms-return-codes
11302 @opindex mvms-return-codes
11303 Return VMS condition codes from main. The default is to return POSIX
11304 style condition (e.g.@: error) codes.
11306 @item -mdebug-main=@var{prefix}
11307 @opindex mdebug-main=@var{prefix}
11308 Flag the first routine whose name starts with @var{prefix} as the main
11309 routine for the debugger.
11313 Default to 64bit memory allocation routines.
11317 @subsection FR30 Options
11318 @cindex FR30 Options
11320 These options are defined specifically for the FR30 port.
11324 @item -msmall-model
11325 @opindex msmall-model
11326 Use the small address space model. This can produce smaller code, but
11327 it does assume that all symbolic values and addresses will fit into a
11332 Assume that run-time support has been provided and so there is no need
11333 to include the simulator library (@file{libsim.a}) on the linker
11339 @subsection FRV Options
11340 @cindex FRV Options
11346 Only use the first 32 general purpose registers.
11351 Use all 64 general purpose registers.
11356 Use only the first 32 floating point registers.
11361 Use all 64 floating point registers
11364 @opindex mhard-float
11366 Use hardware instructions for floating point operations.
11369 @opindex msoft-float
11371 Use library routines for floating point operations.
11376 Dynamically allocate condition code registers.
11381 Do not try to dynamically allocate condition code registers, only
11382 use @code{icc0} and @code{fcc0}.
11387 Change ABI to use double word insns.
11392 Do not use double word instructions.
11397 Use floating point double instructions.
11400 @opindex mno-double
11402 Do not use floating point double instructions.
11407 Use media instructions.
11412 Do not use media instructions.
11417 Use multiply and add/subtract instructions.
11420 @opindex mno-muladd
11422 Do not use multiply and add/subtract instructions.
11427 Select the FDPIC ABI, that uses function descriptors to represent
11428 pointers to functions. Without any PIC/PIE-related options, it
11429 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11430 assumes GOT entries and small data are within a 12-bit range from the
11431 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11432 are computed with 32 bits.
11433 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11436 @opindex minline-plt
11438 Enable inlining of PLT entries in function calls to functions that are
11439 not known to bind locally. It has no effect without @option{-mfdpic}.
11440 It's enabled by default if optimizing for speed and compiling for
11441 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11442 optimization option such as @option{-O3} or above is present in the
11448 Assume a large TLS segment when generating thread-local code.
11453 Do not assume a large TLS segment when generating thread-local code.
11458 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11459 that is known to be in read-only sections. It's enabled by default,
11460 except for @option{-fpic} or @option{-fpie}: even though it may help
11461 make the global offset table smaller, it trades 1 instruction for 4.
11462 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11463 one of which may be shared by multiple symbols, and it avoids the need
11464 for a GOT entry for the referenced symbol, so it's more likely to be a
11465 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11467 @item -multilib-library-pic
11468 @opindex multilib-library-pic
11470 Link with the (library, not FD) pic libraries. It's implied by
11471 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11472 @option{-fpic} without @option{-mfdpic}. You should never have to use
11476 @opindex mlinked-fp
11478 Follow the EABI requirement of always creating a frame pointer whenever
11479 a stack frame is allocated. This option is enabled by default and can
11480 be disabled with @option{-mno-linked-fp}.
11483 @opindex mlong-calls
11485 Use indirect addressing to call functions outside the current
11486 compilation unit. This allows the functions to be placed anywhere
11487 within the 32-bit address space.
11489 @item -malign-labels
11490 @opindex malign-labels
11492 Try to align labels to an 8-byte boundary by inserting nops into the
11493 previous packet. This option only has an effect when VLIW packing
11494 is enabled. It doesn't create new packets; it merely adds nops to
11497 @item -mlibrary-pic
11498 @opindex mlibrary-pic
11500 Generate position-independent EABI code.
11505 Use only the first four media accumulator registers.
11510 Use all eight media accumulator registers.
11515 Pack VLIW instructions.
11520 Do not pack VLIW instructions.
11523 @opindex mno-eflags
11525 Do not mark ABI switches in e_flags.
11528 @opindex mcond-move
11530 Enable the use of conditional-move instructions (default).
11532 This switch is mainly for debugging the compiler and will likely be removed
11533 in a future version.
11535 @item -mno-cond-move
11536 @opindex mno-cond-move
11538 Disable the use of conditional-move instructions.
11540 This switch is mainly for debugging the compiler and will likely be removed
11541 in a future version.
11546 Enable the use of conditional set instructions (default).
11548 This switch is mainly for debugging the compiler and will likely be removed
11549 in a future version.
11554 Disable the use of conditional set instructions.
11556 This switch is mainly for debugging the compiler and will likely be removed
11557 in a future version.
11560 @opindex mcond-exec
11562 Enable the use of conditional execution (default).
11564 This switch is mainly for debugging the compiler and will likely be removed
11565 in a future version.
11567 @item -mno-cond-exec
11568 @opindex mno-cond-exec
11570 Disable the use of conditional execution.
11572 This switch is mainly for debugging the compiler and will likely be removed
11573 in a future version.
11575 @item -mvliw-branch
11576 @opindex mvliw-branch
11578 Run a pass to pack branches into VLIW instructions (default).
11580 This switch is mainly for debugging the compiler and will likely be removed
11581 in a future version.
11583 @item -mno-vliw-branch
11584 @opindex mno-vliw-branch
11586 Do not run a pass to pack branches into VLIW instructions.
11588 This switch is mainly for debugging the compiler and will likely be removed
11589 in a future version.
11591 @item -mmulti-cond-exec
11592 @opindex mmulti-cond-exec
11594 Enable optimization of @code{&&} and @code{||} in conditional execution
11597 This switch is mainly for debugging the compiler and will likely be removed
11598 in a future version.
11600 @item -mno-multi-cond-exec
11601 @opindex mno-multi-cond-exec
11603 Disable optimization of @code{&&} and @code{||} in conditional execution.
11605 This switch is mainly for debugging the compiler and will likely be removed
11606 in a future version.
11608 @item -mnested-cond-exec
11609 @opindex mnested-cond-exec
11611 Enable nested conditional execution optimizations (default).
11613 This switch is mainly for debugging the compiler and will likely be removed
11614 in a future version.
11616 @item -mno-nested-cond-exec
11617 @opindex mno-nested-cond-exec
11619 Disable nested conditional execution optimizations.
11621 This switch is mainly for debugging the compiler and will likely be removed
11622 in a future version.
11624 @item -moptimize-membar
11625 @opindex moptimize-membar
11627 This switch removes redundant @code{membar} instructions from the
11628 compiler generated code. It is enabled by default.
11630 @item -mno-optimize-membar
11631 @opindex mno-optimize-membar
11633 This switch disables the automatic removal of redundant @code{membar}
11634 instructions from the generated code.
11636 @item -mtomcat-stats
11637 @opindex mtomcat-stats
11639 Cause gas to print out tomcat statistics.
11641 @item -mcpu=@var{cpu}
11644 Select the processor type for which to generate code. Possible values are
11645 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11646 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11650 @node GNU/Linux Options
11651 @subsection GNU/Linux Options
11653 These @samp{-m} options are defined for GNU/Linux targets:
11658 Use the GNU C library. This is the default except
11659 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11663 Use uClibc C library. This is the default on
11664 @samp{*-*-linux-*uclibc*} targets.
11668 Use Bionic C library. This is the default on
11669 @samp{*-*-linux-*android*} targets.
11673 Compile code compatible with Android platform. This is the default on
11674 @samp{*-*-linux-*android*} targets.
11676 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11677 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11678 this option makes the GCC driver pass Android-specific options to the linker.
11679 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11682 @item -tno-android-cc
11683 @opindex tno-android-cc
11684 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11685 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11686 @option{-fno-rtti} by default.
11688 @item -tno-android-ld
11689 @opindex tno-android-ld
11690 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11691 linking options to the linker.
11695 @node H8/300 Options
11696 @subsection H8/300 Options
11698 These @samp{-m} options are defined for the H8/300 implementations:
11703 Shorten some address references at link time, when possible; uses the
11704 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11705 ld, Using ld}, for a fuller description.
11709 Generate code for the H8/300H@.
11713 Generate code for the H8S@.
11717 Generate code for the H8S and H8/300H in the normal mode. This switch
11718 must be used either with @option{-mh} or @option{-ms}.
11722 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11726 Make @code{int} data 32 bits by default.
11729 @opindex malign-300
11730 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11731 The default for the H8/300H and H8S is to align longs and floats on 4
11733 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11734 This option has no effect on the H8/300.
11738 @subsection HPPA Options
11739 @cindex HPPA Options
11741 These @samp{-m} options are defined for the HPPA family of computers:
11744 @item -march=@var{architecture-type}
11746 Generate code for the specified architecture. The choices for
11747 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11748 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11749 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11750 architecture option for your machine. Code compiled for lower numbered
11751 architectures will run on higher numbered architectures, but not the
11754 @item -mpa-risc-1-0
11755 @itemx -mpa-risc-1-1
11756 @itemx -mpa-risc-2-0
11757 @opindex mpa-risc-1-0
11758 @opindex mpa-risc-1-1
11759 @opindex mpa-risc-2-0
11760 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11763 @opindex mbig-switch
11764 Generate code suitable for big switch tables. Use this option only if
11765 the assembler/linker complain about out of range branches within a switch
11768 @item -mjump-in-delay
11769 @opindex mjump-in-delay
11770 Fill delay slots of function calls with unconditional jump instructions
11771 by modifying the return pointer for the function call to be the target
11772 of the conditional jump.
11774 @item -mdisable-fpregs
11775 @opindex mdisable-fpregs
11776 Prevent floating point registers from being used in any manner. This is
11777 necessary for compiling kernels which perform lazy context switching of
11778 floating point registers. If you use this option and attempt to perform
11779 floating point operations, the compiler will abort.
11781 @item -mdisable-indexing
11782 @opindex mdisable-indexing
11783 Prevent the compiler from using indexing address modes. This avoids some
11784 rather obscure problems when compiling MIG generated code under MACH@.
11786 @item -mno-space-regs
11787 @opindex mno-space-regs
11788 Generate code that assumes the target has no space registers. This allows
11789 GCC to generate faster indirect calls and use unscaled index address modes.
11791 Such code is suitable for level 0 PA systems and kernels.
11793 @item -mfast-indirect-calls
11794 @opindex mfast-indirect-calls
11795 Generate code that assumes calls never cross space boundaries. This
11796 allows GCC to emit code which performs faster indirect calls.
11798 This option will not work in the presence of shared libraries or nested
11801 @item -mfixed-range=@var{register-range}
11802 @opindex mfixed-range
11803 Generate code treating the given register range as fixed registers.
11804 A fixed register is one that the register allocator can not use. This is
11805 useful when compiling kernel code. A register range is specified as
11806 two registers separated by a dash. Multiple register ranges can be
11807 specified separated by a comma.
11809 @item -mlong-load-store
11810 @opindex mlong-load-store
11811 Generate 3-instruction load and store sequences as sometimes required by
11812 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11815 @item -mportable-runtime
11816 @opindex mportable-runtime
11817 Use the portable calling conventions proposed by HP for ELF systems.
11821 Enable the use of assembler directives only GAS understands.
11823 @item -mschedule=@var{cpu-type}
11825 Schedule code according to the constraints for the machine type
11826 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11827 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11828 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11829 proper scheduling option for your machine. The default scheduling is
11833 @opindex mlinker-opt
11834 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11835 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11836 linkers in which they give bogus error messages when linking some programs.
11839 @opindex msoft-float
11840 Generate output containing library calls for floating point.
11841 @strong{Warning:} the requisite libraries are not available for all HPPA
11842 targets. Normally the facilities of the machine's usual C compiler are
11843 used, but this cannot be done directly in cross-compilation. You must make
11844 your own arrangements to provide suitable library functions for
11847 @option{-msoft-float} changes the calling convention in the output file;
11848 therefore, it is only useful if you compile @emph{all} of a program with
11849 this option. In particular, you need to compile @file{libgcc.a}, the
11850 library that comes with GCC, with @option{-msoft-float} in order for
11855 Generate the predefine, @code{_SIO}, for server IO@. The default is
11856 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11857 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11858 options are available under HP-UX and HI-UX@.
11862 Use GNU ld specific options. This passes @option{-shared} to ld when
11863 building a shared library. It is the default when GCC is configured,
11864 explicitly or implicitly, with the GNU linker. This option does not
11865 have any affect on which ld is called, it only changes what parameters
11866 are passed to that ld. The ld that is called is determined by the
11867 @option{--with-ld} configure option, GCC's program search path, and
11868 finally by the user's @env{PATH}. The linker used by GCC can be printed
11869 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11870 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11874 Use HP ld specific options. This passes @option{-b} to ld when building
11875 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11876 links. It is the default when GCC is configured, explicitly or
11877 implicitly, with the HP linker. This option does not have any affect on
11878 which ld is called, it only changes what parameters are passed to that
11879 ld. The ld that is called is determined by the @option{--with-ld}
11880 configure option, GCC's program search path, and finally by the user's
11881 @env{PATH}. The linker used by GCC can be printed using @samp{which
11882 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11883 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11886 @opindex mno-long-calls
11887 Generate code that uses long call sequences. This ensures that a call
11888 is always able to reach linker generated stubs. The default is to generate
11889 long calls only when the distance from the call site to the beginning
11890 of the function or translation unit, as the case may be, exceeds a
11891 predefined limit set by the branch type being used. The limits for
11892 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11893 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11896 Distances are measured from the beginning of functions when using the
11897 @option{-ffunction-sections} option, or when using the @option{-mgas}
11898 and @option{-mno-portable-runtime} options together under HP-UX with
11901 It is normally not desirable to use this option as it will degrade
11902 performance. However, it may be useful in large applications,
11903 particularly when partial linking is used to build the application.
11905 The types of long calls used depends on the capabilities of the
11906 assembler and linker, and the type of code being generated. The
11907 impact on systems that support long absolute calls, and long pic
11908 symbol-difference or pc-relative calls should be relatively small.
11909 However, an indirect call is used on 32-bit ELF systems in pic code
11910 and it is quite long.
11912 @item -munix=@var{unix-std}
11914 Generate compiler predefines and select a startfile for the specified
11915 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11916 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11917 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11918 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11919 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11922 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11923 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11924 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11925 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11926 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11927 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11929 It is @emph{important} to note that this option changes the interfaces
11930 for various library routines. It also affects the operational behavior
11931 of the C library. Thus, @emph{extreme} care is needed in using this
11934 Library code that is intended to operate with more than one UNIX
11935 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11936 as appropriate. Most GNU software doesn't provide this capability.
11940 Suppress the generation of link options to search libdld.sl when the
11941 @option{-static} option is specified on HP-UX 10 and later.
11945 The HP-UX implementation of setlocale in libc has a dependency on
11946 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11947 when the @option{-static} option is specified, special link options
11948 are needed to resolve this dependency.
11950 On HP-UX 10 and later, the GCC driver adds the necessary options to
11951 link with libdld.sl when the @option{-static} option is specified.
11952 This causes the resulting binary to be dynamic. On the 64-bit port,
11953 the linkers generate dynamic binaries by default in any case. The
11954 @option{-nolibdld} option can be used to prevent the GCC driver from
11955 adding these link options.
11959 Add support for multithreading with the @dfn{dce thread} library
11960 under HP-UX@. This option sets flags for both the preprocessor and
11964 @node i386 and x86-64 Options
11965 @subsection Intel 386 and AMD x86-64 Options
11966 @cindex i386 Options
11967 @cindex x86-64 Options
11968 @cindex Intel 386 Options
11969 @cindex AMD x86-64 Options
11971 These @samp{-m} options are defined for the i386 and x86-64 family of
11975 @item -mtune=@var{cpu-type}
11977 Tune to @var{cpu-type} everything applicable about the generated code, except
11978 for the ABI and the set of available instructions. The choices for
11979 @var{cpu-type} are:
11982 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
11983 If you know the CPU on which your code will run, then you should use
11984 the corresponding @option{-mtune} option instead of
11985 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11986 of your application will have, then you should use this option.
11988 As new processors are deployed in the marketplace, the behavior of this
11989 option will change. Therefore, if you upgrade to a newer version of
11990 GCC, the code generated option will change to reflect the processors
11991 that were most common when that version of GCC was released.
11993 There is no @option{-march=generic} option because @option{-march}
11994 indicates the instruction set the compiler can use, and there is no
11995 generic instruction set applicable to all processors. In contrast,
11996 @option{-mtune} indicates the processor (or, in this case, collection of
11997 processors) for which the code is optimized.
11999 This selects the CPU to tune for at compilation time by determining
12000 the processor type of the compiling machine. Using @option{-mtune=native}
12001 will produce code optimized for the local machine under the constraints
12002 of the selected instruction set. Using @option{-march=native} will
12003 enable all instruction subsets supported by the local machine (hence
12004 the result might not run on different machines).
12006 Original Intel's i386 CPU@.
12008 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12009 @item i586, pentium
12010 Intel Pentium CPU with no MMX support.
12012 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12014 Intel PentiumPro CPU@.
12016 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12017 instruction set will be used, so the code will run on all i686 family chips.
12019 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12020 @item pentium3, pentium3m
12021 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12024 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12025 support. Used by Centrino notebooks.
12026 @item pentium4, pentium4m
12027 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12029 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12032 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12033 SSE2 and SSE3 instruction set support.
12035 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12036 instruction set support.
12038 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12039 instruction set support.
12041 AMD K6 CPU with MMX instruction set support.
12043 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12044 @item athlon, athlon-tbird
12045 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12047 @item athlon-4, athlon-xp, athlon-mp
12048 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12049 instruction set support.
12050 @item k8, opteron, athlon64, athlon-fx
12051 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12052 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12053 @item k8-sse3, opteron-sse3, athlon64-sse3
12054 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12055 @item amdfam10, barcelona
12056 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12057 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12058 instruction set extensions.)
12060 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12063 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12064 instruction set support.
12066 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12067 implemented for this chip.)
12069 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12070 implemented for this chip.)
12072 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12075 While picking a specific @var{cpu-type} will schedule things appropriately
12076 for that particular chip, the compiler will not generate any code that
12077 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12080 @item -march=@var{cpu-type}
12082 Generate instructions for the machine type @var{cpu-type}. The choices
12083 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12084 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12086 @item -mcpu=@var{cpu-type}
12088 A deprecated synonym for @option{-mtune}.
12090 @item -mfpmath=@var{unit}
12092 Generate floating point arithmetics for selected unit @var{unit}. The choices
12093 for @var{unit} are:
12097 Use the standard 387 floating point coprocessor present majority of chips and
12098 emulated otherwise. Code compiled with this option will run almost everywhere.
12099 The temporary results are computed in 80bit precision instead of precision
12100 specified by the type resulting in slightly different results compared to most
12101 of other chips. See @option{-ffloat-store} for more detailed description.
12103 This is the default choice for i386 compiler.
12106 Use scalar floating point instructions present in the SSE instruction set.
12107 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12108 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12109 instruction set supports only single precision arithmetics, thus the double and
12110 extended precision arithmetics is still done using 387. Later version, present
12111 only in Pentium4 and the future AMD x86-64 chips supports double precision
12114 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12115 or @option{-msse2} switches to enable SSE extensions and make this option
12116 effective. For the x86-64 compiler, these extensions are enabled by default.
12118 The resulting code should be considerably faster in the majority of cases and avoid
12119 the numerical instability problems of 387 code, but may break some existing
12120 code that expects temporaries to be 80bit.
12122 This is the default choice for the x86-64 compiler.
12127 Attempt to utilize both instruction sets at once. This effectively double the
12128 amount of available registers and on chips with separate execution units for
12129 387 and SSE the execution resources too. Use this option with care, as it is
12130 still experimental, because the GCC register allocator does not model separate
12131 functional units well resulting in instable performance.
12134 @item -masm=@var{dialect}
12135 @opindex masm=@var{dialect}
12136 Output asm instructions using selected @var{dialect}. Supported
12137 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12138 not support @samp{intel}.
12141 @itemx -mno-ieee-fp
12143 @opindex mno-ieee-fp
12144 Control whether or not the compiler uses IEEE floating point
12145 comparisons. These handle correctly the case where the result of a
12146 comparison is unordered.
12149 @opindex msoft-float
12150 Generate output containing library calls for floating point.
12151 @strong{Warning:} the requisite libraries are not part of GCC@.
12152 Normally the facilities of the machine's usual C compiler are used, but
12153 this can't be done directly in cross-compilation. You must make your
12154 own arrangements to provide suitable library functions for
12157 On machines where a function returns floating point results in the 80387
12158 register stack, some floating point opcodes may be emitted even if
12159 @option{-msoft-float} is used.
12161 @item -mno-fp-ret-in-387
12162 @opindex mno-fp-ret-in-387
12163 Do not use the FPU registers for return values of functions.
12165 The usual calling convention has functions return values of types
12166 @code{float} and @code{double} in an FPU register, even if there
12167 is no FPU@. The idea is that the operating system should emulate
12170 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12171 in ordinary CPU registers instead.
12173 @item -mno-fancy-math-387
12174 @opindex mno-fancy-math-387
12175 Some 387 emulators do not support the @code{sin}, @code{cos} and
12176 @code{sqrt} instructions for the 387. Specify this option to avoid
12177 generating those instructions. This option is the default on FreeBSD,
12178 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12179 indicates that the target cpu will always have an FPU and so the
12180 instruction will not need emulation. As of revision 2.6.1, these
12181 instructions are not generated unless you also use the
12182 @option{-funsafe-math-optimizations} switch.
12184 @item -malign-double
12185 @itemx -mno-align-double
12186 @opindex malign-double
12187 @opindex mno-align-double
12188 Control whether GCC aligns @code{double}, @code{long double}, and
12189 @code{long long} variables on a two word boundary or a one word
12190 boundary. Aligning @code{double} variables on a two word boundary will
12191 produce code that runs somewhat faster on a @samp{Pentium} at the
12192 expense of more memory.
12194 On x86-64, @option{-malign-double} is enabled by default.
12196 @strong{Warning:} if you use the @option{-malign-double} switch,
12197 structures containing the above types will be aligned differently than
12198 the published application binary interface specifications for the 386
12199 and will not be binary compatible with structures in code compiled
12200 without that switch.
12202 @item -m96bit-long-double
12203 @itemx -m128bit-long-double
12204 @opindex m96bit-long-double
12205 @opindex m128bit-long-double
12206 These switches control the size of @code{long double} type. The i386
12207 application binary interface specifies the size to be 96 bits,
12208 so @option{-m96bit-long-double} is the default in 32 bit mode.
12210 Modern architectures (Pentium and newer) would prefer @code{long double}
12211 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12212 conforming to the ABI, this would not be possible. So specifying a
12213 @option{-m128bit-long-double} will align @code{long double}
12214 to a 16 byte boundary by padding the @code{long double} with an additional
12217 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12218 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12220 Notice that neither of these options enable any extra precision over the x87
12221 standard of 80 bits for a @code{long double}.
12223 @strong{Warning:} if you override the default value for your target ABI, the
12224 structures and arrays containing @code{long double} variables will change
12225 their size as well as function calling convention for function taking
12226 @code{long double} will be modified. Hence they will not be binary
12227 compatible with arrays or structures in code compiled without that switch.
12229 @item -mlarge-data-threshold=@var{number}
12230 @opindex mlarge-data-threshold=@var{number}
12231 When @option{-mcmodel=medium} is specified, the data greater than
12232 @var{threshold} are placed in large data section. This value must be the
12233 same across all object linked into the binary and defaults to 65535.
12237 Use a different function-calling convention, in which functions that
12238 take a fixed number of arguments return with the @code{ret} @var{num}
12239 instruction, which pops their arguments while returning. This saves one
12240 instruction in the caller since there is no need to pop the arguments
12243 You can specify that an individual function is called with this calling
12244 sequence with the function attribute @samp{stdcall}. You can also
12245 override the @option{-mrtd} option by using the function attribute
12246 @samp{cdecl}. @xref{Function Attributes}.
12248 @strong{Warning:} this calling convention is incompatible with the one
12249 normally used on Unix, so you cannot use it if you need to call
12250 libraries compiled with the Unix compiler.
12252 Also, you must provide function prototypes for all functions that
12253 take variable numbers of arguments (including @code{printf});
12254 otherwise incorrect code will be generated for calls to those
12257 In addition, seriously incorrect code will result if you call a
12258 function with too many arguments. (Normally, extra arguments are
12259 harmlessly ignored.)
12261 @item -mregparm=@var{num}
12263 Control how many registers are used to pass integer arguments. By
12264 default, no registers are used to pass arguments, and at most 3
12265 registers can be used. You can control this behavior for a specific
12266 function by using the function attribute @samp{regparm}.
12267 @xref{Function Attributes}.
12269 @strong{Warning:} if you use this switch, and
12270 @var{num} is nonzero, then you must build all modules with the same
12271 value, including any libraries. This includes the system libraries and
12275 @opindex msseregparm
12276 Use SSE register passing conventions for float and double arguments
12277 and return values. You can control this behavior for a specific
12278 function by using the function attribute @samp{sseregparm}.
12279 @xref{Function Attributes}.
12281 @strong{Warning:} if you use this switch then you must build all
12282 modules with the same value, including any libraries. This includes
12283 the system libraries and startup modules.
12285 @item -mvect8-ret-in-mem
12286 @opindex mvect8-ret-in-mem
12287 Return 8-byte vectors in memory instead of MMX registers. This is the
12288 default on Solaris~8 and 9 and VxWorks to match the ABI of the Sun
12289 Studio compilers until version 12. Later compiler versions (starting
12290 with Studio 12 Update~1) follow the ABI used by other x86 targets, which
12291 is the default on Solaris~10 and later. @emph{Only} use this option if
12292 you need to remain compatible with existing code produced by those
12293 previous compiler versions or older versions of GCC.
12302 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12303 is specified, the significands of results of floating-point operations are
12304 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12305 significands of results of floating-point operations to 53 bits (double
12306 precision) and @option{-mpc80} rounds the significands of results of
12307 floating-point operations to 64 bits (extended double precision), which is
12308 the default. When this option is used, floating-point operations in higher
12309 precisions are not available to the programmer without setting the FPU
12310 control word explicitly.
12312 Setting the rounding of floating-point operations to less than the default
12313 80 bits can speed some programs by 2% or more. Note that some mathematical
12314 libraries assume that extended precision (80 bit) floating-point operations
12315 are enabled by default; routines in such libraries could suffer significant
12316 loss of accuracy, typically through so-called "catastrophic cancellation",
12317 when this option is used to set the precision to less than extended precision.
12319 @item -mstackrealign
12320 @opindex mstackrealign
12321 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12322 option will generate an alternate prologue and epilogue that realigns the
12323 runtime stack if necessary. This supports mixing legacy codes that keep
12324 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12325 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12326 applicable to individual functions.
12328 @item -mpreferred-stack-boundary=@var{num}
12329 @opindex mpreferred-stack-boundary
12330 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12331 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12332 the default is 4 (16 bytes or 128 bits).
12334 @item -mincoming-stack-boundary=@var{num}
12335 @opindex mincoming-stack-boundary
12336 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12337 boundary. If @option{-mincoming-stack-boundary} is not specified,
12338 the one specified by @option{-mpreferred-stack-boundary} will be used.
12340 On Pentium and PentiumPro, @code{double} and @code{long double} values
12341 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12342 suffer significant run time performance penalties. On Pentium III, the
12343 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12344 properly if it is not 16 byte aligned.
12346 To ensure proper alignment of this values on the stack, the stack boundary
12347 must be as aligned as that required by any value stored on the stack.
12348 Further, every function must be generated such that it keeps the stack
12349 aligned. Thus calling a function compiled with a higher preferred
12350 stack boundary from a function compiled with a lower preferred stack
12351 boundary will most likely misalign the stack. It is recommended that
12352 libraries that use callbacks always use the default setting.
12354 This extra alignment does consume extra stack space, and generally
12355 increases code size. Code that is sensitive to stack space usage, such
12356 as embedded systems and operating system kernels, may want to reduce the
12357 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12384 @itemx -mno-fsgsbase
12410 These switches enable or disable the use of instructions in the MMX,
12411 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12412 F16C, SSE4A, FMA4, XOP, LWP, ABM or 3DNow!@: extended instruction sets.
12413 These extensions are also available as built-in functions: see
12414 @ref{X86 Built-in Functions}, for details of the functions enabled and
12415 disabled by these switches.
12417 To have SSE/SSE2 instructions generated automatically from floating-point
12418 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12420 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12421 generates new AVX instructions or AVX equivalence for all SSEx instructions
12424 These options will enable GCC to use these extended instructions in
12425 generated code, even without @option{-mfpmath=sse}. Applications which
12426 perform runtime CPU detection must compile separate files for each
12427 supported architecture, using the appropriate flags. In particular,
12428 the file containing the CPU detection code should be compiled without
12432 @itemx -mno-fused-madd
12433 @opindex mfused-madd
12434 @opindex mno-fused-madd
12435 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12436 instructions. The default is to use these instructions.
12440 This option instructs GCC to emit a @code{cld} instruction in the prologue
12441 of functions that use string instructions. String instructions depend on
12442 the DF flag to select between autoincrement or autodecrement mode. While the
12443 ABI specifies the DF flag to be cleared on function entry, some operating
12444 systems violate this specification by not clearing the DF flag in their
12445 exception dispatchers. The exception handler can be invoked with the DF flag
12446 set which leads to wrong direction mode, when string instructions are used.
12447 This option can be enabled by default on 32-bit x86 targets by configuring
12448 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12449 instructions can be suppressed with the @option{-mno-cld} compiler option
12454 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12455 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12456 data types. This is useful for high resolution counters that could be updated
12457 by multiple processors (or cores). This instruction is generated as part of
12458 atomic built-in functions: see @ref{Atomic Builtins} for details.
12462 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12463 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12464 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12465 SAHF are load and store instructions, respectively, for certain status flags.
12466 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12467 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12471 This option will enable GCC to use movbe instruction to implement
12472 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12476 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12477 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12478 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12482 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12483 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12484 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12485 variants) for single precision floating point arguments. These instructions
12486 are generated only when @option{-funsafe-math-optimizations} is enabled
12487 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12488 Note that while the throughput of the sequence is higher than the throughput
12489 of the non-reciprocal instruction, the precision of the sequence can be
12490 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12492 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12493 already with @option{-ffast-math} (or the above option combination), and
12494 doesn't need @option{-mrecip}.
12496 @item -mveclibabi=@var{type}
12497 @opindex mveclibabi
12498 Specifies the ABI type to use for vectorizing intrinsics using an
12499 external library. Supported types are @code{svml} for the Intel short
12500 vector math library and @code{acml} for the AMD math core library style
12501 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12502 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12503 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12504 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12505 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12506 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12507 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12508 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12509 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12510 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12511 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12512 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12513 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12514 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12515 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12516 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12517 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12518 compatible library will have to be specified at link time.
12520 @item -mabi=@var{name}
12522 Generate code for the specified calling convention. Permissible values
12523 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12524 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12525 ABI when targeting Windows. On all other systems, the default is the
12526 SYSV ABI. You can control this behavior for a specific function by
12527 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12528 @xref{Function Attributes}.
12531 @itemx -mno-push-args
12532 @opindex mpush-args
12533 @opindex mno-push-args
12534 Use PUSH operations to store outgoing parameters. This method is shorter
12535 and usually equally fast as method using SUB/MOV operations and is enabled
12536 by default. In some cases disabling it may improve performance because of
12537 improved scheduling and reduced dependencies.
12539 @item -maccumulate-outgoing-args
12540 @opindex maccumulate-outgoing-args
12541 If enabled, the maximum amount of space required for outgoing arguments will be
12542 computed in the function prologue. This is faster on most modern CPUs
12543 because of reduced dependencies, improved scheduling and reduced stack usage
12544 when preferred stack boundary is not equal to 2. The drawback is a notable
12545 increase in code size. This switch implies @option{-mno-push-args}.
12549 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12550 on thread-safe exception handling must compile and link all code with the
12551 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12552 @option{-D_MT}; when linking, it links in a special thread helper library
12553 @option{-lmingwthrd} which cleans up per thread exception handling data.
12555 @item -mno-align-stringops
12556 @opindex mno-align-stringops
12557 Do not align destination of inlined string operations. This switch reduces
12558 code size and improves performance in case the destination is already aligned,
12559 but GCC doesn't know about it.
12561 @item -minline-all-stringops
12562 @opindex minline-all-stringops
12563 By default GCC inlines string operations only when destination is known to be
12564 aligned at least to 4 byte boundary. This enables more inlining, increase code
12565 size, but may improve performance of code that depends on fast memcpy, strlen
12566 and memset for short lengths.
12568 @item -minline-stringops-dynamically
12569 @opindex minline-stringops-dynamically
12570 For string operation of unknown size, inline runtime checks so for small
12571 blocks inline code is used, while for large blocks library call is used.
12573 @item -mstringop-strategy=@var{alg}
12574 @opindex mstringop-strategy=@var{alg}
12575 Overwrite internal decision heuristic about particular algorithm to inline
12576 string operation with. The allowed values are @code{rep_byte},
12577 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12578 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12579 expanding inline loop, @code{libcall} for always expanding library call.
12581 @item -momit-leaf-frame-pointer
12582 @opindex momit-leaf-frame-pointer
12583 Don't keep the frame pointer in a register for leaf functions. This
12584 avoids the instructions to save, set up and restore frame pointers and
12585 makes an extra register available in leaf functions. The option
12586 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12587 which might make debugging harder.
12589 @item -mtls-direct-seg-refs
12590 @itemx -mno-tls-direct-seg-refs
12591 @opindex mtls-direct-seg-refs
12592 Controls whether TLS variables may be accessed with offsets from the
12593 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12594 or whether the thread base pointer must be added. Whether or not this
12595 is legal depends on the operating system, and whether it maps the
12596 segment to cover the entire TLS area.
12598 For systems that use GNU libc, the default is on.
12601 @itemx -mno-sse2avx
12603 Specify that the assembler should encode SSE instructions with VEX
12604 prefix. The option @option{-mavx} turns this on by default.
12609 If profiling is active @option{-pg} put the profiling
12610 counter call before prologue.
12611 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12612 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12615 @itemx -mno-8bit-idiv
12617 On some processors, like Intel Atom, 8bit unsigned integer divide is
12618 much faster than 32bit/64bit integer divide. This option will generate a
12619 runt-time check. If both dividend and divisor are within range of 0
12620 to 255, 8bit unsigned integer divide will be used instead of
12621 32bit/64bit integer divide.
12625 These @samp{-m} switches are supported in addition to the above
12626 on AMD x86-64 processors in 64-bit environments.
12633 Generate code for a 32-bit or 64-bit environment.
12634 The 32-bit environment sets int, long and pointer to 32 bits and
12635 generates code that runs on any i386 system.
12636 The 64-bit environment sets int to 32 bits and long and pointer
12637 to 64 bits and generates code for AMD's x86-64 architecture. For
12638 darwin only the -m64 option turns off the @option{-fno-pic} and
12639 @option{-mdynamic-no-pic} options.
12641 @item -mno-red-zone
12642 @opindex mno-red-zone
12643 Do not use a so called red zone for x86-64 code. The red zone is mandated
12644 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12645 stack pointer that will not be modified by signal or interrupt handlers
12646 and therefore can be used for temporary data without adjusting the stack
12647 pointer. The flag @option{-mno-red-zone} disables this red zone.
12649 @item -mcmodel=small
12650 @opindex mcmodel=small
12651 Generate code for the small code model: the program and its symbols must
12652 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12653 Programs can be statically or dynamically linked. This is the default
12656 @item -mcmodel=kernel
12657 @opindex mcmodel=kernel
12658 Generate code for the kernel code model. The kernel runs in the
12659 negative 2 GB of the address space.
12660 This model has to be used for Linux kernel code.
12662 @item -mcmodel=medium
12663 @opindex mcmodel=medium
12664 Generate code for the medium model: The program is linked in the lower 2
12665 GB of the address space. Small symbols are also placed there. Symbols
12666 with sizes larger than @option{-mlarge-data-threshold} are put into
12667 large data or bss sections and can be located above 2GB. Programs can
12668 be statically or dynamically linked.
12670 @item -mcmodel=large
12671 @opindex mcmodel=large
12672 Generate code for the large model: This model makes no assumptions
12673 about addresses and sizes of sections.
12676 @node IA-64 Options
12677 @subsection IA-64 Options
12678 @cindex IA-64 Options
12680 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12684 @opindex mbig-endian
12685 Generate code for a big endian target. This is the default for HP-UX@.
12687 @item -mlittle-endian
12688 @opindex mlittle-endian
12689 Generate code for a little endian target. This is the default for AIX5
12695 @opindex mno-gnu-as
12696 Generate (or don't) code for the GNU assembler. This is the default.
12697 @c Also, this is the default if the configure option @option{--with-gnu-as}
12703 @opindex mno-gnu-ld
12704 Generate (or don't) code for the GNU linker. This is the default.
12705 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12710 Generate code that does not use a global pointer register. The result
12711 is not position independent code, and violates the IA-64 ABI@.
12713 @item -mvolatile-asm-stop
12714 @itemx -mno-volatile-asm-stop
12715 @opindex mvolatile-asm-stop
12716 @opindex mno-volatile-asm-stop
12717 Generate (or don't) a stop bit immediately before and after volatile asm
12720 @item -mregister-names
12721 @itemx -mno-register-names
12722 @opindex mregister-names
12723 @opindex mno-register-names
12724 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12725 the stacked registers. This may make assembler output more readable.
12731 Disable (or enable) optimizations that use the small data section. This may
12732 be useful for working around optimizer bugs.
12734 @item -mconstant-gp
12735 @opindex mconstant-gp
12736 Generate code that uses a single constant global pointer value. This is
12737 useful when compiling kernel code.
12741 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12742 This is useful when compiling firmware code.
12744 @item -minline-float-divide-min-latency
12745 @opindex minline-float-divide-min-latency
12746 Generate code for inline divides of floating point values
12747 using the minimum latency algorithm.
12749 @item -minline-float-divide-max-throughput
12750 @opindex minline-float-divide-max-throughput
12751 Generate code for inline divides of floating point values
12752 using the maximum throughput algorithm.
12754 @item -mno-inline-float-divide
12755 @opindex mno-inline-float-divide
12756 Do not generate inline code for divides of floating point values.
12758 @item -minline-int-divide-min-latency
12759 @opindex minline-int-divide-min-latency
12760 Generate code for inline divides of integer values
12761 using the minimum latency algorithm.
12763 @item -minline-int-divide-max-throughput
12764 @opindex minline-int-divide-max-throughput
12765 Generate code for inline divides of integer values
12766 using the maximum throughput algorithm.
12768 @item -mno-inline-int-divide
12769 @opindex mno-inline-int-divide
12770 Do not generate inline code for divides of integer values.
12772 @item -minline-sqrt-min-latency
12773 @opindex minline-sqrt-min-latency
12774 Generate code for inline square roots
12775 using the minimum latency algorithm.
12777 @item -minline-sqrt-max-throughput
12778 @opindex minline-sqrt-max-throughput
12779 Generate code for inline square roots
12780 using the maximum throughput algorithm.
12782 @item -mno-inline-sqrt
12783 @opindex mno-inline-sqrt
12784 Do not generate inline code for sqrt.
12787 @itemx -mno-fused-madd
12788 @opindex mfused-madd
12789 @opindex mno-fused-madd
12790 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12791 instructions. The default is to use these instructions.
12793 @item -mno-dwarf2-asm
12794 @itemx -mdwarf2-asm
12795 @opindex mno-dwarf2-asm
12796 @opindex mdwarf2-asm
12797 Don't (or do) generate assembler code for the DWARF2 line number debugging
12798 info. This may be useful when not using the GNU assembler.
12800 @item -mearly-stop-bits
12801 @itemx -mno-early-stop-bits
12802 @opindex mearly-stop-bits
12803 @opindex mno-early-stop-bits
12804 Allow stop bits to be placed earlier than immediately preceding the
12805 instruction that triggered the stop bit. This can improve instruction
12806 scheduling, but does not always do so.
12808 @item -mfixed-range=@var{register-range}
12809 @opindex mfixed-range
12810 Generate code treating the given register range as fixed registers.
12811 A fixed register is one that the register allocator can not use. This is
12812 useful when compiling kernel code. A register range is specified as
12813 two registers separated by a dash. Multiple register ranges can be
12814 specified separated by a comma.
12816 @item -mtls-size=@var{tls-size}
12818 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12821 @item -mtune=@var{cpu-type}
12823 Tune the instruction scheduling for a particular CPU, Valid values are
12824 itanium, itanium1, merced, itanium2, and mckinley.
12830 Generate code for a 32-bit or 64-bit environment.
12831 The 32-bit environment sets int, long and pointer to 32 bits.
12832 The 64-bit environment sets int to 32 bits and long and pointer
12833 to 64 bits. These are HP-UX specific flags.
12835 @item -mno-sched-br-data-spec
12836 @itemx -msched-br-data-spec
12837 @opindex mno-sched-br-data-spec
12838 @opindex msched-br-data-spec
12839 (Dis/En)able data speculative scheduling before reload.
12840 This will result in generation of the ld.a instructions and
12841 the corresponding check instructions (ld.c / chk.a).
12842 The default is 'disable'.
12844 @item -msched-ar-data-spec
12845 @itemx -mno-sched-ar-data-spec
12846 @opindex msched-ar-data-spec
12847 @opindex mno-sched-ar-data-spec
12848 (En/Dis)able data speculative scheduling after reload.
12849 This will result in generation of the ld.a instructions and
12850 the corresponding check instructions (ld.c / chk.a).
12851 The default is 'enable'.
12853 @item -mno-sched-control-spec
12854 @itemx -msched-control-spec
12855 @opindex mno-sched-control-spec
12856 @opindex msched-control-spec
12857 (Dis/En)able control speculative scheduling. This feature is
12858 available only during region scheduling (i.e.@: before reload).
12859 This will result in generation of the ld.s instructions and
12860 the corresponding check instructions chk.s .
12861 The default is 'disable'.
12863 @item -msched-br-in-data-spec
12864 @itemx -mno-sched-br-in-data-spec
12865 @opindex msched-br-in-data-spec
12866 @opindex mno-sched-br-in-data-spec
12867 (En/Dis)able speculative scheduling of the instructions that
12868 are dependent on the data speculative loads before reload.
12869 This is effective only with @option{-msched-br-data-spec} enabled.
12870 The default is 'enable'.
12872 @item -msched-ar-in-data-spec
12873 @itemx -mno-sched-ar-in-data-spec
12874 @opindex msched-ar-in-data-spec
12875 @opindex mno-sched-ar-in-data-spec
12876 (En/Dis)able speculative scheduling of the instructions that
12877 are dependent on the data speculative loads after reload.
12878 This is effective only with @option{-msched-ar-data-spec} enabled.
12879 The default is 'enable'.
12881 @item -msched-in-control-spec
12882 @itemx -mno-sched-in-control-spec
12883 @opindex msched-in-control-spec
12884 @opindex mno-sched-in-control-spec
12885 (En/Dis)able speculative scheduling of the instructions that
12886 are dependent on the control speculative loads.
12887 This is effective only with @option{-msched-control-spec} enabled.
12888 The default is 'enable'.
12890 @item -mno-sched-prefer-non-data-spec-insns
12891 @itemx -msched-prefer-non-data-spec-insns
12892 @opindex mno-sched-prefer-non-data-spec-insns
12893 @opindex msched-prefer-non-data-spec-insns
12894 If enabled, data speculative instructions will be chosen for schedule
12895 only if there are no other choices at the moment. This will make
12896 the use of the data speculation much more conservative.
12897 The default is 'disable'.
12899 @item -mno-sched-prefer-non-control-spec-insns
12900 @itemx -msched-prefer-non-control-spec-insns
12901 @opindex mno-sched-prefer-non-control-spec-insns
12902 @opindex msched-prefer-non-control-spec-insns
12903 If enabled, control speculative instructions will be chosen for schedule
12904 only if there are no other choices at the moment. This will make
12905 the use of the control speculation much more conservative.
12906 The default is 'disable'.
12908 @item -mno-sched-count-spec-in-critical-path
12909 @itemx -msched-count-spec-in-critical-path
12910 @opindex mno-sched-count-spec-in-critical-path
12911 @opindex msched-count-spec-in-critical-path
12912 If enabled, speculative dependencies will be considered during
12913 computation of the instructions priorities. This will make the use of the
12914 speculation a bit more conservative.
12915 The default is 'disable'.
12917 @item -msched-spec-ldc
12918 @opindex msched-spec-ldc
12919 Use a simple data speculation check. This option is on by default.
12921 @item -msched-control-spec-ldc
12922 @opindex msched-spec-ldc
12923 Use a simple check for control speculation. This option is on by default.
12925 @item -msched-stop-bits-after-every-cycle
12926 @opindex msched-stop-bits-after-every-cycle
12927 Place a stop bit after every cycle when scheduling. This option is on
12930 @item -msched-fp-mem-deps-zero-cost
12931 @opindex msched-fp-mem-deps-zero-cost
12932 Assume that floating-point stores and loads are not likely to cause a conflict
12933 when placed into the same instruction group. This option is disabled by
12936 @item -msel-sched-dont-check-control-spec
12937 @opindex msel-sched-dont-check-control-spec
12938 Generate checks for control speculation in selective scheduling.
12939 This flag is disabled by default.
12941 @item -msched-max-memory-insns=@var{max-insns}
12942 @opindex msched-max-memory-insns
12943 Limit on the number of memory insns per instruction group, giving lower
12944 priority to subsequent memory insns attempting to schedule in the same
12945 instruction group. Frequently useful to prevent cache bank conflicts.
12946 The default value is 1.
12948 @item -msched-max-memory-insns-hard-limit
12949 @opindex msched-max-memory-insns-hard-limit
12950 Disallow more than `msched-max-memory-insns' in instruction group.
12951 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12952 when limit is reached but may still schedule memory operations.
12956 @node IA-64/VMS Options
12957 @subsection IA-64/VMS Options
12959 These @samp{-m} options are defined for the IA-64/VMS implementations:
12962 @item -mvms-return-codes
12963 @opindex mvms-return-codes
12964 Return VMS condition codes from main. The default is to return POSIX
12965 style condition (e.g.@ error) codes.
12967 @item -mdebug-main=@var{prefix}
12968 @opindex mdebug-main=@var{prefix}
12969 Flag the first routine whose name starts with @var{prefix} as the main
12970 routine for the debugger.
12974 Default to 64bit memory allocation routines.
12978 @subsection LM32 Options
12979 @cindex LM32 options
12981 These @option{-m} options are defined for the Lattice Mico32 architecture:
12984 @item -mbarrel-shift-enabled
12985 @opindex mbarrel-shift-enabled
12986 Enable barrel-shift instructions.
12988 @item -mdivide-enabled
12989 @opindex mdivide-enabled
12990 Enable divide and modulus instructions.
12992 @item -mmultiply-enabled
12993 @opindex multiply-enabled
12994 Enable multiply instructions.
12996 @item -msign-extend-enabled
12997 @opindex msign-extend-enabled
12998 Enable sign extend instructions.
13000 @item -muser-enabled
13001 @opindex muser-enabled
13002 Enable user-defined instructions.
13007 @subsection M32C Options
13008 @cindex M32C options
13011 @item -mcpu=@var{name}
13013 Select the CPU for which code is generated. @var{name} may be one of
13014 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13015 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13016 the M32C/80 series.
13020 Specifies that the program will be run on the simulator. This causes
13021 an alternate runtime library to be linked in which supports, for
13022 example, file I/O@. You must not use this option when generating
13023 programs that will run on real hardware; you must provide your own
13024 runtime library for whatever I/O functions are needed.
13026 @item -memregs=@var{number}
13028 Specifies the number of memory-based pseudo-registers GCC will use
13029 during code generation. These pseudo-registers will be used like real
13030 registers, so there is a tradeoff between GCC's ability to fit the
13031 code into available registers, and the performance penalty of using
13032 memory instead of registers. Note that all modules in a program must
13033 be compiled with the same value for this option. Because of that, you
13034 must not use this option with the default runtime libraries gcc
13039 @node M32R/D Options
13040 @subsection M32R/D Options
13041 @cindex M32R/D options
13043 These @option{-m} options are defined for Renesas M32R/D architectures:
13048 Generate code for the M32R/2@.
13052 Generate code for the M32R/X@.
13056 Generate code for the M32R@. This is the default.
13058 @item -mmodel=small
13059 @opindex mmodel=small
13060 Assume all objects live in the lower 16MB of memory (so that their addresses
13061 can be loaded with the @code{ld24} instruction), and assume all subroutines
13062 are reachable with the @code{bl} instruction.
13063 This is the default.
13065 The addressability of a particular object can be set with the
13066 @code{model} attribute.
13068 @item -mmodel=medium
13069 @opindex mmodel=medium
13070 Assume objects may be anywhere in the 32-bit address space (the compiler
13071 will generate @code{seth/add3} instructions to load their addresses), and
13072 assume all subroutines are reachable with the @code{bl} instruction.
13074 @item -mmodel=large
13075 @opindex mmodel=large
13076 Assume objects may be anywhere in the 32-bit address space (the compiler
13077 will generate @code{seth/add3} instructions to load their addresses), and
13078 assume subroutines may not be reachable with the @code{bl} instruction
13079 (the compiler will generate the much slower @code{seth/add3/jl}
13080 instruction sequence).
13083 @opindex msdata=none
13084 Disable use of the small data area. Variables will be put into
13085 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13086 @code{section} attribute has been specified).
13087 This is the default.
13089 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13090 Objects may be explicitly put in the small data area with the
13091 @code{section} attribute using one of these sections.
13093 @item -msdata=sdata
13094 @opindex msdata=sdata
13095 Put small global and static data in the small data area, but do not
13096 generate special code to reference them.
13099 @opindex msdata=use
13100 Put small global and static data in the small data area, and generate
13101 special instructions to reference them.
13105 @cindex smaller data references
13106 Put global and static objects less than or equal to @var{num} bytes
13107 into the small data or bss sections instead of the normal data or bss
13108 sections. The default value of @var{num} is 8.
13109 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13110 for this option to have any effect.
13112 All modules should be compiled with the same @option{-G @var{num}} value.
13113 Compiling with different values of @var{num} may or may not work; if it
13114 doesn't the linker will give an error message---incorrect code will not be
13119 Makes the M32R specific code in the compiler display some statistics
13120 that might help in debugging programs.
13122 @item -malign-loops
13123 @opindex malign-loops
13124 Align all loops to a 32-byte boundary.
13126 @item -mno-align-loops
13127 @opindex mno-align-loops
13128 Do not enforce a 32-byte alignment for loops. This is the default.
13130 @item -missue-rate=@var{number}
13131 @opindex missue-rate=@var{number}
13132 Issue @var{number} instructions per cycle. @var{number} can only be 1
13135 @item -mbranch-cost=@var{number}
13136 @opindex mbranch-cost=@var{number}
13137 @var{number} can only be 1 or 2. If it is 1 then branches will be
13138 preferred over conditional code, if it is 2, then the opposite will
13141 @item -mflush-trap=@var{number}
13142 @opindex mflush-trap=@var{number}
13143 Specifies the trap number to use to flush the cache. The default is
13144 12. Valid numbers are between 0 and 15 inclusive.
13146 @item -mno-flush-trap
13147 @opindex mno-flush-trap
13148 Specifies that the cache cannot be flushed by using a trap.
13150 @item -mflush-func=@var{name}
13151 @opindex mflush-func=@var{name}
13152 Specifies the name of the operating system function to call to flush
13153 the cache. The default is @emph{_flush_cache}, but a function call
13154 will only be used if a trap is not available.
13156 @item -mno-flush-func
13157 @opindex mno-flush-func
13158 Indicates that there is no OS function for flushing the cache.
13162 @node M680x0 Options
13163 @subsection M680x0 Options
13164 @cindex M680x0 options
13166 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13167 The default settings depend on which architecture was selected when
13168 the compiler was configured; the defaults for the most common choices
13172 @item -march=@var{arch}
13174 Generate code for a specific M680x0 or ColdFire instruction set
13175 architecture. Permissible values of @var{arch} for M680x0
13176 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13177 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13178 architectures are selected according to Freescale's ISA classification
13179 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13180 @samp{isab} and @samp{isac}.
13182 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13183 code for a ColdFire target. The @var{arch} in this macro is one of the
13184 @option{-march} arguments given above.
13186 When used together, @option{-march} and @option{-mtune} select code
13187 that runs on a family of similar processors but that is optimized
13188 for a particular microarchitecture.
13190 @item -mcpu=@var{cpu}
13192 Generate code for a specific M680x0 or ColdFire processor.
13193 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13194 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13195 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13196 below, which also classifies the CPUs into families:
13198 @multitable @columnfractions 0.20 0.80
13199 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13200 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13201 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13202 @item @samp{5206e} @tab @samp{5206e}
13203 @item @samp{5208} @tab @samp{5207} @samp{5208}
13204 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13205 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13206 @item @samp{5216} @tab @samp{5214} @samp{5216}
13207 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13208 @item @samp{5225} @tab @samp{5224} @samp{5225}
13209 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13210 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13211 @item @samp{5249} @tab @samp{5249}
13212 @item @samp{5250} @tab @samp{5250}
13213 @item @samp{5271} @tab @samp{5270} @samp{5271}
13214 @item @samp{5272} @tab @samp{5272}
13215 @item @samp{5275} @tab @samp{5274} @samp{5275}
13216 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13217 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13218 @item @samp{5307} @tab @samp{5307}
13219 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13220 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13221 @item @samp{5407} @tab @samp{5407}
13222 @item @samp{5475} @tab @samp{5470} @samp{5471} @samp{5472} @samp{5473} @samp{5474} @samp{5475} @samp{547x} @samp{5480} @samp{5481} @samp{5482} @samp{5483} @samp{5484} @samp{5485}
13225 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13226 @var{arch} is compatible with @var{cpu}. Other combinations of
13227 @option{-mcpu} and @option{-march} are rejected.
13229 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13230 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13231 where the value of @var{family} is given by the table above.
13233 @item -mtune=@var{tune}
13235 Tune the code for a particular microarchitecture, within the
13236 constraints set by @option{-march} and @option{-mcpu}.
13237 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13238 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13239 and @samp{cpu32}. The ColdFire microarchitectures
13240 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13242 You can also use @option{-mtune=68020-40} for code that needs
13243 to run relatively well on 68020, 68030 and 68040 targets.
13244 @option{-mtune=68020-60} is similar but includes 68060 targets
13245 as well. These two options select the same tuning decisions as
13246 @option{-m68020-40} and @option{-m68020-60} respectively.
13248 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13249 when tuning for 680x0 architecture @var{arch}. It also defines
13250 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13251 option is used. If gcc is tuning for a range of architectures,
13252 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13253 it defines the macros for every architecture in the range.
13255 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13256 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13257 of the arguments given above.
13263 Generate output for a 68000. This is the default
13264 when the compiler is configured for 68000-based systems.
13265 It is equivalent to @option{-march=68000}.
13267 Use this option for microcontrollers with a 68000 or EC000 core,
13268 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13272 Generate output for a 68010. This is the default
13273 when the compiler is configured for 68010-based systems.
13274 It is equivalent to @option{-march=68010}.
13280 Generate output for a 68020. This is the default
13281 when the compiler is configured for 68020-based systems.
13282 It is equivalent to @option{-march=68020}.
13286 Generate output for a 68030. This is the default when the compiler is
13287 configured for 68030-based systems. It is equivalent to
13288 @option{-march=68030}.
13292 Generate output for a 68040. This is the default when the compiler is
13293 configured for 68040-based systems. It is equivalent to
13294 @option{-march=68040}.
13296 This option inhibits the use of 68881/68882 instructions that have to be
13297 emulated by software on the 68040. Use this option if your 68040 does not
13298 have code to emulate those instructions.
13302 Generate output for a 68060. This is the default when the compiler is
13303 configured for 68060-based systems. It is equivalent to
13304 @option{-march=68060}.
13306 This option inhibits the use of 68020 and 68881/68882 instructions that
13307 have to be emulated by software on the 68060. Use this option if your 68060
13308 does not have code to emulate those instructions.
13312 Generate output for a CPU32. This is the default
13313 when the compiler is configured for CPU32-based systems.
13314 It is equivalent to @option{-march=cpu32}.
13316 Use this option for microcontrollers with a
13317 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13318 68336, 68340, 68341, 68349 and 68360.
13322 Generate output for a 520X ColdFire CPU@. This is the default
13323 when the compiler is configured for 520X-based systems.
13324 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13325 in favor of that option.
13327 Use this option for microcontroller with a 5200 core, including
13328 the MCF5202, MCF5203, MCF5204 and MCF5206.
13332 Generate output for a 5206e ColdFire CPU@. The option is now
13333 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13337 Generate output for a member of the ColdFire 528X family.
13338 The option is now deprecated in favor of the equivalent
13339 @option{-mcpu=528x}.
13343 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13344 in favor of the equivalent @option{-mcpu=5307}.
13348 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13349 in favor of the equivalent @option{-mcpu=5407}.
13353 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13354 This includes use of hardware floating point instructions.
13355 The option is equivalent to @option{-mcpu=547x}, and is now
13356 deprecated in favor of that option.
13360 Generate output for a 68040, without using any of the new instructions.
13361 This results in code which can run relatively efficiently on either a
13362 68020/68881 or a 68030 or a 68040. The generated code does use the
13363 68881 instructions that are emulated on the 68040.
13365 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13369 Generate output for a 68060, without using any of the new instructions.
13370 This results in code which can run relatively efficiently on either a
13371 68020/68881 or a 68030 or a 68040. The generated code does use the
13372 68881 instructions that are emulated on the 68060.
13374 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13378 @opindex mhard-float
13380 Generate floating-point instructions. This is the default for 68020
13381 and above, and for ColdFire devices that have an FPU@. It defines the
13382 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13383 on ColdFire targets.
13386 @opindex msoft-float
13387 Do not generate floating-point instructions; use library calls instead.
13388 This is the default for 68000, 68010, and 68832 targets. It is also
13389 the default for ColdFire devices that have no FPU.
13395 Generate (do not generate) ColdFire hardware divide and remainder
13396 instructions. If @option{-march} is used without @option{-mcpu},
13397 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13398 architectures. Otherwise, the default is taken from the target CPU
13399 (either the default CPU, or the one specified by @option{-mcpu}). For
13400 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13401 @option{-mcpu=5206e}.
13403 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13407 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13408 Additionally, parameters passed on the stack are also aligned to a
13409 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13413 Do not consider type @code{int} to be 16 bits wide. This is the default.
13416 @itemx -mno-bitfield
13417 @opindex mnobitfield
13418 @opindex mno-bitfield
13419 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13420 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13424 Do use the bit-field instructions. The @option{-m68020} option implies
13425 @option{-mbitfield}. This is the default if you use a configuration
13426 designed for a 68020.
13430 Use a different function-calling convention, in which functions
13431 that take a fixed number of arguments return with the @code{rtd}
13432 instruction, which pops their arguments while returning. This
13433 saves one instruction in the caller since there is no need to pop
13434 the arguments there.
13436 This calling convention is incompatible with the one normally
13437 used on Unix, so you cannot use it if you need to call libraries
13438 compiled with the Unix compiler.
13440 Also, you must provide function prototypes for all functions that
13441 take variable numbers of arguments (including @code{printf});
13442 otherwise incorrect code will be generated for calls to those
13445 In addition, seriously incorrect code will result if you call a
13446 function with too many arguments. (Normally, extra arguments are
13447 harmlessly ignored.)
13449 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13450 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13454 Do not use the calling conventions selected by @option{-mrtd}.
13455 This is the default.
13458 @itemx -mno-align-int
13459 @opindex malign-int
13460 @opindex mno-align-int
13461 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13462 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13463 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13464 Aligning variables on 32-bit boundaries produces code that runs somewhat
13465 faster on processors with 32-bit busses at the expense of more memory.
13467 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13468 align structures containing the above types differently than
13469 most published application binary interface specifications for the m68k.
13473 Use the pc-relative addressing mode of the 68000 directly, instead of
13474 using a global offset table. At present, this option implies @option{-fpic},
13475 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13476 not presently supported with @option{-mpcrel}, though this could be supported for
13477 68020 and higher processors.
13479 @item -mno-strict-align
13480 @itemx -mstrict-align
13481 @opindex mno-strict-align
13482 @opindex mstrict-align
13483 Do not (do) assume that unaligned memory references will be handled by
13487 Generate code that allows the data segment to be located in a different
13488 area of memory from the text segment. This allows for execute in place in
13489 an environment without virtual memory management. This option implies
13492 @item -mno-sep-data
13493 Generate code that assumes that the data segment follows the text segment.
13494 This is the default.
13496 @item -mid-shared-library
13497 Generate code that supports shared libraries via the library ID method.
13498 This allows for execute in place and shared libraries in an environment
13499 without virtual memory management. This option implies @option{-fPIC}.
13501 @item -mno-id-shared-library
13502 Generate code that doesn't assume ID based shared libraries are being used.
13503 This is the default.
13505 @item -mshared-library-id=n
13506 Specified the identification number of the ID based shared library being
13507 compiled. Specifying a value of 0 will generate more compact code, specifying
13508 other values will force the allocation of that number to the current
13509 library but is no more space or time efficient than omitting this option.
13515 When generating position-independent code for ColdFire, generate code
13516 that works if the GOT has more than 8192 entries. This code is
13517 larger and slower than code generated without this option. On M680x0
13518 processors, this option is not needed; @option{-fPIC} suffices.
13520 GCC normally uses a single instruction to load values from the GOT@.
13521 While this is relatively efficient, it only works if the GOT
13522 is smaller than about 64k. Anything larger causes the linker
13523 to report an error such as:
13525 @cindex relocation truncated to fit (ColdFire)
13527 relocation truncated to fit: R_68K_GOT16O foobar
13530 If this happens, you should recompile your code with @option{-mxgot}.
13531 It should then work with very large GOTs. However, code generated with
13532 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13533 the value of a global symbol.
13535 Note that some linkers, including newer versions of the GNU linker,
13536 can create multiple GOTs and sort GOT entries. If you have such a linker,
13537 you should only need to use @option{-mxgot} when compiling a single
13538 object file that accesses more than 8192 GOT entries. Very few do.
13540 These options have no effect unless GCC is generating
13541 position-independent code.
13545 @node M68hc1x Options
13546 @subsection M68hc1x Options
13547 @cindex M68hc1x options
13549 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13550 microcontrollers. The default values for these options depends on
13551 which style of microcontroller was selected when the compiler was configured;
13552 the defaults for the most common choices are given below.
13559 Generate output for a 68HC11. This is the default
13560 when the compiler is configured for 68HC11-based systems.
13566 Generate output for a 68HC12. This is the default
13567 when the compiler is configured for 68HC12-based systems.
13573 Generate output for a 68HCS12.
13575 @item -mauto-incdec
13576 @opindex mauto-incdec
13577 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13584 Enable the use of 68HC12 min and max instructions.
13587 @itemx -mno-long-calls
13588 @opindex mlong-calls
13589 @opindex mno-long-calls
13590 Treat all calls as being far away (near). If calls are assumed to be
13591 far away, the compiler will use the @code{call} instruction to
13592 call a function and the @code{rtc} instruction for returning.
13596 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13598 @item -msoft-reg-count=@var{count}
13599 @opindex msoft-reg-count
13600 Specify the number of pseudo-soft registers which are used for the
13601 code generation. The maximum number is 32. Using more pseudo-soft
13602 register may or may not result in better code depending on the program.
13603 The default is 4 for 68HC11 and 2 for 68HC12.
13607 @node MCore Options
13608 @subsection MCore Options
13609 @cindex MCore options
13611 These are the @samp{-m} options defined for the Motorola M*Core
13617 @itemx -mno-hardlit
13619 @opindex mno-hardlit
13620 Inline constants into the code stream if it can be done in two
13621 instructions or less.
13627 Use the divide instruction. (Enabled by default).
13629 @item -mrelax-immediate
13630 @itemx -mno-relax-immediate
13631 @opindex mrelax-immediate
13632 @opindex mno-relax-immediate
13633 Allow arbitrary sized immediates in bit operations.
13635 @item -mwide-bitfields
13636 @itemx -mno-wide-bitfields
13637 @opindex mwide-bitfields
13638 @opindex mno-wide-bitfields
13639 Always treat bit-fields as int-sized.
13641 @item -m4byte-functions
13642 @itemx -mno-4byte-functions
13643 @opindex m4byte-functions
13644 @opindex mno-4byte-functions
13645 Force all functions to be aligned to a four byte boundary.
13647 @item -mcallgraph-data
13648 @itemx -mno-callgraph-data
13649 @opindex mcallgraph-data
13650 @opindex mno-callgraph-data
13651 Emit callgraph information.
13654 @itemx -mno-slow-bytes
13655 @opindex mslow-bytes
13656 @opindex mno-slow-bytes
13657 Prefer word access when reading byte quantities.
13659 @item -mlittle-endian
13660 @itemx -mbig-endian
13661 @opindex mlittle-endian
13662 @opindex mbig-endian
13663 Generate code for a little endian target.
13669 Generate code for the 210 processor.
13673 Assume that run-time support has been provided and so omit the
13674 simulator library (@file{libsim.a)} from the linker command line.
13676 @item -mstack-increment=@var{size}
13677 @opindex mstack-increment
13678 Set the maximum amount for a single stack increment operation. Large
13679 values can increase the speed of programs which contain functions
13680 that need a large amount of stack space, but they can also trigger a
13681 segmentation fault if the stack is extended too much. The default
13687 @subsection MeP Options
13688 @cindex MeP options
13694 Enables the @code{abs} instruction, which is the absolute difference
13695 between two registers.
13699 Enables all the optional instructions - average, multiply, divide, bit
13700 operations, leading zero, absolute difference, min/max, clip, and
13706 Enables the @code{ave} instruction, which computes the average of two
13709 @item -mbased=@var{n}
13711 Variables of size @var{n} bytes or smaller will be placed in the
13712 @code{.based} section by default. Based variables use the @code{$tp}
13713 register as a base register, and there is a 128 byte limit to the
13714 @code{.based} section.
13718 Enables the bit operation instructions - bit test (@code{btstm}), set
13719 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13720 test-and-set (@code{tas}).
13722 @item -mc=@var{name}
13724 Selects which section constant data will be placed in. @var{name} may
13725 be @code{tiny}, @code{near}, or @code{far}.
13729 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13730 useful unless you also provide @code{-mminmax}.
13732 @item -mconfig=@var{name}
13734 Selects one of the build-in core configurations. Each MeP chip has
13735 one or more modules in it; each module has a core CPU and a variety of
13736 coprocessors, optional instructions, and peripherals. The
13737 @code{MeP-Integrator} tool, not part of GCC, provides these
13738 configurations through this option; using this option is the same as
13739 using all the corresponding command line options. The default
13740 configuration is @code{default}.
13744 Enables the coprocessor instructions. By default, this is a 32-bit
13745 coprocessor. Note that the coprocessor is normally enabled via the
13746 @code{-mconfig=} option.
13750 Enables the 32-bit coprocessor's instructions.
13754 Enables the 64-bit coprocessor's instructions.
13758 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13762 Causes constant variables to be placed in the @code{.near} section.
13766 Enables the @code{div} and @code{divu} instructions.
13770 Generate big-endian code.
13774 Generate little-endian code.
13776 @item -mio-volatile
13777 @opindex mio-volatile
13778 Tells the compiler that any variable marked with the @code{io}
13779 attribute is to be considered volatile.
13783 Causes variables to be assigned to the @code{.far} section by default.
13787 Enables the @code{leadz} (leading zero) instruction.
13791 Causes variables to be assigned to the @code{.near} section by default.
13795 Enables the @code{min} and @code{max} instructions.
13799 Enables the multiplication and multiply-accumulate instructions.
13803 Disables all the optional instructions enabled by @code{-mall-opts}.
13807 Enables the @code{repeat} and @code{erepeat} instructions, used for
13808 low-overhead looping.
13812 Causes all variables to default to the @code{.tiny} section. Note
13813 that there is a 65536 byte limit to this section. Accesses to these
13814 variables use the @code{%gp} base register.
13818 Enables the saturation instructions. Note that the compiler does not
13819 currently generate these itself, but this option is included for
13820 compatibility with other tools, like @code{as}.
13824 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13828 Link the simulator runtime libraries.
13832 Link the simulator runtime libraries, excluding built-in support
13833 for reset and exception vectors and tables.
13837 Causes all functions to default to the @code{.far} section. Without
13838 this option, functions default to the @code{.near} section.
13840 @item -mtiny=@var{n}
13842 Variables that are @var{n} bytes or smaller will be allocated to the
13843 @code{.tiny} section. These variables use the @code{$gp} base
13844 register. The default for this option is 4, but note that there's a
13845 65536 byte limit to the @code{.tiny} section.
13849 @node MicroBlaze Options
13850 @subsection MicroBlaze Options
13851 @cindex MicroBlaze Options
13856 @opindex msoft-float
13857 Use software emulation for floating point (default).
13860 @opindex mhard-float
13861 Use hardware floating point instructions.
13865 Do not optimize block moves, use @code{memcpy}.
13867 @item -mno-clearbss
13868 @opindex mno-clearbss
13869 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13871 @item -mcpu=@var{cpu-type}
13873 Use features of and schedule code for given CPU.
13874 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13875 where @var{X} is a major version, @var{YY} is the minor version, and
13876 @var{Z} is compatiblity code. Example values are @samp{v3.00.a},
13877 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13879 @item -mxl-soft-mul
13880 @opindex mxl-soft-mul
13881 Use software multiply emulation (default).
13883 @item -mxl-soft-div
13884 @opindex mxl-soft-div
13885 Use software emulation for divides (default).
13887 @item -mxl-barrel-shift
13888 @opindex mxl-barrel-shift
13889 Use the hardware barrel shifter.
13891 @item -mxl-pattern-compare
13892 @opindex mxl-pattern-compare
13893 Use pattern compare instructions.
13895 @item -msmall-divides
13896 @opindex msmall-divides
13897 Use table lookup optimization for small signed integer divisions.
13899 @item -mxl-stack-check
13900 @opindex mxl-stack-check
13901 This option is deprecated. Use -fstack-check instead.
13904 @opindex mxl-gp-opt
13905 Use GP relative sdata/sbss sections.
13907 @item -mxl-multiply-high
13908 @opindex mxl-multiply-high
13909 Use multiply high instructions for high part of 32x32 multiply.
13911 @item -mxl-float-convert
13912 @opindex mxl-float-convert
13913 Use hardware floating point converstion instructions.
13915 @item -mxl-float-sqrt
13916 @opindex mxl-float-sqrt
13917 Use hardware floating point square root instruction.
13919 @item -mxl-mode-@var{app-model}
13920 Select application model @var{app-model}. Valid models are
13923 normal executable (default), uses startup code @file{crt0.o}.
13926 for use with Xilinx Microprocessor Debugger (XMD) based
13927 software intrusive debug agent called xmdstub. This uses startup file
13928 @file{crt1.o} and sets the start address of the program to be 0x800.
13931 for applications that are loaded using a bootloader.
13932 This model uses startup file @file{crt2.o} which does not contain a processor
13933 reset vector handler. This is suitable for transferring control on a
13934 processor reset to the bootloader rather than the application.
13937 for applications that do not require any of the
13938 MicroBlaze vectors. This option may be useful for applications running
13939 within a monitoring application. This model uses @file{crt3.o} as a startup file.
13942 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
13943 @option{-mxl-mode-@var{app-model}}.
13948 @subsection MIPS Options
13949 @cindex MIPS options
13955 Generate big-endian code.
13959 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13962 @item -march=@var{arch}
13964 Generate code that will run on @var{arch}, which can be the name of a
13965 generic MIPS ISA, or the name of a particular processor.
13967 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13968 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13969 The processor names are:
13970 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13971 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13972 @samp{5kc}, @samp{5kf},
13974 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13975 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13976 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13977 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13978 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13979 @samp{loongson2e}, @samp{loongson2f},
13983 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13984 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13985 @samp{rm7000}, @samp{rm9000},
13986 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13989 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13990 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13992 The special value @samp{from-abi} selects the
13993 most compatible architecture for the selected ABI (that is,
13994 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13996 Native Linux/GNU toolchains also support the value @samp{native},
13997 which selects the best architecture option for the host processor.
13998 @option{-march=native} has no effect if GCC does not recognize
14001 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14002 (for example, @samp{-march=r2k}). Prefixes are optional, and
14003 @samp{vr} may be written @samp{r}.
14005 Names of the form @samp{@var{n}f2_1} refer to processors with
14006 FPUs clocked at half the rate of the core, names of the form
14007 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14008 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14009 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14010 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14011 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14012 accepted as synonyms for @samp{@var{n}f1_1}.
14014 GCC defines two macros based on the value of this option. The first
14015 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14016 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14017 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14018 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14019 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14021 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14022 above. In other words, it will have the full prefix and will not
14023 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14024 the macro names the resolved architecture (either @samp{"mips1"} or
14025 @samp{"mips3"}). It names the default architecture when no
14026 @option{-march} option is given.
14028 @item -mtune=@var{arch}
14030 Optimize for @var{arch}. Among other things, this option controls
14031 the way instructions are scheduled, and the perceived cost of arithmetic
14032 operations. The list of @var{arch} values is the same as for
14035 When this option is not used, GCC will optimize for the processor
14036 specified by @option{-march}. By using @option{-march} and
14037 @option{-mtune} together, it is possible to generate code that will
14038 run on a family of processors, but optimize the code for one
14039 particular member of that family.
14041 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14042 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14043 @samp{-march} ones described above.
14047 Equivalent to @samp{-march=mips1}.
14051 Equivalent to @samp{-march=mips2}.
14055 Equivalent to @samp{-march=mips3}.
14059 Equivalent to @samp{-march=mips4}.
14063 Equivalent to @samp{-march=mips32}.
14067 Equivalent to @samp{-march=mips32r2}.
14071 Equivalent to @samp{-march=mips64}.
14075 Equivalent to @samp{-march=mips64r2}.
14080 @opindex mno-mips16
14081 Generate (do not generate) MIPS16 code. If GCC is targetting a
14082 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14084 MIPS16 code generation can also be controlled on a per-function basis
14085 by means of @code{mips16} and @code{nomips16} attributes.
14086 @xref{Function Attributes}, for more information.
14088 @item -mflip-mips16
14089 @opindex mflip-mips16
14090 Generate MIPS16 code on alternating functions. This option is provided
14091 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14092 not intended for ordinary use in compiling user code.
14094 @item -minterlink-mips16
14095 @itemx -mno-interlink-mips16
14096 @opindex minterlink-mips16
14097 @opindex mno-interlink-mips16
14098 Require (do not require) that non-MIPS16 code be link-compatible with
14101 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14102 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14103 therefore disables direct jumps unless GCC knows that the target of the
14104 jump is not MIPS16.
14116 Generate code for the given ABI@.
14118 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14119 generates 64-bit code when you select a 64-bit architecture, but you
14120 can use @option{-mgp32} to get 32-bit code instead.
14122 For information about the O64 ABI, see
14123 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14125 GCC supports a variant of the o32 ABI in which floating-point registers
14126 are 64 rather than 32 bits wide. You can select this combination with
14127 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14128 and @samp{mfhc1} instructions and is therefore only supported for
14129 MIPS32R2 processors.
14131 The register assignments for arguments and return values remain the
14132 same, but each scalar value is passed in a single 64-bit register
14133 rather than a pair of 32-bit registers. For example, scalar
14134 floating-point values are returned in @samp{$f0} only, not a
14135 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14136 remains the same, but all 64 bits are saved.
14139 @itemx -mno-abicalls
14141 @opindex mno-abicalls
14142 Generate (do not generate) code that is suitable for SVR4-style
14143 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14148 Generate (do not generate) code that is fully position-independent,
14149 and that can therefore be linked into shared libraries. This option
14150 only affects @option{-mabicalls}.
14152 All @option{-mabicalls} code has traditionally been position-independent,
14153 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14154 as an extension, the GNU toolchain allows executables to use absolute
14155 accesses for locally-binding symbols. It can also use shorter GP
14156 initialization sequences and generate direct calls to locally-defined
14157 functions. This mode is selected by @option{-mno-shared}.
14159 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14160 objects that can only be linked by the GNU linker. However, the option
14161 does not affect the ABI of the final executable; it only affects the ABI
14162 of relocatable objects. Using @option{-mno-shared} will generally make
14163 executables both smaller and quicker.
14165 @option{-mshared} is the default.
14171 Assume (do not assume) that the static and dynamic linkers
14172 support PLTs and copy relocations. This option only affects
14173 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14174 has no effect without @samp{-msym32}.
14176 You can make @option{-mplt} the default by configuring
14177 GCC with @option{--with-mips-plt}. The default is
14178 @option{-mno-plt} otherwise.
14184 Lift (do not lift) the usual restrictions on the size of the global
14187 GCC normally uses a single instruction to load values from the GOT@.
14188 While this is relatively efficient, it will only work if the GOT
14189 is smaller than about 64k. Anything larger will cause the linker
14190 to report an error such as:
14192 @cindex relocation truncated to fit (MIPS)
14194 relocation truncated to fit: R_MIPS_GOT16 foobar
14197 If this happens, you should recompile your code with @option{-mxgot}.
14198 It should then work with very large GOTs, although it will also be
14199 less efficient, since it will take three instructions to fetch the
14200 value of a global symbol.
14202 Note that some linkers can create multiple GOTs. If you have such a
14203 linker, you should only need to use @option{-mxgot} when a single object
14204 file accesses more than 64k's worth of GOT entries. Very few do.
14206 These options have no effect unless GCC is generating position
14211 Assume that general-purpose registers are 32 bits wide.
14215 Assume that general-purpose registers are 64 bits wide.
14219 Assume that floating-point registers are 32 bits wide.
14223 Assume that floating-point registers are 64 bits wide.
14226 @opindex mhard-float
14227 Use floating-point coprocessor instructions.
14230 @opindex msoft-float
14231 Do not use floating-point coprocessor instructions. Implement
14232 floating-point calculations using library calls instead.
14234 @item -msingle-float
14235 @opindex msingle-float
14236 Assume that the floating-point coprocessor only supports single-precision
14239 @item -mdouble-float
14240 @opindex mdouble-float
14241 Assume that the floating-point coprocessor supports double-precision
14242 operations. This is the default.
14248 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14249 implement atomic memory built-in functions. When neither option is
14250 specified, GCC will use the instructions if the target architecture
14253 @option{-mllsc} is useful if the runtime environment can emulate the
14254 instructions and @option{-mno-llsc} can be useful when compiling for
14255 nonstandard ISAs. You can make either option the default by
14256 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14257 respectively. @option{--with-llsc} is the default for some
14258 configurations; see the installation documentation for details.
14264 Use (do not use) revision 1 of the MIPS DSP ASE@.
14265 @xref{MIPS DSP Built-in Functions}. This option defines the
14266 preprocessor macro @samp{__mips_dsp}. It also defines
14267 @samp{__mips_dsp_rev} to 1.
14273 Use (do not use) revision 2 of the MIPS DSP ASE@.
14274 @xref{MIPS DSP Built-in Functions}. This option defines the
14275 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14276 It also defines @samp{__mips_dsp_rev} to 2.
14279 @itemx -mno-smartmips
14280 @opindex msmartmips
14281 @opindex mno-smartmips
14282 Use (do not use) the MIPS SmartMIPS ASE.
14284 @item -mpaired-single
14285 @itemx -mno-paired-single
14286 @opindex mpaired-single
14287 @opindex mno-paired-single
14288 Use (do not use) paired-single floating-point instructions.
14289 @xref{MIPS Paired-Single Support}. This option requires
14290 hardware floating-point support to be enabled.
14296 Use (do not use) MIPS Digital Media Extension instructions.
14297 This option can only be used when generating 64-bit code and requires
14298 hardware floating-point support to be enabled.
14303 @opindex mno-mips3d
14304 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14305 The option @option{-mips3d} implies @option{-mpaired-single}.
14311 Use (do not use) MT Multithreading instructions.
14315 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14316 an explanation of the default and the way that the pointer size is
14321 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14323 The default size of @code{int}s, @code{long}s and pointers depends on
14324 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14325 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14326 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14327 or the same size as integer registers, whichever is smaller.
14333 Assume (do not assume) that all symbols have 32-bit values, regardless
14334 of the selected ABI@. This option is useful in combination with
14335 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14336 to generate shorter and faster references to symbolic addresses.
14340 Put definitions of externally-visible data in a small data section
14341 if that data is no bigger than @var{num} bytes. GCC can then access
14342 the data more efficiently; see @option{-mgpopt} for details.
14344 The default @option{-G} option depends on the configuration.
14346 @item -mlocal-sdata
14347 @itemx -mno-local-sdata
14348 @opindex mlocal-sdata
14349 @opindex mno-local-sdata
14350 Extend (do not extend) the @option{-G} behavior to local data too,
14351 such as to static variables in C@. @option{-mlocal-sdata} is the
14352 default for all configurations.
14354 If the linker complains that an application is using too much small data,
14355 you might want to try rebuilding the less performance-critical parts with
14356 @option{-mno-local-sdata}. You might also want to build large
14357 libraries with @option{-mno-local-sdata}, so that the libraries leave
14358 more room for the main program.
14360 @item -mextern-sdata
14361 @itemx -mno-extern-sdata
14362 @opindex mextern-sdata
14363 @opindex mno-extern-sdata
14364 Assume (do not assume) that externally-defined data will be in
14365 a small data section if that data is within the @option{-G} limit.
14366 @option{-mextern-sdata} is the default for all configurations.
14368 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14369 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14370 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14371 is placed in a small data section. If @var{Var} is defined by another
14372 module, you must either compile that module with a high-enough
14373 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14374 definition. If @var{Var} is common, you must link the application
14375 with a high-enough @option{-G} setting.
14377 The easiest way of satisfying these restrictions is to compile
14378 and link every module with the same @option{-G} option. However,
14379 you may wish to build a library that supports several different
14380 small data limits. You can do this by compiling the library with
14381 the highest supported @option{-G} setting and additionally using
14382 @option{-mno-extern-sdata} to stop the library from making assumptions
14383 about externally-defined data.
14389 Use (do not use) GP-relative accesses for symbols that are known to be
14390 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14391 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14394 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14395 might not hold the value of @code{_gp}. For example, if the code is
14396 part of a library that might be used in a boot monitor, programs that
14397 call boot monitor routines will pass an unknown value in @code{$gp}.
14398 (In such situations, the boot monitor itself would usually be compiled
14399 with @option{-G0}.)
14401 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14402 @option{-mno-extern-sdata}.
14404 @item -membedded-data
14405 @itemx -mno-embedded-data
14406 @opindex membedded-data
14407 @opindex mno-embedded-data
14408 Allocate variables to the read-only data section first if possible, then
14409 next in the small data section if possible, otherwise in data. This gives
14410 slightly slower code than the default, but reduces the amount of RAM required
14411 when executing, and thus may be preferred for some embedded systems.
14413 @item -muninit-const-in-rodata
14414 @itemx -mno-uninit-const-in-rodata
14415 @opindex muninit-const-in-rodata
14416 @opindex mno-uninit-const-in-rodata
14417 Put uninitialized @code{const} variables in the read-only data section.
14418 This option is only meaningful in conjunction with @option{-membedded-data}.
14420 @item -mcode-readable=@var{setting}
14421 @opindex mcode-readable
14422 Specify whether GCC may generate code that reads from executable sections.
14423 There are three possible settings:
14426 @item -mcode-readable=yes
14427 Instructions may freely access executable sections. This is the
14430 @item -mcode-readable=pcrel
14431 MIPS16 PC-relative load instructions can access executable sections,
14432 but other instructions must not do so. This option is useful on 4KSc
14433 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14434 It is also useful on processors that can be configured to have a dual
14435 instruction/data SRAM interface and that, like the M4K, automatically
14436 redirect PC-relative loads to the instruction RAM.
14438 @item -mcode-readable=no
14439 Instructions must not access executable sections. This option can be
14440 useful on targets that are configured to have a dual instruction/data
14441 SRAM interface but that (unlike the M4K) do not automatically redirect
14442 PC-relative loads to the instruction RAM.
14445 @item -msplit-addresses
14446 @itemx -mno-split-addresses
14447 @opindex msplit-addresses
14448 @opindex mno-split-addresses
14449 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14450 relocation operators. This option has been superseded by
14451 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14453 @item -mexplicit-relocs
14454 @itemx -mno-explicit-relocs
14455 @opindex mexplicit-relocs
14456 @opindex mno-explicit-relocs
14457 Use (do not use) assembler relocation operators when dealing with symbolic
14458 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14459 is to use assembler macros instead.
14461 @option{-mexplicit-relocs} is the default if GCC was configured
14462 to use an assembler that supports relocation operators.
14464 @item -mcheck-zero-division
14465 @itemx -mno-check-zero-division
14466 @opindex mcheck-zero-division
14467 @opindex mno-check-zero-division
14468 Trap (do not trap) on integer division by zero.
14470 The default is @option{-mcheck-zero-division}.
14472 @item -mdivide-traps
14473 @itemx -mdivide-breaks
14474 @opindex mdivide-traps
14475 @opindex mdivide-breaks
14476 MIPS systems check for division by zero by generating either a
14477 conditional trap or a break instruction. Using traps results in
14478 smaller code, but is only supported on MIPS II and later. Also, some
14479 versions of the Linux kernel have a bug that prevents trap from
14480 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14481 allow conditional traps on architectures that support them and
14482 @option{-mdivide-breaks} to force the use of breaks.
14484 The default is usually @option{-mdivide-traps}, but this can be
14485 overridden at configure time using @option{--with-divide=breaks}.
14486 Divide-by-zero checks can be completely disabled using
14487 @option{-mno-check-zero-division}.
14492 @opindex mno-memcpy
14493 Force (do not force) the use of @code{memcpy()} for non-trivial block
14494 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14495 most constant-sized copies.
14498 @itemx -mno-long-calls
14499 @opindex mlong-calls
14500 @opindex mno-long-calls
14501 Disable (do not disable) use of the @code{jal} instruction. Calling
14502 functions using @code{jal} is more efficient but requires the caller
14503 and callee to be in the same 256 megabyte segment.
14505 This option has no effect on abicalls code. The default is
14506 @option{-mno-long-calls}.
14512 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14513 instructions, as provided by the R4650 ISA@.
14516 @itemx -mno-fused-madd
14517 @opindex mfused-madd
14518 @opindex mno-fused-madd
14519 Enable (disable) use of the floating point multiply-accumulate
14520 instructions, when they are available. The default is
14521 @option{-mfused-madd}.
14523 When multiply-accumulate instructions are used, the intermediate
14524 product is calculated to infinite precision and is not subject to
14525 the FCSR Flush to Zero bit. This may be undesirable in some
14530 Tell the MIPS assembler to not run its preprocessor over user
14531 assembler files (with a @samp{.s} suffix) when assembling them.
14534 @itemx -mno-fix-r4000
14535 @opindex mfix-r4000
14536 @opindex mno-fix-r4000
14537 Work around certain R4000 CPU errata:
14540 A double-word or a variable shift may give an incorrect result if executed
14541 immediately after starting an integer division.
14543 A double-word or a variable shift may give an incorrect result if executed
14544 while an integer multiplication is in progress.
14546 An integer division may give an incorrect result if started in a delay slot
14547 of a taken branch or a jump.
14551 @itemx -mno-fix-r4400
14552 @opindex mfix-r4400
14553 @opindex mno-fix-r4400
14554 Work around certain R4400 CPU errata:
14557 A double-word or a variable shift may give an incorrect result if executed
14558 immediately after starting an integer division.
14562 @itemx -mno-fix-r10000
14563 @opindex mfix-r10000
14564 @opindex mno-fix-r10000
14565 Work around certain R10000 errata:
14568 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14569 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14572 This option can only be used if the target architecture supports
14573 branch-likely instructions. @option{-mfix-r10000} is the default when
14574 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14578 @itemx -mno-fix-vr4120
14579 @opindex mfix-vr4120
14580 Work around certain VR4120 errata:
14583 @code{dmultu} does not always produce the correct result.
14585 @code{div} and @code{ddiv} do not always produce the correct result if one
14586 of the operands is negative.
14588 The workarounds for the division errata rely on special functions in
14589 @file{libgcc.a}. At present, these functions are only provided by
14590 the @code{mips64vr*-elf} configurations.
14592 Other VR4120 errata require a nop to be inserted between certain pairs of
14593 instructions. These errata are handled by the assembler, not by GCC itself.
14596 @opindex mfix-vr4130
14597 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14598 workarounds are implemented by the assembler rather than by GCC,
14599 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14600 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14601 instructions are available instead.
14604 @itemx -mno-fix-sb1
14606 Work around certain SB-1 CPU core errata.
14607 (This flag currently works around the SB-1 revision 2
14608 ``F1'' and ``F2'' floating point errata.)
14610 @item -mr10k-cache-barrier=@var{setting}
14611 @opindex mr10k-cache-barrier
14612 Specify whether GCC should insert cache barriers to avoid the
14613 side-effects of speculation on R10K processors.
14615 In common with many processors, the R10K tries to predict the outcome
14616 of a conditional branch and speculatively executes instructions from
14617 the ``taken'' branch. It later aborts these instructions if the
14618 predicted outcome was wrong. However, on the R10K, even aborted
14619 instructions can have side effects.
14621 This problem only affects kernel stores and, depending on the system,
14622 kernel loads. As an example, a speculatively-executed store may load
14623 the target memory into cache and mark the cache line as dirty, even if
14624 the store itself is later aborted. If a DMA operation writes to the
14625 same area of memory before the ``dirty'' line is flushed, the cached
14626 data will overwrite the DMA-ed data. See the R10K processor manual
14627 for a full description, including other potential problems.
14629 One workaround is to insert cache barrier instructions before every memory
14630 access that might be speculatively executed and that might have side
14631 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14632 controls GCC's implementation of this workaround. It assumes that
14633 aborted accesses to any byte in the following regions will not have
14638 the memory occupied by the current function's stack frame;
14641 the memory occupied by an incoming stack argument;
14644 the memory occupied by an object with a link-time-constant address.
14647 It is the kernel's responsibility to ensure that speculative
14648 accesses to these regions are indeed safe.
14650 If the input program contains a function declaration such as:
14656 then the implementation of @code{foo} must allow @code{j foo} and
14657 @code{jal foo} to be executed speculatively. GCC honors this
14658 restriction for functions it compiles itself. It expects non-GCC
14659 functions (such as hand-written assembly code) to do the same.
14661 The option has three forms:
14664 @item -mr10k-cache-barrier=load-store
14665 Insert a cache barrier before a load or store that might be
14666 speculatively executed and that might have side effects even
14669 @item -mr10k-cache-barrier=store
14670 Insert a cache barrier before a store that might be speculatively
14671 executed and that might have side effects even if aborted.
14673 @item -mr10k-cache-barrier=none
14674 Disable the insertion of cache barriers. This is the default setting.
14677 @item -mflush-func=@var{func}
14678 @itemx -mno-flush-func
14679 @opindex mflush-func
14680 Specifies the function to call to flush the I and D caches, or to not
14681 call any such function. If called, the function must take the same
14682 arguments as the common @code{_flush_func()}, that is, the address of the
14683 memory range for which the cache is being flushed, the size of the
14684 memory range, and the number 3 (to flush both caches). The default
14685 depends on the target GCC was configured for, but commonly is either
14686 @samp{_flush_func} or @samp{__cpu_flush}.
14688 @item mbranch-cost=@var{num}
14689 @opindex mbranch-cost
14690 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14691 This cost is only a heuristic and is not guaranteed to produce
14692 consistent results across releases. A zero cost redundantly selects
14693 the default, which is based on the @option{-mtune} setting.
14695 @item -mbranch-likely
14696 @itemx -mno-branch-likely
14697 @opindex mbranch-likely
14698 @opindex mno-branch-likely
14699 Enable or disable use of Branch Likely instructions, regardless of the
14700 default for the selected architecture. By default, Branch Likely
14701 instructions may be generated if they are supported by the selected
14702 architecture. An exception is for the MIPS32 and MIPS64 architectures
14703 and processors which implement those architectures; for those, Branch
14704 Likely instructions will not be generated by default because the MIPS32
14705 and MIPS64 architectures specifically deprecate their use.
14707 @item -mfp-exceptions
14708 @itemx -mno-fp-exceptions
14709 @opindex mfp-exceptions
14710 Specifies whether FP exceptions are enabled. This affects how we schedule
14711 FP instructions for some processors. The default is that FP exceptions are
14714 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14715 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14718 @item -mvr4130-align
14719 @itemx -mno-vr4130-align
14720 @opindex mvr4130-align
14721 The VR4130 pipeline is two-way superscalar, but can only issue two
14722 instructions together if the first one is 8-byte aligned. When this
14723 option is enabled, GCC will align pairs of instructions that it
14724 thinks should execute in parallel.
14726 This option only has an effect when optimizing for the VR4130.
14727 It normally makes code faster, but at the expense of making it bigger.
14728 It is enabled by default at optimization level @option{-O3}.
14733 Enable (disable) generation of @code{synci} instructions on
14734 architectures that support it. The @code{synci} instructions (if
14735 enabled) will be generated when @code{__builtin___clear_cache()} is
14738 This option defaults to @code{-mno-synci}, but the default can be
14739 overridden by configuring with @code{--with-synci}.
14741 When compiling code for single processor systems, it is generally safe
14742 to use @code{synci}. However, on many multi-core (SMP) systems, it
14743 will not invalidate the instruction caches on all cores and may lead
14744 to undefined behavior.
14746 @item -mrelax-pic-calls
14747 @itemx -mno-relax-pic-calls
14748 @opindex mrelax-pic-calls
14749 Try to turn PIC calls that are normally dispatched via register
14750 @code{$25} into direct calls. This is only possible if the linker can
14751 resolve the destination at link-time and if the destination is within
14752 range for a direct call.
14754 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14755 an assembler and a linker that supports the @code{.reloc} assembly
14756 directive and @code{-mexplicit-relocs} is in effect. With
14757 @code{-mno-explicit-relocs}, this optimization can be performed by the
14758 assembler and the linker alone without help from the compiler.
14760 @item -mmcount-ra-address
14761 @itemx -mno-mcount-ra-address
14762 @opindex mmcount-ra-address
14763 @opindex mno-mcount-ra-address
14764 Emit (do not emit) code that allows @code{_mcount} to modify the
14765 calling function's return address. When enabled, this option extends
14766 the usual @code{_mcount} interface with a new @var{ra-address}
14767 parameter, which has type @code{intptr_t *} and is passed in register
14768 @code{$12}. @code{_mcount} can then modify the return address by
14769 doing both of the following:
14772 Returning the new address in register @code{$31}.
14774 Storing the new address in @code{*@var{ra-address}},
14775 if @var{ra-address} is nonnull.
14778 The default is @option{-mno-mcount-ra-address}.
14783 @subsection MMIX Options
14784 @cindex MMIX Options
14786 These options are defined for the MMIX:
14790 @itemx -mno-libfuncs
14792 @opindex mno-libfuncs
14793 Specify that intrinsic library functions are being compiled, passing all
14794 values in registers, no matter the size.
14797 @itemx -mno-epsilon
14799 @opindex mno-epsilon
14800 Generate floating-point comparison instructions that compare with respect
14801 to the @code{rE} epsilon register.
14803 @item -mabi=mmixware
14805 @opindex mabi=mmixware
14807 Generate code that passes function parameters and return values that (in
14808 the called function) are seen as registers @code{$0} and up, as opposed to
14809 the GNU ABI which uses global registers @code{$231} and up.
14811 @item -mzero-extend
14812 @itemx -mno-zero-extend
14813 @opindex mzero-extend
14814 @opindex mno-zero-extend
14815 When reading data from memory in sizes shorter than 64 bits, use (do not
14816 use) zero-extending load instructions by default, rather than
14817 sign-extending ones.
14820 @itemx -mno-knuthdiv
14822 @opindex mno-knuthdiv
14823 Make the result of a division yielding a remainder have the same sign as
14824 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14825 remainder follows the sign of the dividend. Both methods are
14826 arithmetically valid, the latter being almost exclusively used.
14828 @item -mtoplevel-symbols
14829 @itemx -mno-toplevel-symbols
14830 @opindex mtoplevel-symbols
14831 @opindex mno-toplevel-symbols
14832 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14833 code can be used with the @code{PREFIX} assembly directive.
14837 Generate an executable in the ELF format, rather than the default
14838 @samp{mmo} format used by the @command{mmix} simulator.
14840 @item -mbranch-predict
14841 @itemx -mno-branch-predict
14842 @opindex mbranch-predict
14843 @opindex mno-branch-predict
14844 Use (do not use) the probable-branch instructions, when static branch
14845 prediction indicates a probable branch.
14847 @item -mbase-addresses
14848 @itemx -mno-base-addresses
14849 @opindex mbase-addresses
14850 @opindex mno-base-addresses
14851 Generate (do not generate) code that uses @emph{base addresses}. Using a
14852 base address automatically generates a request (handled by the assembler
14853 and the linker) for a constant to be set up in a global register. The
14854 register is used for one or more base address requests within the range 0
14855 to 255 from the value held in the register. The generally leads to short
14856 and fast code, but the number of different data items that can be
14857 addressed is limited. This means that a program that uses lots of static
14858 data may require @option{-mno-base-addresses}.
14860 @item -msingle-exit
14861 @itemx -mno-single-exit
14862 @opindex msingle-exit
14863 @opindex mno-single-exit
14864 Force (do not force) generated code to have a single exit point in each
14868 @node MN10300 Options
14869 @subsection MN10300 Options
14870 @cindex MN10300 options
14872 These @option{-m} options are defined for Matsushita MN10300 architectures:
14877 Generate code to avoid bugs in the multiply instructions for the MN10300
14878 processors. This is the default.
14880 @item -mno-mult-bug
14881 @opindex mno-mult-bug
14882 Do not generate code to avoid bugs in the multiply instructions for the
14883 MN10300 processors.
14887 Generate code which uses features specific to the AM33 processor.
14891 Do not generate code which uses features specific to the AM33 processor. This
14894 @item -mreturn-pointer-on-d0
14895 @opindex mreturn-pointer-on-d0
14896 When generating a function which returns a pointer, return the pointer
14897 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14898 only in a0, and attempts to call such functions without a prototype
14899 would result in errors. Note that this option is on by default; use
14900 @option{-mno-return-pointer-on-d0} to disable it.
14904 Do not link in the C run-time initialization object file.
14908 Indicate to the linker that it should perform a relaxation optimization pass
14909 to shorten branches, calls and absolute memory addresses. This option only
14910 has an effect when used on the command line for the final link step.
14912 This option makes symbolic debugging impossible.
14915 @node PDP-11 Options
14916 @subsection PDP-11 Options
14917 @cindex PDP-11 Options
14919 These options are defined for the PDP-11:
14924 Use hardware FPP floating point. This is the default. (FIS floating
14925 point on the PDP-11/40 is not supported.)
14928 @opindex msoft-float
14929 Do not use hardware floating point.
14933 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14937 Return floating-point results in memory. This is the default.
14941 Generate code for a PDP-11/40.
14945 Generate code for a PDP-11/45. This is the default.
14949 Generate code for a PDP-11/10.
14951 @item -mbcopy-builtin
14952 @opindex mbcopy-builtin
14953 Use inline @code{movmemhi} patterns for copying memory. This is the
14958 Do not use inline @code{movmemhi} patterns for copying memory.
14964 Use 16-bit @code{int}. This is the default.
14970 Use 32-bit @code{int}.
14973 @itemx -mno-float32
14975 @opindex mno-float32
14976 Use 64-bit @code{float}. This is the default.
14979 @itemx -mno-float64
14981 @opindex mno-float64
14982 Use 32-bit @code{float}.
14986 Use @code{abshi2} pattern. This is the default.
14990 Do not use @code{abshi2} pattern.
14992 @item -mbranch-expensive
14993 @opindex mbranch-expensive
14994 Pretend that branches are expensive. This is for experimenting with
14995 code generation only.
14997 @item -mbranch-cheap
14998 @opindex mbranch-cheap
14999 Do not pretend that branches are expensive. This is the default.
15003 Generate code for a system with split I&D@.
15007 Generate code for a system without split I&D@. This is the default.
15011 Use Unix assembler syntax. This is the default when configured for
15012 @samp{pdp11-*-bsd}.
15016 Use DEC assembler syntax. This is the default when configured for any
15017 PDP-11 target other than @samp{pdp11-*-bsd}.
15020 @node picoChip Options
15021 @subsection picoChip Options
15022 @cindex picoChip options
15024 These @samp{-m} options are defined for picoChip implementations:
15028 @item -mae=@var{ae_type}
15030 Set the instruction set, register set, and instruction scheduling
15031 parameters for array element type @var{ae_type}. Supported values
15032 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15034 @option{-mae=ANY} selects a completely generic AE type. Code
15035 generated with this option will run on any of the other AE types. The
15036 code will not be as efficient as it would be if compiled for a specific
15037 AE type, and some types of operation (e.g., multiplication) will not
15038 work properly on all types of AE.
15040 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15041 for compiled code, and is the default.
15043 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15044 option may suffer from poor performance of byte (char) manipulation,
15045 since the DSP AE does not provide hardware support for byte load/stores.
15047 @item -msymbol-as-address
15048 Enable the compiler to directly use a symbol name as an address in a
15049 load/store instruction, without first loading it into a
15050 register. Typically, the use of this option will generate larger
15051 programs, which run faster than when the option isn't used. However, the
15052 results vary from program to program, so it is left as a user option,
15053 rather than being permanently enabled.
15055 @item -mno-inefficient-warnings
15056 Disables warnings about the generation of inefficient code. These
15057 warnings can be generated, for example, when compiling code which
15058 performs byte-level memory operations on the MAC AE type. The MAC AE has
15059 no hardware support for byte-level memory operations, so all byte
15060 load/stores must be synthesized from word load/store operations. This is
15061 inefficient and a warning will be generated indicating to the programmer
15062 that they should rewrite the code to avoid byte operations, or to target
15063 an AE type which has the necessary hardware support. This option enables
15064 the warning to be turned off.
15068 @node PowerPC Options
15069 @subsection PowerPC Options
15070 @cindex PowerPC options
15072 These are listed under @xref{RS/6000 and PowerPC Options}.
15074 @node RS/6000 and PowerPC Options
15075 @subsection IBM RS/6000 and PowerPC Options
15076 @cindex RS/6000 and PowerPC Options
15077 @cindex IBM RS/6000 and PowerPC Options
15079 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15086 @itemx -mno-powerpc
15087 @itemx -mpowerpc-gpopt
15088 @itemx -mno-powerpc-gpopt
15089 @itemx -mpowerpc-gfxopt
15090 @itemx -mno-powerpc-gfxopt
15093 @itemx -mno-powerpc64
15097 @itemx -mno-popcntb
15099 @itemx -mno-popcntd
15108 @itemx -mno-hard-dfp
15112 @opindex mno-power2
15114 @opindex mno-powerpc
15115 @opindex mpowerpc-gpopt
15116 @opindex mno-powerpc-gpopt
15117 @opindex mpowerpc-gfxopt
15118 @opindex mno-powerpc-gfxopt
15119 @opindex mpowerpc64
15120 @opindex mno-powerpc64
15124 @opindex mno-popcntb
15126 @opindex mno-popcntd
15132 @opindex mno-mfpgpr
15134 @opindex mno-hard-dfp
15135 GCC supports two related instruction set architectures for the
15136 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15137 instructions supported by the @samp{rios} chip set used in the original
15138 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15139 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15140 the IBM 4xx, 6xx, and follow-on microprocessors.
15142 Neither architecture is a subset of the other. However there is a
15143 large common subset of instructions supported by both. An MQ
15144 register is included in processors supporting the POWER architecture.
15146 You use these options to specify which instructions are available on the
15147 processor you are using. The default value of these options is
15148 determined when configuring GCC@. Specifying the
15149 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15150 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15151 rather than the options listed above.
15153 The @option{-mpower} option allows GCC to generate instructions that
15154 are found only in the POWER architecture and to use the MQ register.
15155 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15156 to generate instructions that are present in the POWER2 architecture but
15157 not the original POWER architecture.
15159 The @option{-mpowerpc} option allows GCC to generate instructions that
15160 are found only in the 32-bit subset of the PowerPC architecture.
15161 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15162 GCC to use the optional PowerPC architecture instructions in the
15163 General Purpose group, including floating-point square root. Specifying
15164 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15165 use the optional PowerPC architecture instructions in the Graphics
15166 group, including floating-point select.
15168 The @option{-mmfcrf} option allows GCC to generate the move from
15169 condition register field instruction implemented on the POWER4
15170 processor and other processors that support the PowerPC V2.01
15172 The @option{-mpopcntb} option allows GCC to generate the popcount and
15173 double precision FP reciprocal estimate instruction implemented on the
15174 POWER5 processor and other processors that support the PowerPC V2.02
15176 The @option{-mpopcntd} option allows GCC to generate the popcount
15177 instruction implemented on the POWER7 processor and other processors
15178 that support the PowerPC V2.06 architecture.
15179 The @option{-mfprnd} option allows GCC to generate the FP round to
15180 integer instructions implemented on the POWER5+ processor and other
15181 processors that support the PowerPC V2.03 architecture.
15182 The @option{-mcmpb} option allows GCC to generate the compare bytes
15183 instruction implemented on the POWER6 processor and other processors
15184 that support the PowerPC V2.05 architecture.
15185 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15186 general purpose register instructions implemented on the POWER6X
15187 processor and other processors that support the extended PowerPC V2.05
15189 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15190 point instructions implemented on some POWER processors.
15192 The @option{-mpowerpc64} option allows GCC to generate the additional
15193 64-bit instructions that are found in the full PowerPC64 architecture
15194 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15195 @option{-mno-powerpc64}.
15197 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15198 will use only the instructions in the common subset of both
15199 architectures plus some special AIX common-mode calls, and will not use
15200 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15201 permits GCC to use any instruction from either architecture and to
15202 allow use of the MQ register; specify this for the Motorola MPC601.
15204 @item -mnew-mnemonics
15205 @itemx -mold-mnemonics
15206 @opindex mnew-mnemonics
15207 @opindex mold-mnemonics
15208 Select which mnemonics to use in the generated assembler code. With
15209 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15210 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15211 assembler mnemonics defined for the POWER architecture. Instructions
15212 defined in only one architecture have only one mnemonic; GCC uses that
15213 mnemonic irrespective of which of these options is specified.
15215 GCC defaults to the mnemonics appropriate for the architecture in
15216 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15217 value of these option. Unless you are building a cross-compiler, you
15218 should normally not specify either @option{-mnew-mnemonics} or
15219 @option{-mold-mnemonics}, but should instead accept the default.
15221 @item -mcpu=@var{cpu_type}
15223 Set architecture type, register usage, choice of mnemonics, and
15224 instruction scheduling parameters for machine type @var{cpu_type}.
15225 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15226 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15227 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15228 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15229 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15230 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15231 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15232 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15233 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15234 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15235 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15237 @option{-mcpu=common} selects a completely generic processor. Code
15238 generated under this option will run on any POWER or PowerPC processor.
15239 GCC will use only the instructions in the common subset of both
15240 architectures, and will not use the MQ register. GCC assumes a generic
15241 processor model for scheduling purposes.
15243 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15244 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15245 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15246 types, with an appropriate, generic processor model assumed for
15247 scheduling purposes.
15249 The other options specify a specific processor. Code generated under
15250 those options will run best on that processor, and may not run at all on
15253 The @option{-mcpu} options automatically enable or disable the
15256 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15257 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15258 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15259 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15261 The particular options set for any particular CPU will vary between
15262 compiler versions, depending on what setting seems to produce optimal
15263 code for that CPU; it doesn't necessarily reflect the actual hardware's
15264 capabilities. If you wish to set an individual option to a particular
15265 value, you may specify it after the @option{-mcpu} option, like
15266 @samp{-mcpu=970 -mno-altivec}.
15268 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15269 not enabled or disabled by the @option{-mcpu} option at present because
15270 AIX does not have full support for these options. You may still
15271 enable or disable them individually if you're sure it'll work in your
15274 @item -mtune=@var{cpu_type}
15276 Set the instruction scheduling parameters for machine type
15277 @var{cpu_type}, but do not set the architecture type, register usage, or
15278 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15279 values for @var{cpu_type} are used for @option{-mtune} as for
15280 @option{-mcpu}. If both are specified, the code generated will use the
15281 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15282 scheduling parameters set by @option{-mtune}.
15284 @item -mcmodel=small
15285 @opindex mcmodel=small
15286 Generate PowerPC64 code for the small model: The TOC is limited to
15289 @item -mcmodel=medium
15290 @opindex mcmodel=medium
15291 Generate PowerPC64 code for the medium model: The TOC and other static
15292 data may be up to a total of 4G in size.
15294 @item -mcmodel=large
15295 @opindex mcmodel=large
15296 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15297 in size. Other data and code is only limited by the 64-bit address
15301 @itemx -mno-altivec
15303 @opindex mno-altivec
15304 Generate code that uses (does not use) AltiVec instructions, and also
15305 enable the use of built-in functions that allow more direct access to
15306 the AltiVec instruction set. You may also need to set
15307 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15313 @opindex mno-vrsave
15314 Generate VRSAVE instructions when generating AltiVec code.
15316 @item -mgen-cell-microcode
15317 @opindex mgen-cell-microcode
15318 Generate Cell microcode instructions
15320 @item -mwarn-cell-microcode
15321 @opindex mwarn-cell-microcode
15322 Warning when a Cell microcode instruction is going to emitted. An example
15323 of a Cell microcode instruction is a variable shift.
15326 @opindex msecure-plt
15327 Generate code that allows ld and ld.so to build executables and shared
15328 libraries with non-exec .plt and .got sections. This is a PowerPC
15329 32-bit SYSV ABI option.
15333 Generate code that uses a BSS .plt section that ld.so fills in, and
15334 requires .plt and .got sections that are both writable and executable.
15335 This is a PowerPC 32-bit SYSV ABI option.
15341 This switch enables or disables the generation of ISEL instructions.
15343 @item -misel=@var{yes/no}
15344 This switch has been deprecated. Use @option{-misel} and
15345 @option{-mno-isel} instead.
15351 This switch enables or disables the generation of SPE simd
15357 @opindex mno-paired
15358 This switch enables or disables the generation of PAIRED simd
15361 @item -mspe=@var{yes/no}
15362 This option has been deprecated. Use @option{-mspe} and
15363 @option{-mno-spe} instead.
15369 Generate code that uses (does not use) vector/scalar (VSX)
15370 instructions, and also enable the use of built-in functions that allow
15371 more direct access to the VSX instruction set.
15373 @item -mfloat-gprs=@var{yes/single/double/no}
15374 @itemx -mfloat-gprs
15375 @opindex mfloat-gprs
15376 This switch enables or disables the generation of floating point
15377 operations on the general purpose registers for architectures that
15380 The argument @var{yes} or @var{single} enables the use of
15381 single-precision floating point operations.
15383 The argument @var{double} enables the use of single and
15384 double-precision floating point operations.
15386 The argument @var{no} disables floating point operations on the
15387 general purpose registers.
15389 This option is currently only available on the MPC854x.
15395 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15396 targets (including GNU/Linux). The 32-bit environment sets int, long
15397 and pointer to 32 bits and generates code that runs on any PowerPC
15398 variant. The 64-bit environment sets int to 32 bits and long and
15399 pointer to 64 bits, and generates code for PowerPC64, as for
15400 @option{-mpowerpc64}.
15403 @itemx -mno-fp-in-toc
15404 @itemx -mno-sum-in-toc
15405 @itemx -mminimal-toc
15407 @opindex mno-fp-in-toc
15408 @opindex mno-sum-in-toc
15409 @opindex mminimal-toc
15410 Modify generation of the TOC (Table Of Contents), which is created for
15411 every executable file. The @option{-mfull-toc} option is selected by
15412 default. In that case, GCC will allocate at least one TOC entry for
15413 each unique non-automatic variable reference in your program. GCC
15414 will also place floating-point constants in the TOC@. However, only
15415 16,384 entries are available in the TOC@.
15417 If you receive a linker error message that saying you have overflowed
15418 the available TOC space, you can reduce the amount of TOC space used
15419 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15420 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15421 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15422 generate code to calculate the sum of an address and a constant at
15423 run-time instead of putting that sum into the TOC@. You may specify one
15424 or both of these options. Each causes GCC to produce very slightly
15425 slower and larger code at the expense of conserving TOC space.
15427 If you still run out of space in the TOC even when you specify both of
15428 these options, specify @option{-mminimal-toc} instead. This option causes
15429 GCC to make only one TOC entry for every file. When you specify this
15430 option, GCC will produce code that is slower and larger but which
15431 uses extremely little TOC space. You may wish to use this option
15432 only on files that contain less frequently executed code.
15438 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15439 @code{long} type, and the infrastructure needed to support them.
15440 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15441 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15442 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15445 @itemx -mno-xl-compat
15446 @opindex mxl-compat
15447 @opindex mno-xl-compat
15448 Produce code that conforms more closely to IBM XL compiler semantics
15449 when using AIX-compatible ABI@. Pass floating-point arguments to
15450 prototyped functions beyond the register save area (RSA) on the stack
15451 in addition to argument FPRs. Do not assume that most significant
15452 double in 128-bit long double value is properly rounded when comparing
15453 values and converting to double. Use XL symbol names for long double
15456 The AIX calling convention was extended but not initially documented to
15457 handle an obscure K&R C case of calling a function that takes the
15458 address of its arguments with fewer arguments than declared. IBM XL
15459 compilers access floating point arguments which do not fit in the
15460 RSA from the stack when a subroutine is compiled without
15461 optimization. Because always storing floating-point arguments on the
15462 stack is inefficient and rarely needed, this option is not enabled by
15463 default and only is necessary when calling subroutines compiled by IBM
15464 XL compilers without optimization.
15468 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15469 application written to use message passing with special startup code to
15470 enable the application to run. The system must have PE installed in the
15471 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15472 must be overridden with the @option{-specs=} option to specify the
15473 appropriate directory location. The Parallel Environment does not
15474 support threads, so the @option{-mpe} option and the @option{-pthread}
15475 option are incompatible.
15477 @item -malign-natural
15478 @itemx -malign-power
15479 @opindex malign-natural
15480 @opindex malign-power
15481 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15482 @option{-malign-natural} overrides the ABI-defined alignment of larger
15483 types, such as floating-point doubles, on their natural size-based boundary.
15484 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15485 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15487 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15491 @itemx -mhard-float
15492 @opindex msoft-float
15493 @opindex mhard-float
15494 Generate code that does not use (uses) the floating-point register set.
15495 Software floating point emulation is provided if you use the
15496 @option{-msoft-float} option, and pass the option to GCC when linking.
15498 @item -msingle-float
15499 @itemx -mdouble-float
15500 @opindex msingle-float
15501 @opindex mdouble-float
15502 Generate code for single or double-precision floating point operations.
15503 @option{-mdouble-float} implies @option{-msingle-float}.
15506 @opindex msimple-fpu
15507 Do not generate sqrt and div instructions for hardware floating point unit.
15511 Specify type of floating point unit. Valid values are @var{sp_lite}
15512 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15513 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15514 and @var{dp_full} (equivalent to -mdouble-float).
15517 @opindex mxilinx-fpu
15518 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15521 @itemx -mno-multiple
15523 @opindex mno-multiple
15524 Generate code that uses (does not use) the load multiple word
15525 instructions and the store multiple word instructions. These
15526 instructions are generated by default on POWER systems, and not
15527 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15528 endian PowerPC systems, since those instructions do not work when the
15529 processor is in little endian mode. The exceptions are PPC740 and
15530 PPC750 which permit the instructions usage in little endian mode.
15535 @opindex mno-string
15536 Generate code that uses (does not use) the load string instructions
15537 and the store string word instructions to save multiple registers and
15538 do small block moves. These instructions are generated by default on
15539 POWER systems, and not generated on PowerPC systems. Do not use
15540 @option{-mstring} on little endian PowerPC systems, since those
15541 instructions do not work when the processor is in little endian mode.
15542 The exceptions are PPC740 and PPC750 which permit the instructions
15543 usage in little endian mode.
15548 @opindex mno-update
15549 Generate code that uses (does not use) the load or store instructions
15550 that update the base register to the address of the calculated memory
15551 location. These instructions are generated by default. If you use
15552 @option{-mno-update}, there is a small window between the time that the
15553 stack pointer is updated and the address of the previous frame is
15554 stored, which means code that walks the stack frame across interrupts or
15555 signals may get corrupted data.
15557 @item -mavoid-indexed-addresses
15558 @itemx -mno-avoid-indexed-addresses
15559 @opindex mavoid-indexed-addresses
15560 @opindex mno-avoid-indexed-addresses
15561 Generate code that tries to avoid (not avoid) the use of indexed load
15562 or store instructions. These instructions can incur a performance
15563 penalty on Power6 processors in certain situations, such as when
15564 stepping through large arrays that cross a 16M boundary. This option
15565 is enabled by default when targetting Power6 and disabled otherwise.
15568 @itemx -mno-fused-madd
15569 @opindex mfused-madd
15570 @opindex mno-fused-madd
15571 Generate code that uses (does not use) the floating point multiply and
15572 accumulate instructions. These instructions are generated by default if
15573 hardware floating is used.
15579 Generate code that uses (does not use) the half-word multiply and
15580 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15581 These instructions are generated by default when targetting those
15588 Generate code that uses (does not use) the string-search @samp{dlmzb}
15589 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15590 generated by default when targetting those processors.
15592 @item -mno-bit-align
15594 @opindex mno-bit-align
15595 @opindex mbit-align
15596 On System V.4 and embedded PowerPC systems do not (do) force structures
15597 and unions that contain bit-fields to be aligned to the base type of the
15600 For example, by default a structure containing nothing but 8
15601 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15602 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15603 the structure would be aligned to a 1 byte boundary and be one byte in
15606 @item -mno-strict-align
15607 @itemx -mstrict-align
15608 @opindex mno-strict-align
15609 @opindex mstrict-align
15610 On System V.4 and embedded PowerPC systems do not (do) assume that
15611 unaligned memory references will be handled by the system.
15613 @item -mrelocatable
15614 @itemx -mno-relocatable
15615 @opindex mrelocatable
15616 @opindex mno-relocatable
15617 Generate code that allows (does not allow) a static executable to be
15618 relocated to a different address at runtime. A simple embedded
15619 PowerPC system loader should relocate the entire contents of
15620 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15621 a table of 32-bit addresses generated by this option. For this to
15622 work, all objects linked together must be compiled with
15623 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15624 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15626 @item -mrelocatable-lib
15627 @itemx -mno-relocatable-lib
15628 @opindex mrelocatable-lib
15629 @opindex mno-relocatable-lib
15630 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15631 @code{.fixup} section to allow static executables to be relocated at
15632 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15633 alignment of @option{-mrelocatable}. Objects compiled with
15634 @option{-mrelocatable-lib} may be linked with objects compiled with
15635 any combination of the @option{-mrelocatable} options.
15641 On System V.4 and embedded PowerPC systems do not (do) assume that
15642 register 2 contains a pointer to a global area pointing to the addresses
15643 used in the program.
15646 @itemx -mlittle-endian
15648 @opindex mlittle-endian
15649 On System V.4 and embedded PowerPC systems compile code for the
15650 processor in little endian mode. The @option{-mlittle-endian} option is
15651 the same as @option{-mlittle}.
15654 @itemx -mbig-endian
15656 @opindex mbig-endian
15657 On System V.4 and embedded PowerPC systems compile code for the
15658 processor in big endian mode. The @option{-mbig-endian} option is
15659 the same as @option{-mbig}.
15661 @item -mdynamic-no-pic
15662 @opindex mdynamic-no-pic
15663 On Darwin and Mac OS X systems, compile code so that it is not
15664 relocatable, but that its external references are relocatable. The
15665 resulting code is suitable for applications, but not shared
15668 @item -mprioritize-restricted-insns=@var{priority}
15669 @opindex mprioritize-restricted-insns
15670 This option controls the priority that is assigned to
15671 dispatch-slot restricted instructions during the second scheduling
15672 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15673 @var{no/highest/second-highest} priority to dispatch slot restricted
15676 @item -msched-costly-dep=@var{dependence_type}
15677 @opindex msched-costly-dep
15678 This option controls which dependences are considered costly
15679 by the target during instruction scheduling. The argument
15680 @var{dependence_type} takes one of the following values:
15681 @var{no}: no dependence is costly,
15682 @var{all}: all dependences are costly,
15683 @var{true_store_to_load}: a true dependence from store to load is costly,
15684 @var{store_to_load}: any dependence from store to load is costly,
15685 @var{number}: any dependence which latency >= @var{number} is costly.
15687 @item -minsert-sched-nops=@var{scheme}
15688 @opindex minsert-sched-nops
15689 This option controls which nop insertion scheme will be used during
15690 the second scheduling pass. The argument @var{scheme} takes one of the
15692 @var{no}: Don't insert nops.
15693 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15694 according to the scheduler's grouping.
15695 @var{regroup_exact}: Insert nops to force costly dependent insns into
15696 separate groups. Insert exactly as many nops as needed to force an insn
15697 to a new group, according to the estimated processor grouping.
15698 @var{number}: Insert nops to force costly dependent insns into
15699 separate groups. Insert @var{number} nops to force an insn to a new group.
15702 @opindex mcall-sysv
15703 On System V.4 and embedded PowerPC systems compile code using calling
15704 conventions that adheres to the March 1995 draft of the System V
15705 Application Binary Interface, PowerPC processor supplement. This is the
15706 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15708 @item -mcall-sysv-eabi
15710 @opindex mcall-sysv-eabi
15711 @opindex mcall-eabi
15712 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15714 @item -mcall-sysv-noeabi
15715 @opindex mcall-sysv-noeabi
15716 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15718 @item -mcall-aixdesc
15720 On System V.4 and embedded PowerPC systems compile code for the AIX
15724 @opindex mcall-linux
15725 On System V.4 and embedded PowerPC systems compile code for the
15726 Linux-based GNU system.
15730 On System V.4 and embedded PowerPC systems compile code for the
15731 Hurd-based GNU system.
15733 @item -mcall-freebsd
15734 @opindex mcall-freebsd
15735 On System V.4 and embedded PowerPC systems compile code for the
15736 FreeBSD operating system.
15738 @item -mcall-netbsd
15739 @opindex mcall-netbsd
15740 On System V.4 and embedded PowerPC systems compile code for the
15741 NetBSD operating system.
15743 @item -mcall-openbsd
15744 @opindex mcall-netbsd
15745 On System V.4 and embedded PowerPC systems compile code for the
15746 OpenBSD operating system.
15748 @item -maix-struct-return
15749 @opindex maix-struct-return
15750 Return all structures in memory (as specified by the AIX ABI)@.
15752 @item -msvr4-struct-return
15753 @opindex msvr4-struct-return
15754 Return structures smaller than 8 bytes in registers (as specified by the
15757 @item -mabi=@var{abi-type}
15759 Extend the current ABI with a particular extension, or remove such extension.
15760 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15761 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15765 Extend the current ABI with SPE ABI extensions. This does not change
15766 the default ABI, instead it adds the SPE ABI extensions to the current
15770 @opindex mabi=no-spe
15771 Disable Booke SPE ABI extensions for the current ABI@.
15773 @item -mabi=ibmlongdouble
15774 @opindex mabi=ibmlongdouble
15775 Change the current ABI to use IBM extended precision long double.
15776 This is a PowerPC 32-bit SYSV ABI option.
15778 @item -mabi=ieeelongdouble
15779 @opindex mabi=ieeelongdouble
15780 Change the current ABI to use IEEE extended precision long double.
15781 This is a PowerPC 32-bit Linux ABI option.
15784 @itemx -mno-prototype
15785 @opindex mprototype
15786 @opindex mno-prototype
15787 On System V.4 and embedded PowerPC systems assume that all calls to
15788 variable argument functions are properly prototyped. Otherwise, the
15789 compiler must insert an instruction before every non prototyped call to
15790 set or clear bit 6 of the condition code register (@var{CR}) to
15791 indicate whether floating point values were passed in the floating point
15792 registers in case the function takes a variable arguments. With
15793 @option{-mprototype}, only calls to prototyped variable argument functions
15794 will set or clear the bit.
15798 On embedded PowerPC systems, assume that the startup module is called
15799 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15800 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15805 On embedded PowerPC systems, assume that the startup module is called
15806 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15811 On embedded PowerPC systems, assume that the startup module is called
15812 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15815 @item -myellowknife
15816 @opindex myellowknife
15817 On embedded PowerPC systems, assume that the startup module is called
15818 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15823 On System V.4 and embedded PowerPC systems, specify that you are
15824 compiling for a VxWorks system.
15828 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15829 header to indicate that @samp{eabi} extended relocations are used.
15835 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15836 Embedded Applications Binary Interface (eabi) which is a set of
15837 modifications to the System V.4 specifications. Selecting @option{-meabi}
15838 means that the stack is aligned to an 8 byte boundary, a function
15839 @code{__eabi} is called to from @code{main} to set up the eabi
15840 environment, and the @option{-msdata} option can use both @code{r2} and
15841 @code{r13} to point to two separate small data areas. Selecting
15842 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15843 do not call an initialization function from @code{main}, and the
15844 @option{-msdata} option will only use @code{r13} to point to a single
15845 small data area. The @option{-meabi} option is on by default if you
15846 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15849 @opindex msdata=eabi
15850 On System V.4 and embedded PowerPC systems, put small initialized
15851 @code{const} global and static data in the @samp{.sdata2} section, which
15852 is pointed to by register @code{r2}. Put small initialized
15853 non-@code{const} global and static data in the @samp{.sdata} section,
15854 which is pointed to by register @code{r13}. Put small uninitialized
15855 global and static data in the @samp{.sbss} section, which is adjacent to
15856 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15857 incompatible with the @option{-mrelocatable} option. The
15858 @option{-msdata=eabi} option also sets the @option{-memb} option.
15861 @opindex msdata=sysv
15862 On System V.4 and embedded PowerPC systems, put small global and static
15863 data in the @samp{.sdata} section, which is pointed to by register
15864 @code{r13}. Put small uninitialized global and static data in the
15865 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15866 The @option{-msdata=sysv} option is incompatible with the
15867 @option{-mrelocatable} option.
15869 @item -msdata=default
15871 @opindex msdata=default
15873 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15874 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15875 same as @option{-msdata=sysv}.
15878 @opindex msdata=data
15879 On System V.4 and embedded PowerPC systems, put small global
15880 data in the @samp{.sdata} section. Put small uninitialized global
15881 data in the @samp{.sbss} section. Do not use register @code{r13}
15882 to address small data however. This is the default behavior unless
15883 other @option{-msdata} options are used.
15887 @opindex msdata=none
15889 On embedded PowerPC systems, put all initialized global and static data
15890 in the @samp{.data} section, and all uninitialized data in the
15891 @samp{.bss} section.
15893 @item -mblock-move-inline-limit=@var{num}
15894 @opindex mblock-move-inline-limit
15895 Inline all block moves (such as calls to @code{memcpy} or structure
15896 copies) less than or equal to @var{num} bytes. The minimum value for
15897 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15898 targets. The default value is target-specific.
15902 @cindex smaller data references (PowerPC)
15903 @cindex .sdata/.sdata2 references (PowerPC)
15904 On embedded PowerPC systems, put global and static items less than or
15905 equal to @var{num} bytes into the small data or bss sections instead of
15906 the normal data or bss section. By default, @var{num} is 8. The
15907 @option{-G @var{num}} switch is also passed to the linker.
15908 All modules should be compiled with the same @option{-G @var{num}} value.
15911 @itemx -mno-regnames
15913 @opindex mno-regnames
15914 On System V.4 and embedded PowerPC systems do (do not) emit register
15915 names in the assembly language output using symbolic forms.
15918 @itemx -mno-longcall
15920 @opindex mno-longcall
15921 By default assume that all calls are far away so that a longer more
15922 expensive calling sequence is required. This is required for calls
15923 further than 32 megabytes (33,554,432 bytes) from the current location.
15924 A short call will be generated if the compiler knows
15925 the call cannot be that far away. This setting can be overridden by
15926 the @code{shortcall} function attribute, or by @code{#pragma
15929 Some linkers are capable of detecting out-of-range calls and generating
15930 glue code on the fly. On these systems, long calls are unnecessary and
15931 generate slower code. As of this writing, the AIX linker can do this,
15932 as can the GNU linker for PowerPC/64. It is planned to add this feature
15933 to the GNU linker for 32-bit PowerPC systems as well.
15935 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15936 callee, L42'', plus a ``branch island'' (glue code). The two target
15937 addresses represent the callee and the ``branch island''. The
15938 Darwin/PPC linker will prefer the first address and generate a ``bl
15939 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15940 otherwise, the linker will generate ``bl L42'' to call the ``branch
15941 island''. The ``branch island'' is appended to the body of the
15942 calling function; it computes the full 32-bit address of the callee
15945 On Mach-O (Darwin) systems, this option directs the compiler emit to
15946 the glue for every direct call, and the Darwin linker decides whether
15947 to use or discard it.
15949 In the future, we may cause GCC to ignore all longcall specifications
15950 when the linker is known to generate glue.
15952 @item -mtls-markers
15953 @itemx -mno-tls-markers
15954 @opindex mtls-markers
15955 @opindex mno-tls-markers
15956 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15957 specifying the function argument. The relocation allows ld to
15958 reliably associate function call with argument setup instructions for
15959 TLS optimization, which in turn allows gcc to better schedule the
15964 Adds support for multithreading with the @dfn{pthreads} library.
15965 This option sets flags for both the preprocessor and linker.
15970 This option will enable GCC to use the reciprocal estimate and
15971 reciprocal square root estimate instructions with additional
15972 Newton-Raphson steps to increase precision instead of doing a divide or
15973 square root and divide for floating point arguments. You should use
15974 the @option{-ffast-math} option when using @option{-mrecip} (or at
15975 least @option{-funsafe-math-optimizations},
15976 @option{-finite-math-only}, @option{-freciprocal-math} and
15977 @option{-fno-trapping-math}). Note that while the throughput of the
15978 sequence is generally higher than the throughput of the non-reciprocal
15979 instruction, the precision of the sequence can be decreased by up to 2
15980 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
15983 @item -mrecip=@var{opt}
15984 @opindex mrecip=opt
15985 This option allows to control which reciprocal estimate instructions
15986 may be used. @var{opt} is a comma separated list of options, that may
15987 be preceeded by a @code{!} to invert the option:
15988 @code{all}: enable all estimate instructions,
15989 @code{default}: enable the default instructions, equvalent to @option{-mrecip},
15990 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
15991 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
15992 @code{divf}: enable the single precision reciprocal approximation instructions;
15993 @code{divd}: enable the double precision reciprocal approximation instructions;
15994 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
15995 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
15996 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
15998 So for example, @option{-mrecip=all,!rsqrtd} would enable the
15999 all of the reciprocal estimate instructions, except for the
16000 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16001 which handle the double precision reciprocal square root calculations.
16003 @item -mrecip-precision
16004 @itemx -mno-recip-precision
16005 @opindex mrecip-precision
16006 Assume (do not assume) that the reciprocal estimate instructions
16007 provide higher precision estimates than is mandated by the powerpc
16008 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16009 automatically selects @option{-mrecip-precision}. The double
16010 precision square root estimate instructions are not generated by
16011 default on low precision machines, since they do not provide an
16012 estimate that converges after three steps.
16014 @item -mveclibabi=@var{type}
16015 @opindex mveclibabi
16016 Specifies the ABI type to use for vectorizing intrinsics using an
16017 external library. The only type supported at present is @code{mass},
16018 which specifies to use IBM's Mathematical Acceleration Subsystem
16019 (MASS) libraries for vectorizing intrinsics using external libraries.
16020 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16021 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16022 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16023 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16024 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16025 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16026 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16027 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16028 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16029 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16030 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16031 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16032 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16033 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16034 for power7. Both @option{-ftree-vectorize} and
16035 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16036 libraries will have to be specified at link time.
16041 Generate (do not generate) the @code{friz} instruction when the
16042 @option{-funsafe-math-optimizations} option is used to optimize
16043 rounding a floating point value to 64-bit integer and back to floating
16044 point. The @code{friz} instruction does not return the same value if
16045 the floating point number is too large to fit in an integer.
16049 @subsection RX Options
16052 These command line options are defined for RX targets:
16055 @item -m64bit-doubles
16056 @itemx -m32bit-doubles
16057 @opindex m64bit-doubles
16058 @opindex m32bit-doubles
16059 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16060 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16061 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16062 works on 32-bit values, which is why the default is
16063 @option{-m32bit-doubles}.
16069 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16070 floating point hardware. The default is enabled for the @var{RX600}
16071 series and disabled for the @var{RX200} series.
16073 Floating point instructions will only be generated for 32-bit floating
16074 point values however, so if the @option{-m64bit-doubles} option is in
16075 use then the FPU hardware will not be used for doubles.
16077 @emph{Note} If the @option{-fpu} option is enabled then
16078 @option{-funsafe-math-optimizations} is also enabled automatically.
16079 This is because the RX FPU instructions are themselves unsafe.
16081 @item -mcpu=@var{name}
16083 Selects the type of RX CPU to be targeted. Currently three types are
16084 supported, the generic @var{RX600} and @var{RX200} series hardware and
16085 the specific @var{RX610} cpu. The default is @var{RX600}.
16087 The only difference between @var{RX600} and @var{RX610} is that the
16088 @var{RX610} does not support the @code{MVTIPL} instruction.
16090 The @var{RX200} series does not have a hardware floating point unit
16091 and so @option{-nofpu} is enabled by default when this type is
16094 @item -mbig-endian-data
16095 @itemx -mlittle-endian-data
16096 @opindex mbig-endian-data
16097 @opindex mlittle-endian-data
16098 Store data (but not code) in the big-endian format. The default is
16099 @option{-mlittle-endian-data}, ie to store data in the little endian
16102 @item -msmall-data-limit=@var{N}
16103 @opindex msmall-data-limit
16104 Specifies the maximum size in bytes of global and static variables
16105 which can be placed into the small data area. Using the small data
16106 area can lead to smaller and faster code, but the size of area is
16107 limited and it is up to the programmer to ensure that the area does
16108 not overflow. Also when the small data area is used one of the RX's
16109 registers (@code{r13}) is reserved for use pointing to this area, so
16110 it is no longer available for use by the compiler. This could result
16111 in slower and/or larger code if variables which once could have been
16112 held in @code{r13} are now pushed onto the stack.
16114 Note, common variables (variables which have not been initialised) and
16115 constants are not placed into the small data area as they are assigned
16116 to other sections in the output executable.
16118 The default value is zero, which disables this feature. Note, this
16119 feature is not enabled by default with higher optimization levels
16120 (@option{-O2} etc) because of the potentially detrimental effects of
16121 reserving register @code{r13}. It is up to the programmer to
16122 experiment and discover whether this feature is of benefit to their
16129 Use the simulator runtime. The default is to use the libgloss board
16132 @item -mas100-syntax
16133 @itemx -mno-as100-syntax
16134 @opindex mas100-syntax
16135 @opindex mno-as100-syntax
16136 When generating assembler output use a syntax that is compatible with
16137 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16138 assembler but it has some restrictions so generating it is not the
16141 @item -mmax-constant-size=@var{N}
16142 @opindex mmax-constant-size
16143 Specifies the maximum size, in bytes, of a constant that can be used as
16144 an operand in a RX instruction. Although the RX instruction set does
16145 allow constants of up to 4 bytes in length to be used in instructions,
16146 a longer value equates to a longer instruction. Thus in some
16147 circumstances it can be beneficial to restrict the size of constants
16148 that are used in instructions. Constants that are too big are instead
16149 placed into a constant pool and referenced via register indirection.
16151 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16152 or 4 means that constants of any size are allowed.
16156 Enable linker relaxation. Linker relaxation is a process whereby the
16157 linker will attempt to reduce the size of a program by finding shorter
16158 versions of various instructions. Disabled by default.
16160 @item -mint-register=@var{N}
16161 @opindex mint-register
16162 Specify the number of registers to reserve for fast interrupt handler
16163 functions. The value @var{N} can be between 0 and 4. A value of 1
16164 means that register @code{r13} will be reserved for the exclusive use
16165 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16166 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16167 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16168 A value of 0, the default, does not reserve any registers.
16170 @item -msave-acc-in-interrupts
16171 @opindex msave-acc-in-interrupts
16172 Specifies that interrupt handler functions should preserve the
16173 accumulator register. This is only necessary if normal code might use
16174 the accumulator register, for example because it performs 64-bit
16175 multiplications. The default is to ignore the accumulator as this
16176 makes the interrupt handlers faster.
16180 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16181 has special significance to the RX port when used with the
16182 @code{interrupt} function attribute. This attribute indicates a
16183 function intended to process fast interrupts. GCC will will ensure
16184 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16185 and/or @code{r13} and only provided that the normal use of the
16186 corresponding registers have been restricted via the
16187 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16190 @node S/390 and zSeries Options
16191 @subsection S/390 and zSeries Options
16192 @cindex S/390 and zSeries Options
16194 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16198 @itemx -msoft-float
16199 @opindex mhard-float
16200 @opindex msoft-float
16201 Use (do not use) the hardware floating-point instructions and registers
16202 for floating-point operations. When @option{-msoft-float} is specified,
16203 functions in @file{libgcc.a} will be used to perform floating-point
16204 operations. When @option{-mhard-float} is specified, the compiler
16205 generates IEEE floating-point instructions. This is the default.
16208 @itemx -mno-hard-dfp
16210 @opindex mno-hard-dfp
16211 Use (do not use) the hardware decimal-floating-point instructions for
16212 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16213 specified, functions in @file{libgcc.a} will be used to perform
16214 decimal-floating-point operations. When @option{-mhard-dfp} is
16215 specified, the compiler generates decimal-floating-point hardware
16216 instructions. This is the default for @option{-march=z9-ec} or higher.
16218 @item -mlong-double-64
16219 @itemx -mlong-double-128
16220 @opindex mlong-double-64
16221 @opindex mlong-double-128
16222 These switches control the size of @code{long double} type. A size
16223 of 64bit makes the @code{long double} type equivalent to the @code{double}
16224 type. This is the default.
16227 @itemx -mno-backchain
16228 @opindex mbackchain
16229 @opindex mno-backchain
16230 Store (do not store) the address of the caller's frame as backchain pointer
16231 into the callee's stack frame.
16232 A backchain may be needed to allow debugging using tools that do not understand
16233 DWARF-2 call frame information.
16234 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16235 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16236 the backchain is placed into the topmost word of the 96/160 byte register
16239 In general, code compiled with @option{-mbackchain} is call-compatible with
16240 code compiled with @option{-mmo-backchain}; however, use of the backchain
16241 for debugging purposes usually requires that the whole binary is built with
16242 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16243 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16244 to build a linux kernel use @option{-msoft-float}.
16246 The default is to not maintain the backchain.
16248 @item -mpacked-stack
16249 @itemx -mno-packed-stack
16250 @opindex mpacked-stack
16251 @opindex mno-packed-stack
16252 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16253 specified, the compiler uses the all fields of the 96/160 byte register save
16254 area only for their default purpose; unused fields still take up stack space.
16255 When @option{-mpacked-stack} is specified, register save slots are densely
16256 packed at the top of the register save area; unused space is reused for other
16257 purposes, allowing for more efficient use of the available stack space.
16258 However, when @option{-mbackchain} is also in effect, the topmost word of
16259 the save area is always used to store the backchain, and the return address
16260 register is always saved two words below the backchain.
16262 As long as the stack frame backchain is not used, code generated with
16263 @option{-mpacked-stack} is call-compatible with code generated with
16264 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16265 S/390 or zSeries generated code that uses the stack frame backchain at run
16266 time, not just for debugging purposes. Such code is not call-compatible
16267 with code compiled with @option{-mpacked-stack}. Also, note that the
16268 combination of @option{-mbackchain},
16269 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16270 to build a linux kernel use @option{-msoft-float}.
16272 The default is to not use the packed stack layout.
16275 @itemx -mno-small-exec
16276 @opindex msmall-exec
16277 @opindex mno-small-exec
16278 Generate (or do not generate) code using the @code{bras} instruction
16279 to do subroutine calls.
16280 This only works reliably if the total executable size does not
16281 exceed 64k. The default is to use the @code{basr} instruction instead,
16282 which does not have this limitation.
16288 When @option{-m31} is specified, generate code compliant to the
16289 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16290 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16291 particular to generate 64-bit instructions. For the @samp{s390}
16292 targets, the default is @option{-m31}, while the @samp{s390x}
16293 targets default to @option{-m64}.
16299 When @option{-mzarch} is specified, generate code using the
16300 instructions available on z/Architecture.
16301 When @option{-mesa} is specified, generate code using the
16302 instructions available on ESA/390. Note that @option{-mesa} is
16303 not possible with @option{-m64}.
16304 When generating code compliant to the GNU/Linux for S/390 ABI,
16305 the default is @option{-mesa}. When generating code compliant
16306 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16312 Generate (or do not generate) code using the @code{mvcle} instruction
16313 to perform block moves. When @option{-mno-mvcle} is specified,
16314 use a @code{mvc} loop instead. This is the default unless optimizing for
16321 Print (or do not print) additional debug information when compiling.
16322 The default is to not print debug information.
16324 @item -march=@var{cpu-type}
16326 Generate code that will run on @var{cpu-type}, which is the name of a system
16327 representing a certain processor type. Possible values for
16328 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16329 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16330 When generating code using the instructions available on z/Architecture,
16331 the default is @option{-march=z900}. Otherwise, the default is
16332 @option{-march=g5}.
16334 @item -mtune=@var{cpu-type}
16336 Tune to @var{cpu-type} everything applicable about the generated code,
16337 except for the ABI and the set of available instructions.
16338 The list of @var{cpu-type} values is the same as for @option{-march}.
16339 The default is the value used for @option{-march}.
16342 @itemx -mno-tpf-trace
16343 @opindex mtpf-trace
16344 @opindex mno-tpf-trace
16345 Generate code that adds (does not add) in TPF OS specific branches to trace
16346 routines in the operating system. This option is off by default, even
16347 when compiling for the TPF OS@.
16350 @itemx -mno-fused-madd
16351 @opindex mfused-madd
16352 @opindex mno-fused-madd
16353 Generate code that uses (does not use) the floating point multiply and
16354 accumulate instructions. These instructions are generated by default if
16355 hardware floating point is used.
16357 @item -mwarn-framesize=@var{framesize}
16358 @opindex mwarn-framesize
16359 Emit a warning if the current function exceeds the given frame size. Because
16360 this is a compile time check it doesn't need to be a real problem when the program
16361 runs. It is intended to identify functions which most probably cause
16362 a stack overflow. It is useful to be used in an environment with limited stack
16363 size e.g.@: the linux kernel.
16365 @item -mwarn-dynamicstack
16366 @opindex mwarn-dynamicstack
16367 Emit a warning if the function calls alloca or uses dynamically
16368 sized arrays. This is generally a bad idea with a limited stack size.
16370 @item -mstack-guard=@var{stack-guard}
16371 @itemx -mstack-size=@var{stack-size}
16372 @opindex mstack-guard
16373 @opindex mstack-size
16374 If these options are provided the s390 back end emits additional instructions in
16375 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16376 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16377 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16378 the frame size of the compiled function is chosen.
16379 These options are intended to be used to help debugging stack overflow problems.
16380 The additionally emitted code causes only little overhead and hence can also be
16381 used in production like systems without greater performance degradation. The given
16382 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16383 @var{stack-guard} without exceeding 64k.
16384 In order to be efficient the extra code makes the assumption that the stack starts
16385 at an address aligned to the value given by @var{stack-size}.
16386 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16389 @node Score Options
16390 @subsection Score Options
16391 @cindex Score Options
16393 These options are defined for Score implementations:
16398 Compile code for big endian mode. This is the default.
16402 Compile code for little endian mode.
16406 Disable generate bcnz instruction.
16410 Enable generate unaligned load and store instruction.
16414 Enable the use of multiply-accumulate instructions. Disabled by default.
16418 Specify the SCORE5 as the target architecture.
16422 Specify the SCORE5U of the target architecture.
16426 Specify the SCORE7 as the target architecture. This is the default.
16430 Specify the SCORE7D as the target architecture.
16434 @subsection SH Options
16436 These @samp{-m} options are defined for the SH implementations:
16441 Generate code for the SH1.
16445 Generate code for the SH2.
16448 Generate code for the SH2e.
16452 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16453 that the floating-point unit is not used.
16455 @item -m2a-single-only
16456 @opindex m2a-single-only
16457 Generate code for the SH2a-FPU, in such a way that no double-precision
16458 floating point operations are used.
16461 @opindex m2a-single
16462 Generate code for the SH2a-FPU assuming the floating-point unit is in
16463 single-precision mode by default.
16467 Generate code for the SH2a-FPU assuming the floating-point unit is in
16468 double-precision mode by default.
16472 Generate code for the SH3.
16476 Generate code for the SH3e.
16480 Generate code for the SH4 without a floating-point unit.
16482 @item -m4-single-only
16483 @opindex m4-single-only
16484 Generate code for the SH4 with a floating-point unit that only
16485 supports single-precision arithmetic.
16489 Generate code for the SH4 assuming the floating-point unit is in
16490 single-precision mode by default.
16494 Generate code for the SH4.
16498 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16499 floating-point unit is not used.
16501 @item -m4a-single-only
16502 @opindex m4a-single-only
16503 Generate code for the SH4a, in such a way that no double-precision
16504 floating point operations are used.
16507 @opindex m4a-single
16508 Generate code for the SH4a assuming the floating-point unit is in
16509 single-precision mode by default.
16513 Generate code for the SH4a.
16517 Same as @option{-m4a-nofpu}, except that it implicitly passes
16518 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16519 instructions at the moment.
16523 Compile code for the processor in big endian mode.
16527 Compile code for the processor in little endian mode.
16531 Align doubles at 64-bit boundaries. Note that this changes the calling
16532 conventions, and thus some functions from the standard C library will
16533 not work unless you recompile it first with @option{-mdalign}.
16537 Shorten some address references at link time, when possible; uses the
16538 linker option @option{-relax}.
16542 Use 32-bit offsets in @code{switch} tables. The default is to use
16547 Enable the use of bit manipulation instructions on SH2A.
16551 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16552 alignment constraints.
16556 Comply with the calling conventions defined by Renesas.
16560 Comply with the calling conventions defined by Renesas.
16564 Comply with the calling conventions defined for GCC before the Renesas
16565 conventions were available. This option is the default for all
16566 targets of the SH toolchain except for @samp{sh-symbianelf}.
16569 @opindex mnomacsave
16570 Mark the @code{MAC} register as call-clobbered, even if
16571 @option{-mhitachi} is given.
16575 Increase IEEE-compliance of floating-point code.
16576 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16577 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16578 comparisons of NANs / infinities incurs extra overhead in every
16579 floating point comparison, therefore the default is set to
16580 @option{-ffinite-math-only}.
16582 @item -minline-ic_invalidate
16583 @opindex minline-ic_invalidate
16584 Inline code to invalidate instruction cache entries after setting up
16585 nested function trampolines.
16586 This option has no effect if -musermode is in effect and the selected
16587 code generation option (e.g. -m4) does not allow the use of the icbi
16589 If the selected code generation option does not allow the use of the icbi
16590 instruction, and -musermode is not in effect, the inlined code will
16591 manipulate the instruction cache address array directly with an associative
16592 write. This not only requires privileged mode, but it will also
16593 fail if the cache line had been mapped via the TLB and has become unmapped.
16597 Dump instruction size and location in the assembly code.
16600 @opindex mpadstruct
16601 This option is deprecated. It pads structures to multiple of 4 bytes,
16602 which is incompatible with the SH ABI@.
16606 Optimize for space instead of speed. Implied by @option{-Os}.
16609 @opindex mprefergot
16610 When generating position-independent code, emit function calls using
16611 the Global Offset Table instead of the Procedure Linkage Table.
16615 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16616 if the inlined code would not work in user mode.
16617 This is the default when the target is @code{sh-*-linux*}.
16619 @item -multcost=@var{number}
16620 @opindex multcost=@var{number}
16621 Set the cost to assume for a multiply insn.
16623 @item -mdiv=@var{strategy}
16624 @opindex mdiv=@var{strategy}
16625 Set the division strategy to use for SHmedia code. @var{strategy} must be
16626 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16627 inv:call2, inv:fp .
16628 "fp" performs the operation in floating point. This has a very high latency,
16629 but needs only a few instructions, so it might be a good choice if
16630 your code has enough easily exploitable ILP to allow the compiler to
16631 schedule the floating point instructions together with other instructions.
16632 Division by zero causes a floating point exception.
16633 "inv" uses integer operations to calculate the inverse of the divisor,
16634 and then multiplies the dividend with the inverse. This strategy allows
16635 cse and hoisting of the inverse calculation. Division by zero calculates
16636 an unspecified result, but does not trap.
16637 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16638 have been found, or if the entire operation has been hoisted to the same
16639 place, the last stages of the inverse calculation are intertwined with the
16640 final multiply to reduce the overall latency, at the expense of using a few
16641 more instructions, and thus offering fewer scheduling opportunities with
16643 "call" calls a library function that usually implements the inv:minlat
16645 This gives high code density for m5-*media-nofpu compilations.
16646 "call2" uses a different entry point of the same library function, where it
16647 assumes that a pointer to a lookup table has already been set up, which
16648 exposes the pointer load to cse / code hoisting optimizations.
16649 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16650 code generation, but if the code stays unoptimized, revert to the "call",
16651 "call2", or "fp" strategies, respectively. Note that the
16652 potentially-trapping side effect of division by zero is carried by a
16653 separate instruction, so it is possible that all the integer instructions
16654 are hoisted out, but the marker for the side effect stays where it is.
16655 A recombination to fp operations or a call is not possible in that case.
16656 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16657 that the inverse calculation was nor separated from the multiply, they speed
16658 up division where the dividend fits into 20 bits (plus sign where applicable),
16659 by inserting a test to skip a number of operations in this case; this test
16660 slows down the case of larger dividends. inv20u assumes the case of a such
16661 a small dividend to be unlikely, and inv20l assumes it to be likely.
16663 @item -maccumulate-outgoing-args
16664 @opindex maccumulate-outgoing-args
16665 Reserve space once for outgoing arguments in the function prologue rather
16666 than around each call. Generally beneficial for performance and size. Also
16667 needed for unwinding to avoid changing the stack frame around conditional code.
16669 @item -mdivsi3_libfunc=@var{name}
16670 @opindex mdivsi3_libfunc=@var{name}
16671 Set the name of the library function used for 32 bit signed division to
16672 @var{name}. This only affect the name used in the call and inv:call
16673 division strategies, and the compiler will still expect the same
16674 sets of input/output/clobbered registers as if this option was not present.
16676 @item -mfixed-range=@var{register-range}
16677 @opindex mfixed-range
16678 Generate code treating the given register range as fixed registers.
16679 A fixed register is one that the register allocator can not use. This is
16680 useful when compiling kernel code. A register range is specified as
16681 two registers separated by a dash. Multiple register ranges can be
16682 specified separated by a comma.
16684 @item -madjust-unroll
16685 @opindex madjust-unroll
16686 Throttle unrolling to avoid thrashing target registers.
16687 This option only has an effect if the gcc code base supports the
16688 TARGET_ADJUST_UNROLL_MAX target hook.
16690 @item -mindexed-addressing
16691 @opindex mindexed-addressing
16692 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16693 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16694 semantics for the indexed addressing mode. The architecture allows the
16695 implementation of processors with 64 bit MMU, which the OS could use to
16696 get 32 bit addressing, but since no current hardware implementation supports
16697 this or any other way to make the indexed addressing mode safe to use in
16698 the 32 bit ABI, the default is -mno-indexed-addressing.
16700 @item -mgettrcost=@var{number}
16701 @opindex mgettrcost=@var{number}
16702 Set the cost assumed for the gettr instruction to @var{number}.
16703 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16707 Assume pt* instructions won't trap. This will generally generate better
16708 scheduled code, but is unsafe on current hardware. The current architecture
16709 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16710 This has the unintentional effect of making it unsafe to schedule ptabs /
16711 ptrel before a branch, or hoist it out of a loop. For example,
16712 __do_global_ctors, a part of libgcc that runs constructors at program
16713 startup, calls functions in a list which is delimited by @minus{}1. With the
16714 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16715 That means that all the constructors will be run a bit quicker, but when
16716 the loop comes to the end of the list, the program crashes because ptabs
16717 loads @minus{}1 into a target register. Since this option is unsafe for any
16718 hardware implementing the current architecture specification, the default
16719 is -mno-pt-fixed. Unless the user specifies a specific cost with
16720 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16721 this deters register allocation using target registers for storing
16724 @item -minvalid-symbols
16725 @opindex minvalid-symbols
16726 Assume symbols might be invalid. Ordinary function symbols generated by
16727 the compiler will always be valid to load with movi/shori/ptabs or
16728 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16729 to generate symbols that will cause ptabs / ptrel to trap.
16730 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16731 It will then prevent cross-basic-block cse, hoisting and most scheduling
16732 of symbol loads. The default is @option{-mno-invalid-symbols}.
16735 @node Solaris 2 Options
16736 @subsection Solaris 2 Options
16737 @cindex Solaris 2 options
16739 These @samp{-m} options are supported on Solaris 2:
16742 @item -mimpure-text
16743 @opindex mimpure-text
16744 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16745 the compiler to not pass @option{-z text} to the linker when linking a
16746 shared object. Using this option, you can link position-dependent
16747 code into a shared object.
16749 @option{-mimpure-text} suppresses the ``relocations remain against
16750 allocatable but non-writable sections'' linker error message.
16751 However, the necessary relocations will trigger copy-on-write, and the
16752 shared object is not actually shared across processes. Instead of
16753 using @option{-mimpure-text}, you should compile all source code with
16754 @option{-fpic} or @option{-fPIC}.
16758 These switches are supported in addition to the above on Solaris 2:
16763 Add support for multithreading using the Solaris threads library. This
16764 option sets flags for both the preprocessor and linker. This option does
16765 not affect the thread safety of object code produced by the compiler or
16766 that of libraries supplied with it.
16770 Add support for multithreading using the POSIX threads library. This
16771 option sets flags for both the preprocessor and linker. This option does
16772 not affect the thread safety of object code produced by the compiler or
16773 that of libraries supplied with it.
16777 This is a synonym for @option{-pthreads}.
16780 @node SPARC Options
16781 @subsection SPARC Options
16782 @cindex SPARC options
16784 These @samp{-m} options are supported on the SPARC:
16787 @item -mno-app-regs
16789 @opindex mno-app-regs
16791 Specify @option{-mapp-regs} to generate output using the global registers
16792 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16795 To be fully SVR4 ABI compliant at the cost of some performance loss,
16796 specify @option{-mno-app-regs}. You should compile libraries and system
16797 software with this option.
16800 @itemx -mhard-float
16802 @opindex mhard-float
16803 Generate output containing floating point instructions. This is the
16807 @itemx -msoft-float
16809 @opindex msoft-float
16810 Generate output containing library calls for floating point.
16811 @strong{Warning:} the requisite libraries are not available for all SPARC
16812 targets. Normally the facilities of the machine's usual C compiler are
16813 used, but this cannot be done directly in cross-compilation. You must make
16814 your own arrangements to provide suitable library functions for
16815 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16816 @samp{sparclite-*-*} do provide software floating point support.
16818 @option{-msoft-float} changes the calling convention in the output file;
16819 therefore, it is only useful if you compile @emph{all} of a program with
16820 this option. In particular, you need to compile @file{libgcc.a}, the
16821 library that comes with GCC, with @option{-msoft-float} in order for
16824 @item -mhard-quad-float
16825 @opindex mhard-quad-float
16826 Generate output containing quad-word (long double) floating point
16829 @item -msoft-quad-float
16830 @opindex msoft-quad-float
16831 Generate output containing library calls for quad-word (long double)
16832 floating point instructions. The functions called are those specified
16833 in the SPARC ABI@. This is the default.
16835 As of this writing, there are no SPARC implementations that have hardware
16836 support for the quad-word floating point instructions. They all invoke
16837 a trap handler for one of these instructions, and then the trap handler
16838 emulates the effect of the instruction. Because of the trap handler overhead,
16839 this is much slower than calling the ABI library routines. Thus the
16840 @option{-msoft-quad-float} option is the default.
16842 @item -mno-unaligned-doubles
16843 @itemx -munaligned-doubles
16844 @opindex mno-unaligned-doubles
16845 @opindex munaligned-doubles
16846 Assume that doubles have 8 byte alignment. This is the default.
16848 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16849 alignment only if they are contained in another type, or if they have an
16850 absolute address. Otherwise, it assumes they have 4 byte alignment.
16851 Specifying this option avoids some rare compatibility problems with code
16852 generated by other compilers. It is not the default because it results
16853 in a performance loss, especially for floating point code.
16855 @item -mno-faster-structs
16856 @itemx -mfaster-structs
16857 @opindex mno-faster-structs
16858 @opindex mfaster-structs
16859 With @option{-mfaster-structs}, the compiler assumes that structures
16860 should have 8 byte alignment. This enables the use of pairs of
16861 @code{ldd} and @code{std} instructions for copies in structure
16862 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16863 However, the use of this changed alignment directly violates the SPARC
16864 ABI@. Thus, it's intended only for use on targets where the developer
16865 acknowledges that their resulting code will not be directly in line with
16866 the rules of the ABI@.
16868 @item -mcpu=@var{cpu_type}
16870 Set the instruction set, register set, and instruction scheduling parameters
16871 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16872 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16873 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16874 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16875 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16877 Default instruction scheduling parameters are used for values that select
16878 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16879 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16881 Here is a list of each supported architecture and their supported
16886 v8: supersparc, hypersparc
16887 sparclite: f930, f934, sparclite86x
16889 v9: ultrasparc, ultrasparc3, niagara, niagara2
16892 By default (unless configured otherwise), GCC generates code for the V7
16893 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16894 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16895 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16896 SPARCStation 1, 2, IPX etc.
16898 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16899 architecture. The only difference from V7 code is that the compiler emits
16900 the integer multiply and integer divide instructions which exist in SPARC-V8
16901 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16902 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16905 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16906 the SPARC architecture. This adds the integer multiply, integer divide step
16907 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16908 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16909 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16910 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16911 MB86934 chip, which is the more recent SPARClite with FPU@.
16913 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16914 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16915 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16916 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16917 optimizes it for the TEMIC SPARClet chip.
16919 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16920 architecture. This adds 64-bit integer and floating-point move instructions,
16921 3 additional floating-point condition code registers and conditional move
16922 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16923 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16924 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16925 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16926 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16927 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16928 additionally optimizes it for Sun UltraSPARC T2 chips.
16930 @item -mtune=@var{cpu_type}
16932 Set the instruction scheduling parameters for machine type
16933 @var{cpu_type}, but do not set the instruction set or register set that the
16934 option @option{-mcpu=@var{cpu_type}} would.
16936 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16937 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16938 that select a particular cpu implementation. Those are @samp{cypress},
16939 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16940 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16941 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16946 @opindex mno-v8plus
16947 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16948 difference from the V8 ABI is that the global and out registers are
16949 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16950 mode for all SPARC-V9 processors.
16956 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16957 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16960 These @samp{-m} options are supported in addition to the above
16961 on SPARC-V9 processors in 64-bit environments:
16964 @item -mlittle-endian
16965 @opindex mlittle-endian
16966 Generate code for a processor running in little-endian mode. It is only
16967 available for a few configurations and most notably not on Solaris and Linux.
16973 Generate code for a 32-bit or 64-bit environment.
16974 The 32-bit environment sets int, long and pointer to 32 bits.
16975 The 64-bit environment sets int to 32 bits and long and pointer
16978 @item -mcmodel=medlow
16979 @opindex mcmodel=medlow
16980 Generate code for the Medium/Low code model: 64-bit addresses, programs
16981 must be linked in the low 32 bits of memory. Programs can be statically
16982 or dynamically linked.
16984 @item -mcmodel=medmid
16985 @opindex mcmodel=medmid
16986 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16987 must be linked in the low 44 bits of memory, the text and data segments must
16988 be less than 2GB in size and the data segment must be located within 2GB of
16991 @item -mcmodel=medany
16992 @opindex mcmodel=medany
16993 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16994 may be linked anywhere in memory, the text and data segments must be less
16995 than 2GB in size and the data segment must be located within 2GB of the
16998 @item -mcmodel=embmedany
16999 @opindex mcmodel=embmedany
17000 Generate code for the Medium/Anywhere code model for embedded systems:
17001 64-bit addresses, the text and data segments must be less than 2GB in
17002 size, both starting anywhere in memory (determined at link time). The
17003 global register %g4 points to the base of the data segment. Programs
17004 are statically linked and PIC is not supported.
17007 @itemx -mno-stack-bias
17008 @opindex mstack-bias
17009 @opindex mno-stack-bias
17010 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17011 frame pointer if present, are offset by @minus{}2047 which must be added back
17012 when making stack frame references. This is the default in 64-bit mode.
17013 Otherwise, assume no such offset is present.
17017 @subsection SPU Options
17018 @cindex SPU options
17020 These @samp{-m} options are supported on the SPU:
17024 @itemx -merror-reloc
17025 @opindex mwarn-reloc
17026 @opindex merror-reloc
17028 The loader for SPU does not handle dynamic relocations. By default, GCC
17029 will give an error when it generates code that requires a dynamic
17030 relocation. @option{-mno-error-reloc} disables the error,
17031 @option{-mwarn-reloc} will generate a warning instead.
17034 @itemx -munsafe-dma
17036 @opindex munsafe-dma
17038 Instructions which initiate or test completion of DMA must not be
17039 reordered with respect to loads and stores of the memory which is being
17040 accessed. Users typically address this problem using the volatile
17041 keyword, but that can lead to inefficient code in places where the
17042 memory is known to not change. Rather than mark the memory as volatile
17043 we treat the DMA instructions as potentially effecting all memory. With
17044 @option{-munsafe-dma} users must use the volatile keyword to protect
17047 @item -mbranch-hints
17048 @opindex mbranch-hints
17050 By default, GCC will generate a branch hint instruction to avoid
17051 pipeline stalls for always taken or probably taken branches. A hint
17052 will not be generated closer than 8 instructions away from its branch.
17053 There is little reason to disable them, except for debugging purposes,
17054 or to make an object a little bit smaller.
17058 @opindex msmall-mem
17059 @opindex mlarge-mem
17061 By default, GCC generates code assuming that addresses are never larger
17062 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17063 a full 32 bit address.
17068 By default, GCC links against startup code that assumes the SPU-style
17069 main function interface (which has an unconventional parameter list).
17070 With @option{-mstdmain}, GCC will link your program against startup
17071 code that assumes a C99-style interface to @code{main}, including a
17072 local copy of @code{argv} strings.
17074 @item -mfixed-range=@var{register-range}
17075 @opindex mfixed-range
17076 Generate code treating the given register range as fixed registers.
17077 A fixed register is one that the register allocator can not use. This is
17078 useful when compiling kernel code. A register range is specified as
17079 two registers separated by a dash. Multiple register ranges can be
17080 specified separated by a comma.
17086 Compile code assuming that pointers to the PPU address space accessed
17087 via the @code{__ea} named address space qualifier are either 32 or 64
17088 bits wide. The default is 32 bits. As this is an ABI changing option,
17089 all object code in an executable must be compiled with the same setting.
17091 @item -maddress-space-conversion
17092 @itemx -mno-address-space-conversion
17093 @opindex maddress-space-conversion
17094 @opindex mno-address-space-conversion
17095 Allow/disallow treating the @code{__ea} address space as superset
17096 of the generic address space. This enables explicit type casts
17097 between @code{__ea} and generic pointer as well as implicit
17098 conversions of generic pointers to @code{__ea} pointers. The
17099 default is to allow address space pointer conversions.
17101 @item -mcache-size=@var{cache-size}
17102 @opindex mcache-size
17103 This option controls the version of libgcc that the compiler links to an
17104 executable and selects a software-managed cache for accessing variables
17105 in the @code{__ea} address space with a particular cache size. Possible
17106 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17107 and @samp{128}. The default cache size is 64KB.
17109 @item -matomic-updates
17110 @itemx -mno-atomic-updates
17111 @opindex matomic-updates
17112 @opindex mno-atomic-updates
17113 This option controls the version of libgcc that the compiler links to an
17114 executable and selects whether atomic updates to the software-managed
17115 cache of PPU-side variables are used. If you use atomic updates, changes
17116 to a PPU variable from SPU code using the @code{__ea} named address space
17117 qualifier will not interfere with changes to other PPU variables residing
17118 in the same cache line from PPU code. If you do not use atomic updates,
17119 such interference may occur; however, writing back cache lines will be
17120 more efficient. The default behavior is to use atomic updates.
17123 @itemx -mdual-nops=@var{n}
17124 @opindex mdual-nops
17125 By default, GCC will insert nops to increase dual issue when it expects
17126 it to increase performance. @var{n} can be a value from 0 to 10. A
17127 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17128 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17130 @item -mhint-max-nops=@var{n}
17131 @opindex mhint-max-nops
17132 Maximum number of nops to insert for a branch hint. A branch hint must
17133 be at least 8 instructions away from the branch it is effecting. GCC
17134 will insert up to @var{n} nops to enforce this, otherwise it will not
17135 generate the branch hint.
17137 @item -mhint-max-distance=@var{n}
17138 @opindex mhint-max-distance
17139 The encoding of the branch hint instruction limits the hint to be within
17140 256 instructions of the branch it is effecting. By default, GCC makes
17141 sure it is within 125.
17144 @opindex msafe-hints
17145 Work around a hardware bug which causes the SPU to stall indefinitely.
17146 By default, GCC will insert the @code{hbrp} instruction to make sure
17147 this stall won't happen.
17151 @node System V Options
17152 @subsection Options for System V
17154 These additional options are available on System V Release 4 for
17155 compatibility with other compilers on those systems:
17160 Create a shared object.
17161 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17165 Identify the versions of each tool used by the compiler, in a
17166 @code{.ident} assembler directive in the output.
17170 Refrain from adding @code{.ident} directives to the output file (this is
17173 @item -YP,@var{dirs}
17175 Search the directories @var{dirs}, and no others, for libraries
17176 specified with @option{-l}.
17178 @item -Ym,@var{dir}
17180 Look in the directory @var{dir} to find the M4 preprocessor.
17181 The assembler uses this option.
17182 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17183 @c the generic assembler that comes with Solaris takes just -Ym.
17187 @subsection V850 Options
17188 @cindex V850 Options
17190 These @samp{-m} options are defined for V850 implementations:
17194 @itemx -mno-long-calls
17195 @opindex mlong-calls
17196 @opindex mno-long-calls
17197 Treat all calls as being far away (near). If calls are assumed to be
17198 far away, the compiler will always load the functions address up into a
17199 register, and call indirect through the pointer.
17205 Do not optimize (do optimize) basic blocks that use the same index
17206 pointer 4 or more times to copy pointer into the @code{ep} register, and
17207 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17208 option is on by default if you optimize.
17210 @item -mno-prolog-function
17211 @itemx -mprolog-function
17212 @opindex mno-prolog-function
17213 @opindex mprolog-function
17214 Do not use (do use) external functions to save and restore registers
17215 at the prologue and epilogue of a function. The external functions
17216 are slower, but use less code space if more than one function saves
17217 the same number of registers. The @option{-mprolog-function} option
17218 is on by default if you optimize.
17222 Try to make the code as small as possible. At present, this just turns
17223 on the @option{-mep} and @option{-mprolog-function} options.
17225 @item -mtda=@var{n}
17227 Put static or global variables whose size is @var{n} bytes or less into
17228 the tiny data area that register @code{ep} points to. The tiny data
17229 area can hold up to 256 bytes in total (128 bytes for byte references).
17231 @item -msda=@var{n}
17233 Put static or global variables whose size is @var{n} bytes or less into
17234 the small data area that register @code{gp} points to. The small data
17235 area can hold up to 64 kilobytes.
17237 @item -mzda=@var{n}
17239 Put static or global variables whose size is @var{n} bytes or less into
17240 the first 32 kilobytes of memory.
17244 Specify that the target processor is the V850.
17247 @opindex mbig-switch
17248 Generate code suitable for big switch tables. Use this option only if
17249 the assembler/linker complain about out of range branches within a switch
17254 This option will cause r2 and r5 to be used in the code generated by
17255 the compiler. This setting is the default.
17257 @item -mno-app-regs
17258 @opindex mno-app-regs
17259 This option will cause r2 and r5 to be treated as fixed registers.
17263 Specify that the target processor is the V850E2V3. The preprocessor
17264 constants @samp{__v850e2v3__} will be defined if
17265 this option is used.
17269 Specify that the target processor is the V850E2. The preprocessor
17270 constants @samp{__v850e2__} will be defined if
17274 Specify that the target processor is the V850E1. The preprocessor
17275 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17279 Specify that the target processor is the V850E@. The preprocessor
17280 constant @samp{__v850e__} will be defined if this option is used.
17282 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17283 nor @option{-mv850e2} nor @option{-mv850e2v3}
17284 are defined then a default target processor will be chosen and the
17285 relevant @samp{__v850*__} preprocessor constant will be defined.
17287 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17288 defined, regardless of which processor variant is the target.
17290 @item -mdisable-callt
17291 @opindex mdisable-callt
17292 This option will suppress generation of the CALLT instruction for the
17293 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17294 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17299 @subsection VAX Options
17300 @cindex VAX options
17302 These @samp{-m} options are defined for the VAX:
17307 Do not output certain jump instructions (@code{aobleq} and so on)
17308 that the Unix assembler for the VAX cannot handle across long
17313 Do output those jump instructions, on the assumption that you
17314 will assemble with the GNU assembler.
17318 Output code for g-format floating point numbers instead of d-format.
17321 @node VxWorks Options
17322 @subsection VxWorks Options
17323 @cindex VxWorks Options
17325 The options in this section are defined for all VxWorks targets.
17326 Options specific to the target hardware are listed with the other
17327 options for that target.
17332 GCC can generate code for both VxWorks kernels and real time processes
17333 (RTPs). This option switches from the former to the latter. It also
17334 defines the preprocessor macro @code{__RTP__}.
17337 @opindex non-static
17338 Link an RTP executable against shared libraries rather than static
17339 libraries. The options @option{-static} and @option{-shared} can
17340 also be used for RTPs (@pxref{Link Options}); @option{-static}
17347 These options are passed down to the linker. They are defined for
17348 compatibility with Diab.
17351 @opindex Xbind-lazy
17352 Enable lazy binding of function calls. This option is equivalent to
17353 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17357 Disable lazy binding of function calls. This option is the default and
17358 is defined for compatibility with Diab.
17361 @node x86-64 Options
17362 @subsection x86-64 Options
17363 @cindex x86-64 options
17365 These are listed under @xref{i386 and x86-64 Options}.
17367 @node i386 and x86-64 Windows Options
17368 @subsection i386 and x86-64 Windows Options
17369 @cindex i386 and x86-64 Windows Options
17371 These additional options are available for Windows targets:
17376 This option is available for Cygwin and MinGW targets. It
17377 specifies that a console application is to be generated, by
17378 instructing the linker to set the PE header subsystem type
17379 required for console applications.
17380 This is the default behavior for Cygwin and MinGW targets.
17384 This option is available for Cygwin targets. It specifies that
17385 the Cygwin internal interface is to be used for predefined
17386 preprocessor macros, C runtime libraries and related linker
17387 paths and options. For Cygwin targets this is the default behavior.
17388 This option is deprecated and will be removed in a future release.
17391 @opindex mno-cygwin
17392 This option is available for Cygwin targets. It specifies that
17393 the MinGW internal interface is to be used instead of Cygwin's, by
17394 setting MinGW-related predefined macros and linker paths and default
17396 This option is deprecated and will be removed in a future release.
17400 This option is available for Cygwin and MinGW targets. It
17401 specifies that a DLL - a dynamic link library - is to be
17402 generated, enabling the selection of the required runtime
17403 startup object and entry point.
17405 @item -mnop-fun-dllimport
17406 @opindex mnop-fun-dllimport
17407 This option is available for Cygwin and MinGW targets. It
17408 specifies that the dllimport attribute should be ignored.
17412 This option is available for MinGW targets. It specifies
17413 that MinGW-specific thread support is to be used.
17417 This option is available for mingw-w64 targets. It specifies
17418 that the UNICODE macro is getting pre-defined and that the
17419 unicode capable runtime startup code is chosen.
17423 This option is available for Cygwin and MinGW targets. It
17424 specifies that the typical Windows pre-defined macros are to
17425 be set in the pre-processor, but does not influence the choice
17426 of runtime library/startup code.
17430 This option is available for Cygwin and MinGW targets. It
17431 specifies that a GUI application is to be generated by
17432 instructing the linker to set the PE header subsystem type
17435 @item -fno-set-stack-executable
17436 @opindex fno-set-stack-executable
17437 This option is available for MinGW targets. It specifies that
17438 the executable flag for stack used by nested functions isn't
17439 set. This is necessary for binaries running in kernel mode of
17440 Windows, as there the user32 API, which is used to set executable
17441 privileges, isn't available.
17443 @item -mpe-aligned-commons
17444 @opindex mpe-aligned-commons
17445 This option is available for Cygwin and MinGW targets. It
17446 specifies that the GNU extension to the PE file format that
17447 permits the correct alignment of COMMON variables should be
17448 used when generating code. It will be enabled by default if
17449 GCC detects that the target assembler found during configuration
17450 supports the feature.
17453 See also under @ref{i386 and x86-64 Options} for standard options.
17455 @node Xstormy16 Options
17456 @subsection Xstormy16 Options
17457 @cindex Xstormy16 Options
17459 These options are defined for Xstormy16:
17464 Choose startup files and linker script suitable for the simulator.
17467 @node Xtensa Options
17468 @subsection Xtensa Options
17469 @cindex Xtensa Options
17471 These options are supported for Xtensa targets:
17475 @itemx -mno-const16
17477 @opindex mno-const16
17478 Enable or disable use of @code{CONST16} instructions for loading
17479 constant values. The @code{CONST16} instruction is currently not a
17480 standard option from Tensilica. When enabled, @code{CONST16}
17481 instructions are always used in place of the standard @code{L32R}
17482 instructions. The use of @code{CONST16} is enabled by default only if
17483 the @code{L32R} instruction is not available.
17486 @itemx -mno-fused-madd
17487 @opindex mfused-madd
17488 @opindex mno-fused-madd
17489 Enable or disable use of fused multiply/add and multiply/subtract
17490 instructions in the floating-point option. This has no effect if the
17491 floating-point option is not also enabled. Disabling fused multiply/add
17492 and multiply/subtract instructions forces the compiler to use separate
17493 instructions for the multiply and add/subtract operations. This may be
17494 desirable in some cases where strict IEEE 754-compliant results are
17495 required: the fused multiply add/subtract instructions do not round the
17496 intermediate result, thereby producing results with @emph{more} bits of
17497 precision than specified by the IEEE standard. Disabling fused multiply
17498 add/subtract instructions also ensures that the program output is not
17499 sensitive to the compiler's ability to combine multiply and add/subtract
17502 @item -mserialize-volatile
17503 @itemx -mno-serialize-volatile
17504 @opindex mserialize-volatile
17505 @opindex mno-serialize-volatile
17506 When this option is enabled, GCC inserts @code{MEMW} instructions before
17507 @code{volatile} memory references to guarantee sequential consistency.
17508 The default is @option{-mserialize-volatile}. Use
17509 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17511 @item -mforce-no-pic
17512 @opindex mforce-no-pic
17513 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17514 position-independent code (PIC), this option disables PIC for compiling
17517 @item -mtext-section-literals
17518 @itemx -mno-text-section-literals
17519 @opindex mtext-section-literals
17520 @opindex mno-text-section-literals
17521 Control the treatment of literal pools. The default is
17522 @option{-mno-text-section-literals}, which places literals in a separate
17523 section in the output file. This allows the literal pool to be placed
17524 in a data RAM/ROM, and it also allows the linker to combine literal
17525 pools from separate object files to remove redundant literals and
17526 improve code size. With @option{-mtext-section-literals}, the literals
17527 are interspersed in the text section in order to keep them as close as
17528 possible to their references. This may be necessary for large assembly
17531 @item -mtarget-align
17532 @itemx -mno-target-align
17533 @opindex mtarget-align
17534 @opindex mno-target-align
17535 When this option is enabled, GCC instructs the assembler to
17536 automatically align instructions to reduce branch penalties at the
17537 expense of some code density. The assembler attempts to widen density
17538 instructions to align branch targets and the instructions following call
17539 instructions. If there are not enough preceding safe density
17540 instructions to align a target, no widening will be performed. The
17541 default is @option{-mtarget-align}. These options do not affect the
17542 treatment of auto-aligned instructions like @code{LOOP}, which the
17543 assembler will always align, either by widening density instructions or
17544 by inserting no-op instructions.
17547 @itemx -mno-longcalls
17548 @opindex mlongcalls
17549 @opindex mno-longcalls
17550 When this option is enabled, GCC instructs the assembler to translate
17551 direct calls to indirect calls unless it can determine that the target
17552 of a direct call is in the range allowed by the call instruction. This
17553 translation typically occurs for calls to functions in other source
17554 files. Specifically, the assembler translates a direct @code{CALL}
17555 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17556 The default is @option{-mno-longcalls}. This option should be used in
17557 programs where the call target can potentially be out of range. This
17558 option is implemented in the assembler, not the compiler, so the
17559 assembly code generated by GCC will still show direct call
17560 instructions---look at the disassembled object code to see the actual
17561 instructions. Note that the assembler will use an indirect call for
17562 every cross-file call, not just those that really will be out of range.
17565 @node zSeries Options
17566 @subsection zSeries Options
17567 @cindex zSeries options
17569 These are listed under @xref{S/390 and zSeries Options}.
17571 @node Code Gen Options
17572 @section Options for Code Generation Conventions
17573 @cindex code generation conventions
17574 @cindex options, code generation
17575 @cindex run-time options
17577 These machine-independent options control the interface conventions
17578 used in code generation.
17580 Most of them have both positive and negative forms; the negative form
17581 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17582 one of the forms is listed---the one which is not the default. You
17583 can figure out the other form by either removing @samp{no-} or adding
17587 @item -fbounds-check
17588 @opindex fbounds-check
17589 For front-ends that support it, generate additional code to check that
17590 indices used to access arrays are within the declared range. This is
17591 currently only supported by the Java and Fortran front-ends, where
17592 this option defaults to true and false respectively.
17596 This option generates traps for signed overflow on addition, subtraction,
17597 multiplication operations.
17601 This option instructs the compiler to assume that signed arithmetic
17602 overflow of addition, subtraction and multiplication wraps around
17603 using twos-complement representation. This flag enables some optimizations
17604 and disables others. This option is enabled by default for the Java
17605 front-end, as required by the Java language specification.
17608 @opindex fexceptions
17609 Enable exception handling. Generates extra code needed to propagate
17610 exceptions. For some targets, this implies GCC will generate frame
17611 unwind information for all functions, which can produce significant data
17612 size overhead, although it does not affect execution. If you do not
17613 specify this option, GCC will enable it by default for languages like
17614 C++ which normally require exception handling, and disable it for
17615 languages like C that do not normally require it. However, you may need
17616 to enable this option when compiling C code that needs to interoperate
17617 properly with exception handlers written in C++. You may also wish to
17618 disable this option if you are compiling older C++ programs that don't
17619 use exception handling.
17621 @item -fnon-call-exceptions
17622 @opindex fnon-call-exceptions
17623 Generate code that allows trapping instructions to throw exceptions.
17624 Note that this requires platform-specific runtime support that does
17625 not exist everywhere. Moreover, it only allows @emph{trapping}
17626 instructions to throw exceptions, i.e.@: memory references or floating
17627 point instructions. It does not allow exceptions to be thrown from
17628 arbitrary signal handlers such as @code{SIGALRM}.
17630 @item -funwind-tables
17631 @opindex funwind-tables
17632 Similar to @option{-fexceptions}, except that it will just generate any needed
17633 static data, but will not affect the generated code in any other way.
17634 You will normally not enable this option; instead, a language processor
17635 that needs this handling would enable it on your behalf.
17637 @item -fasynchronous-unwind-tables
17638 @opindex fasynchronous-unwind-tables
17639 Generate unwind table in dwarf2 format, if supported by target machine. The
17640 table is exact at each instruction boundary, so it can be used for stack
17641 unwinding from asynchronous events (such as debugger or garbage collector).
17643 @item -fpcc-struct-return
17644 @opindex fpcc-struct-return
17645 Return ``short'' @code{struct} and @code{union} values in memory like
17646 longer ones, rather than in registers. This convention is less
17647 efficient, but it has the advantage of allowing intercallability between
17648 GCC-compiled files and files compiled with other compilers, particularly
17649 the Portable C Compiler (pcc).
17651 The precise convention for returning structures in memory depends
17652 on the target configuration macros.
17654 Short structures and unions are those whose size and alignment match
17655 that of some integer type.
17657 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17658 switch is not binary compatible with code compiled with the
17659 @option{-freg-struct-return} switch.
17660 Use it to conform to a non-default application binary interface.
17662 @item -freg-struct-return
17663 @opindex freg-struct-return
17664 Return @code{struct} and @code{union} values in registers when possible.
17665 This is more efficient for small structures than
17666 @option{-fpcc-struct-return}.
17668 If you specify neither @option{-fpcc-struct-return} nor
17669 @option{-freg-struct-return}, GCC defaults to whichever convention is
17670 standard for the target. If there is no standard convention, GCC
17671 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17672 the principal compiler. In those cases, we can choose the standard, and
17673 we chose the more efficient register return alternative.
17675 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17676 switch is not binary compatible with code compiled with the
17677 @option{-fpcc-struct-return} switch.
17678 Use it to conform to a non-default application binary interface.
17680 @item -fshort-enums
17681 @opindex fshort-enums
17682 Allocate to an @code{enum} type only as many bytes as it needs for the
17683 declared range of possible values. Specifically, the @code{enum} type
17684 will be equivalent to the smallest integer type which has enough room.
17686 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17687 code that is not binary compatible with code generated without that switch.
17688 Use it to conform to a non-default application binary interface.
17690 @item -fshort-double
17691 @opindex fshort-double
17692 Use the same size for @code{double} as for @code{float}.
17694 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17695 code that is not binary compatible with code generated without that switch.
17696 Use it to conform to a non-default application binary interface.
17698 @item -fshort-wchar
17699 @opindex fshort-wchar
17700 Override the underlying type for @samp{wchar_t} to be @samp{short
17701 unsigned int} instead of the default for the target. This option is
17702 useful for building programs to run under WINE@.
17704 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17705 code that is not binary compatible with code generated without that switch.
17706 Use it to conform to a non-default application binary interface.
17709 @opindex fno-common
17710 In C code, controls the placement of uninitialized global variables.
17711 Unix C compilers have traditionally permitted multiple definitions of
17712 such variables in different compilation units by placing the variables
17714 This is the behavior specified by @option{-fcommon}, and is the default
17715 for GCC on most targets.
17716 On the other hand, this behavior is not required by ISO C, and on some
17717 targets may carry a speed or code size penalty on variable references.
17718 The @option{-fno-common} option specifies that the compiler should place
17719 uninitialized global variables in the data section of the object file,
17720 rather than generating them as common blocks.
17721 This has the effect that if the same variable is declared
17722 (without @code{extern}) in two different compilations,
17723 you will get a multiple-definition error when you link them.
17724 In this case, you must compile with @option{-fcommon} instead.
17725 Compiling with @option{-fno-common} is useful on targets for which
17726 it provides better performance, or if you wish to verify that the
17727 program will work on other systems which always treat uninitialized
17728 variable declarations this way.
17732 Ignore the @samp{#ident} directive.
17734 @item -finhibit-size-directive
17735 @opindex finhibit-size-directive
17736 Don't output a @code{.size} assembler directive, or anything else that
17737 would cause trouble if the function is split in the middle, and the
17738 two halves are placed at locations far apart in memory. This option is
17739 used when compiling @file{crtstuff.c}; you should not need to use it
17742 @item -fverbose-asm
17743 @opindex fverbose-asm
17744 Put extra commentary information in the generated assembly code to
17745 make it more readable. This option is generally only of use to those
17746 who actually need to read the generated assembly code (perhaps while
17747 debugging the compiler itself).
17749 @option{-fno-verbose-asm}, the default, causes the
17750 extra information to be omitted and is useful when comparing two assembler
17753 @item -frecord-gcc-switches
17754 @opindex frecord-gcc-switches
17755 This switch causes the command line that was used to invoke the
17756 compiler to be recorded into the object file that is being created.
17757 This switch is only implemented on some targets and the exact format
17758 of the recording is target and binary file format dependent, but it
17759 usually takes the form of a section containing ASCII text. This
17760 switch is related to the @option{-fverbose-asm} switch, but that
17761 switch only records information in the assembler output file as
17762 comments, so it never reaches the object file.
17766 @cindex global offset table
17768 Generate position-independent code (PIC) suitable for use in a shared
17769 library, if supported for the target machine. Such code accesses all
17770 constant addresses through a global offset table (GOT)@. The dynamic
17771 loader resolves the GOT entries when the program starts (the dynamic
17772 loader is not part of GCC; it is part of the operating system). If
17773 the GOT size for the linked executable exceeds a machine-specific
17774 maximum size, you get an error message from the linker indicating that
17775 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17776 instead. (These maximums are 8k on the SPARC and 32k
17777 on the m68k and RS/6000. The 386 has no such limit.)
17779 Position-independent code requires special support, and therefore works
17780 only on certain machines. For the 386, GCC supports PIC for System V
17781 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17782 position-independent.
17784 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17789 If supported for the target machine, emit position-independent code,
17790 suitable for dynamic linking and avoiding any limit on the size of the
17791 global offset table. This option makes a difference on the m68k,
17792 PowerPC and SPARC@.
17794 Position-independent code requires special support, and therefore works
17795 only on certain machines.
17797 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17804 These options are similar to @option{-fpic} and @option{-fPIC}, but
17805 generated position independent code can be only linked into executables.
17806 Usually these options are used when @option{-pie} GCC option will be
17807 used during linking.
17809 @option{-fpie} and @option{-fPIE} both define the macros
17810 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17811 for @option{-fpie} and 2 for @option{-fPIE}.
17813 @item -fno-jump-tables
17814 @opindex fno-jump-tables
17815 Do not use jump tables for switch statements even where it would be
17816 more efficient than other code generation strategies. This option is
17817 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17818 building code which forms part of a dynamic linker and cannot
17819 reference the address of a jump table. On some targets, jump tables
17820 do not require a GOT and this option is not needed.
17822 @item -ffixed-@var{reg}
17824 Treat the register named @var{reg} as a fixed register; generated code
17825 should never refer to it (except perhaps as a stack pointer, frame
17826 pointer or in some other fixed role).
17828 @var{reg} must be the name of a register. The register names accepted
17829 are machine-specific and are defined in the @code{REGISTER_NAMES}
17830 macro in the machine description macro file.
17832 This flag does not have a negative form, because it specifies a
17835 @item -fcall-used-@var{reg}
17836 @opindex fcall-used
17837 Treat the register named @var{reg} as an allocable register that is
17838 clobbered by function calls. It may be allocated for temporaries or
17839 variables that do not live across a call. Functions compiled this way
17840 will not save and restore the register @var{reg}.
17842 It is an error to used this flag with the frame pointer or stack pointer.
17843 Use of this flag for other registers that have fixed pervasive roles in
17844 the machine's execution model will produce disastrous results.
17846 This flag does not have a negative form, because it specifies a
17849 @item -fcall-saved-@var{reg}
17850 @opindex fcall-saved
17851 Treat the register named @var{reg} as an allocable register saved by
17852 functions. It may be allocated even for temporaries or variables that
17853 live across a call. Functions compiled this way will save and restore
17854 the register @var{reg} if they use it.
17856 It is an error to used this flag with the frame pointer or stack pointer.
17857 Use of this flag for other registers that have fixed pervasive roles in
17858 the machine's execution model will produce disastrous results.
17860 A different sort of disaster will result from the use of this flag for
17861 a register in which function values may be returned.
17863 This flag does not have a negative form, because it specifies a
17866 @item -fpack-struct[=@var{n}]
17867 @opindex fpack-struct
17868 Without a value specified, pack all structure members together without
17869 holes. When a value is specified (which must be a small power of two), pack
17870 structure members according to this value, representing the maximum
17871 alignment (that is, objects with default alignment requirements larger than
17872 this will be output potentially unaligned at the next fitting location.
17874 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17875 code that is not binary compatible with code generated without that switch.
17876 Additionally, it makes the code suboptimal.
17877 Use it to conform to a non-default application binary interface.
17879 @item -finstrument-functions
17880 @opindex finstrument-functions
17881 Generate instrumentation calls for entry and exit to functions. Just
17882 after function entry and just before function exit, the following
17883 profiling functions will be called with the address of the current
17884 function and its call site. (On some platforms,
17885 @code{__builtin_return_address} does not work beyond the current
17886 function, so the call site information may not be available to the
17887 profiling functions otherwise.)
17890 void __cyg_profile_func_enter (void *this_fn,
17892 void __cyg_profile_func_exit (void *this_fn,
17896 The first argument is the address of the start of the current function,
17897 which may be looked up exactly in the symbol table.
17899 This instrumentation is also done for functions expanded inline in other
17900 functions. The profiling calls will indicate where, conceptually, the
17901 inline function is entered and exited. This means that addressable
17902 versions of such functions must be available. If all your uses of a
17903 function are expanded inline, this may mean an additional expansion of
17904 code size. If you use @samp{extern inline} in your C code, an
17905 addressable version of such functions must be provided. (This is
17906 normally the case anyways, but if you get lucky and the optimizer always
17907 expands the functions inline, you might have gotten away without
17908 providing static copies.)
17910 A function may be given the attribute @code{no_instrument_function}, in
17911 which case this instrumentation will not be done. This can be used, for
17912 example, for the profiling functions listed above, high-priority
17913 interrupt routines, and any functions from which the profiling functions
17914 cannot safely be called (perhaps signal handlers, if the profiling
17915 routines generate output or allocate memory).
17917 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17918 @opindex finstrument-functions-exclude-file-list
17920 Set the list of functions that are excluded from instrumentation (see
17921 the description of @code{-finstrument-functions}). If the file that
17922 contains a function definition matches with one of @var{file}, then
17923 that function is not instrumented. The match is done on substrings:
17924 if the @var{file} parameter is a substring of the file name, it is
17925 considered to be a match.
17930 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
17934 will exclude any inline function defined in files whose pathnames
17935 contain @code{/bits/stl} or @code{include/sys}.
17937 If, for some reason, you want to include letter @code{','} in one of
17938 @var{sym}, write @code{'\,'}. For example,
17939 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17940 (note the single quote surrounding the option).
17942 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17943 @opindex finstrument-functions-exclude-function-list
17945 This is similar to @code{-finstrument-functions-exclude-file-list},
17946 but this option sets the list of function names to be excluded from
17947 instrumentation. The function name to be matched is its user-visible
17948 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17949 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17950 match is done on substrings: if the @var{sym} parameter is a substring
17951 of the function name, it is considered to be a match. For C99 and C++
17952 extended identifiers, the function name must be given in UTF-8, not
17953 using universal character names.
17955 @item -fstack-check
17956 @opindex fstack-check
17957 Generate code to verify that you do not go beyond the boundary of the
17958 stack. You should specify this flag if you are running in an
17959 environment with multiple threads, but only rarely need to specify it in
17960 a single-threaded environment since stack overflow is automatically
17961 detected on nearly all systems if there is only one stack.
17963 Note that this switch does not actually cause checking to be done; the
17964 operating system or the language runtime must do that. The switch causes
17965 generation of code to ensure that they see the stack being extended.
17967 You can additionally specify a string parameter: @code{no} means no
17968 checking, @code{generic} means force the use of old-style checking,
17969 @code{specific} means use the best checking method and is equivalent
17970 to bare @option{-fstack-check}.
17972 Old-style checking is a generic mechanism that requires no specific
17973 target support in the compiler but comes with the following drawbacks:
17977 Modified allocation strategy for large objects: they will always be
17978 allocated dynamically if their size exceeds a fixed threshold.
17981 Fixed limit on the size of the static frame of functions: when it is
17982 topped by a particular function, stack checking is not reliable and
17983 a warning is issued by the compiler.
17986 Inefficiency: because of both the modified allocation strategy and the
17987 generic implementation, the performances of the code are hampered.
17990 Note that old-style stack checking is also the fallback method for
17991 @code{specific} if no target support has been added in the compiler.
17993 @item -fstack-limit-register=@var{reg}
17994 @itemx -fstack-limit-symbol=@var{sym}
17995 @itemx -fno-stack-limit
17996 @opindex fstack-limit-register
17997 @opindex fstack-limit-symbol
17998 @opindex fno-stack-limit
17999 Generate code to ensure that the stack does not grow beyond a certain value,
18000 either the value of a register or the address of a symbol. If the stack
18001 would grow beyond the value, a signal is raised. For most targets,
18002 the signal is raised before the stack overruns the boundary, so
18003 it is possible to catch the signal without taking special precautions.
18005 For instance, if the stack starts at absolute address @samp{0x80000000}
18006 and grows downwards, you can use the flags
18007 @option{-fstack-limit-symbol=__stack_limit} and
18008 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18009 of 128KB@. Note that this may only work with the GNU linker.
18011 @item -fsplit-stack
18012 @opindex fsplit-stack
18013 Generate code to automatically split the stack before it overflows.
18014 The resulting program has a discontiguous stack which can only
18015 overflow if the program is unable to allocate any more memory. This
18016 is most useful when running threaded programs, as it is no longer
18017 necessary to calculate a good stack size to use for each thread. This
18018 is currently only implemented for the i386 and x86_64 backends running
18021 When code compiled with @option{-fsplit-stack} calls code compiled
18022 without @option{-fsplit-stack}, there may not be much stack space
18023 available for the latter code to run. If compiling all code,
18024 including library code, with @option{-fsplit-stack} is not an option,
18025 then the linker can fix up these calls so that the code compiled
18026 without @option{-fsplit-stack} always has a large stack. Support for
18027 this is implemented in the gold linker in GNU binutils release 2.21
18030 @item -fleading-underscore
18031 @opindex fleading-underscore
18032 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18033 change the way C symbols are represented in the object file. One use
18034 is to help link with legacy assembly code.
18036 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18037 generate code that is not binary compatible with code generated without that
18038 switch. Use it to conform to a non-default application binary interface.
18039 Not all targets provide complete support for this switch.
18041 @item -ftls-model=@var{model}
18042 @opindex ftls-model
18043 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18044 The @var{model} argument should be one of @code{global-dynamic},
18045 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18047 The default without @option{-fpic} is @code{initial-exec}; with
18048 @option{-fpic} the default is @code{global-dynamic}.
18050 @item -fvisibility=@var{default|internal|hidden|protected}
18051 @opindex fvisibility
18052 Set the default ELF image symbol visibility to the specified option---all
18053 symbols will be marked with this unless overridden within the code.
18054 Using this feature can very substantially improve linking and
18055 load times of shared object libraries, produce more optimized
18056 code, provide near-perfect API export and prevent symbol clashes.
18057 It is @strong{strongly} recommended that you use this in any shared objects
18060 Despite the nomenclature, @code{default} always means public ie;
18061 available to be linked against from outside the shared object.
18062 @code{protected} and @code{internal} are pretty useless in real-world
18063 usage so the only other commonly used option will be @code{hidden}.
18064 The default if @option{-fvisibility} isn't specified is
18065 @code{default}, i.e., make every
18066 symbol public---this causes the same behavior as previous versions of
18069 A good explanation of the benefits offered by ensuring ELF
18070 symbols have the correct visibility is given by ``How To Write
18071 Shared Libraries'' by Ulrich Drepper (which can be found at
18072 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18073 solution made possible by this option to marking things hidden when
18074 the default is public is to make the default hidden and mark things
18075 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18076 and @code{__attribute__ ((visibility("default")))} instead of
18077 @code{__declspec(dllexport)} you get almost identical semantics with
18078 identical syntax. This is a great boon to those working with
18079 cross-platform projects.
18081 For those adding visibility support to existing code, you may find
18082 @samp{#pragma GCC visibility} of use. This works by you enclosing
18083 the declarations you wish to set visibility for with (for example)
18084 @samp{#pragma GCC visibility push(hidden)} and
18085 @samp{#pragma GCC visibility pop}.
18086 Bear in mind that symbol visibility should be viewed @strong{as
18087 part of the API interface contract} and thus all new code should
18088 always specify visibility when it is not the default ie; declarations
18089 only for use within the local DSO should @strong{always} be marked explicitly
18090 as hidden as so to avoid PLT indirection overheads---making this
18091 abundantly clear also aids readability and self-documentation of the code.
18092 Note that due to ISO C++ specification requirements, operator new and
18093 operator delete must always be of default visibility.
18095 Be aware that headers from outside your project, in particular system
18096 headers and headers from any other library you use, may not be
18097 expecting to be compiled with visibility other than the default. You
18098 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18099 before including any such headers.
18101 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18102 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18103 no modifications. However, this means that calls to @samp{extern}
18104 functions with no explicit visibility will use the PLT, so it is more
18105 effective to use @samp{__attribute ((visibility))} and/or
18106 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18107 declarations should be treated as hidden.
18109 Note that @samp{-fvisibility} does affect C++ vague linkage
18110 entities. This means that, for instance, an exception class that will
18111 be thrown between DSOs must be explicitly marked with default
18112 visibility so that the @samp{type_info} nodes will be unified between
18115 An overview of these techniques, their benefits and how to use them
18116 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18118 @item -fstrict-volatile-bitfields
18119 This option should be used if accesses to volatile bitfields (or other
18120 structure fields, although the compiler usually honors those types
18121 anyway) should use a single access in a mode of the same size as the
18122 container's type, aligned to a natural alignment if possible. For
18123 example, targets with memory-mapped peripheral registers might require
18124 all such accesses to be 16 bits wide; with this flag the user could
18125 declare all peripheral bitfields as ``unsigned short'' (assuming short
18126 is 16 bits on these targets) to force GCC to use 16 bit accesses
18127 instead of, perhaps, a more efficient 32 bit access.
18129 If this option is disabled, the compiler will use the most efficient
18130 instruction. In the previous example, that might be a 32-bit load
18131 instruction, even though that will access bytes that do not contain
18132 any portion of the bitfield, or memory-mapped registers unrelated to
18133 the one being updated.
18135 If the target requires strict alignment, and honoring the container
18136 type would require violating this alignment, a warning is issued.
18137 However, the access happens as the user requested, under the
18138 assumption that the user knows something about the target hardware
18139 that GCC is unaware of.
18141 The default value of this option is determined by the application binary
18142 interface for the target processor.
18148 @node Environment Variables
18149 @section Environment Variables Affecting GCC
18150 @cindex environment variables
18152 @c man begin ENVIRONMENT
18153 This section describes several environment variables that affect how GCC
18154 operates. Some of them work by specifying directories or prefixes to use
18155 when searching for various kinds of files. Some are used to specify other
18156 aspects of the compilation environment.
18158 Note that you can also specify places to search using options such as
18159 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18160 take precedence over places specified using environment variables, which
18161 in turn take precedence over those specified by the configuration of GCC@.
18162 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18163 GNU Compiler Collection (GCC) Internals}.
18168 @c @itemx LC_COLLATE
18170 @c @itemx LC_MONETARY
18171 @c @itemx LC_NUMERIC
18176 @c @findex LC_COLLATE
18177 @findex LC_MESSAGES
18178 @c @findex LC_MONETARY
18179 @c @findex LC_NUMERIC
18183 These environment variables control the way that GCC uses
18184 localization information that allow GCC to work with different
18185 national conventions. GCC inspects the locale categories
18186 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18187 so. These locale categories can be set to any value supported by your
18188 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18189 Kingdom encoded in UTF-8.
18191 The @env{LC_CTYPE} environment variable specifies character
18192 classification. GCC uses it to determine the character boundaries in
18193 a string; this is needed for some multibyte encodings that contain quote
18194 and escape characters that would otherwise be interpreted as a string
18197 The @env{LC_MESSAGES} environment variable specifies the language to
18198 use in diagnostic messages.
18200 If the @env{LC_ALL} environment variable is set, it overrides the value
18201 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18202 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18203 environment variable. If none of these variables are set, GCC
18204 defaults to traditional C English behavior.
18208 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18209 files. GCC uses temporary files to hold the output of one stage of
18210 compilation which is to be used as input to the next stage: for example,
18211 the output of the preprocessor, which is the input to the compiler
18214 @item GCC_EXEC_PREFIX
18215 @findex GCC_EXEC_PREFIX
18216 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18217 names of the subprograms executed by the compiler. No slash is added
18218 when this prefix is combined with the name of a subprogram, but you can
18219 specify a prefix that ends with a slash if you wish.
18221 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18222 an appropriate prefix to use based on the pathname it was invoked with.
18224 If GCC cannot find the subprogram using the specified prefix, it
18225 tries looking in the usual places for the subprogram.
18227 The default value of @env{GCC_EXEC_PREFIX} is
18228 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18229 the installed compiler. In many cases @var{prefix} is the value
18230 of @code{prefix} when you ran the @file{configure} script.
18232 Other prefixes specified with @option{-B} take precedence over this prefix.
18234 This prefix is also used for finding files such as @file{crt0.o} that are
18237 In addition, the prefix is used in an unusual way in finding the
18238 directories to search for header files. For each of the standard
18239 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18240 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18241 replacing that beginning with the specified prefix to produce an
18242 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18243 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18244 These alternate directories are searched first; the standard directories
18245 come next. If a standard directory begins with the configured
18246 @var{prefix} then the value of @var{prefix} is replaced by
18247 @env{GCC_EXEC_PREFIX} when looking for header files.
18249 @item COMPILER_PATH
18250 @findex COMPILER_PATH
18251 The value of @env{COMPILER_PATH} is a colon-separated list of
18252 directories, much like @env{PATH}. GCC tries the directories thus
18253 specified when searching for subprograms, if it can't find the
18254 subprograms using @env{GCC_EXEC_PREFIX}.
18257 @findex LIBRARY_PATH
18258 The value of @env{LIBRARY_PATH} is a colon-separated list of
18259 directories, much like @env{PATH}. When configured as a native compiler,
18260 GCC tries the directories thus specified when searching for special
18261 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18262 using GCC also uses these directories when searching for ordinary
18263 libraries for the @option{-l} option (but directories specified with
18264 @option{-L} come first).
18268 @cindex locale definition
18269 This variable is used to pass locale information to the compiler. One way in
18270 which this information is used is to determine the character set to be used
18271 when character literals, string literals and comments are parsed in C and C++.
18272 When the compiler is configured to allow multibyte characters,
18273 the following values for @env{LANG} are recognized:
18277 Recognize JIS characters.
18279 Recognize SJIS characters.
18281 Recognize EUCJP characters.
18284 If @env{LANG} is not defined, or if it has some other value, then the
18285 compiler will use mblen and mbtowc as defined by the default locale to
18286 recognize and translate multibyte characters.
18290 Some additional environments variables affect the behavior of the
18293 @include cppenv.texi
18297 @node Precompiled Headers
18298 @section Using Precompiled Headers
18299 @cindex precompiled headers
18300 @cindex speed of compilation
18302 Often large projects have many header files that are included in every
18303 source file. The time the compiler takes to process these header files
18304 over and over again can account for nearly all of the time required to
18305 build the project. To make builds faster, GCC allows users to
18306 `precompile' a header file; then, if builds can use the precompiled
18307 header file they will be much faster.
18309 To create a precompiled header file, simply compile it as you would any
18310 other file, if necessary using the @option{-x} option to make the driver
18311 treat it as a C or C++ header file. You will probably want to use a
18312 tool like @command{make} to keep the precompiled header up-to-date when
18313 the headers it contains change.
18315 A precompiled header file will be searched for when @code{#include} is
18316 seen in the compilation. As it searches for the included file
18317 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18318 compiler looks for a precompiled header in each directory just before it
18319 looks for the include file in that directory. The name searched for is
18320 the name specified in the @code{#include} with @samp{.gch} appended. If
18321 the precompiled header file can't be used, it is ignored.
18323 For instance, if you have @code{#include "all.h"}, and you have
18324 @file{all.h.gch} in the same directory as @file{all.h}, then the
18325 precompiled header file will be used if possible, and the original
18326 header will be used otherwise.
18328 Alternatively, you might decide to put the precompiled header file in a
18329 directory and use @option{-I} to ensure that directory is searched
18330 before (or instead of) the directory containing the original header.
18331 Then, if you want to check that the precompiled header file is always
18332 used, you can put a file of the same name as the original header in this
18333 directory containing an @code{#error} command.
18335 This also works with @option{-include}. So yet another way to use
18336 precompiled headers, good for projects not designed with precompiled
18337 header files in mind, is to simply take most of the header files used by
18338 a project, include them from another header file, precompile that header
18339 file, and @option{-include} the precompiled header. If the header files
18340 have guards against multiple inclusion, they will be skipped because
18341 they've already been included (in the precompiled header).
18343 If you need to precompile the same header file for different
18344 languages, targets, or compiler options, you can instead make a
18345 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18346 header in the directory, perhaps using @option{-o}. It doesn't matter
18347 what you call the files in the directory, every precompiled header in
18348 the directory will be considered. The first precompiled header
18349 encountered in the directory that is valid for this compilation will
18350 be used; they're searched in no particular order.
18352 There are many other possibilities, limited only by your imagination,
18353 good sense, and the constraints of your build system.
18355 A precompiled header file can be used only when these conditions apply:
18359 Only one precompiled header can be used in a particular compilation.
18362 A precompiled header can't be used once the first C token is seen. You
18363 can have preprocessor directives before a precompiled header; you can
18364 even include a precompiled header from inside another header, so long as
18365 there are no C tokens before the @code{#include}.
18368 The precompiled header file must be produced for the same language as
18369 the current compilation. You can't use a C precompiled header for a C++
18373 The precompiled header file must have been produced by the same compiler
18374 binary as the current compilation is using.
18377 Any macros defined before the precompiled header is included must
18378 either be defined in the same way as when the precompiled header was
18379 generated, or must not affect the precompiled header, which usually
18380 means that they don't appear in the precompiled header at all.
18382 The @option{-D} option is one way to define a macro before a
18383 precompiled header is included; using a @code{#define} can also do it.
18384 There are also some options that define macros implicitly, like
18385 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18388 @item If debugging information is output when using the precompiled
18389 header, using @option{-g} or similar, the same kind of debugging information
18390 must have been output when building the precompiled header. However,
18391 a precompiled header built using @option{-g} can be used in a compilation
18392 when no debugging information is being output.
18394 @item The same @option{-m} options must generally be used when building
18395 and using the precompiled header. @xref{Submodel Options},
18396 for any cases where this rule is relaxed.
18398 @item Each of the following options must be the same when building and using
18399 the precompiled header:
18401 @gccoptlist{-fexceptions}
18404 Some other command-line options starting with @option{-f},
18405 @option{-p}, or @option{-O} must be defined in the same way as when
18406 the precompiled header was generated. At present, it's not clear
18407 which options are safe to change and which are not; the safest choice
18408 is to use exactly the same options when generating and using the
18409 precompiled header. The following are known to be safe:
18411 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18412 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18413 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
18418 For all of these except the last, the compiler will automatically
18419 ignore the precompiled header if the conditions aren't met. If you
18420 find an option combination that doesn't work and doesn't cause the
18421 precompiled header to be ignored, please consider filing a bug report,
18424 If you do use differing options when generating and using the
18425 precompiled header, the actual behavior will be a mixture of the
18426 behavior for the options. For instance, if you use @option{-g} to
18427 generate the precompiled header but not when using it, you may or may
18428 not get debugging information for routines in the precompiled header.