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
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
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 1.2 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} -combine -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}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
311 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
312 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
313 -ftest-coverage -ftime-report -fvar-tracking @gol
314 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
315 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
316 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
317 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
318 -fdebug-prefix-map=@var{old}=@var{new} @gol
319 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
320 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
321 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
322 -print-multi-directory -print-multi-lib @gol
323 -print-prog-name=@var{program} -print-search-dirs -Q @gol
324 -print-sysroot -print-sysroot-headers-suffix @gol
325 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
327 @item Optimization Options
328 @xref{Optimize Options,,Options that Control Optimization}.
330 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
331 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
332 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
333 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
334 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
335 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
336 -fdata-sections -fdce -fdce @gol
337 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
338 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
339 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
340 -fforward-propagate -ffunction-sections @gol
341 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
342 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
343 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
344 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
345 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
346 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
347 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
348 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
349 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
350 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
351 -floop-parallelize-all @gol
352 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
353 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
354 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
355 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
356 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
357 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
358 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
359 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
360 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
361 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
362 -fprofile-generate=@var{path} @gol
363 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
364 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
365 -freorder-blocks-and-partition -freorder-functions @gol
366 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
367 -frounding-math -fsched2-use-superblocks @gol
368 -fsched2-use-traces -fsched-pressure @gol
369 -fsched-spec-load -fsched-spec-load-dangerous @gol
370 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
371 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
372 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
373 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
374 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
375 -fselective-scheduling -fselective-scheduling2 @gol
376 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
377 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
378 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
379 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
380 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
381 -ftree-copyrename -ftree-dce @gol
382 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
383 -ftree-phiprop -ftree-loop-distribution @gol
384 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
385 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
386 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
387 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
388 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
389 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
390 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
392 --param @var{name}=@var{value}
393 -O -O0 -O1 -O2 -O3 -Os}
395 @item Preprocessor Options
396 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
397 @gccoptlist{-A@var{question}=@var{answer} @gol
398 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
399 -C -dD -dI -dM -dN @gol
400 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
401 -idirafter @var{dir} @gol
402 -include @var{file} -imacros @var{file} @gol
403 -iprefix @var{file} -iwithprefix @var{dir} @gol
404 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
405 -imultilib @var{dir} -isysroot @var{dir} @gol
406 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
407 -P -fworking-directory -remap @gol
408 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
409 -Xpreprocessor @var{option}}
411 @item Assembler Option
412 @xref{Assembler Options,,Passing Options to the Assembler}.
413 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
416 @xref{Link Options,,Options for Linking}.
417 @gccoptlist{@var{object-file-name} -l@var{library} @gol
418 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
419 -s -static -static-libgcc -static-libstdc++ -shared @gol
420 -shared-libgcc -symbolic @gol
421 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
424 @item Directory Options
425 @xref{Directory Options,,Options for Directory Search}.
426 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
427 -specs=@var{file} -I- --sysroot=@var{dir}}
430 @c I wrote this xref this way to avoid overfull hbox. -- rms
431 @xref{Target Options}.
432 @gccoptlist{-V @var{version} -b @var{machine}}
434 @item Machine Dependent Options
435 @xref{Submodel Options,,Hardware Models and Configurations}.
436 @c This list is ordered alphanumerically by subsection name.
437 @c Try and put the significant identifier (CPU or system) first,
438 @c so users have a clue at guessing where the ones they want will be.
441 @gccoptlist{-EB -EL @gol
442 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
443 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
446 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
447 -mabi=@var{name} @gol
448 -mapcs-stack-check -mno-apcs-stack-check @gol
449 -mapcs-float -mno-apcs-float @gol
450 -mapcs-reentrant -mno-apcs-reentrant @gol
451 -msched-prolog -mno-sched-prolog @gol
452 -mlittle-endian -mbig-endian -mwords-little-endian @gol
453 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
454 -mfp16-format=@var{name}
455 -mthumb-interwork -mno-thumb-interwork @gol
456 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
457 -mstructure-size-boundary=@var{n} @gol
458 -mabort-on-noreturn @gol
459 -mlong-calls -mno-long-calls @gol
460 -msingle-pic-base -mno-single-pic-base @gol
461 -mpic-register=@var{reg} @gol
462 -mnop-fun-dllimport @gol
463 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
464 -mpoke-function-name @gol
466 -mtpcs-frame -mtpcs-leaf-frame @gol
467 -mcaller-super-interworking -mcallee-super-interworking @gol
469 -mword-relocations @gol
470 -mfix-cortex-m3-ldrd}
473 @gccoptlist{-mmcu=@var{mcu} -msize -mno-interrupts @gol
474 -mcall-prologues -mtiny-stack -mint8}
476 @emph{Blackfin Options}
477 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
478 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
479 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
480 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
481 -mno-id-shared-library -mshared-library-id=@var{n} @gol
482 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
483 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
484 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
488 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
489 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
490 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
491 -mstack-align -mdata-align -mconst-align @gol
492 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
493 -melf -maout -melinux -mlinux -sim -sim2 @gol
494 -mmul-bug-workaround -mno-mul-bug-workaround}
497 @gccoptlist{-mmac -mpush-args}
499 @emph{Darwin Options}
500 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
501 -arch_only -bind_at_load -bundle -bundle_loader @gol
502 -client_name -compatibility_version -current_version @gol
504 -dependency-file -dylib_file -dylinker_install_name @gol
505 -dynamic -dynamiclib -exported_symbols_list @gol
506 -filelist -flat_namespace -force_cpusubtype_ALL @gol
507 -force_flat_namespace -headerpad_max_install_names @gol
509 -image_base -init -install_name -keep_private_externs @gol
510 -multi_module -multiply_defined -multiply_defined_unused @gol
511 -noall_load -no_dead_strip_inits_and_terms @gol
512 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
513 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
514 -private_bundle -read_only_relocs -sectalign @gol
515 -sectobjectsymbols -whyload -seg1addr @gol
516 -sectcreate -sectobjectsymbols -sectorder @gol
517 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
518 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
519 -segprot -segs_read_only_addr -segs_read_write_addr @gol
520 -single_module -static -sub_library -sub_umbrella @gol
521 -twolevel_namespace -umbrella -undefined @gol
522 -unexported_symbols_list -weak_reference_mismatches @gol
523 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
524 -mkernel -mone-byte-bool}
526 @emph{DEC Alpha Options}
527 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
528 -mieee -mieee-with-inexact -mieee-conformant @gol
529 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
530 -mtrap-precision=@var{mode} -mbuild-constants @gol
531 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
532 -mbwx -mmax -mfix -mcix @gol
533 -mfloat-vax -mfloat-ieee @gol
534 -mexplicit-relocs -msmall-data -mlarge-data @gol
535 -msmall-text -mlarge-text @gol
536 -mmemory-latency=@var{time}}
538 @emph{DEC Alpha/VMS Options}
539 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
542 @gccoptlist{-msmall-model -mno-lsim}
545 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
546 -mhard-float -msoft-float @gol
547 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
548 -mdouble -mno-double @gol
549 -mmedia -mno-media -mmuladd -mno-muladd @gol
550 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
551 -mlinked-fp -mlong-calls -malign-labels @gol
552 -mlibrary-pic -macc-4 -macc-8 @gol
553 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
554 -moptimize-membar -mno-optimize-membar @gol
555 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
556 -mvliw-branch -mno-vliw-branch @gol
557 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
558 -mno-nested-cond-exec -mtomcat-stats @gol
562 @emph{GNU/Linux Options}
563 @gccoptlist{-muclibc}
565 @emph{H8/300 Options}
566 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
569 @gccoptlist{-march=@var{architecture-type} @gol
570 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
571 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
572 -mfixed-range=@var{register-range} @gol
573 -mjump-in-delay -mlinker-opt -mlong-calls @gol
574 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
575 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
576 -mno-jump-in-delay -mno-long-load-store @gol
577 -mno-portable-runtime -mno-soft-float @gol
578 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
579 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
580 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
581 -munix=@var{unix-std} -nolibdld -static -threads}
583 @emph{i386 and x86-64 Options}
584 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
585 -mfpmath=@var{unit} @gol
586 -masm=@var{dialect} -mno-fancy-math-387 @gol
587 -mno-fp-ret-in-387 -msoft-float @gol
588 -mno-wide-multiply -mrtd -malign-double @gol
589 -mpreferred-stack-boundary=@var{num}
590 -mincoming-stack-boundary=@var{num}
591 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
592 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
594 -msse4a -m3dnow -mpopcnt -mabm @gol
595 -mthreads -mno-align-stringops -minline-all-stringops @gol
596 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
597 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
598 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
599 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
600 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
601 -mcmodel=@var{code-model} -mabi=@var{name} @gol
602 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
606 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
607 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
608 -mconstant-gp -mauto-pic -mfused-madd @gol
609 -minline-float-divide-min-latency @gol
610 -minline-float-divide-max-throughput @gol
611 -mno-inline-float-divide @gol
612 -minline-int-divide-min-latency @gol
613 -minline-int-divide-max-throughput @gol
614 -mno-inline-int-divide @gol
615 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
616 -mno-inline-sqrt @gol
617 -mdwarf2-asm -mearly-stop-bits @gol
618 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
619 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
620 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
621 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
622 -msched-spec-ldc -msched-spec-control-ldc @gol
623 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
624 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
625 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
626 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
628 @emph{IA-64/VMS Options}
629 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
631 @emph{M32R/D Options}
632 @gccoptlist{-m32r2 -m32rx -m32r @gol
634 -malign-loops -mno-align-loops @gol
635 -missue-rate=@var{number} @gol
636 -mbranch-cost=@var{number} @gol
637 -mmodel=@var{code-size-model-type} @gol
638 -msdata=@var{sdata-type} @gol
639 -mno-flush-func -mflush-func=@var{name} @gol
640 -mno-flush-trap -mflush-trap=@var{number} @gol
644 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
646 @emph{M680x0 Options}
647 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
648 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
649 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
650 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
651 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
652 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
653 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
654 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
657 @emph{M68hc1x Options}
658 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
659 -mauto-incdec -minmax -mlong-calls -mshort @gol
660 -msoft-reg-count=@var{count}}
663 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
664 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
665 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
666 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
667 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
670 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
671 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
672 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
673 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
677 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
678 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
679 -mips64 -mips64r2 @gol
680 -mips16 -mno-mips16 -mflip-mips16 @gol
681 -minterlink-mips16 -mno-interlink-mips16 @gol
682 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
683 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
684 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
685 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
686 -mfpu=@var{fpu-type} @gol
687 -msmartmips -mno-smartmips @gol
688 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
689 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
690 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
691 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
692 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
693 -membedded-data -mno-embedded-data @gol
694 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
695 -mcode-readable=@var{setting} @gol
696 -msplit-addresses -mno-split-addresses @gol
697 -mexplicit-relocs -mno-explicit-relocs @gol
698 -mcheck-zero-division -mno-check-zero-division @gol
699 -mdivide-traps -mdivide-breaks @gol
700 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
701 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
702 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
703 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
704 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
705 -mflush-func=@var{func} -mno-flush-func @gol
706 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
707 -mfp-exceptions -mno-fp-exceptions @gol
708 -mvr4130-align -mno-vr4130-align -msynci -mno-synci}
711 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
712 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
713 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
714 -mno-base-addresses -msingle-exit -mno-single-exit}
716 @emph{MN10300 Options}
717 @gccoptlist{-mmult-bug -mno-mult-bug @gol
718 -mam33 -mno-am33 @gol
719 -mam33-2 -mno-am33-2 @gol
720 -mreturn-pointer-on-d0 @gol
723 @emph{PDP-11 Options}
724 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
725 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
726 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
727 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
728 -mbranch-expensive -mbranch-cheap @gol
729 -msplit -mno-split -munix-asm -mdec-asm}
731 @emph{picoChip Options}
732 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
733 -msymbol-as-address -mno-inefficient-warnings}
735 @emph{PowerPC Options}
736 See RS/6000 and PowerPC Options.
738 @emph{RS/6000 and PowerPC Options}
739 @gccoptlist{-mcpu=@var{cpu-type} @gol
740 -mtune=@var{cpu-type} @gol
741 -mpower -mno-power -mpower2 -mno-power2 @gol
742 -mpowerpc -mpowerpc64 -mno-powerpc @gol
743 -maltivec -mno-altivec @gol
744 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
745 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
746 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
747 -mfprnd -mno-fprnd @gol
748 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
749 -mnew-mnemonics -mold-mnemonics @gol
750 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
751 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
752 -malign-power -malign-natural @gol
753 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
754 -msingle-float -mdouble-float -msimple-fpu @gol
755 -mstring -mno-string -mupdate -mno-update @gol
756 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
757 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
758 -mstrict-align -mno-strict-align -mrelocatable @gol
759 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
760 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
761 -mdynamic-no-pic -maltivec -mswdiv @gol
762 -mprioritize-restricted-insns=@var{priority} @gol
763 -msched-costly-dep=@var{dependence_type} @gol
764 -minsert-sched-nops=@var{scheme} @gol
765 -mcall-sysv -mcall-netbsd @gol
766 -maix-struct-return -msvr4-struct-return @gol
767 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
768 -misel -mno-isel @gol
769 -misel=yes -misel=no @gol
771 -mspe=yes -mspe=no @gol
773 -mgen-cell-microcode -mwarn-cell-microcode @gol
774 -mvrsave -mno-vrsave @gol
775 -mmulhw -mno-mulhw @gol
776 -mdlmzb -mno-dlmzb @gol
777 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
778 -mprototype -mno-prototype @gol
779 -msim -mmvme -mads -myellowknife -memb -msdata @gol
780 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
782 @emph{S/390 and zSeries Options}
783 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
784 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
785 -mlong-double-64 -mlong-double-128 @gol
786 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
787 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
788 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
789 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
790 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
793 @gccoptlist{-meb -mel @gol
797 -mscore5 -mscore5u -mscore7 -mscore7d}
800 @gccoptlist{-m1 -m2 -m2e @gol
801 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
803 -m4-nofpu -m4-single-only -m4-single -m4 @gol
804 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
805 -m5-64media -m5-64media-nofpu @gol
806 -m5-32media -m5-32media-nofpu @gol
807 -m5-compact -m5-compact-nofpu @gol
808 -mb -ml -mdalign -mrelax @gol
809 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
810 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
811 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
812 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
813 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
817 @gccoptlist{-mcpu=@var{cpu-type} @gol
818 -mtune=@var{cpu-type} @gol
819 -mcmodel=@var{code-model} @gol
820 -m32 -m64 -mapp-regs -mno-app-regs @gol
821 -mfaster-structs -mno-faster-structs @gol
822 -mfpu -mno-fpu -mhard-float -msoft-float @gol
823 -mhard-quad-float -msoft-quad-float @gol
824 -mimpure-text -mno-impure-text -mlittle-endian @gol
825 -mstack-bias -mno-stack-bias @gol
826 -munaligned-doubles -mno-unaligned-doubles @gol
827 -mv8plus -mno-v8plus -mvis -mno-vis
828 -threads -pthreads -pthread}
831 @gccoptlist{-mwarn-reloc -merror-reloc @gol
832 -msafe-dma -munsafe-dma @gol
834 -msmall-mem -mlarge-mem -mstdmain @gol
835 -mfixed-range=@var{register-range}}
837 @emph{System V Options}
838 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
841 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
842 -mprolog-function -mno-prolog-function -mspace @gol
843 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
844 -mapp-regs -mno-app-regs @gol
845 -mdisable-callt -mno-disable-callt @gol
851 @gccoptlist{-mg -mgnu -munix}
853 @emph{VxWorks Options}
854 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
855 -Xbind-lazy -Xbind-now}
857 @emph{x86-64 Options}
858 See i386 and x86-64 Options.
860 @emph{i386 and x86-64 Windows Options}
861 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
862 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
864 @emph{Xstormy16 Options}
867 @emph{Xtensa Options}
868 @gccoptlist{-mconst16 -mno-const16 @gol
869 -mfused-madd -mno-fused-madd @gol
870 -mserialize-volatile -mno-serialize-volatile @gol
871 -mtext-section-literals -mno-text-section-literals @gol
872 -mtarget-align -mno-target-align @gol
873 -mlongcalls -mno-longcalls}
875 @emph{zSeries Options}
876 See S/390 and zSeries Options.
878 @item Code Generation Options
879 @xref{Code Gen Options,,Options for Code Generation Conventions}.
880 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
881 -ffixed-@var{reg} -fexceptions @gol
882 -fnon-call-exceptions -funwind-tables @gol
883 -fasynchronous-unwind-tables @gol
884 -finhibit-size-directive -finstrument-functions @gol
885 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
886 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
887 -fno-common -fno-ident @gol
888 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
889 -fno-jump-tables @gol
890 -frecord-gcc-switches @gol
891 -freg-struct-return -fshort-enums @gol
892 -fshort-double -fshort-wchar @gol
893 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
894 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
895 -fno-stack-limit -fargument-alias -fargument-noalias @gol
896 -fargument-noalias-global -fargument-noalias-anything @gol
897 -fleading-underscore -ftls-model=@var{model} @gol
898 -ftrapv -fwrapv -fbounds-check @gol
903 * Overall Options:: Controlling the kind of output:
904 an executable, object files, assembler files,
905 or preprocessed source.
906 * C Dialect Options:: Controlling the variant of C language compiled.
907 * C++ Dialect Options:: Variations on C++.
908 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
910 * Language Independent Options:: Controlling how diagnostics should be
912 * Warning Options:: How picky should the compiler be?
913 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
914 * Optimize Options:: How much optimization?
915 * Preprocessor Options:: Controlling header files and macro definitions.
916 Also, getting dependency information for Make.
917 * Assembler Options:: Passing options to the assembler.
918 * Link Options:: Specifying libraries and so on.
919 * Directory Options:: Where to find header files and libraries.
920 Where to find the compiler executable files.
921 * Spec Files:: How to pass switches to sub-processes.
922 * Target Options:: Running a cross-compiler, or an old version of GCC.
925 @node Overall Options
926 @section Options Controlling the Kind of Output
928 Compilation can involve up to four stages: preprocessing, compilation
929 proper, assembly and linking, always in that order. GCC is capable of
930 preprocessing and compiling several files either into several
931 assembler input files, or into one assembler input file; then each
932 assembler input file produces an object file, and linking combines all
933 the object files (those newly compiled, and those specified as input)
934 into an executable file.
936 @cindex file name suffix
937 For any given input file, the file name suffix determines what kind of
942 C source code which must be preprocessed.
945 C source code which should not be preprocessed.
948 C++ source code which should not be preprocessed.
951 Objective-C source code. Note that you must link with the @file{libobjc}
952 library to make an Objective-C program work.
955 Objective-C source code which should not be preprocessed.
959 Objective-C++ source code. Note that you must link with the @file{libobjc}
960 library to make an Objective-C++ program work. Note that @samp{.M} refers
961 to a literal capital M@.
964 Objective-C++ source code which should not be preprocessed.
967 C, C++, Objective-C or Objective-C++ header file to be turned into a
972 @itemx @var{file}.cxx
973 @itemx @var{file}.cpp
974 @itemx @var{file}.CPP
975 @itemx @var{file}.c++
977 C++ source code which must be preprocessed. Note that in @samp{.cxx},
978 the last two letters must both be literally @samp{x}. Likewise,
979 @samp{.C} refers to a literal capital C@.
983 Objective-C++ source code which must be preprocessed.
986 Objective-C++ source code which should not be preprocessed.
991 @itemx @var{file}.hxx
992 @itemx @var{file}.hpp
993 @itemx @var{file}.HPP
994 @itemx @var{file}.h++
995 @itemx @var{file}.tcc
996 C++ header file to be turned into a precompiled header.
999 @itemx @var{file}.for
1000 @itemx @var{file}.ftn
1001 Fixed form Fortran source code which should not be preprocessed.
1004 @itemx @var{file}.FOR
1005 @itemx @var{file}.fpp
1006 @itemx @var{file}.FPP
1007 @itemx @var{file}.FTN
1008 Fixed form Fortran source code which must be preprocessed (with the traditional
1011 @item @var{file}.f90
1012 @itemx @var{file}.f95
1013 @itemx @var{file}.f03
1014 @itemx @var{file}.f08
1015 Free form Fortran source code which should not be preprocessed.
1017 @item @var{file}.F90
1018 @itemx @var{file}.F95
1019 @itemx @var{file}.F03
1020 @itemx @var{file}.F08
1021 Free form Fortran source code which must be preprocessed (with the
1022 traditional preprocessor).
1024 @c FIXME: Descriptions of Java file types.
1030 @item @var{file}.ads
1031 Ada source code file which contains a library unit declaration (a
1032 declaration of a package, subprogram, or generic, or a generic
1033 instantiation), or a library unit renaming declaration (a package,
1034 generic, or subprogram renaming declaration). Such files are also
1037 @item @var{file}.adb
1038 Ada source code file containing a library unit body (a subprogram or
1039 package body). Such files are also called @dfn{bodies}.
1041 @c GCC also knows about some suffixes for languages not yet included:
1052 @itemx @var{file}.sx
1053 Assembler code which must be preprocessed.
1056 An object file to be fed straight into linking.
1057 Any file name with no recognized suffix is treated this way.
1061 You can specify the input language explicitly with the @option{-x} option:
1064 @item -x @var{language}
1065 Specify explicitly the @var{language} for the following input files
1066 (rather than letting the compiler choose a default based on the file
1067 name suffix). This option applies to all following input files until
1068 the next @option{-x} option. Possible values for @var{language} are:
1070 c c-header c-cpp-output
1071 c++ c++-header c++-cpp-output
1072 objective-c objective-c-header objective-c-cpp-output
1073 objective-c++ objective-c++-header objective-c++-cpp-output
1074 assembler assembler-with-cpp
1076 f77 f77-cpp-input f95 f95-cpp-input
1081 Turn off any specification of a language, so that subsequent files are
1082 handled according to their file name suffixes (as they are if @option{-x}
1083 has not been used at all).
1085 @item -pass-exit-codes
1086 @opindex pass-exit-codes
1087 Normally the @command{gcc} program will exit with the code of 1 if any
1088 phase of the compiler returns a non-success return code. If you specify
1089 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1090 numerically highest error produced by any phase that returned an error
1091 indication. The C, C++, and Fortran frontends return 4, if an internal
1092 compiler error is encountered.
1095 If you only want some of the stages of compilation, you can use
1096 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1097 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1098 @command{gcc} is to stop. Note that some combinations (for example,
1099 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1104 Compile or assemble the source files, but do not link. The linking
1105 stage simply is not done. The ultimate output is in the form of an
1106 object file for each source file.
1108 By default, the object file name for a source file is made by replacing
1109 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1111 Unrecognized input files, not requiring compilation or assembly, are
1116 Stop after the stage of compilation proper; do not assemble. The output
1117 is in the form of an assembler code file for each non-assembler input
1120 By default, the assembler file name for a source file is made by
1121 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1123 Input files that don't require compilation are ignored.
1127 Stop after the preprocessing stage; do not run the compiler proper. The
1128 output is in the form of preprocessed source code, which is sent to the
1131 Input files which don't require preprocessing are ignored.
1133 @cindex output file option
1136 Place output in file @var{file}. This applies regardless to whatever
1137 sort of output is being produced, whether it be an executable file,
1138 an object file, an assembler file or preprocessed C code.
1140 If @option{-o} is not specified, the default is to put an executable
1141 file in @file{a.out}, the object file for
1142 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1143 assembler file in @file{@var{source}.s}, a precompiled header file in
1144 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1149 Print (on standard error output) the commands executed to run the stages
1150 of compilation. Also print the version number of the compiler driver
1151 program and of the preprocessor and the compiler proper.
1155 Like @option{-v} except the commands are not executed and all command
1156 arguments are quoted. This is useful for shell scripts to capture the
1157 driver-generated command lines.
1161 Use pipes rather than temporary files for communication between the
1162 various stages of compilation. This fails to work on some systems where
1163 the assembler is unable to read from a pipe; but the GNU assembler has
1168 If you are compiling multiple source files, this option tells the driver
1169 to pass all the source files to the compiler at once (for those
1170 languages for which the compiler can handle this). This will allow
1171 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1172 language for which this is supported is C@. If you pass source files for
1173 multiple languages to the driver, using this option, the driver will invoke
1174 the compiler(s) that support IMA once each, passing each compiler all the
1175 source files appropriate for it. For those languages that do not support
1176 IMA this option will be ignored, and the compiler will be invoked once for
1177 each source file in that language. If you use this option in conjunction
1178 with @option{-save-temps}, the compiler will generate multiple
1180 (one for each source file), but only one (combined) @file{.o} or
1185 Print (on the standard output) a description of the command line options
1186 understood by @command{gcc}. If the @option{-v} option is also specified
1187 then @option{--help} will also be passed on to the various processes
1188 invoked by @command{gcc}, so that they can display the command line options
1189 they accept. If the @option{-Wextra} option has also been specified
1190 (prior to the @option{--help} option), then command line options which
1191 have no documentation associated with them will also be displayed.
1194 @opindex target-help
1195 Print (on the standard output) a description of target-specific command
1196 line options for each tool. For some targets extra target-specific
1197 information may also be printed.
1199 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1200 Print (on the standard output) a description of the command line
1201 options understood by the compiler that fit into all specified classes
1202 and qualifiers. These are the supported classes:
1205 @item @samp{optimizers}
1206 This will display all of the optimization options supported by the
1209 @item @samp{warnings}
1210 This will display all of the options controlling warning messages
1211 produced by the compiler.
1214 This will display target-specific options. Unlike the
1215 @option{--target-help} option however, target-specific options of the
1216 linker and assembler will not be displayed. This is because those
1217 tools do not currently support the extended @option{--help=} syntax.
1220 This will display the values recognized by the @option{--param}
1223 @item @var{language}
1224 This will display the options supported for @var{language}, where
1225 @var{language} is the name of one of the languages supported in this
1229 This will display the options that are common to all languages.
1232 These are the supported qualifiers:
1235 @item @samp{undocumented}
1236 Display only those options which are undocumented.
1239 Display options which take an argument that appears after an equal
1240 sign in the same continuous piece of text, such as:
1241 @samp{--help=target}.
1243 @item @samp{separate}
1244 Display options which take an argument that appears as a separate word
1245 following the original option, such as: @samp{-o output-file}.
1248 Thus for example to display all the undocumented target-specific
1249 switches supported by the compiler the following can be used:
1252 --help=target,undocumented
1255 The sense of a qualifier can be inverted by prefixing it with the
1256 @samp{^} character, so for example to display all binary warning
1257 options (i.e., ones that are either on or off and that do not take an
1258 argument), which have a description the following can be used:
1261 --help=warnings,^joined,^undocumented
1264 The argument to @option{--help=} should not consist solely of inverted
1267 Combining several classes is possible, although this usually
1268 restricts the output by so much that there is nothing to display. One
1269 case where it does work however is when one of the classes is
1270 @var{target}. So for example to display all the target-specific
1271 optimization options the following can be used:
1274 --help=target,optimizers
1277 The @option{--help=} option can be repeated on the command line. Each
1278 successive use will display its requested class of options, skipping
1279 those that have already been displayed.
1281 If the @option{-Q} option appears on the command line before the
1282 @option{--help=} option, then the descriptive text displayed by
1283 @option{--help=} is changed. Instead of describing the displayed
1284 options, an indication is given as to whether the option is enabled,
1285 disabled or set to a specific value (assuming that the compiler
1286 knows this at the point where the @option{--help=} option is used).
1288 Here is a truncated example from the ARM port of @command{gcc}:
1291 % gcc -Q -mabi=2 --help=target -c
1292 The following options are target specific:
1294 -mabort-on-noreturn [disabled]
1298 The output is sensitive to the effects of previous command line
1299 options, so for example it is possible to find out which optimizations
1300 are enabled at @option{-O2} by using:
1303 -Q -O2 --help=optimizers
1306 Alternatively you can discover which binary optimizations are enabled
1307 by @option{-O3} by using:
1310 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1311 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1312 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1315 @item -no-canonical-prefixes
1316 @opindex no-canonical-prefixes
1317 Do not expand any symbolic links, resolve references to @samp{/../}
1318 or @samp{/./}, or make the path absolute when generating a relative
1323 Display the version number and copyrights of the invoked GCC@.
1327 Invoke all subcommands under a wrapper program. It takes a single
1328 comma separated list as an argument, which will be used to invoke
1332 gcc -c t.c -wrapper gdb,--args
1335 This will invoke all subprograms of gcc under "gdb --args",
1336 thus cc1 invocation will be "gdb --args cc1 ...".
1338 @item -fplugin=@var{name}.so
1339 Load the plugin code in file @var{name}.so, assumed to be a
1340 shared object to be dlopen'd by the compiler. The base name of
1341 the shared object file is used to identify the plugin for the
1342 purposes of argument parsing (See
1343 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1344 Each plugin should define the callback functions specified in the
1347 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1348 Define an argument called @var{key} with a value of @var{value}
1349 for the plugin called @var{name}.
1351 @include @value{srcdir}/../libiberty/at-file.texi
1355 @section Compiling C++ Programs
1357 @cindex suffixes for C++ source
1358 @cindex C++ source file suffixes
1359 C++ source files conventionally use one of the suffixes @samp{.C},
1360 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1361 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1362 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1363 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1364 files with these names and compiles them as C++ programs even if you
1365 call the compiler the same way as for compiling C programs (usually
1366 with the name @command{gcc}).
1370 However, the use of @command{gcc} does not add the C++ library.
1371 @command{g++} is a program that calls GCC and treats @samp{.c},
1372 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1373 files unless @option{-x} is used, and automatically specifies linking
1374 against the C++ library. This program is also useful when
1375 precompiling a C header file with a @samp{.h} extension for use in C++
1376 compilations. On many systems, @command{g++} is also installed with
1377 the name @command{c++}.
1379 @cindex invoking @command{g++}
1380 When you compile C++ programs, you may specify many of the same
1381 command-line options that you use for compiling programs in any
1382 language; or command-line options meaningful for C and related
1383 languages; or options that are meaningful only for C++ programs.
1384 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1385 explanations of options for languages related to C@.
1386 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1387 explanations of options that are meaningful only for C++ programs.
1389 @node C Dialect Options
1390 @section Options Controlling C Dialect
1391 @cindex dialect options
1392 @cindex language dialect options
1393 @cindex options, dialect
1395 The following options control the dialect of C (or languages derived
1396 from C, such as C++, Objective-C and Objective-C++) that the compiler
1400 @cindex ANSI support
1404 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1405 equivalent to @samp{-std=c++98}.
1407 This turns off certain features of GCC that are incompatible with ISO
1408 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1409 such as the @code{asm} and @code{typeof} keywords, and
1410 predefined macros such as @code{unix} and @code{vax} that identify the
1411 type of system you are using. It also enables the undesirable and
1412 rarely used ISO trigraph feature. For the C compiler,
1413 it disables recognition of C++ style @samp{//} comments as well as
1414 the @code{inline} keyword.
1416 The alternate keywords @code{__asm__}, @code{__extension__},
1417 @code{__inline__} and @code{__typeof__} continue to work despite
1418 @option{-ansi}. You would not want to use them in an ISO C program, of
1419 course, but it is useful to put them in header files that might be included
1420 in compilations done with @option{-ansi}. Alternate predefined macros
1421 such as @code{__unix__} and @code{__vax__} are also available, with or
1422 without @option{-ansi}.
1424 The @option{-ansi} option does not cause non-ISO programs to be
1425 rejected gratuitously. For that, @option{-pedantic} is required in
1426 addition to @option{-ansi}. @xref{Warning Options}.
1428 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1429 option is used. Some header files may notice this macro and refrain
1430 from declaring certain functions or defining certain macros that the
1431 ISO standard doesn't call for; this is to avoid interfering with any
1432 programs that might use these names for other things.
1434 Functions that would normally be built in but do not have semantics
1435 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1436 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1437 built-in functions provided by GCC}, for details of the functions
1442 Determine the language standard. @xref{Standards,,Language Standards
1443 Supported by GCC}, for details of these standard versions. This option
1444 is currently only supported when compiling C or C++.
1446 The compiler can accept several base standards, such as @samp{c89} or
1447 @samp{c++98}, and GNU dialects of those standards, such as
1448 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1449 compiler will accept all programs following that standard and those
1450 using GNU extensions that do not contradict it. For example,
1451 @samp{-std=c89} turns off certain features of GCC that are
1452 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1453 keywords, but not other GNU extensions that do not have a meaning in
1454 ISO C90, such as omitting the middle term of a @code{?:}
1455 expression. On the other hand, by specifying a GNU dialect of a
1456 standard, all features the compiler support are enabled, even when
1457 those features change the meaning of the base standard and some
1458 strict-conforming programs may be rejected. The particular standard
1459 is used by @option{-pedantic} to identify which features are GNU
1460 extensions given that version of the standard. For example
1461 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1462 comments, while @samp{-std=gnu99 -pedantic} would not.
1464 A value for this option must be provided; possible values are
1469 Support all ISO C90 programs (certain GNU extensions that conflict
1470 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1472 @item iso9899:199409
1473 ISO C90 as modified in amendment 1.
1479 ISO C99. Note that this standard is not yet fully supported; see
1480 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1481 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1484 GNU dialect of ISO C90 (including some C99 features). This
1485 is the default for C code.
1489 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1490 this will become the default. The name @samp{gnu9x} is deprecated.
1493 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1497 GNU dialect of @option{-std=c++98}. This is the default for
1501 The working draft of the upcoming ISO C++0x standard. This option
1502 enables experimental features that are likely to be included in
1503 C++0x. The working draft is constantly changing, and any feature that is
1504 enabled by this flag may be removed from future versions of GCC if it is
1505 not part of the C++0x standard.
1508 GNU dialect of @option{-std=c++0x}. This option enables
1509 experimental features that may be removed in future versions of GCC.
1512 @item -fgnu89-inline
1513 @opindex fgnu89-inline
1514 The option @option{-fgnu89-inline} tells GCC to use the traditional
1515 GNU semantics for @code{inline} functions when in C99 mode.
1516 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1517 is accepted and ignored by GCC versions 4.1.3 up to but not including
1518 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1519 C99 mode. Using this option is roughly equivalent to adding the
1520 @code{gnu_inline} function attribute to all inline functions
1521 (@pxref{Function Attributes}).
1523 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1524 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1525 specifies the default behavior). This option was first supported in
1526 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1528 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1529 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1530 in effect for @code{inline} functions. @xref{Common Predefined
1531 Macros,,,cpp,The C Preprocessor}.
1533 @item -aux-info @var{filename}
1535 Output to the given filename prototyped declarations for all functions
1536 declared and/or defined in a translation unit, including those in header
1537 files. This option is silently ignored in any language other than C@.
1539 Besides declarations, the file indicates, in comments, the origin of
1540 each declaration (source file and line), whether the declaration was
1541 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1542 @samp{O} for old, respectively, in the first character after the line
1543 number and the colon), and whether it came from a declaration or a
1544 definition (@samp{C} or @samp{F}, respectively, in the following
1545 character). In the case of function definitions, a K&R-style list of
1546 arguments followed by their declarations is also provided, inside
1547 comments, after the declaration.
1551 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1552 keyword, so that code can use these words as identifiers. You can use
1553 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1554 instead. @option{-ansi} implies @option{-fno-asm}.
1556 In C++, this switch only affects the @code{typeof} keyword, since
1557 @code{asm} and @code{inline} are standard keywords. You may want to
1558 use the @option{-fno-gnu-keywords} flag instead, which has the same
1559 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1560 switch only affects the @code{asm} and @code{typeof} keywords, since
1561 @code{inline} is a standard keyword in ISO C99.
1564 @itemx -fno-builtin-@var{function}
1565 @opindex fno-builtin
1566 @cindex built-in functions
1567 Don't recognize built-in functions that do not begin with
1568 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1569 functions provided by GCC}, for details of the functions affected,
1570 including those which are not built-in functions when @option{-ansi} or
1571 @option{-std} options for strict ISO C conformance are used because they
1572 do not have an ISO standard meaning.
1574 GCC normally generates special code to handle certain built-in functions
1575 more efficiently; for instance, calls to @code{alloca} may become single
1576 instructions that adjust the stack directly, and calls to @code{memcpy}
1577 may become inline copy loops. The resulting code is often both smaller
1578 and faster, but since the function calls no longer appear as such, you
1579 cannot set a breakpoint on those calls, nor can you change the behavior
1580 of the functions by linking with a different library. In addition,
1581 when a function is recognized as a built-in function, GCC may use
1582 information about that function to warn about problems with calls to
1583 that function, or to generate more efficient code, even if the
1584 resulting code still contains calls to that function. For example,
1585 warnings are given with @option{-Wformat} for bad calls to
1586 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1587 known not to modify global memory.
1589 With the @option{-fno-builtin-@var{function}} option
1590 only the built-in function @var{function} is
1591 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1592 function is named that is not built-in in this version of GCC, this
1593 option is ignored. There is no corresponding
1594 @option{-fbuiltin-@var{function}} option; if you wish to enable
1595 built-in functions selectively when using @option{-fno-builtin} or
1596 @option{-ffreestanding}, you may define macros such as:
1599 #define abs(n) __builtin_abs ((n))
1600 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1605 @cindex hosted environment
1607 Assert that compilation takes place in a hosted environment. This implies
1608 @option{-fbuiltin}. A hosted environment is one in which the
1609 entire standard library is available, and in which @code{main} has a return
1610 type of @code{int}. Examples are nearly everything except a kernel.
1611 This is equivalent to @option{-fno-freestanding}.
1613 @item -ffreestanding
1614 @opindex ffreestanding
1615 @cindex hosted environment
1617 Assert that compilation takes place in a freestanding environment. This
1618 implies @option{-fno-builtin}. A freestanding environment
1619 is one in which the standard library may not exist, and program startup may
1620 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1621 This is equivalent to @option{-fno-hosted}.
1623 @xref{Standards,,Language Standards Supported by GCC}, for details of
1624 freestanding and hosted environments.
1628 @cindex openmp parallel
1629 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1630 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1631 compiler generates parallel code according to the OpenMP Application
1632 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1633 implies @option{-pthread}, and thus is only supported on targets that
1634 have support for @option{-pthread}.
1636 @item -fms-extensions
1637 @opindex fms-extensions
1638 Accept some non-standard constructs used in Microsoft header files.
1640 Some cases of unnamed fields in structures and unions are only
1641 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1642 fields within structs/unions}, for details.
1646 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1647 options for strict ISO C conformance) implies @option{-trigraphs}.
1649 @item -no-integrated-cpp
1650 @opindex no-integrated-cpp
1651 Performs a compilation in two passes: preprocessing and compiling. This
1652 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1653 @option{-B} option. The user supplied compilation step can then add in
1654 an additional preprocessing step after normal preprocessing but before
1655 compiling. The default is to use the integrated cpp (internal cpp)
1657 The semantics of this option will change if "cc1", "cc1plus", and
1658 "cc1obj" are merged.
1660 @cindex traditional C language
1661 @cindex C language, traditional
1663 @itemx -traditional-cpp
1664 @opindex traditional-cpp
1665 @opindex traditional
1666 Formerly, these options caused GCC to attempt to emulate a pre-standard
1667 C compiler. They are now only supported with the @option{-E} switch.
1668 The preprocessor continues to support a pre-standard mode. See the GNU
1669 CPP manual for details.
1671 @item -fcond-mismatch
1672 @opindex fcond-mismatch
1673 Allow conditional expressions with mismatched types in the second and
1674 third arguments. The value of such an expression is void. This option
1675 is not supported for C++.
1677 @item -flax-vector-conversions
1678 @opindex flax-vector-conversions
1679 Allow implicit conversions between vectors with differing numbers of
1680 elements and/or incompatible element types. This option should not be
1683 @item -funsigned-char
1684 @opindex funsigned-char
1685 Let the type @code{char} be unsigned, like @code{unsigned char}.
1687 Each kind of machine has a default for what @code{char} should
1688 be. It is either like @code{unsigned char} by default or like
1689 @code{signed char} by default.
1691 Ideally, a portable program should always use @code{signed char} or
1692 @code{unsigned char} when it depends on the signedness of an object.
1693 But many programs have been written to use plain @code{char} and
1694 expect it to be signed, or expect it to be unsigned, depending on the
1695 machines they were written for. This option, and its inverse, let you
1696 make such a program work with the opposite default.
1698 The type @code{char} is always a distinct type from each of
1699 @code{signed char} or @code{unsigned char}, even though its behavior
1700 is always just like one of those two.
1703 @opindex fsigned-char
1704 Let the type @code{char} be signed, like @code{signed char}.
1706 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1707 the negative form of @option{-funsigned-char}. Likewise, the option
1708 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1710 @item -fsigned-bitfields
1711 @itemx -funsigned-bitfields
1712 @itemx -fno-signed-bitfields
1713 @itemx -fno-unsigned-bitfields
1714 @opindex fsigned-bitfields
1715 @opindex funsigned-bitfields
1716 @opindex fno-signed-bitfields
1717 @opindex fno-unsigned-bitfields
1718 These options control whether a bit-field is signed or unsigned, when the
1719 declaration does not use either @code{signed} or @code{unsigned}. By
1720 default, such a bit-field is signed, because this is consistent: the
1721 basic integer types such as @code{int} are signed types.
1724 @node C++ Dialect Options
1725 @section Options Controlling C++ Dialect
1727 @cindex compiler options, C++
1728 @cindex C++ options, command line
1729 @cindex options, C++
1730 This section describes the command-line options that are only meaningful
1731 for C++ programs; but you can also use most of the GNU compiler options
1732 regardless of what language your program is in. For example, you
1733 might compile a file @code{firstClass.C} like this:
1736 g++ -g -frepo -O -c firstClass.C
1740 In this example, only @option{-frepo} is an option meant
1741 only for C++ programs; you can use the other options with any
1742 language supported by GCC@.
1744 Here is a list of options that are @emph{only} for compiling C++ programs:
1748 @item -fabi-version=@var{n}
1749 @opindex fabi-version
1750 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1751 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1752 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1753 the version that conforms most closely to the C++ ABI specification.
1754 Therefore, the ABI obtained using version 0 will change as ABI bugs
1757 The default is version 2.
1759 @item -fno-access-control
1760 @opindex fno-access-control
1761 Turn off all access checking. This switch is mainly useful for working
1762 around bugs in the access control code.
1766 Check that the pointer returned by @code{operator new} is non-null
1767 before attempting to modify the storage allocated. This check is
1768 normally unnecessary because the C++ standard specifies that
1769 @code{operator new} will only return @code{0} if it is declared
1770 @samp{throw()}, in which case the compiler will always check the
1771 return value even without this option. In all other cases, when
1772 @code{operator new} has a non-empty exception specification, memory
1773 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1774 @samp{new (nothrow)}.
1776 @item -fconserve-space
1777 @opindex fconserve-space
1778 Put uninitialized or runtime-initialized global variables into the
1779 common segment, as C does. This saves space in the executable at the
1780 cost of not diagnosing duplicate definitions. If you compile with this
1781 flag and your program mysteriously crashes after @code{main()} has
1782 completed, you may have an object that is being destroyed twice because
1783 two definitions were merged.
1785 This option is no longer useful on most targets, now that support has
1786 been added for putting variables into BSS without making them common.
1788 @item -ffriend-injection
1789 @opindex ffriend-injection
1790 Inject friend functions into the enclosing namespace, so that they are
1791 visible outside the scope of the class in which they are declared.
1792 Friend functions were documented to work this way in the old Annotated
1793 C++ Reference Manual, and versions of G++ before 4.1 always worked
1794 that way. However, in ISO C++ a friend function which is not declared
1795 in an enclosing scope can only be found using argument dependent
1796 lookup. This option causes friends to be injected as they were in
1799 This option is for compatibility, and may be removed in a future
1802 @item -fno-elide-constructors
1803 @opindex fno-elide-constructors
1804 The C++ standard allows an implementation to omit creating a temporary
1805 which is only used to initialize another object of the same type.
1806 Specifying this option disables that optimization, and forces G++ to
1807 call the copy constructor in all cases.
1809 @item -fno-enforce-eh-specs
1810 @opindex fno-enforce-eh-specs
1811 Don't generate code to check for violation of exception specifications
1812 at runtime. This option violates the C++ standard, but may be useful
1813 for reducing code size in production builds, much like defining
1814 @samp{NDEBUG}. This does not give user code permission to throw
1815 exceptions in violation of the exception specifications; the compiler
1816 will still optimize based on the specifications, so throwing an
1817 unexpected exception will result in undefined behavior.
1820 @itemx -fno-for-scope
1822 @opindex fno-for-scope
1823 If @option{-ffor-scope} is specified, the scope of variables declared in
1824 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1825 as specified by the C++ standard.
1826 If @option{-fno-for-scope} is specified, the scope of variables declared in
1827 a @i{for-init-statement} extends to the end of the enclosing scope,
1828 as was the case in old versions of G++, and other (traditional)
1829 implementations of C++.
1831 The default if neither flag is given to follow the standard,
1832 but to allow and give a warning for old-style code that would
1833 otherwise be invalid, or have different behavior.
1835 @item -fno-gnu-keywords
1836 @opindex fno-gnu-keywords
1837 Do not recognize @code{typeof} as a keyword, so that code can use this
1838 word as an identifier. You can use the keyword @code{__typeof__} instead.
1839 @option{-ansi} implies @option{-fno-gnu-keywords}.
1841 @item -fno-implicit-templates
1842 @opindex fno-implicit-templates
1843 Never emit code for non-inline templates which are instantiated
1844 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1845 @xref{Template Instantiation}, for more information.
1847 @item -fno-implicit-inline-templates
1848 @opindex fno-implicit-inline-templates
1849 Don't emit code for implicit instantiations of inline templates, either.
1850 The default is to handle inlines differently so that compiles with and
1851 without optimization will need the same set of explicit instantiations.
1853 @item -fno-implement-inlines
1854 @opindex fno-implement-inlines
1855 To save space, do not emit out-of-line copies of inline functions
1856 controlled by @samp{#pragma implementation}. This will cause linker
1857 errors if these functions are not inlined everywhere they are called.
1859 @item -fms-extensions
1860 @opindex fms-extensions
1861 Disable pedantic warnings about constructs used in MFC, such as implicit
1862 int and getting a pointer to member function via non-standard syntax.
1864 @item -fno-nonansi-builtins
1865 @opindex fno-nonansi-builtins
1866 Disable built-in declarations of functions that are not mandated by
1867 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1868 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1870 @item -fno-operator-names
1871 @opindex fno-operator-names
1872 Do not treat the operator name keywords @code{and}, @code{bitand},
1873 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1874 synonyms as keywords.
1876 @item -fno-optional-diags
1877 @opindex fno-optional-diags
1878 Disable diagnostics that the standard says a compiler does not need to
1879 issue. Currently, the only such diagnostic issued by G++ is the one for
1880 a name having multiple meanings within a class.
1883 @opindex fpermissive
1884 Downgrade some diagnostics about nonconformant code from errors to
1885 warnings. Thus, using @option{-fpermissive} will allow some
1886 nonconforming code to compile.
1888 @item -fno-pretty-templates
1889 @opindex fno-pretty-templates
1890 When an error message refers to a specialization of a function
1891 template, the compiler will normally print the signature of the
1892 template followed by the template arguments and any typedefs or
1893 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1894 rather than @code{void f(int)}) so that it's clear which template is
1895 involved. When an error message refers to a specialization of a class
1896 template, the compiler will omit any template arguments which match
1897 the default template arguments for that template. If either of these
1898 behaviors make it harder to understand the error message rather than
1899 easier, using @option{-fno-pretty-templates} will disable them.
1903 Enable automatic template instantiation at link time. This option also
1904 implies @option{-fno-implicit-templates}. @xref{Template
1905 Instantiation}, for more information.
1909 Disable generation of information about every class with virtual
1910 functions for use by the C++ runtime type identification features
1911 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1912 of the language, you can save some space by using this flag. Note that
1913 exception handling uses the same information, but it will generate it as
1914 needed. The @samp{dynamic_cast} operator can still be used for casts that
1915 do not require runtime type information, i.e.@: casts to @code{void *} or to
1916 unambiguous base classes.
1920 Emit statistics about front-end processing at the end of the compilation.
1921 This information is generally only useful to the G++ development team.
1923 @item -ftemplate-depth-@var{n}
1924 @opindex ftemplate-depth
1925 Set the maximum instantiation depth for template classes to @var{n}.
1926 A limit on the template instantiation depth is needed to detect
1927 endless recursions during template class instantiation. ANSI/ISO C++
1928 conforming programs must not rely on a maximum depth greater than 17
1929 (changed to 1024 in C++0x).
1931 @item -fno-threadsafe-statics
1932 @opindex fno-threadsafe-statics
1933 Do not emit the extra code to use the routines specified in the C++
1934 ABI for thread-safe initialization of local statics. You can use this
1935 option to reduce code size slightly in code that doesn't need to be
1938 @item -fuse-cxa-atexit
1939 @opindex fuse-cxa-atexit
1940 Register destructors for objects with static storage duration with the
1941 @code{__cxa_atexit} function rather than the @code{atexit} function.
1942 This option is required for fully standards-compliant handling of static
1943 destructors, but will only work if your C library supports
1944 @code{__cxa_atexit}.
1946 @item -fno-use-cxa-get-exception-ptr
1947 @opindex fno-use-cxa-get-exception-ptr
1948 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1949 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1950 if the runtime routine is not available.
1952 @item -fvisibility-inlines-hidden
1953 @opindex fvisibility-inlines-hidden
1954 This switch declares that the user does not attempt to compare
1955 pointers to inline methods where the addresses of the two functions
1956 were taken in different shared objects.
1958 The effect of this is that GCC may, effectively, mark inline methods with
1959 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1960 appear in the export table of a DSO and do not require a PLT indirection
1961 when used within the DSO@. Enabling this option can have a dramatic effect
1962 on load and link times of a DSO as it massively reduces the size of the
1963 dynamic export table when the library makes heavy use of templates.
1965 The behavior of this switch is not quite the same as marking the
1966 methods as hidden directly, because it does not affect static variables
1967 local to the function or cause the compiler to deduce that
1968 the function is defined in only one shared object.
1970 You may mark a method as having a visibility explicitly to negate the
1971 effect of the switch for that method. For example, if you do want to
1972 compare pointers to a particular inline method, you might mark it as
1973 having default visibility. Marking the enclosing class with explicit
1974 visibility will have no effect.
1976 Explicitly instantiated inline methods are unaffected by this option
1977 as their linkage might otherwise cross a shared library boundary.
1978 @xref{Template Instantiation}.
1980 @item -fvisibility-ms-compat
1981 @opindex fvisibility-ms-compat
1982 This flag attempts to use visibility settings to make GCC's C++
1983 linkage model compatible with that of Microsoft Visual Studio.
1985 The flag makes these changes to GCC's linkage model:
1989 It sets the default visibility to @code{hidden}, like
1990 @option{-fvisibility=hidden}.
1993 Types, but not their members, are not hidden by default.
1996 The One Definition Rule is relaxed for types without explicit
1997 visibility specifications which are defined in more than one different
1998 shared object: those declarations are permitted if they would have
1999 been permitted when this option was not used.
2002 In new code it is better to use @option{-fvisibility=hidden} and
2003 export those classes which are intended to be externally visible.
2004 Unfortunately it is possible for code to rely, perhaps accidentally,
2005 on the Visual Studio behavior.
2007 Among the consequences of these changes are that static data members
2008 of the same type with the same name but defined in different shared
2009 objects will be different, so changing one will not change the other;
2010 and that pointers to function members defined in different shared
2011 objects may not compare equal. When this flag is given, it is a
2012 violation of the ODR to define types with the same name differently.
2016 Do not use weak symbol support, even if it is provided by the linker.
2017 By default, G++ will use weak symbols if they are available. This
2018 option exists only for testing, and should not be used by end-users;
2019 it will result in inferior code and has no benefits. This option may
2020 be removed in a future release of G++.
2024 Do not search for header files in the standard directories specific to
2025 C++, but do still search the other standard directories. (This option
2026 is used when building the C++ library.)
2029 In addition, these optimization, warning, and code generation options
2030 have meanings only for C++ programs:
2033 @item -fno-default-inline
2034 @opindex fno-default-inline
2035 Do not assume @samp{inline} for functions defined inside a class scope.
2036 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2037 functions will have linkage like inline functions; they just won't be
2040 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2043 Warn when G++ generates code that is probably not compatible with the
2044 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2045 all such cases, there are probably some cases that are not warned about,
2046 even though G++ is generating incompatible code. There may also be
2047 cases where warnings are emitted even though the code that is generated
2050 You should rewrite your code to avoid these warnings if you are
2051 concerned about the fact that code generated by G++ may not be binary
2052 compatible with code generated by other compilers.
2054 The known incompatibilities at this point include:
2059 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2060 pack data into the same byte as a base class. For example:
2063 struct A @{ virtual void f(); int f1 : 1; @};
2064 struct B : public A @{ int f2 : 1; @};
2068 In this case, G++ will place @code{B::f2} into the same byte
2069 as@code{A::f1}; other compilers will not. You can avoid this problem
2070 by explicitly padding @code{A} so that its size is a multiple of the
2071 byte size on your platform; that will cause G++ and other compilers to
2072 layout @code{B} identically.
2075 Incorrect handling of tail-padding for virtual bases. G++ does not use
2076 tail padding when laying out virtual bases. For example:
2079 struct A @{ virtual void f(); char c1; @};
2080 struct B @{ B(); char c2; @};
2081 struct C : public A, public virtual B @{@};
2085 In this case, G++ will not place @code{B} into the tail-padding for
2086 @code{A}; other compilers will. You can avoid this problem by
2087 explicitly padding @code{A} so that its size is a multiple of its
2088 alignment (ignoring virtual base classes); that will cause G++ and other
2089 compilers to layout @code{C} identically.
2092 Incorrect handling of bit-fields with declared widths greater than that
2093 of their underlying types, when the bit-fields appear in a union. For
2097 union U @{ int i : 4096; @};
2101 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2102 union too small by the number of bits in an @code{int}.
2105 Empty classes can be placed at incorrect offsets. For example:
2115 struct C : public B, public A @{@};
2119 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2120 it should be placed at offset zero. G++ mistakenly believes that the
2121 @code{A} data member of @code{B} is already at offset zero.
2124 Names of template functions whose types involve @code{typename} or
2125 template template parameters can be mangled incorrectly.
2128 template <typename Q>
2129 void f(typename Q::X) @{@}
2131 template <template <typename> class Q>
2132 void f(typename Q<int>::X) @{@}
2136 Instantiations of these templates may be mangled incorrectly.
2140 It also warns psABI related changes. The known psABI changes at this
2146 For SYSV/x86-64, when passing union with long double, it is changed to
2147 pass in memory as specified in psABI. For example:
2157 @code{union U} will always be passed in memory.
2161 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2162 @opindex Wctor-dtor-privacy
2163 @opindex Wno-ctor-dtor-privacy
2164 Warn when a class seems unusable because all the constructors or
2165 destructors in that class are private, and it has neither friends nor
2166 public static member functions.
2168 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2169 @opindex Wnon-virtual-dtor
2170 @opindex Wno-non-virtual-dtor
2171 Warn when a class has virtual functions and accessible non-virtual
2172 destructor, in which case it would be possible but unsafe to delete
2173 an instance of a derived class through a pointer to the base class.
2174 This warning is also enabled if -Weffc++ is specified.
2176 @item -Wreorder @r{(C++ and Objective-C++ only)}
2178 @opindex Wno-reorder
2179 @cindex reordering, warning
2180 @cindex warning for reordering of member initializers
2181 Warn when the order of member initializers given in the code does not
2182 match the order in which they must be executed. For instance:
2188 A(): j (0), i (1) @{ @}
2192 The compiler will rearrange the member initializers for @samp{i}
2193 and @samp{j} to match the declaration order of the members, emitting
2194 a warning to that effect. This warning is enabled by @option{-Wall}.
2197 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2200 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2203 Warn about violations of the following style guidelines from Scott Meyers'
2204 @cite{Effective C++} book:
2208 Item 11: Define a copy constructor and an assignment operator for classes
2209 with dynamically allocated memory.
2212 Item 12: Prefer initialization to assignment in constructors.
2215 Item 14: Make destructors virtual in base classes.
2218 Item 15: Have @code{operator=} return a reference to @code{*this}.
2221 Item 23: Don't try to return a reference when you must return an object.
2225 Also warn about violations of the following style guidelines from
2226 Scott Meyers' @cite{More Effective C++} book:
2230 Item 6: Distinguish between prefix and postfix forms of increment and
2231 decrement operators.
2234 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2238 When selecting this option, be aware that the standard library
2239 headers do not obey all of these guidelines; use @samp{grep -v}
2240 to filter out those warnings.
2242 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2243 @opindex Wstrict-null-sentinel
2244 @opindex Wno-strict-null-sentinel
2245 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2246 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2247 to @code{__null}. Although it is a null pointer constant not a null pointer,
2248 it is guaranteed to be of the same size as a pointer. But this use is
2249 not portable across different compilers.
2251 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2252 @opindex Wno-non-template-friend
2253 @opindex Wnon-template-friend
2254 Disable warnings when non-templatized friend functions are declared
2255 within a template. Since the advent of explicit template specification
2256 support in G++, if the name of the friend is an unqualified-id (i.e.,
2257 @samp{friend foo(int)}), the C++ language specification demands that the
2258 friend declare or define an ordinary, nontemplate function. (Section
2259 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2260 could be interpreted as a particular specialization of a templatized
2261 function. Because this non-conforming behavior is no longer the default
2262 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2263 check existing code for potential trouble spots and is on by default.
2264 This new compiler behavior can be turned off with
2265 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2266 but disables the helpful warning.
2268 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2269 @opindex Wold-style-cast
2270 @opindex Wno-old-style-cast
2271 Warn if an old-style (C-style) cast to a non-void type is used within
2272 a C++ program. The new-style casts (@samp{dynamic_cast},
2273 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2274 less vulnerable to unintended effects and much easier to search for.
2276 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2277 @opindex Woverloaded-virtual
2278 @opindex Wno-overloaded-virtual
2279 @cindex overloaded virtual fn, warning
2280 @cindex warning for overloaded virtual fn
2281 Warn when a function declaration hides virtual functions from a
2282 base class. For example, in:
2289 struct B: public A @{
2294 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2302 will fail to compile.
2304 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2305 @opindex Wno-pmf-conversions
2306 @opindex Wpmf-conversions
2307 Disable the diagnostic for converting a bound pointer to member function
2310 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2311 @opindex Wsign-promo
2312 @opindex Wno-sign-promo
2313 Warn when overload resolution chooses a promotion from unsigned or
2314 enumerated type to a signed type, over a conversion to an unsigned type of
2315 the same size. Previous versions of G++ would try to preserve
2316 unsignedness, but the standard mandates the current behavior.
2321 A& operator = (int);
2331 In this example, G++ will synthesize a default @samp{A& operator =
2332 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2335 @node Objective-C and Objective-C++ Dialect Options
2336 @section Options Controlling Objective-C and Objective-C++ Dialects
2338 @cindex compiler options, Objective-C and Objective-C++
2339 @cindex Objective-C and Objective-C++ options, command line
2340 @cindex options, Objective-C and Objective-C++
2341 (NOTE: This manual does not describe the Objective-C and Objective-C++
2342 languages themselves. See @xref{Standards,,Language Standards
2343 Supported by GCC}, for references.)
2345 This section describes the command-line options that are only meaningful
2346 for Objective-C and Objective-C++ programs, but you can also use most of
2347 the language-independent GNU compiler options.
2348 For example, you might compile a file @code{some_class.m} like this:
2351 gcc -g -fgnu-runtime -O -c some_class.m
2355 In this example, @option{-fgnu-runtime} is an option meant only for
2356 Objective-C and Objective-C++ programs; you can use the other options with
2357 any language supported by GCC@.
2359 Note that since Objective-C is an extension of the C language, Objective-C
2360 compilations may also use options specific to the C front-end (e.g.,
2361 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2362 C++-specific options (e.g., @option{-Wabi}).
2364 Here is a list of options that are @emph{only} for compiling Objective-C
2365 and Objective-C++ programs:
2368 @item -fconstant-string-class=@var{class-name}
2369 @opindex fconstant-string-class
2370 Use @var{class-name} as the name of the class to instantiate for each
2371 literal string specified with the syntax @code{@@"@dots{}"}. The default
2372 class name is @code{NXConstantString} if the GNU runtime is being used, and
2373 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2374 @option{-fconstant-cfstrings} option, if also present, will override the
2375 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2376 to be laid out as constant CoreFoundation strings.
2379 @opindex fgnu-runtime
2380 Generate object code compatible with the standard GNU Objective-C
2381 runtime. This is the default for most types of systems.
2383 @item -fnext-runtime
2384 @opindex fnext-runtime
2385 Generate output compatible with the NeXT runtime. This is the default
2386 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2387 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2390 @item -fno-nil-receivers
2391 @opindex fno-nil-receivers
2392 Assume that all Objective-C message dispatches (e.g.,
2393 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2394 is not @code{nil}. This allows for more efficient entry points in the runtime
2395 to be used. Currently, this option is only available in conjunction with
2396 the NeXT runtime on Mac OS X 10.3 and later.
2398 @item -fobjc-call-cxx-cdtors
2399 @opindex fobjc-call-cxx-cdtors
2400 For each Objective-C class, check if any of its instance variables is a
2401 C++ object with a non-trivial default constructor. If so, synthesize a
2402 special @code{- (id) .cxx_construct} instance method that will run
2403 non-trivial default constructors on any such instance variables, in order,
2404 and then return @code{self}. Similarly, check if any instance variable
2405 is a C++ object with a non-trivial destructor, and if so, synthesize a
2406 special @code{- (void) .cxx_destruct} method that will run
2407 all such default destructors, in reverse order.
2409 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2410 thusly generated will only operate on instance variables declared in the
2411 current Objective-C class, and not those inherited from superclasses. It
2412 is the responsibility of the Objective-C runtime to invoke all such methods
2413 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2414 will be invoked by the runtime immediately after a new object
2415 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2416 be invoked immediately before the runtime deallocates an object instance.
2418 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2419 support for invoking the @code{- (id) .cxx_construct} and
2420 @code{- (void) .cxx_destruct} methods.
2422 @item -fobjc-direct-dispatch
2423 @opindex fobjc-direct-dispatch
2424 Allow fast jumps to the message dispatcher. On Darwin this is
2425 accomplished via the comm page.
2427 @item -fobjc-exceptions
2428 @opindex fobjc-exceptions
2429 Enable syntactic support for structured exception handling in Objective-C,
2430 similar to what is offered by C++ and Java. This option is
2431 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2440 @@catch (AnObjCClass *exc) @{
2447 @@catch (AnotherClass *exc) @{
2450 @@catch (id allOthers) @{
2460 The @code{@@throw} statement may appear anywhere in an Objective-C or
2461 Objective-C++ program; when used inside of a @code{@@catch} block, the
2462 @code{@@throw} may appear without an argument (as shown above), in which case
2463 the object caught by the @code{@@catch} will be rethrown.
2465 Note that only (pointers to) Objective-C objects may be thrown and
2466 caught using this scheme. When an object is thrown, it will be caught
2467 by the nearest @code{@@catch} clause capable of handling objects of that type,
2468 analogously to how @code{catch} blocks work in C++ and Java. A
2469 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2470 any and all Objective-C exceptions not caught by previous @code{@@catch}
2473 The @code{@@finally} clause, if present, will be executed upon exit from the
2474 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2475 regardless of whether any exceptions are thrown, caught or rethrown
2476 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2477 of the @code{finally} clause in Java.
2479 There are several caveats to using the new exception mechanism:
2483 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2484 idioms provided by the @code{NSException} class, the new
2485 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2486 systems, due to additional functionality needed in the (NeXT) Objective-C
2490 As mentioned above, the new exceptions do not support handling
2491 types other than Objective-C objects. Furthermore, when used from
2492 Objective-C++, the Objective-C exception model does not interoperate with C++
2493 exceptions at this time. This means you cannot @code{@@throw} an exception
2494 from Objective-C and @code{catch} it in C++, or vice versa
2495 (i.e., @code{throw @dots{} @@catch}).
2498 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2499 blocks for thread-safe execution:
2502 @@synchronized (ObjCClass *guard) @{
2507 Upon entering the @code{@@synchronized} block, a thread of execution shall
2508 first check whether a lock has been placed on the corresponding @code{guard}
2509 object by another thread. If it has, the current thread shall wait until
2510 the other thread relinquishes its lock. Once @code{guard} becomes available,
2511 the current thread will place its own lock on it, execute the code contained in
2512 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2513 making @code{guard} available to other threads).
2515 Unlike Java, Objective-C does not allow for entire methods to be marked
2516 @code{@@synchronized}. Note that throwing exceptions out of
2517 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2518 to be unlocked properly.
2522 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2524 @item -freplace-objc-classes
2525 @opindex freplace-objc-classes
2526 Emit a special marker instructing @command{ld(1)} not to statically link in
2527 the resulting object file, and allow @command{dyld(1)} to load it in at
2528 run time instead. This is used in conjunction with the Fix-and-Continue
2529 debugging mode, where the object file in question may be recompiled and
2530 dynamically reloaded in the course of program execution, without the need
2531 to restart the program itself. Currently, Fix-and-Continue functionality
2532 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2537 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2538 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2539 compile time) with static class references that get initialized at load time,
2540 which improves run-time performance. Specifying the @option{-fzero-link} flag
2541 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2542 to be retained. This is useful in Zero-Link debugging mode, since it allows
2543 for individual class implementations to be modified during program execution.
2547 Dump interface declarations for all classes seen in the source file to a
2548 file named @file{@var{sourcename}.decl}.
2550 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2551 @opindex Wassign-intercept
2552 @opindex Wno-assign-intercept
2553 Warn whenever an Objective-C assignment is being intercepted by the
2556 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2557 @opindex Wno-protocol
2559 If a class is declared to implement a protocol, a warning is issued for
2560 every method in the protocol that is not implemented by the class. The
2561 default behavior is to issue a warning for every method not explicitly
2562 implemented in the class, even if a method implementation is inherited
2563 from the superclass. If you use the @option{-Wno-protocol} option, then
2564 methods inherited from the superclass are considered to be implemented,
2565 and no warning is issued for them.
2567 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2569 @opindex Wno-selector
2570 Warn if multiple methods of different types for the same selector are
2571 found during compilation. The check is performed on the list of methods
2572 in the final stage of compilation. Additionally, a check is performed
2573 for each selector appearing in a @code{@@selector(@dots{})}
2574 expression, and a corresponding method for that selector has been found
2575 during compilation. Because these checks scan the method table only at
2576 the end of compilation, these warnings are not produced if the final
2577 stage of compilation is not reached, for example because an error is
2578 found during compilation, or because the @option{-fsyntax-only} option is
2581 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2582 @opindex Wstrict-selector-match
2583 @opindex Wno-strict-selector-match
2584 Warn if multiple methods with differing argument and/or return types are
2585 found for a given selector when attempting to send a message using this
2586 selector to a receiver of type @code{id} or @code{Class}. When this flag
2587 is off (which is the default behavior), the compiler will omit such warnings
2588 if any differences found are confined to types which share the same size
2591 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2592 @opindex Wundeclared-selector
2593 @opindex Wno-undeclared-selector
2594 Warn if a @code{@@selector(@dots{})} expression referring to an
2595 undeclared selector is found. A selector is considered undeclared if no
2596 method with that name has been declared before the
2597 @code{@@selector(@dots{})} expression, either explicitly in an
2598 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2599 an @code{@@implementation} section. This option always performs its
2600 checks as soon as a @code{@@selector(@dots{})} expression is found,
2601 while @option{-Wselector} only performs its checks in the final stage of
2602 compilation. This also enforces the coding style convention
2603 that methods and selectors must be declared before being used.
2605 @item -print-objc-runtime-info
2606 @opindex print-objc-runtime-info
2607 Generate C header describing the largest structure that is passed by
2612 @node Language Independent Options
2613 @section Options to Control Diagnostic Messages Formatting
2614 @cindex options to control diagnostics formatting
2615 @cindex diagnostic messages
2616 @cindex message formatting
2618 Traditionally, diagnostic messages have been formatted irrespective of
2619 the output device's aspect (e.g.@: its width, @dots{}). The options described
2620 below can be used to control the diagnostic messages formatting
2621 algorithm, e.g.@: how many characters per line, how often source location
2622 information should be reported. Right now, only the C++ front end can
2623 honor these options. However it is expected, in the near future, that
2624 the remaining front ends would be able to digest them correctly.
2627 @item -fmessage-length=@var{n}
2628 @opindex fmessage-length
2629 Try to format error messages so that they fit on lines of about @var{n}
2630 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2631 the front ends supported by GCC@. If @var{n} is zero, then no
2632 line-wrapping will be done; each error message will appear on a single
2635 @opindex fdiagnostics-show-location
2636 @item -fdiagnostics-show-location=once
2637 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2638 reporter to emit @emph{once} source location information; that is, in
2639 case the message is too long to fit on a single physical line and has to
2640 be wrapped, the source location won't be emitted (as prefix) again,
2641 over and over, in subsequent continuation lines. This is the default
2644 @item -fdiagnostics-show-location=every-line
2645 Only meaningful in line-wrapping mode. Instructs the diagnostic
2646 messages reporter to emit the same source location information (as
2647 prefix) for physical lines that result from the process of breaking
2648 a message which is too long to fit on a single line.
2650 @item -fdiagnostics-show-option
2651 @opindex fdiagnostics-show-option
2652 This option instructs the diagnostic machinery to add text to each
2653 diagnostic emitted, which indicates which command line option directly
2654 controls that diagnostic, when such an option is known to the
2655 diagnostic machinery.
2657 @item -Wcoverage-mismatch
2658 @opindex Wcoverage-mismatch
2659 Warn if feedback profiles do not match when using the
2660 @option{-fprofile-use} option.
2661 If a source file was changed between @option{-fprofile-gen} and
2662 @option{-fprofile-use}, the files with the profile feedback can fail
2663 to match the source file and GCC can not use the profile feedback
2664 information. By default, GCC emits an error message in this case.
2665 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2666 error. GCC does not use appropriate feedback profiles, so using this
2667 option can result in poorly optimized code. This option is useful
2668 only in the case of very minor changes such as bug fixes to an
2673 @node Warning Options
2674 @section Options to Request or Suppress Warnings
2675 @cindex options to control warnings
2676 @cindex warning messages
2677 @cindex messages, warning
2678 @cindex suppressing warnings
2680 Warnings are diagnostic messages that report constructions which
2681 are not inherently erroneous but which are risky or suggest there
2682 may have been an error.
2684 The following language-independent options do not enable specific
2685 warnings but control the kinds of diagnostics produced by GCC.
2688 @cindex syntax checking
2690 @opindex fsyntax-only
2691 Check the code for syntax errors, but don't do anything beyond that.
2695 Inhibit all warning messages.
2700 Make all warnings into errors.
2705 Make the specified warning into an error. The specifier for a warning
2706 is appended, for example @option{-Werror=switch} turns the warnings
2707 controlled by @option{-Wswitch} into errors. This switch takes a
2708 negative form, to be used to negate @option{-Werror} for specific
2709 warnings, for example @option{-Wno-error=switch} makes
2710 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2711 is in effect. You can use the @option{-fdiagnostics-show-option}
2712 option to have each controllable warning amended with the option which
2713 controls it, to determine what to use with this option.
2715 Note that specifying @option{-Werror=}@var{foo} automatically implies
2716 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2719 @item -Wfatal-errors
2720 @opindex Wfatal-errors
2721 @opindex Wno-fatal-errors
2722 This option causes the compiler to abort compilation on the first error
2723 occurred rather than trying to keep going and printing further error
2728 You can request many specific warnings with options beginning
2729 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2730 implicit declarations. Each of these specific warning options also
2731 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2732 example, @option{-Wno-implicit}. This manual lists only one of the
2733 two forms, whichever is not the default. For further,
2734 language-specific options also refer to @ref{C++ Dialect Options} and
2735 @ref{Objective-C and Objective-C++ Dialect Options}.
2740 Issue all the warnings demanded by strict ISO C and ISO C++;
2741 reject all programs that use forbidden extensions, and some other
2742 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2743 version of the ISO C standard specified by any @option{-std} option used.
2745 Valid ISO C and ISO C++ programs should compile properly with or without
2746 this option (though a rare few will require @option{-ansi} or a
2747 @option{-std} option specifying the required version of ISO C)@. However,
2748 without this option, certain GNU extensions and traditional C and C++
2749 features are supported as well. With this option, they are rejected.
2751 @option{-pedantic} does not cause warning messages for use of the
2752 alternate keywords whose names begin and end with @samp{__}. Pedantic
2753 warnings are also disabled in the expression that follows
2754 @code{__extension__}. However, only system header files should use
2755 these escape routes; application programs should avoid them.
2756 @xref{Alternate Keywords}.
2758 Some users try to use @option{-pedantic} to check programs for strict ISO
2759 C conformance. They soon find that it does not do quite what they want:
2760 it finds some non-ISO practices, but not all---only those for which
2761 ISO C @emph{requires} a diagnostic, and some others for which
2762 diagnostics have been added.
2764 A feature to report any failure to conform to ISO C might be useful in
2765 some instances, but would require considerable additional work and would
2766 be quite different from @option{-pedantic}. We don't have plans to
2767 support such a feature in the near future.
2769 Where the standard specified with @option{-std} represents a GNU
2770 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2771 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2772 extended dialect is based. Warnings from @option{-pedantic} are given
2773 where they are required by the base standard. (It would not make sense
2774 for such warnings to be given only for features not in the specified GNU
2775 C dialect, since by definition the GNU dialects of C include all
2776 features the compiler supports with the given option, and there would be
2777 nothing to warn about.)
2779 @item -pedantic-errors
2780 @opindex pedantic-errors
2781 Like @option{-pedantic}, except that errors are produced rather than
2787 This enables all the warnings about constructions that some users
2788 consider questionable, and that are easy to avoid (or modify to
2789 prevent the warning), even in conjunction with macros. This also
2790 enables some language-specific warnings described in @ref{C++ Dialect
2791 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2793 @option{-Wall} turns on the following warning flags:
2795 @gccoptlist{-Waddress @gol
2796 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2798 -Wchar-subscripts @gol
2799 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2801 -Wimplicit-function-declaration @gol
2804 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2805 -Wmissing-braces @gol
2811 -Wsequence-point @gol
2812 -Wsign-compare @r{(only in C++)} @gol
2813 -Wstrict-aliasing @gol
2814 -Wstrict-overflow=1 @gol
2817 -Wuninitialized @gol
2818 -Wunknown-pragmas @gol
2819 -Wunused-function @gol
2822 -Wunused-variable @gol
2823 -Wvolatile-register-var @gol
2826 Note that some warning flags are not implied by @option{-Wall}. Some of
2827 them warn about constructions that users generally do not consider
2828 questionable, but which occasionally you might wish to check for;
2829 others warn about constructions that are necessary or hard to avoid in
2830 some cases, and there is no simple way to modify the code to suppress
2831 the warning. Some of them are enabled by @option{-Wextra} but many of
2832 them must be enabled individually.
2838 This enables some extra warning flags that are not enabled by
2839 @option{-Wall}. (This option used to be called @option{-W}. The older
2840 name is still supported, but the newer name is more descriptive.)
2842 @gccoptlist{-Wclobbered @gol
2844 -Wignored-qualifiers @gol
2845 -Wmissing-field-initializers @gol
2846 -Wmissing-parameter-type @r{(C only)} @gol
2847 -Wold-style-declaration @r{(C only)} @gol
2848 -Woverride-init @gol
2851 -Wuninitialized @gol
2852 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2855 The option @option{-Wextra} also prints warning messages for the
2861 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2862 @samp{>}, or @samp{>=}.
2865 (C++ only) An enumerator and a non-enumerator both appear in a
2866 conditional expression.
2869 (C++ only) Ambiguous virtual bases.
2872 (C++ only) Subscripting an array which has been declared @samp{register}.
2875 (C++ only) Taking the address of a variable which has been declared
2879 (C++ only) A base class is not initialized in a derived class' copy
2884 @item -Wchar-subscripts
2885 @opindex Wchar-subscripts
2886 @opindex Wno-char-subscripts
2887 Warn if an array subscript has type @code{char}. This is a common cause
2888 of error, as programmers often forget that this type is signed on some
2890 This warning is enabled by @option{-Wall}.
2894 @opindex Wno-comment
2895 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2896 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2897 This warning is enabled by @option{-Wall}.
2902 @opindex ffreestanding
2903 @opindex fno-builtin
2904 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2905 the arguments supplied have types appropriate to the format string
2906 specified, and that the conversions specified in the format string make
2907 sense. This includes standard functions, and others specified by format
2908 attributes (@pxref{Function Attributes}), in the @code{printf},
2909 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2910 not in the C standard) families (or other target-specific families).
2911 Which functions are checked without format attributes having been
2912 specified depends on the standard version selected, and such checks of
2913 functions without the attribute specified are disabled by
2914 @option{-ffreestanding} or @option{-fno-builtin}.
2916 The formats are checked against the format features supported by GNU
2917 libc version 2.2. These include all ISO C90 and C99 features, as well
2918 as features from the Single Unix Specification and some BSD and GNU
2919 extensions. Other library implementations may not support all these
2920 features; GCC does not support warning about features that go beyond a
2921 particular library's limitations. However, if @option{-pedantic} is used
2922 with @option{-Wformat}, warnings will be given about format features not
2923 in the selected standard version (but not for @code{strfmon} formats,
2924 since those are not in any version of the C standard). @xref{C Dialect
2925 Options,,Options Controlling C Dialect}.
2927 Since @option{-Wformat} also checks for null format arguments for
2928 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2930 @option{-Wformat} is included in @option{-Wall}. For more control over some
2931 aspects of format checking, the options @option{-Wformat-y2k},
2932 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2933 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2934 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2937 @opindex Wformat-y2k
2938 @opindex Wno-format-y2k
2939 If @option{-Wformat} is specified, also warn about @code{strftime}
2940 formats which may yield only a two-digit year.
2942 @item -Wno-format-contains-nul
2943 @opindex Wno-format-contains-nul
2944 @opindex Wformat-contains-nul
2945 If @option{-Wformat} is specified, do not warn about format strings that
2948 @item -Wno-format-extra-args
2949 @opindex Wno-format-extra-args
2950 @opindex Wformat-extra-args
2951 If @option{-Wformat} is specified, do not warn about excess arguments to a
2952 @code{printf} or @code{scanf} format function. The C standard specifies
2953 that such arguments are ignored.
2955 Where the unused arguments lie between used arguments that are
2956 specified with @samp{$} operand number specifications, normally
2957 warnings are still given, since the implementation could not know what
2958 type to pass to @code{va_arg} to skip the unused arguments. However,
2959 in the case of @code{scanf} formats, this option will suppress the
2960 warning if the unused arguments are all pointers, since the Single
2961 Unix Specification says that such unused arguments are allowed.
2963 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2964 @opindex Wno-format-zero-length
2965 @opindex Wformat-zero-length
2966 If @option{-Wformat} is specified, do not warn about zero-length formats.
2967 The C standard specifies that zero-length formats are allowed.
2969 @item -Wformat-nonliteral
2970 @opindex Wformat-nonliteral
2971 @opindex Wno-format-nonliteral
2972 If @option{-Wformat} is specified, also warn if the format string is not a
2973 string literal and so cannot be checked, unless the format function
2974 takes its format arguments as a @code{va_list}.
2976 @item -Wformat-security
2977 @opindex Wformat-security
2978 @opindex Wno-format-security
2979 If @option{-Wformat} is specified, also warn about uses of format
2980 functions that represent possible security problems. At present, this
2981 warns about calls to @code{printf} and @code{scanf} functions where the
2982 format string is not a string literal and there are no format arguments,
2983 as in @code{printf (foo);}. This may be a security hole if the format
2984 string came from untrusted input and contains @samp{%n}. (This is
2985 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2986 in future warnings may be added to @option{-Wformat-security} that are not
2987 included in @option{-Wformat-nonliteral}.)
2991 @opindex Wno-format=2
2992 Enable @option{-Wformat} plus format checks not included in
2993 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2994 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2996 @item -Wnonnull @r{(C and Objective-C only)}
2998 @opindex Wno-nonnull
2999 Warn about passing a null pointer for arguments marked as
3000 requiring a non-null value by the @code{nonnull} function attribute.
3002 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3003 can be disabled with the @option{-Wno-nonnull} option.
3005 @item -Wjump-misses-init @r{(C, Objective-C only)}
3006 @opindex Wjump-misses-init
3007 @opindex Wno-jump-misses-init
3008 Warn if a @code{goto} statement or a @code{switch} statement jumps
3009 forward across the initialization of a variable, or jumps backward to a
3010 label after the variable has been initialized. This only warns about
3011 variables which are initialized when they are declared. This warning is
3012 only supported for C and Objective C; in C++ this sort of branch is an
3015 @option{-Wjump-misses-init} is included in @option{-Wall} and
3016 @option{-Wc++-compat}. It can be disabled with the
3017 @option{-Wno-jump-misses-init} option.
3019 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3021 @opindex Wno-init-self
3022 Warn about uninitialized variables which are initialized with themselves.
3023 Note this option can only be used with the @option{-Wuninitialized} option.
3025 For example, GCC will warn about @code{i} being uninitialized in the
3026 following snippet only when @option{-Winit-self} has been specified:
3037 @item -Wimplicit-int @r{(C and Objective-C only)}
3038 @opindex Wimplicit-int
3039 @opindex Wno-implicit-int
3040 Warn when a declaration does not specify a type.
3041 This warning is enabled by @option{-Wall}.
3043 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3044 @opindex Wimplicit-function-declaration
3045 @opindex Wno-implicit-function-declaration
3046 Give a warning whenever a function is used before being declared. In
3047 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3048 enabled by default and it is made into an error by
3049 @option{-pedantic-errors}. This warning is also enabled by
3054 @opindex Wno-implicit
3055 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3056 This warning is enabled by @option{-Wall}.
3058 @item -Wignored-qualifiers @r{(C and C++ only)}
3059 @opindex Wignored-qualifiers
3060 @opindex Wno-ignored-qualifiers
3061 Warn if the return type of a function has a type qualifier
3062 such as @code{const}. For ISO C such a type qualifier has no effect,
3063 since the value returned by a function is not an lvalue.
3064 For C++, the warning is only emitted for scalar types or @code{void}.
3065 ISO C prohibits qualified @code{void} return types on function
3066 definitions, so such return types always receive a warning
3067 even without this option.
3069 This warning is also enabled by @option{-Wextra}.
3074 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3075 a function with external linkage, returning int, taking either zero
3076 arguments, two, or three arguments of appropriate types. This warning
3077 is enabled by default in C++ and is enabled by either @option{-Wall}
3078 or @option{-pedantic}.
3080 @item -Wmissing-braces
3081 @opindex Wmissing-braces
3082 @opindex Wno-missing-braces
3083 Warn if an aggregate or union initializer is not fully bracketed. In
3084 the following example, the initializer for @samp{a} is not fully
3085 bracketed, but that for @samp{b} is fully bracketed.
3088 int a[2][2] = @{ 0, 1, 2, 3 @};
3089 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3092 This warning is enabled by @option{-Wall}.
3094 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3095 @opindex Wmissing-include-dirs
3096 @opindex Wno-missing-include-dirs
3097 Warn if a user-supplied include directory does not exist.
3100 @opindex Wparentheses
3101 @opindex Wno-parentheses
3102 Warn if parentheses are omitted in certain contexts, such
3103 as when there is an assignment in a context where a truth value
3104 is expected, or when operators are nested whose precedence people
3105 often get confused about.
3107 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3108 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3109 interpretation from that of ordinary mathematical notation.
3111 Also warn about constructions where there may be confusion to which
3112 @code{if} statement an @code{else} branch belongs. Here is an example of
3127 In C/C++, every @code{else} branch belongs to the innermost possible
3128 @code{if} statement, which in this example is @code{if (b)}. This is
3129 often not what the programmer expected, as illustrated in the above
3130 example by indentation the programmer chose. When there is the
3131 potential for this confusion, GCC will issue a warning when this flag
3132 is specified. To eliminate the warning, add explicit braces around
3133 the innermost @code{if} statement so there is no way the @code{else}
3134 could belong to the enclosing @code{if}. The resulting code would
3151 This warning is enabled by @option{-Wall}.
3153 @item -Wsequence-point
3154 @opindex Wsequence-point
3155 @opindex Wno-sequence-point
3156 Warn about code that may have undefined semantics because of violations
3157 of sequence point rules in the C and C++ standards.
3159 The C and C++ standards defines the order in which expressions in a C/C++
3160 program are evaluated in terms of @dfn{sequence points}, which represent
3161 a partial ordering between the execution of parts of the program: those
3162 executed before the sequence point, and those executed after it. These
3163 occur after the evaluation of a full expression (one which is not part
3164 of a larger expression), after the evaluation of the first operand of a
3165 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3166 function is called (but after the evaluation of its arguments and the
3167 expression denoting the called function), and in certain other places.
3168 Other than as expressed by the sequence point rules, the order of
3169 evaluation of subexpressions of an expression is not specified. All
3170 these rules describe only a partial order rather than a total order,
3171 since, for example, if two functions are called within one expression
3172 with no sequence point between them, the order in which the functions
3173 are called is not specified. However, the standards committee have
3174 ruled that function calls do not overlap.
3176 It is not specified when between sequence points modifications to the
3177 values of objects take effect. Programs whose behavior depends on this
3178 have undefined behavior; the C and C++ standards specify that ``Between
3179 the previous and next sequence point an object shall have its stored
3180 value modified at most once by the evaluation of an expression.
3181 Furthermore, the prior value shall be read only to determine the value
3182 to be stored.''. If a program breaks these rules, the results on any
3183 particular implementation are entirely unpredictable.
3185 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3186 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3187 diagnosed by this option, and it may give an occasional false positive
3188 result, but in general it has been found fairly effective at detecting
3189 this sort of problem in programs.
3191 The standard is worded confusingly, therefore there is some debate
3192 over the precise meaning of the sequence point rules in subtle cases.
3193 Links to discussions of the problem, including proposed formal
3194 definitions, may be found on the GCC readings page, at
3195 @w{@uref{http://gcc.gnu.org/readings.html}}.
3197 This warning is enabled by @option{-Wall} for C and C++.
3200 @opindex Wreturn-type
3201 @opindex Wno-return-type
3202 Warn whenever a function is defined with a return-type that defaults
3203 to @code{int}. Also warn about any @code{return} statement with no
3204 return-value in a function whose return-type is not @code{void}
3205 (falling off the end of the function body is considered returning
3206 without a value), and about a @code{return} statement with an
3207 expression in a function whose return-type is @code{void}.
3209 For C++, a function without return type always produces a diagnostic
3210 message, even when @option{-Wno-return-type} is specified. The only
3211 exceptions are @samp{main} and functions defined in system headers.
3213 This warning is enabled by @option{-Wall}.
3218 Warn whenever a @code{switch} statement has an index of enumerated type
3219 and lacks a @code{case} for one or more of the named codes of that
3220 enumeration. (The presence of a @code{default} label prevents this
3221 warning.) @code{case} labels outside the enumeration range also
3222 provoke warnings when this option is used (even if there is a
3223 @code{default} label).
3224 This warning is enabled by @option{-Wall}.
3226 @item -Wswitch-default
3227 @opindex Wswitch-default
3228 @opindex Wno-switch-default
3229 Warn whenever a @code{switch} statement does not have a @code{default}
3233 @opindex Wswitch-enum
3234 @opindex Wno-switch-enum
3235 Warn whenever a @code{switch} statement has an index of enumerated type
3236 and lacks a @code{case} for one or more of the named codes of that
3237 enumeration. @code{case} labels outside the enumeration range also
3238 provoke warnings when this option is used. The only difference
3239 between @option{-Wswitch} and this option is that this option gives a
3240 warning about an omitted enumeration code even if there is a
3241 @code{default} label.
3243 @item -Wsync-nand @r{(C and C++ only)}
3245 @opindex Wno-sync-nand
3246 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3247 built-in functions are used. These functions changed semantics in GCC 4.4.
3251 @opindex Wno-trigraphs
3252 Warn if any trigraphs are encountered that might change the meaning of
3253 the program (trigraphs within comments are not warned about).
3254 This warning is enabled by @option{-Wall}.
3256 @item -Wunused-function
3257 @opindex Wunused-function
3258 @opindex Wno-unused-function
3259 Warn whenever a static function is declared but not defined or a
3260 non-inline static function is unused.
3261 This warning is enabled by @option{-Wall}.
3263 @item -Wunused-label
3264 @opindex Wunused-label
3265 @opindex Wno-unused-label
3266 Warn whenever a label is declared but not used.
3267 This warning is enabled by @option{-Wall}.
3269 To suppress this warning use the @samp{unused} attribute
3270 (@pxref{Variable Attributes}).
3272 @item -Wunused-parameter
3273 @opindex Wunused-parameter
3274 @opindex Wno-unused-parameter
3275 Warn whenever a function parameter is unused aside from its declaration.
3277 To suppress this warning use the @samp{unused} attribute
3278 (@pxref{Variable Attributes}).
3280 @item -Wno-unused-result
3281 @opindex Wunused-result
3282 @opindex Wno-unused-result
3283 Do not warn if a caller of a function marked with attribute
3284 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3285 its return value. The default is @option{-Wunused-result}.
3287 @item -Wunused-variable
3288 @opindex Wunused-variable
3289 @opindex Wno-unused-variable
3290 Warn whenever a local variable or non-constant static variable is unused
3291 aside from its declaration.
3292 This warning is enabled by @option{-Wall}.
3294 To suppress this warning use the @samp{unused} attribute
3295 (@pxref{Variable Attributes}).
3297 @item -Wunused-value
3298 @opindex Wunused-value
3299 @opindex Wno-unused-value
3300 Warn whenever a statement computes a result that is explicitly not
3301 used. To suppress this warning cast the unused expression to
3302 @samp{void}. This includes an expression-statement or the left-hand
3303 side of a comma expression that contains no side effects. For example,
3304 an expression such as @samp{x[i,j]} will cause a warning, while
3305 @samp{x[(void)i,j]} will not.
3307 This warning is enabled by @option{-Wall}.
3312 All the above @option{-Wunused} options combined.
3314 In order to get a warning about an unused function parameter, you must
3315 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3316 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3318 @item -Wuninitialized
3319 @opindex Wuninitialized
3320 @opindex Wno-uninitialized
3321 Warn if an automatic variable is used without first being initialized
3322 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3323 warn if a non-static reference or non-static @samp{const} member
3324 appears in a class without constructors.
3326 If you want to warn about code which uses the uninitialized value of the
3327 variable in its own initializer, use the @option{-Winit-self} option.
3329 These warnings occur for individual uninitialized or clobbered
3330 elements of structure, union or array variables as well as for
3331 variables which are uninitialized or clobbered as a whole. They do
3332 not occur for variables or elements declared @code{volatile}. Because
3333 these warnings depend on optimization, the exact variables or elements
3334 for which there are warnings will depend on the precise optimization
3335 options and version of GCC used.
3337 Note that there may be no warning about a variable that is used only
3338 to compute a value that itself is never used, because such
3339 computations may be deleted by data flow analysis before the warnings
3342 These warnings are made optional because GCC is not smart
3343 enough to see all the reasons why the code might be correct
3344 despite appearing to have an error. Here is one example of how
3365 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3366 always initialized, but GCC doesn't know this. Here is
3367 another common case:
3372 if (change_y) save_y = y, y = new_y;
3374 if (change_y) y = save_y;
3379 This has no bug because @code{save_y} is used only if it is set.
3381 @cindex @code{longjmp} warnings
3382 This option also warns when a non-volatile automatic variable might be
3383 changed by a call to @code{longjmp}. These warnings as well are possible
3384 only in optimizing compilation.
3386 The compiler sees only the calls to @code{setjmp}. It cannot know
3387 where @code{longjmp} will be called; in fact, a signal handler could
3388 call it at any point in the code. As a result, you may get a warning
3389 even when there is in fact no problem because @code{longjmp} cannot
3390 in fact be called at the place which would cause a problem.
3392 Some spurious warnings can be avoided if you declare all the functions
3393 you use that never return as @code{noreturn}. @xref{Function
3396 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3398 @item -Wunknown-pragmas
3399 @opindex Wunknown-pragmas
3400 @opindex Wno-unknown-pragmas
3401 @cindex warning for unknown pragmas
3402 @cindex unknown pragmas, warning
3403 @cindex pragmas, warning of unknown
3404 Warn when a #pragma directive is encountered which is not understood by
3405 GCC@. If this command line option is used, warnings will even be issued
3406 for unknown pragmas in system header files. This is not the case if
3407 the warnings were only enabled by the @option{-Wall} command line option.
3410 @opindex Wno-pragmas
3412 Do not warn about misuses of pragmas, such as incorrect parameters,
3413 invalid syntax, or conflicts between pragmas. See also
3414 @samp{-Wunknown-pragmas}.
3416 @item -Wstrict-aliasing
3417 @opindex Wstrict-aliasing
3418 @opindex Wno-strict-aliasing
3419 This option is only active when @option{-fstrict-aliasing} is active.
3420 It warns about code which might break the strict aliasing rules that the
3421 compiler is using for optimization. The warning does not catch all
3422 cases, but does attempt to catch the more common pitfalls. It is
3423 included in @option{-Wall}.
3424 It is equivalent to @option{-Wstrict-aliasing=3}
3426 @item -Wstrict-aliasing=n
3427 @opindex Wstrict-aliasing=n
3428 @opindex Wno-strict-aliasing=n
3429 This option is only active when @option{-fstrict-aliasing} is active.
3430 It warns about code which might break the strict aliasing rules that the
3431 compiler is using for optimization.
3432 Higher levels correspond to higher accuracy (fewer false positives).
3433 Higher levels also correspond to more effort, similar to the way -O works.
3434 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3437 Level 1: Most aggressive, quick, least accurate.
3438 Possibly useful when higher levels
3439 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3440 false negatives. However, it has many false positives.
3441 Warns for all pointer conversions between possibly incompatible types,
3442 even if never dereferenced. Runs in the frontend only.
3444 Level 2: Aggressive, quick, not too precise.
3445 May still have many false positives (not as many as level 1 though),
3446 and few false negatives (but possibly more than level 1).
3447 Unlike level 1, it only warns when an address is taken. Warns about
3448 incomplete types. Runs in the frontend only.
3450 Level 3 (default for @option{-Wstrict-aliasing}):
3451 Should have very few false positives and few false
3452 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3453 Takes care of the common punn+dereference pattern in the frontend:
3454 @code{*(int*)&some_float}.
3455 If optimization is enabled, it also runs in the backend, where it deals
3456 with multiple statement cases using flow-sensitive points-to information.
3457 Only warns when the converted pointer is dereferenced.
3458 Does not warn about incomplete types.
3460 @item -Wstrict-overflow
3461 @itemx -Wstrict-overflow=@var{n}
3462 @opindex Wstrict-overflow
3463 @opindex Wno-strict-overflow
3464 This option is only active when @option{-fstrict-overflow} is active.
3465 It warns about cases where the compiler optimizes based on the
3466 assumption that signed overflow does not occur. Note that it does not
3467 warn about all cases where the code might overflow: it only warns
3468 about cases where the compiler implements some optimization. Thus
3469 this warning depends on the optimization level.
3471 An optimization which assumes that signed overflow does not occur is
3472 perfectly safe if the values of the variables involved are such that
3473 overflow never does, in fact, occur. Therefore this warning can
3474 easily give a false positive: a warning about code which is not
3475 actually a problem. To help focus on important issues, several
3476 warning levels are defined. No warnings are issued for the use of
3477 undefined signed overflow when estimating how many iterations a loop
3478 will require, in particular when determining whether a loop will be
3482 @item -Wstrict-overflow=1
3483 Warn about cases which are both questionable and easy to avoid. For
3484 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3485 compiler will simplify this to @code{1}. This level of
3486 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3487 are not, and must be explicitly requested.
3489 @item -Wstrict-overflow=2
3490 Also warn about other cases where a comparison is simplified to a
3491 constant. For example: @code{abs (x) >= 0}. This can only be
3492 simplified when @option{-fstrict-overflow} is in effect, because
3493 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3494 zero. @option{-Wstrict-overflow} (with no level) is the same as
3495 @option{-Wstrict-overflow=2}.
3497 @item -Wstrict-overflow=3
3498 Also warn about other cases where a comparison is simplified. For
3499 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3501 @item -Wstrict-overflow=4
3502 Also warn about other simplifications not covered by the above cases.
3503 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3505 @item -Wstrict-overflow=5
3506 Also warn about cases where the compiler reduces the magnitude of a
3507 constant involved in a comparison. For example: @code{x + 2 > y} will
3508 be simplified to @code{x + 1 >= y}. This is reported only at the
3509 highest warning level because this simplification applies to many
3510 comparisons, so this warning level will give a very large number of
3514 @item -Warray-bounds
3515 @opindex Wno-array-bounds
3516 @opindex Warray-bounds
3517 This option is only active when @option{-ftree-vrp} is active
3518 (default for -O2 and above). It warns about subscripts to arrays
3519 that are always out of bounds. This warning is enabled by @option{-Wall}.
3521 @item -Wno-div-by-zero
3522 @opindex Wno-div-by-zero
3523 @opindex Wdiv-by-zero
3524 Do not warn about compile-time integer division by zero. Floating point
3525 division by zero is not warned about, as it can be a legitimate way of
3526 obtaining infinities and NaNs.
3528 @item -Wsystem-headers
3529 @opindex Wsystem-headers
3530 @opindex Wno-system-headers
3531 @cindex warnings from system headers
3532 @cindex system headers, warnings from
3533 Print warning messages for constructs found in system header files.
3534 Warnings from system headers are normally suppressed, on the assumption
3535 that they usually do not indicate real problems and would only make the
3536 compiler output harder to read. Using this command line option tells
3537 GCC to emit warnings from system headers as if they occurred in user
3538 code. However, note that using @option{-Wall} in conjunction with this
3539 option will @emph{not} warn about unknown pragmas in system
3540 headers---for that, @option{-Wunknown-pragmas} must also be used.
3543 @opindex Wfloat-equal
3544 @opindex Wno-float-equal
3545 Warn if floating point values are used in equality comparisons.
3547 The idea behind this is that sometimes it is convenient (for the
3548 programmer) to consider floating-point values as approximations to
3549 infinitely precise real numbers. If you are doing this, then you need
3550 to compute (by analyzing the code, or in some other way) the maximum or
3551 likely maximum error that the computation introduces, and allow for it
3552 when performing comparisons (and when producing output, but that's a
3553 different problem). In particular, instead of testing for equality, you
3554 would check to see whether the two values have ranges that overlap; and
3555 this is done with the relational operators, so equality comparisons are
3558 @item -Wtraditional @r{(C and Objective-C only)}
3559 @opindex Wtraditional
3560 @opindex Wno-traditional
3561 Warn about certain constructs that behave differently in traditional and
3562 ISO C@. Also warn about ISO C constructs that have no traditional C
3563 equivalent, and/or problematic constructs which should be avoided.
3567 Macro parameters that appear within string literals in the macro body.
3568 In traditional C macro replacement takes place within string literals,
3569 but does not in ISO C@.
3572 In traditional C, some preprocessor directives did not exist.
3573 Traditional preprocessors would only consider a line to be a directive
3574 if the @samp{#} appeared in column 1 on the line. Therefore
3575 @option{-Wtraditional} warns about directives that traditional C
3576 understands but would ignore because the @samp{#} does not appear as the
3577 first character on the line. It also suggests you hide directives like
3578 @samp{#pragma} not understood by traditional C by indenting them. Some
3579 traditional implementations would not recognize @samp{#elif}, so it
3580 suggests avoiding it altogether.
3583 A function-like macro that appears without arguments.
3586 The unary plus operator.
3589 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3590 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3591 constants.) Note, these suffixes appear in macros defined in the system
3592 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3593 Use of these macros in user code might normally lead to spurious
3594 warnings, however GCC's integrated preprocessor has enough context to
3595 avoid warning in these cases.
3598 A function declared external in one block and then used after the end of
3602 A @code{switch} statement has an operand of type @code{long}.
3605 A non-@code{static} function declaration follows a @code{static} one.
3606 This construct is not accepted by some traditional C compilers.
3609 The ISO type of an integer constant has a different width or
3610 signedness from its traditional type. This warning is only issued if
3611 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3612 typically represent bit patterns, are not warned about.
3615 Usage of ISO string concatenation is detected.
3618 Initialization of automatic aggregates.
3621 Identifier conflicts with labels. Traditional C lacks a separate
3622 namespace for labels.
3625 Initialization of unions. If the initializer is zero, the warning is
3626 omitted. This is done under the assumption that the zero initializer in
3627 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3628 initializer warnings and relies on default initialization to zero in the
3632 Conversions by prototypes between fixed/floating point values and vice
3633 versa. The absence of these prototypes when compiling with traditional
3634 C would cause serious problems. This is a subset of the possible
3635 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3638 Use of ISO C style function definitions. This warning intentionally is
3639 @emph{not} issued for prototype declarations or variadic functions
3640 because these ISO C features will appear in your code when using
3641 libiberty's traditional C compatibility macros, @code{PARAMS} and
3642 @code{VPARAMS}. This warning is also bypassed for nested functions
3643 because that feature is already a GCC extension and thus not relevant to
3644 traditional C compatibility.
3647 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3648 @opindex Wtraditional-conversion
3649 @opindex Wno-traditional-conversion
3650 Warn if a prototype causes a type conversion that is different from what
3651 would happen to the same argument in the absence of a prototype. This
3652 includes conversions of fixed point to floating and vice versa, and
3653 conversions changing the width or signedness of a fixed point argument
3654 except when the same as the default promotion.
3656 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3657 @opindex Wdeclaration-after-statement
3658 @opindex Wno-declaration-after-statement
3659 Warn when a declaration is found after a statement in a block. This
3660 construct, known from C++, was introduced with ISO C99 and is by default
3661 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3662 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3667 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3669 @item -Wno-endif-labels
3670 @opindex Wno-endif-labels
3671 @opindex Wendif-labels
3672 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3677 Warn whenever a local variable shadows another local variable, parameter or
3678 global variable or whenever a built-in function is shadowed.
3680 @item -Wlarger-than=@var{len}
3681 @opindex Wlarger-than=@var{len}
3682 @opindex Wlarger-than-@var{len}
3683 Warn whenever an object of larger than @var{len} bytes is defined.
3685 @item -Wframe-larger-than=@var{len}
3686 @opindex Wframe-larger-than
3687 Warn if the size of a function frame is larger than @var{len} bytes.
3688 The computation done to determine the stack frame size is approximate
3689 and not conservative.
3690 The actual requirements may be somewhat greater than @var{len}
3691 even if you do not get a warning. In addition, any space allocated
3692 via @code{alloca}, variable-length arrays, or related constructs
3693 is not included by the compiler when determining
3694 whether or not to issue a warning.
3696 @item -Wunsafe-loop-optimizations
3697 @opindex Wunsafe-loop-optimizations
3698 @opindex Wno-unsafe-loop-optimizations
3699 Warn if the loop cannot be optimized because the compiler could not
3700 assume anything on the bounds of the loop indices. With
3701 @option{-funsafe-loop-optimizations} warn if the compiler made
3704 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3705 @opindex Wno-pedantic-ms-format
3706 @opindex Wpedantic-ms-format
3707 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3708 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3709 depending on the MS runtime, when you are using the options @option{-Wformat}
3710 and @option{-pedantic} without gnu-extensions.
3712 @item -Wpointer-arith
3713 @opindex Wpointer-arith
3714 @opindex Wno-pointer-arith
3715 Warn about anything that depends on the ``size of'' a function type or
3716 of @code{void}. GNU C assigns these types a size of 1, for
3717 convenience in calculations with @code{void *} pointers and pointers
3718 to functions. In C++, warn also when an arithmetic operation involves
3719 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3722 @opindex Wtype-limits
3723 @opindex Wno-type-limits
3724 Warn if a comparison is always true or always false due to the limited
3725 range of the data type, but do not warn for constant expressions. For
3726 example, warn if an unsigned variable is compared against zero with
3727 @samp{<} or @samp{>=}. This warning is also enabled by
3730 @item -Wbad-function-cast @r{(C and Objective-C only)}
3731 @opindex Wbad-function-cast
3732 @opindex Wno-bad-function-cast
3733 Warn whenever a function call is cast to a non-matching type.
3734 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3736 @item -Wc++-compat @r{(C and Objective-C only)}
3737 Warn about ISO C constructs that are outside of the common subset of
3738 ISO C and ISO C++, e.g.@: request for implicit conversion from
3739 @code{void *} to a pointer to non-@code{void} type.
3741 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3742 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3743 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3744 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3748 @opindex Wno-cast-qual
3749 Warn whenever a pointer is cast so as to remove a type qualifier from
3750 the target type. For example, warn if a @code{const char *} is cast
3751 to an ordinary @code{char *}.
3753 Also warn when making a cast which introduces a type qualifier in an
3754 unsafe way. For example, casting @code{char **} to @code{const char **}
3755 is unsafe, as in this example:
3758 /* p is char ** value. */
3759 const char **q = (const char **) p;
3760 /* Assignment of readonly string to const char * is OK. */
3762 /* Now char** pointer points to read-only memory. */
3767 @opindex Wcast-align
3768 @opindex Wno-cast-align
3769 Warn whenever a pointer is cast such that the required alignment of the
3770 target is increased. For example, warn if a @code{char *} is cast to
3771 an @code{int *} on machines where integers can only be accessed at
3772 two- or four-byte boundaries.
3774 @item -Wwrite-strings
3775 @opindex Wwrite-strings
3776 @opindex Wno-write-strings
3777 When compiling C, give string constants the type @code{const
3778 char[@var{length}]} so that copying the address of one into a
3779 non-@code{const} @code{char *} pointer will get a warning. These
3780 warnings will help you find at compile time code that can try to write
3781 into a string constant, but only if you have been very careful about
3782 using @code{const} in declarations and prototypes. Otherwise, it will
3783 just be a nuisance. This is why we did not make @option{-Wall} request
3786 When compiling C++, warn about the deprecated conversion from string
3787 literals to @code{char *}. This warning is enabled by default for C++
3792 @opindex Wno-clobbered
3793 Warn for variables that might be changed by @samp{longjmp} or
3794 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3797 @opindex Wconversion
3798 @opindex Wno-conversion
3799 Warn for implicit conversions that may alter a value. This includes
3800 conversions between real and integer, like @code{abs (x)} when
3801 @code{x} is @code{double}; conversions between signed and unsigned,
3802 like @code{unsigned ui = -1}; and conversions to smaller types, like
3803 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3804 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3805 changed by the conversion like in @code{abs (2.0)}. Warnings about
3806 conversions between signed and unsigned integers can be disabled by
3807 using @option{-Wno-sign-conversion}.
3809 For C++, also warn for conversions between @code{NULL} and non-pointer
3810 types; confusing overload resolution for user-defined conversions; and
3811 conversions that will never use a type conversion operator:
3812 conversions to @code{void}, the same type, a base class or a reference
3813 to them. Warnings about conversions between signed and unsigned
3814 integers are disabled by default in C++ unless
3815 @option{-Wsign-conversion} is explicitly enabled.
3818 @opindex Wempty-body
3819 @opindex Wno-empty-body
3820 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3821 while} statement. This warning is also enabled by @option{-Wextra}.
3823 @item -Wenum-compare
3824 @opindex Wenum-compare
3825 @opindex Wno-enum-compare
3826 Warn about a comparison between values of different enum types. In C++
3827 this warning is enabled by default. In C this warning is enabled by
3830 @item -Wsign-compare
3831 @opindex Wsign-compare
3832 @opindex Wno-sign-compare
3833 @cindex warning for comparison of signed and unsigned values
3834 @cindex comparison of signed and unsigned values, warning
3835 @cindex signed and unsigned values, comparison warning
3836 Warn when a comparison between signed and unsigned values could produce
3837 an incorrect result when the signed value is converted to unsigned.
3838 This warning is also enabled by @option{-Wextra}; to get the other warnings
3839 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3841 @item -Wsign-conversion
3842 @opindex Wsign-conversion
3843 @opindex Wno-sign-conversion
3844 Warn for implicit conversions that may change the sign of an integer
3845 value, like assigning a signed integer expression to an unsigned
3846 integer variable. An explicit cast silences the warning. In C, this
3847 option is enabled also by @option{-Wconversion}.
3851 @opindex Wno-address
3852 Warn about suspicious uses of memory addresses. These include using
3853 the address of a function in a conditional expression, such as
3854 @code{void func(void); if (func)}, and comparisons against the memory
3855 address of a string literal, such as @code{if (x == "abc")}. Such
3856 uses typically indicate a programmer error: the address of a function
3857 always evaluates to true, so their use in a conditional usually
3858 indicate that the programmer forgot the parentheses in a function
3859 call; and comparisons against string literals result in unspecified
3860 behavior and are not portable in C, so they usually indicate that the
3861 programmer intended to use @code{strcmp}. This warning is enabled by
3865 @opindex Wlogical-op
3866 @opindex Wno-logical-op
3867 Warn about suspicious uses of logical operators in expressions.
3868 This includes using logical operators in contexts where a
3869 bit-wise operator is likely to be expected.
3871 @item -Waggregate-return
3872 @opindex Waggregate-return
3873 @opindex Wno-aggregate-return
3874 Warn if any functions that return structures or unions are defined or
3875 called. (In languages where you can return an array, this also elicits
3878 @item -Wno-attributes
3879 @opindex Wno-attributes
3880 @opindex Wattributes
3881 Do not warn if an unexpected @code{__attribute__} is used, such as
3882 unrecognized attributes, function attributes applied to variables,
3883 etc. This will not stop errors for incorrect use of supported
3886 @item -Wno-builtin-macro-redefined
3887 @opindex Wno-builtin-macro-redefined
3888 @opindex Wbuiltin-macro-redefined
3889 Do not warn if certain built-in macros are redefined. This suppresses
3890 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3891 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3893 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3894 @opindex Wstrict-prototypes
3895 @opindex Wno-strict-prototypes
3896 Warn if a function is declared or defined without specifying the
3897 argument types. (An old-style function definition is permitted without
3898 a warning if preceded by a declaration which specifies the argument
3901 @item -Wold-style-declaration @r{(C and Objective-C only)}
3902 @opindex Wold-style-declaration
3903 @opindex Wno-old-style-declaration
3904 Warn for obsolescent usages, according to the C Standard, in a
3905 declaration. For example, warn if storage-class specifiers like
3906 @code{static} are not the first things in a declaration. This warning
3907 is also enabled by @option{-Wextra}.
3909 @item -Wold-style-definition @r{(C and Objective-C only)}
3910 @opindex Wold-style-definition
3911 @opindex Wno-old-style-definition
3912 Warn if an old-style function definition is used. A warning is given
3913 even if there is a previous prototype.
3915 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3916 @opindex Wmissing-parameter-type
3917 @opindex Wno-missing-parameter-type
3918 A function parameter is declared without a type specifier in K&R-style
3925 This warning is also enabled by @option{-Wextra}.
3927 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3928 @opindex Wmissing-prototypes
3929 @opindex Wno-missing-prototypes
3930 Warn if a global function is defined without a previous prototype
3931 declaration. This warning is issued even if the definition itself
3932 provides a prototype. The aim is to detect global functions that fail
3933 to be declared in header files.
3935 @item -Wmissing-declarations
3936 @opindex Wmissing-declarations
3937 @opindex Wno-missing-declarations
3938 Warn if a global function is defined without a previous declaration.
3939 Do so even if the definition itself provides a prototype.
3940 Use this option to detect global functions that are not declared in
3941 header files. In C++, no warnings are issued for function templates,
3942 or for inline functions, or for functions in anonymous namespaces.
3944 @item -Wmissing-field-initializers
3945 @opindex Wmissing-field-initializers
3946 @opindex Wno-missing-field-initializers
3950 Warn if a structure's initializer has some fields missing. For
3951 example, the following code would cause such a warning, because
3952 @code{x.h} is implicitly zero:
3955 struct s @{ int f, g, h; @};
3956 struct s x = @{ 3, 4 @};
3959 This option does not warn about designated initializers, so the following
3960 modification would not trigger a warning:
3963 struct s @{ int f, g, h; @};
3964 struct s x = @{ .f = 3, .g = 4 @};
3967 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3968 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3970 @item -Wmissing-noreturn
3971 @opindex Wmissing-noreturn
3972 @opindex Wno-missing-noreturn
3973 Warn about functions which might be candidates for attribute @code{noreturn}.
3974 Note these are only possible candidates, not absolute ones. Care should
3975 be taken to manually verify functions actually do not ever return before
3976 adding the @code{noreturn} attribute, otherwise subtle code generation
3977 bugs could be introduced. You will not get a warning for @code{main} in
3978 hosted C environments.
3980 @item -Wmissing-format-attribute
3981 @opindex Wmissing-format-attribute
3982 @opindex Wno-missing-format-attribute
3985 Warn about function pointers which might be candidates for @code{format}
3986 attributes. Note these are only possible candidates, not absolute ones.
3987 GCC will guess that function pointers with @code{format} attributes that
3988 are used in assignment, initialization, parameter passing or return
3989 statements should have a corresponding @code{format} attribute in the
3990 resulting type. I.e.@: the left-hand side of the assignment or
3991 initialization, the type of the parameter variable, or the return type
3992 of the containing function respectively should also have a @code{format}
3993 attribute to avoid the warning.
3995 GCC will also warn about function definitions which might be
3996 candidates for @code{format} attributes. Again, these are only
3997 possible candidates. GCC will guess that @code{format} attributes
3998 might be appropriate for any function that calls a function like
3999 @code{vprintf} or @code{vscanf}, but this might not always be the
4000 case, and some functions for which @code{format} attributes are
4001 appropriate may not be detected.
4003 @item -Wno-multichar
4004 @opindex Wno-multichar
4006 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4007 Usually they indicate a typo in the user's code, as they have
4008 implementation-defined values, and should not be used in portable code.
4010 @item -Wnormalized=<none|id|nfc|nfkc>
4011 @opindex Wnormalized=
4014 @cindex character set, input normalization
4015 In ISO C and ISO C++, two identifiers are different if they are
4016 different sequences of characters. However, sometimes when characters
4017 outside the basic ASCII character set are used, you can have two
4018 different character sequences that look the same. To avoid confusion,
4019 the ISO 10646 standard sets out some @dfn{normalization rules} which
4020 when applied ensure that two sequences that look the same are turned into
4021 the same sequence. GCC can warn you if you are using identifiers which
4022 have not been normalized; this option controls that warning.
4024 There are four levels of warning that GCC supports. The default is
4025 @option{-Wnormalized=nfc}, which warns about any identifier which is
4026 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4027 recommended form for most uses.
4029 Unfortunately, there are some characters which ISO C and ISO C++ allow
4030 in identifiers that when turned into NFC aren't allowable as
4031 identifiers. That is, there's no way to use these symbols in portable
4032 ISO C or C++ and have all your identifiers in NFC@.
4033 @option{-Wnormalized=id} suppresses the warning for these characters.
4034 It is hoped that future versions of the standards involved will correct
4035 this, which is why this option is not the default.
4037 You can switch the warning off for all characters by writing
4038 @option{-Wnormalized=none}. You would only want to do this if you
4039 were using some other normalization scheme (like ``D''), because
4040 otherwise you can easily create bugs that are literally impossible to see.
4042 Some characters in ISO 10646 have distinct meanings but look identical
4043 in some fonts or display methodologies, especially once formatting has
4044 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4045 LETTER N'', will display just like a regular @code{n} which has been
4046 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4047 normalization scheme to convert all these into a standard form as
4048 well, and GCC will warn if your code is not in NFKC if you use
4049 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4050 about every identifier that contains the letter O because it might be
4051 confused with the digit 0, and so is not the default, but may be
4052 useful as a local coding convention if the programming environment is
4053 unable to be fixed to display these characters distinctly.
4055 @item -Wno-deprecated
4056 @opindex Wno-deprecated
4057 @opindex Wdeprecated
4058 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4060 @item -Wno-deprecated-declarations
4061 @opindex Wno-deprecated-declarations
4062 @opindex Wdeprecated-declarations
4063 Do not warn about uses of functions (@pxref{Function Attributes}),
4064 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4065 Attributes}) marked as deprecated by using the @code{deprecated}
4069 @opindex Wno-overflow
4071 Do not warn about compile-time overflow in constant expressions.
4073 @item -Woverride-init @r{(C and Objective-C only)}
4074 @opindex Woverride-init
4075 @opindex Wno-override-init
4079 Warn if an initialized field without side effects is overridden when
4080 using designated initializers (@pxref{Designated Inits, , Designated
4083 This warning is included in @option{-Wextra}. To get other
4084 @option{-Wextra} warnings without this one, use @samp{-Wextra
4085 -Wno-override-init}.
4090 Warn if a structure is given the packed attribute, but the packed
4091 attribute has no effect on the layout or size of the structure.
4092 Such structures may be mis-aligned for little benefit. For
4093 instance, in this code, the variable @code{f.x} in @code{struct bar}
4094 will be misaligned even though @code{struct bar} does not itself
4095 have the packed attribute:
4102 @} __attribute__((packed));
4110 @item -Wpacked-bitfield-compat
4111 @opindex Wpacked-bitfield-compat
4112 @opindex Wno-packed-bitfield-compat
4113 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4114 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4115 the change can lead to differences in the structure layout. GCC
4116 informs you when the offset of such a field has changed in GCC 4.4.
4117 For example there is no longer a 4-bit padding between field @code{a}
4118 and @code{b} in this structure:
4125 @} __attribute__ ((packed));
4128 This warning is enabled by default. Use
4129 @option{-Wno-packed-bitfield-compat} to disable this warning.
4134 Warn if padding is included in a structure, either to align an element
4135 of the structure or to align the whole structure. Sometimes when this
4136 happens it is possible to rearrange the fields of the structure to
4137 reduce the padding and so make the structure smaller.
4139 @item -Wredundant-decls
4140 @opindex Wredundant-decls
4141 @opindex Wno-redundant-decls
4142 Warn if anything is declared more than once in the same scope, even in
4143 cases where multiple declaration is valid and changes nothing.
4145 @item -Wnested-externs @r{(C and Objective-C only)}
4146 @opindex Wnested-externs
4147 @opindex Wno-nested-externs
4148 Warn if an @code{extern} declaration is encountered within a function.
4150 @item -Wunreachable-code
4151 @opindex Wunreachable-code
4152 @opindex Wno-unreachable-code
4153 Warn if the compiler detects that code will never be executed.
4155 This option is intended to warn when the compiler detects that at
4156 least a whole line of source code will never be executed, because
4157 some condition is never satisfied or because it is after a
4158 procedure that never returns.
4160 It is possible for this option to produce a warning even though there
4161 are circumstances under which part of the affected line can be executed,
4162 so care should be taken when removing apparently-unreachable code.
4164 For instance, when a function is inlined, a warning may mean that the
4165 line is unreachable in only one inlined copy of the function.
4167 This option is not made part of @option{-Wall} because in a debugging
4168 version of a program there is often substantial code which checks
4169 correct functioning of the program and is, hopefully, unreachable
4170 because the program does work. Another common use of unreachable
4171 code is to provide behavior which is selectable at compile-time.
4176 Warn if a function can not be inlined and it was declared as inline.
4177 Even with this option, the compiler will not warn about failures to
4178 inline functions declared in system headers.
4180 The compiler uses a variety of heuristics to determine whether or not
4181 to inline a function. For example, the compiler takes into account
4182 the size of the function being inlined and the amount of inlining
4183 that has already been done in the current function. Therefore,
4184 seemingly insignificant changes in the source program can cause the
4185 warnings produced by @option{-Winline} to appear or disappear.
4187 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4188 @opindex Wno-invalid-offsetof
4189 @opindex Winvalid-offsetof
4190 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4191 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4192 to a non-POD type is undefined. In existing C++ implementations,
4193 however, @samp{offsetof} typically gives meaningful results even when
4194 applied to certain kinds of non-POD types. (Such as a simple
4195 @samp{struct} that fails to be a POD type only by virtue of having a
4196 constructor.) This flag is for users who are aware that they are
4197 writing nonportable code and who have deliberately chosen to ignore the
4200 The restrictions on @samp{offsetof} may be relaxed in a future version
4201 of the C++ standard.
4203 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4204 @opindex Wno-int-to-pointer-cast
4205 @opindex Wint-to-pointer-cast
4206 Suppress warnings from casts to pointer type of an integer of a
4209 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4210 @opindex Wno-pointer-to-int-cast
4211 @opindex Wpointer-to-int-cast
4212 Suppress warnings from casts from a pointer to an integer type of a
4216 @opindex Winvalid-pch
4217 @opindex Wno-invalid-pch
4218 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4219 the search path but can't be used.
4223 @opindex Wno-long-long
4224 Warn if @samp{long long} type is used. This is enabled by either
4225 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4226 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4228 @item -Wvariadic-macros
4229 @opindex Wvariadic-macros
4230 @opindex Wno-variadic-macros
4231 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4232 alternate syntax when in pedantic ISO C99 mode. This is default.
4233 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4238 Warn if variable length array is used in the code.
4239 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4240 the variable length array.
4242 @item -Wvolatile-register-var
4243 @opindex Wvolatile-register-var
4244 @opindex Wno-volatile-register-var
4245 Warn if a register variable is declared volatile. The volatile
4246 modifier does not inhibit all optimizations that may eliminate reads
4247 and/or writes to register variables. This warning is enabled by
4250 @item -Wdisabled-optimization
4251 @opindex Wdisabled-optimization
4252 @opindex Wno-disabled-optimization
4253 Warn if a requested optimization pass is disabled. This warning does
4254 not generally indicate that there is anything wrong with your code; it
4255 merely indicates that GCC's optimizers were unable to handle the code
4256 effectively. Often, the problem is that your code is too big or too
4257 complex; GCC will refuse to optimize programs when the optimization
4258 itself is likely to take inordinate amounts of time.
4260 @item -Wpointer-sign @r{(C and Objective-C only)}
4261 @opindex Wpointer-sign
4262 @opindex Wno-pointer-sign
4263 Warn for pointer argument passing or assignment with different signedness.
4264 This option is only supported for C and Objective-C@. It is implied by
4265 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4266 @option{-Wno-pointer-sign}.
4268 @item -Wstack-protector
4269 @opindex Wstack-protector
4270 @opindex Wno-stack-protector
4271 This option is only active when @option{-fstack-protector} is active. It
4272 warns about functions that will not be protected against stack smashing.
4275 @opindex Wno-mudflap
4276 Suppress warnings about constructs that cannot be instrumented by
4279 @item -Woverlength-strings
4280 @opindex Woverlength-strings
4281 @opindex Wno-overlength-strings
4282 Warn about string constants which are longer than the ``minimum
4283 maximum'' length specified in the C standard. Modern compilers
4284 generally allow string constants which are much longer than the
4285 standard's minimum limit, but very portable programs should avoid
4286 using longer strings.
4288 The limit applies @emph{after} string constant concatenation, and does
4289 not count the trailing NUL@. In C89, the limit was 509 characters; in
4290 C99, it was raised to 4095. C++98 does not specify a normative
4291 minimum maximum, so we do not diagnose overlength strings in C++@.
4293 This option is implied by @option{-pedantic}, and can be disabled with
4294 @option{-Wno-overlength-strings}.
4296 @item -Wunsuffixed-float-constants
4297 @opindex Wunsuffixed-float-constants
4299 GCC will issue a warning for any floating constant that does not have
4300 a suffix. When used together with @option{-Wsystem-headers} it will
4301 warn about such constants in system header files. This can be useful
4302 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4303 from the decimal floating-point extension to C99.
4306 @node Debugging Options
4307 @section Options for Debugging Your Program or GCC
4308 @cindex options, debugging
4309 @cindex debugging information options
4311 GCC has various special options that are used for debugging
4312 either your program or GCC:
4317 Produce debugging information in the operating system's native format
4318 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4321 On most systems that use stabs format, @option{-g} enables use of extra
4322 debugging information that only GDB can use; this extra information
4323 makes debugging work better in GDB but will probably make other debuggers
4325 refuse to read the program. If you want to control for certain whether
4326 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4327 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4329 GCC allows you to use @option{-g} with
4330 @option{-O}. The shortcuts taken by optimized code may occasionally
4331 produce surprising results: some variables you declared may not exist
4332 at all; flow of control may briefly move where you did not expect it;
4333 some statements may not be executed because they compute constant
4334 results or their values were already at hand; some statements may
4335 execute in different places because they were moved out of loops.
4337 Nevertheless it proves possible to debug optimized output. This makes
4338 it reasonable to use the optimizer for programs that might have bugs.
4340 The following options are useful when GCC is generated with the
4341 capability for more than one debugging format.
4345 Produce debugging information for use by GDB@. This means to use the
4346 most expressive format available (DWARF 2, stabs, or the native format
4347 if neither of those are supported), including GDB extensions if at all
4352 Produce debugging information in stabs format (if that is supported),
4353 without GDB extensions. This is the format used by DBX on most BSD
4354 systems. On MIPS, Alpha and System V Release 4 systems this option
4355 produces stabs debugging output which is not understood by DBX or SDB@.
4356 On System V Release 4 systems this option requires the GNU assembler.
4358 @item -feliminate-unused-debug-symbols
4359 @opindex feliminate-unused-debug-symbols
4360 Produce debugging information in stabs format (if that is supported),
4361 for only symbols that are actually used.
4363 @item -femit-class-debug-always
4364 Instead of emitting debugging information for a C++ class in only one
4365 object file, emit it in all object files using the class. This option
4366 should be used only with debuggers that are unable to handle the way GCC
4367 normally emits debugging information for classes because using this
4368 option will increase the size of debugging information by as much as a
4373 Produce debugging information in stabs format (if that is supported),
4374 using GNU extensions understood only by the GNU debugger (GDB)@. The
4375 use of these extensions is likely to make other debuggers crash or
4376 refuse to read the program.
4380 Produce debugging information in COFF format (if that is supported).
4381 This is the format used by SDB on most System V systems prior to
4386 Produce debugging information in XCOFF format (if that is supported).
4387 This is the format used by the DBX debugger on IBM RS/6000 systems.
4391 Produce debugging information in XCOFF format (if that is supported),
4392 using GNU extensions understood only by the GNU debugger (GDB)@. The
4393 use of these extensions is likely to make other debuggers crash or
4394 refuse to read the program, and may cause assemblers other than the GNU
4395 assembler (GAS) to fail with an error.
4397 @item -gdwarf-@var{version}
4398 @opindex gdwarf-@var{version}
4399 Produce debugging information in DWARF format (if that is
4400 supported). This is the format used by DBX on IRIX 6. The value
4401 of @var{version} may be either 2, 3 or 4; the default version is 2.
4403 Note that with DWARF version 2 some ports require, and will always
4404 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4406 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4407 for maximum benefit.
4411 Produce debugging information in VMS debug format (if that is
4412 supported). This is the format used by DEBUG on VMS systems.
4415 @itemx -ggdb@var{level}
4416 @itemx -gstabs@var{level}
4417 @itemx -gcoff@var{level}
4418 @itemx -gxcoff@var{level}
4419 @itemx -gvms@var{level}
4420 Request debugging information and also use @var{level} to specify how
4421 much information. The default level is 2.
4423 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4426 Level 1 produces minimal information, enough for making backtraces in
4427 parts of the program that you don't plan to debug. This includes
4428 descriptions of functions and external variables, but no information
4429 about local variables and no line numbers.
4431 Level 3 includes extra information, such as all the macro definitions
4432 present in the program. Some debuggers support macro expansion when
4433 you use @option{-g3}.
4435 @option{-gdwarf-2} does not accept a concatenated debug level, because
4436 GCC used to support an option @option{-gdwarf} that meant to generate
4437 debug information in version 1 of the DWARF format (which is very
4438 different from version 2), and it would have been too confusing. That
4439 debug format is long obsolete, but the option cannot be changed now.
4440 Instead use an additional @option{-g@var{level}} option to change the
4441 debug level for DWARF.
4445 Turn off generation of debug info, if leaving out this option would have
4446 generated it, or turn it on at level 2 otherwise. The position of this
4447 argument in the command line does not matter, it takes effect after all
4448 other options are processed, and it does so only once, no matter how
4449 many times it is given. This is mainly intended to be used with
4450 @option{-fcompare-debug}.
4452 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4453 @opindex fdump-final-insns
4454 Dump the final internal representation (RTL) to @var{file}. If the
4455 optional argument is omitted (or if @var{file} is @code{.}), the name
4456 of the dump file will be determined by appending @code{.gkd} to the
4457 compilation output file name.
4459 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4460 @opindex fcompare-debug
4461 @opindex fno-compare-debug
4462 If no error occurs during compilation, run the compiler a second time,
4463 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4464 passed to the second compilation. Dump the final internal
4465 representation in both compilations, and print an error if they differ.
4467 If the equal sign is omitted, the default @option{-gtoggle} is used.
4469 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4470 and nonzero, implicitly enables @option{-fcompare-debug}. If
4471 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4472 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4475 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4476 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4477 of the final representation and the second compilation, preventing even
4478 @env{GCC_COMPARE_DEBUG} from taking effect.
4480 To verify full coverage during @option{-fcompare-debug} testing, set
4481 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4482 which GCC will reject as an invalid option in any actual compilation
4483 (rather than preprocessing, assembly or linking). To get just a
4484 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4485 not overridden} will do.
4487 @item -fcompare-debug-second
4488 @opindex fcompare-debug-second
4489 This option is implicitly passed to the compiler for the second
4490 compilation requested by @option{-fcompare-debug}, along with options to
4491 silence warnings, and omitting other options that would cause
4492 side-effect compiler outputs to files or to the standard output. Dump
4493 files and preserved temporary files are renamed so as to contain the
4494 @code{.gk} additional extension during the second compilation, to avoid
4495 overwriting those generated by the first.
4497 When this option is passed to the compiler driver, it causes the
4498 @emph{first} compilation to be skipped, which makes it useful for little
4499 other than debugging the compiler proper.
4501 @item -feliminate-dwarf2-dups
4502 @opindex feliminate-dwarf2-dups
4503 Compress DWARF2 debugging information by eliminating duplicated
4504 information about each symbol. This option only makes sense when
4505 generating DWARF2 debugging information with @option{-gdwarf-2}.
4507 @item -femit-struct-debug-baseonly
4508 Emit debug information for struct-like types
4509 only when the base name of the compilation source file
4510 matches the base name of file in which the struct was defined.
4512 This option substantially reduces the size of debugging information,
4513 but at significant potential loss in type information to the debugger.
4514 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4515 See @option{-femit-struct-debug-detailed} for more detailed control.
4517 This option works only with DWARF 2.
4519 @item -femit-struct-debug-reduced
4520 Emit debug information for struct-like types
4521 only when the base name of the compilation source file
4522 matches the base name of file in which the type was defined,
4523 unless the struct is a template or defined in a system header.
4525 This option significantly reduces the size of debugging information,
4526 with some potential loss in type information to the debugger.
4527 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4528 See @option{-femit-struct-debug-detailed} for more detailed control.
4530 This option works only with DWARF 2.
4532 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4533 Specify the struct-like types
4534 for which the compiler will generate debug information.
4535 The intent is to reduce duplicate struct debug information
4536 between different object files within the same program.
4538 This option is a detailed version of
4539 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4540 which will serve for most needs.
4542 A specification has the syntax
4543 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4545 The optional first word limits the specification to
4546 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4547 A struct type is used directly when it is the type of a variable, member.
4548 Indirect uses arise through pointers to structs.
4549 That is, when use of an incomplete struct would be legal, the use is indirect.
4551 @samp{struct one direct; struct two * indirect;}.
4553 The optional second word limits the specification to
4554 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4555 Generic structs are a bit complicated to explain.
4556 For C++, these are non-explicit specializations of template classes,
4557 or non-template classes within the above.
4558 Other programming languages have generics,
4559 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4561 The third word specifies the source files for those
4562 structs for which the compiler will emit debug information.
4563 The values @samp{none} and @samp{any} have the normal meaning.
4564 The value @samp{base} means that
4565 the base of name of the file in which the type declaration appears
4566 must match the base of the name of the main compilation file.
4567 In practice, this means that
4568 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4569 but types declared in other header will not.
4570 The value @samp{sys} means those types satisfying @samp{base}
4571 or declared in system or compiler headers.
4573 You may need to experiment to determine the best settings for your application.
4575 The default is @samp{-femit-struct-debug-detailed=all}.
4577 This option works only with DWARF 2.
4579 @item -fno-merge-debug-strings
4580 @opindex fmerge-debug-strings
4581 @opindex fno-merge-debug-strings
4582 Direct the linker to not merge together strings in the debugging
4583 information which are identical in different object files. Merging is
4584 not supported by all assemblers or linkers. Merging decreases the size
4585 of the debug information in the output file at the cost of increasing
4586 link processing time. Merging is enabled by default.
4588 @item -fdebug-prefix-map=@var{old}=@var{new}
4589 @opindex fdebug-prefix-map
4590 When compiling files in directory @file{@var{old}}, record debugging
4591 information describing them as in @file{@var{new}} instead.
4593 @item -fno-dwarf2-cfi-asm
4594 @opindex fdwarf2-cfi-asm
4595 @opindex fno-dwarf2-cfi-asm
4596 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4597 instead of using GAS @code{.cfi_*} directives.
4599 @cindex @command{prof}
4602 Generate extra code to write profile information suitable for the
4603 analysis program @command{prof}. You must use this option when compiling
4604 the source files you want data about, and you must also use it when
4607 @cindex @command{gprof}
4610 Generate extra code to write profile information suitable for the
4611 analysis program @command{gprof}. You must use this option when compiling
4612 the source files you want data about, and you must also use it when
4617 Makes the compiler print out each function name as it is compiled, and
4618 print some statistics about each pass when it finishes.
4621 @opindex ftime-report
4622 Makes the compiler print some statistics about the time consumed by each
4623 pass when it finishes.
4626 @opindex fmem-report
4627 Makes the compiler print some statistics about permanent memory
4628 allocation when it finishes.
4630 @item -fpre-ipa-mem-report
4631 @opindex fpre-ipa-mem-report
4632 @item -fpost-ipa-mem-report
4633 @opindex fpost-ipa-mem-report
4634 Makes the compiler print some statistics about permanent memory
4635 allocation before or after interprocedural optimization.
4637 @item -fprofile-arcs
4638 @opindex fprofile-arcs
4639 Add code so that program flow @dfn{arcs} are instrumented. During
4640 execution the program records how many times each branch and call is
4641 executed and how many times it is taken or returns. When the compiled
4642 program exits it saves this data to a file called
4643 @file{@var{auxname}.gcda} for each source file. The data may be used for
4644 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4645 test coverage analysis (@option{-ftest-coverage}). Each object file's
4646 @var{auxname} is generated from the name of the output file, if
4647 explicitly specified and it is not the final executable, otherwise it is
4648 the basename of the source file. In both cases any suffix is removed
4649 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4650 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4651 @xref{Cross-profiling}.
4653 @cindex @command{gcov}
4657 This option is used to compile and link code instrumented for coverage
4658 analysis. The option is a synonym for @option{-fprofile-arcs}
4659 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4660 linking). See the documentation for those options for more details.
4665 Compile the source files with @option{-fprofile-arcs} plus optimization
4666 and code generation options. For test coverage analysis, use the
4667 additional @option{-ftest-coverage} option. You do not need to profile
4668 every source file in a program.
4671 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4672 (the latter implies the former).
4675 Run the program on a representative workload to generate the arc profile
4676 information. This may be repeated any number of times. You can run
4677 concurrent instances of your program, and provided that the file system
4678 supports locking, the data files will be correctly updated. Also
4679 @code{fork} calls are detected and correctly handled (double counting
4683 For profile-directed optimizations, compile the source files again with
4684 the same optimization and code generation options plus
4685 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4686 Control Optimization}).
4689 For test coverage analysis, use @command{gcov} to produce human readable
4690 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4691 @command{gcov} documentation for further information.
4695 With @option{-fprofile-arcs}, for each function of your program GCC
4696 creates a program flow graph, then finds a spanning tree for the graph.
4697 Only arcs that are not on the spanning tree have to be instrumented: the
4698 compiler adds code to count the number of times that these arcs are
4699 executed. When an arc is the only exit or only entrance to a block, the
4700 instrumentation code can be added to the block; otherwise, a new basic
4701 block must be created to hold the instrumentation code.
4704 @item -ftest-coverage
4705 @opindex ftest-coverage
4706 Produce a notes file that the @command{gcov} code-coverage utility
4707 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4708 show program coverage. Each source file's note file is called
4709 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4710 above for a description of @var{auxname} and instructions on how to
4711 generate test coverage data. Coverage data will match the source files
4712 more closely, if you do not optimize.
4714 @item -fdbg-cnt-list
4715 @opindex fdbg-cnt-list
4716 Print the name and the counter upperbound for all debug counters.
4718 @item -fdbg-cnt=@var{counter-value-list}
4720 Set the internal debug counter upperbound. @var{counter-value-list}
4721 is a comma-separated list of @var{name}:@var{value} pairs
4722 which sets the upperbound of each debug counter @var{name} to @var{value}.
4723 All debug counters have the initial upperbound of @var{UINT_MAX},
4724 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4725 e.g. With -fdbg-cnt=dce:10,tail_call:0
4726 dbg_cnt(dce) will return true only for first 10 invocations
4727 and dbg_cnt(tail_call) will return false always.
4729 @item -d@var{letters}
4730 @itemx -fdump-rtl-@var{pass}
4732 Says to make debugging dumps during compilation at times specified by
4733 @var{letters}. This is used for debugging the RTL-based passes of the
4734 compiler. The file names for most of the dumps are made by appending
4735 a pass number and a word to the @var{dumpname}, and the files are
4736 created in the directory of the output file. @var{dumpname} is
4737 generated from the name of the output file, if explicitly specified
4738 and it is not an executable, otherwise it is the basename of the
4739 source file. These switches may have different effects when
4740 @option{-E} is used for preprocessing.
4742 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4743 @option{-d} option @var{letters}. Here are the possible
4744 letters for use in @var{pass} and @var{letters}, and their meanings:
4748 @item -fdump-rtl-alignments
4749 @opindex fdump-rtl-alignments
4750 Dump after branch alignments have been computed.
4752 @item -fdump-rtl-asmcons
4753 @opindex fdump-rtl-asmcons
4754 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4756 @item -fdump-rtl-auto_inc_dec
4757 @opindex fdump-rtl-auto_inc_dec
4758 Dump after auto-inc-dec discovery. This pass is only run on
4759 architectures that have auto inc or auto dec instructions.
4761 @item -fdump-rtl-barriers
4762 @opindex fdump-rtl-barriers
4763 Dump after cleaning up the barrier instructions.
4765 @item -fdump-rtl-bbpart
4766 @opindex fdump-rtl-bbpart
4767 Dump after partitioning hot and cold basic blocks.
4769 @item -fdump-rtl-bbro
4770 @opindex fdump-rtl-bbro
4771 Dump after block reordering.
4773 @item -fdump-rtl-btl1
4774 @itemx -fdump-rtl-btl2
4775 @opindex fdump-rtl-btl2
4776 @opindex fdump-rtl-btl2
4777 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4778 after the two branch
4779 target load optimization passes.
4781 @item -fdump-rtl-bypass
4782 @opindex fdump-rtl-bypass
4783 Dump after jump bypassing and control flow optimizations.
4785 @item -fdump-rtl-combine
4786 @opindex fdump-rtl-combine
4787 Dump after the RTL instruction combination pass.
4789 @item -fdump-rtl-compgotos
4790 @opindex fdump-rtl-compgotos
4791 Dump after duplicating the computed gotos.
4793 @item -fdump-rtl-ce1
4794 @itemx -fdump-rtl-ce2
4795 @itemx -fdump-rtl-ce3
4796 @opindex fdump-rtl-ce1
4797 @opindex fdump-rtl-ce2
4798 @opindex fdump-rtl-ce3
4799 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4800 @option{-fdump-rtl-ce3} enable dumping after the three
4801 if conversion passes.
4803 @itemx -fdump-rtl-cprop_hardreg
4804 @opindex fdump-rtl-cprop_hardreg
4805 Dump after hard register copy propagation.
4807 @itemx -fdump-rtl-csa
4808 @opindex fdump-rtl-csa
4809 Dump after combining stack adjustments.
4811 @item -fdump-rtl-cse1
4812 @itemx -fdump-rtl-cse2
4813 @opindex fdump-rtl-cse1
4814 @opindex fdump-rtl-cse2
4815 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4816 the two common sub-expression elimination passes.
4818 @itemx -fdump-rtl-dce
4819 @opindex fdump-rtl-dce
4820 Dump after the standalone dead code elimination passes.
4822 @itemx -fdump-rtl-dbr
4823 @opindex fdump-rtl-dbr
4824 Dump after delayed branch scheduling.
4826 @item -fdump-rtl-dce1
4827 @itemx -fdump-rtl-dce2
4828 @opindex fdump-rtl-dce1
4829 @opindex fdump-rtl-dce2
4830 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4831 the two dead store elimination passes.
4834 @opindex fdump-rtl-eh
4835 Dump after finalization of EH handling code.
4837 @item -fdump-rtl-eh_ranges
4838 @opindex fdump-rtl-eh_ranges
4839 Dump after conversion of EH handling range regions.
4841 @item -fdump-rtl-expand
4842 @opindex fdump-rtl-expand
4843 Dump after RTL generation.
4845 @item -fdump-rtl-fwprop1
4846 @itemx -fdump-rtl-fwprop2
4847 @opindex fdump-rtl-fwprop1
4848 @opindex fdump-rtl-fwprop2
4849 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4850 dumping after the two forward propagation passes.
4852 @item -fdump-rtl-gcse1
4853 @itemx -fdump-rtl-gcse2
4854 @opindex fdump-rtl-gcse1
4855 @opindex fdump-rtl-gcse2
4856 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4857 after global common subexpression elimination.
4859 @item -fdump-rtl-init-regs
4860 @opindex fdump-rtl-init-regs
4861 Dump after the initialization of the registers.
4863 @item -fdump-rtl-initvals
4864 @opindex fdump-rtl-initvals
4865 Dump after the computation of the initial value sets.
4867 @itemx -fdump-rtl-into_cfglayout
4868 @opindex fdump-rtl-into_cfglayout
4869 Dump after converting to cfglayout mode.
4871 @item -fdump-rtl-ira
4872 @opindex fdump-rtl-ira
4873 Dump after iterated register allocation.
4875 @item -fdump-rtl-jump
4876 @opindex fdump-rtl-jump
4877 Dump after the second jump optimization.
4879 @item -fdump-rtl-loop2
4880 @opindex fdump-rtl-loop2
4881 @option{-fdump-rtl-loop2} enables dumping after the rtl
4882 loop optimization passes.
4884 @item -fdump-rtl-mach
4885 @opindex fdump-rtl-mach
4886 Dump after performing the machine dependent reorganization pass, if that
4889 @item -fdump-rtl-mode_sw
4890 @opindex fdump-rtl-mode_sw
4891 Dump after removing redundant mode switches.
4893 @item -fdump-rtl-rnreg
4894 @opindex fdump-rtl-rnreg
4895 Dump after register renumbering.
4897 @itemx -fdump-rtl-outof_cfglayout
4898 @opindex fdump-rtl-outof_cfglayout
4899 Dump after converting from cfglayout mode.
4901 @item -fdump-rtl-peephole2
4902 @opindex fdump-rtl-peephole2
4903 Dump after the peephole pass.
4905 @item -fdump-rtl-postreload
4906 @opindex fdump-rtl-postreload
4907 Dump after post-reload optimizations.
4909 @itemx -fdump-rtl-pro_and_epilogue
4910 @opindex fdump-rtl-pro_and_epilogue
4911 Dump after generating the function pro and epilogues.
4913 @item -fdump-rtl-regmove
4914 @opindex fdump-rtl-regmove
4915 Dump after the register move pass.
4917 @item -fdump-rtl-sched1
4918 @itemx -fdump-rtl-sched2
4919 @opindex fdump-rtl-sched1
4920 @opindex fdump-rtl-sched2
4921 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4922 after the basic block scheduling passes.
4924 @item -fdump-rtl-see
4925 @opindex fdump-rtl-see
4926 Dump after sign extension elimination.
4928 @item -fdump-rtl-seqabstr
4929 @opindex fdump-rtl-seqabstr
4930 Dump after common sequence discovery.
4932 @item -fdump-rtl-shorten
4933 @opindex fdump-rtl-shorten
4934 Dump after shortening branches.
4936 @item -fdump-rtl-sibling
4937 @opindex fdump-rtl-sibling
4938 Dump after sibling call optimizations.
4940 @item -fdump-rtl-split1
4941 @itemx -fdump-rtl-split2
4942 @itemx -fdump-rtl-split3
4943 @itemx -fdump-rtl-split4
4944 @itemx -fdump-rtl-split5
4945 @opindex fdump-rtl-split1
4946 @opindex fdump-rtl-split2
4947 @opindex fdump-rtl-split3
4948 @opindex fdump-rtl-split4
4949 @opindex fdump-rtl-split5
4950 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4951 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4952 @option{-fdump-rtl-split5} enable dumping after five rounds of
4953 instruction splitting.
4955 @item -fdump-rtl-sms
4956 @opindex fdump-rtl-sms
4957 Dump after modulo scheduling. This pass is only run on some
4960 @item -fdump-rtl-stack
4961 @opindex fdump-rtl-stack
4962 Dump after conversion from GCC's "flat register file" registers to the
4963 x87's stack-like registers. This pass is only run on x86 variants.
4965 @item -fdump-rtl-subreg1
4966 @itemx -fdump-rtl-subreg2
4967 @opindex fdump-rtl-subreg1
4968 @opindex fdump-rtl-subreg2
4969 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4970 the two subreg expansion passes.
4972 @item -fdump-rtl-unshare
4973 @opindex fdump-rtl-unshare
4974 Dump after all rtl has been unshared.
4976 @item -fdump-rtl-vartrack
4977 @opindex fdump-rtl-vartrack
4978 Dump after variable tracking.
4980 @item -fdump-rtl-vregs
4981 @opindex fdump-rtl-vregs
4982 Dump after converting virtual registers to hard registers.
4984 @item -fdump-rtl-web
4985 @opindex fdump-rtl-web
4986 Dump after live range splitting.
4988 @item -fdump-rtl-regclass
4989 @itemx -fdump-rtl-subregs_of_mode_init
4990 @itemx -fdump-rtl-subregs_of_mode_finish
4991 @itemx -fdump-rtl-dfinit
4992 @itemx -fdump-rtl-dfinish
4993 @opindex fdump-rtl-regclass
4994 @opindex fdump-rtl-subregs_of_mode_init
4995 @opindex fdump-rtl-subregs_of_mode_finish
4996 @opindex fdump-rtl-dfinit
4997 @opindex fdump-rtl-dfinish
4998 These dumps are defined but always produce empty files.
5000 @item -fdump-rtl-all
5001 @opindex fdump-rtl-all
5002 Produce all the dumps listed above.
5006 Annotate the assembler output with miscellaneous debugging information.
5010 Dump all macro definitions, at the end of preprocessing, in addition to
5015 Produce a core dump whenever an error occurs.
5019 Print statistics on memory usage, at the end of the run, to
5024 Annotate the assembler output with a comment indicating which
5025 pattern and alternative was used. The length of each instruction is
5030 Dump the RTL in the assembler output as a comment before each instruction.
5031 Also turns on @option{-dp} annotation.
5035 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5036 dump a representation of the control flow graph suitable for viewing with VCG
5037 to @file{@var{file}.@var{pass}.vcg}.
5041 Just generate RTL for a function instead of compiling it. Usually used
5042 with @option{-fdump-rtl-expand}.
5046 Dump debugging information during parsing, to standard error.
5050 @opindex fdump-noaddr
5051 When doing debugging dumps, suppress address output. This makes it more
5052 feasible to use diff on debugging dumps for compiler invocations with
5053 different compiler binaries and/or different
5054 text / bss / data / heap / stack / dso start locations.
5056 @item -fdump-unnumbered
5057 @opindex fdump-unnumbered
5058 When doing debugging dumps, suppress instruction numbers and address output.
5059 This makes it more feasible to use diff on debugging dumps for compiler
5060 invocations with different options, in particular with and without
5063 @item -fdump-unnumbered-links
5064 @opindex fdump-unnumbered-links
5065 When doing debugging dumps (see @option{-d} option above), suppress
5066 instruction numbers for the links to the previous and next instructions
5069 @item -fdump-translation-unit @r{(C++ only)}
5070 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5071 @opindex fdump-translation-unit
5072 Dump a representation of the tree structure for the entire translation
5073 unit to a file. The file name is made by appending @file{.tu} to the
5074 source file name, and the file is created in the same directory as the
5075 output file. If the @samp{-@var{options}} form is used, @var{options}
5076 controls the details of the dump as described for the
5077 @option{-fdump-tree} options.
5079 @item -fdump-class-hierarchy @r{(C++ only)}
5080 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5081 @opindex fdump-class-hierarchy
5082 Dump a representation of each class's hierarchy and virtual function
5083 table layout to a file. The file name is made by appending
5084 @file{.class} to the source file name, and the file is created in the
5085 same directory as the output file. If the @samp{-@var{options}} form
5086 is used, @var{options} controls the details of the dump as described
5087 for the @option{-fdump-tree} options.
5089 @item -fdump-ipa-@var{switch}
5091 Control the dumping at various stages of inter-procedural analysis
5092 language tree to a file. The file name is generated by appending a
5093 switch specific suffix to the source file name, and the file is created
5094 in the same directory as the output file. The following dumps are
5099 Enables all inter-procedural analysis dumps.
5102 Dumps information about call-graph optimization, unused function removal,
5103 and inlining decisions.
5106 Dump after function inlining.
5110 @item -fdump-statistics-@var{option}
5111 @opindex fdump-statistics
5112 Enable and control dumping of pass statistics in a separate file. The
5113 file name is generated by appending a suffix ending in
5114 @samp{.statistics} to the source file name, and the file is created in
5115 the same directory as the output file. If the @samp{-@var{option}}
5116 form is used, @samp{-stats} will cause counters to be summed over the
5117 whole compilation unit while @samp{-details} will dump every event as
5118 the passes generate them. The default with no option is to sum
5119 counters for each function compiled.
5121 @item -fdump-tree-@var{switch}
5122 @itemx -fdump-tree-@var{switch}-@var{options}
5124 Control the dumping at various stages of processing the intermediate
5125 language tree to a file. The file name is generated by appending a
5126 switch specific suffix to the source file name, and the file is
5127 created in the same directory as the output file. If the
5128 @samp{-@var{options}} form is used, @var{options} is a list of
5129 @samp{-} separated options that control the details of the dump. Not
5130 all options are applicable to all dumps, those which are not
5131 meaningful will be ignored. The following options are available
5135 Print the address of each node. Usually this is not meaningful as it
5136 changes according to the environment and source file. Its primary use
5137 is for tying up a dump file with a debug environment.
5139 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5140 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5141 use working backward from mangled names in the assembly file.
5143 Inhibit dumping of members of a scope or body of a function merely
5144 because that scope has been reached. Only dump such items when they
5145 are directly reachable by some other path. When dumping pretty-printed
5146 trees, this option inhibits dumping the bodies of control structures.
5148 Print a raw representation of the tree. By default, trees are
5149 pretty-printed into a C-like representation.
5151 Enable more detailed dumps (not honored by every dump option).
5153 Enable dumping various statistics about the pass (not honored by every dump
5156 Enable showing basic block boundaries (disabled in raw dumps).
5158 Enable showing virtual operands for every statement.
5160 Enable showing line numbers for statements.
5162 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5164 Enable showing the tree dump for each statement.
5166 Enable showing the EH region number holding each statement.
5168 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5169 and @option{lineno}.
5172 The following tree dumps are possible:
5176 @opindex fdump-tree-original
5177 Dump before any tree based optimization, to @file{@var{file}.original}.
5180 @opindex fdump-tree-optimized
5181 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5184 @opindex fdump-tree-gimple
5185 Dump each function before and after the gimplification pass to a file. The
5186 file name is made by appending @file{.gimple} to the source file name.
5189 @opindex fdump-tree-cfg
5190 Dump the control flow graph of each function to a file. The file name is
5191 made by appending @file{.cfg} to the source file name.
5194 @opindex fdump-tree-vcg
5195 Dump the control flow graph of each function to a file in VCG format. The
5196 file name is made by appending @file{.vcg} to the source file name. Note
5197 that if the file contains more than one function, the generated file cannot
5198 be used directly by VCG@. You will need to cut and paste each function's
5199 graph into its own separate file first.
5202 @opindex fdump-tree-ch
5203 Dump each function after copying loop headers. The file name is made by
5204 appending @file{.ch} to the source file name.
5207 @opindex fdump-tree-ssa
5208 Dump SSA related information to a file. The file name is made by appending
5209 @file{.ssa} to the source file name.
5212 @opindex fdump-tree-alias
5213 Dump aliasing information for each function. The file name is made by
5214 appending @file{.alias} to the source file name.
5217 @opindex fdump-tree-ccp
5218 Dump each function after CCP@. The file name is made by appending
5219 @file{.ccp} to the source file name.
5222 @opindex fdump-tree-storeccp
5223 Dump each function after STORE-CCP@. The file name is made by appending
5224 @file{.storeccp} to the source file name.
5227 @opindex fdump-tree-pre
5228 Dump trees after partial redundancy elimination. The file name is made
5229 by appending @file{.pre} to the source file name.
5232 @opindex fdump-tree-fre
5233 Dump trees after full redundancy elimination. The file name is made
5234 by appending @file{.fre} to the source file name.
5237 @opindex fdump-tree-copyprop
5238 Dump trees after copy propagation. The file name is made
5239 by appending @file{.copyprop} to the source file name.
5241 @item store_copyprop
5242 @opindex fdump-tree-store_copyprop
5243 Dump trees after store copy-propagation. The file name is made
5244 by appending @file{.store_copyprop} to the source file name.
5247 @opindex fdump-tree-dce
5248 Dump each function after dead code elimination. The file name is made by
5249 appending @file{.dce} to the source file name.
5252 @opindex fdump-tree-mudflap
5253 Dump each function after adding mudflap instrumentation. The file name is
5254 made by appending @file{.mudflap} to the source file name.
5257 @opindex fdump-tree-sra
5258 Dump each function after performing scalar replacement of aggregates. The
5259 file name is made by appending @file{.sra} to the source file name.
5262 @opindex fdump-tree-sink
5263 Dump each function after performing code sinking. The file name is made
5264 by appending @file{.sink} to the source file name.
5267 @opindex fdump-tree-dom
5268 Dump each function after applying dominator tree optimizations. The file
5269 name is made by appending @file{.dom} to the source file name.
5272 @opindex fdump-tree-dse
5273 Dump each function after applying dead store elimination. The file
5274 name is made by appending @file{.dse} to the source file name.
5277 @opindex fdump-tree-phiopt
5278 Dump each function after optimizing PHI nodes into straightline code. The file
5279 name is made by appending @file{.phiopt} to the source file name.
5282 @opindex fdump-tree-forwprop
5283 Dump each function after forward propagating single use variables. The file
5284 name is made by appending @file{.forwprop} to the source file name.
5287 @opindex fdump-tree-copyrename
5288 Dump each function after applying the copy rename optimization. The file
5289 name is made by appending @file{.copyrename} to the source file name.
5292 @opindex fdump-tree-nrv
5293 Dump each function after applying the named return value optimization on
5294 generic trees. The file name is made by appending @file{.nrv} to the source
5298 @opindex fdump-tree-vect
5299 Dump each function after applying vectorization of loops. The file name is
5300 made by appending @file{.vect} to the source file name.
5303 @opindex fdump-tree-vrp
5304 Dump each function after Value Range Propagation (VRP). The file name
5305 is made by appending @file{.vrp} to the source file name.
5308 @opindex fdump-tree-all
5309 Enable all the available tree dumps with the flags provided in this option.
5312 @item -ftree-vectorizer-verbose=@var{n}
5313 @opindex ftree-vectorizer-verbose
5314 This option controls the amount of debugging output the vectorizer prints.
5315 This information is written to standard error, unless
5316 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5317 in which case it is output to the usual dump listing file, @file{.vect}.
5318 For @var{n}=0 no diagnostic information is reported.
5319 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5320 and the total number of loops that got vectorized.
5321 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5322 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5323 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5324 level that @option{-fdump-tree-vect-stats} uses.
5325 Higher verbosity levels mean either more information dumped for each
5326 reported loop, or same amount of information reported for more loops:
5327 If @var{n}=3, alignment related information is added to the reports.
5328 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5329 memory access-patterns) is added to the reports.
5330 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5331 that did not pass the first analysis phase (i.e., may not be countable, or
5332 may have complicated control-flow).
5333 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5334 For @var{n}=7, all the information the vectorizer generates during its
5335 analysis and transformation is reported. This is the same verbosity level
5336 that @option{-fdump-tree-vect-details} uses.
5338 @item -frandom-seed=@var{string}
5339 @opindex frandom-seed
5340 This option provides a seed that GCC uses when it would otherwise use
5341 random numbers. It is used to generate certain symbol names
5342 that have to be different in every compiled file. It is also used to
5343 place unique stamps in coverage data files and the object files that
5344 produce them. You can use the @option{-frandom-seed} option to produce
5345 reproducibly identical object files.
5347 The @var{string} should be different for every file you compile.
5349 @item -fsched-verbose=@var{n}
5350 @opindex fsched-verbose
5351 On targets that use instruction scheduling, this option controls the
5352 amount of debugging output the scheduler prints. This information is
5353 written to standard error, unless @option{-fdump-rtl-sched1} or
5354 @option{-fdump-rtl-sched2} is specified, in which case it is output
5355 to the usual dump listing file, @file{.sched} or @file{.sched2}
5356 respectively. However for @var{n} greater than nine, the output is
5357 always printed to standard error.
5359 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5360 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5361 For @var{n} greater than one, it also output basic block probabilities,
5362 detailed ready list information and unit/insn info. For @var{n} greater
5363 than two, it includes RTL at abort point, control-flow and regions info.
5364 And for @var{n} over four, @option{-fsched-verbose} also includes
5368 @itemx -save-temps=cwd
5370 Store the usual ``temporary'' intermediate files permanently; place them
5371 in the current directory and name them based on the source file. Thus,
5372 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5373 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5374 preprocessed @file{foo.i} output file even though the compiler now
5375 normally uses an integrated preprocessor.
5377 When used in combination with the @option{-x} command line option,
5378 @option{-save-temps} is sensible enough to avoid over writing an
5379 input source file with the same extension as an intermediate file.
5380 The corresponding intermediate file may be obtained by renaming the
5381 source file before using @option{-save-temps}.
5383 If you invoke GCC in parallel, compiling several different source
5384 files that share a common base name in different subdirectories or the
5385 same source file compiled for multiple output destinations, it is
5386 likely that the different parallel compilers will interfere with each
5387 other, and overwrite the temporary files. For instance:
5390 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5391 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5394 may result in @file{foo.i} and @file{foo.o} being written to
5395 simultaneously by both compilers.
5397 @item -save-temps=obj
5398 @opindex save-temps=obj
5399 Store the usual ``temporary'' intermediate files permanently. If the
5400 @option{-o} option is used, the temporary files are based on the
5401 object file. If the @option{-o} option is not used, the
5402 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5407 gcc -save-temps=obj -c foo.c
5408 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5409 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5412 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5413 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5414 @file{dir2/yfoobar.o}.
5416 @item -time@r{[}=@var{file}@r{]}
5418 Report the CPU time taken by each subprocess in the compilation
5419 sequence. For C source files, this is the compiler proper and assembler
5420 (plus the linker if linking is done).
5422 Without the specification of an output file, the output looks like this:
5429 The first number on each line is the ``user time'', that is time spent
5430 executing the program itself. The second number is ``system time'',
5431 time spent executing operating system routines on behalf of the program.
5432 Both numbers are in seconds.
5434 With the specification of an output file, the output is appended to the
5435 named file, and it looks like this:
5438 0.12 0.01 cc1 @var{options}
5439 0.00 0.01 as @var{options}
5442 The ``user time'' and the ``system time'' are moved before the program
5443 name, and the options passed to the program are displayed, so that one
5444 can later tell what file was being compiled, and with which options.
5446 @item -fvar-tracking
5447 @opindex fvar-tracking
5448 Run variable tracking pass. It computes where variables are stored at each
5449 position in code. Better debugging information is then generated
5450 (if the debugging information format supports this information).
5452 It is enabled by default when compiling with optimization (@option{-Os},
5453 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5454 the debug info format supports it.
5456 @item -fvar-tracking-assignments
5457 @opindex fvar-tracking-assignments
5458 @opindex fno-var-tracking-assignments
5459 Annotate assignments to user variables early in the compilation and
5460 attempt to carry the annotations over throughout the compilation all the
5461 way to the end, in an attempt to improve debug information while
5462 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5464 It can be enabled even if var-tracking is disabled, in which case
5465 annotations will be created and maintained, but discarded at the end.
5467 @item -fvar-tracking-assignments-toggle
5468 @opindex fvar-tracking-assignments-toggle
5469 @opindex fno-var-tracking-assignments-toggle
5470 Toggle @option{-fvar-tracking-assignments}, in the same way that
5471 @option{-gtoggle} toggles @option{-g}.
5473 @item -print-file-name=@var{library}
5474 @opindex print-file-name
5475 Print the full absolute name of the library file @var{library} that
5476 would be used when linking---and don't do anything else. With this
5477 option, GCC does not compile or link anything; it just prints the
5480 @item -print-multi-directory
5481 @opindex print-multi-directory
5482 Print the directory name corresponding to the multilib selected by any
5483 other switches present in the command line. This directory is supposed
5484 to exist in @env{GCC_EXEC_PREFIX}.
5486 @item -print-multi-lib
5487 @opindex print-multi-lib
5488 Print the mapping from multilib directory names to compiler switches
5489 that enable them. The directory name is separated from the switches by
5490 @samp{;}, and each switch starts with an @samp{@@} instead of the
5491 @samp{-}, without spaces between multiple switches. This is supposed to
5492 ease shell-processing.
5494 @item -print-prog-name=@var{program}
5495 @opindex print-prog-name
5496 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5498 @item -print-libgcc-file-name
5499 @opindex print-libgcc-file-name
5500 Same as @option{-print-file-name=libgcc.a}.
5502 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5503 but you do want to link with @file{libgcc.a}. You can do
5506 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5509 @item -print-search-dirs
5510 @opindex print-search-dirs
5511 Print the name of the configured installation directory and a list of
5512 program and library directories @command{gcc} will search---and don't do anything else.
5514 This is useful when @command{gcc} prints the error message
5515 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5516 To resolve this you either need to put @file{cpp0} and the other compiler
5517 components where @command{gcc} expects to find them, or you can set the environment
5518 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5519 Don't forget the trailing @samp{/}.
5520 @xref{Environment Variables}.
5522 @item -print-sysroot
5523 @opindex print-sysroot
5524 Print the target sysroot directory that will be used during
5525 compilation. This is the target sysroot specified either at configure
5526 time or using the @option{--sysroot} option, possibly with an extra
5527 suffix that depends on compilation options. If no target sysroot is
5528 specified, the option prints nothing.
5530 @item -print-sysroot-headers-suffix
5531 @opindex print-sysroot-headers-suffix
5532 Print the suffix added to the target sysroot when searching for
5533 headers, or give an error if the compiler is not configured with such
5534 a suffix---and don't do anything else.
5537 @opindex dumpmachine
5538 Print the compiler's target machine (for example,
5539 @samp{i686-pc-linux-gnu})---and don't do anything else.
5542 @opindex dumpversion
5543 Print the compiler version (for example, @samp{3.0})---and don't do
5548 Print the compiler's built-in specs---and don't do anything else. (This
5549 is used when GCC itself is being built.) @xref{Spec Files}.
5551 @item -feliminate-unused-debug-types
5552 @opindex feliminate-unused-debug-types
5553 Normally, when producing DWARF2 output, GCC will emit debugging
5554 information for all types declared in a compilation
5555 unit, regardless of whether or not they are actually used
5556 in that compilation unit. Sometimes this is useful, such as
5557 if, in the debugger, you want to cast a value to a type that is
5558 not actually used in your program (but is declared). More often,
5559 however, this results in a significant amount of wasted space.
5560 With this option, GCC will avoid producing debug symbol output
5561 for types that are nowhere used in the source file being compiled.
5564 @node Optimize Options
5565 @section Options That Control Optimization
5566 @cindex optimize options
5567 @cindex options, optimization
5569 These options control various sorts of optimizations.
5571 Without any optimization option, the compiler's goal is to reduce the
5572 cost of compilation and to make debugging produce the expected
5573 results. Statements are independent: if you stop the program with a
5574 breakpoint between statements, you can then assign a new value to any
5575 variable or change the program counter to any other statement in the
5576 function and get exactly the results you would expect from the source
5579 Turning on optimization flags makes the compiler attempt to improve
5580 the performance and/or code size at the expense of compilation time
5581 and possibly the ability to debug the program.
5583 The compiler performs optimization based on the knowledge it has of the
5584 program. Compiling multiple files at once to a single output file mode allows
5585 the compiler to use information gained from all of the files when compiling
5588 Not all optimizations are controlled directly by a flag. Only
5589 optimizations that have a flag are listed in this section.
5591 Depending on the target and how GCC was configured, a slightly different
5592 set of optimizations may be enabled at each @option{-O} level than
5593 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5594 to find out the exact set of optimizations that are enabled at each level.
5595 @xref{Overall Options}, for examples.
5602 Optimize. Optimizing compilation takes somewhat more time, and a lot
5603 more memory for a large function.
5605 With @option{-O}, the compiler tries to reduce code size and execution
5606 time, without performing any optimizations that take a great deal of
5609 @option{-O} turns on the following optimization flags:
5612 -fcprop-registers @gol
5615 -fdelayed-branch @gol
5617 -fguess-branch-probability @gol
5618 -fif-conversion2 @gol
5619 -fif-conversion @gol
5620 -fipa-pure-const @gol
5621 -fipa-reference @gol
5623 -fsplit-wide-types @gol
5624 -ftree-builtin-call-dce @gol
5627 -ftree-copyrename @gol
5629 -ftree-dominator-opts @gol
5631 -ftree-forwprop @gol
5639 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5640 where doing so does not interfere with debugging.
5644 Optimize even more. GCC performs nearly all supported optimizations
5645 that do not involve a space-speed tradeoff.
5646 As compared to @option{-O}, this option increases both compilation time
5647 and the performance of the generated code.
5649 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5650 also turns on the following optimization flags:
5651 @gccoptlist{-fthread-jumps @gol
5652 -falign-functions -falign-jumps @gol
5653 -falign-loops -falign-labels @gol
5656 -fcse-follow-jumps -fcse-skip-blocks @gol
5657 -fdelete-null-pointer-checks @gol
5658 -fexpensive-optimizations @gol
5659 -fgcse -fgcse-lm @gol
5660 -finline-small-functions @gol
5661 -findirect-inlining @gol
5662 -foptimize-sibling-calls @gol
5665 -freorder-blocks -freorder-functions @gol
5666 -frerun-cse-after-loop @gol
5667 -fsched-interblock -fsched-spec @gol
5668 -fschedule-insns -fschedule-insns2 @gol
5669 -fstrict-aliasing -fstrict-overflow @gol
5670 -ftree-switch-conversion @gol
5674 Please note the warning under @option{-fgcse} about
5675 invoking @option{-O2} on programs that use computed gotos.
5679 Optimize yet more. @option{-O3} turns on all optimizations specified
5680 by @option{-O2} and also turns on the @option{-finline-functions},
5681 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5682 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5686 Reduce compilation time and make debugging produce the expected
5687 results. This is the default.
5691 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5692 do not typically increase code size. It also performs further
5693 optimizations designed to reduce code size.
5695 @option{-Os} disables the following optimization flags:
5696 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5697 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5698 -fprefetch-loop-arrays -ftree-vect-loop-version}
5700 If you use multiple @option{-O} options, with or without level numbers,
5701 the last such option is the one that is effective.
5704 Options of the form @option{-f@var{flag}} specify machine-independent
5705 flags. Most flags have both positive and negative forms; the negative
5706 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5707 below, only one of the forms is listed---the one you typically will
5708 use. You can figure out the other form by either removing @samp{no-}
5711 The following options control specific optimizations. They are either
5712 activated by @option{-O} options or are related to ones that are. You
5713 can use the following flags in the rare cases when ``fine-tuning'' of
5714 optimizations to be performed is desired.
5717 @item -fno-default-inline
5718 @opindex fno-default-inline
5719 Do not make member functions inline by default merely because they are
5720 defined inside the class scope (C++ only). Otherwise, when you specify
5721 @w{@option{-O}}, member functions defined inside class scope are compiled
5722 inline by default; i.e., you don't need to add @samp{inline} in front of
5723 the member function name.
5725 @item -fno-defer-pop
5726 @opindex fno-defer-pop
5727 Always pop the arguments to each function call as soon as that function
5728 returns. For machines which must pop arguments after a function call,
5729 the compiler normally lets arguments accumulate on the stack for several
5730 function calls and pops them all at once.
5732 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5734 @item -fforward-propagate
5735 @opindex fforward-propagate
5736 Perform a forward propagation pass on RTL@. The pass tries to combine two
5737 instructions and checks if the result can be simplified. If loop unrolling
5738 is active, two passes are performed and the second is scheduled after
5741 This option is enabled by default at optimization levels @option{-O},
5742 @option{-O2}, @option{-O3}, @option{-Os}.
5744 @item -fomit-frame-pointer
5745 @opindex fomit-frame-pointer
5746 Don't keep the frame pointer in a register for functions that
5747 don't need one. This avoids the instructions to save, set up and
5748 restore frame pointers; it also makes an extra register available
5749 in many functions. @strong{It also makes debugging impossible on
5752 On some machines, such as the VAX, this flag has no effect, because
5753 the standard calling sequence automatically handles the frame pointer
5754 and nothing is saved by pretending it doesn't exist. The
5755 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5756 whether a target machine supports this flag. @xref{Registers,,Register
5757 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5759 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5761 @item -foptimize-sibling-calls
5762 @opindex foptimize-sibling-calls
5763 Optimize sibling and tail recursive calls.
5765 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5769 Don't pay attention to the @code{inline} keyword. Normally this option
5770 is used to keep the compiler from expanding any functions inline.
5771 Note that if you are not optimizing, no functions can be expanded inline.
5773 @item -finline-small-functions
5774 @opindex finline-small-functions
5775 Integrate functions into their callers when their body is smaller than expected
5776 function call code (so overall size of program gets smaller). The compiler
5777 heuristically decides which functions are simple enough to be worth integrating
5780 Enabled at level @option{-O2}.
5782 @item -findirect-inlining
5783 @opindex findirect-inlining
5784 Inline also indirect calls that are discovered to be known at compile
5785 time thanks to previous inlining. This option has any effect only
5786 when inlining itself is turned on by the @option{-finline-functions}
5787 or @option{-finline-small-functions} options.
5789 Enabled at level @option{-O2}.
5791 @item -finline-functions
5792 @opindex finline-functions
5793 Integrate all simple functions into their callers. The compiler
5794 heuristically decides which functions are simple enough to be worth
5795 integrating in this way.
5797 If all calls to a given function are integrated, and the function is
5798 declared @code{static}, then the function is normally not output as
5799 assembler code in its own right.
5801 Enabled at level @option{-O3}.
5803 @item -finline-functions-called-once
5804 @opindex finline-functions-called-once
5805 Consider all @code{static} functions called once for inlining into their
5806 caller even if they are not marked @code{inline}. If a call to a given
5807 function is integrated, then the function is not output as assembler code
5810 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5812 @item -fearly-inlining
5813 @opindex fearly-inlining
5814 Inline functions marked by @code{always_inline} and functions whose body seems
5815 smaller than the function call overhead early before doing
5816 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5817 makes profiling significantly cheaper and usually inlining faster on programs
5818 having large chains of nested wrapper functions.
5822 @item -finline-limit=@var{n}
5823 @opindex finline-limit
5824 By default, GCC limits the size of functions that can be inlined. This flag
5825 allows coarse control of this limit. @var{n} is the size of functions that
5826 can be inlined in number of pseudo instructions.
5828 Inlining is actually controlled by a number of parameters, which may be
5829 specified individually by using @option{--param @var{name}=@var{value}}.
5830 The @option{-finline-limit=@var{n}} option sets some of these parameters
5834 @item max-inline-insns-single
5835 is set to @var{n}/2.
5836 @item max-inline-insns-auto
5837 is set to @var{n}/2.
5840 See below for a documentation of the individual
5841 parameters controlling inlining and for the defaults of these parameters.
5843 @emph{Note:} there may be no value to @option{-finline-limit} that results
5844 in default behavior.
5846 @emph{Note:} pseudo instruction represents, in this particular context, an
5847 abstract measurement of function's size. In no way does it represent a count
5848 of assembly instructions and as such its exact meaning might change from one
5849 release to an another.
5851 @item -fkeep-inline-functions
5852 @opindex fkeep-inline-functions
5853 In C, emit @code{static} functions that are declared @code{inline}
5854 into the object file, even if the function has been inlined into all
5855 of its callers. This switch does not affect functions using the
5856 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5857 inline functions into the object file.
5859 @item -fkeep-static-consts
5860 @opindex fkeep-static-consts
5861 Emit variables declared @code{static const} when optimization isn't turned
5862 on, even if the variables aren't referenced.
5864 GCC enables this option by default. If you want to force the compiler to
5865 check if the variable was referenced, regardless of whether or not
5866 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5868 @item -fmerge-constants
5869 @opindex fmerge-constants
5870 Attempt to merge identical constants (string constants and floating point
5871 constants) across compilation units.
5873 This option is the default for optimized compilation if the assembler and
5874 linker support it. Use @option{-fno-merge-constants} to inhibit this
5877 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5879 @item -fmerge-all-constants
5880 @opindex fmerge-all-constants
5881 Attempt to merge identical constants and identical variables.
5883 This option implies @option{-fmerge-constants}. In addition to
5884 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5885 arrays or initialized constant variables with integral or floating point
5886 types. Languages like C or C++ require each variable, including multiple
5887 instances of the same variable in recursive calls, to have distinct locations,
5888 so using this option will result in non-conforming
5891 @item -fmodulo-sched
5892 @opindex fmodulo-sched
5893 Perform swing modulo scheduling immediately before the first scheduling
5894 pass. This pass looks at innermost loops and reorders their
5895 instructions by overlapping different iterations.
5897 @item -fmodulo-sched-allow-regmoves
5898 @opindex fmodulo-sched-allow-regmoves
5899 Perform more aggressive SMS based modulo scheduling with register moves
5900 allowed. By setting this flag certain anti-dependences edges will be
5901 deleted which will trigger the generation of reg-moves based on the
5902 life-range analysis. This option is effective only with
5903 @option{-fmodulo-sched} enabled.
5905 @item -fno-branch-count-reg
5906 @opindex fno-branch-count-reg
5907 Do not use ``decrement and branch'' instructions on a count register,
5908 but instead generate a sequence of instructions that decrement a
5909 register, compare it against zero, then branch based upon the result.
5910 This option is only meaningful on architectures that support such
5911 instructions, which include x86, PowerPC, IA-64 and S/390.
5913 The default is @option{-fbranch-count-reg}.
5915 @item -fno-function-cse
5916 @opindex fno-function-cse
5917 Do not put function addresses in registers; make each instruction that
5918 calls a constant function contain the function's address explicitly.
5920 This option results in less efficient code, but some strange hacks
5921 that alter the assembler output may be confused by the optimizations
5922 performed when this option is not used.
5924 The default is @option{-ffunction-cse}
5926 @item -fno-zero-initialized-in-bss
5927 @opindex fno-zero-initialized-in-bss
5928 If the target supports a BSS section, GCC by default puts variables that
5929 are initialized to zero into BSS@. This can save space in the resulting
5932 This option turns off this behavior because some programs explicitly
5933 rely on variables going to the data section. E.g., so that the
5934 resulting executable can find the beginning of that section and/or make
5935 assumptions based on that.
5937 The default is @option{-fzero-initialized-in-bss}.
5939 @item -fmudflap -fmudflapth -fmudflapir
5943 @cindex bounds checking
5945 For front-ends that support it (C and C++), instrument all risky
5946 pointer/array dereferencing operations, some standard library
5947 string/heap functions, and some other associated constructs with
5948 range/validity tests. Modules so instrumented should be immune to
5949 buffer overflows, invalid heap use, and some other classes of C/C++
5950 programming errors. The instrumentation relies on a separate runtime
5951 library (@file{libmudflap}), which will be linked into a program if
5952 @option{-fmudflap} is given at link time. Run-time behavior of the
5953 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5954 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5957 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5958 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5959 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5960 instrumentation should ignore pointer reads. This produces less
5961 instrumentation (and therefore faster execution) and still provides
5962 some protection against outright memory corrupting writes, but allows
5963 erroneously read data to propagate within a program.
5965 @item -fthread-jumps
5966 @opindex fthread-jumps
5967 Perform optimizations where we check to see if a jump branches to a
5968 location where another comparison subsumed by the first is found. If
5969 so, the first branch is redirected to either the destination of the
5970 second branch or a point immediately following it, depending on whether
5971 the condition is known to be true or false.
5973 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5975 @item -fsplit-wide-types
5976 @opindex fsplit-wide-types
5977 When using a type that occupies multiple registers, such as @code{long
5978 long} on a 32-bit system, split the registers apart and allocate them
5979 independently. This normally generates better code for those types,
5980 but may make debugging more difficult.
5982 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5985 @item -fcse-follow-jumps
5986 @opindex fcse-follow-jumps
5987 In common subexpression elimination (CSE), scan through jump instructions
5988 when the target of the jump is not reached by any other path. For
5989 example, when CSE encounters an @code{if} statement with an
5990 @code{else} clause, CSE will follow the jump when the condition
5993 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5995 @item -fcse-skip-blocks
5996 @opindex fcse-skip-blocks
5997 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5998 follow jumps which conditionally skip over blocks. When CSE
5999 encounters a simple @code{if} statement with no else clause,
6000 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6001 body of the @code{if}.
6003 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6005 @item -frerun-cse-after-loop
6006 @opindex frerun-cse-after-loop
6007 Re-run common subexpression elimination after loop optimizations has been
6010 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6014 Perform a global common subexpression elimination pass.
6015 This pass also performs global constant and copy propagation.
6017 @emph{Note:} When compiling a program using computed gotos, a GCC
6018 extension, you may get better runtime performance if you disable
6019 the global common subexpression elimination pass by adding
6020 @option{-fno-gcse} to the command line.
6022 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6026 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6027 attempt to move loads which are only killed by stores into themselves. This
6028 allows a loop containing a load/store sequence to be changed to a load outside
6029 the loop, and a copy/store within the loop.
6031 Enabled by default when gcse is enabled.
6035 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6036 global common subexpression elimination. This pass will attempt to move
6037 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6038 loops containing a load/store sequence can be changed to a load before
6039 the loop and a store after the loop.
6041 Not enabled at any optimization level.
6045 When @option{-fgcse-las} is enabled, the global common subexpression
6046 elimination pass eliminates redundant loads that come after stores to the
6047 same memory location (both partial and full redundancies).
6049 Not enabled at any optimization level.
6051 @item -fgcse-after-reload
6052 @opindex fgcse-after-reload
6053 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6054 pass is performed after reload. The purpose of this pass is to cleanup
6057 @item -funsafe-loop-optimizations
6058 @opindex funsafe-loop-optimizations
6059 If given, the loop optimizer will assume that loop indices do not
6060 overflow, and that the loops with nontrivial exit condition are not
6061 infinite. This enables a wider range of loop optimizations even if
6062 the loop optimizer itself cannot prove that these assumptions are valid.
6063 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6064 if it finds this kind of loop.
6066 @item -fcrossjumping
6067 @opindex fcrossjumping
6068 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6069 resulting code may or may not perform better than without cross-jumping.
6071 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6073 @item -fauto-inc-dec
6074 @opindex fauto-inc-dec
6075 Combine increments or decrements of addresses with memory accesses.
6076 This pass is always skipped on architectures that do not have
6077 instructions to support this. Enabled by default at @option{-O} and
6078 higher on architectures that support this.
6082 Perform dead code elimination (DCE) on RTL@.
6083 Enabled by default at @option{-O} and higher.
6087 Perform dead store elimination (DSE) on RTL@.
6088 Enabled by default at @option{-O} and higher.
6090 @item -fif-conversion
6091 @opindex fif-conversion
6092 Attempt to transform conditional jumps into branch-less equivalents. This
6093 include use of conditional moves, min, max, set flags and abs instructions, and
6094 some tricks doable by standard arithmetics. The use of conditional execution
6095 on chips where it is available is controlled by @code{if-conversion2}.
6097 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6099 @item -fif-conversion2
6100 @opindex fif-conversion2
6101 Use conditional execution (where available) to transform conditional jumps into
6102 branch-less equivalents.
6104 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6106 @item -fdelete-null-pointer-checks
6107 @opindex fdelete-null-pointer-checks
6108 Assume that programs cannot safely dereference null pointers, and that
6109 no code or data element resides there. This enables simple constant
6110 folding optimizations at all optimization levels. In addition, other
6111 optimization passes in GCC use this flag to control global dataflow
6112 analyses that eliminate useless checks for null pointers; these assume
6113 that if a pointer is checked after it has already been dereferenced,
6116 Note however that in some environments this assumption is not true.
6117 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6118 for programs which depend on that behavior.
6120 Some targets, especially embedded ones, disable this option at all levels.
6121 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6122 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6123 are enabled independently at different optimization levels.
6125 @item -fexpensive-optimizations
6126 @opindex fexpensive-optimizations
6127 Perform a number of minor optimizations that are relatively expensive.
6129 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6131 @item -foptimize-register-move
6133 @opindex foptimize-register-move
6135 Attempt to reassign register numbers in move instructions and as
6136 operands of other simple instructions in order to maximize the amount of
6137 register tying. This is especially helpful on machines with two-operand
6140 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6143 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6145 @item -fira-algorithm=@var{algorithm}
6146 Use specified coloring algorithm for the integrated register
6147 allocator. The @var{algorithm} argument should be @code{priority} or
6148 @code{CB}. The first algorithm specifies Chow's priority coloring,
6149 the second one specifies Chaitin-Briggs coloring. The second
6150 algorithm can be unimplemented for some architectures. If it is
6151 implemented, it is the default because Chaitin-Briggs coloring as a
6152 rule generates a better code.
6154 @item -fira-region=@var{region}
6155 Use specified regions for the integrated register allocator. The
6156 @var{region} argument should be one of @code{all}, @code{mixed}, or
6157 @code{one}. The first value means using all loops as register
6158 allocation regions, the second value which is the default means using
6159 all loops except for loops with small register pressure as the
6160 regions, and third one means using all function as a single region.
6161 The first value can give best result for machines with small size and
6162 irregular register set, the third one results in faster and generates
6163 decent code and the smallest size code, and the default value usually
6164 give the best results in most cases and for most architectures.
6166 @item -fira-coalesce
6167 @opindex fira-coalesce
6168 Do optimistic register coalescing. This option might be profitable for
6169 architectures with big regular register files.
6171 @item -fno-ira-share-save-slots
6172 @opindex fno-ira-share-save-slots
6173 Switch off sharing stack slots used for saving call used hard
6174 registers living through a call. Each hard register will get a
6175 separate stack slot and as a result function stack frame will be
6178 @item -fno-ira-share-spill-slots
6179 @opindex fno-ira-share-spill-slots
6180 Switch off sharing stack slots allocated for pseudo-registers. Each
6181 pseudo-register which did not get a hard register will get a separate
6182 stack slot and as a result function stack frame will be bigger.
6184 @item -fira-verbose=@var{n}
6185 @opindex fira-verbose
6186 Set up how verbose dump file for the integrated register allocator
6187 will be. Default value is 5. If the value is greater or equal to 10,
6188 the dump file will be stderr as if the value were @var{n} minus 10.
6190 @item -fdelayed-branch
6191 @opindex fdelayed-branch
6192 If supported for the target machine, attempt to reorder instructions
6193 to exploit instruction slots available after delayed branch
6196 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6198 @item -fschedule-insns
6199 @opindex fschedule-insns
6200 If supported for the target machine, attempt to reorder instructions to
6201 eliminate execution stalls due to required data being unavailable. This
6202 helps machines that have slow floating point or memory load instructions
6203 by allowing other instructions to be issued until the result of the load
6204 or floating point instruction is required.
6206 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6208 @item -fschedule-insns2
6209 @opindex fschedule-insns2
6210 Similar to @option{-fschedule-insns}, but requests an additional pass of
6211 instruction scheduling after register allocation has been done. This is
6212 especially useful on machines with a relatively small number of
6213 registers and where memory load instructions take more than one cycle.
6215 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6217 @item -fno-sched-interblock
6218 @opindex fno-sched-interblock
6219 Don't schedule instructions across basic blocks. This is normally
6220 enabled by default when scheduling before register allocation, i.e.@:
6221 with @option{-fschedule-insns} or at @option{-O2} or higher.
6223 @item -fno-sched-spec
6224 @opindex fno-sched-spec
6225 Don't allow speculative motion of non-load instructions. This is normally
6226 enabled by default when scheduling before register allocation, i.e.@:
6227 with @option{-fschedule-insns} or at @option{-O2} or higher.
6229 @item -fsched-pressure
6230 @opindex fsched-pressure
6231 Enable register pressure sensitive insn scheduling before the register
6232 allocation. This only makes sense when scheduling before register
6233 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6234 @option{-O2} or higher. Usage of this option can improve the
6235 generated code and decrease its size by preventing register pressure
6236 increase above the number of available hard registers and as a
6237 consequence register spills in the register allocation.
6239 @item -fsched-spec-load
6240 @opindex fsched-spec-load
6241 Allow speculative motion of some load instructions. This only makes
6242 sense when scheduling before register allocation, i.e.@: with
6243 @option{-fschedule-insns} or at @option{-O2} or higher.
6245 @item -fsched-spec-load-dangerous
6246 @opindex fsched-spec-load-dangerous
6247 Allow speculative motion of more load instructions. This only makes
6248 sense when scheduling before register allocation, i.e.@: with
6249 @option{-fschedule-insns} or at @option{-O2} or higher.
6251 @item -fsched-stalled-insns
6252 @itemx -fsched-stalled-insns=@var{n}
6253 @opindex fsched-stalled-insns
6254 Define how many insns (if any) can be moved prematurely from the queue
6255 of stalled insns into the ready list, during the second scheduling pass.
6256 @option{-fno-sched-stalled-insns} means that no insns will be moved
6257 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6258 on how many queued insns can be moved prematurely.
6259 @option{-fsched-stalled-insns} without a value is equivalent to
6260 @option{-fsched-stalled-insns=1}.
6262 @item -fsched-stalled-insns-dep
6263 @itemx -fsched-stalled-insns-dep=@var{n}
6264 @opindex fsched-stalled-insns-dep
6265 Define how many insn groups (cycles) will be examined for a dependency
6266 on a stalled insn that is candidate for premature removal from the queue
6267 of stalled insns. This has an effect only during the second scheduling pass,
6268 and only if @option{-fsched-stalled-insns} is used.
6269 @option{-fno-sched-stalled-insns-dep} is equivalent to
6270 @option{-fsched-stalled-insns-dep=0}.
6271 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6272 @option{-fsched-stalled-insns-dep=1}.
6274 @item -fsched2-use-superblocks
6275 @opindex fsched2-use-superblocks
6276 When scheduling after register allocation, do use superblock scheduling
6277 algorithm. Superblock scheduling allows motion across basic block boundaries
6278 resulting on faster schedules. This option is experimental, as not all machine
6279 descriptions used by GCC model the CPU closely enough to avoid unreliable
6280 results from the algorithm.
6282 This only makes sense when scheduling after register allocation, i.e.@: with
6283 @option{-fschedule-insns2} or at @option{-O2} or higher.
6285 @item -fsched-group-heuristic
6286 @opindex fsched-group-heuristic
6287 Enable the group heuristic in the scheduler. This heuristic favors
6288 the instruction that belongs to a schedule group. This is enabled
6289 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6290 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6292 @item -fsched-critical-path-heuristic
6293 @opindex fsched-critical-path-heuristic
6294 Enable the critical-path heuristic in the scheduler. This heuristic favors
6295 instructions on the critical path. This is enabled by default when
6296 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6297 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6299 @item -fsched-spec-insn-heuristic
6300 @opindex fsched-spec-insn-heuristic
6301 Enable the speculative instruction heuristic in the scheduler. This
6302 heuristic favors speculative instructions with greater dependency weakness.
6303 This is enabled by default when scheduling is enabled, i.e.@:
6304 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6305 or at @option{-O2} or higher.
6307 @item -fsched-rank-heuristic
6308 @opindex fsched-rank-heuristic
6309 Enable the rank heuristic in the scheduler. This heuristic favors
6310 the instruction belonging to a basic block with greater size or frequency.
6311 This is enabled by default when scheduling is enabled, i.e.@:
6312 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6313 at @option{-O2} or higher.
6315 @item -fsched-last-insn-heuristic
6316 @opindex fsched-last-insn-heuristic
6317 Enable the last-instruction heuristic in the scheduler. This heuristic
6318 favors the instruction that is less dependent on the last instruction
6319 scheduled. This is enabled by default when scheduling is enabled,
6320 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6321 at @option{-O2} or higher.
6323 @item -fsched-dep-count-heuristic
6324 @opindex fsched-dep-count-heuristic
6325 Enable the dependent-count heuristic in the scheduler. This heuristic
6326 favors the instruction that has more instructions depending on it.
6327 This is enabled by default when scheduling is enabled, i.e.@:
6328 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6329 at @option{-O2} or higher.
6331 @item -fsched2-use-traces
6332 @opindex fsched2-use-traces
6333 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6334 allocation and additionally perform code duplication in order to increase the
6335 size of superblocks using tracer pass. See @option{-ftracer} for details on
6338 This mode should produce faster but significantly longer programs. Also
6339 without @option{-fbranch-probabilities} the traces constructed may not
6340 match the reality and hurt the performance. This only makes
6341 sense when scheduling after register allocation, i.e.@: with
6342 @option{-fschedule-insns2} or at @option{-O2} or higher.
6344 @item -freschedule-modulo-scheduled-loops
6345 @opindex freschedule-modulo-scheduled-loops
6346 The modulo scheduling comes before the traditional scheduling, if a loop
6347 was modulo scheduled we may want to prevent the later scheduling passes
6348 from changing its schedule, we use this option to control that.
6350 @item -fselective-scheduling
6351 @opindex fselective-scheduling
6352 Schedule instructions using selective scheduling algorithm. Selective
6353 scheduling runs instead of the first scheduler pass.
6355 @item -fselective-scheduling2
6356 @opindex fselective-scheduling2
6357 Schedule instructions using selective scheduling algorithm. Selective
6358 scheduling runs instead of the second scheduler pass.
6360 @item -fsel-sched-pipelining
6361 @opindex fsel-sched-pipelining
6362 Enable software pipelining of innermost loops during selective scheduling.
6363 This option has no effect until one of @option{-fselective-scheduling} or
6364 @option{-fselective-scheduling2} is turned on.
6366 @item -fsel-sched-pipelining-outer-loops
6367 @opindex fsel-sched-pipelining-outer-loops
6368 When pipelining loops during selective scheduling, also pipeline outer loops.
6369 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6371 @item -fcaller-saves
6372 @opindex fcaller-saves
6373 Enable values to be allocated in registers that will be clobbered by
6374 function calls, by emitting extra instructions to save and restore the
6375 registers around such calls. Such allocation is done only when it
6376 seems to result in better code than would otherwise be produced.
6378 This option is always enabled by default on certain machines, usually
6379 those which have no call-preserved registers to use instead.
6381 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6383 @item -fconserve-stack
6384 @opindex fconserve-stack
6385 Attempt to minimize stack usage. The compiler will attempt to use less
6386 stack space, even if that makes the program slower. This option
6387 implies setting the @option{large-stack-frame} parameter to 100
6388 and the @option{large-stack-frame-growth} parameter to 400.
6390 @item -ftree-reassoc
6391 @opindex ftree-reassoc
6392 Perform reassociation on trees. This flag is enabled by default
6393 at @option{-O} and higher.
6397 Perform partial redundancy elimination (PRE) on trees. This flag is
6398 enabled by default at @option{-O2} and @option{-O3}.
6400 @item -ftree-forwprop
6401 @opindex ftree-forwprop
6402 Perform forward propagation on trees. This flag is enabled by default
6403 at @option{-O} and higher.
6407 Perform full redundancy elimination (FRE) on trees. The difference
6408 between FRE and PRE is that FRE only considers expressions
6409 that are computed on all paths leading to the redundant computation.
6410 This analysis is faster than PRE, though it exposes fewer redundancies.
6411 This flag is enabled by default at @option{-O} and higher.
6413 @item -ftree-phiprop
6414 @opindex ftree-phiprop
6415 Perform hoisting of loads from conditional pointers on trees. This
6416 pass is enabled by default at @option{-O} and higher.
6418 @item -ftree-copy-prop
6419 @opindex ftree-copy-prop
6420 Perform copy propagation on trees. This pass eliminates unnecessary
6421 copy operations. This flag is enabled by default at @option{-O} and
6424 @item -fipa-pure-const
6425 @opindex fipa-pure-const
6426 Discover which functions are pure or constant.
6427 Enabled by default at @option{-O} and higher.
6429 @item -fipa-reference
6430 @opindex fipa-reference
6431 Discover which static variables do not escape cannot escape the
6433 Enabled by default at @option{-O} and higher.
6435 @item -fipa-struct-reorg
6436 @opindex fipa-struct-reorg
6437 Perform structure reorganization optimization, that change C-like structures
6438 layout in order to better utilize spatial locality. This transformation is
6439 affective for programs containing arrays of structures. Available in two
6440 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6441 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6442 to provide the safety of this transformation. It works only in whole program
6443 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6444 enabled. Structures considered @samp{cold} by this transformation are not
6445 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6447 With this flag, the program debug info reflects a new structure layout.
6451 Perform interprocedural pointer analysis. This option is experimental
6452 and does not affect generated code.
6456 Perform interprocedural constant propagation.
6457 This optimization analyzes the program to determine when values passed
6458 to functions are constants and then optimizes accordingly.
6459 This optimization can substantially increase performance
6460 if the application has constants passed to functions.
6461 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6463 @item -fipa-cp-clone
6464 @opindex fipa-cp-clone
6465 Perform function cloning to make interprocedural constant propagation stronger.
6466 When enabled, interprocedural constant propagation will perform function cloning
6467 when externally visible function can be called with constant arguments.
6468 Because this optimization can create multiple copies of functions,
6469 it may significantly increase code size
6470 (see @option{--param ipcp-unit-growth=@var{value}}).
6471 This flag is enabled by default at @option{-O3}.
6473 @item -fipa-matrix-reorg
6474 @opindex fipa-matrix-reorg
6475 Perform matrix flattening and transposing.
6476 Matrix flattening tries to replace an @math{m}-dimensional matrix
6477 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6478 This reduces the level of indirection needed for accessing the elements
6479 of the matrix. The second optimization is matrix transposing that
6480 attempts to change the order of the matrix's dimensions in order to
6481 improve cache locality.
6482 Both optimizations need the @option{-fwhole-program} flag.
6483 Transposing is enabled only if profiling information is available.
6487 Perform forward store motion on trees. This flag is
6488 enabled by default at @option{-O} and higher.
6492 Perform sparse conditional constant propagation (CCP) on trees. This
6493 pass only operates on local scalar variables and is enabled by default
6494 at @option{-O} and higher.
6496 @item -ftree-switch-conversion
6497 Perform conversion of simple initializations in a switch to
6498 initializations from a scalar array. This flag is enabled by default
6499 at @option{-O2} and higher.
6503 Perform dead code elimination (DCE) on trees. This flag is enabled by
6504 default at @option{-O} and higher.
6506 @item -ftree-builtin-call-dce
6507 @opindex ftree-builtin-call-dce
6508 Perform conditional dead code elimination (DCE) for calls to builtin functions
6509 that may set @code{errno} but are otherwise side-effect free. This flag is
6510 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6513 @item -ftree-dominator-opts
6514 @opindex ftree-dominator-opts
6515 Perform a variety of simple scalar cleanups (constant/copy
6516 propagation, redundancy elimination, range propagation and expression
6517 simplification) based on a dominator tree traversal. This also
6518 performs jump threading (to reduce jumps to jumps). This flag is
6519 enabled by default at @option{-O} and higher.
6523 Perform dead store elimination (DSE) on trees. A dead store is a store into
6524 a memory location which will later be overwritten by another store without
6525 any intervening loads. In this case the earlier store can be deleted. This
6526 flag is enabled by default at @option{-O} and higher.
6530 Perform loop header copying on trees. This is beneficial since it increases
6531 effectiveness of code motion optimizations. It also saves one jump. This flag
6532 is enabled by default at @option{-O} and higher. It is not enabled
6533 for @option{-Os}, since it usually increases code size.
6535 @item -ftree-loop-optimize
6536 @opindex ftree-loop-optimize
6537 Perform loop optimizations on trees. This flag is enabled by default
6538 at @option{-O} and higher.
6540 @item -ftree-loop-linear
6541 @opindex ftree-loop-linear
6542 Perform linear loop transformations on tree. This flag can improve cache
6543 performance and allow further loop optimizations to take place.
6545 @item -floop-interchange
6546 Perform loop interchange transformations on loops. Interchanging two
6547 nested loops switches the inner and outer loops. For example, given a
6552 A(J, I) = A(J, I) * C
6556 loop interchange will transform the loop as if the user had written:
6560 A(J, I) = A(J, I) * C
6564 which can be beneficial when @code{N} is larger than the caches,
6565 because in Fortran, the elements of an array are stored in memory
6566 contiguously by column, and the original loop iterates over rows,
6567 potentially creating at each access a cache miss. This optimization
6568 applies to all the languages supported by GCC and is not limited to
6569 Fortran. To use this code transformation, GCC has to be configured
6570 with @option{--with-ppl} and @option{--with-cloog} to enable the
6571 Graphite loop transformation infrastructure.
6573 @item -floop-strip-mine
6574 Perform loop strip mining transformations on loops. Strip mining
6575 splits a loop into two nested loops. The outer loop has strides
6576 equal to the strip size and the inner loop has strides of the
6577 original loop within a strip. For example, given a loop like:
6583 loop strip mining will transform the loop as if the user had written:
6586 DO I = II, min (II + 3, N)
6591 This optimization applies to all the languages supported by GCC and is
6592 not limited to Fortran. To use this code transformation, GCC has to
6593 be configured with @option{--with-ppl} and @option{--with-cloog} to
6594 enable the Graphite loop transformation infrastructure.
6597 Perform loop blocking transformations on loops. Blocking strip mines
6598 each loop in the loop nest such that the memory accesses of the
6599 element loops fit inside caches. For example, given a loop like:
6603 A(J, I) = B(I) + C(J)
6607 loop blocking will transform the loop as if the user had written:
6611 DO I = II, min (II + 63, N)
6612 DO J = JJ, min (JJ + 63, M)
6613 A(J, I) = B(I) + C(J)
6619 which can be beneficial when @code{M} is larger than the caches,
6620 because the innermost loop will iterate over a smaller amount of data
6621 that can be kept in the caches. This optimization applies to all the
6622 languages supported by GCC and is not limited to Fortran. To use this
6623 code transformation, GCC has to be configured with @option{--with-ppl}
6624 and @option{--with-cloog} to enable the Graphite loop transformation
6627 @item -fgraphite-identity
6628 @opindex fgraphite-identity
6629 Enable the identity transformation for graphite. For every SCoP we generate
6630 the polyhedral representation and transform it back to gimple. Using
6631 @option{-fgraphite-identity} we can check the costs or benefits of the
6632 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6633 are also performed by the code generator CLooG, like index splitting and
6634 dead code elimination in loops.
6636 @item -floop-parallelize-all
6637 Use the Graphite data dependence analysis to identify loops that can
6638 be parallelized. Parallelize all the loops that can be analyzed to
6639 not contain loop carried dependences without checking that it is
6640 profitable to parallelize the loops.
6642 @item -fcheck-data-deps
6643 @opindex fcheck-data-deps
6644 Compare the results of several data dependence analyzers. This option
6645 is used for debugging the data dependence analyzers.
6647 @item -ftree-loop-distribution
6648 Perform loop distribution. This flag can improve cache performance on
6649 big loop bodies and allow further loop optimizations, like
6650 parallelization or vectorization, to take place. For example, the loop
6667 @item -ftree-loop-im
6668 @opindex ftree-loop-im
6669 Perform loop invariant motion on trees. This pass moves only invariants that
6670 would be hard to handle at RTL level (function calls, operations that expand to
6671 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6672 operands of conditions that are invariant out of the loop, so that we can use
6673 just trivial invariantness analysis in loop unswitching. The pass also includes
6676 @item -ftree-loop-ivcanon
6677 @opindex ftree-loop-ivcanon
6678 Create a canonical counter for number of iterations in the loop for that
6679 determining number of iterations requires complicated analysis. Later
6680 optimizations then may determine the number easily. Useful especially
6681 in connection with unrolling.
6685 Perform induction variable optimizations (strength reduction, induction
6686 variable merging and induction variable elimination) on trees.
6688 @item -ftree-parallelize-loops=n
6689 @opindex ftree-parallelize-loops
6690 Parallelize loops, i.e., split their iteration space to run in n threads.
6691 This is only possible for loops whose iterations are independent
6692 and can be arbitrarily reordered. The optimization is only
6693 profitable on multiprocessor machines, for loops that are CPU-intensive,
6694 rather than constrained e.g.@: by memory bandwidth. This option
6695 implies @option{-pthread}, and thus is only supported on targets
6696 that have support for @option{-pthread}.
6700 Perform function-local points-to analysis on trees. This flag is
6701 enabled by default at @option{-O} and higher.
6705 Perform scalar replacement of aggregates. This pass replaces structure
6706 references with scalars to prevent committing structures to memory too
6707 early. This flag is enabled by default at @option{-O} and higher.
6709 @item -ftree-copyrename
6710 @opindex ftree-copyrename
6711 Perform copy renaming on trees. This pass attempts to rename compiler
6712 temporaries to other variables at copy locations, usually resulting in
6713 variable names which more closely resemble the original variables. This flag
6714 is enabled by default at @option{-O} and higher.
6718 Perform temporary expression replacement during the SSA->normal phase. Single
6719 use/single def temporaries are replaced at their use location with their
6720 defining expression. This results in non-GIMPLE code, but gives the expanders
6721 much more complex trees to work on resulting in better RTL generation. This is
6722 enabled by default at @option{-O} and higher.
6724 @item -ftree-vectorize
6725 @opindex ftree-vectorize
6726 Perform loop vectorization on trees. This flag is enabled by default at
6729 @item -ftree-vect-loop-version
6730 @opindex ftree-vect-loop-version
6731 Perform loop versioning when doing loop vectorization on trees. When a loop
6732 appears to be vectorizable except that data alignment or data dependence cannot
6733 be determined at compile time then vectorized and non-vectorized versions of
6734 the loop are generated along with runtime checks for alignment or dependence
6735 to control which version is executed. This option is enabled by default
6736 except at level @option{-Os} where it is disabled.
6738 @item -fvect-cost-model
6739 @opindex fvect-cost-model
6740 Enable cost model for vectorization.
6744 Perform Value Range Propagation on trees. This is similar to the
6745 constant propagation pass, but instead of values, ranges of values are
6746 propagated. This allows the optimizers to remove unnecessary range
6747 checks like array bound checks and null pointer checks. This is
6748 enabled by default at @option{-O2} and higher. Null pointer check
6749 elimination is only done if @option{-fdelete-null-pointer-checks} is
6754 Perform tail duplication to enlarge superblock size. This transformation
6755 simplifies the control flow of the function allowing other optimizations to do
6758 @item -funroll-loops
6759 @opindex funroll-loops
6760 Unroll loops whose number of iterations can be determined at compile
6761 time or upon entry to the loop. @option{-funroll-loops} implies
6762 @option{-frerun-cse-after-loop}. This option makes code larger,
6763 and may or may not make it run faster.
6765 @item -funroll-all-loops
6766 @opindex funroll-all-loops
6767 Unroll all loops, even if their number of iterations is uncertain when
6768 the loop is entered. This usually makes programs run more slowly.
6769 @option{-funroll-all-loops} implies the same options as
6770 @option{-funroll-loops},
6772 @item -fsplit-ivs-in-unroller
6773 @opindex fsplit-ivs-in-unroller
6774 Enables expressing of values of induction variables in later iterations
6775 of the unrolled loop using the value in the first iteration. This breaks
6776 long dependency chains, thus improving efficiency of the scheduling passes.
6778 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6779 same effect. However in cases the loop body is more complicated than
6780 a single basic block, this is not reliable. It also does not work at all
6781 on some of the architectures due to restrictions in the CSE pass.
6783 This optimization is enabled by default.
6785 @item -fvariable-expansion-in-unroller
6786 @opindex fvariable-expansion-in-unroller
6787 With this option, the compiler will create multiple copies of some
6788 local variables when unrolling a loop which can result in superior code.
6790 @item -fpredictive-commoning
6791 @opindex fpredictive-commoning
6792 Perform predictive commoning optimization, i.e., reusing computations
6793 (especially memory loads and stores) performed in previous
6794 iterations of loops.
6796 This option is enabled at level @option{-O3}.
6798 @item -fprefetch-loop-arrays
6799 @opindex fprefetch-loop-arrays
6800 If supported by the target machine, generate instructions to prefetch
6801 memory to improve the performance of loops that access large arrays.
6803 This option may generate better or worse code; results are highly
6804 dependent on the structure of loops within the source code.
6806 Disabled at level @option{-Os}.
6809 @itemx -fno-peephole2
6810 @opindex fno-peephole
6811 @opindex fno-peephole2
6812 Disable any machine-specific peephole optimizations. The difference
6813 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6814 are implemented in the compiler; some targets use one, some use the
6815 other, a few use both.
6817 @option{-fpeephole} is enabled by default.
6818 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6820 @item -fno-guess-branch-probability
6821 @opindex fno-guess-branch-probability
6822 Do not guess branch probabilities using heuristics.
6824 GCC will use heuristics to guess branch probabilities if they are
6825 not provided by profiling feedback (@option{-fprofile-arcs}). These
6826 heuristics are based on the control flow graph. If some branch probabilities
6827 are specified by @samp{__builtin_expect}, then the heuristics will be
6828 used to guess branch probabilities for the rest of the control flow graph,
6829 taking the @samp{__builtin_expect} info into account. The interactions
6830 between the heuristics and @samp{__builtin_expect} can be complex, and in
6831 some cases, it may be useful to disable the heuristics so that the effects
6832 of @samp{__builtin_expect} are easier to understand.
6834 The default is @option{-fguess-branch-probability} at levels
6835 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6837 @item -freorder-blocks
6838 @opindex freorder-blocks
6839 Reorder basic blocks in the compiled function in order to reduce number of
6840 taken branches and improve code locality.
6842 Enabled at levels @option{-O2}, @option{-O3}.
6844 @item -freorder-blocks-and-partition
6845 @opindex freorder-blocks-and-partition
6846 In addition to reordering basic blocks in the compiled function, in order
6847 to reduce number of taken branches, partitions hot and cold basic blocks
6848 into separate sections of the assembly and .o files, to improve
6849 paging and cache locality performance.
6851 This optimization is automatically turned off in the presence of
6852 exception handling, for linkonce sections, for functions with a user-defined
6853 section attribute and on any architecture that does not support named
6856 @item -freorder-functions
6857 @opindex freorder-functions
6858 Reorder functions in the object file in order to
6859 improve code locality. This is implemented by using special
6860 subsections @code{.text.hot} for most frequently executed functions and
6861 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6862 the linker so object file format must support named sections and linker must
6863 place them in a reasonable way.
6865 Also profile feedback must be available in to make this option effective. See
6866 @option{-fprofile-arcs} for details.
6868 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6870 @item -fstrict-aliasing
6871 @opindex fstrict-aliasing
6872 Allows the compiler to assume the strictest aliasing rules applicable to
6873 the language being compiled. For C (and C++), this activates
6874 optimizations based on the type of expressions. In particular, an
6875 object of one type is assumed never to reside at the same address as an
6876 object of a different type, unless the types are almost the same. For
6877 example, an @code{unsigned int} can alias an @code{int}, but not a
6878 @code{void*} or a @code{double}. A character type may alias any other
6881 @anchor{Type-punning}Pay special attention to code like this:
6894 The practice of reading from a different union member than the one most
6895 recently written to (called ``type-punning'') is common. Even with
6896 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6897 is accessed through the union type. So, the code above will work as
6898 expected. @xref{Structures unions enumerations and bit-fields
6899 implementation}. However, this code might not:
6910 Similarly, access by taking the address, casting the resulting pointer
6911 and dereferencing the result has undefined behavior, even if the cast
6912 uses a union type, e.g.:
6916 return ((union a_union *) &d)->i;
6920 The @option{-fstrict-aliasing} option is enabled at levels
6921 @option{-O2}, @option{-O3}, @option{-Os}.
6923 @item -fstrict-overflow
6924 @opindex fstrict-overflow
6925 Allow the compiler to assume strict signed overflow rules, depending
6926 on the language being compiled. For C (and C++) this means that
6927 overflow when doing arithmetic with signed numbers is undefined, which
6928 means that the compiler may assume that it will not happen. This
6929 permits various optimizations. For example, the compiler will assume
6930 that an expression like @code{i + 10 > i} will always be true for
6931 signed @code{i}. This assumption is only valid if signed overflow is
6932 undefined, as the expression is false if @code{i + 10} overflows when
6933 using twos complement arithmetic. When this option is in effect any
6934 attempt to determine whether an operation on signed numbers will
6935 overflow must be written carefully to not actually involve overflow.
6937 This option also allows the compiler to assume strict pointer
6938 semantics: given a pointer to an object, if adding an offset to that
6939 pointer does not produce a pointer to the same object, the addition is
6940 undefined. This permits the compiler to conclude that @code{p + u >
6941 p} is always true for a pointer @code{p} and unsigned integer
6942 @code{u}. This assumption is only valid because pointer wraparound is
6943 undefined, as the expression is false if @code{p + u} overflows using
6944 twos complement arithmetic.
6946 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6947 that integer signed overflow is fully defined: it wraps. When
6948 @option{-fwrapv} is used, there is no difference between
6949 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6950 integers. With @option{-fwrapv} certain types of overflow are
6951 permitted. For example, if the compiler gets an overflow when doing
6952 arithmetic on constants, the overflowed value can still be used with
6953 @option{-fwrapv}, but not otherwise.
6955 The @option{-fstrict-overflow} option is enabled at levels
6956 @option{-O2}, @option{-O3}, @option{-Os}.
6958 @item -falign-functions
6959 @itemx -falign-functions=@var{n}
6960 @opindex falign-functions
6961 Align the start of functions to the next power-of-two greater than
6962 @var{n}, skipping up to @var{n} bytes. For instance,
6963 @option{-falign-functions=32} aligns functions to the next 32-byte
6964 boundary, but @option{-falign-functions=24} would align to the next
6965 32-byte boundary only if this can be done by skipping 23 bytes or less.
6967 @option{-fno-align-functions} and @option{-falign-functions=1} are
6968 equivalent and mean that functions will not be aligned.
6970 Some assemblers only support this flag when @var{n} is a power of two;
6971 in that case, it is rounded up.
6973 If @var{n} is not specified or is zero, use a machine-dependent default.
6975 Enabled at levels @option{-O2}, @option{-O3}.
6977 @item -falign-labels
6978 @itemx -falign-labels=@var{n}
6979 @opindex falign-labels
6980 Align all branch targets to a power-of-two boundary, skipping up to
6981 @var{n} bytes like @option{-falign-functions}. This option can easily
6982 make code slower, because it must insert dummy operations for when the
6983 branch target is reached in the usual flow of the code.
6985 @option{-fno-align-labels} and @option{-falign-labels=1} are
6986 equivalent and mean that labels will not be aligned.
6988 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6989 are greater than this value, then their values are used instead.
6991 If @var{n} is not specified or is zero, use a machine-dependent default
6992 which is very likely to be @samp{1}, meaning no alignment.
6994 Enabled at levels @option{-O2}, @option{-O3}.
6997 @itemx -falign-loops=@var{n}
6998 @opindex falign-loops
6999 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7000 like @option{-falign-functions}. The hope is that the loop will be
7001 executed many times, which will make up for any execution of the dummy
7004 @option{-fno-align-loops} and @option{-falign-loops=1} are
7005 equivalent and mean that loops will not be aligned.
7007 If @var{n} is not specified or is zero, use a machine-dependent default.
7009 Enabled at levels @option{-O2}, @option{-O3}.
7012 @itemx -falign-jumps=@var{n}
7013 @opindex falign-jumps
7014 Align branch targets to a power-of-two boundary, for branch targets
7015 where the targets can only be reached by jumping, skipping up to @var{n}
7016 bytes like @option{-falign-functions}. In this case, no dummy operations
7019 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7020 equivalent and mean that loops will not be aligned.
7022 If @var{n} is not specified or is zero, use a machine-dependent default.
7024 Enabled at levels @option{-O2}, @option{-O3}.
7026 @item -funit-at-a-time
7027 @opindex funit-at-a-time
7028 This option is left for compatibility reasons. @option{-funit-at-a-time}
7029 has no effect, while @option{-fno-unit-at-a-time} implies
7030 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7034 @item -fno-toplevel-reorder
7035 @opindex fno-toplevel-reorder
7036 Do not reorder top-level functions, variables, and @code{asm}
7037 statements. Output them in the same order that they appear in the
7038 input file. When this option is used, unreferenced static variables
7039 will not be removed. This option is intended to support existing code
7040 which relies on a particular ordering. For new code, it is better to
7043 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7044 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7049 Constructs webs as commonly used for register allocation purposes and assign
7050 each web individual pseudo register. This allows the register allocation pass
7051 to operate on pseudos directly, but also strengthens several other optimization
7052 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7053 however, make debugging impossible, since variables will no longer stay in a
7056 Enabled by default with @option{-funroll-loops}.
7058 @item -fwhole-program
7059 @opindex fwhole-program
7060 Assume that the current compilation unit represents the whole program being
7061 compiled. All public functions and variables with the exception of @code{main}
7062 and those merged by attribute @code{externally_visible} become static functions
7063 and in effect are optimized more aggressively by interprocedural optimizers.
7064 While this option is equivalent to proper use of the @code{static} keyword for
7065 programs consisting of a single file, in combination with option
7066 @option{--combine} this flag can be used to compile many smaller scale C
7067 programs since the functions and variables become local for the whole combined
7068 compilation unit, not for the single source file itself.
7070 This option implies @option{-fwhole-file} for Fortran programs.
7072 @item -fcprop-registers
7073 @opindex fcprop-registers
7074 After register allocation and post-register allocation instruction splitting,
7075 we perform a copy-propagation pass to try to reduce scheduling dependencies
7076 and occasionally eliminate the copy.
7078 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7080 @item -fprofile-correction
7081 @opindex fprofile-correction
7082 Profiles collected using an instrumented binary for multi-threaded programs may
7083 be inconsistent due to missed counter updates. When this option is specified,
7084 GCC will use heuristics to correct or smooth out such inconsistencies. By
7085 default, GCC will emit an error message when an inconsistent profile is detected.
7087 @item -fprofile-dir=@var{path}
7088 @opindex fprofile-dir
7090 Set the directory to search the profile data files in to @var{path}.
7091 This option affects only the profile data generated by
7092 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7093 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7094 and its related options.
7095 By default, GCC will use the current directory as @var{path}
7096 thus the profile data file will appear in the same directory as the object file.
7098 @item -fprofile-generate
7099 @itemx -fprofile-generate=@var{path}
7100 @opindex fprofile-generate
7102 Enable options usually used for instrumenting application to produce
7103 profile useful for later recompilation with profile feedback based
7104 optimization. You must use @option{-fprofile-generate} both when
7105 compiling and when linking your program.
7107 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7109 If @var{path} is specified, GCC will look at the @var{path} to find
7110 the profile feedback data files. See @option{-fprofile-dir}.
7113 @itemx -fprofile-use=@var{path}
7114 @opindex fprofile-use
7115 Enable profile feedback directed optimizations, and optimizations
7116 generally profitable only with profile feedback available.
7118 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7119 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7121 By default, GCC emits an error message if the feedback profiles do not
7122 match the source code. This error can be turned into a warning by using
7123 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7126 If @var{path} is specified, GCC will look at the @var{path} to find
7127 the profile feedback data files. See @option{-fprofile-dir}.
7130 The following options control compiler behavior regarding floating
7131 point arithmetic. These options trade off between speed and
7132 correctness. All must be specifically enabled.
7136 @opindex ffloat-store
7137 Do not store floating point variables in registers, and inhibit other
7138 options that might change whether a floating point value is taken from a
7141 @cindex floating point precision
7142 This option prevents undesirable excess precision on machines such as
7143 the 68000 where the floating registers (of the 68881) keep more
7144 precision than a @code{double} is supposed to have. Similarly for the
7145 x86 architecture. For most programs, the excess precision does only
7146 good, but a few programs rely on the precise definition of IEEE floating
7147 point. Use @option{-ffloat-store} for such programs, after modifying
7148 them to store all pertinent intermediate computations into variables.
7150 @item -fexcess-precision=@var{style}
7151 @opindex fexcess-precision
7152 This option allows further control over excess precision on machines
7153 where floating-point registers have more precision than the IEEE
7154 @code{float} and @code{double} types and the processor does not
7155 support operations rounding to those types. By default,
7156 @option{-fexcess-precision=fast} is in effect; this means that
7157 operations are carried out in the precision of the registers and that
7158 it is unpredictable when rounding to the types specified in the source
7159 code takes place. When compiling C, if
7160 @option{-fexcess-precision=standard} is specified then excess
7161 precision will follow the rules specified in ISO C99; in particular,
7162 both casts and assignments cause values to be rounded to their
7163 semantic types (whereas @option{-ffloat-store} only affects
7164 assignments). This option is enabled by default for C if a strict
7165 conformance option such as @option{-std=c99} is used.
7168 @option{-fexcess-precision=standard} is not implemented for languages
7169 other than C, and has no effect if
7170 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7171 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7172 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7173 semantics apply without excess precision, and in the latter, rounding
7178 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7179 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7180 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7182 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7184 This option is not turned on by any @option{-O} option since
7185 it can result in incorrect output for programs which depend on
7186 an exact implementation of IEEE or ISO rules/specifications for
7187 math functions. It may, however, yield faster code for programs
7188 that do not require the guarantees of these specifications.
7190 @item -fno-math-errno
7191 @opindex fno-math-errno
7192 Do not set ERRNO after calling math functions that are executed
7193 with a single instruction, e.g., sqrt. A program that relies on
7194 IEEE exceptions for math error handling may want to use this flag
7195 for speed while maintaining IEEE arithmetic compatibility.
7197 This option is not turned on by any @option{-O} option since
7198 it can result in incorrect output for programs which depend on
7199 an exact implementation of IEEE or ISO rules/specifications for
7200 math functions. It may, however, yield faster code for programs
7201 that do not require the guarantees of these specifications.
7203 The default is @option{-fmath-errno}.
7205 On Darwin systems, the math library never sets @code{errno}. There is
7206 therefore no reason for the compiler to consider the possibility that
7207 it might, and @option{-fno-math-errno} is the default.
7209 @item -funsafe-math-optimizations
7210 @opindex funsafe-math-optimizations
7212 Allow optimizations for floating-point arithmetic that (a) assume
7213 that arguments and results are valid and (b) may violate IEEE or
7214 ANSI standards. When used at link-time, it may include libraries
7215 or startup files that change the default FPU control word or other
7216 similar optimizations.
7218 This option is not turned on by any @option{-O} option since
7219 it can result in incorrect output for programs which depend on
7220 an exact implementation of IEEE or ISO rules/specifications for
7221 math functions. It may, however, yield faster code for programs
7222 that do not require the guarantees of these specifications.
7223 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7224 @option{-fassociative-math} and @option{-freciprocal-math}.
7226 The default is @option{-fno-unsafe-math-optimizations}.
7228 @item -fassociative-math
7229 @opindex fassociative-math
7231 Allow re-association of operands in series of floating-point operations.
7232 This violates the ISO C and C++ language standard by possibly changing
7233 computation result. NOTE: re-ordering may change the sign of zero as
7234 well as ignore NaNs and inhibit or create underflow or overflow (and
7235 thus cannot be used on a code which relies on rounding behavior like
7236 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7237 and thus may not be used when ordered comparisons are required.
7238 This option requires that both @option{-fno-signed-zeros} and
7239 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7240 much sense with @option{-frounding-math}.
7242 The default is @option{-fno-associative-math}.
7244 @item -freciprocal-math
7245 @opindex freciprocal-math
7247 Allow the reciprocal of a value to be used instead of dividing by
7248 the value if this enables optimizations. For example @code{x / y}
7249 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7250 is subject to common subexpression elimination. Note that this loses
7251 precision and increases the number of flops operating on the value.
7253 The default is @option{-fno-reciprocal-math}.
7255 @item -ffinite-math-only
7256 @opindex ffinite-math-only
7257 Allow optimizations for floating-point arithmetic that assume
7258 that arguments and results are not NaNs or +-Infs.
7260 This option is not turned on by any @option{-O} option since
7261 it can result in incorrect output for programs which depend on
7262 an exact implementation of IEEE or ISO rules/specifications for
7263 math functions. It may, however, yield faster code for programs
7264 that do not require the guarantees of these specifications.
7266 The default is @option{-fno-finite-math-only}.
7268 @item -fno-signed-zeros
7269 @opindex fno-signed-zeros
7270 Allow optimizations for floating point arithmetic that ignore the
7271 signedness of zero. IEEE arithmetic specifies the behavior of
7272 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7273 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7274 This option implies that the sign of a zero result isn't significant.
7276 The default is @option{-fsigned-zeros}.
7278 @item -fno-trapping-math
7279 @opindex fno-trapping-math
7280 Compile code assuming that floating-point operations cannot generate
7281 user-visible traps. These traps include division by zero, overflow,
7282 underflow, inexact result and invalid operation. This option requires
7283 that @option{-fno-signaling-nans} be in effect. Setting this option may
7284 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7286 This option should never be turned on by any @option{-O} option since
7287 it can result in incorrect output for programs which depend on
7288 an exact implementation of IEEE or ISO rules/specifications for
7291 The default is @option{-ftrapping-math}.
7293 @item -frounding-math
7294 @opindex frounding-math
7295 Disable transformations and optimizations that assume default floating
7296 point rounding behavior. This is round-to-zero for all floating point
7297 to integer conversions, and round-to-nearest for all other arithmetic
7298 truncations. This option should be specified for programs that change
7299 the FP rounding mode dynamically, or that may be executed with a
7300 non-default rounding mode. This option disables constant folding of
7301 floating point expressions at compile-time (which may be affected by
7302 rounding mode) and arithmetic transformations that are unsafe in the
7303 presence of sign-dependent rounding modes.
7305 The default is @option{-fno-rounding-math}.
7307 This option is experimental and does not currently guarantee to
7308 disable all GCC optimizations that are affected by rounding mode.
7309 Future versions of GCC may provide finer control of this setting
7310 using C99's @code{FENV_ACCESS} pragma. This command line option
7311 will be used to specify the default state for @code{FENV_ACCESS}.
7313 @item -fsignaling-nans
7314 @opindex fsignaling-nans
7315 Compile code assuming that IEEE signaling NaNs may generate user-visible
7316 traps during floating-point operations. Setting this option disables
7317 optimizations that may change the number of exceptions visible with
7318 signaling NaNs. This option implies @option{-ftrapping-math}.
7320 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7323 The default is @option{-fno-signaling-nans}.
7325 This option is experimental and does not currently guarantee to
7326 disable all GCC optimizations that affect signaling NaN behavior.
7328 @item -fsingle-precision-constant
7329 @opindex fsingle-precision-constant
7330 Treat floating point constant as single precision constant instead of
7331 implicitly converting it to double precision constant.
7333 @item -fcx-limited-range
7334 @opindex fcx-limited-range
7335 When enabled, this option states that a range reduction step is not
7336 needed when performing complex division. Also, there is no checking
7337 whether the result of a complex multiplication or division is @code{NaN
7338 + I*NaN}, with an attempt to rescue the situation in that case. The
7339 default is @option{-fno-cx-limited-range}, but is enabled by
7340 @option{-ffast-math}.
7342 This option controls the default setting of the ISO C99
7343 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7346 @item -fcx-fortran-rules
7347 @opindex fcx-fortran-rules
7348 Complex multiplication and division follow Fortran rules. Range
7349 reduction is done as part of complex division, but there is no checking
7350 whether the result of a complex multiplication or division is @code{NaN
7351 + I*NaN}, with an attempt to rescue the situation in that case.
7353 The default is @option{-fno-cx-fortran-rules}.
7357 The following options control optimizations that may improve
7358 performance, but are not enabled by any @option{-O} options. This
7359 section includes experimental options that may produce broken code.
7362 @item -fbranch-probabilities
7363 @opindex fbranch-probabilities
7364 After running a program compiled with @option{-fprofile-arcs}
7365 (@pxref{Debugging Options,, Options for Debugging Your Program or
7366 @command{gcc}}), you can compile it a second time using
7367 @option{-fbranch-probabilities}, to improve optimizations based on
7368 the number of times each branch was taken. When the program
7369 compiled with @option{-fprofile-arcs} exits it saves arc execution
7370 counts to a file called @file{@var{sourcename}.gcda} for each source
7371 file. The information in this data file is very dependent on the
7372 structure of the generated code, so you must use the same source code
7373 and the same optimization options for both compilations.
7375 With @option{-fbranch-probabilities}, GCC puts a
7376 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7377 These can be used to improve optimization. Currently, they are only
7378 used in one place: in @file{reorg.c}, instead of guessing which path a
7379 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7380 exactly determine which path is taken more often.
7382 @item -fprofile-values
7383 @opindex fprofile-values
7384 If combined with @option{-fprofile-arcs}, it adds code so that some
7385 data about values of expressions in the program is gathered.
7387 With @option{-fbranch-probabilities}, it reads back the data gathered
7388 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7389 notes to instructions for their later usage in optimizations.
7391 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7395 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7396 a code to gather information about values of expressions.
7398 With @option{-fbranch-probabilities}, it reads back the data gathered
7399 and actually performs the optimizations based on them.
7400 Currently the optimizations include specialization of division operation
7401 using the knowledge about the value of the denominator.
7403 @item -frename-registers
7404 @opindex frename-registers
7405 Attempt to avoid false dependencies in scheduled code by making use
7406 of registers left over after register allocation. This optimization
7407 will most benefit processors with lots of registers. Depending on the
7408 debug information format adopted by the target, however, it can
7409 make debugging impossible, since variables will no longer stay in
7410 a ``home register''.
7412 Enabled by default with @option{-funroll-loops}.
7416 Perform tail duplication to enlarge superblock size. This transformation
7417 simplifies the control flow of the function allowing other optimizations to do
7420 Enabled with @option{-fprofile-use}.
7422 @item -funroll-loops
7423 @opindex funroll-loops
7424 Unroll loops whose number of iterations can be determined at compile time or
7425 upon entry to the loop. @option{-funroll-loops} implies
7426 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7427 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7428 small constant number of iterations). This option makes code larger, and may
7429 or may not make it run faster.
7431 Enabled with @option{-fprofile-use}.
7433 @item -funroll-all-loops
7434 @opindex funroll-all-loops
7435 Unroll all loops, even if their number of iterations is uncertain when
7436 the loop is entered. This usually makes programs run more slowly.
7437 @option{-funroll-all-loops} implies the same options as
7438 @option{-funroll-loops}.
7441 @opindex fpeel-loops
7442 Peels the loops for that there is enough information that they do not
7443 roll much (from profile feedback). It also turns on complete loop peeling
7444 (i.e.@: complete removal of loops with small constant number of iterations).
7446 Enabled with @option{-fprofile-use}.
7448 @item -fmove-loop-invariants
7449 @opindex fmove-loop-invariants
7450 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7451 at level @option{-O1}
7453 @item -funswitch-loops
7454 @opindex funswitch-loops
7455 Move branches with loop invariant conditions out of the loop, with duplicates
7456 of the loop on both branches (modified according to result of the condition).
7458 @item -ffunction-sections
7459 @itemx -fdata-sections
7460 @opindex ffunction-sections
7461 @opindex fdata-sections
7462 Place each function or data item into its own section in the output
7463 file if the target supports arbitrary sections. The name of the
7464 function or the name of the data item determines the section's name
7467 Use these options on systems where the linker can perform optimizations
7468 to improve locality of reference in the instruction space. Most systems
7469 using the ELF object format and SPARC processors running Solaris 2 have
7470 linkers with such optimizations. AIX may have these optimizations in
7473 Only use these options when there are significant benefits from doing
7474 so. When you specify these options, the assembler and linker will
7475 create larger object and executable files and will also be slower.
7476 You will not be able to use @code{gprof} on all systems if you
7477 specify this option and you may have problems with debugging if
7478 you specify both this option and @option{-g}.
7480 @item -fbranch-target-load-optimize
7481 @opindex fbranch-target-load-optimize
7482 Perform branch target register load optimization before prologue / epilogue
7484 The use of target registers can typically be exposed only during reload,
7485 thus hoisting loads out of loops and doing inter-block scheduling needs
7486 a separate optimization pass.
7488 @item -fbranch-target-load-optimize2
7489 @opindex fbranch-target-load-optimize2
7490 Perform branch target register load optimization after prologue / epilogue
7493 @item -fbtr-bb-exclusive
7494 @opindex fbtr-bb-exclusive
7495 When performing branch target register load optimization, don't reuse
7496 branch target registers in within any basic block.
7498 @item -fstack-protector
7499 @opindex fstack-protector
7500 Emit extra code to check for buffer overflows, such as stack smashing
7501 attacks. This is done by adding a guard variable to functions with
7502 vulnerable objects. This includes functions that call alloca, and
7503 functions with buffers larger than 8 bytes. The guards are initialized
7504 when a function is entered and then checked when the function exits.
7505 If a guard check fails, an error message is printed and the program exits.
7507 @item -fstack-protector-all
7508 @opindex fstack-protector-all
7509 Like @option{-fstack-protector} except that all functions are protected.
7511 @item -fsection-anchors
7512 @opindex fsection-anchors
7513 Try to reduce the number of symbolic address calculations by using
7514 shared ``anchor'' symbols to address nearby objects. This transformation
7515 can help to reduce the number of GOT entries and GOT accesses on some
7518 For example, the implementation of the following function @code{foo}:
7522 int foo (void) @{ return a + b + c; @}
7525 would usually calculate the addresses of all three variables, but if you
7526 compile it with @option{-fsection-anchors}, it will access the variables
7527 from a common anchor point instead. The effect is similar to the
7528 following pseudocode (which isn't valid C):
7533 register int *xr = &x;
7534 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7538 Not all targets support this option.
7540 @item --param @var{name}=@var{value}
7542 In some places, GCC uses various constants to control the amount of
7543 optimization that is done. For example, GCC will not inline functions
7544 that contain more that a certain number of instructions. You can
7545 control some of these constants on the command-line using the
7546 @option{--param} option.
7548 The names of specific parameters, and the meaning of the values, are
7549 tied to the internals of the compiler, and are subject to change
7550 without notice in future releases.
7552 In each case, the @var{value} is an integer. The allowable choices for
7553 @var{name} are given in the following table:
7556 @item struct-reorg-cold-struct-ratio
7557 The threshold ratio (as a percentage) between a structure frequency
7558 and the frequency of the hottest structure in the program. This parameter
7559 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7560 We say that if the ratio of a structure frequency, calculated by profiling,
7561 to the hottest structure frequency in the program is less than this
7562 parameter, then structure reorganization is not applied to this structure.
7565 @item predictable-branch-cost-outcome
7566 When branch is predicted to be taken with probability lower than this threshold
7567 (in percent), then it is considered well predictable. The default is 10.
7569 @item max-crossjump-edges
7570 The maximum number of incoming edges to consider for crossjumping.
7571 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7572 the number of edges incoming to each block. Increasing values mean
7573 more aggressive optimization, making the compile time increase with
7574 probably small improvement in executable size.
7576 @item min-crossjump-insns
7577 The minimum number of instructions which must be matched at the end
7578 of two blocks before crossjumping will be performed on them. This
7579 value is ignored in the case where all instructions in the block being
7580 crossjumped from are matched. The default value is 5.
7582 @item max-grow-copy-bb-insns
7583 The maximum code size expansion factor when copying basic blocks
7584 instead of jumping. The expansion is relative to a jump instruction.
7585 The default value is 8.
7587 @item max-goto-duplication-insns
7588 The maximum number of instructions to duplicate to a block that jumps
7589 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7590 passes, GCC factors computed gotos early in the compilation process,
7591 and unfactors them as late as possible. Only computed jumps at the
7592 end of a basic blocks with no more than max-goto-duplication-insns are
7593 unfactored. The default value is 8.
7595 @item max-delay-slot-insn-search
7596 The maximum number of instructions to consider when looking for an
7597 instruction to fill a delay slot. If more than this arbitrary number of
7598 instructions is searched, the time savings from filling the delay slot
7599 will be minimal so stop searching. Increasing values mean more
7600 aggressive optimization, making the compile time increase with probably
7601 small improvement in executable run time.
7603 @item max-delay-slot-live-search
7604 When trying to fill delay slots, the maximum number of instructions to
7605 consider when searching for a block with valid live register
7606 information. Increasing this arbitrarily chosen value means more
7607 aggressive optimization, increasing the compile time. This parameter
7608 should be removed when the delay slot code is rewritten to maintain the
7611 @item max-gcse-memory
7612 The approximate maximum amount of memory that will be allocated in
7613 order to perform the global common subexpression elimination
7614 optimization. If more memory than specified is required, the
7615 optimization will not be done.
7617 @item max-pending-list-length
7618 The maximum number of pending dependencies scheduling will allow
7619 before flushing the current state and starting over. Large functions
7620 with few branches or calls can create excessively large lists which
7621 needlessly consume memory and resources.
7623 @item max-inline-insns-single
7624 Several parameters control the tree inliner used in gcc.
7625 This number sets the maximum number of instructions (counted in GCC's
7626 internal representation) in a single function that the tree inliner
7627 will consider for inlining. This only affects functions declared
7628 inline and methods implemented in a class declaration (C++).
7629 The default value is 300.
7631 @item max-inline-insns-auto
7632 When you use @option{-finline-functions} (included in @option{-O3}),
7633 a lot of functions that would otherwise not be considered for inlining
7634 by the compiler will be investigated. To those functions, a different
7635 (more restrictive) limit compared to functions declared inline can
7637 The default value is 60.
7639 @item large-function-insns
7640 The limit specifying really large functions. For functions larger than this
7641 limit after inlining, inlining is constrained by
7642 @option{--param large-function-growth}. This parameter is useful primarily
7643 to avoid extreme compilation time caused by non-linear algorithms used by the
7645 The default value is 2700.
7647 @item large-function-growth
7648 Specifies maximal growth of large function caused by inlining in percents.
7649 The default value is 100 which limits large function growth to 2.0 times
7652 @item large-unit-insns
7653 The limit specifying large translation unit. Growth caused by inlining of
7654 units larger than this limit is limited by @option{--param inline-unit-growth}.
7655 For small units this might be too tight (consider unit consisting of function A
7656 that is inline and B that just calls A three time. If B is small relative to
7657 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7658 large units consisting of small inlineable functions however the overall unit
7659 growth limit is needed to avoid exponential explosion of code size. Thus for
7660 smaller units, the size is increased to @option{--param large-unit-insns}
7661 before applying @option{--param inline-unit-growth}. The default is 10000
7663 @item inline-unit-growth
7664 Specifies maximal overall growth of the compilation unit caused by inlining.
7665 The default value is 30 which limits unit growth to 1.3 times the original
7668 @item ipcp-unit-growth
7669 Specifies maximal overall growth of the compilation unit caused by
7670 interprocedural constant propagation. The default value is 10 which limits
7671 unit growth to 1.1 times the original size.
7673 @item large-stack-frame
7674 The limit specifying large stack frames. While inlining the algorithm is trying
7675 to not grow past this limit too much. Default value is 256 bytes.
7677 @item large-stack-frame-growth
7678 Specifies maximal growth of large stack frames caused by inlining in percents.
7679 The default value is 1000 which limits large stack frame growth to 11 times
7682 @item max-inline-insns-recursive
7683 @itemx max-inline-insns-recursive-auto
7684 Specifies maximum number of instructions out-of-line copy of self recursive inline
7685 function can grow into by performing recursive inlining.
7687 For functions declared inline @option{--param max-inline-insns-recursive} is
7688 taken into account. For function not declared inline, recursive inlining
7689 happens only when @option{-finline-functions} (included in @option{-O3}) is
7690 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7691 default value is 450.
7693 @item max-inline-recursive-depth
7694 @itemx max-inline-recursive-depth-auto
7695 Specifies maximum recursion depth used by the recursive inlining.
7697 For functions declared inline @option{--param max-inline-recursive-depth} is
7698 taken into account. For function not declared inline, recursive inlining
7699 happens only when @option{-finline-functions} (included in @option{-O3}) is
7700 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7703 @item min-inline-recursive-probability
7704 Recursive inlining is profitable only for function having deep recursion
7705 in average and can hurt for function having little recursion depth by
7706 increasing the prologue size or complexity of function body to other
7709 When profile feedback is available (see @option{-fprofile-generate}) the actual
7710 recursion depth can be guessed from probability that function will recurse via
7711 given call expression. This parameter limits inlining only to call expression
7712 whose probability exceeds given threshold (in percents). The default value is
7715 @item early-inlining-insns
7716 Specify growth that early inliner can make. In effect it increases amount of
7717 inlining for code having large abstraction penalty. The default value is 12.
7719 @item max-early-inliner-iterations
7720 @itemx max-early-inliner-iterations
7721 Limit of iterations of early inliner. This basically bounds number of nested
7722 indirect calls early inliner can resolve. Deeper chains are still handled by
7725 @item min-vect-loop-bound
7726 The minimum number of iterations under which a loop will not get vectorized
7727 when @option{-ftree-vectorize} is used. The number of iterations after
7728 vectorization needs to be greater than the value specified by this option
7729 to allow vectorization. The default value is 0.
7731 @item max-unrolled-insns
7732 The maximum number of instructions that a loop should have if that loop
7733 is unrolled, and if the loop is unrolled, it determines how many times
7734 the loop code is unrolled.
7736 @item max-average-unrolled-insns
7737 The maximum number of instructions biased by probabilities of their execution
7738 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7739 it determines how many times the loop code is unrolled.
7741 @item max-unroll-times
7742 The maximum number of unrollings of a single loop.
7744 @item max-peeled-insns
7745 The maximum number of instructions that a loop should have if that loop
7746 is peeled, and if the loop is peeled, it determines how many times
7747 the loop code is peeled.
7749 @item max-peel-times
7750 The maximum number of peelings of a single loop.
7752 @item max-completely-peeled-insns
7753 The maximum number of insns of a completely peeled loop.
7755 @item max-completely-peel-times
7756 The maximum number of iterations of a loop to be suitable for complete peeling.
7758 @item max-unswitch-insns
7759 The maximum number of insns of an unswitched loop.
7761 @item max-unswitch-level
7762 The maximum number of branches unswitched in a single loop.
7765 The minimum cost of an expensive expression in the loop invariant motion.
7767 @item iv-consider-all-candidates-bound
7768 Bound on number of candidates for induction variables below that
7769 all candidates are considered for each use in induction variable
7770 optimizations. Only the most relevant candidates are considered
7771 if there are more candidates, to avoid quadratic time complexity.
7773 @item iv-max-considered-uses
7774 The induction variable optimizations give up on loops that contain more
7775 induction variable uses.
7777 @item iv-always-prune-cand-set-bound
7778 If number of candidates in the set is smaller than this value,
7779 we always try to remove unnecessary ivs from the set during its
7780 optimization when a new iv is added to the set.
7782 @item scev-max-expr-size
7783 Bound on size of expressions used in the scalar evolutions analyzer.
7784 Large expressions slow the analyzer.
7786 @item omega-max-vars
7787 The maximum number of variables in an Omega constraint system.
7788 The default value is 128.
7790 @item omega-max-geqs
7791 The maximum number of inequalities in an Omega constraint system.
7792 The default value is 256.
7795 The maximum number of equalities in an Omega constraint system.
7796 The default value is 128.
7798 @item omega-max-wild-cards
7799 The maximum number of wildcard variables that the Omega solver will
7800 be able to insert. The default value is 18.
7802 @item omega-hash-table-size
7803 The size of the hash table in the Omega solver. The default value is
7806 @item omega-max-keys
7807 The maximal number of keys used by the Omega solver. The default
7810 @item omega-eliminate-redundant-constraints
7811 When set to 1, use expensive methods to eliminate all redundant
7812 constraints. The default value is 0.
7814 @item vect-max-version-for-alignment-checks
7815 The maximum number of runtime checks that can be performed when
7816 doing loop versioning for alignment in the vectorizer. See option
7817 ftree-vect-loop-version for more information.
7819 @item vect-max-version-for-alias-checks
7820 The maximum number of runtime checks that can be performed when
7821 doing loop versioning for alias in the vectorizer. See option
7822 ftree-vect-loop-version for more information.
7824 @item max-iterations-to-track
7826 The maximum number of iterations of a loop the brute force algorithm
7827 for analysis of # of iterations of the loop tries to evaluate.
7829 @item hot-bb-count-fraction
7830 Select fraction of the maximal count of repetitions of basic block in program
7831 given basic block needs to have to be considered hot.
7833 @item hot-bb-frequency-fraction
7834 Select fraction of the maximal frequency of executions of basic block in
7835 function given basic block needs to have to be considered hot
7837 @item max-predicted-iterations
7838 The maximum number of loop iterations we predict statically. This is useful
7839 in cases where function contain single loop with known bound and other loop
7840 with unknown. We predict the known number of iterations correctly, while
7841 the unknown number of iterations average to roughly 10. This means that the
7842 loop without bounds would appear artificially cold relative to the other one.
7844 @item align-threshold
7846 Select fraction of the maximal frequency of executions of basic block in
7847 function given basic block will get aligned.
7849 @item align-loop-iterations
7851 A loop expected to iterate at lest the selected number of iterations will get
7854 @item tracer-dynamic-coverage
7855 @itemx tracer-dynamic-coverage-feedback
7857 This value is used to limit superblock formation once the given percentage of
7858 executed instructions is covered. This limits unnecessary code size
7861 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7862 feedback is available. The real profiles (as opposed to statically estimated
7863 ones) are much less balanced allowing the threshold to be larger value.
7865 @item tracer-max-code-growth
7866 Stop tail duplication once code growth has reached given percentage. This is
7867 rather hokey argument, as most of the duplicates will be eliminated later in
7868 cross jumping, so it may be set to much higher values than is the desired code
7871 @item tracer-min-branch-ratio
7873 Stop reverse growth when the reverse probability of best edge is less than this
7874 threshold (in percent).
7876 @item tracer-min-branch-ratio
7877 @itemx tracer-min-branch-ratio-feedback
7879 Stop forward growth if the best edge do have probability lower than this
7882 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7883 compilation for profile feedback and one for compilation without. The value
7884 for compilation with profile feedback needs to be more conservative (higher) in
7885 order to make tracer effective.
7887 @item max-cse-path-length
7889 Maximum number of basic blocks on path that cse considers. The default is 10.
7892 The maximum instructions CSE process before flushing. The default is 1000.
7894 @item ggc-min-expand
7896 GCC uses a garbage collector to manage its own memory allocation. This
7897 parameter specifies the minimum percentage by which the garbage
7898 collector's heap should be allowed to expand between collections.
7899 Tuning this may improve compilation speed; it has no effect on code
7902 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7903 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7904 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7905 GCC is not able to calculate RAM on a particular platform, the lower
7906 bound of 30% is used. Setting this parameter and
7907 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7908 every opportunity. This is extremely slow, but can be useful for
7911 @item ggc-min-heapsize
7913 Minimum size of the garbage collector's heap before it begins bothering
7914 to collect garbage. The first collection occurs after the heap expands
7915 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7916 tuning this may improve compilation speed, and has no effect on code
7919 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7920 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7921 with a lower bound of 4096 (four megabytes) and an upper bound of
7922 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7923 particular platform, the lower bound is used. Setting this parameter
7924 very large effectively disables garbage collection. Setting this
7925 parameter and @option{ggc-min-expand} to zero causes a full collection
7926 to occur at every opportunity.
7928 @item max-reload-search-insns
7929 The maximum number of instruction reload should look backward for equivalent
7930 register. Increasing values mean more aggressive optimization, making the
7931 compile time increase with probably slightly better performance. The default
7934 @item max-cselib-memory-locations
7935 The maximum number of memory locations cselib should take into account.
7936 Increasing values mean more aggressive optimization, making the compile time
7937 increase with probably slightly better performance. The default value is 500.
7939 @item reorder-blocks-duplicate
7940 @itemx reorder-blocks-duplicate-feedback
7942 Used by basic block reordering pass to decide whether to use unconditional
7943 branch or duplicate the code on its destination. Code is duplicated when its
7944 estimated size is smaller than this value multiplied by the estimated size of
7945 unconditional jump in the hot spots of the program.
7947 The @option{reorder-block-duplicate-feedback} is used only when profile
7948 feedback is available and may be set to higher values than
7949 @option{reorder-block-duplicate} since information about the hot spots is more
7952 @item max-sched-ready-insns
7953 The maximum number of instructions ready to be issued the scheduler should
7954 consider at any given time during the first scheduling pass. Increasing
7955 values mean more thorough searches, making the compilation time increase
7956 with probably little benefit. The default value is 100.
7958 @item max-sched-region-blocks
7959 The maximum number of blocks in a region to be considered for
7960 interblock scheduling. The default value is 10.
7962 @item max-pipeline-region-blocks
7963 The maximum number of blocks in a region to be considered for
7964 pipelining in the selective scheduler. The default value is 15.
7966 @item max-sched-region-insns
7967 The maximum number of insns in a region to be considered for
7968 interblock scheduling. The default value is 100.
7970 @item max-pipeline-region-insns
7971 The maximum number of insns in a region to be considered for
7972 pipelining in the selective scheduler. The default value is 200.
7975 The minimum probability (in percents) of reaching a source block
7976 for interblock speculative scheduling. The default value is 40.
7978 @item max-sched-extend-regions-iters
7979 The maximum number of iterations through CFG to extend regions.
7980 0 - disable region extension,
7981 N - do at most N iterations.
7982 The default value is 0.
7984 @item max-sched-insn-conflict-delay
7985 The maximum conflict delay for an insn to be considered for speculative motion.
7986 The default value is 3.
7988 @item sched-spec-prob-cutoff
7989 The minimal probability of speculation success (in percents), so that
7990 speculative insn will be scheduled.
7991 The default value is 40.
7993 @item sched-mem-true-dep-cost
7994 Minimal distance (in CPU cycles) between store and load targeting same
7995 memory locations. The default value is 1.
7997 @item selsched-max-lookahead
7998 The maximum size of the lookahead window of selective scheduling. It is a
7999 depth of search for available instructions.
8000 The default value is 50.
8002 @item selsched-max-sched-times
8003 The maximum number of times that an instruction will be scheduled during
8004 selective scheduling. This is the limit on the number of iterations
8005 through which the instruction may be pipelined. The default value is 2.
8007 @item selsched-max-insns-to-rename
8008 The maximum number of best instructions in the ready list that are considered
8009 for renaming in the selective scheduler. The default value is 2.
8011 @item max-last-value-rtl
8012 The maximum size measured as number of RTLs that can be recorded in an expression
8013 in combiner for a pseudo register as last known value of that register. The default
8016 @item integer-share-limit
8017 Small integer constants can use a shared data structure, reducing the
8018 compiler's memory usage and increasing its speed. This sets the maximum
8019 value of a shared integer constant. The default value is 256.
8021 @item min-virtual-mappings
8022 Specifies the minimum number of virtual mappings in the incremental
8023 SSA updater that should be registered to trigger the virtual mappings
8024 heuristic defined by virtual-mappings-ratio. The default value is
8027 @item virtual-mappings-ratio
8028 If the number of virtual mappings is virtual-mappings-ratio bigger
8029 than the number of virtual symbols to be updated, then the incremental
8030 SSA updater switches to a full update for those symbols. The default
8033 @item ssp-buffer-size
8034 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8035 protection when @option{-fstack-protection} is used.
8037 @item max-jump-thread-duplication-stmts
8038 Maximum number of statements allowed in a block that needs to be
8039 duplicated when threading jumps.
8041 @item max-fields-for-field-sensitive
8042 Maximum number of fields in a structure we will treat in
8043 a field sensitive manner during pointer analysis. The default is zero
8044 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8046 @item prefetch-latency
8047 Estimate on average number of instructions that are executed before
8048 prefetch finishes. The distance we prefetch ahead is proportional
8049 to this constant. Increasing this number may also lead to less
8050 streams being prefetched (see @option{simultaneous-prefetches}).
8052 @item simultaneous-prefetches
8053 Maximum number of prefetches that can run at the same time.
8055 @item l1-cache-line-size
8056 The size of cache line in L1 cache, in bytes.
8059 The size of L1 cache, in kilobytes.
8062 The size of L2 cache, in kilobytes.
8064 @item min-insn-to-prefetch-ratio
8065 The minimum ratio between the number of instructions and the
8066 number of prefetches to enable prefetching in a loop with an
8069 @item prefetch-min-insn-to-mem-ratio
8070 The minimum ratio between the number of instructions and the
8071 number of memory references to enable prefetching in a loop.
8073 @item use-canonical-types
8074 Whether the compiler should use the ``canonical'' type system. By
8075 default, this should always be 1, which uses a more efficient internal
8076 mechanism for comparing types in C++ and Objective-C++. However, if
8077 bugs in the canonical type system are causing compilation failures,
8078 set this value to 0 to disable canonical types.
8080 @item switch-conversion-max-branch-ratio
8081 Switch initialization conversion will refuse to create arrays that are
8082 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8083 branches in the switch.
8085 @item max-partial-antic-length
8086 Maximum length of the partial antic set computed during the tree
8087 partial redundancy elimination optimization (@option{-ftree-pre}) when
8088 optimizing at @option{-O3} and above. For some sorts of source code
8089 the enhanced partial redundancy elimination optimization can run away,
8090 consuming all of the memory available on the host machine. This
8091 parameter sets a limit on the length of the sets that are computed,
8092 which prevents the runaway behavior. Setting a value of 0 for
8093 this parameter will allow an unlimited set length.
8095 @item sccvn-max-scc-size
8096 Maximum size of a strongly connected component (SCC) during SCCVN
8097 processing. If this limit is hit, SCCVN processing for the whole
8098 function will not be done and optimizations depending on it will
8099 be disabled. The default maximum SCC size is 10000.
8101 @item ira-max-loops-num
8102 IRA uses a regional register allocation by default. If a function
8103 contains loops more than number given by the parameter, only at most
8104 given number of the most frequently executed loops will form regions
8105 for the regional register allocation. The default value of the
8108 @item ira-max-conflict-table-size
8109 Although IRA uses a sophisticated algorithm of compression conflict
8110 table, the table can be still big for huge functions. If the conflict
8111 table for a function could be more than size in MB given by the
8112 parameter, the conflict table is not built and faster, simpler, and
8113 lower quality register allocation algorithm will be used. The
8114 algorithm do not use pseudo-register conflicts. The default value of
8115 the parameter is 2000.
8117 @item loop-invariant-max-bbs-in-loop
8118 Loop invariant motion can be very expensive, both in compile time and
8119 in amount of needed compile time memory, with very large loops. Loops
8120 with more basic blocks than this parameter won't have loop invariant
8121 motion optimization performed on them. The default value of the
8122 parameter is 1000 for -O1 and 10000 for -O2 and above.
8124 @item min-nondebug-insn-uid
8125 Use uids starting at this parameter for nondebug insns. The range below
8126 the parameter is reserved exclusively for debug insns created by
8127 @option{-fvar-tracking-assignments}, but debug insns may get
8128 (non-overlapping) uids above it if the reserved range is exhausted.
8133 @node Preprocessor Options
8134 @section Options Controlling the Preprocessor
8135 @cindex preprocessor options
8136 @cindex options, preprocessor
8138 These options control the C preprocessor, which is run on each C source
8139 file before actual compilation.
8141 If you use the @option{-E} option, nothing is done except preprocessing.
8142 Some of these options make sense only together with @option{-E} because
8143 they cause the preprocessor output to be unsuitable for actual
8147 @item -Wp,@var{option}
8149 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8150 and pass @var{option} directly through to the preprocessor. If
8151 @var{option} contains commas, it is split into multiple options at the
8152 commas. However, many options are modified, translated or interpreted
8153 by the compiler driver before being passed to the preprocessor, and
8154 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8155 interface is undocumented and subject to change, so whenever possible
8156 you should avoid using @option{-Wp} and let the driver handle the
8159 @item -Xpreprocessor @var{option}
8160 @opindex Xpreprocessor
8161 Pass @var{option} as an option to the preprocessor. You can use this to
8162 supply system-specific preprocessor options which GCC does not know how to
8165 If you want to pass an option that takes an argument, you must use
8166 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8169 @include cppopts.texi
8171 @node Assembler Options
8172 @section Passing Options to the Assembler
8174 @c prevent bad page break with this line
8175 You can pass options to the assembler.
8178 @item -Wa,@var{option}
8180 Pass @var{option} as an option to the assembler. If @var{option}
8181 contains commas, it is split into multiple options at the commas.
8183 @item -Xassembler @var{option}
8185 Pass @var{option} as an option to the assembler. You can use this to
8186 supply system-specific assembler options which GCC does not know how to
8189 If you want to pass an option that takes an argument, you must use
8190 @option{-Xassembler} twice, once for the option and once for the argument.
8195 @section Options for Linking
8196 @cindex link options
8197 @cindex options, linking
8199 These options come into play when the compiler links object files into
8200 an executable output file. They are meaningless if the compiler is
8201 not doing a link step.
8205 @item @var{object-file-name}
8206 A file name that does not end in a special recognized suffix is
8207 considered to name an object file or library. (Object files are
8208 distinguished from libraries by the linker according to the file
8209 contents.) If linking is done, these object files are used as input
8218 If any of these options is used, then the linker is not run, and
8219 object file names should not be used as arguments. @xref{Overall
8223 @item -l@var{library}
8224 @itemx -l @var{library}
8226 Search the library named @var{library} when linking. (The second
8227 alternative with the library as a separate argument is only for
8228 POSIX compliance and is not recommended.)
8230 It makes a difference where in the command you write this option; the
8231 linker searches and processes libraries and object files in the order they
8232 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8233 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8234 to functions in @samp{z}, those functions may not be loaded.
8236 The linker searches a standard list of directories for the library,
8237 which is actually a file named @file{lib@var{library}.a}. The linker
8238 then uses this file as if it had been specified precisely by name.
8240 The directories searched include several standard system directories
8241 plus any that you specify with @option{-L}.
8243 Normally the files found this way are library files---archive files
8244 whose members are object files. The linker handles an archive file by
8245 scanning through it for members which define symbols that have so far
8246 been referenced but not defined. But if the file that is found is an
8247 ordinary object file, it is linked in the usual fashion. The only
8248 difference between using an @option{-l} option and specifying a file name
8249 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8250 and searches several directories.
8254 You need this special case of the @option{-l} option in order to
8255 link an Objective-C or Objective-C++ program.
8258 @opindex nostartfiles
8259 Do not use the standard system startup files when linking.
8260 The standard system libraries are used normally, unless @option{-nostdlib}
8261 or @option{-nodefaultlibs} is used.
8263 @item -nodefaultlibs
8264 @opindex nodefaultlibs
8265 Do not use the standard system libraries when linking.
8266 Only the libraries you specify will be passed to the linker, options
8267 specifying linkage of the system libraries, such as @code{-static-libgcc}
8268 or @code{-shared-libgcc}, will be ignored.
8269 The standard startup files are used normally, unless @option{-nostartfiles}
8270 is used. The compiler may generate calls to @code{memcmp},
8271 @code{memset}, @code{memcpy} and @code{memmove}.
8272 These entries are usually resolved by entries in
8273 libc. These entry points should be supplied through some other
8274 mechanism when this option is specified.
8278 Do not use the standard system startup files or libraries when linking.
8279 No startup files and only the libraries you specify will be passed to
8280 the linker, options specifying linkage of the system libraries, such as
8281 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8282 The compiler may generate calls to @code{memcmp}, @code{memset},
8283 @code{memcpy} and @code{memmove}.
8284 These entries are usually resolved by entries in
8285 libc. These entry points should be supplied through some other
8286 mechanism when this option is specified.
8288 @cindex @option{-lgcc}, use with @option{-nostdlib}
8289 @cindex @option{-nostdlib} and unresolved references
8290 @cindex unresolved references and @option{-nostdlib}
8291 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8292 @cindex @option{-nodefaultlibs} and unresolved references
8293 @cindex unresolved references and @option{-nodefaultlibs}
8294 One of the standard libraries bypassed by @option{-nostdlib} and
8295 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8296 that GCC uses to overcome shortcomings of particular machines, or special
8297 needs for some languages.
8298 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8299 Collection (GCC) Internals},
8300 for more discussion of @file{libgcc.a}.)
8301 In most cases, you need @file{libgcc.a} even when you want to avoid
8302 other standard libraries. In other words, when you specify @option{-nostdlib}
8303 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8304 This ensures that you have no unresolved references to internal GCC
8305 library subroutines. (For example, @samp{__main}, used to ensure C++
8306 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8307 GNU Compiler Collection (GCC) Internals}.)
8311 Produce a position independent executable on targets which support it.
8312 For predictable results, you must also specify the same set of options
8313 that were used to generate code (@option{-fpie}, @option{-fPIE},
8314 or model suboptions) when you specify this option.
8318 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8319 that support it. This instructs the linker to add all symbols, not
8320 only used ones, to the dynamic symbol table. This option is needed
8321 for some uses of @code{dlopen} or to allow obtaining backtraces
8322 from within a program.
8326 Remove all symbol table and relocation information from the executable.
8330 On systems that support dynamic linking, this prevents linking with the shared
8331 libraries. On other systems, this option has no effect.
8335 Produce a shared object which can then be linked with other objects to
8336 form an executable. Not all systems support this option. For predictable
8337 results, you must also specify the same set of options that were used to
8338 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8339 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8340 needs to build supplementary stub code for constructors to work. On
8341 multi-libbed systems, @samp{gcc -shared} must select the correct support
8342 libraries to link against. Failing to supply the correct flags may lead
8343 to subtle defects. Supplying them in cases where they are not necessary
8346 @item -shared-libgcc
8347 @itemx -static-libgcc
8348 @opindex shared-libgcc
8349 @opindex static-libgcc
8350 On systems that provide @file{libgcc} as a shared library, these options
8351 force the use of either the shared or static version respectively.
8352 If no shared version of @file{libgcc} was built when the compiler was
8353 configured, these options have no effect.
8355 There are several situations in which an application should use the
8356 shared @file{libgcc} instead of the static version. The most common
8357 of these is when the application wishes to throw and catch exceptions
8358 across different shared libraries. In that case, each of the libraries
8359 as well as the application itself should use the shared @file{libgcc}.
8361 Therefore, the G++ and GCJ drivers automatically add
8362 @option{-shared-libgcc} whenever you build a shared library or a main
8363 executable, because C++ and Java programs typically use exceptions, so
8364 this is the right thing to do.
8366 If, instead, you use the GCC driver to create shared libraries, you may
8367 find that they will not always be linked with the shared @file{libgcc}.
8368 If GCC finds, at its configuration time, that you have a non-GNU linker
8369 or a GNU linker that does not support option @option{--eh-frame-hdr},
8370 it will link the shared version of @file{libgcc} into shared libraries
8371 by default. Otherwise, it will take advantage of the linker and optimize
8372 away the linking with the shared version of @file{libgcc}, linking with
8373 the static version of libgcc by default. This allows exceptions to
8374 propagate through such shared libraries, without incurring relocation
8375 costs at library load time.
8377 However, if a library or main executable is supposed to throw or catch
8378 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8379 for the languages used in the program, or using the option
8380 @option{-shared-libgcc}, such that it is linked with the shared
8383 @item -static-libstdc++
8384 When the @command{g++} program is used to link a C++ program, it will
8385 normally automatically link against @option{libstdc++}. If
8386 @file{libstdc++} is available as a shared library, and the
8387 @option{-static} option is not used, then this will link against the
8388 shared version of @file{libstdc++}. That is normally fine. However, it
8389 is sometimes useful to freeze the version of @file{libstdc++} used by
8390 the program without going all the way to a fully static link. The
8391 @option{-static-libstdc++} option directs the @command{g++} driver to
8392 link @file{libstdc++} statically, without necessarily linking other
8393 libraries statically.
8397 Bind references to global symbols when building a shared object. Warn
8398 about any unresolved references (unless overridden by the link editor
8399 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8402 @item -T @var{script}
8404 @cindex linker script
8405 Use @var{script} as the linker script. This option is supported by most
8406 systems using the GNU linker. On some targets, such as bare-board
8407 targets without an operating system, the @option{-T} option may be required
8408 when linking to avoid references to undefined symbols.
8410 @item -Xlinker @var{option}
8412 Pass @var{option} as an option to the linker. You can use this to
8413 supply system-specific linker options which GCC does not know how to
8416 If you want to pass an option that takes a separate argument, you must use
8417 @option{-Xlinker} twice, once for the option and once for the argument.
8418 For example, to pass @option{-assert definitions}, you must write
8419 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8420 @option{-Xlinker "-assert definitions"}, because this passes the entire
8421 string as a single argument, which is not what the linker expects.
8423 When using the GNU linker, it is usually more convenient to pass
8424 arguments to linker options using the @option{@var{option}=@var{value}}
8425 syntax than as separate arguments. For example, you can specify
8426 @samp{-Xlinker -Map=output.map} rather than
8427 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8428 this syntax for command-line options.
8430 @item -Wl,@var{option}
8432 Pass @var{option} as an option to the linker. If @var{option} contains
8433 commas, it is split into multiple options at the commas. You can use this
8434 syntax to pass an argument to the option.
8435 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8436 linker. When using the GNU linker, you can also get the same effect with
8437 @samp{-Wl,-Map=output.map}.
8439 @item -u @var{symbol}
8441 Pretend the symbol @var{symbol} is undefined, to force linking of
8442 library modules to define it. You can use @option{-u} multiple times with
8443 different symbols to force loading of additional library modules.
8446 @node Directory Options
8447 @section Options for Directory Search
8448 @cindex directory options
8449 @cindex options, directory search
8452 These options specify directories to search for header files, for
8453 libraries and for parts of the compiler:
8458 Add the directory @var{dir} to the head of the list of directories to be
8459 searched for header files. This can be used to override a system header
8460 file, substituting your own version, since these directories are
8461 searched before the system header file directories. However, you should
8462 not use this option to add directories that contain vendor-supplied
8463 system header files (use @option{-isystem} for that). If you use more than
8464 one @option{-I} option, the directories are scanned in left-to-right
8465 order; the standard system directories come after.
8467 If a standard system include directory, or a directory specified with
8468 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8469 option will be ignored. The directory will still be searched but as a
8470 system directory at its normal position in the system include chain.
8471 This is to ensure that GCC's procedure to fix buggy system headers and
8472 the ordering for the include_next directive are not inadvertently changed.
8473 If you really need to change the search order for system directories,
8474 use the @option{-nostdinc} and/or @option{-isystem} options.
8476 @item -iquote@var{dir}
8478 Add the directory @var{dir} to the head of the list of directories to
8479 be searched for header files only for the case of @samp{#include
8480 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8481 otherwise just like @option{-I}.
8485 Add directory @var{dir} to the list of directories to be searched
8488 @item -B@var{prefix}
8490 This option specifies where to find the executables, libraries,
8491 include files, and data files of the compiler itself.
8493 The compiler driver program runs one or more of the subprograms
8494 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8495 @var{prefix} as a prefix for each program it tries to run, both with and
8496 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8498 For each subprogram to be run, the compiler driver first tries the
8499 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8500 was not specified, the driver tries two standard prefixes, which are
8501 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8502 those results in a file name that is found, the unmodified program
8503 name is searched for using the directories specified in your
8504 @env{PATH} environment variable.
8506 The compiler will check to see if the path provided by the @option{-B}
8507 refers to a directory, and if necessary it will add a directory
8508 separator character at the end of the path.
8510 @option{-B} prefixes that effectively specify directory names also apply
8511 to libraries in the linker, because the compiler translates these
8512 options into @option{-L} options for the linker. They also apply to
8513 includes files in the preprocessor, because the compiler translates these
8514 options into @option{-isystem} options for the preprocessor. In this case,
8515 the compiler appends @samp{include} to the prefix.
8517 The run-time support file @file{libgcc.a} can also be searched for using
8518 the @option{-B} prefix, if needed. If it is not found there, the two
8519 standard prefixes above are tried, and that is all. The file is left
8520 out of the link if it is not found by those means.
8522 Another way to specify a prefix much like the @option{-B} prefix is to use
8523 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8526 As a special kludge, if the path provided by @option{-B} is
8527 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8528 9, then it will be replaced by @file{[dir/]include}. This is to help
8529 with boot-strapping the compiler.
8531 @item -specs=@var{file}
8533 Process @var{file} after the compiler reads in the standard @file{specs}
8534 file, in order to override the defaults that the @file{gcc} driver
8535 program uses when determining what switches to pass to @file{cc1},
8536 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8537 @option{-specs=@var{file}} can be specified on the command line, and they
8538 are processed in order, from left to right.
8540 @item --sysroot=@var{dir}
8542 Use @var{dir} as the logical root directory for headers and libraries.
8543 For example, if the compiler would normally search for headers in
8544 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8545 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8547 If you use both this option and the @option{-isysroot} option, then
8548 the @option{--sysroot} option will apply to libraries, but the
8549 @option{-isysroot} option will apply to header files.
8551 The GNU linker (beginning with version 2.16) has the necessary support
8552 for this option. If your linker does not support this option, the
8553 header file aspect of @option{--sysroot} will still work, but the
8554 library aspect will not.
8558 This option has been deprecated. Please use @option{-iquote} instead for
8559 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8560 Any directories you specify with @option{-I} options before the @option{-I-}
8561 option are searched only for the case of @samp{#include "@var{file}"};
8562 they are not searched for @samp{#include <@var{file}>}.
8564 If additional directories are specified with @option{-I} options after
8565 the @option{-I-}, these directories are searched for all @samp{#include}
8566 directives. (Ordinarily @emph{all} @option{-I} directories are used
8569 In addition, the @option{-I-} option inhibits the use of the current
8570 directory (where the current input file came from) as the first search
8571 directory for @samp{#include "@var{file}"}. There is no way to
8572 override this effect of @option{-I-}. With @option{-I.} you can specify
8573 searching the directory which was current when the compiler was
8574 invoked. That is not exactly the same as what the preprocessor does
8575 by default, but it is often satisfactory.
8577 @option{-I-} does not inhibit the use of the standard system directories
8578 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8585 @section Specifying subprocesses and the switches to pass to them
8588 @command{gcc} is a driver program. It performs its job by invoking a
8589 sequence of other programs to do the work of compiling, assembling and
8590 linking. GCC interprets its command-line parameters and uses these to
8591 deduce which programs it should invoke, and which command-line options
8592 it ought to place on their command lines. This behavior is controlled
8593 by @dfn{spec strings}. In most cases there is one spec string for each
8594 program that GCC can invoke, but a few programs have multiple spec
8595 strings to control their behavior. The spec strings built into GCC can
8596 be overridden by using the @option{-specs=} command-line switch to specify
8599 @dfn{Spec files} are plaintext files that are used to construct spec
8600 strings. They consist of a sequence of directives separated by blank
8601 lines. The type of directive is determined by the first non-whitespace
8602 character on the line and it can be one of the following:
8605 @item %@var{command}
8606 Issues a @var{command} to the spec file processor. The commands that can
8610 @item %include <@var{file}>
8612 Search for @var{file} and insert its text at the current point in the
8615 @item %include_noerr <@var{file}>
8616 @cindex %include_noerr
8617 Just like @samp{%include}, but do not generate an error message if the include
8618 file cannot be found.
8620 @item %rename @var{old_name} @var{new_name}
8622 Rename the spec string @var{old_name} to @var{new_name}.
8626 @item *[@var{spec_name}]:
8627 This tells the compiler to create, override or delete the named spec
8628 string. All lines after this directive up to the next directive or
8629 blank line are considered to be the text for the spec string. If this
8630 results in an empty string then the spec will be deleted. (Or, if the
8631 spec did not exist, then nothing will happened.) Otherwise, if the spec
8632 does not currently exist a new spec will be created. If the spec does
8633 exist then its contents will be overridden by the text of this
8634 directive, unless the first character of that text is the @samp{+}
8635 character, in which case the text will be appended to the spec.
8637 @item [@var{suffix}]:
8638 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8639 and up to the next directive or blank line are considered to make up the
8640 spec string for the indicated suffix. When the compiler encounters an
8641 input file with the named suffix, it will processes the spec string in
8642 order to work out how to compile that file. For example:
8649 This says that any input file whose name ends in @samp{.ZZ} should be
8650 passed to the program @samp{z-compile}, which should be invoked with the
8651 command-line switch @option{-input} and with the result of performing the
8652 @samp{%i} substitution. (See below.)
8654 As an alternative to providing a spec string, the text that follows a
8655 suffix directive can be one of the following:
8658 @item @@@var{language}
8659 This says that the suffix is an alias for a known @var{language}. This is
8660 similar to using the @option{-x} command-line switch to GCC to specify a
8661 language explicitly. For example:
8668 Says that .ZZ files are, in fact, C++ source files.
8671 This causes an error messages saying:
8674 @var{name} compiler not installed on this system.
8678 GCC already has an extensive list of suffixes built into it.
8679 This directive will add an entry to the end of the list of suffixes, but
8680 since the list is searched from the end backwards, it is effectively
8681 possible to override earlier entries using this technique.
8685 GCC has the following spec strings built into it. Spec files can
8686 override these strings or create their own. Note that individual
8687 targets can also add their own spec strings to this list.
8690 asm Options to pass to the assembler
8691 asm_final Options to pass to the assembler post-processor
8692 cpp Options to pass to the C preprocessor
8693 cc1 Options to pass to the C compiler
8694 cc1plus Options to pass to the C++ compiler
8695 endfile Object files to include at the end of the link
8696 link Options to pass to the linker
8697 lib Libraries to include on the command line to the linker
8698 libgcc Decides which GCC support library to pass to the linker
8699 linker Sets the name of the linker
8700 predefines Defines to be passed to the C preprocessor
8701 signed_char Defines to pass to CPP to say whether @code{char} is signed
8703 startfile Object files to include at the start of the link
8706 Here is a small example of a spec file:
8712 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8715 This example renames the spec called @samp{lib} to @samp{old_lib} and
8716 then overrides the previous definition of @samp{lib} with a new one.
8717 The new definition adds in some extra command-line options before
8718 including the text of the old definition.
8720 @dfn{Spec strings} are a list of command-line options to be passed to their
8721 corresponding program. In addition, the spec strings can contain
8722 @samp{%}-prefixed sequences to substitute variable text or to
8723 conditionally insert text into the command line. Using these constructs
8724 it is possible to generate quite complex command lines.
8726 Here is a table of all defined @samp{%}-sequences for spec
8727 strings. Note that spaces are not generated automatically around the
8728 results of expanding these sequences. Therefore you can concatenate them
8729 together or combine them with constant text in a single argument.
8733 Substitute one @samp{%} into the program name or argument.
8736 Substitute the name of the input file being processed.
8739 Substitute the basename of the input file being processed.
8740 This is the substring up to (and not including) the last period
8741 and not including the directory.
8744 This is the same as @samp{%b}, but include the file suffix (text after
8748 Marks the argument containing or following the @samp{%d} as a
8749 temporary file name, so that that file will be deleted if GCC exits
8750 successfully. Unlike @samp{%g}, this contributes no text to the
8753 @item %g@var{suffix}
8754 Substitute a file name that has suffix @var{suffix} and is chosen
8755 once per compilation, and mark the argument in the same way as
8756 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8757 name is now chosen in a way that is hard to predict even when previously
8758 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8759 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8760 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8761 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8762 was simply substituted with a file name chosen once per compilation,
8763 without regard to any appended suffix (which was therefore treated
8764 just like ordinary text), making such attacks more likely to succeed.
8766 @item %u@var{suffix}
8767 Like @samp{%g}, but generates a new temporary file name even if
8768 @samp{%u@var{suffix}} was already seen.
8770 @item %U@var{suffix}
8771 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8772 new one if there is no such last file name. In the absence of any
8773 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8774 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8775 would involve the generation of two distinct file names, one
8776 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8777 simply substituted with a file name chosen for the previous @samp{%u},
8778 without regard to any appended suffix.
8780 @item %j@var{suffix}
8781 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8782 writable, and if save-temps is off; otherwise, substitute the name
8783 of a temporary file, just like @samp{%u}. This temporary file is not
8784 meant for communication between processes, but rather as a junk
8787 @item %|@var{suffix}
8788 @itemx %m@var{suffix}
8789 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8790 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8791 all. These are the two most common ways to instruct a program that it
8792 should read from standard input or write to standard output. If you
8793 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8794 construct: see for example @file{f/lang-specs.h}.
8796 @item %.@var{SUFFIX}
8797 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8798 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8799 terminated by the next space or %.
8802 Marks the argument containing or following the @samp{%w} as the
8803 designated output file of this compilation. This puts the argument
8804 into the sequence of arguments that @samp{%o} will substitute later.
8807 Substitutes the names of all the output files, with spaces
8808 automatically placed around them. You should write spaces
8809 around the @samp{%o} as well or the results are undefined.
8810 @samp{%o} is for use in the specs for running the linker.
8811 Input files whose names have no recognized suffix are not compiled
8812 at all, but they are included among the output files, so they will
8816 Substitutes the suffix for object files. Note that this is
8817 handled specially when it immediately follows @samp{%g, %u, or %U},
8818 because of the need for those to form complete file names. The
8819 handling is such that @samp{%O} is treated exactly as if it had already
8820 been substituted, except that @samp{%g, %u, and %U} do not currently
8821 support additional @var{suffix} characters following @samp{%O} as they would
8822 following, for example, @samp{.o}.
8825 Substitutes the standard macro predefinitions for the
8826 current target machine. Use this when running @code{cpp}.
8829 Like @samp{%p}, but puts @samp{__} before and after the name of each
8830 predefined macro, except for macros that start with @samp{__} or with
8831 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8835 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8836 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8837 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8838 and @option{-imultilib} as necessary.
8841 Current argument is the name of a library or startup file of some sort.
8842 Search for that file in a standard list of directories and substitute
8843 the full name found. The current working directory is included in the
8844 list of directories scanned.
8847 Current argument is the name of a linker script. Search for that file
8848 in the current list of directories to scan for libraries. If the file
8849 is located insert a @option{--script} option into the command line
8850 followed by the full path name found. If the file is not found then
8851 generate an error message. Note: the current working directory is not
8855 Print @var{str} as an error message. @var{str} is terminated by a newline.
8856 Use this when inconsistent options are detected.
8859 Substitute the contents of spec string @var{name} at this point.
8862 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8864 @item %x@{@var{option}@}
8865 Accumulate an option for @samp{%X}.
8868 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8872 Output the accumulated assembler options specified by @option{-Wa}.
8875 Output the accumulated preprocessor options specified by @option{-Wp}.
8878 Process the @code{asm} spec. This is used to compute the
8879 switches to be passed to the assembler.
8882 Process the @code{asm_final} spec. This is a spec string for
8883 passing switches to an assembler post-processor, if such a program is
8887 Process the @code{link} spec. This is the spec for computing the
8888 command line passed to the linker. Typically it will make use of the
8889 @samp{%L %G %S %D and %E} sequences.
8892 Dump out a @option{-L} option for each directory that GCC believes might
8893 contain startup files. If the target supports multilibs then the
8894 current multilib directory will be prepended to each of these paths.
8897 Process the @code{lib} spec. This is a spec string for deciding which
8898 libraries should be included on the command line to the linker.
8901 Process the @code{libgcc} spec. This is a spec string for deciding
8902 which GCC support library should be included on the command line to the linker.
8905 Process the @code{startfile} spec. This is a spec for deciding which
8906 object files should be the first ones passed to the linker. Typically
8907 this might be a file named @file{crt0.o}.
8910 Process the @code{endfile} spec. This is a spec string that specifies
8911 the last object files that will be passed to the linker.
8914 Process the @code{cpp} spec. This is used to construct the arguments
8915 to be passed to the C preprocessor.
8918 Process the @code{cc1} spec. This is used to construct the options to be
8919 passed to the actual C compiler (@samp{cc1}).
8922 Process the @code{cc1plus} spec. This is used to construct the options to be
8923 passed to the actual C++ compiler (@samp{cc1plus}).
8926 Substitute the variable part of a matched option. See below.
8927 Note that each comma in the substituted string is replaced by
8931 Remove all occurrences of @code{-S} from the command line. Note---this
8932 command is position dependent. @samp{%} commands in the spec string
8933 before this one will see @code{-S}, @samp{%} commands in the spec string
8934 after this one will not.
8936 @item %:@var{function}(@var{args})
8937 Call the named function @var{function}, passing it @var{args}.
8938 @var{args} is first processed as a nested spec string, then split
8939 into an argument vector in the usual fashion. The function returns
8940 a string which is processed as if it had appeared literally as part
8941 of the current spec.
8943 The following built-in spec functions are provided:
8947 The @code{getenv} spec function takes two arguments: an environment
8948 variable name and a string. If the environment variable is not
8949 defined, a fatal error is issued. Otherwise, the return value is the
8950 value of the environment variable concatenated with the string. For
8951 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8954 %:getenv(TOPDIR /include)
8957 expands to @file{/path/to/top/include}.
8959 @item @code{if-exists}
8960 The @code{if-exists} spec function takes one argument, an absolute
8961 pathname to a file. If the file exists, @code{if-exists} returns the
8962 pathname. Here is a small example of its usage:
8966 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8969 @item @code{if-exists-else}
8970 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8971 spec function, except that it takes two arguments. The first argument is
8972 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8973 returns the pathname. If it does not exist, it returns the second argument.
8974 This way, @code{if-exists-else} can be used to select one file or another,
8975 based on the existence of the first. Here is a small example of its usage:
8979 crt0%O%s %:if-exists(crti%O%s) \
8980 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8983 @item @code{replace-outfile}
8984 The @code{replace-outfile} spec function takes two arguments. It looks for the
8985 first argument in the outfiles array and replaces it with the second argument. Here
8986 is a small example of its usage:
8989 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8992 @item @code{print-asm-header}
8993 The @code{print-asm-header} function takes no arguments and simply
8994 prints a banner like:
9000 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9003 It is used to separate compiler options from assembler options
9004 in the @option{--target-help} output.
9008 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9009 If that switch was not specified, this substitutes nothing. Note that
9010 the leading dash is omitted when specifying this option, and it is
9011 automatically inserted if the substitution is performed. Thus the spec
9012 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9013 and would output the command line option @option{-foo}.
9015 @item %W@{@code{S}@}
9016 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9019 @item %@{@code{S}*@}
9020 Substitutes all the switches specified to GCC whose names start
9021 with @code{-S}, but which also take an argument. This is used for
9022 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9023 GCC considers @option{-o foo} as being
9024 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9025 text, including the space. Thus two arguments would be generated.
9027 @item %@{@code{S}*&@code{T}*@}
9028 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9029 (the order of @code{S} and @code{T} in the spec is not significant).
9030 There can be any number of ampersand-separated variables; for each the
9031 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9033 @item %@{@code{S}:@code{X}@}
9034 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9036 @item %@{!@code{S}:@code{X}@}
9037 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9039 @item %@{@code{S}*:@code{X}@}
9040 Substitutes @code{X} if one or more switches whose names start with
9041 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9042 once, no matter how many such switches appeared. However, if @code{%*}
9043 appears somewhere in @code{X}, then @code{X} will be substituted once
9044 for each matching switch, with the @code{%*} replaced by the part of
9045 that switch that matched the @code{*}.
9047 @item %@{.@code{S}:@code{X}@}
9048 Substitutes @code{X}, if processing a file with suffix @code{S}.
9050 @item %@{!.@code{S}:@code{X}@}
9051 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9053 @item %@{,@code{S}:@code{X}@}
9054 Substitutes @code{X}, if processing a file for language @code{S}.
9056 @item %@{!,@code{S}:@code{X}@}
9057 Substitutes @code{X}, if not processing a file for language @code{S}.
9059 @item %@{@code{S}|@code{P}:@code{X}@}
9060 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9061 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9062 @code{*} sequences as well, although they have a stronger binding than
9063 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9064 alternatives must be starred, and only the first matching alternative
9067 For example, a spec string like this:
9070 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9073 will output the following command-line options from the following input
9074 command-line options:
9079 -d fred.c -foo -baz -boggle
9080 -d jim.d -bar -baz -boggle
9083 @item %@{S:X; T:Y; :D@}
9085 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9086 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9087 be as many clauses as you need. This may be combined with @code{.},
9088 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9093 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9094 construct may contain other nested @samp{%} constructs or spaces, or
9095 even newlines. They are processed as usual, as described above.
9096 Trailing white space in @code{X} is ignored. White space may also
9097 appear anywhere on the left side of the colon in these constructs,
9098 except between @code{.} or @code{*} and the corresponding word.
9100 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9101 handled specifically in these constructs. If another value of
9102 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9103 @option{-W} switch is found later in the command line, the earlier
9104 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9105 just one letter, which passes all matching options.
9107 The character @samp{|} at the beginning of the predicate text is used to
9108 indicate that a command should be piped to the following command, but
9109 only if @option{-pipe} is specified.
9111 It is built into GCC which switches take arguments and which do not.
9112 (You might think it would be useful to generalize this to allow each
9113 compiler's spec to say which switches take arguments. But this cannot
9114 be done in a consistent fashion. GCC cannot even decide which input
9115 files have been specified without knowing which switches take arguments,
9116 and it must know which input files to compile in order to tell which
9119 GCC also knows implicitly that arguments starting in @option{-l} are to be
9120 treated as compiler output files, and passed to the linker in their
9121 proper position among the other output files.
9123 @c man begin OPTIONS
9125 @node Target Options
9126 @section Specifying Target Machine and Compiler Version
9127 @cindex target options
9128 @cindex cross compiling
9129 @cindex specifying machine version
9130 @cindex specifying compiler version and target machine
9131 @cindex compiler version, specifying
9132 @cindex target machine, specifying
9134 The usual way to run GCC is to run the executable called @file{gcc}, or
9135 @file{<machine>-gcc} when cross-compiling, or
9136 @file{<machine>-gcc-<version>} to run a version other than the one that
9137 was installed last. Sometimes this is inconvenient, so GCC provides
9138 options that will switch to another cross-compiler or version.
9141 @item -b @var{machine}
9143 The argument @var{machine} specifies the target machine for compilation.
9145 The value to use for @var{machine} is the same as was specified as the
9146 machine type when configuring GCC as a cross-compiler. For
9147 example, if a cross-compiler was configured with @samp{configure
9148 arm-elf}, meaning to compile for an arm processor with elf binaries,
9149 then you would specify @option{-b arm-elf} to run that cross compiler.
9150 Because there are other options beginning with @option{-b}, the
9151 configuration must contain a hyphen, or @option{-b} alone should be one
9152 argument followed by the configuration in the next argument.
9154 @item -V @var{version}
9156 The argument @var{version} specifies which version of GCC to run.
9157 This is useful when multiple versions are installed. For example,
9158 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9161 The @option{-V} and @option{-b} options work by running the
9162 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9163 use them if you can just run that directly.
9165 @node Submodel Options
9166 @section Hardware Models and Configurations
9167 @cindex submodel options
9168 @cindex specifying hardware config
9169 @cindex hardware models and configurations, specifying
9170 @cindex machine dependent options
9172 Earlier we discussed the standard option @option{-b} which chooses among
9173 different installed compilers for completely different target
9174 machines, such as VAX vs.@: 68000 vs.@: 80386.
9176 In addition, each of these target machine types can have its own
9177 special options, starting with @samp{-m}, to choose among various
9178 hardware models or configurations---for example, 68010 vs 68020,
9179 floating coprocessor or none. A single installed version of the
9180 compiler can compile for any model or configuration, according to the
9183 Some configurations of the compiler also support additional special
9184 options, usually for compatibility with other compilers on the same
9187 @c This list is ordered alphanumerically by subsection name.
9188 @c It should be the same order and spelling as these options are listed
9189 @c in Machine Dependent Options
9195 * Blackfin Options::
9199 * DEC Alpha Options::
9200 * DEC Alpha/VMS Options::
9203 * GNU/Linux Options::
9206 * i386 and x86-64 Options::
9207 * i386 and x86-64 Windows Options::
9209 * IA-64/VMS Options::
9220 * picoChip Options::
9222 * RS/6000 and PowerPC Options::
9223 * S/390 and zSeries Options::
9228 * System V Options::
9233 * Xstormy16 Options::
9239 @subsection ARC Options
9242 These options are defined for ARC implementations:
9247 Compile code for little endian mode. This is the default.
9251 Compile code for big endian mode.
9254 @opindex mmangle-cpu
9255 Prepend the name of the cpu to all public symbol names.
9256 In multiple-processor systems, there are many ARC variants with different
9257 instruction and register set characteristics. This flag prevents code
9258 compiled for one cpu to be linked with code compiled for another.
9259 No facility exists for handling variants that are ``almost identical''.
9260 This is an all or nothing option.
9262 @item -mcpu=@var{cpu}
9264 Compile code for ARC variant @var{cpu}.
9265 Which variants are supported depend on the configuration.
9266 All variants support @option{-mcpu=base}, this is the default.
9268 @item -mtext=@var{text-section}
9269 @itemx -mdata=@var{data-section}
9270 @itemx -mrodata=@var{readonly-data-section}
9274 Put functions, data, and readonly data in @var{text-section},
9275 @var{data-section}, and @var{readonly-data-section} respectively
9276 by default. This can be overridden with the @code{section} attribute.
9277 @xref{Variable Attributes}.
9279 @item -mfix-cortex-m3-ldrd
9280 @opindex mfix-cortex-m3-ldrd
9281 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9282 with overlapping destination and base registers are used. This option avoids
9283 generating these instructions. This option is enabled by default when
9284 @option{-mcpu=cortex-m3} is specified.
9289 @subsection ARM Options
9292 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9296 @item -mabi=@var{name}
9298 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9299 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9302 @opindex mapcs-frame
9303 Generate a stack frame that is compliant with the ARM Procedure Call
9304 Standard for all functions, even if this is not strictly necessary for
9305 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9306 with this option will cause the stack frames not to be generated for
9307 leaf functions. The default is @option{-mno-apcs-frame}.
9311 This is a synonym for @option{-mapcs-frame}.
9314 @c not currently implemented
9315 @item -mapcs-stack-check
9316 @opindex mapcs-stack-check
9317 Generate code to check the amount of stack space available upon entry to
9318 every function (that actually uses some stack space). If there is
9319 insufficient space available then either the function
9320 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9321 called, depending upon the amount of stack space required. The run time
9322 system is required to provide these functions. The default is
9323 @option{-mno-apcs-stack-check}, since this produces smaller code.
9325 @c not currently implemented
9327 @opindex mapcs-float
9328 Pass floating point arguments using the float point registers. This is
9329 one of the variants of the APCS@. This option is recommended if the
9330 target hardware has a floating point unit or if a lot of floating point
9331 arithmetic is going to be performed by the code. The default is
9332 @option{-mno-apcs-float}, since integer only code is slightly increased in
9333 size if @option{-mapcs-float} is used.
9335 @c not currently implemented
9336 @item -mapcs-reentrant
9337 @opindex mapcs-reentrant
9338 Generate reentrant, position independent code. The default is
9339 @option{-mno-apcs-reentrant}.
9342 @item -mthumb-interwork
9343 @opindex mthumb-interwork
9344 Generate code which supports calling between the ARM and Thumb
9345 instruction sets. Without this option the two instruction sets cannot
9346 be reliably used inside one program. The default is
9347 @option{-mno-thumb-interwork}, since slightly larger code is generated
9348 when @option{-mthumb-interwork} is specified.
9350 @item -mno-sched-prolog
9351 @opindex mno-sched-prolog
9352 Prevent the reordering of instructions in the function prolog, or the
9353 merging of those instruction with the instructions in the function's
9354 body. This means that all functions will start with a recognizable set
9355 of instructions (or in fact one of a choice from a small set of
9356 different function prologues), and this information can be used to
9357 locate the start if functions inside an executable piece of code. The
9358 default is @option{-msched-prolog}.
9360 @item -mfloat-abi=@var{name}
9362 Specifies which floating-point ABI to use. Permissible values
9363 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9365 Specifying @samp{soft} causes GCC to generate output containing
9366 library calls for floating-point operations.
9367 @samp{softfp} allows the generation of code using hardware floating-point
9368 instructions, but still uses the soft-float calling conventions.
9369 @samp{hard} allows generation of floating-point instructions
9370 and uses FPU-specific calling conventions.
9372 The default depends on the specific target configuration. Note that
9373 the hard-float and soft-float ABIs are not link-compatible; you must
9374 compile your entire program with the same ABI, and link with a
9375 compatible set of libraries.
9378 @opindex mhard-float
9379 Equivalent to @option{-mfloat-abi=hard}.
9382 @opindex msoft-float
9383 Equivalent to @option{-mfloat-abi=soft}.
9385 @item -mlittle-endian
9386 @opindex mlittle-endian
9387 Generate code for a processor running in little-endian mode. This is
9388 the default for all standard configurations.
9391 @opindex mbig-endian
9392 Generate code for a processor running in big-endian mode; the default is
9393 to compile code for a little-endian processor.
9395 @item -mwords-little-endian
9396 @opindex mwords-little-endian
9397 This option only applies when generating code for big-endian processors.
9398 Generate code for a little-endian word order but a big-endian byte
9399 order. That is, a byte order of the form @samp{32107654}. Note: this
9400 option should only be used if you require compatibility with code for
9401 big-endian ARM processors generated by versions of the compiler prior to
9404 @item -mcpu=@var{name}
9406 This specifies the name of the target ARM processor. GCC uses this name
9407 to determine what kind of instructions it can emit when generating
9408 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9409 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9410 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9411 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9412 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9414 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9415 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9416 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9417 @samp{strongarm1110},
9418 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9419 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9420 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9421 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9422 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9423 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9424 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9425 @samp{cortex-a8}, @samp{cortex-a9},
9426 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9429 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9431 @item -mtune=@var{name}
9433 This option is very similar to the @option{-mcpu=} option, except that
9434 instead of specifying the actual target processor type, and hence
9435 restricting which instructions can be used, it specifies that GCC should
9436 tune the performance of the code as if the target were of the type
9437 specified in this option, but still choosing the instructions that it
9438 will generate based on the cpu specified by a @option{-mcpu=} option.
9439 For some ARM implementations better performance can be obtained by using
9442 @item -march=@var{name}
9444 This specifies the name of the target ARM architecture. GCC uses this
9445 name to determine what kind of instructions it can emit when generating
9446 assembly code. This option can be used in conjunction with or instead
9447 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9448 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9449 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9450 @samp{armv6}, @samp{armv6j},
9451 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9452 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9453 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9455 @item -mfpu=@var{name}
9456 @itemx -mfpe=@var{number}
9457 @itemx -mfp=@var{number}
9461 This specifies what floating point hardware (or hardware emulation) is
9462 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9463 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9464 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9465 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9466 with older versions of GCC@.
9468 If @option{-msoft-float} is specified this specifies the format of
9469 floating point values.
9471 @item -mfp16-format=@var{name}
9472 @opindex mfp16-format
9473 Specify the format of the @code{__fp16} half-precision floating-point type.
9474 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9475 the default is @samp{none}, in which case the @code{__fp16} type is not
9476 defined. @xref{Half-Precision}, for more information.
9478 @item -mstructure-size-boundary=@var{n}
9479 @opindex mstructure-size-boundary
9480 The size of all structures and unions will be rounded up to a multiple
9481 of the number of bits set by this option. Permissible values are 8, 32
9482 and 64. The default value varies for different toolchains. For the COFF
9483 targeted toolchain the default value is 8. A value of 64 is only allowed
9484 if the underlying ABI supports it.
9486 Specifying the larger number can produce faster, more efficient code, but
9487 can also increase the size of the program. Different values are potentially
9488 incompatible. Code compiled with one value cannot necessarily expect to
9489 work with code or libraries compiled with another value, if they exchange
9490 information using structures or unions.
9492 @item -mabort-on-noreturn
9493 @opindex mabort-on-noreturn
9494 Generate a call to the function @code{abort} at the end of a
9495 @code{noreturn} function. It will be executed if the function tries to
9499 @itemx -mno-long-calls
9500 @opindex mlong-calls
9501 @opindex mno-long-calls
9502 Tells the compiler to perform function calls by first loading the
9503 address of the function into a register and then performing a subroutine
9504 call on this register. This switch is needed if the target function
9505 will lie outside of the 64 megabyte addressing range of the offset based
9506 version of subroutine call instruction.
9508 Even if this switch is enabled, not all function calls will be turned
9509 into long calls. The heuristic is that static functions, functions
9510 which have the @samp{short-call} attribute, functions that are inside
9511 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9512 definitions have already been compiled within the current compilation
9513 unit, will not be turned into long calls. The exception to this rule is
9514 that weak function definitions, functions with the @samp{long-call}
9515 attribute or the @samp{section} attribute, and functions that are within
9516 the scope of a @samp{#pragma long_calls} directive, will always be
9517 turned into long calls.
9519 This feature is not enabled by default. Specifying
9520 @option{-mno-long-calls} will restore the default behavior, as will
9521 placing the function calls within the scope of a @samp{#pragma
9522 long_calls_off} directive. Note these switches have no effect on how
9523 the compiler generates code to handle function calls via function
9526 @item -msingle-pic-base
9527 @opindex msingle-pic-base
9528 Treat the register used for PIC addressing as read-only, rather than
9529 loading it in the prologue for each function. The run-time system is
9530 responsible for initializing this register with an appropriate value
9531 before execution begins.
9533 @item -mpic-register=@var{reg}
9534 @opindex mpic-register
9535 Specify the register to be used for PIC addressing. The default is R10
9536 unless stack-checking is enabled, when R9 is used.
9538 @item -mcirrus-fix-invalid-insns
9539 @opindex mcirrus-fix-invalid-insns
9540 @opindex mno-cirrus-fix-invalid-insns
9541 Insert NOPs into the instruction stream to in order to work around
9542 problems with invalid Maverick instruction combinations. This option
9543 is only valid if the @option{-mcpu=ep9312} option has been used to
9544 enable generation of instructions for the Cirrus Maverick floating
9545 point co-processor. This option is not enabled by default, since the
9546 problem is only present in older Maverick implementations. The default
9547 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9550 @item -mpoke-function-name
9551 @opindex mpoke-function-name
9552 Write the name of each function into the text section, directly
9553 preceding the function prologue. The generated code is similar to this:
9557 .ascii "arm_poke_function_name", 0
9560 .word 0xff000000 + (t1 - t0)
9561 arm_poke_function_name
9563 stmfd sp!, @{fp, ip, lr, pc@}
9567 When performing a stack backtrace, code can inspect the value of
9568 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9569 location @code{pc - 12} and the top 8 bits are set, then we know that
9570 there is a function name embedded immediately preceding this location
9571 and has length @code{((pc[-3]) & 0xff000000)}.
9575 Generate code for the Thumb instruction set. The default is to
9576 use the 32-bit ARM instruction set.
9577 This option automatically enables either 16-bit Thumb-1 or
9578 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9579 and @option{-march=@var{name}} options. This option is not passed to the
9580 assembler. If you want to force assembler files to be interpreted as Thumb code,
9581 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9582 option directly to the assembler by prefixing it with @option{-Wa}.
9585 @opindex mtpcs-frame
9586 Generate a stack frame that is compliant with the Thumb Procedure Call
9587 Standard for all non-leaf functions. (A leaf function is one that does
9588 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9590 @item -mtpcs-leaf-frame
9591 @opindex mtpcs-leaf-frame
9592 Generate a stack frame that is compliant with the Thumb Procedure Call
9593 Standard for all leaf functions. (A leaf function is one that does
9594 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9596 @item -mcallee-super-interworking
9597 @opindex mcallee-super-interworking
9598 Gives all externally visible functions in the file being compiled an ARM
9599 instruction set header which switches to Thumb mode before executing the
9600 rest of the function. This allows these functions to be called from
9601 non-interworking code. This option is not valid in AAPCS configurations
9602 because interworking is enabled by default.
9604 @item -mcaller-super-interworking
9605 @opindex mcaller-super-interworking
9606 Allows calls via function pointers (including virtual functions) to
9607 execute correctly regardless of whether the target code has been
9608 compiled for interworking or not. There is a small overhead in the cost
9609 of executing a function pointer if this option is enabled. This option
9610 is not valid in AAPCS configurations because interworking is enabled
9613 @item -mtp=@var{name}
9615 Specify the access model for the thread local storage pointer. The valid
9616 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9617 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9618 (supported in the arm6k architecture), and @option{auto}, which uses the
9619 best available method for the selected processor. The default setting is
9622 @item -mword-relocations
9623 @opindex mword-relocations
9624 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9625 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9626 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9632 @subsection AVR Options
9635 These options are defined for AVR implementations:
9638 @item -mmcu=@var{mcu}
9640 Specify ATMEL AVR instruction set or MCU type.
9642 Instruction set avr1 is for the minimal AVR core, not supported by the C
9643 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9644 attiny11, attiny12, attiny15, attiny28).
9646 Instruction set avr2 (default) is for the classic AVR core with up to
9647 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9648 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9649 at90c8534, at90s8535).
9651 Instruction set avr3 is for the classic AVR core with up to 128K program
9652 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9654 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9655 memory space (MCU types: atmega8, atmega83, atmega85).
9657 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9658 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9659 atmega64, atmega128, at43usb355, at94k).
9663 Output instruction sizes to the asm file.
9665 @item -mno-interrupts
9666 @opindex mno-interrupts
9667 Generated code is not compatible with hardware interrupts.
9668 Code size will be smaller.
9670 @item -mcall-prologues
9671 @opindex mcall-prologues
9672 Functions prologues/epilogues expanded as call to appropriate
9673 subroutines. Code size will be smaller.
9676 @opindex mtiny-stack
9677 Change only the low 8 bits of the stack pointer.
9681 Assume int to be 8 bit integer. This affects the sizes of all types: A
9682 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9683 and long long will be 4 bytes. Please note that this option does not
9684 comply to the C standards, but it will provide you with smaller code
9688 @node Blackfin Options
9689 @subsection Blackfin Options
9690 @cindex Blackfin Options
9693 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9695 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9696 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9697 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9698 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9699 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9700 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9702 The optional @var{sirevision} specifies the silicon revision of the target
9703 Blackfin processor. Any workarounds available for the targeted silicon revision
9704 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9705 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9706 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9707 hexadecimal digits representing the major and minor numbers in the silicon
9708 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9709 is not defined. If @var{sirevision} is @samp{any}, the
9710 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9711 If this optional @var{sirevision} is not used, GCC assumes the latest known
9712 silicon revision of the targeted Blackfin processor.
9714 Support for @samp{bf561} is incomplete. For @samp{bf561},
9715 Only the processor macro is defined.
9716 Without this option, @samp{bf532} is used as the processor by default.
9717 The corresponding predefined processor macros for @var{cpu} is to
9718 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9719 provided by libgloss to be linked in if @option{-msim} is not given.
9723 Specifies that the program will be run on the simulator. This causes
9724 the simulator BSP provided by libgloss to be linked in. This option
9725 has effect only for @samp{bfin-elf} toolchain.
9726 Certain other options, such as @option{-mid-shared-library} and
9727 @option{-mfdpic}, imply @option{-msim}.
9729 @item -momit-leaf-frame-pointer
9730 @opindex momit-leaf-frame-pointer
9731 Don't keep the frame pointer in a register for leaf functions. This
9732 avoids the instructions to save, set up and restore frame pointers and
9733 makes an extra register available in leaf functions. The option
9734 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9735 which might make debugging harder.
9737 @item -mspecld-anomaly
9738 @opindex mspecld-anomaly
9739 When enabled, the compiler will ensure that the generated code does not
9740 contain speculative loads after jump instructions. If this option is used,
9741 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9743 @item -mno-specld-anomaly
9744 @opindex mno-specld-anomaly
9745 Don't generate extra code to prevent speculative loads from occurring.
9747 @item -mcsync-anomaly
9748 @opindex mcsync-anomaly
9749 When enabled, the compiler will ensure that the generated code does not
9750 contain CSYNC or SSYNC instructions too soon after conditional branches.
9751 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9753 @item -mno-csync-anomaly
9754 @opindex mno-csync-anomaly
9755 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9756 occurring too soon after a conditional branch.
9760 When enabled, the compiler is free to take advantage of the knowledge that
9761 the entire program fits into the low 64k of memory.
9764 @opindex mno-low-64k
9765 Assume that the program is arbitrarily large. This is the default.
9767 @item -mstack-check-l1
9768 @opindex mstack-check-l1
9769 Do stack checking using information placed into L1 scratchpad memory by the
9772 @item -mid-shared-library
9773 @opindex mid-shared-library
9774 Generate code that supports shared libraries via the library ID method.
9775 This allows for execute in place and shared libraries in an environment
9776 without virtual memory management. This option implies @option{-fPIC}.
9777 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9779 @item -mno-id-shared-library
9780 @opindex mno-id-shared-library
9781 Generate code that doesn't assume ID based shared libraries are being used.
9782 This is the default.
9784 @item -mleaf-id-shared-library
9785 @opindex mleaf-id-shared-library
9786 Generate code that supports shared libraries via the library ID method,
9787 but assumes that this library or executable won't link against any other
9788 ID shared libraries. That allows the compiler to use faster code for jumps
9791 @item -mno-leaf-id-shared-library
9792 @opindex mno-leaf-id-shared-library
9793 Do not assume that the code being compiled won't link against any ID shared
9794 libraries. Slower code will be generated for jump and call insns.
9796 @item -mshared-library-id=n
9797 @opindex mshared-library-id
9798 Specified the identification number of the ID based shared library being
9799 compiled. Specifying a value of 0 will generate more compact code, specifying
9800 other values will force the allocation of that number to the current
9801 library but is no more space or time efficient than omitting this option.
9805 Generate code that allows the data segment to be located in a different
9806 area of memory from the text segment. This allows for execute in place in
9807 an environment without virtual memory management by eliminating relocations
9808 against the text section.
9811 @opindex mno-sep-data
9812 Generate code that assumes that the data segment follows the text segment.
9813 This is the default.
9816 @itemx -mno-long-calls
9817 @opindex mlong-calls
9818 @opindex mno-long-calls
9819 Tells the compiler to perform function calls by first loading the
9820 address of the function into a register and then performing a subroutine
9821 call on this register. This switch is needed if the target function
9822 will lie outside of the 24 bit addressing range of the offset based
9823 version of subroutine call instruction.
9825 This feature is not enabled by default. Specifying
9826 @option{-mno-long-calls} will restore the default behavior. Note these
9827 switches have no effect on how the compiler generates code to handle
9828 function calls via function pointers.
9832 Link with the fast floating-point library. This library relaxes some of
9833 the IEEE floating-point standard's rules for checking inputs against
9834 Not-a-Number (NAN), in the interest of performance.
9837 @opindex minline-plt
9838 Enable inlining of PLT entries in function calls to functions that are
9839 not known to bind locally. It has no effect without @option{-mfdpic}.
9843 Build standalone application for multicore Blackfin processor. Proper
9844 start files and link scripts will be used to support multicore.
9845 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9846 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9847 @option{-mcorea} or @option{-mcoreb}. If it's used without
9848 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9849 programming model is used. In this model, the main function of Core B
9850 should be named as coreb_main. If it's used with @option{-mcorea} or
9851 @option{-mcoreb}, one application per core programming model is used.
9852 If this option is not used, single core application programming
9857 Build standalone application for Core A of BF561 when using
9858 one application per core programming model. Proper start files
9859 and link scripts will be used to support Core A. This option
9860 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9864 Build standalone application for Core B of BF561 when using
9865 one application per core programming model. Proper start files
9866 and link scripts will be used to support Core B. This option
9867 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9868 should be used instead of main. It must be used with
9869 @option{-mmulticore}.
9873 Build standalone application for SDRAM. Proper start files and
9874 link scripts will be used to put the application into SDRAM.
9875 Loader should initialize SDRAM before loading the application
9876 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9880 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9881 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9882 are enabled; for standalone applications the default is off.
9886 @subsection CRIS Options
9887 @cindex CRIS Options
9889 These options are defined specifically for the CRIS ports.
9892 @item -march=@var{architecture-type}
9893 @itemx -mcpu=@var{architecture-type}
9896 Generate code for the specified architecture. The choices for
9897 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9898 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9899 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9902 @item -mtune=@var{architecture-type}
9904 Tune to @var{architecture-type} everything applicable about the generated
9905 code, except for the ABI and the set of available instructions. The
9906 choices for @var{architecture-type} are the same as for
9907 @option{-march=@var{architecture-type}}.
9909 @item -mmax-stack-frame=@var{n}
9910 @opindex mmax-stack-frame
9911 Warn when the stack frame of a function exceeds @var{n} bytes.
9917 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9918 @option{-march=v3} and @option{-march=v8} respectively.
9920 @item -mmul-bug-workaround
9921 @itemx -mno-mul-bug-workaround
9922 @opindex mmul-bug-workaround
9923 @opindex mno-mul-bug-workaround
9924 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9925 models where it applies. This option is active by default.
9929 Enable CRIS-specific verbose debug-related information in the assembly
9930 code. This option also has the effect to turn off the @samp{#NO_APP}
9931 formatted-code indicator to the assembler at the beginning of the
9936 Do not use condition-code results from previous instruction; always emit
9937 compare and test instructions before use of condition codes.
9939 @item -mno-side-effects
9940 @opindex mno-side-effects
9941 Do not emit instructions with side-effects in addressing modes other than
9945 @itemx -mno-stack-align
9947 @itemx -mno-data-align
9948 @itemx -mconst-align
9949 @itemx -mno-const-align
9950 @opindex mstack-align
9951 @opindex mno-stack-align
9952 @opindex mdata-align
9953 @opindex mno-data-align
9954 @opindex mconst-align
9955 @opindex mno-const-align
9956 These options (no-options) arranges (eliminate arrangements) for the
9957 stack-frame, individual data and constants to be aligned for the maximum
9958 single data access size for the chosen CPU model. The default is to
9959 arrange for 32-bit alignment. ABI details such as structure layout are
9960 not affected by these options.
9968 Similar to the stack- data- and const-align options above, these options
9969 arrange for stack-frame, writable data and constants to all be 32-bit,
9970 16-bit or 8-bit aligned. The default is 32-bit alignment.
9972 @item -mno-prologue-epilogue
9973 @itemx -mprologue-epilogue
9974 @opindex mno-prologue-epilogue
9975 @opindex mprologue-epilogue
9976 With @option{-mno-prologue-epilogue}, the normal function prologue and
9977 epilogue that sets up the stack-frame are omitted and no return
9978 instructions or return sequences are generated in the code. Use this
9979 option only together with visual inspection of the compiled code: no
9980 warnings or errors are generated when call-saved registers must be saved,
9981 or storage for local variable needs to be allocated.
9987 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9988 instruction sequences that load addresses for functions from the PLT part
9989 of the GOT rather than (traditional on other architectures) calls to the
9990 PLT@. The default is @option{-mgotplt}.
9994 Legacy no-op option only recognized with the cris-axis-elf and
9995 cris-axis-linux-gnu targets.
9999 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10003 This option, recognized for the cris-axis-elf arranges
10004 to link with input-output functions from a simulator library. Code,
10005 initialized data and zero-initialized data are allocated consecutively.
10009 Like @option{-sim}, but pass linker options to locate initialized data at
10010 0x40000000 and zero-initialized data at 0x80000000.
10014 @subsection CRX Options
10015 @cindex CRX Options
10017 These options are defined specifically for the CRX ports.
10023 Enable the use of multiply-accumulate instructions. Disabled by default.
10026 @opindex mpush-args
10027 Push instructions will be used to pass outgoing arguments when functions
10028 are called. Enabled by default.
10031 @node Darwin Options
10032 @subsection Darwin Options
10033 @cindex Darwin options
10035 These options are defined for all architectures running the Darwin operating
10038 FSF GCC on Darwin does not create ``fat'' object files; it will create
10039 an object file for the single architecture that it was built to
10040 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10041 @option{-arch} options are used; it does so by running the compiler or
10042 linker multiple times and joining the results together with
10045 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10046 @samp{i686}) is determined by the flags that specify the ISA
10047 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10048 @option{-force_cpusubtype_ALL} option can be used to override this.
10050 The Darwin tools vary in their behavior when presented with an ISA
10051 mismatch. The assembler, @file{as}, will only permit instructions to
10052 be used that are valid for the subtype of the file it is generating,
10053 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10054 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10055 and print an error if asked to create a shared library with a less
10056 restrictive subtype than its input files (for instance, trying to put
10057 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10058 for executables, @file{ld}, will quietly give the executable the most
10059 restrictive subtype of any of its input files.
10064 Add the framework directory @var{dir} to the head of the list of
10065 directories to be searched for header files. These directories are
10066 interleaved with those specified by @option{-I} options and are
10067 scanned in a left-to-right order.
10069 A framework directory is a directory with frameworks in it. A
10070 framework is a directory with a @samp{"Headers"} and/or
10071 @samp{"PrivateHeaders"} directory contained directly in it that ends
10072 in @samp{".framework"}. The name of a framework is the name of this
10073 directory excluding the @samp{".framework"}. Headers associated with
10074 the framework are found in one of those two directories, with
10075 @samp{"Headers"} being searched first. A subframework is a framework
10076 directory that is in a framework's @samp{"Frameworks"} directory.
10077 Includes of subframework headers can only appear in a header of a
10078 framework that contains the subframework, or in a sibling subframework
10079 header. Two subframeworks are siblings if they occur in the same
10080 framework. A subframework should not have the same name as a
10081 framework, a warning will be issued if this is violated. Currently a
10082 subframework cannot have subframeworks, in the future, the mechanism
10083 may be extended to support this. The standard frameworks can be found
10084 in @samp{"/System/Library/Frameworks"} and
10085 @samp{"/Library/Frameworks"}. An example include looks like
10086 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10087 the name of the framework and header.h is found in the
10088 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10090 @item -iframework@var{dir}
10091 @opindex iframework
10092 Like @option{-F} except the directory is a treated as a system
10093 directory. The main difference between this @option{-iframework} and
10094 @option{-F} is that with @option{-iframework} the compiler does not
10095 warn about constructs contained within header files found via
10096 @var{dir}. This option is valid only for the C family of languages.
10100 Emit debugging information for symbols that are used. For STABS
10101 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10102 This is by default ON@.
10106 Emit debugging information for all symbols and types.
10108 @item -mmacosx-version-min=@var{version}
10109 The earliest version of MacOS X that this executable will run on
10110 is @var{version}. Typical values of @var{version} include @code{10.1},
10111 @code{10.2}, and @code{10.3.9}.
10113 If the compiler was built to use the system's headers by default,
10114 then the default for this option is the system version on which the
10115 compiler is running, otherwise the default is to make choices which
10116 are compatible with as many systems and code bases as possible.
10120 Enable kernel development mode. The @option{-mkernel} option sets
10121 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10122 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10123 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10124 applicable. This mode also sets @option{-mno-altivec},
10125 @option{-msoft-float}, @option{-fno-builtin} and
10126 @option{-mlong-branch} for PowerPC targets.
10128 @item -mone-byte-bool
10129 @opindex mone-byte-bool
10130 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10131 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10132 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10133 option has no effect on x86.
10135 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10136 to generate code that is not binary compatible with code generated
10137 without that switch. Using this switch may require recompiling all
10138 other modules in a program, including system libraries. Use this
10139 switch to conform to a non-default data model.
10141 @item -mfix-and-continue
10142 @itemx -ffix-and-continue
10143 @itemx -findirect-data
10144 @opindex mfix-and-continue
10145 @opindex ffix-and-continue
10146 @opindex findirect-data
10147 Generate code suitable for fast turn around development. Needed to
10148 enable gdb to dynamically load @code{.o} files into already running
10149 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10150 are provided for backwards compatibility.
10154 Loads all members of static archive libraries.
10155 See man ld(1) for more information.
10157 @item -arch_errors_fatal
10158 @opindex arch_errors_fatal
10159 Cause the errors having to do with files that have the wrong architecture
10162 @item -bind_at_load
10163 @opindex bind_at_load
10164 Causes the output file to be marked such that the dynamic linker will
10165 bind all undefined references when the file is loaded or launched.
10169 Produce a Mach-o bundle format file.
10170 See man ld(1) for more information.
10172 @item -bundle_loader @var{executable}
10173 @opindex bundle_loader
10174 This option specifies the @var{executable} that will be loading the build
10175 output file being linked. See man ld(1) for more information.
10178 @opindex dynamiclib
10179 When passed this option, GCC will produce a dynamic library instead of
10180 an executable when linking, using the Darwin @file{libtool} command.
10182 @item -force_cpusubtype_ALL
10183 @opindex force_cpusubtype_ALL
10184 This causes GCC's output file to have the @var{ALL} subtype, instead of
10185 one controlled by the @option{-mcpu} or @option{-march} option.
10187 @item -allowable_client @var{client_name}
10188 @itemx -client_name
10189 @itemx -compatibility_version
10190 @itemx -current_version
10192 @itemx -dependency-file
10194 @itemx -dylinker_install_name
10196 @itemx -exported_symbols_list
10198 @itemx -flat_namespace
10199 @itemx -force_flat_namespace
10200 @itemx -headerpad_max_install_names
10203 @itemx -install_name
10204 @itemx -keep_private_externs
10205 @itemx -multi_module
10206 @itemx -multiply_defined
10207 @itemx -multiply_defined_unused
10209 @itemx -no_dead_strip_inits_and_terms
10210 @itemx -nofixprebinding
10211 @itemx -nomultidefs
10213 @itemx -noseglinkedit
10214 @itemx -pagezero_size
10216 @itemx -prebind_all_twolevel_modules
10217 @itemx -private_bundle
10218 @itemx -read_only_relocs
10220 @itemx -sectobjectsymbols
10224 @itemx -sectobjectsymbols
10227 @itemx -segs_read_only_addr
10228 @itemx -segs_read_write_addr
10229 @itemx -seg_addr_table
10230 @itemx -seg_addr_table_filename
10231 @itemx -seglinkedit
10233 @itemx -segs_read_only_addr
10234 @itemx -segs_read_write_addr
10235 @itemx -single_module
10237 @itemx -sub_library
10238 @itemx -sub_umbrella
10239 @itemx -twolevel_namespace
10242 @itemx -unexported_symbols_list
10243 @itemx -weak_reference_mismatches
10244 @itemx -whatsloaded
10245 @opindex allowable_client
10246 @opindex client_name
10247 @opindex compatibility_version
10248 @opindex current_version
10249 @opindex dead_strip
10250 @opindex dependency-file
10251 @opindex dylib_file
10252 @opindex dylinker_install_name
10254 @opindex exported_symbols_list
10256 @opindex flat_namespace
10257 @opindex force_flat_namespace
10258 @opindex headerpad_max_install_names
10259 @opindex image_base
10261 @opindex install_name
10262 @opindex keep_private_externs
10263 @opindex multi_module
10264 @opindex multiply_defined
10265 @opindex multiply_defined_unused
10266 @opindex noall_load
10267 @opindex no_dead_strip_inits_and_terms
10268 @opindex nofixprebinding
10269 @opindex nomultidefs
10271 @opindex noseglinkedit
10272 @opindex pagezero_size
10274 @opindex prebind_all_twolevel_modules
10275 @opindex private_bundle
10276 @opindex read_only_relocs
10278 @opindex sectobjectsymbols
10281 @opindex sectcreate
10282 @opindex sectobjectsymbols
10285 @opindex segs_read_only_addr
10286 @opindex segs_read_write_addr
10287 @opindex seg_addr_table
10288 @opindex seg_addr_table_filename
10289 @opindex seglinkedit
10291 @opindex segs_read_only_addr
10292 @opindex segs_read_write_addr
10293 @opindex single_module
10295 @opindex sub_library
10296 @opindex sub_umbrella
10297 @opindex twolevel_namespace
10300 @opindex unexported_symbols_list
10301 @opindex weak_reference_mismatches
10302 @opindex whatsloaded
10303 These options are passed to the Darwin linker. The Darwin linker man page
10304 describes them in detail.
10307 @node DEC Alpha Options
10308 @subsection DEC Alpha Options
10310 These @samp{-m} options are defined for the DEC Alpha implementations:
10313 @item -mno-soft-float
10314 @itemx -msoft-float
10315 @opindex mno-soft-float
10316 @opindex msoft-float
10317 Use (do not use) the hardware floating-point instructions for
10318 floating-point operations. When @option{-msoft-float} is specified,
10319 functions in @file{libgcc.a} will be used to perform floating-point
10320 operations. Unless they are replaced by routines that emulate the
10321 floating-point operations, or compiled in such a way as to call such
10322 emulations routines, these routines will issue floating-point
10323 operations. If you are compiling for an Alpha without floating-point
10324 operations, you must ensure that the library is built so as not to call
10327 Note that Alpha implementations without floating-point operations are
10328 required to have floating-point registers.
10331 @itemx -mno-fp-regs
10333 @opindex mno-fp-regs
10334 Generate code that uses (does not use) the floating-point register set.
10335 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10336 register set is not used, floating point operands are passed in integer
10337 registers as if they were integers and floating-point results are passed
10338 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10339 so any function with a floating-point argument or return value called by code
10340 compiled with @option{-mno-fp-regs} must also be compiled with that
10343 A typical use of this option is building a kernel that does not use,
10344 and hence need not save and restore, any floating-point registers.
10348 The Alpha architecture implements floating-point hardware optimized for
10349 maximum performance. It is mostly compliant with the IEEE floating
10350 point standard. However, for full compliance, software assistance is
10351 required. This option generates code fully IEEE compliant code
10352 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10353 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10354 defined during compilation. The resulting code is less efficient but is
10355 able to correctly support denormalized numbers and exceptional IEEE
10356 values such as not-a-number and plus/minus infinity. Other Alpha
10357 compilers call this option @option{-ieee_with_no_inexact}.
10359 @item -mieee-with-inexact
10360 @opindex mieee-with-inexact
10361 This is like @option{-mieee} except the generated code also maintains
10362 the IEEE @var{inexact-flag}. Turning on this option causes the
10363 generated code to implement fully-compliant IEEE math. In addition to
10364 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10365 macro. On some Alpha implementations the resulting code may execute
10366 significantly slower than the code generated by default. Since there is
10367 very little code that depends on the @var{inexact-flag}, you should
10368 normally not specify this option. Other Alpha compilers call this
10369 option @option{-ieee_with_inexact}.
10371 @item -mfp-trap-mode=@var{trap-mode}
10372 @opindex mfp-trap-mode
10373 This option controls what floating-point related traps are enabled.
10374 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10375 The trap mode can be set to one of four values:
10379 This is the default (normal) setting. The only traps that are enabled
10380 are the ones that cannot be disabled in software (e.g., division by zero
10384 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10388 Like @samp{u}, but the instructions are marked to be safe for software
10389 completion (see Alpha architecture manual for details).
10392 Like @samp{su}, but inexact traps are enabled as well.
10395 @item -mfp-rounding-mode=@var{rounding-mode}
10396 @opindex mfp-rounding-mode
10397 Selects the IEEE rounding mode. Other Alpha compilers call this option
10398 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10403 Normal IEEE rounding mode. Floating point numbers are rounded towards
10404 the nearest machine number or towards the even machine number in case
10408 Round towards minus infinity.
10411 Chopped rounding mode. Floating point numbers are rounded towards zero.
10414 Dynamic rounding mode. A field in the floating point control register
10415 (@var{fpcr}, see Alpha architecture reference manual) controls the
10416 rounding mode in effect. The C library initializes this register for
10417 rounding towards plus infinity. Thus, unless your program modifies the
10418 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10421 @item -mtrap-precision=@var{trap-precision}
10422 @opindex mtrap-precision
10423 In the Alpha architecture, floating point traps are imprecise. This
10424 means without software assistance it is impossible to recover from a
10425 floating trap and program execution normally needs to be terminated.
10426 GCC can generate code that can assist operating system trap handlers
10427 in determining the exact location that caused a floating point trap.
10428 Depending on the requirements of an application, different levels of
10429 precisions can be selected:
10433 Program precision. This option is the default and means a trap handler
10434 can only identify which program caused a floating point exception.
10437 Function precision. The trap handler can determine the function that
10438 caused a floating point exception.
10441 Instruction precision. The trap handler can determine the exact
10442 instruction that caused a floating point exception.
10445 Other Alpha compilers provide the equivalent options called
10446 @option{-scope_safe} and @option{-resumption_safe}.
10448 @item -mieee-conformant
10449 @opindex mieee-conformant
10450 This option marks the generated code as IEEE conformant. You must not
10451 use this option unless you also specify @option{-mtrap-precision=i} and either
10452 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10453 is to emit the line @samp{.eflag 48} in the function prologue of the
10454 generated assembly file. Under DEC Unix, this has the effect that
10455 IEEE-conformant math library routines will be linked in.
10457 @item -mbuild-constants
10458 @opindex mbuild-constants
10459 Normally GCC examines a 32- or 64-bit integer constant to
10460 see if it can construct it from smaller constants in two or three
10461 instructions. If it cannot, it will output the constant as a literal and
10462 generate code to load it from the data segment at runtime.
10464 Use this option to require GCC to construct @emph{all} integer constants
10465 using code, even if it takes more instructions (the maximum is six).
10467 You would typically use this option to build a shared library dynamic
10468 loader. Itself a shared library, it must relocate itself in memory
10469 before it can find the variables and constants in its own data segment.
10475 Select whether to generate code to be assembled by the vendor-supplied
10476 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10494 Indicate whether GCC should generate code to use the optional BWX,
10495 CIX, FIX and MAX instruction sets. The default is to use the instruction
10496 sets supported by the CPU type specified via @option{-mcpu=} option or that
10497 of the CPU on which GCC was built if none was specified.
10500 @itemx -mfloat-ieee
10501 @opindex mfloat-vax
10502 @opindex mfloat-ieee
10503 Generate code that uses (does not use) VAX F and G floating point
10504 arithmetic instead of IEEE single and double precision.
10506 @item -mexplicit-relocs
10507 @itemx -mno-explicit-relocs
10508 @opindex mexplicit-relocs
10509 @opindex mno-explicit-relocs
10510 Older Alpha assemblers provided no way to generate symbol relocations
10511 except via assembler macros. Use of these macros does not allow
10512 optimal instruction scheduling. GNU binutils as of version 2.12
10513 supports a new syntax that allows the compiler to explicitly mark
10514 which relocations should apply to which instructions. This option
10515 is mostly useful for debugging, as GCC detects the capabilities of
10516 the assembler when it is built and sets the default accordingly.
10519 @itemx -mlarge-data
10520 @opindex msmall-data
10521 @opindex mlarge-data
10522 When @option{-mexplicit-relocs} is in effect, static data is
10523 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10524 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10525 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10526 16-bit relocations off of the @code{$gp} register. This limits the
10527 size of the small data area to 64KB, but allows the variables to be
10528 directly accessed via a single instruction.
10530 The default is @option{-mlarge-data}. With this option the data area
10531 is limited to just below 2GB@. Programs that require more than 2GB of
10532 data must use @code{malloc} or @code{mmap} to allocate the data in the
10533 heap instead of in the program's data segment.
10535 When generating code for shared libraries, @option{-fpic} implies
10536 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10539 @itemx -mlarge-text
10540 @opindex msmall-text
10541 @opindex mlarge-text
10542 When @option{-msmall-text} is used, the compiler assumes that the
10543 code of the entire program (or shared library) fits in 4MB, and is
10544 thus reachable with a branch instruction. When @option{-msmall-data}
10545 is used, the compiler can assume that all local symbols share the
10546 same @code{$gp} value, and thus reduce the number of instructions
10547 required for a function call from 4 to 1.
10549 The default is @option{-mlarge-text}.
10551 @item -mcpu=@var{cpu_type}
10553 Set the instruction set and instruction scheduling parameters for
10554 machine type @var{cpu_type}. You can specify either the @samp{EV}
10555 style name or the corresponding chip number. GCC supports scheduling
10556 parameters for the EV4, EV5 and EV6 family of processors and will
10557 choose the default values for the instruction set from the processor
10558 you specify. If you do not specify a processor type, GCC will default
10559 to the processor on which the compiler was built.
10561 Supported values for @var{cpu_type} are
10567 Schedules as an EV4 and has no instruction set extensions.
10571 Schedules as an EV5 and has no instruction set extensions.
10575 Schedules as an EV5 and supports the BWX extension.
10580 Schedules as an EV5 and supports the BWX and MAX extensions.
10584 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10588 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10591 Native Linux/GNU toolchains also support the value @samp{native},
10592 which selects the best architecture option for the host processor.
10593 @option{-mcpu=native} has no effect if GCC does not recognize
10596 @item -mtune=@var{cpu_type}
10598 Set only the instruction scheduling parameters for machine type
10599 @var{cpu_type}. The instruction set is not changed.
10601 Native Linux/GNU toolchains also support the value @samp{native},
10602 which selects the best architecture option for the host processor.
10603 @option{-mtune=native} has no effect if GCC does not recognize
10606 @item -mmemory-latency=@var{time}
10607 @opindex mmemory-latency
10608 Sets the latency the scheduler should assume for typical memory
10609 references as seen by the application. This number is highly
10610 dependent on the memory access patterns used by the application
10611 and the size of the external cache on the machine.
10613 Valid options for @var{time} are
10617 A decimal number representing clock cycles.
10623 The compiler contains estimates of the number of clock cycles for
10624 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10625 (also called Dcache, Scache, and Bcache), as well as to main memory.
10626 Note that L3 is only valid for EV5.
10631 @node DEC Alpha/VMS Options
10632 @subsection DEC Alpha/VMS Options
10634 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10637 @item -mvms-return-codes
10638 @opindex mvms-return-codes
10639 Return VMS condition codes from main. The default is to return POSIX
10640 style condition (e.g.@: error) codes.
10642 @item -mdebug-main=@var{prefix}
10643 @opindex mdebug-main=@var{prefix}
10644 Flag the first routine whose name starts with @var{prefix} as the main
10645 routine for the debugger.
10649 Default to 64bit memory allocation routines.
10653 @subsection FR30 Options
10654 @cindex FR30 Options
10656 These options are defined specifically for the FR30 port.
10660 @item -msmall-model
10661 @opindex msmall-model
10662 Use the small address space model. This can produce smaller code, but
10663 it does assume that all symbolic values and addresses will fit into a
10668 Assume that run-time support has been provided and so there is no need
10669 to include the simulator library (@file{libsim.a}) on the linker
10675 @subsection FRV Options
10676 @cindex FRV Options
10682 Only use the first 32 general purpose registers.
10687 Use all 64 general purpose registers.
10692 Use only the first 32 floating point registers.
10697 Use all 64 floating point registers
10700 @opindex mhard-float
10702 Use hardware instructions for floating point operations.
10705 @opindex msoft-float
10707 Use library routines for floating point operations.
10712 Dynamically allocate condition code registers.
10717 Do not try to dynamically allocate condition code registers, only
10718 use @code{icc0} and @code{fcc0}.
10723 Change ABI to use double word insns.
10728 Do not use double word instructions.
10733 Use floating point double instructions.
10736 @opindex mno-double
10738 Do not use floating point double instructions.
10743 Use media instructions.
10748 Do not use media instructions.
10753 Use multiply and add/subtract instructions.
10756 @opindex mno-muladd
10758 Do not use multiply and add/subtract instructions.
10763 Select the FDPIC ABI, that uses function descriptors to represent
10764 pointers to functions. Without any PIC/PIE-related options, it
10765 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10766 assumes GOT entries and small data are within a 12-bit range from the
10767 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10768 are computed with 32 bits.
10769 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10772 @opindex minline-plt
10774 Enable inlining of PLT entries in function calls to functions that are
10775 not known to bind locally. It has no effect without @option{-mfdpic}.
10776 It's enabled by default if optimizing for speed and compiling for
10777 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10778 optimization option such as @option{-O3} or above is present in the
10784 Assume a large TLS segment when generating thread-local code.
10789 Do not assume a large TLS segment when generating thread-local code.
10794 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10795 that is known to be in read-only sections. It's enabled by default,
10796 except for @option{-fpic} or @option{-fpie}: even though it may help
10797 make the global offset table smaller, it trades 1 instruction for 4.
10798 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10799 one of which may be shared by multiple symbols, and it avoids the need
10800 for a GOT entry for the referenced symbol, so it's more likely to be a
10801 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10803 @item -multilib-library-pic
10804 @opindex multilib-library-pic
10806 Link with the (library, not FD) pic libraries. It's implied by
10807 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10808 @option{-fpic} without @option{-mfdpic}. You should never have to use
10812 @opindex mlinked-fp
10814 Follow the EABI requirement of always creating a frame pointer whenever
10815 a stack frame is allocated. This option is enabled by default and can
10816 be disabled with @option{-mno-linked-fp}.
10819 @opindex mlong-calls
10821 Use indirect addressing to call functions outside the current
10822 compilation unit. This allows the functions to be placed anywhere
10823 within the 32-bit address space.
10825 @item -malign-labels
10826 @opindex malign-labels
10828 Try to align labels to an 8-byte boundary by inserting nops into the
10829 previous packet. This option only has an effect when VLIW packing
10830 is enabled. It doesn't create new packets; it merely adds nops to
10833 @item -mlibrary-pic
10834 @opindex mlibrary-pic
10836 Generate position-independent EABI code.
10841 Use only the first four media accumulator registers.
10846 Use all eight media accumulator registers.
10851 Pack VLIW instructions.
10856 Do not pack VLIW instructions.
10859 @opindex mno-eflags
10861 Do not mark ABI switches in e_flags.
10864 @opindex mcond-move
10866 Enable the use of conditional-move instructions (default).
10868 This switch is mainly for debugging the compiler and will likely be removed
10869 in a future version.
10871 @item -mno-cond-move
10872 @opindex mno-cond-move
10874 Disable the use of conditional-move instructions.
10876 This switch is mainly for debugging the compiler and will likely be removed
10877 in a future version.
10882 Enable the use of conditional set instructions (default).
10884 This switch is mainly for debugging the compiler and will likely be removed
10885 in a future version.
10890 Disable the use of conditional set instructions.
10892 This switch is mainly for debugging the compiler and will likely be removed
10893 in a future version.
10896 @opindex mcond-exec
10898 Enable the use of conditional execution (default).
10900 This switch is mainly for debugging the compiler and will likely be removed
10901 in a future version.
10903 @item -mno-cond-exec
10904 @opindex mno-cond-exec
10906 Disable the use of conditional execution.
10908 This switch is mainly for debugging the compiler and will likely be removed
10909 in a future version.
10911 @item -mvliw-branch
10912 @opindex mvliw-branch
10914 Run a pass to pack branches into VLIW instructions (default).
10916 This switch is mainly for debugging the compiler and will likely be removed
10917 in a future version.
10919 @item -mno-vliw-branch
10920 @opindex mno-vliw-branch
10922 Do not run a pass to pack branches into VLIW instructions.
10924 This switch is mainly for debugging the compiler and will likely be removed
10925 in a future version.
10927 @item -mmulti-cond-exec
10928 @opindex mmulti-cond-exec
10930 Enable optimization of @code{&&} and @code{||} in conditional execution
10933 This switch is mainly for debugging the compiler and will likely be removed
10934 in a future version.
10936 @item -mno-multi-cond-exec
10937 @opindex mno-multi-cond-exec
10939 Disable optimization of @code{&&} and @code{||} in conditional execution.
10941 This switch is mainly for debugging the compiler and will likely be removed
10942 in a future version.
10944 @item -mnested-cond-exec
10945 @opindex mnested-cond-exec
10947 Enable nested conditional execution optimizations (default).
10949 This switch is mainly for debugging the compiler and will likely be removed
10950 in a future version.
10952 @item -mno-nested-cond-exec
10953 @opindex mno-nested-cond-exec
10955 Disable nested conditional execution optimizations.
10957 This switch is mainly for debugging the compiler and will likely be removed
10958 in a future version.
10960 @item -moptimize-membar
10961 @opindex moptimize-membar
10963 This switch removes redundant @code{membar} instructions from the
10964 compiler generated code. It is enabled by default.
10966 @item -mno-optimize-membar
10967 @opindex mno-optimize-membar
10969 This switch disables the automatic removal of redundant @code{membar}
10970 instructions from the generated code.
10972 @item -mtomcat-stats
10973 @opindex mtomcat-stats
10975 Cause gas to print out tomcat statistics.
10977 @item -mcpu=@var{cpu}
10980 Select the processor type for which to generate code. Possible values are
10981 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10982 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10986 @node GNU/Linux Options
10987 @subsection GNU/Linux Options
10989 These @samp{-m} options are defined for GNU/Linux targets:
10994 Use the GNU C library instead of uClibc. This is the default except
10995 on @samp{*-*-linux-*uclibc*} targets.
10999 Use uClibc instead of the GNU C library. This is the default on
11000 @samp{*-*-linux-*uclibc*} targets.
11003 @node H8/300 Options
11004 @subsection H8/300 Options
11006 These @samp{-m} options are defined for the H8/300 implementations:
11011 Shorten some address references at link time, when possible; uses the
11012 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11013 ld, Using ld}, for a fuller description.
11017 Generate code for the H8/300H@.
11021 Generate code for the H8S@.
11025 Generate code for the H8S and H8/300H in the normal mode. This switch
11026 must be used either with @option{-mh} or @option{-ms}.
11030 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11034 Make @code{int} data 32 bits by default.
11037 @opindex malign-300
11038 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11039 The default for the H8/300H and H8S is to align longs and floats on 4
11041 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11042 This option has no effect on the H8/300.
11046 @subsection HPPA Options
11047 @cindex HPPA Options
11049 These @samp{-m} options are defined for the HPPA family of computers:
11052 @item -march=@var{architecture-type}
11054 Generate code for the specified architecture. The choices for
11055 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11056 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11057 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11058 architecture option for your machine. Code compiled for lower numbered
11059 architectures will run on higher numbered architectures, but not the
11062 @item -mpa-risc-1-0
11063 @itemx -mpa-risc-1-1
11064 @itemx -mpa-risc-2-0
11065 @opindex mpa-risc-1-0
11066 @opindex mpa-risc-1-1
11067 @opindex mpa-risc-2-0
11068 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11071 @opindex mbig-switch
11072 Generate code suitable for big switch tables. Use this option only if
11073 the assembler/linker complain about out of range branches within a switch
11076 @item -mjump-in-delay
11077 @opindex mjump-in-delay
11078 Fill delay slots of function calls with unconditional jump instructions
11079 by modifying the return pointer for the function call to be the target
11080 of the conditional jump.
11082 @item -mdisable-fpregs
11083 @opindex mdisable-fpregs
11084 Prevent floating point registers from being used in any manner. This is
11085 necessary for compiling kernels which perform lazy context switching of
11086 floating point registers. If you use this option and attempt to perform
11087 floating point operations, the compiler will abort.
11089 @item -mdisable-indexing
11090 @opindex mdisable-indexing
11091 Prevent the compiler from using indexing address modes. This avoids some
11092 rather obscure problems when compiling MIG generated code under MACH@.
11094 @item -mno-space-regs
11095 @opindex mno-space-regs
11096 Generate code that assumes the target has no space registers. This allows
11097 GCC to generate faster indirect calls and use unscaled index address modes.
11099 Such code is suitable for level 0 PA systems and kernels.
11101 @item -mfast-indirect-calls
11102 @opindex mfast-indirect-calls
11103 Generate code that assumes calls never cross space boundaries. This
11104 allows GCC to emit code which performs faster indirect calls.
11106 This option will not work in the presence of shared libraries or nested
11109 @item -mfixed-range=@var{register-range}
11110 @opindex mfixed-range
11111 Generate code treating the given register range as fixed registers.
11112 A fixed register is one that the register allocator can not use. This is
11113 useful when compiling kernel code. A register range is specified as
11114 two registers separated by a dash. Multiple register ranges can be
11115 specified separated by a comma.
11117 @item -mlong-load-store
11118 @opindex mlong-load-store
11119 Generate 3-instruction load and store sequences as sometimes required by
11120 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11123 @item -mportable-runtime
11124 @opindex mportable-runtime
11125 Use the portable calling conventions proposed by HP for ELF systems.
11129 Enable the use of assembler directives only GAS understands.
11131 @item -mschedule=@var{cpu-type}
11133 Schedule code according to the constraints for the machine type
11134 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11135 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11136 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11137 proper scheduling option for your machine. The default scheduling is
11141 @opindex mlinker-opt
11142 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11143 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11144 linkers in which they give bogus error messages when linking some programs.
11147 @opindex msoft-float
11148 Generate output containing library calls for floating point.
11149 @strong{Warning:} the requisite libraries are not available for all HPPA
11150 targets. Normally the facilities of the machine's usual C compiler are
11151 used, but this cannot be done directly in cross-compilation. You must make
11152 your own arrangements to provide suitable library functions for
11155 @option{-msoft-float} changes the calling convention in the output file;
11156 therefore, it is only useful if you compile @emph{all} of a program with
11157 this option. In particular, you need to compile @file{libgcc.a}, the
11158 library that comes with GCC, with @option{-msoft-float} in order for
11163 Generate the predefine, @code{_SIO}, for server IO@. The default is
11164 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11165 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11166 options are available under HP-UX and HI-UX@.
11170 Use GNU ld specific options. This passes @option{-shared} to ld when
11171 building a shared library. It is the default when GCC is configured,
11172 explicitly or implicitly, with the GNU linker. This option does not
11173 have any affect on which ld is called, it only changes what parameters
11174 are passed to that ld. The ld that is called is determined by the
11175 @option{--with-ld} configure option, GCC's program search path, and
11176 finally by the user's @env{PATH}. The linker used by GCC can be printed
11177 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11178 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11182 Use HP ld specific options. This passes @option{-b} to ld when building
11183 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11184 links. It is the default when GCC is configured, explicitly or
11185 implicitly, with the HP linker. This option does not have any affect on
11186 which ld is called, it only changes what parameters are passed to that
11187 ld. The ld that is called is determined by the @option{--with-ld}
11188 configure option, GCC's program search path, and finally by the user's
11189 @env{PATH}. The linker used by GCC can be printed using @samp{which
11190 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11191 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11194 @opindex mno-long-calls
11195 Generate code that uses long call sequences. This ensures that a call
11196 is always able to reach linker generated stubs. The default is to generate
11197 long calls only when the distance from the call site to the beginning
11198 of the function or translation unit, as the case may be, exceeds a
11199 predefined limit set by the branch type being used. The limits for
11200 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11201 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11204 Distances are measured from the beginning of functions when using the
11205 @option{-ffunction-sections} option, or when using the @option{-mgas}
11206 and @option{-mno-portable-runtime} options together under HP-UX with
11209 It is normally not desirable to use this option as it will degrade
11210 performance. However, it may be useful in large applications,
11211 particularly when partial linking is used to build the application.
11213 The types of long calls used depends on the capabilities of the
11214 assembler and linker, and the type of code being generated. The
11215 impact on systems that support long absolute calls, and long pic
11216 symbol-difference or pc-relative calls should be relatively small.
11217 However, an indirect call is used on 32-bit ELF systems in pic code
11218 and it is quite long.
11220 @item -munix=@var{unix-std}
11222 Generate compiler predefines and select a startfile for the specified
11223 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11224 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11225 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11226 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11227 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11230 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11231 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11232 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11233 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11234 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11235 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11237 It is @emph{important} to note that this option changes the interfaces
11238 for various library routines. It also affects the operational behavior
11239 of the C library. Thus, @emph{extreme} care is needed in using this
11242 Library code that is intended to operate with more than one UNIX
11243 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11244 as appropriate. Most GNU software doesn't provide this capability.
11248 Suppress the generation of link options to search libdld.sl when the
11249 @option{-static} option is specified on HP-UX 10 and later.
11253 The HP-UX implementation of setlocale in libc has a dependency on
11254 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11255 when the @option{-static} option is specified, special link options
11256 are needed to resolve this dependency.
11258 On HP-UX 10 and later, the GCC driver adds the necessary options to
11259 link with libdld.sl when the @option{-static} option is specified.
11260 This causes the resulting binary to be dynamic. On the 64-bit port,
11261 the linkers generate dynamic binaries by default in any case. The
11262 @option{-nolibdld} option can be used to prevent the GCC driver from
11263 adding these link options.
11267 Add support for multithreading with the @dfn{dce thread} library
11268 under HP-UX@. This option sets flags for both the preprocessor and
11272 @node i386 and x86-64 Options
11273 @subsection Intel 386 and AMD x86-64 Options
11274 @cindex i386 Options
11275 @cindex x86-64 Options
11276 @cindex Intel 386 Options
11277 @cindex AMD x86-64 Options
11279 These @samp{-m} options are defined for the i386 and x86-64 family of
11283 @item -mtune=@var{cpu-type}
11285 Tune to @var{cpu-type} everything applicable about the generated code, except
11286 for the ABI and the set of available instructions. The choices for
11287 @var{cpu-type} are:
11290 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11291 If you know the CPU on which your code will run, then you should use
11292 the corresponding @option{-mtune} option instead of
11293 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11294 of your application will have, then you should use this option.
11296 As new processors are deployed in the marketplace, the behavior of this
11297 option will change. Therefore, if you upgrade to a newer version of
11298 GCC, the code generated option will change to reflect the processors
11299 that were most common when that version of GCC was released.
11301 There is no @option{-march=generic} option because @option{-march}
11302 indicates the instruction set the compiler can use, and there is no
11303 generic instruction set applicable to all processors. In contrast,
11304 @option{-mtune} indicates the processor (or, in this case, collection of
11305 processors) for which the code is optimized.
11307 This selects the CPU to tune for at compilation time by determining
11308 the processor type of the compiling machine. Using @option{-mtune=native}
11309 will produce code optimized for the local machine under the constraints
11310 of the selected instruction set. Using @option{-march=native} will
11311 enable all instruction subsets supported by the local machine (hence
11312 the result might not run on different machines).
11314 Original Intel's i386 CPU@.
11316 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11317 @item i586, pentium
11318 Intel Pentium CPU with no MMX support.
11320 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11322 Intel PentiumPro CPU@.
11324 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11325 instruction set will be used, so the code will run on all i686 family chips.
11327 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11328 @item pentium3, pentium3m
11329 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11332 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11333 support. Used by Centrino notebooks.
11334 @item pentium4, pentium4m
11335 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11337 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11340 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11341 SSE2 and SSE3 instruction set support.
11343 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11344 instruction set support.
11346 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11347 instruction set support.
11349 AMD K6 CPU with MMX instruction set support.
11351 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11352 @item athlon, athlon-tbird
11353 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11355 @item athlon-4, athlon-xp, athlon-mp
11356 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11357 instruction set support.
11358 @item k8, opteron, athlon64, athlon-fx
11359 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11360 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11361 @item k8-sse3, opteron-sse3, athlon64-sse3
11362 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11363 @item amdfam10, barcelona
11364 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11365 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11366 instruction set extensions.)
11368 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11371 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11372 instruction set support.
11374 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11375 implemented for this chip.)
11377 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11378 implemented for this chip.)
11380 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11383 While picking a specific @var{cpu-type} will schedule things appropriately
11384 for that particular chip, the compiler will not generate any code that
11385 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11388 @item -march=@var{cpu-type}
11390 Generate instructions for the machine type @var{cpu-type}. The choices
11391 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11392 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11394 @item -mcpu=@var{cpu-type}
11396 A deprecated synonym for @option{-mtune}.
11398 @item -mfpmath=@var{unit}
11400 Generate floating point arithmetics for selected unit @var{unit}. The choices
11401 for @var{unit} are:
11405 Use the standard 387 floating point coprocessor present majority of chips and
11406 emulated otherwise. Code compiled with this option will run almost everywhere.
11407 The temporary results are computed in 80bit precision instead of precision
11408 specified by the type resulting in slightly different results compared to most
11409 of other chips. See @option{-ffloat-store} for more detailed description.
11411 This is the default choice for i386 compiler.
11414 Use scalar floating point instructions present in the SSE instruction set.
11415 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11416 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11417 instruction set supports only single precision arithmetics, thus the double and
11418 extended precision arithmetics is still done using 387. Later version, present
11419 only in Pentium4 and the future AMD x86-64 chips supports double precision
11422 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11423 or @option{-msse2} switches to enable SSE extensions and make this option
11424 effective. For the x86-64 compiler, these extensions are enabled by default.
11426 The resulting code should be considerably faster in the majority of cases and avoid
11427 the numerical instability problems of 387 code, but may break some existing
11428 code that expects temporaries to be 80bit.
11430 This is the default choice for the x86-64 compiler.
11435 Attempt to utilize both instruction sets at once. This effectively double the
11436 amount of available registers and on chips with separate execution units for
11437 387 and SSE the execution resources too. Use this option with care, as it is
11438 still experimental, because the GCC register allocator does not model separate
11439 functional units well resulting in instable performance.
11442 @item -masm=@var{dialect}
11443 @opindex masm=@var{dialect}
11444 Output asm instructions using selected @var{dialect}. Supported
11445 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11446 not support @samp{intel}.
11449 @itemx -mno-ieee-fp
11451 @opindex mno-ieee-fp
11452 Control whether or not the compiler uses IEEE floating point
11453 comparisons. These handle correctly the case where the result of a
11454 comparison is unordered.
11457 @opindex msoft-float
11458 Generate output containing library calls for floating point.
11459 @strong{Warning:} the requisite libraries are not part of GCC@.
11460 Normally the facilities of the machine's usual C compiler are used, but
11461 this can't be done directly in cross-compilation. You must make your
11462 own arrangements to provide suitable library functions for
11465 On machines where a function returns floating point results in the 80387
11466 register stack, some floating point opcodes may be emitted even if
11467 @option{-msoft-float} is used.
11469 @item -mno-fp-ret-in-387
11470 @opindex mno-fp-ret-in-387
11471 Do not use the FPU registers for return values of functions.
11473 The usual calling convention has functions return values of types
11474 @code{float} and @code{double} in an FPU register, even if there
11475 is no FPU@. The idea is that the operating system should emulate
11478 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11479 in ordinary CPU registers instead.
11481 @item -mno-fancy-math-387
11482 @opindex mno-fancy-math-387
11483 Some 387 emulators do not support the @code{sin}, @code{cos} and
11484 @code{sqrt} instructions for the 387. Specify this option to avoid
11485 generating those instructions. This option is the default on FreeBSD,
11486 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11487 indicates that the target cpu will always have an FPU and so the
11488 instruction will not need emulation. As of revision 2.6.1, these
11489 instructions are not generated unless you also use the
11490 @option{-funsafe-math-optimizations} switch.
11492 @item -malign-double
11493 @itemx -mno-align-double
11494 @opindex malign-double
11495 @opindex mno-align-double
11496 Control whether GCC aligns @code{double}, @code{long double}, and
11497 @code{long long} variables on a two word boundary or a one word
11498 boundary. Aligning @code{double} variables on a two word boundary will
11499 produce code that runs somewhat faster on a @samp{Pentium} at the
11500 expense of more memory.
11502 On x86-64, @option{-malign-double} is enabled by default.
11504 @strong{Warning:} if you use the @option{-malign-double} switch,
11505 structures containing the above types will be aligned differently than
11506 the published application binary interface specifications for the 386
11507 and will not be binary compatible with structures in code compiled
11508 without that switch.
11510 @item -m96bit-long-double
11511 @itemx -m128bit-long-double
11512 @opindex m96bit-long-double
11513 @opindex m128bit-long-double
11514 These switches control the size of @code{long double} type. The i386
11515 application binary interface specifies the size to be 96 bits,
11516 so @option{-m96bit-long-double} is the default in 32 bit mode.
11518 Modern architectures (Pentium and newer) would prefer @code{long double}
11519 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11520 conforming to the ABI, this would not be possible. So specifying a
11521 @option{-m128bit-long-double} will align @code{long double}
11522 to a 16 byte boundary by padding the @code{long double} with an additional
11525 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11526 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11528 Notice that neither of these options enable any extra precision over the x87
11529 standard of 80 bits for a @code{long double}.
11531 @strong{Warning:} if you override the default value for your target ABI, the
11532 structures and arrays containing @code{long double} variables will change
11533 their size as well as function calling convention for function taking
11534 @code{long double} will be modified. Hence they will not be binary
11535 compatible with arrays or structures in code compiled without that switch.
11537 @item -mlarge-data-threshold=@var{number}
11538 @opindex mlarge-data-threshold=@var{number}
11539 When @option{-mcmodel=medium} is specified, the data greater than
11540 @var{threshold} are placed in large data section. This value must be the
11541 same across all object linked into the binary and defaults to 65535.
11545 Use a different function-calling convention, in which functions that
11546 take a fixed number of arguments return with the @code{ret} @var{num}
11547 instruction, which pops their arguments while returning. This saves one
11548 instruction in the caller since there is no need to pop the arguments
11551 You can specify that an individual function is called with this calling
11552 sequence with the function attribute @samp{stdcall}. You can also
11553 override the @option{-mrtd} option by using the function attribute
11554 @samp{cdecl}. @xref{Function Attributes}.
11556 @strong{Warning:} this calling convention is incompatible with the one
11557 normally used on Unix, so you cannot use it if you need to call
11558 libraries compiled with the Unix compiler.
11560 Also, you must provide function prototypes for all functions that
11561 take variable numbers of arguments (including @code{printf});
11562 otherwise incorrect code will be generated for calls to those
11565 In addition, seriously incorrect code will result if you call a
11566 function with too many arguments. (Normally, extra arguments are
11567 harmlessly ignored.)
11569 @item -mregparm=@var{num}
11571 Control how many registers are used to pass integer arguments. By
11572 default, no registers are used to pass arguments, and at most 3
11573 registers can be used. You can control this behavior for a specific
11574 function by using the function attribute @samp{regparm}.
11575 @xref{Function Attributes}.
11577 @strong{Warning:} if you use this switch, and
11578 @var{num} is nonzero, then you must build all modules with the same
11579 value, including any libraries. This includes the system libraries and
11583 @opindex msseregparm
11584 Use SSE register passing conventions for float and double arguments
11585 and return values. You can control this behavior for a specific
11586 function by using the function attribute @samp{sseregparm}.
11587 @xref{Function Attributes}.
11589 @strong{Warning:} if you use this switch then you must build all
11590 modules with the same value, including any libraries. This includes
11591 the system libraries and startup modules.
11600 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11601 is specified, the significands of results of floating-point operations are
11602 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11603 significands of results of floating-point operations to 53 bits (double
11604 precision) and @option{-mpc80} rounds the significands of results of
11605 floating-point operations to 64 bits (extended double precision), which is
11606 the default. When this option is used, floating-point operations in higher
11607 precisions are not available to the programmer without setting the FPU
11608 control word explicitly.
11610 Setting the rounding of floating-point operations to less than the default
11611 80 bits can speed some programs by 2% or more. Note that some mathematical
11612 libraries assume that extended precision (80 bit) floating-point operations
11613 are enabled by default; routines in such libraries could suffer significant
11614 loss of accuracy, typically through so-called "catastrophic cancellation",
11615 when this option is used to set the precision to less than extended precision.
11617 @item -mstackrealign
11618 @opindex mstackrealign
11619 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11620 option will generate an alternate prologue and epilogue that realigns the
11621 runtime stack if necessary. This supports mixing legacy codes that keep
11622 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11623 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11624 applicable to individual functions.
11626 @item -mpreferred-stack-boundary=@var{num}
11627 @opindex mpreferred-stack-boundary
11628 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11629 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11630 the default is 4 (16 bytes or 128 bits).
11632 @item -mincoming-stack-boundary=@var{num}
11633 @opindex mincoming-stack-boundary
11634 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11635 boundary. If @option{-mincoming-stack-boundary} is not specified,
11636 the one specified by @option{-mpreferred-stack-boundary} will be used.
11638 On Pentium and PentiumPro, @code{double} and @code{long double} values
11639 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11640 suffer significant run time performance penalties. On Pentium III, the
11641 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11642 properly if it is not 16 byte aligned.
11644 To ensure proper alignment of this values on the stack, the stack boundary
11645 must be as aligned as that required by any value stored on the stack.
11646 Further, every function must be generated such that it keeps the stack
11647 aligned. Thus calling a function compiled with a higher preferred
11648 stack boundary from a function compiled with a lower preferred stack
11649 boundary will most likely misalign the stack. It is recommended that
11650 libraries that use callbacks always use the default setting.
11652 This extra alignment does consume extra stack space, and generally
11653 increases code size. Code that is sensitive to stack space usage, such
11654 as embedded systems and operating system kernels, may want to reduce the
11655 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11693 These switches enable or disable the use of instructions in the MMX,
11694 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, ABM or
11695 3DNow!@: extended instruction sets.
11696 These extensions are also available as built-in functions: see
11697 @ref{X86 Built-in Functions}, for details of the functions enabled and
11698 disabled by these switches.
11700 To have SSE/SSE2 instructions generated automatically from floating-point
11701 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11703 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11704 generates new AVX instructions or AVX equivalence for all SSEx instructions
11707 These options will enable GCC to use these extended instructions in
11708 generated code, even without @option{-mfpmath=sse}. Applications which
11709 perform runtime CPU detection must compile separate files for each
11710 supported architecture, using the appropriate flags. In particular,
11711 the file containing the CPU detection code should be compiled without
11716 This option instructs GCC to emit a @code{cld} instruction in the prologue
11717 of functions that use string instructions. String instructions depend on
11718 the DF flag to select between autoincrement or autodecrement mode. While the
11719 ABI specifies the DF flag to be cleared on function entry, some operating
11720 systems violate this specification by not clearing the DF flag in their
11721 exception dispatchers. The exception handler can be invoked with the DF flag
11722 set which leads to wrong direction mode, when string instructions are used.
11723 This option can be enabled by default on 32-bit x86 targets by configuring
11724 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11725 instructions can be suppressed with the @option{-mno-cld} compiler option
11730 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11731 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11732 data types. This is useful for high resolution counters that could be updated
11733 by multiple processors (or cores). This instruction is generated as part of
11734 atomic built-in functions: see @ref{Atomic Builtins} for details.
11738 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11739 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11740 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11741 SAHF are load and store instructions, respectively, for certain status flags.
11742 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11743 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11747 This option will enable GCC to use movbe instruction to implement
11748 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
11752 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
11753 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
11754 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
11758 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11759 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11760 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11761 variants) for single precision floating point arguments. These instructions
11762 are generated only when @option{-funsafe-math-optimizations} is enabled
11763 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11764 Note that while the throughput of the sequence is higher than the throughput
11765 of the non-reciprocal instruction, the precision of the sequence can be
11766 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11768 @item -mveclibabi=@var{type}
11769 @opindex mveclibabi
11770 Specifies the ABI type to use for vectorizing intrinsics using an
11771 external library. Supported types are @code{svml} for the Intel short
11772 vector math library and @code{acml} for the AMD math core library style
11773 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11774 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11775 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11776 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11777 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11778 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11779 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11780 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11781 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11782 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11783 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11784 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11785 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11786 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11787 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11788 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11789 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11790 compatible library will have to be specified at link time.
11792 @item -mabi=@var{name}
11794 Generate code for the specified calling convention. Permissible values
11795 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11796 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11797 ABI when targeting Windows. On all other systems, the default is the
11798 SYSV ABI. You can control this behavior for a specific function by
11799 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11800 @xref{Function Attributes}.
11803 @itemx -mno-push-args
11804 @opindex mpush-args
11805 @opindex mno-push-args
11806 Use PUSH operations to store outgoing parameters. This method is shorter
11807 and usually equally fast as method using SUB/MOV operations and is enabled
11808 by default. In some cases disabling it may improve performance because of
11809 improved scheduling and reduced dependencies.
11811 @item -maccumulate-outgoing-args
11812 @opindex maccumulate-outgoing-args
11813 If enabled, the maximum amount of space required for outgoing arguments will be
11814 computed in the function prologue. This is faster on most modern CPUs
11815 because of reduced dependencies, improved scheduling and reduced stack usage
11816 when preferred stack boundary is not equal to 2. The drawback is a notable
11817 increase in code size. This switch implies @option{-mno-push-args}.
11821 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11822 on thread-safe exception handling must compile and link all code with the
11823 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11824 @option{-D_MT}; when linking, it links in a special thread helper library
11825 @option{-lmingwthrd} which cleans up per thread exception handling data.
11827 @item -mno-align-stringops
11828 @opindex mno-align-stringops
11829 Do not align destination of inlined string operations. This switch reduces
11830 code size and improves performance in case the destination is already aligned,
11831 but GCC doesn't know about it.
11833 @item -minline-all-stringops
11834 @opindex minline-all-stringops
11835 By default GCC inlines string operations only when destination is known to be
11836 aligned at least to 4 byte boundary. This enables more inlining, increase code
11837 size, but may improve performance of code that depends on fast memcpy, strlen
11838 and memset for short lengths.
11840 @item -minline-stringops-dynamically
11841 @opindex minline-stringops-dynamically
11842 For string operation of unknown size, inline runtime checks so for small
11843 blocks inline code is used, while for large blocks library call is used.
11845 @item -mstringop-strategy=@var{alg}
11846 @opindex mstringop-strategy=@var{alg}
11847 Overwrite internal decision heuristic about particular algorithm to inline
11848 string operation with. The allowed values are @code{rep_byte},
11849 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11850 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11851 expanding inline loop, @code{libcall} for always expanding library call.
11853 @item -momit-leaf-frame-pointer
11854 @opindex momit-leaf-frame-pointer
11855 Don't keep the frame pointer in a register for leaf functions. This
11856 avoids the instructions to save, set up and restore frame pointers and
11857 makes an extra register available in leaf functions. The option
11858 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11859 which might make debugging harder.
11861 @item -mtls-direct-seg-refs
11862 @itemx -mno-tls-direct-seg-refs
11863 @opindex mtls-direct-seg-refs
11864 Controls whether TLS variables may be accessed with offsets from the
11865 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11866 or whether the thread base pointer must be added. Whether or not this
11867 is legal depends on the operating system, and whether it maps the
11868 segment to cover the entire TLS area.
11870 For systems that use GNU libc, the default is on.
11873 @itemx -mno-sse2avx
11875 Specify that the assembler should encode SSE instructions with VEX
11876 prefix. The option @option{-mavx} turns this on by default.
11879 These @samp{-m} switches are supported in addition to the above
11880 on AMD x86-64 processors in 64-bit environments.
11887 Generate code for a 32-bit or 64-bit environment.
11888 The 32-bit environment sets int, long and pointer to 32 bits and
11889 generates code that runs on any i386 system.
11890 The 64-bit environment sets int to 32 bits and long and pointer
11891 to 64 bits and generates code for AMD's x86-64 architecture. For
11892 darwin only the -m64 option turns off the @option{-fno-pic} and
11893 @option{-mdynamic-no-pic} options.
11895 @item -mno-red-zone
11896 @opindex mno-red-zone
11897 Do not use a so called red zone for x86-64 code. The red zone is mandated
11898 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11899 stack pointer that will not be modified by signal or interrupt handlers
11900 and therefore can be used for temporary data without adjusting the stack
11901 pointer. The flag @option{-mno-red-zone} disables this red zone.
11903 @item -mcmodel=small
11904 @opindex mcmodel=small
11905 Generate code for the small code model: the program and its symbols must
11906 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11907 Programs can be statically or dynamically linked. This is the default
11910 @item -mcmodel=kernel
11911 @opindex mcmodel=kernel
11912 Generate code for the kernel code model. The kernel runs in the
11913 negative 2 GB of the address space.
11914 This model has to be used for Linux kernel code.
11916 @item -mcmodel=medium
11917 @opindex mcmodel=medium
11918 Generate code for the medium model: The program is linked in the lower 2
11919 GB of the address space. Small symbols are also placed there. Symbols
11920 with sizes larger than @option{-mlarge-data-threshold} are put into
11921 large data or bss sections and can be located above 2GB. Programs can
11922 be statically or dynamically linked.
11924 @item -mcmodel=large
11925 @opindex mcmodel=large
11926 Generate code for the large model: This model makes no assumptions
11927 about addresses and sizes of sections.
11930 @node IA-64 Options
11931 @subsection IA-64 Options
11932 @cindex IA-64 Options
11934 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11938 @opindex mbig-endian
11939 Generate code for a big endian target. This is the default for HP-UX@.
11941 @item -mlittle-endian
11942 @opindex mlittle-endian
11943 Generate code for a little endian target. This is the default for AIX5
11949 @opindex mno-gnu-as
11950 Generate (or don't) code for the GNU assembler. This is the default.
11951 @c Also, this is the default if the configure option @option{--with-gnu-as}
11957 @opindex mno-gnu-ld
11958 Generate (or don't) code for the GNU linker. This is the default.
11959 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11964 Generate code that does not use a global pointer register. The result
11965 is not position independent code, and violates the IA-64 ABI@.
11967 @item -mvolatile-asm-stop
11968 @itemx -mno-volatile-asm-stop
11969 @opindex mvolatile-asm-stop
11970 @opindex mno-volatile-asm-stop
11971 Generate (or don't) a stop bit immediately before and after volatile asm
11974 @item -mregister-names
11975 @itemx -mno-register-names
11976 @opindex mregister-names
11977 @opindex mno-register-names
11978 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11979 the stacked registers. This may make assembler output more readable.
11985 Disable (or enable) optimizations that use the small data section. This may
11986 be useful for working around optimizer bugs.
11988 @item -mconstant-gp
11989 @opindex mconstant-gp
11990 Generate code that uses a single constant global pointer value. This is
11991 useful when compiling kernel code.
11995 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11996 This is useful when compiling firmware code.
11998 @item -minline-float-divide-min-latency
11999 @opindex minline-float-divide-min-latency
12000 Generate code for inline divides of floating point values
12001 using the minimum latency algorithm.
12003 @item -minline-float-divide-max-throughput
12004 @opindex minline-float-divide-max-throughput
12005 Generate code for inline divides of floating point values
12006 using the maximum throughput algorithm.
12008 @item -mno-inline-float-divide
12009 @opindex mno-inline-float-divide
12010 Do not generate inline code for divides of floating point values.
12012 @item -minline-int-divide-min-latency
12013 @opindex minline-int-divide-min-latency
12014 Generate code for inline divides of integer values
12015 using the minimum latency algorithm.
12017 @item -minline-int-divide-max-throughput
12018 @opindex minline-int-divide-max-throughput
12019 Generate code for inline divides of integer values
12020 using the maximum throughput algorithm.
12022 @item -mno-inline-int-divide
12023 @opindex mno-inline-int-divide
12024 Do not generate inline code for divides of integer values.
12026 @item -minline-sqrt-min-latency
12027 @opindex minline-sqrt-min-latency
12028 Generate code for inline square roots
12029 using the minimum latency algorithm.
12031 @item -minline-sqrt-max-throughput
12032 @opindex minline-sqrt-max-throughput
12033 Generate code for inline square roots
12034 using the maximum throughput algorithm.
12036 @item -mno-inline-sqrt
12037 @opindex mno-inline-sqrt
12038 Do not generate inline code for sqrt.
12041 @itemx -mno-fused-madd
12042 @opindex mfused-madd
12043 @opindex mno-fused-madd
12044 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12045 instructions. The default is to use these instructions.
12047 @item -mno-dwarf2-asm
12048 @itemx -mdwarf2-asm
12049 @opindex mno-dwarf2-asm
12050 @opindex mdwarf2-asm
12051 Don't (or do) generate assembler code for the DWARF2 line number debugging
12052 info. This may be useful when not using the GNU assembler.
12054 @item -mearly-stop-bits
12055 @itemx -mno-early-stop-bits
12056 @opindex mearly-stop-bits
12057 @opindex mno-early-stop-bits
12058 Allow stop bits to be placed earlier than immediately preceding the
12059 instruction that triggered the stop bit. This can improve instruction
12060 scheduling, but does not always do so.
12062 @item -mfixed-range=@var{register-range}
12063 @opindex mfixed-range
12064 Generate code treating the given register range as fixed registers.
12065 A fixed register is one that the register allocator can not use. This is
12066 useful when compiling kernel code. A register range is specified as
12067 two registers separated by a dash. Multiple register ranges can be
12068 specified separated by a comma.
12070 @item -mtls-size=@var{tls-size}
12072 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12075 @item -mtune=@var{cpu-type}
12077 Tune the instruction scheduling for a particular CPU, Valid values are
12078 itanium, itanium1, merced, itanium2, and mckinley.
12084 Generate code for a 32-bit or 64-bit environment.
12085 The 32-bit environment sets int, long and pointer to 32 bits.
12086 The 64-bit environment sets int to 32 bits and long and pointer
12087 to 64 bits. These are HP-UX specific flags.
12089 @item -mno-sched-br-data-spec
12090 @itemx -msched-br-data-spec
12091 @opindex mno-sched-br-data-spec
12092 @opindex msched-br-data-spec
12093 (Dis/En)able data speculative scheduling before reload.
12094 This will result in generation of the ld.a instructions and
12095 the corresponding check instructions (ld.c / chk.a).
12096 The default is 'disable'.
12098 @item -msched-ar-data-spec
12099 @itemx -mno-sched-ar-data-spec
12100 @opindex msched-ar-data-spec
12101 @opindex mno-sched-ar-data-spec
12102 (En/Dis)able data speculative scheduling after reload.
12103 This will result in generation of the ld.a instructions and
12104 the corresponding check instructions (ld.c / chk.a).
12105 The default is 'enable'.
12107 @item -mno-sched-control-spec
12108 @itemx -msched-control-spec
12109 @opindex mno-sched-control-spec
12110 @opindex msched-control-spec
12111 (Dis/En)able control speculative scheduling. This feature is
12112 available only during region scheduling (i.e.@: before reload).
12113 This will result in generation of the ld.s instructions and
12114 the corresponding check instructions chk.s .
12115 The default is 'disable'.
12117 @item -msched-br-in-data-spec
12118 @itemx -mno-sched-br-in-data-spec
12119 @opindex msched-br-in-data-spec
12120 @opindex mno-sched-br-in-data-spec
12121 (En/Dis)able speculative scheduling of the instructions that
12122 are dependent on the data speculative loads before reload.
12123 This is effective only with @option{-msched-br-data-spec} enabled.
12124 The default is 'enable'.
12126 @item -msched-ar-in-data-spec
12127 @itemx -mno-sched-ar-in-data-spec
12128 @opindex msched-ar-in-data-spec
12129 @opindex mno-sched-ar-in-data-spec
12130 (En/Dis)able speculative scheduling of the instructions that
12131 are dependent on the data speculative loads after reload.
12132 This is effective only with @option{-msched-ar-data-spec} enabled.
12133 The default is 'enable'.
12135 @item -msched-in-control-spec
12136 @itemx -mno-sched-in-control-spec
12137 @opindex msched-in-control-spec
12138 @opindex mno-sched-in-control-spec
12139 (En/Dis)able speculative scheduling of the instructions that
12140 are dependent on the control speculative loads.
12141 This is effective only with @option{-msched-control-spec} enabled.
12142 The default is 'enable'.
12144 @item -mno-sched-prefer-non-data-spec-insns
12145 @itemx -msched-prefer-non-data-spec-insns
12146 @opindex mno-sched-prefer-non-data-spec-insns
12147 @opindex msched-prefer-non-data-spec-insns
12148 If enabled, data speculative instructions will be chosen for schedule
12149 only if there are no other choices at the moment. This will make
12150 the use of the data speculation much more conservative.
12151 The default is 'disable'.
12153 @item -mno-sched-prefer-non-control-spec-insns
12154 @itemx -msched-prefer-non-control-spec-insns
12155 @opindex mno-sched-prefer-non-control-spec-insns
12156 @opindex msched-prefer-non-control-spec-insns
12157 If enabled, control speculative instructions will be chosen for schedule
12158 only if there are no other choices at the moment. This will make
12159 the use of the control speculation much more conservative.
12160 The default is 'disable'.
12162 @item -mno-sched-count-spec-in-critical-path
12163 @itemx -msched-count-spec-in-critical-path
12164 @opindex mno-sched-count-spec-in-critical-path
12165 @opindex msched-count-spec-in-critical-path
12166 If enabled, speculative dependencies will be considered during
12167 computation of the instructions priorities. This will make the use of the
12168 speculation a bit more conservative.
12169 The default is 'disable'.
12171 @item -msched-spec-ldc
12172 @opindex msched-spec-ldc
12173 Use a simple data speculation check. This option is on by default.
12175 @item -msched-control-spec-ldc
12176 @opindex msched-spec-ldc
12177 Use a simple check for control speculation. This option is on by default.
12179 @item -msched-stop-bits-after-every-cycle
12180 @opindex msched-stop-bits-after-every-cycle
12181 Place a stop bit after every cycle when scheduling. This option is on
12184 @item -msched-fp-mem-deps-zero-cost
12185 @opindex msched-fp-mem-deps-zero-cost
12186 Assume that floating-point stores and loads are not likely to cause a conflict
12187 when placed into the same instruction group. This option is disabled by
12190 @item -msel-sched-dont-check-control-spec
12191 @opindex msel-sched-dont-check-control-spec
12192 Generate checks for control speculation in selective scheduling.
12193 This flag is disabled by default.
12195 @item -msched-max-memory-insns=@var{max-insns}
12196 @opindex msched-max-memory-insns
12197 Limit on the number of memory insns per instruction group, giving lower
12198 priority to subsequent memory insns attempting to schedule in the same
12199 instruction group. Frequently useful to prevent cache bank conflicts.
12200 The default value is 1.
12202 @item -msched-max-memory-insns-hard-limit
12203 @opindex msched-max-memory-insns-hard-limit
12204 Disallow more than `msched-max-memory-insns' in instruction group.
12205 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12206 when limit is reached but may still schedule memory operations.
12210 @node IA-64/VMS Options
12211 @subsection IA-64/VMS Options
12213 These @samp{-m} options are defined for the IA-64/VMS implementations:
12216 @item -mvms-return-codes
12217 @opindex mvms-return-codes
12218 Return VMS condition codes from main. The default is to return POSIX
12219 style condition (e.g.@ error) codes.
12221 @item -mdebug-main=@var{prefix}
12222 @opindex mdebug-main=@var{prefix}
12223 Flag the first routine whose name starts with @var{prefix} as the main
12224 routine for the debugger.
12228 Default to 64bit memory allocation routines.
12232 @subsection M32C Options
12233 @cindex M32C options
12236 @item -mcpu=@var{name}
12238 Select the CPU for which code is generated. @var{name} may be one of
12239 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12240 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12241 the M32C/80 series.
12245 Specifies that the program will be run on the simulator. This causes
12246 an alternate runtime library to be linked in which supports, for
12247 example, file I/O@. You must not use this option when generating
12248 programs that will run on real hardware; you must provide your own
12249 runtime library for whatever I/O functions are needed.
12251 @item -memregs=@var{number}
12253 Specifies the number of memory-based pseudo-registers GCC will use
12254 during code generation. These pseudo-registers will be used like real
12255 registers, so there is a tradeoff between GCC's ability to fit the
12256 code into available registers, and the performance penalty of using
12257 memory instead of registers. Note that all modules in a program must
12258 be compiled with the same value for this option. Because of that, you
12259 must not use this option with the default runtime libraries gcc
12264 @node M32R/D Options
12265 @subsection M32R/D Options
12266 @cindex M32R/D options
12268 These @option{-m} options are defined for Renesas M32R/D architectures:
12273 Generate code for the M32R/2@.
12277 Generate code for the M32R/X@.
12281 Generate code for the M32R@. This is the default.
12283 @item -mmodel=small
12284 @opindex mmodel=small
12285 Assume all objects live in the lower 16MB of memory (so that their addresses
12286 can be loaded with the @code{ld24} instruction), and assume all subroutines
12287 are reachable with the @code{bl} instruction.
12288 This is the default.
12290 The addressability of a particular object can be set with the
12291 @code{model} attribute.
12293 @item -mmodel=medium
12294 @opindex mmodel=medium
12295 Assume objects may be anywhere in the 32-bit address space (the compiler
12296 will generate @code{seth/add3} instructions to load their addresses), and
12297 assume all subroutines are reachable with the @code{bl} instruction.
12299 @item -mmodel=large
12300 @opindex mmodel=large
12301 Assume objects may be anywhere in the 32-bit address space (the compiler
12302 will generate @code{seth/add3} instructions to load their addresses), and
12303 assume subroutines may not be reachable with the @code{bl} instruction
12304 (the compiler will generate the much slower @code{seth/add3/jl}
12305 instruction sequence).
12308 @opindex msdata=none
12309 Disable use of the small data area. Variables will be put into
12310 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12311 @code{section} attribute has been specified).
12312 This is the default.
12314 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12315 Objects may be explicitly put in the small data area with the
12316 @code{section} attribute using one of these sections.
12318 @item -msdata=sdata
12319 @opindex msdata=sdata
12320 Put small global and static data in the small data area, but do not
12321 generate special code to reference them.
12324 @opindex msdata=use
12325 Put small global and static data in the small data area, and generate
12326 special instructions to reference them.
12330 @cindex smaller data references
12331 Put global and static objects less than or equal to @var{num} bytes
12332 into the small data or bss sections instead of the normal data or bss
12333 sections. The default value of @var{num} is 8.
12334 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12335 for this option to have any effect.
12337 All modules should be compiled with the same @option{-G @var{num}} value.
12338 Compiling with different values of @var{num} may or may not work; if it
12339 doesn't the linker will give an error message---incorrect code will not be
12344 Makes the M32R specific code in the compiler display some statistics
12345 that might help in debugging programs.
12347 @item -malign-loops
12348 @opindex malign-loops
12349 Align all loops to a 32-byte boundary.
12351 @item -mno-align-loops
12352 @opindex mno-align-loops
12353 Do not enforce a 32-byte alignment for loops. This is the default.
12355 @item -missue-rate=@var{number}
12356 @opindex missue-rate=@var{number}
12357 Issue @var{number} instructions per cycle. @var{number} can only be 1
12360 @item -mbranch-cost=@var{number}
12361 @opindex mbranch-cost=@var{number}
12362 @var{number} can only be 1 or 2. If it is 1 then branches will be
12363 preferred over conditional code, if it is 2, then the opposite will
12366 @item -mflush-trap=@var{number}
12367 @opindex mflush-trap=@var{number}
12368 Specifies the trap number to use to flush the cache. The default is
12369 12. Valid numbers are between 0 and 15 inclusive.
12371 @item -mno-flush-trap
12372 @opindex mno-flush-trap
12373 Specifies that the cache cannot be flushed by using a trap.
12375 @item -mflush-func=@var{name}
12376 @opindex mflush-func=@var{name}
12377 Specifies the name of the operating system function to call to flush
12378 the cache. The default is @emph{_flush_cache}, but a function call
12379 will only be used if a trap is not available.
12381 @item -mno-flush-func
12382 @opindex mno-flush-func
12383 Indicates that there is no OS function for flushing the cache.
12387 @node M680x0 Options
12388 @subsection M680x0 Options
12389 @cindex M680x0 options
12391 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12392 The default settings depend on which architecture was selected when
12393 the compiler was configured; the defaults for the most common choices
12397 @item -march=@var{arch}
12399 Generate code for a specific M680x0 or ColdFire instruction set
12400 architecture. Permissible values of @var{arch} for M680x0
12401 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12402 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12403 architectures are selected according to Freescale's ISA classification
12404 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12405 @samp{isab} and @samp{isac}.
12407 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12408 code for a ColdFire target. The @var{arch} in this macro is one of the
12409 @option{-march} arguments given above.
12411 When used together, @option{-march} and @option{-mtune} select code
12412 that runs on a family of similar processors but that is optimized
12413 for a particular microarchitecture.
12415 @item -mcpu=@var{cpu}
12417 Generate code for a specific M680x0 or ColdFire processor.
12418 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12419 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12420 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12421 below, which also classifies the CPUs into families:
12423 @multitable @columnfractions 0.20 0.80
12424 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12425 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12426 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12427 @item @samp{5206e} @tab @samp{5206e}
12428 @item @samp{5208} @tab @samp{5207} @samp{5208}
12429 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12430 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12431 @item @samp{5216} @tab @samp{5214} @samp{5216}
12432 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12433 @item @samp{5225} @tab @samp{5224} @samp{5225}
12434 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12435 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12436 @item @samp{5249} @tab @samp{5249}
12437 @item @samp{5250} @tab @samp{5250}
12438 @item @samp{5271} @tab @samp{5270} @samp{5271}
12439 @item @samp{5272} @tab @samp{5272}
12440 @item @samp{5275} @tab @samp{5274} @samp{5275}
12441 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12442 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12443 @item @samp{5307} @tab @samp{5307}
12444 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12445 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12446 @item @samp{5407} @tab @samp{5407}
12447 @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}
12450 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12451 @var{arch} is compatible with @var{cpu}. Other combinations of
12452 @option{-mcpu} and @option{-march} are rejected.
12454 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12455 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12456 where the value of @var{family} is given by the table above.
12458 @item -mtune=@var{tune}
12460 Tune the code for a particular microarchitecture, within the
12461 constraints set by @option{-march} and @option{-mcpu}.
12462 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12463 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12464 and @samp{cpu32}. The ColdFire microarchitectures
12465 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12467 You can also use @option{-mtune=68020-40} for code that needs
12468 to run relatively well on 68020, 68030 and 68040 targets.
12469 @option{-mtune=68020-60} is similar but includes 68060 targets
12470 as well. These two options select the same tuning decisions as
12471 @option{-m68020-40} and @option{-m68020-60} respectively.
12473 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12474 when tuning for 680x0 architecture @var{arch}. It also defines
12475 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12476 option is used. If gcc is tuning for a range of architectures,
12477 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12478 it defines the macros for every architecture in the range.
12480 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12481 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12482 of the arguments given above.
12488 Generate output for a 68000. This is the default
12489 when the compiler is configured for 68000-based systems.
12490 It is equivalent to @option{-march=68000}.
12492 Use this option for microcontrollers with a 68000 or EC000 core,
12493 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12497 Generate output for a 68010. This is the default
12498 when the compiler is configured for 68010-based systems.
12499 It is equivalent to @option{-march=68010}.
12505 Generate output for a 68020. This is the default
12506 when the compiler is configured for 68020-based systems.
12507 It is equivalent to @option{-march=68020}.
12511 Generate output for a 68030. This is the default when the compiler is
12512 configured for 68030-based systems. It is equivalent to
12513 @option{-march=68030}.
12517 Generate output for a 68040. This is the default when the compiler is
12518 configured for 68040-based systems. It is equivalent to
12519 @option{-march=68040}.
12521 This option inhibits the use of 68881/68882 instructions that have to be
12522 emulated by software on the 68040. Use this option if your 68040 does not
12523 have code to emulate those instructions.
12527 Generate output for a 68060. This is the default when the compiler is
12528 configured for 68060-based systems. It is equivalent to
12529 @option{-march=68060}.
12531 This option inhibits the use of 68020 and 68881/68882 instructions that
12532 have to be emulated by software on the 68060. Use this option if your 68060
12533 does not have code to emulate those instructions.
12537 Generate output for a CPU32. This is the default
12538 when the compiler is configured for CPU32-based systems.
12539 It is equivalent to @option{-march=cpu32}.
12541 Use this option for microcontrollers with a
12542 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12543 68336, 68340, 68341, 68349 and 68360.
12547 Generate output for a 520X ColdFire CPU@. This is the default
12548 when the compiler is configured for 520X-based systems.
12549 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12550 in favor of that option.
12552 Use this option for microcontroller with a 5200 core, including
12553 the MCF5202, MCF5203, MCF5204 and MCF5206.
12557 Generate output for a 5206e ColdFire CPU@. The option is now
12558 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12562 Generate output for a member of the ColdFire 528X family.
12563 The option is now deprecated in favor of the equivalent
12564 @option{-mcpu=528x}.
12568 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12569 in favor of the equivalent @option{-mcpu=5307}.
12573 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12574 in favor of the equivalent @option{-mcpu=5407}.
12578 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12579 This includes use of hardware floating point instructions.
12580 The option is equivalent to @option{-mcpu=547x}, and is now
12581 deprecated in favor of that option.
12585 Generate output for a 68040, without using any of the new instructions.
12586 This results in code which can run relatively efficiently on either a
12587 68020/68881 or a 68030 or a 68040. The generated code does use the
12588 68881 instructions that are emulated on the 68040.
12590 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12594 Generate output for a 68060, without using any of the new instructions.
12595 This results in code which can run relatively efficiently on either a
12596 68020/68881 or a 68030 or a 68040. The generated code does use the
12597 68881 instructions that are emulated on the 68060.
12599 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12603 @opindex mhard-float
12605 Generate floating-point instructions. This is the default for 68020
12606 and above, and for ColdFire devices that have an FPU@. It defines the
12607 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12608 on ColdFire targets.
12611 @opindex msoft-float
12612 Do not generate floating-point instructions; use library calls instead.
12613 This is the default for 68000, 68010, and 68832 targets. It is also
12614 the default for ColdFire devices that have no FPU.
12620 Generate (do not generate) ColdFire hardware divide and remainder
12621 instructions. If @option{-march} is used without @option{-mcpu},
12622 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12623 architectures. Otherwise, the default is taken from the target CPU
12624 (either the default CPU, or the one specified by @option{-mcpu}). For
12625 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12626 @option{-mcpu=5206e}.
12628 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12632 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12633 Additionally, parameters passed on the stack are also aligned to a
12634 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12638 Do not consider type @code{int} to be 16 bits wide. This is the default.
12641 @itemx -mno-bitfield
12642 @opindex mnobitfield
12643 @opindex mno-bitfield
12644 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12645 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12649 Do use the bit-field instructions. The @option{-m68020} option implies
12650 @option{-mbitfield}. This is the default if you use a configuration
12651 designed for a 68020.
12655 Use a different function-calling convention, in which functions
12656 that take a fixed number of arguments return with the @code{rtd}
12657 instruction, which pops their arguments while returning. This
12658 saves one instruction in the caller since there is no need to pop
12659 the arguments there.
12661 This calling convention is incompatible with the one normally
12662 used on Unix, so you cannot use it if you need to call libraries
12663 compiled with the Unix compiler.
12665 Also, you must provide function prototypes for all functions that
12666 take variable numbers of arguments (including @code{printf});
12667 otherwise incorrect code will be generated for calls to those
12670 In addition, seriously incorrect code will result if you call a
12671 function with too many arguments. (Normally, extra arguments are
12672 harmlessly ignored.)
12674 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12675 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12679 Do not use the calling conventions selected by @option{-mrtd}.
12680 This is the default.
12683 @itemx -mno-align-int
12684 @opindex malign-int
12685 @opindex mno-align-int
12686 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12687 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12688 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12689 Aligning variables on 32-bit boundaries produces code that runs somewhat
12690 faster on processors with 32-bit busses at the expense of more memory.
12692 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12693 align structures containing the above types differently than
12694 most published application binary interface specifications for the m68k.
12698 Use the pc-relative addressing mode of the 68000 directly, instead of
12699 using a global offset table. At present, this option implies @option{-fpic},
12700 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12701 not presently supported with @option{-mpcrel}, though this could be supported for
12702 68020 and higher processors.
12704 @item -mno-strict-align
12705 @itemx -mstrict-align
12706 @opindex mno-strict-align
12707 @opindex mstrict-align
12708 Do not (do) assume that unaligned memory references will be handled by
12712 Generate code that allows the data segment to be located in a different
12713 area of memory from the text segment. This allows for execute in place in
12714 an environment without virtual memory management. This option implies
12717 @item -mno-sep-data
12718 Generate code that assumes that the data segment follows the text segment.
12719 This is the default.
12721 @item -mid-shared-library
12722 Generate code that supports shared libraries via the library ID method.
12723 This allows for execute in place and shared libraries in an environment
12724 without virtual memory management. This option implies @option{-fPIC}.
12726 @item -mno-id-shared-library
12727 Generate code that doesn't assume ID based shared libraries are being used.
12728 This is the default.
12730 @item -mshared-library-id=n
12731 Specified the identification number of the ID based shared library being
12732 compiled. Specifying a value of 0 will generate more compact code, specifying
12733 other values will force the allocation of that number to the current
12734 library but is no more space or time efficient than omitting this option.
12740 When generating position-independent code for ColdFire, generate code
12741 that works if the GOT has more than 8192 entries. This code is
12742 larger and slower than code generated without this option. On M680x0
12743 processors, this option is not needed; @option{-fPIC} suffices.
12745 GCC normally uses a single instruction to load values from the GOT@.
12746 While this is relatively efficient, it only works if the GOT
12747 is smaller than about 64k. Anything larger causes the linker
12748 to report an error such as:
12750 @cindex relocation truncated to fit (ColdFire)
12752 relocation truncated to fit: R_68K_GOT16O foobar
12755 If this happens, you should recompile your code with @option{-mxgot}.
12756 It should then work with very large GOTs. However, code generated with
12757 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12758 the value of a global symbol.
12760 Note that some linkers, including newer versions of the GNU linker,
12761 can create multiple GOTs and sort GOT entries. If you have such a linker,
12762 you should only need to use @option{-mxgot} when compiling a single
12763 object file that accesses more than 8192 GOT entries. Very few do.
12765 These options have no effect unless GCC is generating
12766 position-independent code.
12770 @node M68hc1x Options
12771 @subsection M68hc1x Options
12772 @cindex M68hc1x options
12774 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12775 microcontrollers. The default values for these options depends on
12776 which style of microcontroller was selected when the compiler was configured;
12777 the defaults for the most common choices are given below.
12784 Generate output for a 68HC11. This is the default
12785 when the compiler is configured for 68HC11-based systems.
12791 Generate output for a 68HC12. This is the default
12792 when the compiler is configured for 68HC12-based systems.
12798 Generate output for a 68HCS12.
12800 @item -mauto-incdec
12801 @opindex mauto-incdec
12802 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12809 Enable the use of 68HC12 min and max instructions.
12812 @itemx -mno-long-calls
12813 @opindex mlong-calls
12814 @opindex mno-long-calls
12815 Treat all calls as being far away (near). If calls are assumed to be
12816 far away, the compiler will use the @code{call} instruction to
12817 call a function and the @code{rtc} instruction for returning.
12821 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12823 @item -msoft-reg-count=@var{count}
12824 @opindex msoft-reg-count
12825 Specify the number of pseudo-soft registers which are used for the
12826 code generation. The maximum number is 32. Using more pseudo-soft
12827 register may or may not result in better code depending on the program.
12828 The default is 4 for 68HC11 and 2 for 68HC12.
12832 @node MCore Options
12833 @subsection MCore Options
12834 @cindex MCore options
12836 These are the @samp{-m} options defined for the Motorola M*Core
12842 @itemx -mno-hardlit
12844 @opindex mno-hardlit
12845 Inline constants into the code stream if it can be done in two
12846 instructions or less.
12852 Use the divide instruction. (Enabled by default).
12854 @item -mrelax-immediate
12855 @itemx -mno-relax-immediate
12856 @opindex mrelax-immediate
12857 @opindex mno-relax-immediate
12858 Allow arbitrary sized immediates in bit operations.
12860 @item -mwide-bitfields
12861 @itemx -mno-wide-bitfields
12862 @opindex mwide-bitfields
12863 @opindex mno-wide-bitfields
12864 Always treat bit-fields as int-sized.
12866 @item -m4byte-functions
12867 @itemx -mno-4byte-functions
12868 @opindex m4byte-functions
12869 @opindex mno-4byte-functions
12870 Force all functions to be aligned to a four byte boundary.
12872 @item -mcallgraph-data
12873 @itemx -mno-callgraph-data
12874 @opindex mcallgraph-data
12875 @opindex mno-callgraph-data
12876 Emit callgraph information.
12879 @itemx -mno-slow-bytes
12880 @opindex mslow-bytes
12881 @opindex mno-slow-bytes
12882 Prefer word access when reading byte quantities.
12884 @item -mlittle-endian
12885 @itemx -mbig-endian
12886 @opindex mlittle-endian
12887 @opindex mbig-endian
12888 Generate code for a little endian target.
12894 Generate code for the 210 processor.
12898 Assume that run-time support has been provided and so omit the
12899 simulator library (@file{libsim.a)} from the linker command line.
12901 @item -mstack-increment=@var{size}
12902 @opindex mstack-increment
12903 Set the maximum amount for a single stack increment operation. Large
12904 values can increase the speed of programs which contain functions
12905 that need a large amount of stack space, but they can also trigger a
12906 segmentation fault if the stack is extended too much. The default
12912 @subsection MeP Options
12913 @cindex MeP options
12919 Enables the @code{abs} instruction, which is the absolute difference
12920 between two registers.
12924 Enables all the optional instructions - average, multiply, divide, bit
12925 operations, leading zero, absolute difference, min/max, clip, and
12931 Enables the @code{ave} instruction, which computes the average of two
12934 @item -mbased=@var{n}
12936 Variables of size @var{n} bytes or smaller will be placed in the
12937 @code{.based} section by default. Based variables use the @code{$tp}
12938 register as a base register, and there is a 128 byte limit to the
12939 @code{.based} section.
12943 Enables the bit operation instructions - bit test (@code{btstm}), set
12944 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
12945 test-and-set (@code{tas}).
12947 @item -mc=@var{name}
12949 Selects which section constant data will be placed in. @var{name} may
12950 be @code{tiny}, @code{near}, or @code{far}.
12954 Enables the @code{clip} instruction. Note that @code{-mclip} is not
12955 useful unless you also provide @code{-mminmax}.
12957 @item -mconfig=@var{name}
12959 Selects one of the build-in core configurations. Each MeP chip has
12960 one or more modules in it; each module has a core CPU and a variety of
12961 coprocessors, optional instructions, and peripherals. The
12962 @code{MeP-Integrator} tool, not part of GCC, provides these
12963 configurations through this option; using this option is the same as
12964 using all the corresponding command line options. The default
12965 configuration is @code{default}.
12969 Enables the coprocessor instructions. By default, this is a 32-bit
12970 coprocessor. Note that the coprocessor is normally enabled via the
12971 @code{-mconfig=} option.
12975 Enables the 32-bit coprocessor's instructions.
12979 Enables the 64-bit coprocessor's instructions.
12983 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
12987 Causes constant variables to be placed in the @code{.near} section.
12991 Enables the @code{div} and @code{divu} instructions.
12995 Generate big-endian code.
12999 Generate little-endian code.
13001 @item -mio-volatile
13002 @opindex mio-volatile
13003 Tells the compiler that any variable marked with the @code{io}
13004 attribute is to be considered volatile.
13008 Causes variables to be assigned to the @code{.far} section by default.
13012 Enables the @code{leadz} (leading zero) instruction.
13016 Causes variables to be assigned to the @code{.near} section by default.
13020 Enables the @code{min} and @code{max} instructions.
13024 Enables the multiplication and multiply-accumulate instructions.
13028 Disables all the optional instructions enabled by @code{-mall-opts}.
13032 Enables the @code{repeat} and @code{erepeat} instructions, used for
13033 low-overhead looping.
13037 Causes all variables to default to the @code{.tiny} section. Note
13038 that there is a 65536 byte limit to this section. Accesses to these
13039 variables use the @code{%gp} base register.
13043 Enables the saturation instructions. Note that the compiler does not
13044 currently generate these itself, but this option is included for
13045 compatibility with other tools, like @code{as}.
13049 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13053 Link the simulator runtime libraries.
13057 Link the simulator runtime libraries, excluding built-in support
13058 for reset and exception vectors and tables.
13062 Causes all functions to default to the @code{.far} section. Without
13063 this option, functions default to the @code{.near} section.
13065 @item -mtiny=@var{n}
13067 Variables that are @var{n} bytes or smaller will be allocated to the
13068 @code{.tiny} section. These variables use the @code{$gp} base
13069 register. The default for this option is 4, but note that there's a
13070 65536 byte limit to the @code{.tiny} section.
13075 @subsection MIPS Options
13076 @cindex MIPS options
13082 Generate big-endian code.
13086 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13089 @item -march=@var{arch}
13091 Generate code that will run on @var{arch}, which can be the name of a
13092 generic MIPS ISA, or the name of a particular processor.
13094 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13095 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13096 The processor names are:
13097 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13098 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13099 @samp{5kc}, @samp{5kf},
13101 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13102 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13103 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13104 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13105 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13106 @samp{loongson2e}, @samp{loongson2f},
13110 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13111 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13112 @samp{rm7000}, @samp{rm9000},
13113 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13116 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13117 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13119 The special value @samp{from-abi} selects the
13120 most compatible architecture for the selected ABI (that is,
13121 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13123 Native Linux/GNU toolchains also support the value @samp{native},
13124 which selects the best architecture option for the host processor.
13125 @option{-march=native} has no effect if GCC does not recognize
13128 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13129 (for example, @samp{-march=r2k}). Prefixes are optional, and
13130 @samp{vr} may be written @samp{r}.
13132 Names of the form @samp{@var{n}f2_1} refer to processors with
13133 FPUs clocked at half the rate of the core, names of the form
13134 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13135 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13136 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13137 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13138 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13139 accepted as synonyms for @samp{@var{n}f1_1}.
13141 GCC defines two macros based on the value of this option. The first
13142 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13143 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13144 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13145 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13146 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13148 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13149 above. In other words, it will have the full prefix and will not
13150 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13151 the macro names the resolved architecture (either @samp{"mips1"} or
13152 @samp{"mips3"}). It names the default architecture when no
13153 @option{-march} option is given.
13155 @item -mtune=@var{arch}
13157 Optimize for @var{arch}. Among other things, this option controls
13158 the way instructions are scheduled, and the perceived cost of arithmetic
13159 operations. The list of @var{arch} values is the same as for
13162 When this option is not used, GCC will optimize for the processor
13163 specified by @option{-march}. By using @option{-march} and
13164 @option{-mtune} together, it is possible to generate code that will
13165 run on a family of processors, but optimize the code for one
13166 particular member of that family.
13168 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13169 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13170 @samp{-march} ones described above.
13174 Equivalent to @samp{-march=mips1}.
13178 Equivalent to @samp{-march=mips2}.
13182 Equivalent to @samp{-march=mips3}.
13186 Equivalent to @samp{-march=mips4}.
13190 Equivalent to @samp{-march=mips32}.
13194 Equivalent to @samp{-march=mips32r2}.
13198 Equivalent to @samp{-march=mips64}.
13202 Equivalent to @samp{-march=mips64r2}.
13207 @opindex mno-mips16
13208 Generate (do not generate) MIPS16 code. If GCC is targetting a
13209 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13211 MIPS16 code generation can also be controlled on a per-function basis
13212 by means of @code{mips16} and @code{nomips16} attributes.
13213 @xref{Function Attributes}, for more information.
13215 @item -mflip-mips16
13216 @opindex mflip-mips16
13217 Generate MIPS16 code on alternating functions. This option is provided
13218 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13219 not intended for ordinary use in compiling user code.
13221 @item -minterlink-mips16
13222 @itemx -mno-interlink-mips16
13223 @opindex minterlink-mips16
13224 @opindex mno-interlink-mips16
13225 Require (do not require) that non-MIPS16 code be link-compatible with
13228 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13229 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13230 therefore disables direct jumps unless GCC knows that the target of the
13231 jump is not MIPS16.
13243 Generate code for the given ABI@.
13245 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13246 generates 64-bit code when you select a 64-bit architecture, but you
13247 can use @option{-mgp32} to get 32-bit code instead.
13249 For information about the O64 ABI, see
13250 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13252 GCC supports a variant of the o32 ABI in which floating-point registers
13253 are 64 rather than 32 bits wide. You can select this combination with
13254 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13255 and @samp{mfhc1} instructions and is therefore only supported for
13256 MIPS32R2 processors.
13258 The register assignments for arguments and return values remain the
13259 same, but each scalar value is passed in a single 64-bit register
13260 rather than a pair of 32-bit registers. For example, scalar
13261 floating-point values are returned in @samp{$f0} only, not a
13262 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13263 remains the same, but all 64 bits are saved.
13266 @itemx -mno-abicalls
13268 @opindex mno-abicalls
13269 Generate (do not generate) code that is suitable for SVR4-style
13270 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13275 Generate (do not generate) code that is fully position-independent,
13276 and that can therefore be linked into shared libraries. This option
13277 only affects @option{-mabicalls}.
13279 All @option{-mabicalls} code has traditionally been position-independent,
13280 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13281 as an extension, the GNU toolchain allows executables to use absolute
13282 accesses for locally-binding symbols. It can also use shorter GP
13283 initialization sequences and generate direct calls to locally-defined
13284 functions. This mode is selected by @option{-mno-shared}.
13286 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13287 objects that can only be linked by the GNU linker. However, the option
13288 does not affect the ABI of the final executable; it only affects the ABI
13289 of relocatable objects. Using @option{-mno-shared} will generally make
13290 executables both smaller and quicker.
13292 @option{-mshared} is the default.
13298 Assume (do not assume) that the static and dynamic linkers
13299 support PLTs and copy relocations. This option only affects
13300 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13301 has no effect without @samp{-msym32}.
13303 You can make @option{-mplt} the default by configuring
13304 GCC with @option{--with-mips-plt}. The default is
13305 @option{-mno-plt} otherwise.
13311 Lift (do not lift) the usual restrictions on the size of the global
13314 GCC normally uses a single instruction to load values from the GOT@.
13315 While this is relatively efficient, it will only work if the GOT
13316 is smaller than about 64k. Anything larger will cause the linker
13317 to report an error such as:
13319 @cindex relocation truncated to fit (MIPS)
13321 relocation truncated to fit: R_MIPS_GOT16 foobar
13324 If this happens, you should recompile your code with @option{-mxgot}.
13325 It should then work with very large GOTs, although it will also be
13326 less efficient, since it will take three instructions to fetch the
13327 value of a global symbol.
13329 Note that some linkers can create multiple GOTs. If you have such a
13330 linker, you should only need to use @option{-mxgot} when a single object
13331 file accesses more than 64k's worth of GOT entries. Very few do.
13333 These options have no effect unless GCC is generating position
13338 Assume that general-purpose registers are 32 bits wide.
13342 Assume that general-purpose registers are 64 bits wide.
13346 Assume that floating-point registers are 32 bits wide.
13350 Assume that floating-point registers are 64 bits wide.
13353 @opindex mhard-float
13354 Use floating-point coprocessor instructions.
13357 @opindex msoft-float
13358 Do not use floating-point coprocessor instructions. Implement
13359 floating-point calculations using library calls instead.
13361 @item -msingle-float
13362 @opindex msingle-float
13363 Assume that the floating-point coprocessor only supports single-precision
13366 @item -mdouble-float
13367 @opindex mdouble-float
13368 Assume that the floating-point coprocessor supports double-precision
13369 operations. This is the default.
13375 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13376 implement atomic memory built-in functions. When neither option is
13377 specified, GCC will use the instructions if the target architecture
13380 @option{-mllsc} is useful if the runtime environment can emulate the
13381 instructions and @option{-mno-llsc} can be useful when compiling for
13382 nonstandard ISAs. You can make either option the default by
13383 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13384 respectively. @option{--with-llsc} is the default for some
13385 configurations; see the installation documentation for details.
13391 Use (do not use) revision 1 of the MIPS DSP ASE@.
13392 @xref{MIPS DSP Built-in Functions}. This option defines the
13393 preprocessor macro @samp{__mips_dsp}. It also defines
13394 @samp{__mips_dsp_rev} to 1.
13400 Use (do not use) revision 2 of the MIPS DSP ASE@.
13401 @xref{MIPS DSP Built-in Functions}. This option defines the
13402 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13403 It also defines @samp{__mips_dsp_rev} to 2.
13406 @itemx -mno-smartmips
13407 @opindex msmartmips
13408 @opindex mno-smartmips
13409 Use (do not use) the MIPS SmartMIPS ASE.
13411 @item -mpaired-single
13412 @itemx -mno-paired-single
13413 @opindex mpaired-single
13414 @opindex mno-paired-single
13415 Use (do not use) paired-single floating-point instructions.
13416 @xref{MIPS Paired-Single Support}. This option requires
13417 hardware floating-point support to be enabled.
13423 Use (do not use) MIPS Digital Media Extension instructions.
13424 This option can only be used when generating 64-bit code and requires
13425 hardware floating-point support to be enabled.
13430 @opindex mno-mips3d
13431 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13432 The option @option{-mips3d} implies @option{-mpaired-single}.
13438 Use (do not use) MT Multithreading instructions.
13442 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13443 an explanation of the default and the way that the pointer size is
13448 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13450 The default size of @code{int}s, @code{long}s and pointers depends on
13451 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13452 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13453 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13454 or the same size as integer registers, whichever is smaller.
13460 Assume (do not assume) that all symbols have 32-bit values, regardless
13461 of the selected ABI@. This option is useful in combination with
13462 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13463 to generate shorter and faster references to symbolic addresses.
13467 Put definitions of externally-visible data in a small data section
13468 if that data is no bigger than @var{num} bytes. GCC can then access
13469 the data more efficiently; see @option{-mgpopt} for details.
13471 The default @option{-G} option depends on the configuration.
13473 @item -mlocal-sdata
13474 @itemx -mno-local-sdata
13475 @opindex mlocal-sdata
13476 @opindex mno-local-sdata
13477 Extend (do not extend) the @option{-G} behavior to local data too,
13478 such as to static variables in C@. @option{-mlocal-sdata} is the
13479 default for all configurations.
13481 If the linker complains that an application is using too much small data,
13482 you might want to try rebuilding the less performance-critical parts with
13483 @option{-mno-local-sdata}. You might also want to build large
13484 libraries with @option{-mno-local-sdata}, so that the libraries leave
13485 more room for the main program.
13487 @item -mextern-sdata
13488 @itemx -mno-extern-sdata
13489 @opindex mextern-sdata
13490 @opindex mno-extern-sdata
13491 Assume (do not assume) that externally-defined data will be in
13492 a small data section if that data is within the @option{-G} limit.
13493 @option{-mextern-sdata} is the default for all configurations.
13495 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13496 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13497 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13498 is placed in a small data section. If @var{Var} is defined by another
13499 module, you must either compile that module with a high-enough
13500 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13501 definition. If @var{Var} is common, you must link the application
13502 with a high-enough @option{-G} setting.
13504 The easiest way of satisfying these restrictions is to compile
13505 and link every module with the same @option{-G} option. However,
13506 you may wish to build a library that supports several different
13507 small data limits. You can do this by compiling the library with
13508 the highest supported @option{-G} setting and additionally using
13509 @option{-mno-extern-sdata} to stop the library from making assumptions
13510 about externally-defined data.
13516 Use (do not use) GP-relative accesses for symbols that are known to be
13517 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13518 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13521 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13522 might not hold the value of @code{_gp}. For example, if the code is
13523 part of a library that might be used in a boot monitor, programs that
13524 call boot monitor routines will pass an unknown value in @code{$gp}.
13525 (In such situations, the boot monitor itself would usually be compiled
13526 with @option{-G0}.)
13528 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13529 @option{-mno-extern-sdata}.
13531 @item -membedded-data
13532 @itemx -mno-embedded-data
13533 @opindex membedded-data
13534 @opindex mno-embedded-data
13535 Allocate variables to the read-only data section first if possible, then
13536 next in the small data section if possible, otherwise in data. This gives
13537 slightly slower code than the default, but reduces the amount of RAM required
13538 when executing, and thus may be preferred for some embedded systems.
13540 @item -muninit-const-in-rodata
13541 @itemx -mno-uninit-const-in-rodata
13542 @opindex muninit-const-in-rodata
13543 @opindex mno-uninit-const-in-rodata
13544 Put uninitialized @code{const} variables in the read-only data section.
13545 This option is only meaningful in conjunction with @option{-membedded-data}.
13547 @item -mcode-readable=@var{setting}
13548 @opindex mcode-readable
13549 Specify whether GCC may generate code that reads from executable sections.
13550 There are three possible settings:
13553 @item -mcode-readable=yes
13554 Instructions may freely access executable sections. This is the
13557 @item -mcode-readable=pcrel
13558 MIPS16 PC-relative load instructions can access executable sections,
13559 but other instructions must not do so. This option is useful on 4KSc
13560 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13561 It is also useful on processors that can be configured to have a dual
13562 instruction/data SRAM interface and that, like the M4K, automatically
13563 redirect PC-relative loads to the instruction RAM.
13565 @item -mcode-readable=no
13566 Instructions must not access executable sections. This option can be
13567 useful on targets that are configured to have a dual instruction/data
13568 SRAM interface but that (unlike the M4K) do not automatically redirect
13569 PC-relative loads to the instruction RAM.
13572 @item -msplit-addresses
13573 @itemx -mno-split-addresses
13574 @opindex msplit-addresses
13575 @opindex mno-split-addresses
13576 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13577 relocation operators. This option has been superseded by
13578 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13580 @item -mexplicit-relocs
13581 @itemx -mno-explicit-relocs
13582 @opindex mexplicit-relocs
13583 @opindex mno-explicit-relocs
13584 Use (do not use) assembler relocation operators when dealing with symbolic
13585 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13586 is to use assembler macros instead.
13588 @option{-mexplicit-relocs} is the default if GCC was configured
13589 to use an assembler that supports relocation operators.
13591 @item -mcheck-zero-division
13592 @itemx -mno-check-zero-division
13593 @opindex mcheck-zero-division
13594 @opindex mno-check-zero-division
13595 Trap (do not trap) on integer division by zero.
13597 The default is @option{-mcheck-zero-division}.
13599 @item -mdivide-traps
13600 @itemx -mdivide-breaks
13601 @opindex mdivide-traps
13602 @opindex mdivide-breaks
13603 MIPS systems check for division by zero by generating either a
13604 conditional trap or a break instruction. Using traps results in
13605 smaller code, but is only supported on MIPS II and later. Also, some
13606 versions of the Linux kernel have a bug that prevents trap from
13607 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13608 allow conditional traps on architectures that support them and
13609 @option{-mdivide-breaks} to force the use of breaks.
13611 The default is usually @option{-mdivide-traps}, but this can be
13612 overridden at configure time using @option{--with-divide=breaks}.
13613 Divide-by-zero checks can be completely disabled using
13614 @option{-mno-check-zero-division}.
13619 @opindex mno-memcpy
13620 Force (do not force) the use of @code{memcpy()} for non-trivial block
13621 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13622 most constant-sized copies.
13625 @itemx -mno-long-calls
13626 @opindex mlong-calls
13627 @opindex mno-long-calls
13628 Disable (do not disable) use of the @code{jal} instruction. Calling
13629 functions using @code{jal} is more efficient but requires the caller
13630 and callee to be in the same 256 megabyte segment.
13632 This option has no effect on abicalls code. The default is
13633 @option{-mno-long-calls}.
13639 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13640 instructions, as provided by the R4650 ISA@.
13643 @itemx -mno-fused-madd
13644 @opindex mfused-madd
13645 @opindex mno-fused-madd
13646 Enable (disable) use of the floating point multiply-accumulate
13647 instructions, when they are available. The default is
13648 @option{-mfused-madd}.
13650 When multiply-accumulate instructions are used, the intermediate
13651 product is calculated to infinite precision and is not subject to
13652 the FCSR Flush to Zero bit. This may be undesirable in some
13657 Tell the MIPS assembler to not run its preprocessor over user
13658 assembler files (with a @samp{.s} suffix) when assembling them.
13661 @itemx -mno-fix-r4000
13662 @opindex mfix-r4000
13663 @opindex mno-fix-r4000
13664 Work around certain R4000 CPU errata:
13667 A double-word or a variable shift may give an incorrect result if executed
13668 immediately after starting an integer division.
13670 A double-word or a variable shift may give an incorrect result if executed
13671 while an integer multiplication is in progress.
13673 An integer division may give an incorrect result if started in a delay slot
13674 of a taken branch or a jump.
13678 @itemx -mno-fix-r4400
13679 @opindex mfix-r4400
13680 @opindex mno-fix-r4400
13681 Work around certain R4400 CPU errata:
13684 A double-word or a variable shift may give an incorrect result if executed
13685 immediately after starting an integer division.
13689 @itemx -mno-fix-r10000
13690 @opindex mfix-r10000
13691 @opindex mno-fix-r10000
13692 Work around certain R10000 errata:
13695 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13696 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13699 This option can only be used if the target architecture supports
13700 branch-likely instructions. @option{-mfix-r10000} is the default when
13701 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13705 @itemx -mno-fix-vr4120
13706 @opindex mfix-vr4120
13707 Work around certain VR4120 errata:
13710 @code{dmultu} does not always produce the correct result.
13712 @code{div} and @code{ddiv} do not always produce the correct result if one
13713 of the operands is negative.
13715 The workarounds for the division errata rely on special functions in
13716 @file{libgcc.a}. At present, these functions are only provided by
13717 the @code{mips64vr*-elf} configurations.
13719 Other VR4120 errata require a nop to be inserted between certain pairs of
13720 instructions. These errata are handled by the assembler, not by GCC itself.
13723 @opindex mfix-vr4130
13724 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13725 workarounds are implemented by the assembler rather than by GCC,
13726 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13727 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13728 instructions are available instead.
13731 @itemx -mno-fix-sb1
13733 Work around certain SB-1 CPU core errata.
13734 (This flag currently works around the SB-1 revision 2
13735 ``F1'' and ``F2'' floating point errata.)
13737 @item -mr10k-cache-barrier=@var{setting}
13738 @opindex mr10k-cache-barrier
13739 Specify whether GCC should insert cache barriers to avoid the
13740 side-effects of speculation on R10K processors.
13742 In common with many processors, the R10K tries to predict the outcome
13743 of a conditional branch and speculatively executes instructions from
13744 the ``taken'' branch. It later aborts these instructions if the
13745 predicted outcome was wrong. However, on the R10K, even aborted
13746 instructions can have side effects.
13748 This problem only affects kernel stores and, depending on the system,
13749 kernel loads. As an example, a speculatively-executed store may load
13750 the target memory into cache and mark the cache line as dirty, even if
13751 the store itself is later aborted. If a DMA operation writes to the
13752 same area of memory before the ``dirty'' line is flushed, the cached
13753 data will overwrite the DMA-ed data. See the R10K processor manual
13754 for a full description, including other potential problems.
13756 One workaround is to insert cache barrier instructions before every memory
13757 access that might be speculatively executed and that might have side
13758 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13759 controls GCC's implementation of this workaround. It assumes that
13760 aborted accesses to any byte in the following regions will not have
13765 the memory occupied by the current function's stack frame;
13768 the memory occupied by an incoming stack argument;
13771 the memory occupied by an object with a link-time-constant address.
13774 It is the kernel's responsibility to ensure that speculative
13775 accesses to these regions are indeed safe.
13777 If the input program contains a function declaration such as:
13783 then the implementation of @code{foo} must allow @code{j foo} and
13784 @code{jal foo} to be executed speculatively. GCC honors this
13785 restriction for functions it compiles itself. It expects non-GCC
13786 functions (such as hand-written assembly code) to do the same.
13788 The option has three forms:
13791 @item -mr10k-cache-barrier=load-store
13792 Insert a cache barrier before a load or store that might be
13793 speculatively executed and that might have side effects even
13796 @item -mr10k-cache-barrier=store
13797 Insert a cache barrier before a store that might be speculatively
13798 executed and that might have side effects even if aborted.
13800 @item -mr10k-cache-barrier=none
13801 Disable the insertion of cache barriers. This is the default setting.
13804 @item -mflush-func=@var{func}
13805 @itemx -mno-flush-func
13806 @opindex mflush-func
13807 Specifies the function to call to flush the I and D caches, or to not
13808 call any such function. If called, the function must take the same
13809 arguments as the common @code{_flush_func()}, that is, the address of the
13810 memory range for which the cache is being flushed, the size of the
13811 memory range, and the number 3 (to flush both caches). The default
13812 depends on the target GCC was configured for, but commonly is either
13813 @samp{_flush_func} or @samp{__cpu_flush}.
13815 @item mbranch-cost=@var{num}
13816 @opindex mbranch-cost
13817 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13818 This cost is only a heuristic and is not guaranteed to produce
13819 consistent results across releases. A zero cost redundantly selects
13820 the default, which is based on the @option{-mtune} setting.
13822 @item -mbranch-likely
13823 @itemx -mno-branch-likely
13824 @opindex mbranch-likely
13825 @opindex mno-branch-likely
13826 Enable or disable use of Branch Likely instructions, regardless of the
13827 default for the selected architecture. By default, Branch Likely
13828 instructions may be generated if they are supported by the selected
13829 architecture. An exception is for the MIPS32 and MIPS64 architectures
13830 and processors which implement those architectures; for those, Branch
13831 Likely instructions will not be generated by default because the MIPS32
13832 and MIPS64 architectures specifically deprecate their use.
13834 @item -mfp-exceptions
13835 @itemx -mno-fp-exceptions
13836 @opindex mfp-exceptions
13837 Specifies whether FP exceptions are enabled. This affects how we schedule
13838 FP instructions for some processors. The default is that FP exceptions are
13841 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13842 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13845 @item -mvr4130-align
13846 @itemx -mno-vr4130-align
13847 @opindex mvr4130-align
13848 The VR4130 pipeline is two-way superscalar, but can only issue two
13849 instructions together if the first one is 8-byte aligned. When this
13850 option is enabled, GCC will align pairs of instructions that it
13851 thinks should execute in parallel.
13853 This option only has an effect when optimizing for the VR4130.
13854 It normally makes code faster, but at the expense of making it bigger.
13855 It is enabled by default at optimization level @option{-O3}.
13860 Enable (disable) generation of @code{synci} instructions on
13861 architectures that support it. The @code{synci} instructions (if
13862 enabled) will be generated when @code{__builtin___clear_cache()} is
13865 This option defaults to @code{-mno-synci}, but the default can be
13866 overridden by configuring with @code{--with-synci}.
13868 When compiling code for single processor systems, it is generally safe
13869 to use @code{synci}. However, on many multi-core (SMP) systems, it
13870 will not invalidate the instruction caches on all cores and may lead
13871 to undefined behavior.
13875 @subsection MMIX Options
13876 @cindex MMIX Options
13878 These options are defined for the MMIX:
13882 @itemx -mno-libfuncs
13884 @opindex mno-libfuncs
13885 Specify that intrinsic library functions are being compiled, passing all
13886 values in registers, no matter the size.
13889 @itemx -mno-epsilon
13891 @opindex mno-epsilon
13892 Generate floating-point comparison instructions that compare with respect
13893 to the @code{rE} epsilon register.
13895 @item -mabi=mmixware
13897 @opindex mabi=mmixware
13899 Generate code that passes function parameters and return values that (in
13900 the called function) are seen as registers @code{$0} and up, as opposed to
13901 the GNU ABI which uses global registers @code{$231} and up.
13903 @item -mzero-extend
13904 @itemx -mno-zero-extend
13905 @opindex mzero-extend
13906 @opindex mno-zero-extend
13907 When reading data from memory in sizes shorter than 64 bits, use (do not
13908 use) zero-extending load instructions by default, rather than
13909 sign-extending ones.
13912 @itemx -mno-knuthdiv
13914 @opindex mno-knuthdiv
13915 Make the result of a division yielding a remainder have the same sign as
13916 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13917 remainder follows the sign of the dividend. Both methods are
13918 arithmetically valid, the latter being almost exclusively used.
13920 @item -mtoplevel-symbols
13921 @itemx -mno-toplevel-symbols
13922 @opindex mtoplevel-symbols
13923 @opindex mno-toplevel-symbols
13924 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13925 code can be used with the @code{PREFIX} assembly directive.
13929 Generate an executable in the ELF format, rather than the default
13930 @samp{mmo} format used by the @command{mmix} simulator.
13932 @item -mbranch-predict
13933 @itemx -mno-branch-predict
13934 @opindex mbranch-predict
13935 @opindex mno-branch-predict
13936 Use (do not use) the probable-branch instructions, when static branch
13937 prediction indicates a probable branch.
13939 @item -mbase-addresses
13940 @itemx -mno-base-addresses
13941 @opindex mbase-addresses
13942 @opindex mno-base-addresses
13943 Generate (do not generate) code that uses @emph{base addresses}. Using a
13944 base address automatically generates a request (handled by the assembler
13945 and the linker) for a constant to be set up in a global register. The
13946 register is used for one or more base address requests within the range 0
13947 to 255 from the value held in the register. The generally leads to short
13948 and fast code, but the number of different data items that can be
13949 addressed is limited. This means that a program that uses lots of static
13950 data may require @option{-mno-base-addresses}.
13952 @item -msingle-exit
13953 @itemx -mno-single-exit
13954 @opindex msingle-exit
13955 @opindex mno-single-exit
13956 Force (do not force) generated code to have a single exit point in each
13960 @node MN10300 Options
13961 @subsection MN10300 Options
13962 @cindex MN10300 options
13964 These @option{-m} options are defined for Matsushita MN10300 architectures:
13969 Generate code to avoid bugs in the multiply instructions for the MN10300
13970 processors. This is the default.
13972 @item -mno-mult-bug
13973 @opindex mno-mult-bug
13974 Do not generate code to avoid bugs in the multiply instructions for the
13975 MN10300 processors.
13979 Generate code which uses features specific to the AM33 processor.
13983 Do not generate code which uses features specific to the AM33 processor. This
13986 @item -mreturn-pointer-on-d0
13987 @opindex mreturn-pointer-on-d0
13988 When generating a function which returns a pointer, return the pointer
13989 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13990 only in a0, and attempts to call such functions without a prototype
13991 would result in errors. Note that this option is on by default; use
13992 @option{-mno-return-pointer-on-d0} to disable it.
13996 Do not link in the C run-time initialization object file.
14000 Indicate to the linker that it should perform a relaxation optimization pass
14001 to shorten branches, calls and absolute memory addresses. This option only
14002 has an effect when used on the command line for the final link step.
14004 This option makes symbolic debugging impossible.
14007 @node PDP-11 Options
14008 @subsection PDP-11 Options
14009 @cindex PDP-11 Options
14011 These options are defined for the PDP-11:
14016 Use hardware FPP floating point. This is the default. (FIS floating
14017 point on the PDP-11/40 is not supported.)
14020 @opindex msoft-float
14021 Do not use hardware floating point.
14025 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14029 Return floating-point results in memory. This is the default.
14033 Generate code for a PDP-11/40.
14037 Generate code for a PDP-11/45. This is the default.
14041 Generate code for a PDP-11/10.
14043 @item -mbcopy-builtin
14044 @opindex mbcopy-builtin
14045 Use inline @code{movmemhi} patterns for copying memory. This is the
14050 Do not use inline @code{movmemhi} patterns for copying memory.
14056 Use 16-bit @code{int}. This is the default.
14062 Use 32-bit @code{int}.
14065 @itemx -mno-float32
14067 @opindex mno-float32
14068 Use 64-bit @code{float}. This is the default.
14071 @itemx -mno-float64
14073 @opindex mno-float64
14074 Use 32-bit @code{float}.
14078 Use @code{abshi2} pattern. This is the default.
14082 Do not use @code{abshi2} pattern.
14084 @item -mbranch-expensive
14085 @opindex mbranch-expensive
14086 Pretend that branches are expensive. This is for experimenting with
14087 code generation only.
14089 @item -mbranch-cheap
14090 @opindex mbranch-cheap
14091 Do not pretend that branches are expensive. This is the default.
14095 Generate code for a system with split I&D@.
14099 Generate code for a system without split I&D@. This is the default.
14103 Use Unix assembler syntax. This is the default when configured for
14104 @samp{pdp11-*-bsd}.
14108 Use DEC assembler syntax. This is the default when configured for any
14109 PDP-11 target other than @samp{pdp11-*-bsd}.
14112 @node picoChip Options
14113 @subsection picoChip Options
14114 @cindex picoChip options
14116 These @samp{-m} options are defined for picoChip implementations:
14120 @item -mae=@var{ae_type}
14122 Set the instruction set, register set, and instruction scheduling
14123 parameters for array element type @var{ae_type}. Supported values
14124 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14126 @option{-mae=ANY} selects a completely generic AE type. Code
14127 generated with this option will run on any of the other AE types. The
14128 code will not be as efficient as it would be if compiled for a specific
14129 AE type, and some types of operation (e.g., multiplication) will not
14130 work properly on all types of AE.
14132 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14133 for compiled code, and is the default.
14135 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14136 option may suffer from poor performance of byte (char) manipulation,
14137 since the DSP AE does not provide hardware support for byte load/stores.
14139 @item -msymbol-as-address
14140 Enable the compiler to directly use a symbol name as an address in a
14141 load/store instruction, without first loading it into a
14142 register. Typically, the use of this option will generate larger
14143 programs, which run faster than when the option isn't used. However, the
14144 results vary from program to program, so it is left as a user option,
14145 rather than being permanently enabled.
14147 @item -mno-inefficient-warnings
14148 Disables warnings about the generation of inefficient code. These
14149 warnings can be generated, for example, when compiling code which
14150 performs byte-level memory operations on the MAC AE type. The MAC AE has
14151 no hardware support for byte-level memory operations, so all byte
14152 load/stores must be synthesized from word load/store operations. This is
14153 inefficient and a warning will be generated indicating to the programmer
14154 that they should rewrite the code to avoid byte operations, or to target
14155 an AE type which has the necessary hardware support. This option enables
14156 the warning to be turned off.
14160 @node PowerPC Options
14161 @subsection PowerPC Options
14162 @cindex PowerPC options
14164 These are listed under @xref{RS/6000 and PowerPC Options}.
14166 @node RS/6000 and PowerPC Options
14167 @subsection IBM RS/6000 and PowerPC Options
14168 @cindex RS/6000 and PowerPC Options
14169 @cindex IBM RS/6000 and PowerPC Options
14171 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14178 @itemx -mno-powerpc
14179 @itemx -mpowerpc-gpopt
14180 @itemx -mno-powerpc-gpopt
14181 @itemx -mpowerpc-gfxopt
14182 @itemx -mno-powerpc-gfxopt
14184 @itemx -mno-powerpc64
14188 @itemx -mno-popcntb
14190 @itemx -mno-popcntd
14198 @itemx -mno-hard-dfp
14202 @opindex mno-power2
14204 @opindex mno-powerpc
14205 @opindex mpowerpc-gpopt
14206 @opindex mno-powerpc-gpopt
14207 @opindex mpowerpc-gfxopt
14208 @opindex mno-powerpc-gfxopt
14209 @opindex mpowerpc64
14210 @opindex mno-powerpc64
14214 @opindex mno-popcntb
14216 @opindex mno-popcntd
14222 @opindex mno-mfpgpr
14224 @opindex mno-hard-dfp
14225 GCC supports two related instruction set architectures for the
14226 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14227 instructions supported by the @samp{rios} chip set used in the original
14228 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14229 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14230 the IBM 4xx, 6xx, and follow-on microprocessors.
14232 Neither architecture is a subset of the other. However there is a
14233 large common subset of instructions supported by both. An MQ
14234 register is included in processors supporting the POWER architecture.
14236 You use these options to specify which instructions are available on the
14237 processor you are using. The default value of these options is
14238 determined when configuring GCC@. Specifying the
14239 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14240 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14241 rather than the options listed above.
14243 The @option{-mpower} option allows GCC to generate instructions that
14244 are found only in the POWER architecture and to use the MQ register.
14245 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14246 to generate instructions that are present in the POWER2 architecture but
14247 not the original POWER architecture.
14249 The @option{-mpowerpc} option allows GCC to generate instructions that
14250 are found only in the 32-bit subset of the PowerPC architecture.
14251 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14252 GCC to use the optional PowerPC architecture instructions in the
14253 General Purpose group, including floating-point square root. Specifying
14254 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14255 use the optional PowerPC architecture instructions in the Graphics
14256 group, including floating-point select.
14258 The @option{-mmfcrf} option allows GCC to generate the move from
14259 condition register field instruction implemented on the POWER4
14260 processor and other processors that support the PowerPC V2.01
14262 The @option{-mpopcntb} option allows GCC to generate the popcount and
14263 double precision FP reciprocal estimate instruction implemented on the
14264 POWER5 processor and other processors that support the PowerPC V2.02
14266 The @option{-mpopcntd} option allows GCC to generate the popcount
14267 instruction implemented on the POWER7 processor and other processors
14268 that support the PowerPC V2.06 architecture.
14269 The @option{-mfprnd} option allows GCC to generate the FP round to
14270 integer instructions implemented on the POWER5+ processor and other
14271 processors that support the PowerPC V2.03 architecture.
14272 The @option{-mcmpb} option allows GCC to generate the compare bytes
14273 instruction implemented on the POWER6 processor and other processors
14274 that support the PowerPC V2.05 architecture.
14275 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14276 general purpose register instructions implemented on the POWER6X
14277 processor and other processors that support the extended PowerPC V2.05
14279 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14280 point instructions implemented on some POWER processors.
14282 The @option{-mpowerpc64} option allows GCC to generate the additional
14283 64-bit instructions that are found in the full PowerPC64 architecture
14284 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14285 @option{-mno-powerpc64}.
14287 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14288 will use only the instructions in the common subset of both
14289 architectures plus some special AIX common-mode calls, and will not use
14290 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14291 permits GCC to use any instruction from either architecture and to
14292 allow use of the MQ register; specify this for the Motorola MPC601.
14294 @item -mnew-mnemonics
14295 @itemx -mold-mnemonics
14296 @opindex mnew-mnemonics
14297 @opindex mold-mnemonics
14298 Select which mnemonics to use in the generated assembler code. With
14299 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14300 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14301 assembler mnemonics defined for the POWER architecture. Instructions
14302 defined in only one architecture have only one mnemonic; GCC uses that
14303 mnemonic irrespective of which of these options is specified.
14305 GCC defaults to the mnemonics appropriate for the architecture in
14306 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14307 value of these option. Unless you are building a cross-compiler, you
14308 should normally not specify either @option{-mnew-mnemonics} or
14309 @option{-mold-mnemonics}, but should instead accept the default.
14311 @item -mcpu=@var{cpu_type}
14313 Set architecture type, register usage, choice of mnemonics, and
14314 instruction scheduling parameters for machine type @var{cpu_type}.
14315 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14316 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14317 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
14318 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
14319 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14320 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
14321 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14322 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14323 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
14324 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14325 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14327 @option{-mcpu=common} selects a completely generic processor. Code
14328 generated under this option will run on any POWER or PowerPC processor.
14329 GCC will use only the instructions in the common subset of both
14330 architectures, and will not use the MQ register. GCC assumes a generic
14331 processor model for scheduling purposes.
14333 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14334 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14335 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14336 types, with an appropriate, generic processor model assumed for
14337 scheduling purposes.
14339 The other options specify a specific processor. Code generated under
14340 those options will run best on that processor, and may not run at all on
14343 The @option{-mcpu} options automatically enable or disable the
14346 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14347 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14348 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14349 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14351 The particular options set for any particular CPU will vary between
14352 compiler versions, depending on what setting seems to produce optimal
14353 code for that CPU; it doesn't necessarily reflect the actual hardware's
14354 capabilities. If you wish to set an individual option to a particular
14355 value, you may specify it after the @option{-mcpu} option, like
14356 @samp{-mcpu=970 -mno-altivec}.
14358 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14359 not enabled or disabled by the @option{-mcpu} option at present because
14360 AIX does not have full support for these options. You may still
14361 enable or disable them individually if you're sure it'll work in your
14364 @item -mtune=@var{cpu_type}
14366 Set the instruction scheduling parameters for machine type
14367 @var{cpu_type}, but do not set the architecture type, register usage, or
14368 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14369 values for @var{cpu_type} are used for @option{-mtune} as for
14370 @option{-mcpu}. If both are specified, the code generated will use the
14371 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14372 scheduling parameters set by @option{-mtune}.
14378 Generate code to compute division as reciprocal estimate and iterative
14379 refinement, creating opportunities for increased throughput. This
14380 feature requires: optional PowerPC Graphics instruction set for single
14381 precision and FRE instruction for double precision, assuming divides
14382 cannot generate user-visible traps, and the domain values not include
14383 Infinities, denormals or zero denominator.
14386 @itemx -mno-altivec
14388 @opindex mno-altivec
14389 Generate code that uses (does not use) AltiVec instructions, and also
14390 enable the use of built-in functions that allow more direct access to
14391 the AltiVec instruction set. You may also need to set
14392 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14398 @opindex mno-vrsave
14399 Generate VRSAVE instructions when generating AltiVec code.
14401 @item -mgen-cell-microcode
14402 @opindex mgen-cell-microcode
14403 Generate Cell microcode instructions
14405 @item -mwarn-cell-microcode
14406 @opindex mwarn-cell-microcode
14407 Warning when a Cell microcode instruction is going to emitted. An example
14408 of a Cell microcode instruction is a variable shift.
14411 @opindex msecure-plt
14412 Generate code that allows ld and ld.so to build executables and shared
14413 libraries with non-exec .plt and .got sections. This is a PowerPC
14414 32-bit SYSV ABI option.
14418 Generate code that uses a BSS .plt section that ld.so fills in, and
14419 requires .plt and .got sections that are both writable and executable.
14420 This is a PowerPC 32-bit SYSV ABI option.
14426 This switch enables or disables the generation of ISEL instructions.
14428 @item -misel=@var{yes/no}
14429 This switch has been deprecated. Use @option{-misel} and
14430 @option{-mno-isel} instead.
14436 This switch enables or disables the generation of SPE simd
14442 @opindex mno-paired
14443 This switch enables or disables the generation of PAIRED simd
14446 @item -mspe=@var{yes/no}
14447 This option has been deprecated. Use @option{-mspe} and
14448 @option{-mno-spe} instead.
14454 Generate code that uses (does not use) vector/scalar (VSX)
14455 instructions, and also enable the use of built-in functions that allow
14456 more direct access to the VSX instruction set.
14458 @item -mfloat-gprs=@var{yes/single/double/no}
14459 @itemx -mfloat-gprs
14460 @opindex mfloat-gprs
14461 This switch enables or disables the generation of floating point
14462 operations on the general purpose registers for architectures that
14465 The argument @var{yes} or @var{single} enables the use of
14466 single-precision floating point operations.
14468 The argument @var{double} enables the use of single and
14469 double-precision floating point operations.
14471 The argument @var{no} disables floating point operations on the
14472 general purpose registers.
14474 This option is currently only available on the MPC854x.
14480 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14481 targets (including GNU/Linux). The 32-bit environment sets int, long
14482 and pointer to 32 bits and generates code that runs on any PowerPC
14483 variant. The 64-bit environment sets int to 32 bits and long and
14484 pointer to 64 bits, and generates code for PowerPC64, as for
14485 @option{-mpowerpc64}.
14488 @itemx -mno-fp-in-toc
14489 @itemx -mno-sum-in-toc
14490 @itemx -mminimal-toc
14492 @opindex mno-fp-in-toc
14493 @opindex mno-sum-in-toc
14494 @opindex mminimal-toc
14495 Modify generation of the TOC (Table Of Contents), which is created for
14496 every executable file. The @option{-mfull-toc} option is selected by
14497 default. In that case, GCC will allocate at least one TOC entry for
14498 each unique non-automatic variable reference in your program. GCC
14499 will also place floating-point constants in the TOC@. However, only
14500 16,384 entries are available in the TOC@.
14502 If you receive a linker error message that saying you have overflowed
14503 the available TOC space, you can reduce the amount of TOC space used
14504 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14505 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14506 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14507 generate code to calculate the sum of an address and a constant at
14508 run-time instead of putting that sum into the TOC@. You may specify one
14509 or both of these options. Each causes GCC to produce very slightly
14510 slower and larger code at the expense of conserving TOC space.
14512 If you still run out of space in the TOC even when you specify both of
14513 these options, specify @option{-mminimal-toc} instead. This option causes
14514 GCC to make only one TOC entry for every file. When you specify this
14515 option, GCC will produce code that is slower and larger but which
14516 uses extremely little TOC space. You may wish to use this option
14517 only on files that contain less frequently executed code.
14523 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14524 @code{long} type, and the infrastructure needed to support them.
14525 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14526 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14527 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14530 @itemx -mno-xl-compat
14531 @opindex mxl-compat
14532 @opindex mno-xl-compat
14533 Produce code that conforms more closely to IBM XL compiler semantics
14534 when using AIX-compatible ABI@. Pass floating-point arguments to
14535 prototyped functions beyond the register save area (RSA) on the stack
14536 in addition to argument FPRs. Do not assume that most significant
14537 double in 128-bit long double value is properly rounded when comparing
14538 values and converting to double. Use XL symbol names for long double
14541 The AIX calling convention was extended but not initially documented to
14542 handle an obscure K&R C case of calling a function that takes the
14543 address of its arguments with fewer arguments than declared. IBM XL
14544 compilers access floating point arguments which do not fit in the
14545 RSA from the stack when a subroutine is compiled without
14546 optimization. Because always storing floating-point arguments on the
14547 stack is inefficient and rarely needed, this option is not enabled by
14548 default and only is necessary when calling subroutines compiled by IBM
14549 XL compilers without optimization.
14553 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14554 application written to use message passing with special startup code to
14555 enable the application to run. The system must have PE installed in the
14556 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14557 must be overridden with the @option{-specs=} option to specify the
14558 appropriate directory location. The Parallel Environment does not
14559 support threads, so the @option{-mpe} option and the @option{-pthread}
14560 option are incompatible.
14562 @item -malign-natural
14563 @itemx -malign-power
14564 @opindex malign-natural
14565 @opindex malign-power
14566 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14567 @option{-malign-natural} overrides the ABI-defined alignment of larger
14568 types, such as floating-point doubles, on their natural size-based boundary.
14569 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14570 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14572 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14576 @itemx -mhard-float
14577 @opindex msoft-float
14578 @opindex mhard-float
14579 Generate code that does not use (uses) the floating-point register set.
14580 Software floating point emulation is provided if you use the
14581 @option{-msoft-float} option, and pass the option to GCC when linking.
14583 @item -msingle-float
14584 @itemx -mdouble-float
14585 @opindex msingle-float
14586 @opindex mdouble-float
14587 Generate code for single or double-precision floating point operations.
14588 @option{-mdouble-float} implies @option{-msingle-float}.
14591 @opindex msimple-fpu
14592 Do not generate sqrt and div instructions for hardware floating point unit.
14596 Specify type of floating point unit. Valid values are @var{sp_lite}
14597 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14598 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14599 and @var{dp_full} (equivalent to -mdouble-float).
14602 @opindex mxilinx-fpu
14603 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14606 @itemx -mno-multiple
14608 @opindex mno-multiple
14609 Generate code that uses (does not use) the load multiple word
14610 instructions and the store multiple word instructions. These
14611 instructions are generated by default on POWER systems, and not
14612 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14613 endian PowerPC systems, since those instructions do not work when the
14614 processor is in little endian mode. The exceptions are PPC740 and
14615 PPC750 which permit the instructions usage in little endian mode.
14620 @opindex mno-string
14621 Generate code that uses (does not use) the load string instructions
14622 and the store string word instructions to save multiple registers and
14623 do small block moves. These instructions are generated by default on
14624 POWER systems, and not generated on PowerPC systems. Do not use
14625 @option{-mstring} on little endian PowerPC systems, since those
14626 instructions do not work when the processor is in little endian mode.
14627 The exceptions are PPC740 and PPC750 which permit the instructions
14628 usage in little endian mode.
14633 @opindex mno-update
14634 Generate code that uses (does not use) the load or store instructions
14635 that update the base register to the address of the calculated memory
14636 location. These instructions are generated by default. If you use
14637 @option{-mno-update}, there is a small window between the time that the
14638 stack pointer is updated and the address of the previous frame is
14639 stored, which means code that walks the stack frame across interrupts or
14640 signals may get corrupted data.
14642 @item -mavoid-indexed-addresses
14643 @item -mno-avoid-indexed-addresses
14644 @opindex mavoid-indexed-addresses
14645 @opindex mno-avoid-indexed-addresses
14646 Generate code that tries to avoid (not avoid) the use of indexed load
14647 or store instructions. These instructions can incur a performance
14648 penalty on Power6 processors in certain situations, such as when
14649 stepping through large arrays that cross a 16M boundary. This option
14650 is enabled by default when targetting Power6 and disabled otherwise.
14653 @itemx -mno-fused-madd
14654 @opindex mfused-madd
14655 @opindex mno-fused-madd
14656 Generate code that uses (does not use) the floating point multiply and
14657 accumulate instructions. These instructions are generated by default if
14658 hardware floating is used.
14664 Generate code that uses (does not use) the half-word multiply and
14665 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14666 These instructions are generated by default when targetting those
14673 Generate code that uses (does not use) the string-search @samp{dlmzb}
14674 instruction on the IBM 405, 440 and 464 processors. This instruction is
14675 generated by default when targetting those processors.
14677 @item -mno-bit-align
14679 @opindex mno-bit-align
14680 @opindex mbit-align
14681 On System V.4 and embedded PowerPC systems do not (do) force structures
14682 and unions that contain bit-fields to be aligned to the base type of the
14685 For example, by default a structure containing nothing but 8
14686 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14687 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14688 the structure would be aligned to a 1 byte boundary and be one byte in
14691 @item -mno-strict-align
14692 @itemx -mstrict-align
14693 @opindex mno-strict-align
14694 @opindex mstrict-align
14695 On System V.4 and embedded PowerPC systems do not (do) assume that
14696 unaligned memory references will be handled by the system.
14698 @item -mrelocatable
14699 @itemx -mno-relocatable
14700 @opindex mrelocatable
14701 @opindex mno-relocatable
14702 On embedded PowerPC systems generate code that allows (does not allow)
14703 the program to be relocated to a different address at runtime. If you
14704 use @option{-mrelocatable} on any module, all objects linked together must
14705 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14707 @item -mrelocatable-lib
14708 @itemx -mno-relocatable-lib
14709 @opindex mrelocatable-lib
14710 @opindex mno-relocatable-lib
14711 On embedded PowerPC systems generate code that allows (does not allow)
14712 the program to be relocated to a different address at runtime. Modules
14713 compiled with @option{-mrelocatable-lib} can be linked with either modules
14714 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14715 with modules compiled with the @option{-mrelocatable} options.
14721 On System V.4 and embedded PowerPC systems do not (do) assume that
14722 register 2 contains a pointer to a global area pointing to the addresses
14723 used in the program.
14726 @itemx -mlittle-endian
14728 @opindex mlittle-endian
14729 On System V.4 and embedded PowerPC systems compile code for the
14730 processor in little endian mode. The @option{-mlittle-endian} option is
14731 the same as @option{-mlittle}.
14734 @itemx -mbig-endian
14736 @opindex mbig-endian
14737 On System V.4 and embedded PowerPC systems compile code for the
14738 processor in big endian mode. The @option{-mbig-endian} option is
14739 the same as @option{-mbig}.
14741 @item -mdynamic-no-pic
14742 @opindex mdynamic-no-pic
14743 On Darwin and Mac OS X systems, compile code so that it is not
14744 relocatable, but that its external references are relocatable. The
14745 resulting code is suitable for applications, but not shared
14748 @item -mprioritize-restricted-insns=@var{priority}
14749 @opindex mprioritize-restricted-insns
14750 This option controls the priority that is assigned to
14751 dispatch-slot restricted instructions during the second scheduling
14752 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14753 @var{no/highest/second-highest} priority to dispatch slot restricted
14756 @item -msched-costly-dep=@var{dependence_type}
14757 @opindex msched-costly-dep
14758 This option controls which dependences are considered costly
14759 by the target during instruction scheduling. The argument
14760 @var{dependence_type} takes one of the following values:
14761 @var{no}: no dependence is costly,
14762 @var{all}: all dependences are costly,
14763 @var{true_store_to_load}: a true dependence from store to load is costly,
14764 @var{store_to_load}: any dependence from store to load is costly,
14765 @var{number}: any dependence which latency >= @var{number} is costly.
14767 @item -minsert-sched-nops=@var{scheme}
14768 @opindex minsert-sched-nops
14769 This option controls which nop insertion scheme will be used during
14770 the second scheduling pass. The argument @var{scheme} takes one of the
14772 @var{no}: Don't insert nops.
14773 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14774 according to the scheduler's grouping.
14775 @var{regroup_exact}: Insert nops to force costly dependent insns into
14776 separate groups. Insert exactly as many nops as needed to force an insn
14777 to a new group, according to the estimated processor grouping.
14778 @var{number}: Insert nops to force costly dependent insns into
14779 separate groups. Insert @var{number} nops to force an insn to a new group.
14782 @opindex mcall-sysv
14783 On System V.4 and embedded PowerPC systems compile code using calling
14784 conventions that adheres to the March 1995 draft of the System V
14785 Application Binary Interface, PowerPC processor supplement. This is the
14786 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14788 @item -mcall-sysv-eabi
14790 @opindex mcall-sysv-eabi
14791 @opindex mcall-eabi
14792 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14794 @item -mcall-sysv-noeabi
14795 @opindex mcall-sysv-noeabi
14796 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14798 @item -mcall-aixdesc
14800 On System V.4 and embedded PowerPC systems compile code for the AIX
14804 @opindex mcall-linux
14805 On System V.4 and embedded PowerPC systems compile code for the
14806 Linux-based GNU system.
14810 On System V.4 and embedded PowerPC systems compile code for the
14811 Hurd-based GNU system.
14813 @item -mcall-freebsd
14814 @opindex mcall-freebsd
14815 On System V.4 and embedded PowerPC systems compile code for the
14816 FreeBSD operating system.
14818 @item -mcall-netbsd
14819 @opindex mcall-netbsd
14820 On System V.4 and embedded PowerPC systems compile code for the
14821 NetBSD operating system.
14823 @item -mcall-openbsd
14824 @opindex mcall-netbsd
14825 On System V.4 and embedded PowerPC systems compile code for the
14826 OpenBSD operating system.
14828 @item -maix-struct-return
14829 @opindex maix-struct-return
14830 Return all structures in memory (as specified by the AIX ABI)@.
14832 @item -msvr4-struct-return
14833 @opindex msvr4-struct-return
14834 Return structures smaller than 8 bytes in registers (as specified by the
14837 @item -mabi=@var{abi-type}
14839 Extend the current ABI with a particular extension, or remove such extension.
14840 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14841 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14845 Extend the current ABI with SPE ABI extensions. This does not change
14846 the default ABI, instead it adds the SPE ABI extensions to the current
14850 @opindex mabi=no-spe
14851 Disable Booke SPE ABI extensions for the current ABI@.
14853 @item -mabi=ibmlongdouble
14854 @opindex mabi=ibmlongdouble
14855 Change the current ABI to use IBM extended precision long double.
14856 This is a PowerPC 32-bit SYSV ABI option.
14858 @item -mabi=ieeelongdouble
14859 @opindex mabi=ieeelongdouble
14860 Change the current ABI to use IEEE extended precision long double.
14861 This is a PowerPC 32-bit Linux ABI option.
14864 @itemx -mno-prototype
14865 @opindex mprototype
14866 @opindex mno-prototype
14867 On System V.4 and embedded PowerPC systems assume that all calls to
14868 variable argument functions are properly prototyped. Otherwise, the
14869 compiler must insert an instruction before every non prototyped call to
14870 set or clear bit 6 of the condition code register (@var{CR}) to
14871 indicate whether floating point values were passed in the floating point
14872 registers in case the function takes a variable arguments. With
14873 @option{-mprototype}, only calls to prototyped variable argument functions
14874 will set or clear the bit.
14878 On embedded PowerPC systems, assume that the startup module is called
14879 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14880 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14885 On embedded PowerPC systems, assume that the startup module is called
14886 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14891 On embedded PowerPC systems, assume that the startup module is called
14892 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14895 @item -myellowknife
14896 @opindex myellowknife
14897 On embedded PowerPC systems, assume that the startup module is called
14898 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14903 On System V.4 and embedded PowerPC systems, specify that you are
14904 compiling for a VxWorks system.
14908 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14909 header to indicate that @samp{eabi} extended relocations are used.
14915 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14916 Embedded Applications Binary Interface (eabi) which is a set of
14917 modifications to the System V.4 specifications. Selecting @option{-meabi}
14918 means that the stack is aligned to an 8 byte boundary, a function
14919 @code{__eabi} is called to from @code{main} to set up the eabi
14920 environment, and the @option{-msdata} option can use both @code{r2} and
14921 @code{r13} to point to two separate small data areas. Selecting
14922 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14923 do not call an initialization function from @code{main}, and the
14924 @option{-msdata} option will only use @code{r13} to point to a single
14925 small data area. The @option{-meabi} option is on by default if you
14926 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14929 @opindex msdata=eabi
14930 On System V.4 and embedded PowerPC systems, put small initialized
14931 @code{const} global and static data in the @samp{.sdata2} section, which
14932 is pointed to by register @code{r2}. Put small initialized
14933 non-@code{const} global and static data in the @samp{.sdata} section,
14934 which is pointed to by register @code{r13}. Put small uninitialized
14935 global and static data in the @samp{.sbss} section, which is adjacent to
14936 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14937 incompatible with the @option{-mrelocatable} option. The
14938 @option{-msdata=eabi} option also sets the @option{-memb} option.
14941 @opindex msdata=sysv
14942 On System V.4 and embedded PowerPC systems, put small global and static
14943 data in the @samp{.sdata} section, which is pointed to by register
14944 @code{r13}. Put small uninitialized global and static data in the
14945 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14946 The @option{-msdata=sysv} option is incompatible with the
14947 @option{-mrelocatable} option.
14949 @item -msdata=default
14951 @opindex msdata=default
14953 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14954 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14955 same as @option{-msdata=sysv}.
14958 @opindex msdata=data
14959 On System V.4 and embedded PowerPC systems, put small global
14960 data in the @samp{.sdata} section. Put small uninitialized global
14961 data in the @samp{.sbss} section. Do not use register @code{r13}
14962 to address small data however. This is the default behavior unless
14963 other @option{-msdata} options are used.
14967 @opindex msdata=none
14969 On embedded PowerPC systems, put all initialized global and static data
14970 in the @samp{.data} section, and all uninitialized data in the
14971 @samp{.bss} section.
14975 @cindex smaller data references (PowerPC)
14976 @cindex .sdata/.sdata2 references (PowerPC)
14977 On embedded PowerPC systems, put global and static items less than or
14978 equal to @var{num} bytes into the small data or bss sections instead of
14979 the normal data or bss section. By default, @var{num} is 8. The
14980 @option{-G @var{num}} switch is also passed to the linker.
14981 All modules should be compiled with the same @option{-G @var{num}} value.
14984 @itemx -mno-regnames
14986 @opindex mno-regnames
14987 On System V.4 and embedded PowerPC systems do (do not) emit register
14988 names in the assembly language output using symbolic forms.
14991 @itemx -mno-longcall
14993 @opindex mno-longcall
14994 By default assume that all calls are far away so that a longer more
14995 expensive calling sequence is required. This is required for calls
14996 further than 32 megabytes (33,554,432 bytes) from the current location.
14997 A short call will be generated if the compiler knows
14998 the call cannot be that far away. This setting can be overridden by
14999 the @code{shortcall} function attribute, or by @code{#pragma
15002 Some linkers are capable of detecting out-of-range calls and generating
15003 glue code on the fly. On these systems, long calls are unnecessary and
15004 generate slower code. As of this writing, the AIX linker can do this,
15005 as can the GNU linker for PowerPC/64. It is planned to add this feature
15006 to the GNU linker for 32-bit PowerPC systems as well.
15008 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15009 callee, L42'', plus a ``branch island'' (glue code). The two target
15010 addresses represent the callee and the ``branch island''. The
15011 Darwin/PPC linker will prefer the first address and generate a ``bl
15012 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15013 otherwise, the linker will generate ``bl L42'' to call the ``branch
15014 island''. The ``branch island'' is appended to the body of the
15015 calling function; it computes the full 32-bit address of the callee
15018 On Mach-O (Darwin) systems, this option directs the compiler emit to
15019 the glue for every direct call, and the Darwin linker decides whether
15020 to use or discard it.
15022 In the future, we may cause GCC to ignore all longcall specifications
15023 when the linker is known to generate glue.
15025 @item -mtls-markers
15026 @itemx -mno-tls-markers
15027 @opindex mtls-markers
15028 @opindex mno-tls-markers
15029 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15030 specifying the function argument. The relocation allows ld to
15031 reliably associate function call with argument setup instructions for
15032 TLS optimization, which in turn allows gcc to better schedule the
15037 Adds support for multithreading with the @dfn{pthreads} library.
15038 This option sets flags for both the preprocessor and linker.
15042 @node S/390 and zSeries Options
15043 @subsection S/390 and zSeries Options
15044 @cindex S/390 and zSeries Options
15046 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15050 @itemx -msoft-float
15051 @opindex mhard-float
15052 @opindex msoft-float
15053 Use (do not use) the hardware floating-point instructions and registers
15054 for floating-point operations. When @option{-msoft-float} is specified,
15055 functions in @file{libgcc.a} will be used to perform floating-point
15056 operations. When @option{-mhard-float} is specified, the compiler
15057 generates IEEE floating-point instructions. This is the default.
15060 @itemx -mno-hard-dfp
15062 @opindex mno-hard-dfp
15063 Use (do not use) the hardware decimal-floating-point instructions for
15064 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15065 specified, functions in @file{libgcc.a} will be used to perform
15066 decimal-floating-point operations. When @option{-mhard-dfp} is
15067 specified, the compiler generates decimal-floating-point hardware
15068 instructions. This is the default for @option{-march=z9-ec} or higher.
15070 @item -mlong-double-64
15071 @itemx -mlong-double-128
15072 @opindex mlong-double-64
15073 @opindex mlong-double-128
15074 These switches control the size of @code{long double} type. A size
15075 of 64bit makes the @code{long double} type equivalent to the @code{double}
15076 type. This is the default.
15079 @itemx -mno-backchain
15080 @opindex mbackchain
15081 @opindex mno-backchain
15082 Store (do not store) the address of the caller's frame as backchain pointer
15083 into the callee's stack frame.
15084 A backchain may be needed to allow debugging using tools that do not understand
15085 DWARF-2 call frame information.
15086 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15087 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15088 the backchain is placed into the topmost word of the 96/160 byte register
15091 In general, code compiled with @option{-mbackchain} is call-compatible with
15092 code compiled with @option{-mmo-backchain}; however, use of the backchain
15093 for debugging purposes usually requires that the whole binary is built with
15094 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15095 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15096 to build a linux kernel use @option{-msoft-float}.
15098 The default is to not maintain the backchain.
15100 @item -mpacked-stack
15101 @itemx -mno-packed-stack
15102 @opindex mpacked-stack
15103 @opindex mno-packed-stack
15104 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15105 specified, the compiler uses the all fields of the 96/160 byte register save
15106 area only for their default purpose; unused fields still take up stack space.
15107 When @option{-mpacked-stack} is specified, register save slots are densely
15108 packed at the top of the register save area; unused space is reused for other
15109 purposes, allowing for more efficient use of the available stack space.
15110 However, when @option{-mbackchain} is also in effect, the topmost word of
15111 the save area is always used to store the backchain, and the return address
15112 register is always saved two words below the backchain.
15114 As long as the stack frame backchain is not used, code generated with
15115 @option{-mpacked-stack} is call-compatible with code generated with
15116 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15117 S/390 or zSeries generated code that uses the stack frame backchain at run
15118 time, not just for debugging purposes. Such code is not call-compatible
15119 with code compiled with @option{-mpacked-stack}. Also, note that the
15120 combination of @option{-mbackchain},
15121 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15122 to build a linux kernel use @option{-msoft-float}.
15124 The default is to not use the packed stack layout.
15127 @itemx -mno-small-exec
15128 @opindex msmall-exec
15129 @opindex mno-small-exec
15130 Generate (or do not generate) code using the @code{bras} instruction
15131 to do subroutine calls.
15132 This only works reliably if the total executable size does not
15133 exceed 64k. The default is to use the @code{basr} instruction instead,
15134 which does not have this limitation.
15140 When @option{-m31} is specified, generate code compliant to the
15141 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15142 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15143 particular to generate 64-bit instructions. For the @samp{s390}
15144 targets, the default is @option{-m31}, while the @samp{s390x}
15145 targets default to @option{-m64}.
15151 When @option{-mzarch} is specified, generate code using the
15152 instructions available on z/Architecture.
15153 When @option{-mesa} is specified, generate code using the
15154 instructions available on ESA/390. Note that @option{-mesa} is
15155 not possible with @option{-m64}.
15156 When generating code compliant to the GNU/Linux for S/390 ABI,
15157 the default is @option{-mesa}. When generating code compliant
15158 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15164 Generate (or do not generate) code using the @code{mvcle} instruction
15165 to perform block moves. When @option{-mno-mvcle} is specified,
15166 use a @code{mvc} loop instead. This is the default unless optimizing for
15173 Print (or do not print) additional debug information when compiling.
15174 The default is to not print debug information.
15176 @item -march=@var{cpu-type}
15178 Generate code that will run on @var{cpu-type}, which is the name of a system
15179 representing a certain processor type. Possible values for
15180 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15181 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15182 When generating code using the instructions available on z/Architecture,
15183 the default is @option{-march=z900}. Otherwise, the default is
15184 @option{-march=g5}.
15186 @item -mtune=@var{cpu-type}
15188 Tune to @var{cpu-type} everything applicable about the generated code,
15189 except for the ABI and the set of available instructions.
15190 The list of @var{cpu-type} values is the same as for @option{-march}.
15191 The default is the value used for @option{-march}.
15194 @itemx -mno-tpf-trace
15195 @opindex mtpf-trace
15196 @opindex mno-tpf-trace
15197 Generate code that adds (does not add) in TPF OS specific branches to trace
15198 routines in the operating system. This option is off by default, even
15199 when compiling for the TPF OS@.
15202 @itemx -mno-fused-madd
15203 @opindex mfused-madd
15204 @opindex mno-fused-madd
15205 Generate code that uses (does not use) the floating point multiply and
15206 accumulate instructions. These instructions are generated by default if
15207 hardware floating point is used.
15209 @item -mwarn-framesize=@var{framesize}
15210 @opindex mwarn-framesize
15211 Emit a warning if the current function exceeds the given frame size. Because
15212 this is a compile time check it doesn't need to be a real problem when the program
15213 runs. It is intended to identify functions which most probably cause
15214 a stack overflow. It is useful to be used in an environment with limited stack
15215 size e.g.@: the linux kernel.
15217 @item -mwarn-dynamicstack
15218 @opindex mwarn-dynamicstack
15219 Emit a warning if the function calls alloca or uses dynamically
15220 sized arrays. This is generally a bad idea with a limited stack size.
15222 @item -mstack-guard=@var{stack-guard}
15223 @itemx -mstack-size=@var{stack-size}
15224 @opindex mstack-guard
15225 @opindex mstack-size
15226 If these options are provided the s390 back end emits additional instructions in
15227 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15228 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15229 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15230 the frame size of the compiled function is chosen.
15231 These options are intended to be used to help debugging stack overflow problems.
15232 The additionally emitted code causes only little overhead and hence can also be
15233 used in production like systems without greater performance degradation. The given
15234 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15235 @var{stack-guard} without exceeding 64k.
15236 In order to be efficient the extra code makes the assumption that the stack starts
15237 at an address aligned to the value given by @var{stack-size}.
15238 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15241 @node Score Options
15242 @subsection Score Options
15243 @cindex Score Options
15245 These options are defined for Score implementations:
15250 Compile code for big endian mode. This is the default.
15254 Compile code for little endian mode.
15258 Disable generate bcnz instruction.
15262 Enable generate unaligned load and store instruction.
15266 Enable the use of multiply-accumulate instructions. Disabled by default.
15270 Specify the SCORE5 as the target architecture.
15274 Specify the SCORE5U of the target architecture.
15278 Specify the SCORE7 as the target architecture. This is the default.
15282 Specify the SCORE7D as the target architecture.
15286 @subsection SH Options
15288 These @samp{-m} options are defined for the SH implementations:
15293 Generate code for the SH1.
15297 Generate code for the SH2.
15300 Generate code for the SH2e.
15304 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15305 that the floating-point unit is not used.
15307 @item -m2a-single-only
15308 @opindex m2a-single-only
15309 Generate code for the SH2a-FPU, in such a way that no double-precision
15310 floating point operations are used.
15313 @opindex m2a-single
15314 Generate code for the SH2a-FPU assuming the floating-point unit is in
15315 single-precision mode by default.
15319 Generate code for the SH2a-FPU assuming the floating-point unit is in
15320 double-precision mode by default.
15324 Generate code for the SH3.
15328 Generate code for the SH3e.
15332 Generate code for the SH4 without a floating-point unit.
15334 @item -m4-single-only
15335 @opindex m4-single-only
15336 Generate code for the SH4 with a floating-point unit that only
15337 supports single-precision arithmetic.
15341 Generate code for the SH4 assuming the floating-point unit is in
15342 single-precision mode by default.
15346 Generate code for the SH4.
15350 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15351 floating-point unit is not used.
15353 @item -m4a-single-only
15354 @opindex m4a-single-only
15355 Generate code for the SH4a, in such a way that no double-precision
15356 floating point operations are used.
15359 @opindex m4a-single
15360 Generate code for the SH4a assuming the floating-point unit is in
15361 single-precision mode by default.
15365 Generate code for the SH4a.
15369 Same as @option{-m4a-nofpu}, except that it implicitly passes
15370 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15371 instructions at the moment.
15375 Compile code for the processor in big endian mode.
15379 Compile code for the processor in little endian mode.
15383 Align doubles at 64-bit boundaries. Note that this changes the calling
15384 conventions, and thus some functions from the standard C library will
15385 not work unless you recompile it first with @option{-mdalign}.
15389 Shorten some address references at link time, when possible; uses the
15390 linker option @option{-relax}.
15394 Use 32-bit offsets in @code{switch} tables. The default is to use
15399 Enable the use of bit manipulation instructions on SH2A.
15403 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15404 alignment constraints.
15408 Comply with the calling conventions defined by Renesas.
15412 Comply with the calling conventions defined by Renesas.
15416 Comply with the calling conventions defined for GCC before the Renesas
15417 conventions were available. This option is the default for all
15418 targets of the SH toolchain except for @samp{sh-symbianelf}.
15421 @opindex mnomacsave
15422 Mark the @code{MAC} register as call-clobbered, even if
15423 @option{-mhitachi} is given.
15427 Increase IEEE-compliance of floating-point code.
15428 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15429 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15430 comparisons of NANs / infinities incurs extra overhead in every
15431 floating point comparison, therefore the default is set to
15432 @option{-ffinite-math-only}.
15434 @item -minline-ic_invalidate
15435 @opindex minline-ic_invalidate
15436 Inline code to invalidate instruction cache entries after setting up
15437 nested function trampolines.
15438 This option has no effect if -musermode is in effect and the selected
15439 code generation option (e.g. -m4) does not allow the use of the icbi
15441 If the selected code generation option does not allow the use of the icbi
15442 instruction, and -musermode is not in effect, the inlined code will
15443 manipulate the instruction cache address array directly with an associative
15444 write. This not only requires privileged mode, but it will also
15445 fail if the cache line had been mapped via the TLB and has become unmapped.
15449 Dump instruction size and location in the assembly code.
15452 @opindex mpadstruct
15453 This option is deprecated. It pads structures to multiple of 4 bytes,
15454 which is incompatible with the SH ABI@.
15458 Optimize for space instead of speed. Implied by @option{-Os}.
15461 @opindex mprefergot
15462 When generating position-independent code, emit function calls using
15463 the Global Offset Table instead of the Procedure Linkage Table.
15467 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15468 if the inlined code would not work in user mode.
15469 This is the default when the target is @code{sh-*-linux*}.
15471 @item -multcost=@var{number}
15472 @opindex multcost=@var{number}
15473 Set the cost to assume for a multiply insn.
15475 @item -mdiv=@var{strategy}
15476 @opindex mdiv=@var{strategy}
15477 Set the division strategy to use for SHmedia code. @var{strategy} must be
15478 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15479 inv:call2, inv:fp .
15480 "fp" performs the operation in floating point. This has a very high latency,
15481 but needs only a few instructions, so it might be a good choice if
15482 your code has enough easily exploitable ILP to allow the compiler to
15483 schedule the floating point instructions together with other instructions.
15484 Division by zero causes a floating point exception.
15485 "inv" uses integer operations to calculate the inverse of the divisor,
15486 and then multiplies the dividend with the inverse. This strategy allows
15487 cse and hoisting of the inverse calculation. Division by zero calculates
15488 an unspecified result, but does not trap.
15489 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15490 have been found, or if the entire operation has been hoisted to the same
15491 place, the last stages of the inverse calculation are intertwined with the
15492 final multiply to reduce the overall latency, at the expense of using a few
15493 more instructions, and thus offering fewer scheduling opportunities with
15495 "call" calls a library function that usually implements the inv:minlat
15497 This gives high code density for m5-*media-nofpu compilations.
15498 "call2" uses a different entry point of the same library function, where it
15499 assumes that a pointer to a lookup table has already been set up, which
15500 exposes the pointer load to cse / code hoisting optimizations.
15501 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15502 code generation, but if the code stays unoptimized, revert to the "call",
15503 "call2", or "fp" strategies, respectively. Note that the
15504 potentially-trapping side effect of division by zero is carried by a
15505 separate instruction, so it is possible that all the integer instructions
15506 are hoisted out, but the marker for the side effect stays where it is.
15507 A recombination to fp operations or a call is not possible in that case.
15508 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15509 that the inverse calculation was nor separated from the multiply, they speed
15510 up division where the dividend fits into 20 bits (plus sign where applicable),
15511 by inserting a test to skip a number of operations in this case; this test
15512 slows down the case of larger dividends. inv20u assumes the case of a such
15513 a small dividend to be unlikely, and inv20l assumes it to be likely.
15515 @item -mdivsi3_libfunc=@var{name}
15516 @opindex mdivsi3_libfunc=@var{name}
15517 Set the name of the library function used for 32 bit signed division to
15518 @var{name}. This only affect the name used in the call and inv:call
15519 division strategies, and the compiler will still expect the same
15520 sets of input/output/clobbered registers as if this option was not present.
15522 @item -mfixed-range=@var{register-range}
15523 @opindex mfixed-range
15524 Generate code treating the given register range as fixed registers.
15525 A fixed register is one that the register allocator can not use. This is
15526 useful when compiling kernel code. A register range is specified as
15527 two registers separated by a dash. Multiple register ranges can be
15528 specified separated by a comma.
15530 @item -madjust-unroll
15531 @opindex madjust-unroll
15532 Throttle unrolling to avoid thrashing target registers.
15533 This option only has an effect if the gcc code base supports the
15534 TARGET_ADJUST_UNROLL_MAX target hook.
15536 @item -mindexed-addressing
15537 @opindex mindexed-addressing
15538 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15539 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15540 semantics for the indexed addressing mode. The architecture allows the
15541 implementation of processors with 64 bit MMU, which the OS could use to
15542 get 32 bit addressing, but since no current hardware implementation supports
15543 this or any other way to make the indexed addressing mode safe to use in
15544 the 32 bit ABI, the default is -mno-indexed-addressing.
15546 @item -mgettrcost=@var{number}
15547 @opindex mgettrcost=@var{number}
15548 Set the cost assumed for the gettr instruction to @var{number}.
15549 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15553 Assume pt* instructions won't trap. This will generally generate better
15554 scheduled code, but is unsafe on current hardware. The current architecture
15555 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15556 This has the unintentional effect of making it unsafe to schedule ptabs /
15557 ptrel before a branch, or hoist it out of a loop. For example,
15558 __do_global_ctors, a part of libgcc that runs constructors at program
15559 startup, calls functions in a list which is delimited by @minus{}1. With the
15560 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15561 That means that all the constructors will be run a bit quicker, but when
15562 the loop comes to the end of the list, the program crashes because ptabs
15563 loads @minus{}1 into a target register. Since this option is unsafe for any
15564 hardware implementing the current architecture specification, the default
15565 is -mno-pt-fixed. Unless the user specifies a specific cost with
15566 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15567 this deters register allocation using target registers for storing
15570 @item -minvalid-symbols
15571 @opindex minvalid-symbols
15572 Assume symbols might be invalid. Ordinary function symbols generated by
15573 the compiler will always be valid to load with movi/shori/ptabs or
15574 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15575 to generate symbols that will cause ptabs / ptrel to trap.
15576 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15577 It will then prevent cross-basic-block cse, hoisting and most scheduling
15578 of symbol loads. The default is @option{-mno-invalid-symbols}.
15581 @node SPARC Options
15582 @subsection SPARC Options
15583 @cindex SPARC options
15585 These @samp{-m} options are supported on the SPARC:
15588 @item -mno-app-regs
15590 @opindex mno-app-regs
15592 Specify @option{-mapp-regs} to generate output using the global registers
15593 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15596 To be fully SVR4 ABI compliant at the cost of some performance loss,
15597 specify @option{-mno-app-regs}. You should compile libraries and system
15598 software with this option.
15601 @itemx -mhard-float
15603 @opindex mhard-float
15604 Generate output containing floating point instructions. This is the
15608 @itemx -msoft-float
15610 @opindex msoft-float
15611 Generate output containing library calls for floating point.
15612 @strong{Warning:} the requisite libraries are not available for all SPARC
15613 targets. Normally the facilities of the machine's usual C compiler are
15614 used, but this cannot be done directly in cross-compilation. You must make
15615 your own arrangements to provide suitable library functions for
15616 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15617 @samp{sparclite-*-*} do provide software floating point support.
15619 @option{-msoft-float} changes the calling convention in the output file;
15620 therefore, it is only useful if you compile @emph{all} of a program with
15621 this option. In particular, you need to compile @file{libgcc.a}, the
15622 library that comes with GCC, with @option{-msoft-float} in order for
15625 @item -mhard-quad-float
15626 @opindex mhard-quad-float
15627 Generate output containing quad-word (long double) floating point
15630 @item -msoft-quad-float
15631 @opindex msoft-quad-float
15632 Generate output containing library calls for quad-word (long double)
15633 floating point instructions. The functions called are those specified
15634 in the SPARC ABI@. This is the default.
15636 As of this writing, there are no SPARC implementations that have hardware
15637 support for the quad-word floating point instructions. They all invoke
15638 a trap handler for one of these instructions, and then the trap handler
15639 emulates the effect of the instruction. Because of the trap handler overhead,
15640 this is much slower than calling the ABI library routines. Thus the
15641 @option{-msoft-quad-float} option is the default.
15643 @item -mno-unaligned-doubles
15644 @itemx -munaligned-doubles
15645 @opindex mno-unaligned-doubles
15646 @opindex munaligned-doubles
15647 Assume that doubles have 8 byte alignment. This is the default.
15649 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15650 alignment only if they are contained in another type, or if they have an
15651 absolute address. Otherwise, it assumes they have 4 byte alignment.
15652 Specifying this option avoids some rare compatibility problems with code
15653 generated by other compilers. It is not the default because it results
15654 in a performance loss, especially for floating point code.
15656 @item -mno-faster-structs
15657 @itemx -mfaster-structs
15658 @opindex mno-faster-structs
15659 @opindex mfaster-structs
15660 With @option{-mfaster-structs}, the compiler assumes that structures
15661 should have 8 byte alignment. This enables the use of pairs of
15662 @code{ldd} and @code{std} instructions for copies in structure
15663 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15664 However, the use of this changed alignment directly violates the SPARC
15665 ABI@. Thus, it's intended only for use on targets where the developer
15666 acknowledges that their resulting code will not be directly in line with
15667 the rules of the ABI@.
15669 @item -mimpure-text
15670 @opindex mimpure-text
15671 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15672 the compiler to not pass @option{-z text} to the linker when linking a
15673 shared object. Using this option, you can link position-dependent
15674 code into a shared object.
15676 @option{-mimpure-text} suppresses the ``relocations remain against
15677 allocatable but non-writable sections'' linker error message.
15678 However, the necessary relocations will trigger copy-on-write, and the
15679 shared object is not actually shared across processes. Instead of
15680 using @option{-mimpure-text}, you should compile all source code with
15681 @option{-fpic} or @option{-fPIC}.
15683 This option is only available on SunOS and Solaris.
15685 @item -mcpu=@var{cpu_type}
15687 Set the instruction set, register set, and instruction scheduling parameters
15688 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15689 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15690 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15691 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15692 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15694 Default instruction scheduling parameters are used for values that select
15695 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15696 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15698 Here is a list of each supported architecture and their supported
15703 v8: supersparc, hypersparc
15704 sparclite: f930, f934, sparclite86x
15706 v9: ultrasparc, ultrasparc3, niagara, niagara2
15709 By default (unless configured otherwise), GCC generates code for the V7
15710 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15711 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15712 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15713 SPARCStation 1, 2, IPX etc.
15715 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15716 architecture. The only difference from V7 code is that the compiler emits
15717 the integer multiply and integer divide instructions which exist in SPARC-V8
15718 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15719 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15722 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15723 the SPARC architecture. This adds the integer multiply, integer divide step
15724 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15725 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15726 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15727 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15728 MB86934 chip, which is the more recent SPARClite with FPU@.
15730 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15731 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15732 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15733 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15734 optimizes it for the TEMIC SPARClet chip.
15736 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15737 architecture. This adds 64-bit integer and floating-point move instructions,
15738 3 additional floating-point condition code registers and conditional move
15739 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15740 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15741 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15742 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15743 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15744 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15745 additionally optimizes it for Sun UltraSPARC T2 chips.
15747 @item -mtune=@var{cpu_type}
15749 Set the instruction scheduling parameters for machine type
15750 @var{cpu_type}, but do not set the instruction set or register set that the
15751 option @option{-mcpu=@var{cpu_type}} would.
15753 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15754 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15755 that select a particular cpu implementation. Those are @samp{cypress},
15756 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15757 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15758 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15763 @opindex mno-v8plus
15764 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15765 difference from the V8 ABI is that the global and out registers are
15766 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15767 mode for all SPARC-V9 processors.
15773 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15774 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15777 These @samp{-m} options are supported in addition to the above
15778 on SPARC-V9 processors in 64-bit environments:
15781 @item -mlittle-endian
15782 @opindex mlittle-endian
15783 Generate code for a processor running in little-endian mode. It is only
15784 available for a few configurations and most notably not on Solaris and Linux.
15790 Generate code for a 32-bit or 64-bit environment.
15791 The 32-bit environment sets int, long and pointer to 32 bits.
15792 The 64-bit environment sets int to 32 bits and long and pointer
15795 @item -mcmodel=medlow
15796 @opindex mcmodel=medlow
15797 Generate code for the Medium/Low code model: 64-bit addresses, programs
15798 must be linked in the low 32 bits of memory. Programs can be statically
15799 or dynamically linked.
15801 @item -mcmodel=medmid
15802 @opindex mcmodel=medmid
15803 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15804 must be linked in the low 44 bits of memory, the text and data segments must
15805 be less than 2GB in size and the data segment must be located within 2GB of
15808 @item -mcmodel=medany
15809 @opindex mcmodel=medany
15810 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15811 may be linked anywhere in memory, the text and data segments must be less
15812 than 2GB in size and the data segment must be located within 2GB of the
15815 @item -mcmodel=embmedany
15816 @opindex mcmodel=embmedany
15817 Generate code for the Medium/Anywhere code model for embedded systems:
15818 64-bit addresses, the text and data segments must be less than 2GB in
15819 size, both starting anywhere in memory (determined at link time). The
15820 global register %g4 points to the base of the data segment. Programs
15821 are statically linked and PIC is not supported.
15824 @itemx -mno-stack-bias
15825 @opindex mstack-bias
15826 @opindex mno-stack-bias
15827 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15828 frame pointer if present, are offset by @minus{}2047 which must be added back
15829 when making stack frame references. This is the default in 64-bit mode.
15830 Otherwise, assume no such offset is present.
15833 These switches are supported in addition to the above on Solaris:
15838 Add support for multithreading using the Solaris threads library. This
15839 option sets flags for both the preprocessor and linker. This option does
15840 not affect the thread safety of object code produced by the compiler or
15841 that of libraries supplied with it.
15845 Add support for multithreading using the POSIX threads library. This
15846 option sets flags for both the preprocessor and linker. This option does
15847 not affect the thread safety of object code produced by the compiler or
15848 that of libraries supplied with it.
15852 This is a synonym for @option{-pthreads}.
15856 @subsection SPU Options
15857 @cindex SPU options
15859 These @samp{-m} options are supported on the SPU:
15863 @itemx -merror-reloc
15864 @opindex mwarn-reloc
15865 @opindex merror-reloc
15867 The loader for SPU does not handle dynamic relocations. By default, GCC
15868 will give an error when it generates code that requires a dynamic
15869 relocation. @option{-mno-error-reloc} disables the error,
15870 @option{-mwarn-reloc} will generate a warning instead.
15873 @itemx -munsafe-dma
15875 @opindex munsafe-dma
15877 Instructions which initiate or test completion of DMA must not be
15878 reordered with respect to loads and stores of the memory which is being
15879 accessed. Users typically address this problem using the volatile
15880 keyword, but that can lead to inefficient code in places where the
15881 memory is known to not change. Rather than mark the memory as volatile
15882 we treat the DMA instructions as potentially effecting all memory. With
15883 @option{-munsafe-dma} users must use the volatile keyword to protect
15886 @item -mbranch-hints
15887 @opindex mbranch-hints
15889 By default, GCC will generate a branch hint instruction to avoid
15890 pipeline stalls for always taken or probably taken branches. A hint
15891 will not be generated closer than 8 instructions away from its branch.
15892 There is little reason to disable them, except for debugging purposes,
15893 or to make an object a little bit smaller.
15897 @opindex msmall-mem
15898 @opindex mlarge-mem
15900 By default, GCC generates code assuming that addresses are never larger
15901 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15902 a full 32 bit address.
15907 By default, GCC links against startup code that assumes the SPU-style
15908 main function interface (which has an unconventional parameter list).
15909 With @option{-mstdmain}, GCC will link your program against startup
15910 code that assumes a C99-style interface to @code{main}, including a
15911 local copy of @code{argv} strings.
15913 @item -mfixed-range=@var{register-range}
15914 @opindex mfixed-range
15915 Generate code treating the given register range as fixed registers.
15916 A fixed register is one that the register allocator can not use. This is
15917 useful when compiling kernel code. A register range is specified as
15918 two registers separated by a dash. Multiple register ranges can be
15919 specified separated by a comma.
15922 @itemx -mdual-nops=@var{n}
15923 @opindex mdual-nops
15924 By default, GCC will insert nops to increase dual issue when it expects
15925 it to increase performance. @var{n} can be a value from 0 to 10. A
15926 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15927 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15929 @item -mhint-max-nops=@var{n}
15930 @opindex mhint-max-nops
15931 Maximum number of nops to insert for a branch hint. A branch hint must
15932 be at least 8 instructions away from the branch it is effecting. GCC
15933 will insert up to @var{n} nops to enforce this, otherwise it will not
15934 generate the branch hint.
15936 @item -mhint-max-distance=@var{n}
15937 @opindex mhint-max-distance
15938 The encoding of the branch hint instruction limits the hint to be within
15939 256 instructions of the branch it is effecting. By default, GCC makes
15940 sure it is within 125.
15943 @opindex msafe-hints
15944 Work around a hardware bug which causes the SPU to stall indefinitely.
15945 By default, GCC will insert the @code{hbrp} instruction to make sure
15946 this stall won't happen.
15950 @node System V Options
15951 @subsection Options for System V
15953 These additional options are available on System V Release 4 for
15954 compatibility with other compilers on those systems:
15959 Create a shared object.
15960 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15964 Identify the versions of each tool used by the compiler, in a
15965 @code{.ident} assembler directive in the output.
15969 Refrain from adding @code{.ident} directives to the output file (this is
15972 @item -YP,@var{dirs}
15974 Search the directories @var{dirs}, and no others, for libraries
15975 specified with @option{-l}.
15977 @item -Ym,@var{dir}
15979 Look in the directory @var{dir} to find the M4 preprocessor.
15980 The assembler uses this option.
15981 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15982 @c the generic assembler that comes with Solaris takes just -Ym.
15986 @subsection V850 Options
15987 @cindex V850 Options
15989 These @samp{-m} options are defined for V850 implementations:
15993 @itemx -mno-long-calls
15994 @opindex mlong-calls
15995 @opindex mno-long-calls
15996 Treat all calls as being far away (near). If calls are assumed to be
15997 far away, the compiler will always load the functions address up into a
15998 register, and call indirect through the pointer.
16004 Do not optimize (do optimize) basic blocks that use the same index
16005 pointer 4 or more times to copy pointer into the @code{ep} register, and
16006 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16007 option is on by default if you optimize.
16009 @item -mno-prolog-function
16010 @itemx -mprolog-function
16011 @opindex mno-prolog-function
16012 @opindex mprolog-function
16013 Do not use (do use) external functions to save and restore registers
16014 at the prologue and epilogue of a function. The external functions
16015 are slower, but use less code space if more than one function saves
16016 the same number of registers. The @option{-mprolog-function} option
16017 is on by default if you optimize.
16021 Try to make the code as small as possible. At present, this just turns
16022 on the @option{-mep} and @option{-mprolog-function} options.
16024 @item -mtda=@var{n}
16026 Put static or global variables whose size is @var{n} bytes or less into
16027 the tiny data area that register @code{ep} points to. The tiny data
16028 area can hold up to 256 bytes in total (128 bytes for byte references).
16030 @item -msda=@var{n}
16032 Put static or global variables whose size is @var{n} bytes or less into
16033 the small data area that register @code{gp} points to. The small data
16034 area can hold up to 64 kilobytes.
16036 @item -mzda=@var{n}
16038 Put static or global variables whose size is @var{n} bytes or less into
16039 the first 32 kilobytes of memory.
16043 Specify that the target processor is the V850.
16046 @opindex mbig-switch
16047 Generate code suitable for big switch tables. Use this option only if
16048 the assembler/linker complain about out of range branches within a switch
16053 This option will cause r2 and r5 to be used in the code generated by
16054 the compiler. This setting is the default.
16056 @item -mno-app-regs
16057 @opindex mno-app-regs
16058 This option will cause r2 and r5 to be treated as fixed registers.
16062 Specify that the target processor is the V850E1. The preprocessor
16063 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16064 this option is used.
16068 Specify that the target processor is the V850E@. The preprocessor
16069 constant @samp{__v850e__} will be defined if this option is used.
16071 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16072 are defined then a default target processor will be chosen and the
16073 relevant @samp{__v850*__} preprocessor constant will be defined.
16075 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16076 defined, regardless of which processor variant is the target.
16078 @item -mdisable-callt
16079 @opindex mdisable-callt
16080 This option will suppress generation of the CALLT instruction for the
16081 v850e and v850e1 flavors of the v850 architecture. The default is
16082 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16087 @subsection VAX Options
16088 @cindex VAX options
16090 These @samp{-m} options are defined for the VAX:
16095 Do not output certain jump instructions (@code{aobleq} and so on)
16096 that the Unix assembler for the VAX cannot handle across long
16101 Do output those jump instructions, on the assumption that you
16102 will assemble with the GNU assembler.
16106 Output code for g-format floating point numbers instead of d-format.
16109 @node VxWorks Options
16110 @subsection VxWorks Options
16111 @cindex VxWorks Options
16113 The options in this section are defined for all VxWorks targets.
16114 Options specific to the target hardware are listed with the other
16115 options for that target.
16120 GCC can generate code for both VxWorks kernels and real time processes
16121 (RTPs). This option switches from the former to the latter. It also
16122 defines the preprocessor macro @code{__RTP__}.
16125 @opindex non-static
16126 Link an RTP executable against shared libraries rather than static
16127 libraries. The options @option{-static} and @option{-shared} can
16128 also be used for RTPs (@pxref{Link Options}); @option{-static}
16135 These options are passed down to the linker. They are defined for
16136 compatibility with Diab.
16139 @opindex Xbind-lazy
16140 Enable lazy binding of function calls. This option is equivalent to
16141 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16145 Disable lazy binding of function calls. This option is the default and
16146 is defined for compatibility with Diab.
16149 @node x86-64 Options
16150 @subsection x86-64 Options
16151 @cindex x86-64 options
16153 These are listed under @xref{i386 and x86-64 Options}.
16155 @node i386 and x86-64 Windows Options
16156 @subsection i386 and x86-64 Windows Options
16157 @cindex i386 and x86-64 Windows Options
16159 These additional options are available for Windows targets:
16164 This option is available for Cygwin and MinGW targets. It
16165 specifies that a console application is to be generated, by
16166 instructing the linker to set the PE header subsystem type
16167 required for console applications.
16168 This is the default behavior for Cygwin and MinGW targets.
16172 This option is available for Cygwin targets. It specifies that
16173 the Cygwin internal interface is to be used for predefined
16174 preprocessor macros, C runtime libraries and related linker
16175 paths and options. For Cygwin targets this is the default behavior.
16176 This option is deprecated and will be removed in a future release.
16179 @opindex mno-cygwin
16180 This option is available for Cygwin targets. It specifies that
16181 the MinGW internal interface is to be used instead of Cygwin's, by
16182 setting MinGW-related predefined macros and linker paths and default
16184 This option is deprecated and will be removed in a future release.
16188 This option is available for Cygwin and MinGW targets. It
16189 specifies that a DLL - a dynamic link library - is to be
16190 generated, enabling the selection of the required runtime
16191 startup object and entry point.
16193 @item -mnop-fun-dllimport
16194 @opindex mnop-fun-dllimport
16195 This option is available for Cygwin and MinGW targets. It
16196 specifies that the dllimport attribute should be ignored.
16200 This option is available for MinGW targets. It specifies
16201 that MinGW-specific thread support is to be used.
16205 This option is available for mingw-w64 targets. It specifies
16206 that the UNICODE macro is getting pre-defined and that the
16207 unicode capable runtime startup code is choosen.
16211 This option is available for Cygwin and MinGW targets. It
16212 specifies that the typical Windows pre-defined macros are to
16213 be set in the pre-processor, but does not influence the choice
16214 of runtime library/startup code.
16218 This option is available for Cygwin and MinGW targets. It
16219 specifies that a GUI application is to be generated by
16220 instructing the linker to set the PE header subsystem type
16223 @item -mpe-aligned-commons
16224 @opindex mpe-aligned-commons
16225 This option is available for Cygwin and MinGW targets. It
16226 specifies that the GNU extension to the PE file format that
16227 permits the correct alignment of COMMON variables should be
16228 used when generating code. It will be enabled by default if
16229 GCC detects that the target assembler found during configuration
16230 supports the feature.
16233 See also under @ref{i386 and x86-64 Options} for standard options.
16235 @node Xstormy16 Options
16236 @subsection Xstormy16 Options
16237 @cindex Xstormy16 Options
16239 These options are defined for Xstormy16:
16244 Choose startup files and linker script suitable for the simulator.
16247 @node Xtensa Options
16248 @subsection Xtensa Options
16249 @cindex Xtensa Options
16251 These options are supported for Xtensa targets:
16255 @itemx -mno-const16
16257 @opindex mno-const16
16258 Enable or disable use of @code{CONST16} instructions for loading
16259 constant values. The @code{CONST16} instruction is currently not a
16260 standard option from Tensilica. When enabled, @code{CONST16}
16261 instructions are always used in place of the standard @code{L32R}
16262 instructions. The use of @code{CONST16} is enabled by default only if
16263 the @code{L32R} instruction is not available.
16266 @itemx -mno-fused-madd
16267 @opindex mfused-madd
16268 @opindex mno-fused-madd
16269 Enable or disable use of fused multiply/add and multiply/subtract
16270 instructions in the floating-point option. This has no effect if the
16271 floating-point option is not also enabled. Disabling fused multiply/add
16272 and multiply/subtract instructions forces the compiler to use separate
16273 instructions for the multiply and add/subtract operations. This may be
16274 desirable in some cases where strict IEEE 754-compliant results are
16275 required: the fused multiply add/subtract instructions do not round the
16276 intermediate result, thereby producing results with @emph{more} bits of
16277 precision than specified by the IEEE standard. Disabling fused multiply
16278 add/subtract instructions also ensures that the program output is not
16279 sensitive to the compiler's ability to combine multiply and add/subtract
16282 @item -mserialize-volatile
16283 @itemx -mno-serialize-volatile
16284 @opindex mserialize-volatile
16285 @opindex mno-serialize-volatile
16286 When this option is enabled, GCC inserts @code{MEMW} instructions before
16287 @code{volatile} memory references to guarantee sequential consistency.
16288 The default is @option{-mserialize-volatile}. Use
16289 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16291 @item -mtext-section-literals
16292 @itemx -mno-text-section-literals
16293 @opindex mtext-section-literals
16294 @opindex mno-text-section-literals
16295 Control the treatment of literal pools. The default is
16296 @option{-mno-text-section-literals}, which places literals in a separate
16297 section in the output file. This allows the literal pool to be placed
16298 in a data RAM/ROM, and it also allows the linker to combine literal
16299 pools from separate object files to remove redundant literals and
16300 improve code size. With @option{-mtext-section-literals}, the literals
16301 are interspersed in the text section in order to keep them as close as
16302 possible to their references. This may be necessary for large assembly
16305 @item -mtarget-align
16306 @itemx -mno-target-align
16307 @opindex mtarget-align
16308 @opindex mno-target-align
16309 When this option is enabled, GCC instructs the assembler to
16310 automatically align instructions to reduce branch penalties at the
16311 expense of some code density. The assembler attempts to widen density
16312 instructions to align branch targets and the instructions following call
16313 instructions. If there are not enough preceding safe density
16314 instructions to align a target, no widening will be performed. The
16315 default is @option{-mtarget-align}. These options do not affect the
16316 treatment of auto-aligned instructions like @code{LOOP}, which the
16317 assembler will always align, either by widening density instructions or
16318 by inserting no-op instructions.
16321 @itemx -mno-longcalls
16322 @opindex mlongcalls
16323 @opindex mno-longcalls
16324 When this option is enabled, GCC instructs the assembler to translate
16325 direct calls to indirect calls unless it can determine that the target
16326 of a direct call is in the range allowed by the call instruction. This
16327 translation typically occurs for calls to functions in other source
16328 files. Specifically, the assembler translates a direct @code{CALL}
16329 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16330 The default is @option{-mno-longcalls}. This option should be used in
16331 programs where the call target can potentially be out of range. This
16332 option is implemented in the assembler, not the compiler, so the
16333 assembly code generated by GCC will still show direct call
16334 instructions---look at the disassembled object code to see the actual
16335 instructions. Note that the assembler will use an indirect call for
16336 every cross-file call, not just those that really will be out of range.
16339 @node zSeries Options
16340 @subsection zSeries Options
16341 @cindex zSeries options
16343 These are listed under @xref{S/390 and zSeries Options}.
16345 @node Code Gen Options
16346 @section Options for Code Generation Conventions
16347 @cindex code generation conventions
16348 @cindex options, code generation
16349 @cindex run-time options
16351 These machine-independent options control the interface conventions
16352 used in code generation.
16354 Most of them have both positive and negative forms; the negative form
16355 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16356 one of the forms is listed---the one which is not the default. You
16357 can figure out the other form by either removing @samp{no-} or adding
16361 @item -fbounds-check
16362 @opindex fbounds-check
16363 For front-ends that support it, generate additional code to check that
16364 indices used to access arrays are within the declared range. This is
16365 currently only supported by the Java and Fortran front-ends, where
16366 this option defaults to true and false respectively.
16370 This option generates traps for signed overflow on addition, subtraction,
16371 multiplication operations.
16375 This option instructs the compiler to assume that signed arithmetic
16376 overflow of addition, subtraction and multiplication wraps around
16377 using twos-complement representation. This flag enables some optimizations
16378 and disables others. This option is enabled by default for the Java
16379 front-end, as required by the Java language specification.
16382 @opindex fexceptions
16383 Enable exception handling. Generates extra code needed to propagate
16384 exceptions. For some targets, this implies GCC will generate frame
16385 unwind information for all functions, which can produce significant data
16386 size overhead, although it does not affect execution. If you do not
16387 specify this option, GCC will enable it by default for languages like
16388 C++ which normally require exception handling, and disable it for
16389 languages like C that do not normally require it. However, you may need
16390 to enable this option when compiling C code that needs to interoperate
16391 properly with exception handlers written in C++. You may also wish to
16392 disable this option if you are compiling older C++ programs that don't
16393 use exception handling.
16395 @item -fnon-call-exceptions
16396 @opindex fnon-call-exceptions
16397 Generate code that allows trapping instructions to throw exceptions.
16398 Note that this requires platform-specific runtime support that does
16399 not exist everywhere. Moreover, it only allows @emph{trapping}
16400 instructions to throw exceptions, i.e.@: memory references or floating
16401 point instructions. It does not allow exceptions to be thrown from
16402 arbitrary signal handlers such as @code{SIGALRM}.
16404 @item -funwind-tables
16405 @opindex funwind-tables
16406 Similar to @option{-fexceptions}, except that it will just generate any needed
16407 static data, but will not affect the generated code in any other way.
16408 You will normally not enable this option; instead, a language processor
16409 that needs this handling would enable it on your behalf.
16411 @item -fasynchronous-unwind-tables
16412 @opindex fasynchronous-unwind-tables
16413 Generate unwind table in dwarf2 format, if supported by target machine. The
16414 table is exact at each instruction boundary, so it can be used for stack
16415 unwinding from asynchronous events (such as debugger or garbage collector).
16417 @item -fpcc-struct-return
16418 @opindex fpcc-struct-return
16419 Return ``short'' @code{struct} and @code{union} values in memory like
16420 longer ones, rather than in registers. This convention is less
16421 efficient, but it has the advantage of allowing intercallability between
16422 GCC-compiled files and files compiled with other compilers, particularly
16423 the Portable C Compiler (pcc).
16425 The precise convention for returning structures in memory depends
16426 on the target configuration macros.
16428 Short structures and unions are those whose size and alignment match
16429 that of some integer type.
16431 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16432 switch is not binary compatible with code compiled with the
16433 @option{-freg-struct-return} switch.
16434 Use it to conform to a non-default application binary interface.
16436 @item -freg-struct-return
16437 @opindex freg-struct-return
16438 Return @code{struct} and @code{union} values in registers when possible.
16439 This is more efficient for small structures than
16440 @option{-fpcc-struct-return}.
16442 If you specify neither @option{-fpcc-struct-return} nor
16443 @option{-freg-struct-return}, GCC defaults to whichever convention is
16444 standard for the target. If there is no standard convention, GCC
16445 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16446 the principal compiler. In those cases, we can choose the standard, and
16447 we chose the more efficient register return alternative.
16449 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16450 switch is not binary compatible with code compiled with the
16451 @option{-fpcc-struct-return} switch.
16452 Use it to conform to a non-default application binary interface.
16454 @item -fshort-enums
16455 @opindex fshort-enums
16456 Allocate to an @code{enum} type only as many bytes as it needs for the
16457 declared range of possible values. Specifically, the @code{enum} type
16458 will be equivalent to the smallest integer type which has enough room.
16460 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16461 code that is not binary compatible with code generated without that switch.
16462 Use it to conform to a non-default application binary interface.
16464 @item -fshort-double
16465 @opindex fshort-double
16466 Use the same size for @code{double} as for @code{float}.
16468 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16469 code that is not binary compatible with code generated without that switch.
16470 Use it to conform to a non-default application binary interface.
16472 @item -fshort-wchar
16473 @opindex fshort-wchar
16474 Override the underlying type for @samp{wchar_t} to be @samp{short
16475 unsigned int} instead of the default for the target. This option is
16476 useful for building programs to run under WINE@.
16478 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16479 code that is not binary compatible with code generated without that switch.
16480 Use it to conform to a non-default application binary interface.
16483 @opindex fno-common
16484 In C code, controls the placement of uninitialized global variables.
16485 Unix C compilers have traditionally permitted multiple definitions of
16486 such variables in different compilation units by placing the variables
16488 This is the behavior specified by @option{-fcommon}, and is the default
16489 for GCC on most targets.
16490 On the other hand, this behavior is not required by ISO C, and on some
16491 targets may carry a speed or code size penalty on variable references.
16492 The @option{-fno-common} option specifies that the compiler should place
16493 uninitialized global variables in the data section of the object file,
16494 rather than generating them as common blocks.
16495 This has the effect that if the same variable is declared
16496 (without @code{extern}) in two different compilations,
16497 you will get a multiple-definition error when you link them.
16498 In this case, you must compile with @option{-fcommon} instead.
16499 Compiling with @option{-fno-common} is useful on targets for which
16500 it provides better performance, or if you wish to verify that the
16501 program will work on other systems which always treat uninitialized
16502 variable declarations this way.
16506 Ignore the @samp{#ident} directive.
16508 @item -finhibit-size-directive
16509 @opindex finhibit-size-directive
16510 Don't output a @code{.size} assembler directive, or anything else that
16511 would cause trouble if the function is split in the middle, and the
16512 two halves are placed at locations far apart in memory. This option is
16513 used when compiling @file{crtstuff.c}; you should not need to use it
16516 @item -fverbose-asm
16517 @opindex fverbose-asm
16518 Put extra commentary information in the generated assembly code to
16519 make it more readable. This option is generally only of use to those
16520 who actually need to read the generated assembly code (perhaps while
16521 debugging the compiler itself).
16523 @option{-fno-verbose-asm}, the default, causes the
16524 extra information to be omitted and is useful when comparing two assembler
16527 @item -frecord-gcc-switches
16528 @opindex frecord-gcc-switches
16529 This switch causes the command line that was used to invoke the
16530 compiler to be recorded into the object file that is being created.
16531 This switch is only implemented on some targets and the exact format
16532 of the recording is target and binary file format dependent, but it
16533 usually takes the form of a section containing ASCII text. This
16534 switch is related to the @option{-fverbose-asm} switch, but that
16535 switch only records information in the assembler output file as
16536 comments, so it never reaches the object file.
16540 @cindex global offset table
16542 Generate position-independent code (PIC) suitable for use in a shared
16543 library, if supported for the target machine. Such code accesses all
16544 constant addresses through a global offset table (GOT)@. The dynamic
16545 loader resolves the GOT entries when the program starts (the dynamic
16546 loader is not part of GCC; it is part of the operating system). If
16547 the GOT size for the linked executable exceeds a machine-specific
16548 maximum size, you get an error message from the linker indicating that
16549 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16550 instead. (These maximums are 8k on the SPARC and 32k
16551 on the m68k and RS/6000. The 386 has no such limit.)
16553 Position-independent code requires special support, and therefore works
16554 only on certain machines. For the 386, GCC supports PIC for System V
16555 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16556 position-independent.
16558 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16563 If supported for the target machine, emit position-independent code,
16564 suitable for dynamic linking and avoiding any limit on the size of the
16565 global offset table. This option makes a difference on the m68k,
16566 PowerPC and SPARC@.
16568 Position-independent code requires special support, and therefore works
16569 only on certain machines.
16571 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16578 These options are similar to @option{-fpic} and @option{-fPIC}, but
16579 generated position independent code can be only linked into executables.
16580 Usually these options are used when @option{-pie} GCC option will be
16581 used during linking.
16583 @option{-fpie} and @option{-fPIE} both define the macros
16584 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16585 for @option{-fpie} and 2 for @option{-fPIE}.
16587 @item -fno-jump-tables
16588 @opindex fno-jump-tables
16589 Do not use jump tables for switch statements even where it would be
16590 more efficient than other code generation strategies. This option is
16591 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16592 building code which forms part of a dynamic linker and cannot
16593 reference the address of a jump table. On some targets, jump tables
16594 do not require a GOT and this option is not needed.
16596 @item -ffixed-@var{reg}
16598 Treat the register named @var{reg} as a fixed register; generated code
16599 should never refer to it (except perhaps as a stack pointer, frame
16600 pointer or in some other fixed role).
16602 @var{reg} must be the name of a register. The register names accepted
16603 are machine-specific and are defined in the @code{REGISTER_NAMES}
16604 macro in the machine description macro file.
16606 This flag does not have a negative form, because it specifies a
16609 @item -fcall-used-@var{reg}
16610 @opindex fcall-used
16611 Treat the register named @var{reg} as an allocable register that is
16612 clobbered by function calls. It may be allocated for temporaries or
16613 variables that do not live across a call. Functions compiled this way
16614 will not save and restore the register @var{reg}.
16616 It is an error to used this flag with the frame pointer or stack pointer.
16617 Use of this flag for other registers that have fixed pervasive roles in
16618 the machine's execution model will produce disastrous results.
16620 This flag does not have a negative form, because it specifies a
16623 @item -fcall-saved-@var{reg}
16624 @opindex fcall-saved
16625 Treat the register named @var{reg} as an allocable register saved by
16626 functions. It may be allocated even for temporaries or variables that
16627 live across a call. Functions compiled this way will save and restore
16628 the register @var{reg} if they use it.
16630 It is an error to used this flag with the frame pointer or stack pointer.
16631 Use of this flag for other registers that have fixed pervasive roles in
16632 the machine's execution model will produce disastrous results.
16634 A different sort of disaster will result from the use of this flag for
16635 a register in which function values may be returned.
16637 This flag does not have a negative form, because it specifies a
16640 @item -fpack-struct[=@var{n}]
16641 @opindex fpack-struct
16642 Without a value specified, pack all structure members together without
16643 holes. When a value is specified (which must be a small power of two), pack
16644 structure members according to this value, representing the maximum
16645 alignment (that is, objects with default alignment requirements larger than
16646 this will be output potentially unaligned at the next fitting location.
16648 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16649 code that is not binary compatible with code generated without that switch.
16650 Additionally, it makes the code suboptimal.
16651 Use it to conform to a non-default application binary interface.
16653 @item -finstrument-functions
16654 @opindex finstrument-functions
16655 Generate instrumentation calls for entry and exit to functions. Just
16656 after function entry and just before function exit, the following
16657 profiling functions will be called with the address of the current
16658 function and its call site. (On some platforms,
16659 @code{__builtin_return_address} does not work beyond the current
16660 function, so the call site information may not be available to the
16661 profiling functions otherwise.)
16664 void __cyg_profile_func_enter (void *this_fn,
16666 void __cyg_profile_func_exit (void *this_fn,
16670 The first argument is the address of the start of the current function,
16671 which may be looked up exactly in the symbol table.
16673 This instrumentation is also done for functions expanded inline in other
16674 functions. The profiling calls will indicate where, conceptually, the
16675 inline function is entered and exited. This means that addressable
16676 versions of such functions must be available. If all your uses of a
16677 function are expanded inline, this may mean an additional expansion of
16678 code size. If you use @samp{extern inline} in your C code, an
16679 addressable version of such functions must be provided. (This is
16680 normally the case anyways, but if you get lucky and the optimizer always
16681 expands the functions inline, you might have gotten away without
16682 providing static copies.)
16684 A function may be given the attribute @code{no_instrument_function}, in
16685 which case this instrumentation will not be done. This can be used, for
16686 example, for the profiling functions listed above, high-priority
16687 interrupt routines, and any functions from which the profiling functions
16688 cannot safely be called (perhaps signal handlers, if the profiling
16689 routines generate output or allocate memory).
16691 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16692 @opindex finstrument-functions-exclude-file-list
16694 Set the list of functions that are excluded from instrumentation (see
16695 the description of @code{-finstrument-functions}). If the file that
16696 contains a function definition matches with one of @var{file}, then
16697 that function is not instrumented. The match is done on substrings:
16698 if the @var{file} parameter is a substring of the file name, it is
16699 considered to be a match.
16702 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16703 will exclude any inline function defined in files whose pathnames
16704 contain @code{/bits/stl} or @code{include/sys}.
16706 If, for some reason, you want to include letter @code{','} in one of
16707 @var{sym}, write @code{'\,'}. For example,
16708 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16709 (note the single quote surrounding the option).
16711 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16712 @opindex finstrument-functions-exclude-function-list
16714 This is similar to @code{-finstrument-functions-exclude-file-list},
16715 but this option sets the list of function names to be excluded from
16716 instrumentation. The function name to be matched is its user-visible
16717 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16718 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16719 match is done on substrings: if the @var{sym} parameter is a substring
16720 of the function name, it is considered to be a match. For C99 and C++
16721 extended identifiers, the function name must be given in UTF-8, not
16722 using universal character names.
16724 @item -fstack-check
16725 @opindex fstack-check
16726 Generate code to verify that you do not go beyond the boundary of the
16727 stack. You should specify this flag if you are running in an
16728 environment with multiple threads, but only rarely need to specify it in
16729 a single-threaded environment since stack overflow is automatically
16730 detected on nearly all systems if there is only one stack.
16732 Note that this switch does not actually cause checking to be done; the
16733 operating system or the language runtime must do that. The switch causes
16734 generation of code to ensure that they see the stack being extended.
16736 You can additionally specify a string parameter: @code{no} means no
16737 checking, @code{generic} means force the use of old-style checking,
16738 @code{specific} means use the best checking method and is equivalent
16739 to bare @option{-fstack-check}.
16741 Old-style checking is a generic mechanism that requires no specific
16742 target support in the compiler but comes with the following drawbacks:
16746 Modified allocation strategy for large objects: they will always be
16747 allocated dynamically if their size exceeds a fixed threshold.
16750 Fixed limit on the size of the static frame of functions: when it is
16751 topped by a particular function, stack checking is not reliable and
16752 a warning is issued by the compiler.
16755 Inefficiency: because of both the modified allocation strategy and the
16756 generic implementation, the performances of the code are hampered.
16759 Note that old-style stack checking is also the fallback method for
16760 @code{specific} if no target support has been added in the compiler.
16762 @item -fstack-limit-register=@var{reg}
16763 @itemx -fstack-limit-symbol=@var{sym}
16764 @itemx -fno-stack-limit
16765 @opindex fstack-limit-register
16766 @opindex fstack-limit-symbol
16767 @opindex fno-stack-limit
16768 Generate code to ensure that the stack does not grow beyond a certain value,
16769 either the value of a register or the address of a symbol. If the stack
16770 would grow beyond the value, a signal is raised. For most targets,
16771 the signal is raised before the stack overruns the boundary, so
16772 it is possible to catch the signal without taking special precautions.
16774 For instance, if the stack starts at absolute address @samp{0x80000000}
16775 and grows downwards, you can use the flags
16776 @option{-fstack-limit-symbol=__stack_limit} and
16777 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16778 of 128KB@. Note that this may only work with the GNU linker.
16780 @cindex aliasing of parameters
16781 @cindex parameters, aliased
16782 @item -fargument-alias
16783 @itemx -fargument-noalias
16784 @itemx -fargument-noalias-global
16785 @itemx -fargument-noalias-anything
16786 @opindex fargument-alias
16787 @opindex fargument-noalias
16788 @opindex fargument-noalias-global
16789 @opindex fargument-noalias-anything
16790 Specify the possible relationships among parameters and between
16791 parameters and global data.
16793 @option{-fargument-alias} specifies that arguments (parameters) may
16794 alias each other and may alias global storage.@*
16795 @option{-fargument-noalias} specifies that arguments do not alias
16796 each other, but may alias global storage.@*
16797 @option{-fargument-noalias-global} specifies that arguments do not
16798 alias each other and do not alias global storage.
16799 @option{-fargument-noalias-anything} specifies that arguments do not
16800 alias any other storage.
16802 Each language will automatically use whatever option is required by
16803 the language standard. You should not need to use these options yourself.
16805 @item -fleading-underscore
16806 @opindex fleading-underscore
16807 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16808 change the way C symbols are represented in the object file. One use
16809 is to help link with legacy assembly code.
16811 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16812 generate code that is not binary compatible with code generated without that
16813 switch. Use it to conform to a non-default application binary interface.
16814 Not all targets provide complete support for this switch.
16816 @item -ftls-model=@var{model}
16817 @opindex ftls-model
16818 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16819 The @var{model} argument should be one of @code{global-dynamic},
16820 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16822 The default without @option{-fpic} is @code{initial-exec}; with
16823 @option{-fpic} the default is @code{global-dynamic}.
16825 @item -fvisibility=@var{default|internal|hidden|protected}
16826 @opindex fvisibility
16827 Set the default ELF image symbol visibility to the specified option---all
16828 symbols will be marked with this unless overridden within the code.
16829 Using this feature can very substantially improve linking and
16830 load times of shared object libraries, produce more optimized
16831 code, provide near-perfect API export and prevent symbol clashes.
16832 It is @strong{strongly} recommended that you use this in any shared objects
16835 Despite the nomenclature, @code{default} always means public ie;
16836 available to be linked against from outside the shared object.
16837 @code{protected} and @code{internal} are pretty useless in real-world
16838 usage so the only other commonly used option will be @code{hidden}.
16839 The default if @option{-fvisibility} isn't specified is
16840 @code{default}, i.e., make every
16841 symbol public---this causes the same behavior as previous versions of
16844 A good explanation of the benefits offered by ensuring ELF
16845 symbols have the correct visibility is given by ``How To Write
16846 Shared Libraries'' by Ulrich Drepper (which can be found at
16847 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16848 solution made possible by this option to marking things hidden when
16849 the default is public is to make the default hidden and mark things
16850 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16851 and @code{__attribute__ ((visibility("default")))} instead of
16852 @code{__declspec(dllexport)} you get almost identical semantics with
16853 identical syntax. This is a great boon to those working with
16854 cross-platform projects.
16856 For those adding visibility support to existing code, you may find
16857 @samp{#pragma GCC visibility} of use. This works by you enclosing
16858 the declarations you wish to set visibility for with (for example)
16859 @samp{#pragma GCC visibility push(hidden)} and
16860 @samp{#pragma GCC visibility pop}.
16861 Bear in mind that symbol visibility should be viewed @strong{as
16862 part of the API interface contract} and thus all new code should
16863 always specify visibility when it is not the default ie; declarations
16864 only for use within the local DSO should @strong{always} be marked explicitly
16865 as hidden as so to avoid PLT indirection overheads---making this
16866 abundantly clear also aids readability and self-documentation of the code.
16867 Note that due to ISO C++ specification requirements, operator new and
16868 operator delete must always be of default visibility.
16870 Be aware that headers from outside your project, in particular system
16871 headers and headers from any other library you use, may not be
16872 expecting to be compiled with visibility other than the default. You
16873 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16874 before including any such headers.
16876 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16877 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16878 no modifications. However, this means that calls to @samp{extern}
16879 functions with no explicit visibility will use the PLT, so it is more
16880 effective to use @samp{__attribute ((visibility))} and/or
16881 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16882 declarations should be treated as hidden.
16884 Note that @samp{-fvisibility} does affect C++ vague linkage
16885 entities. This means that, for instance, an exception class that will
16886 be thrown between DSOs must be explicitly marked with default
16887 visibility so that the @samp{type_info} nodes will be unified between
16890 An overview of these techniques, their benefits and how to use them
16891 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16897 @node Environment Variables
16898 @section Environment Variables Affecting GCC
16899 @cindex environment variables
16901 @c man begin ENVIRONMENT
16902 This section describes several environment variables that affect how GCC
16903 operates. Some of them work by specifying directories or prefixes to use
16904 when searching for various kinds of files. Some are used to specify other
16905 aspects of the compilation environment.
16907 Note that you can also specify places to search using options such as
16908 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16909 take precedence over places specified using environment variables, which
16910 in turn take precedence over those specified by the configuration of GCC@.
16911 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16912 GNU Compiler Collection (GCC) Internals}.
16917 @c @itemx LC_COLLATE
16919 @c @itemx LC_MONETARY
16920 @c @itemx LC_NUMERIC
16925 @c @findex LC_COLLATE
16926 @findex LC_MESSAGES
16927 @c @findex LC_MONETARY
16928 @c @findex LC_NUMERIC
16932 These environment variables control the way that GCC uses
16933 localization information that allow GCC to work with different
16934 national conventions. GCC inspects the locale categories
16935 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16936 so. These locale categories can be set to any value supported by your
16937 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16938 Kingdom encoded in UTF-8.
16940 The @env{LC_CTYPE} environment variable specifies character
16941 classification. GCC uses it to determine the character boundaries in
16942 a string; this is needed for some multibyte encodings that contain quote
16943 and escape characters that would otherwise be interpreted as a string
16946 The @env{LC_MESSAGES} environment variable specifies the language to
16947 use in diagnostic messages.
16949 If the @env{LC_ALL} environment variable is set, it overrides the value
16950 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16951 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16952 environment variable. If none of these variables are set, GCC
16953 defaults to traditional C English behavior.
16957 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16958 files. GCC uses temporary files to hold the output of one stage of
16959 compilation which is to be used as input to the next stage: for example,
16960 the output of the preprocessor, which is the input to the compiler
16963 @item GCC_EXEC_PREFIX
16964 @findex GCC_EXEC_PREFIX
16965 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16966 names of the subprograms executed by the compiler. No slash is added
16967 when this prefix is combined with the name of a subprogram, but you can
16968 specify a prefix that ends with a slash if you wish.
16970 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16971 an appropriate prefix to use based on the pathname it was invoked with.
16973 If GCC cannot find the subprogram using the specified prefix, it
16974 tries looking in the usual places for the subprogram.
16976 The default value of @env{GCC_EXEC_PREFIX} is
16977 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16978 the installed compiler. In many cases @var{prefix} is the value
16979 of @code{prefix} when you ran the @file{configure} script.
16981 Other prefixes specified with @option{-B} take precedence over this prefix.
16983 This prefix is also used for finding files such as @file{crt0.o} that are
16986 In addition, the prefix is used in an unusual way in finding the
16987 directories to search for header files. For each of the standard
16988 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16989 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16990 replacing that beginning with the specified prefix to produce an
16991 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16992 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16993 These alternate directories are searched first; the standard directories
16994 come next. If a standard directory begins with the configured
16995 @var{prefix} then the value of @var{prefix} is replaced by
16996 @env{GCC_EXEC_PREFIX} when looking for header files.
16998 @item COMPILER_PATH
16999 @findex COMPILER_PATH
17000 The value of @env{COMPILER_PATH} is a colon-separated list of
17001 directories, much like @env{PATH}. GCC tries the directories thus
17002 specified when searching for subprograms, if it can't find the
17003 subprograms using @env{GCC_EXEC_PREFIX}.
17006 @findex LIBRARY_PATH
17007 The value of @env{LIBRARY_PATH} is a colon-separated list of
17008 directories, much like @env{PATH}. When configured as a native compiler,
17009 GCC tries the directories thus specified when searching for special
17010 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17011 using GCC also uses these directories when searching for ordinary
17012 libraries for the @option{-l} option (but directories specified with
17013 @option{-L} come first).
17017 @cindex locale definition
17018 This variable is used to pass locale information to the compiler. One way in
17019 which this information is used is to determine the character set to be used
17020 when character literals, string literals and comments are parsed in C and C++.
17021 When the compiler is configured to allow multibyte characters,
17022 the following values for @env{LANG} are recognized:
17026 Recognize JIS characters.
17028 Recognize SJIS characters.
17030 Recognize EUCJP characters.
17033 If @env{LANG} is not defined, or if it has some other value, then the
17034 compiler will use mblen and mbtowc as defined by the default locale to
17035 recognize and translate multibyte characters.
17039 Some additional environments variables affect the behavior of the
17042 @include cppenv.texi
17046 @node Precompiled Headers
17047 @section Using Precompiled Headers
17048 @cindex precompiled headers
17049 @cindex speed of compilation
17051 Often large projects have many header files that are included in every
17052 source file. The time the compiler takes to process these header files
17053 over and over again can account for nearly all of the time required to
17054 build the project. To make builds faster, GCC allows users to
17055 `precompile' a header file; then, if builds can use the precompiled
17056 header file they will be much faster.
17058 To create a precompiled header file, simply compile it as you would any
17059 other file, if necessary using the @option{-x} option to make the driver
17060 treat it as a C or C++ header file. You will probably want to use a
17061 tool like @command{make} to keep the precompiled header up-to-date when
17062 the headers it contains change.
17064 A precompiled header file will be searched for when @code{#include} is
17065 seen in the compilation. As it searches for the included file
17066 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17067 compiler looks for a precompiled header in each directory just before it
17068 looks for the include file in that directory. The name searched for is
17069 the name specified in the @code{#include} with @samp{.gch} appended. If
17070 the precompiled header file can't be used, it is ignored.
17072 For instance, if you have @code{#include "all.h"}, and you have
17073 @file{all.h.gch} in the same directory as @file{all.h}, then the
17074 precompiled header file will be used if possible, and the original
17075 header will be used otherwise.
17077 Alternatively, you might decide to put the precompiled header file in a
17078 directory and use @option{-I} to ensure that directory is searched
17079 before (or instead of) the directory containing the original header.
17080 Then, if you want to check that the precompiled header file is always
17081 used, you can put a file of the same name as the original header in this
17082 directory containing an @code{#error} command.
17084 This also works with @option{-include}. So yet another way to use
17085 precompiled headers, good for projects not designed with precompiled
17086 header files in mind, is to simply take most of the header files used by
17087 a project, include them from another header file, precompile that header
17088 file, and @option{-include} the precompiled header. If the header files
17089 have guards against multiple inclusion, they will be skipped because
17090 they've already been included (in the precompiled header).
17092 If you need to precompile the same header file for different
17093 languages, targets, or compiler options, you can instead make a
17094 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17095 header in the directory, perhaps using @option{-o}. It doesn't matter
17096 what you call the files in the directory, every precompiled header in
17097 the directory will be considered. The first precompiled header
17098 encountered in the directory that is valid for this compilation will
17099 be used; they're searched in no particular order.
17101 There are many other possibilities, limited only by your imagination,
17102 good sense, and the constraints of your build system.
17104 A precompiled header file can be used only when these conditions apply:
17108 Only one precompiled header can be used in a particular compilation.
17111 A precompiled header can't be used once the first C token is seen. You
17112 can have preprocessor directives before a precompiled header; you can
17113 even include a precompiled header from inside another header, so long as
17114 there are no C tokens before the @code{#include}.
17117 The precompiled header file must be produced for the same language as
17118 the current compilation. You can't use a C precompiled header for a C++
17122 The precompiled header file must have been produced by the same compiler
17123 binary as the current compilation is using.
17126 Any macros defined before the precompiled header is included must
17127 either be defined in the same way as when the precompiled header was
17128 generated, or must not affect the precompiled header, which usually
17129 means that they don't appear in the precompiled header at all.
17131 The @option{-D} option is one way to define a macro before a
17132 precompiled header is included; using a @code{#define} can also do it.
17133 There are also some options that define macros implicitly, like
17134 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17137 @item If debugging information is output when using the precompiled
17138 header, using @option{-g} or similar, the same kind of debugging information
17139 must have been output when building the precompiled header. However,
17140 a precompiled header built using @option{-g} can be used in a compilation
17141 when no debugging information is being output.
17143 @item The same @option{-m} options must generally be used when building
17144 and using the precompiled header. @xref{Submodel Options},
17145 for any cases where this rule is relaxed.
17147 @item Each of the following options must be the same when building and using
17148 the precompiled header:
17150 @gccoptlist{-fexceptions}
17153 Some other command-line options starting with @option{-f},
17154 @option{-p}, or @option{-O} must be defined in the same way as when
17155 the precompiled header was generated. At present, it's not clear
17156 which options are safe to change and which are not; the safest choice
17157 is to use exactly the same options when generating and using the
17158 precompiled header. The following are known to be safe:
17160 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17161 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17162 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17167 For all of these except the last, the compiler will automatically
17168 ignore the precompiled header if the conditions aren't met. If you
17169 find an option combination that doesn't work and doesn't cause the
17170 precompiled header to be ignored, please consider filing a bug report,
17173 If you do use differing options when generating and using the
17174 precompiled header, the actual behavior will be a mixture of the
17175 behavior for the options. For instance, if you use @option{-g} to
17176 generate the precompiled header but not when using it, you may or may
17177 not get debugging information for routines in the precompiled header.