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 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
315 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
316 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
317 -fdebug-prefix-map=@var{old}=@var{new} @gol
318 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
319 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
320 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
321 -print-multi-directory -print-multi-lib @gol
322 -print-prog-name=@var{program} -print-search-dirs -Q @gol
323 -print-sysroot -print-sysroot-headers-suffix @gol
324 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
326 @item Optimization Options
327 @xref{Optimize Options,,Options that Control Optimization}.
329 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
330 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
331 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
332 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
333 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
334 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
335 -fdata-sections -fdce -fdce @gol
336 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
337 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
338 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
339 -fforward-propagate -ffunction-sections @gol
340 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
341 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
342 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
343 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
344 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
345 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
346 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
347 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
348 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
349 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
350 -floop-parallelize-all @gol
351 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
352 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
353 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
354 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
355 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
356 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
357 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
358 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
359 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
360 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
361 -fprofile-generate=@var{path} @gol
362 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
363 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
364 -freorder-blocks-and-partition -freorder-functions @gol
365 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
366 -frounding-math -fsched2-use-superblocks @gol
367 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
368 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
369 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
370 -fsched-spec-insn-heuristic -fsched-reg-pressure-heuristic @gol
371 -fsched-rank-heuristic -fsched-last-insn-heuristic @gol
372 -fsched-dep-count-heuristic @gol
373 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
374 -fselective-scheduling -fselective-scheduling2 @gol
375 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
376 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
377 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
378 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
379 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
380 -ftree-copyrename -ftree-dce @gol
381 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
382 -ftree-phiprop -ftree-loop-distribution @gol
383 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
384 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
385 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
386 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
387 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
388 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
389 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
391 --param @var{name}=@var{value}
392 -O -O0 -O1 -O2 -O3 -Os}
394 @item Preprocessor Options
395 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
396 @gccoptlist{-A@var{question}=@var{answer} @gol
397 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
398 -C -dD -dI -dM -dN @gol
399 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
400 -idirafter @var{dir} @gol
401 -include @var{file} -imacros @var{file} @gol
402 -iprefix @var{file} -iwithprefix @var{dir} @gol
403 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
404 -imultilib @var{dir} -isysroot @var{dir} @gol
405 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
406 -P -fworking-directory -remap @gol
407 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
408 -Xpreprocessor @var{option}}
410 @item Assembler Option
411 @xref{Assembler Options,,Passing Options to the Assembler}.
412 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
415 @xref{Link Options,,Options for Linking}.
416 @gccoptlist{@var{object-file-name} -l@var{library} @gol
417 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
418 -s -static -static-libgcc -static-libstdc++ -shared @gol
419 -shared-libgcc -symbolic @gol
420 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
423 @item Directory Options
424 @xref{Directory Options,,Options for Directory Search}.
425 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
426 -specs=@var{file} -I- --sysroot=@var{dir}}
429 @c I wrote this xref this way to avoid overfull hbox. -- rms
430 @xref{Target Options}.
431 @gccoptlist{-V @var{version} -b @var{machine}}
433 @item Machine Dependent Options
434 @xref{Submodel Options,,Hardware Models and Configurations}.
435 @c This list is ordered alphanumerically by subsection name.
436 @c Try and put the significant identifier (CPU or system) first,
437 @c so users have a clue at guessing where the ones they want will be.
440 @gccoptlist{-EB -EL @gol
441 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
442 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
445 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
446 -mabi=@var{name} @gol
447 -mapcs-stack-check -mno-apcs-stack-check @gol
448 -mapcs-float -mno-apcs-float @gol
449 -mapcs-reentrant -mno-apcs-reentrant @gol
450 -msched-prolog -mno-sched-prolog @gol
451 -mlittle-endian -mbig-endian -mwords-little-endian @gol
452 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
453 -mfp16-format=@var{name}
454 -mthumb-interwork -mno-thumb-interwork @gol
455 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
456 -mstructure-size-boundary=@var{n} @gol
457 -mabort-on-noreturn @gol
458 -mlong-calls -mno-long-calls @gol
459 -msingle-pic-base -mno-single-pic-base @gol
460 -mpic-register=@var{reg} @gol
461 -mnop-fun-dllimport @gol
462 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
463 -mpoke-function-name @gol
465 -mtpcs-frame -mtpcs-leaf-frame @gol
466 -mcaller-super-interworking -mcallee-super-interworking @gol
468 -mword-relocations @gol
469 -mfix-cortex-m3-ldrd}
472 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
473 -mcall-prologues -mtiny-stack -mint8}
475 @emph{Blackfin Options}
476 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
477 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
478 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
479 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
480 -mno-id-shared-library -mshared-library-id=@var{n} @gol
481 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
482 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
483 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
487 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
488 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
489 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
490 -mstack-align -mdata-align -mconst-align @gol
491 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
492 -melf -maout -melinux -mlinux -sim -sim2 @gol
493 -mmul-bug-workaround -mno-mul-bug-workaround}
496 @gccoptlist{-mmac -mpush-args}
498 @emph{Darwin Options}
499 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
500 -arch_only -bind_at_load -bundle -bundle_loader @gol
501 -client_name -compatibility_version -current_version @gol
503 -dependency-file -dylib_file -dylinker_install_name @gol
504 -dynamic -dynamiclib -exported_symbols_list @gol
505 -filelist -flat_namespace -force_cpusubtype_ALL @gol
506 -force_flat_namespace -headerpad_max_install_names @gol
508 -image_base -init -install_name -keep_private_externs @gol
509 -multi_module -multiply_defined -multiply_defined_unused @gol
510 -noall_load -no_dead_strip_inits_and_terms @gol
511 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
512 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
513 -private_bundle -read_only_relocs -sectalign @gol
514 -sectobjectsymbols -whyload -seg1addr @gol
515 -sectcreate -sectobjectsymbols -sectorder @gol
516 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
517 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
518 -segprot -segs_read_only_addr -segs_read_write_addr @gol
519 -single_module -static -sub_library -sub_umbrella @gol
520 -twolevel_namespace -umbrella -undefined @gol
521 -unexported_symbols_list -weak_reference_mismatches @gol
522 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
523 -mkernel -mone-byte-bool}
525 @emph{DEC Alpha Options}
526 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
527 -mieee -mieee-with-inexact -mieee-conformant @gol
528 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
529 -mtrap-precision=@var{mode} -mbuild-constants @gol
530 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
531 -mbwx -mmax -mfix -mcix @gol
532 -mfloat-vax -mfloat-ieee @gol
533 -mexplicit-relocs -msmall-data -mlarge-data @gol
534 -msmall-text -mlarge-text @gol
535 -mmemory-latency=@var{time}}
537 @emph{DEC Alpha/VMS Options}
538 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
541 @gccoptlist{-msmall-model -mno-lsim}
544 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
545 -mhard-float -msoft-float @gol
546 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
547 -mdouble -mno-double @gol
548 -mmedia -mno-media -mmuladd -mno-muladd @gol
549 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
550 -mlinked-fp -mlong-calls -malign-labels @gol
551 -mlibrary-pic -macc-4 -macc-8 @gol
552 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
553 -moptimize-membar -mno-optimize-membar @gol
554 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
555 -mvliw-branch -mno-vliw-branch @gol
556 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
557 -mno-nested-cond-exec -mtomcat-stats @gol
561 @emph{GNU/Linux Options}
562 @gccoptlist{-muclibc}
564 @emph{H8/300 Options}
565 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
568 @gccoptlist{-march=@var{architecture-type} @gol
569 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
570 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
571 -mfixed-range=@var{register-range} @gol
572 -mjump-in-delay -mlinker-opt -mlong-calls @gol
573 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
574 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
575 -mno-jump-in-delay -mno-long-load-store @gol
576 -mno-portable-runtime -mno-soft-float @gol
577 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
578 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
579 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
580 -munix=@var{unix-std} -nolibdld -static -threads}
582 @emph{i386 and x86-64 Options}
583 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
584 -mfpmath=@var{unit} @gol
585 -masm=@var{dialect} -mno-fancy-math-387 @gol
586 -mno-fp-ret-in-387 -msoft-float @gol
587 -mno-wide-multiply -mrtd -malign-double @gol
588 -mpreferred-stack-boundary=@var{num}
589 -mincoming-stack-boundary=@var{num}
590 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
591 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
593 -msse4a -m3dnow -mpopcnt -mabm @gol
594 -mthreads -mno-align-stringops -minline-all-stringops @gol
595 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
596 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
597 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
598 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
599 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
600 -mcmodel=@var{code-model} -mabi=@var{name} @gol
601 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
605 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
606 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
607 -mconstant-gp -mauto-pic -mfused-madd @gol
608 -minline-float-divide-min-latency @gol
609 -minline-float-divide-max-throughput @gol
610 -mno-inline-float-divide @gol
611 -minline-int-divide-min-latency @gol
612 -minline-int-divide-max-throughput @gol
613 -mno-inline-int-divide @gol
614 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
615 -mno-inline-sqrt @gol
616 -mdwarf2-asm -mearly-stop-bits @gol
617 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
618 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
619 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
620 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
621 -msched-spec-ldc -msched-spec-control-ldc @gol
622 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
623 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
624 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
625 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
627 @emph{IA-64/VMS Options}
628 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
630 @emph{M32R/D Options}
631 @gccoptlist{-m32r2 -m32rx -m32r @gol
633 -malign-loops -mno-align-loops @gol
634 -missue-rate=@var{number} @gol
635 -mbranch-cost=@var{number} @gol
636 -mmodel=@var{code-size-model-type} @gol
637 -msdata=@var{sdata-type} @gol
638 -mno-flush-func -mflush-func=@var{name} @gol
639 -mno-flush-trap -mflush-trap=@var{number} @gol
643 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
645 @emph{M680x0 Options}
646 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
647 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
648 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
649 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
650 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
651 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
652 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
653 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
656 @emph{M68hc1x Options}
657 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
658 -mauto-incdec -minmax -mlong-calls -mshort @gol
659 -msoft-reg-count=@var{count}}
662 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
663 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
664 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
665 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
666 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
669 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
670 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
671 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
672 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
676 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
677 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
678 -mips64 -mips64r2 @gol
679 -mips16 -mno-mips16 -mflip-mips16 @gol
680 -minterlink-mips16 -mno-interlink-mips16 @gol
681 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
682 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
683 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
684 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
685 -mfpu=@var{fpu-type} @gol
686 -msmartmips -mno-smartmips @gol
687 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
688 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
689 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
690 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
691 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
692 -membedded-data -mno-embedded-data @gol
693 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
694 -mcode-readable=@var{setting} @gol
695 -msplit-addresses -mno-split-addresses @gol
696 -mexplicit-relocs -mno-explicit-relocs @gol
697 -mcheck-zero-division -mno-check-zero-division @gol
698 -mdivide-traps -mdivide-breaks @gol
699 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
700 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
701 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
702 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
703 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
704 -mflush-func=@var{func} -mno-flush-func @gol
705 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
706 -mfp-exceptions -mno-fp-exceptions @gol
707 -mvr4130-align -mno-vr4130-align -msynci -mno-synci}
710 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
711 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
712 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
713 -mno-base-addresses -msingle-exit -mno-single-exit}
715 @emph{MN10300 Options}
716 @gccoptlist{-mmult-bug -mno-mult-bug @gol
717 -mam33 -mno-am33 @gol
718 -mam33-2 -mno-am33-2 @gol
719 -mreturn-pointer-on-d0 @gol
722 @emph{PDP-11 Options}
723 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
724 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
725 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
726 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
727 -mbranch-expensive -mbranch-cheap @gol
728 -msplit -mno-split -munix-asm -mdec-asm}
730 @emph{picoChip Options}
731 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
732 -msymbol-as-address -mno-inefficient-warnings}
734 @emph{PowerPC Options}
735 See RS/6000 and PowerPC Options.
737 @emph{RS/6000 and PowerPC Options}
738 @gccoptlist{-mcpu=@var{cpu-type} @gol
739 -mtune=@var{cpu-type} @gol
740 -mpower -mno-power -mpower2 -mno-power2 @gol
741 -mpowerpc -mpowerpc64 -mno-powerpc @gol
742 -maltivec -mno-altivec @gol
743 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
744 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
745 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
746 -mfprnd -mno-fprnd @gol
747 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
748 -mnew-mnemonics -mold-mnemonics @gol
749 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
750 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
751 -malign-power -malign-natural @gol
752 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
753 -msingle-float -mdouble-float -msimple-fpu @gol
754 -mstring -mno-string -mupdate -mno-update @gol
755 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
756 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
757 -mstrict-align -mno-strict-align -mrelocatable @gol
758 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
759 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
760 -mdynamic-no-pic -maltivec -mswdiv @gol
761 -mprioritize-restricted-insns=@var{priority} @gol
762 -msched-costly-dep=@var{dependence_type} @gol
763 -minsert-sched-nops=@var{scheme} @gol
764 -mcall-sysv -mcall-netbsd @gol
765 -maix-struct-return -msvr4-struct-return @gol
766 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
767 -misel -mno-isel @gol
768 -misel=yes -misel=no @gol
770 -mspe=yes -mspe=no @gol
772 -mgen-cell-microcode -mwarn-cell-microcode @gol
773 -mvrsave -mno-vrsave @gol
774 -mmulhw -mno-mulhw @gol
775 -mdlmzb -mno-dlmzb @gol
776 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
777 -mprototype -mno-prototype @gol
778 -msim -mmvme -mads -myellowknife -memb -msdata @gol
779 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
781 @emph{S/390 and zSeries Options}
782 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
783 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
784 -mlong-double-64 -mlong-double-128 @gol
785 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
786 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
787 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
788 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
789 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
792 @gccoptlist{-meb -mel @gol
796 -mscore5 -mscore5u -mscore7 -mscore7d}
799 @gccoptlist{-m1 -m2 -m2e @gol
800 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
802 -m4-nofpu -m4-single-only -m4-single -m4 @gol
803 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
804 -m5-64media -m5-64media-nofpu @gol
805 -m5-32media -m5-32media-nofpu @gol
806 -m5-compact -m5-compact-nofpu @gol
807 -mb -ml -mdalign -mrelax @gol
808 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
809 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
810 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
811 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
812 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
816 @gccoptlist{-mcpu=@var{cpu-type} @gol
817 -mtune=@var{cpu-type} @gol
818 -mcmodel=@var{code-model} @gol
819 -m32 -m64 -mapp-regs -mno-app-regs @gol
820 -mfaster-structs -mno-faster-structs @gol
821 -mfpu -mno-fpu -mhard-float -msoft-float @gol
822 -mhard-quad-float -msoft-quad-float @gol
823 -mimpure-text -mno-impure-text -mlittle-endian @gol
824 -mstack-bias -mno-stack-bias @gol
825 -munaligned-doubles -mno-unaligned-doubles @gol
826 -mv8plus -mno-v8plus -mvis -mno-vis
827 -threads -pthreads -pthread}
830 @gccoptlist{-mwarn-reloc -merror-reloc @gol
831 -msafe-dma -munsafe-dma @gol
833 -msmall-mem -mlarge-mem -mstdmain @gol
834 -mfixed-range=@var{register-range}}
836 @emph{System V Options}
837 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
840 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
841 -mprolog-function -mno-prolog-function -mspace @gol
842 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
843 -mapp-regs -mno-app-regs @gol
844 -mdisable-callt -mno-disable-callt @gol
850 @gccoptlist{-mg -mgnu -munix}
852 @emph{VxWorks Options}
853 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
854 -Xbind-lazy -Xbind-now}
856 @emph{x86-64 Options}
857 See i386 and x86-64 Options.
859 @emph{i386 and x86-64 Windows Options}
860 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
861 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
863 @emph{Xstormy16 Options}
866 @emph{Xtensa Options}
867 @gccoptlist{-mconst16 -mno-const16 @gol
868 -mfused-madd -mno-fused-madd @gol
869 -mserialize-volatile -mno-serialize-volatile @gol
870 -mtext-section-literals -mno-text-section-literals @gol
871 -mtarget-align -mno-target-align @gol
872 -mlongcalls -mno-longcalls}
874 @emph{zSeries Options}
875 See S/390 and zSeries Options.
877 @item Code Generation Options
878 @xref{Code Gen Options,,Options for Code Generation Conventions}.
879 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
880 -ffixed-@var{reg} -fexceptions @gol
881 -fnon-call-exceptions -funwind-tables @gol
882 -fasynchronous-unwind-tables @gol
883 -finhibit-size-directive -finstrument-functions @gol
884 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
885 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
886 -fno-common -fno-ident @gol
887 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
888 -fno-jump-tables @gol
889 -frecord-gcc-switches @gol
890 -freg-struct-return -fshort-enums @gol
891 -fshort-double -fshort-wchar @gol
892 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
893 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
894 -fno-stack-limit -fargument-alias -fargument-noalias @gol
895 -fargument-noalias-global -fargument-noalias-anything @gol
896 -fleading-underscore -ftls-model=@var{model} @gol
897 -ftrapv -fwrapv -fbounds-check @gol
902 * Overall Options:: Controlling the kind of output:
903 an executable, object files, assembler files,
904 or preprocessed source.
905 * C Dialect Options:: Controlling the variant of C language compiled.
906 * C++ Dialect Options:: Variations on C++.
907 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
909 * Language Independent Options:: Controlling how diagnostics should be
911 * Warning Options:: How picky should the compiler be?
912 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
913 * Optimize Options:: How much optimization?
914 * Preprocessor Options:: Controlling header files and macro definitions.
915 Also, getting dependency information for Make.
916 * Assembler Options:: Passing options to the assembler.
917 * Link Options:: Specifying libraries and so on.
918 * Directory Options:: Where to find header files and libraries.
919 Where to find the compiler executable files.
920 * Spec Files:: How to pass switches to sub-processes.
921 * Target Options:: Running a cross-compiler, or an old version of GCC.
924 @node Overall Options
925 @section Options Controlling the Kind of Output
927 Compilation can involve up to four stages: preprocessing, compilation
928 proper, assembly and linking, always in that order. GCC is capable of
929 preprocessing and compiling several files either into several
930 assembler input files, or into one assembler input file; then each
931 assembler input file produces an object file, and linking combines all
932 the object files (those newly compiled, and those specified as input)
933 into an executable file.
935 @cindex file name suffix
936 For any given input file, the file name suffix determines what kind of
941 C source code which must be preprocessed.
944 C source code which should not be preprocessed.
947 C++ source code which should not be preprocessed.
950 Objective-C source code. Note that you must link with the @file{libobjc}
951 library to make an Objective-C program work.
954 Objective-C source code which should not be preprocessed.
958 Objective-C++ source code. Note that you must link with the @file{libobjc}
959 library to make an Objective-C++ program work. Note that @samp{.M} refers
960 to a literal capital M@.
963 Objective-C++ source code which should not be preprocessed.
966 C, C++, Objective-C or Objective-C++ header file to be turned into a
971 @itemx @var{file}.cxx
972 @itemx @var{file}.cpp
973 @itemx @var{file}.CPP
974 @itemx @var{file}.c++
976 C++ source code which must be preprocessed. Note that in @samp{.cxx},
977 the last two letters must both be literally @samp{x}. Likewise,
978 @samp{.C} refers to a literal capital C@.
982 Objective-C++ source code which must be preprocessed.
985 Objective-C++ source code which should not be preprocessed.
990 @itemx @var{file}.hxx
991 @itemx @var{file}.hpp
992 @itemx @var{file}.HPP
993 @itemx @var{file}.h++
994 @itemx @var{file}.tcc
995 C++ header file to be turned into a precompiled header.
998 @itemx @var{file}.for
999 @itemx @var{file}.ftn
1000 Fixed form Fortran source code which should not be preprocessed.
1003 @itemx @var{file}.FOR
1004 @itemx @var{file}.fpp
1005 @itemx @var{file}.FPP
1006 @itemx @var{file}.FTN
1007 Fixed form Fortran source code which must be preprocessed (with the traditional
1010 @item @var{file}.f90
1011 @itemx @var{file}.f95
1012 @itemx @var{file}.f03
1013 @itemx @var{file}.f08
1014 Free form Fortran source code which should not be preprocessed.
1016 @item @var{file}.F90
1017 @itemx @var{file}.F95
1018 @itemx @var{file}.F03
1019 @itemx @var{file}.F08
1020 Free form Fortran source code which must be preprocessed (with the
1021 traditional preprocessor).
1023 @c FIXME: Descriptions of Java file types.
1029 @item @var{file}.ads
1030 Ada source code file which contains a library unit declaration (a
1031 declaration of a package, subprogram, or generic, or a generic
1032 instantiation), or a library unit renaming declaration (a package,
1033 generic, or subprogram renaming declaration). Such files are also
1036 @item @var{file}.adb
1037 Ada source code file containing a library unit body (a subprogram or
1038 package body). Such files are also called @dfn{bodies}.
1040 @c GCC also knows about some suffixes for languages not yet included:
1051 @itemx @var{file}.sx
1052 Assembler code which must be preprocessed.
1055 An object file to be fed straight into linking.
1056 Any file name with no recognized suffix is treated this way.
1060 You can specify the input language explicitly with the @option{-x} option:
1063 @item -x @var{language}
1064 Specify explicitly the @var{language} for the following input files
1065 (rather than letting the compiler choose a default based on the file
1066 name suffix). This option applies to all following input files until
1067 the next @option{-x} option. Possible values for @var{language} are:
1069 c c-header c-cpp-output
1070 c++ c++-header c++-cpp-output
1071 objective-c objective-c-header objective-c-cpp-output
1072 objective-c++ objective-c++-header objective-c++-cpp-output
1073 assembler assembler-with-cpp
1075 f77 f77-cpp-input f95 f95-cpp-input
1080 Turn off any specification of a language, so that subsequent files are
1081 handled according to their file name suffixes (as they are if @option{-x}
1082 has not been used at all).
1084 @item -pass-exit-codes
1085 @opindex pass-exit-codes
1086 Normally the @command{gcc} program will exit with the code of 1 if any
1087 phase of the compiler returns a non-success return code. If you specify
1088 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1089 numerically highest error produced by any phase that returned an error
1090 indication. The C, C++, and Fortran frontends return 4, if an internal
1091 compiler error is encountered.
1094 If you only want some of the stages of compilation, you can use
1095 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1096 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1097 @command{gcc} is to stop. Note that some combinations (for example,
1098 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1103 Compile or assemble the source files, but do not link. The linking
1104 stage simply is not done. The ultimate output is in the form of an
1105 object file for each source file.
1107 By default, the object file name for a source file is made by replacing
1108 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1110 Unrecognized input files, not requiring compilation or assembly, are
1115 Stop after the stage of compilation proper; do not assemble. The output
1116 is in the form of an assembler code file for each non-assembler input
1119 By default, the assembler file name for a source file is made by
1120 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1122 Input files that don't require compilation are ignored.
1126 Stop after the preprocessing stage; do not run the compiler proper. The
1127 output is in the form of preprocessed source code, which is sent to the
1130 Input files which don't require preprocessing are ignored.
1132 @cindex output file option
1135 Place output in file @var{file}. This applies regardless to whatever
1136 sort of output is being produced, whether it be an executable file,
1137 an object file, an assembler file or preprocessed C code.
1139 If @option{-o} is not specified, the default is to put an executable
1140 file in @file{a.out}, the object file for
1141 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1142 assembler file in @file{@var{source}.s}, a precompiled header file in
1143 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1148 Print (on standard error output) the commands executed to run the stages
1149 of compilation. Also print the version number of the compiler driver
1150 program and of the preprocessor and the compiler proper.
1154 Like @option{-v} except the commands are not executed and all command
1155 arguments are quoted. This is useful for shell scripts to capture the
1156 driver-generated command lines.
1160 Use pipes rather than temporary files for communication between the
1161 various stages of compilation. This fails to work on some systems where
1162 the assembler is unable to read from a pipe; but the GNU assembler has
1167 If you are compiling multiple source files, this option tells the driver
1168 to pass all the source files to the compiler at once (for those
1169 languages for which the compiler can handle this). This will allow
1170 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1171 language for which this is supported is C@. If you pass source files for
1172 multiple languages to the driver, using this option, the driver will invoke
1173 the compiler(s) that support IMA once each, passing each compiler all the
1174 source files appropriate for it. For those languages that do not support
1175 IMA this option will be ignored, and the compiler will be invoked once for
1176 each source file in that language. If you use this option in conjunction
1177 with @option{-save-temps}, the compiler will generate multiple
1179 (one for each source file), but only one (combined) @file{.o} or
1184 Print (on the standard output) a description of the command line options
1185 understood by @command{gcc}. If the @option{-v} option is also specified
1186 then @option{--help} will also be passed on to the various processes
1187 invoked by @command{gcc}, so that they can display the command line options
1188 they accept. If the @option{-Wextra} option has also been specified
1189 (prior to the @option{--help} option), then command line options which
1190 have no documentation associated with them will also be displayed.
1193 @opindex target-help
1194 Print (on the standard output) a description of target-specific command
1195 line options for each tool. For some targets extra target-specific
1196 information may also be printed.
1198 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1199 Print (on the standard output) a description of the command line
1200 options understood by the compiler that fit into all specified classes
1201 and qualifiers. These are the supported classes:
1204 @item @samp{optimizers}
1205 This will display all of the optimization options supported by the
1208 @item @samp{warnings}
1209 This will display all of the options controlling warning messages
1210 produced by the compiler.
1213 This will display target-specific options. Unlike the
1214 @option{--target-help} option however, target-specific options of the
1215 linker and assembler will not be displayed. This is because those
1216 tools do not currently support the extended @option{--help=} syntax.
1219 This will display the values recognized by the @option{--param}
1222 @item @var{language}
1223 This will display the options supported for @var{language}, where
1224 @var{language} is the name of one of the languages supported in this
1228 This will display the options that are common to all languages.
1231 These are the supported qualifiers:
1234 @item @samp{undocumented}
1235 Display only those options which are undocumented.
1238 Display options which take an argument that appears after an equal
1239 sign in the same continuous piece of text, such as:
1240 @samp{--help=target}.
1242 @item @samp{separate}
1243 Display options which take an argument that appears as a separate word
1244 following the original option, such as: @samp{-o output-file}.
1247 Thus for example to display all the undocumented target-specific
1248 switches supported by the compiler the following can be used:
1251 --help=target,undocumented
1254 The sense of a qualifier can be inverted by prefixing it with the
1255 @samp{^} character, so for example to display all binary warning
1256 options (i.e., ones that are either on or off and that do not take an
1257 argument), which have a description the following can be used:
1260 --help=warnings,^joined,^undocumented
1263 The argument to @option{--help=} should not consist solely of inverted
1266 Combining several classes is possible, although this usually
1267 restricts the output by so much that there is nothing to display. One
1268 case where it does work however is when one of the classes is
1269 @var{target}. So for example to display all the target-specific
1270 optimization options the following can be used:
1273 --help=target,optimizers
1276 The @option{--help=} option can be repeated on the command line. Each
1277 successive use will display its requested class of options, skipping
1278 those that have already been displayed.
1280 If the @option{-Q} option appears on the command line before the
1281 @option{--help=} option, then the descriptive text displayed by
1282 @option{--help=} is changed. Instead of describing the displayed
1283 options, an indication is given as to whether the option is enabled,
1284 disabled or set to a specific value (assuming that the compiler
1285 knows this at the point where the @option{--help=} option is used).
1287 Here is a truncated example from the ARM port of @command{gcc}:
1290 % gcc -Q -mabi=2 --help=target -c
1291 The following options are target specific:
1293 -mabort-on-noreturn [disabled]
1297 The output is sensitive to the effects of previous command line
1298 options, so for example it is possible to find out which optimizations
1299 are enabled at @option{-O2} by using:
1302 -Q -O2 --help=optimizers
1305 Alternatively you can discover which binary optimizations are enabled
1306 by @option{-O3} by using:
1309 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1310 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1311 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1314 @item -no-canonical-prefixes
1315 @opindex no-canonical-prefixes
1316 Do not expand any symbolic links, resolve references to @samp{/../}
1317 or @samp{/./}, or make the path absolute when generating a relative
1322 Display the version number and copyrights of the invoked GCC@.
1326 Invoke all subcommands under a wrapper program. It takes a single
1327 comma separated list as an argument, which will be used to invoke
1331 gcc -c t.c -wrapper gdb,--args
1334 This will invoke all subprograms of gcc under "gdb --args",
1335 thus cc1 invocation will be "gdb --args cc1 ...".
1337 @item -fplugin=@var{name}.so
1338 Load the plugin code in file @var{name}.so, assumed to be a
1339 shared object to be dlopen'd by the compiler. The base name of
1340 the shared object file is used to identify the plugin for the
1341 purposes of argument parsing (See
1342 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1343 Each plugin should define the callback functions specified in the
1346 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1347 Define an argument called @var{key} with a value of @var{value}
1348 for the plugin called @var{name}.
1350 @include @value{srcdir}/../libiberty/at-file.texi
1354 @section Compiling C++ Programs
1356 @cindex suffixes for C++ source
1357 @cindex C++ source file suffixes
1358 C++ source files conventionally use one of the suffixes @samp{.C},
1359 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1360 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1361 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1362 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1363 files with these names and compiles them as C++ programs even if you
1364 call the compiler the same way as for compiling C programs (usually
1365 with the name @command{gcc}).
1369 However, the use of @command{gcc} does not add the C++ library.
1370 @command{g++} is a program that calls GCC and treats @samp{.c},
1371 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1372 files unless @option{-x} is used, and automatically specifies linking
1373 against the C++ library. This program is also useful when
1374 precompiling a C header file with a @samp{.h} extension for use in C++
1375 compilations. On many systems, @command{g++} is also installed with
1376 the name @command{c++}.
1378 @cindex invoking @command{g++}
1379 When you compile C++ programs, you may specify many of the same
1380 command-line options that you use for compiling programs in any
1381 language; or command-line options meaningful for C and related
1382 languages; or options that are meaningful only for C++ programs.
1383 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1384 explanations of options for languages related to C@.
1385 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1386 explanations of options that are meaningful only for C++ programs.
1388 @node C Dialect Options
1389 @section Options Controlling C Dialect
1390 @cindex dialect options
1391 @cindex language dialect options
1392 @cindex options, dialect
1394 The following options control the dialect of C (or languages derived
1395 from C, such as C++, Objective-C and Objective-C++) that the compiler
1399 @cindex ANSI support
1403 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1404 equivalent to @samp{-std=c++98}.
1406 This turns off certain features of GCC that are incompatible with ISO
1407 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1408 such as the @code{asm} and @code{typeof} keywords, and
1409 predefined macros such as @code{unix} and @code{vax} that identify the
1410 type of system you are using. It also enables the undesirable and
1411 rarely used ISO trigraph feature. For the C compiler,
1412 it disables recognition of C++ style @samp{//} comments as well as
1413 the @code{inline} keyword.
1415 The alternate keywords @code{__asm__}, @code{__extension__},
1416 @code{__inline__} and @code{__typeof__} continue to work despite
1417 @option{-ansi}. You would not want to use them in an ISO C program, of
1418 course, but it is useful to put them in header files that might be included
1419 in compilations done with @option{-ansi}. Alternate predefined macros
1420 such as @code{__unix__} and @code{__vax__} are also available, with or
1421 without @option{-ansi}.
1423 The @option{-ansi} option does not cause non-ISO programs to be
1424 rejected gratuitously. For that, @option{-pedantic} is required in
1425 addition to @option{-ansi}. @xref{Warning Options}.
1427 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1428 option is used. Some header files may notice this macro and refrain
1429 from declaring certain functions or defining certain macros that the
1430 ISO standard doesn't call for; this is to avoid interfering with any
1431 programs that might use these names for other things.
1433 Functions that would normally be built in but do not have semantics
1434 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1435 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1436 built-in functions provided by GCC}, for details of the functions
1441 Determine the language standard. @xref{Standards,,Language Standards
1442 Supported by GCC}, for details of these standard versions. This option
1443 is currently only supported when compiling C or C++.
1445 The compiler can accept several base standards, such as @samp{c89} or
1446 @samp{c++98}, and GNU dialects of those standards, such as
1447 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1448 compiler will accept all programs following that standard and those
1449 using GNU extensions that do not contradict it. For example,
1450 @samp{-std=c89} turns off certain features of GCC that are
1451 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1452 keywords, but not other GNU extensions that do not have a meaning in
1453 ISO C90, such as omitting the middle term of a @code{?:}
1454 expression. On the other hand, by specifying a GNU dialect of a
1455 standard, all features the compiler support are enabled, even when
1456 those features change the meaning of the base standard and some
1457 strict-conforming programs may be rejected. The particular standard
1458 is used by @option{-pedantic} to identify which features are GNU
1459 extensions given that version of the standard. For example
1460 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1461 comments, while @samp{-std=gnu99 -pedantic} would not.
1463 A value for this option must be provided; possible values are
1468 Support all ISO C90 programs (certain GNU extensions that conflict
1469 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1471 @item iso9899:199409
1472 ISO C90 as modified in amendment 1.
1478 ISO C99. Note that this standard is not yet fully supported; see
1479 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1480 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1483 GNU dialect of ISO C90 (including some C99 features). This
1484 is the default for C code.
1488 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1489 this will become the default. The name @samp{gnu9x} is deprecated.
1492 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1496 GNU dialect of @option{-std=c++98}. This is the default for
1500 The working draft of the upcoming ISO C++0x standard. This option
1501 enables experimental features that are likely to be included in
1502 C++0x. The working draft is constantly changing, and any feature that is
1503 enabled by this flag may be removed from future versions of GCC if it is
1504 not part of the C++0x standard.
1507 GNU dialect of @option{-std=c++0x}. This option enables
1508 experimental features that may be removed in future versions of GCC.
1511 @item -fgnu89-inline
1512 @opindex fgnu89-inline
1513 The option @option{-fgnu89-inline} tells GCC to use the traditional
1514 GNU semantics for @code{inline} functions when in C99 mode.
1515 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1516 is accepted and ignored by GCC versions 4.1.3 up to but not including
1517 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1518 C99 mode. Using this option is roughly equivalent to adding the
1519 @code{gnu_inline} function attribute to all inline functions
1520 (@pxref{Function Attributes}).
1522 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1523 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1524 specifies the default behavior). This option was first supported in
1525 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1527 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1528 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1529 in effect for @code{inline} functions. @xref{Common Predefined
1530 Macros,,,cpp,The C Preprocessor}.
1532 @item -aux-info @var{filename}
1534 Output to the given filename prototyped declarations for all functions
1535 declared and/or defined in a translation unit, including those in header
1536 files. This option is silently ignored in any language other than C@.
1538 Besides declarations, the file indicates, in comments, the origin of
1539 each declaration (source file and line), whether the declaration was
1540 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1541 @samp{O} for old, respectively, in the first character after the line
1542 number and the colon), and whether it came from a declaration or a
1543 definition (@samp{C} or @samp{F}, respectively, in the following
1544 character). In the case of function definitions, a K&R-style list of
1545 arguments followed by their declarations is also provided, inside
1546 comments, after the declaration.
1550 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1551 keyword, so that code can use these words as identifiers. You can use
1552 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1553 instead. @option{-ansi} implies @option{-fno-asm}.
1555 In C++, this switch only affects the @code{typeof} keyword, since
1556 @code{asm} and @code{inline} are standard keywords. You may want to
1557 use the @option{-fno-gnu-keywords} flag instead, which has the same
1558 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1559 switch only affects the @code{asm} and @code{typeof} keywords, since
1560 @code{inline} is a standard keyword in ISO C99.
1563 @itemx -fno-builtin-@var{function}
1564 @opindex fno-builtin
1565 @cindex built-in functions
1566 Don't recognize built-in functions that do not begin with
1567 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1568 functions provided by GCC}, for details of the functions affected,
1569 including those which are not built-in functions when @option{-ansi} or
1570 @option{-std} options for strict ISO C conformance are used because they
1571 do not have an ISO standard meaning.
1573 GCC normally generates special code to handle certain built-in functions
1574 more efficiently; for instance, calls to @code{alloca} may become single
1575 instructions that adjust the stack directly, and calls to @code{memcpy}
1576 may become inline copy loops. The resulting code is often both smaller
1577 and faster, but since the function calls no longer appear as such, you
1578 cannot set a breakpoint on those calls, nor can you change the behavior
1579 of the functions by linking with a different library. In addition,
1580 when a function is recognized as a built-in function, GCC may use
1581 information about that function to warn about problems with calls to
1582 that function, or to generate more efficient code, even if the
1583 resulting code still contains calls to that function. For example,
1584 warnings are given with @option{-Wformat} for bad calls to
1585 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1586 known not to modify global memory.
1588 With the @option{-fno-builtin-@var{function}} option
1589 only the built-in function @var{function} is
1590 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1591 function is named that is not built-in in this version of GCC, this
1592 option is ignored. There is no corresponding
1593 @option{-fbuiltin-@var{function}} option; if you wish to enable
1594 built-in functions selectively when using @option{-fno-builtin} or
1595 @option{-ffreestanding}, you may define macros such as:
1598 #define abs(n) __builtin_abs ((n))
1599 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1604 @cindex hosted environment
1606 Assert that compilation takes place in a hosted environment. This implies
1607 @option{-fbuiltin}. A hosted environment is one in which the
1608 entire standard library is available, and in which @code{main} has a return
1609 type of @code{int}. Examples are nearly everything except a kernel.
1610 This is equivalent to @option{-fno-freestanding}.
1612 @item -ffreestanding
1613 @opindex ffreestanding
1614 @cindex hosted environment
1616 Assert that compilation takes place in a freestanding environment. This
1617 implies @option{-fno-builtin}. A freestanding environment
1618 is one in which the standard library may not exist, and program startup may
1619 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1620 This is equivalent to @option{-fno-hosted}.
1622 @xref{Standards,,Language Standards Supported by GCC}, for details of
1623 freestanding and hosted environments.
1627 @cindex openmp parallel
1628 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1629 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1630 compiler generates parallel code according to the OpenMP Application
1631 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1632 implies @option{-pthread}, and thus is only supported on targets that
1633 have support for @option{-pthread}.
1635 @item -fms-extensions
1636 @opindex fms-extensions
1637 Accept some non-standard constructs used in Microsoft header files.
1639 Some cases of unnamed fields in structures and unions are only
1640 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1641 fields within structs/unions}, for details.
1645 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1646 options for strict ISO C conformance) implies @option{-trigraphs}.
1648 @item -no-integrated-cpp
1649 @opindex no-integrated-cpp
1650 Performs a compilation in two passes: preprocessing and compiling. This
1651 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1652 @option{-B} option. The user supplied compilation step can then add in
1653 an additional preprocessing step after normal preprocessing but before
1654 compiling. The default is to use the integrated cpp (internal cpp)
1656 The semantics of this option will change if "cc1", "cc1plus", and
1657 "cc1obj" are merged.
1659 @cindex traditional C language
1660 @cindex C language, traditional
1662 @itemx -traditional-cpp
1663 @opindex traditional-cpp
1664 @opindex traditional
1665 Formerly, these options caused GCC to attempt to emulate a pre-standard
1666 C compiler. They are now only supported with the @option{-E} switch.
1667 The preprocessor continues to support a pre-standard mode. See the GNU
1668 CPP manual for details.
1670 @item -fcond-mismatch
1671 @opindex fcond-mismatch
1672 Allow conditional expressions with mismatched types in the second and
1673 third arguments. The value of such an expression is void. This option
1674 is not supported for C++.
1676 @item -flax-vector-conversions
1677 @opindex flax-vector-conversions
1678 Allow implicit conversions between vectors with differing numbers of
1679 elements and/or incompatible element types. This option should not be
1682 @item -funsigned-char
1683 @opindex funsigned-char
1684 Let the type @code{char} be unsigned, like @code{unsigned char}.
1686 Each kind of machine has a default for what @code{char} should
1687 be. It is either like @code{unsigned char} by default or like
1688 @code{signed char} by default.
1690 Ideally, a portable program should always use @code{signed char} or
1691 @code{unsigned char} when it depends on the signedness of an object.
1692 But many programs have been written to use plain @code{char} and
1693 expect it to be signed, or expect it to be unsigned, depending on the
1694 machines they were written for. This option, and its inverse, let you
1695 make such a program work with the opposite default.
1697 The type @code{char} is always a distinct type from each of
1698 @code{signed char} or @code{unsigned char}, even though its behavior
1699 is always just like one of those two.
1702 @opindex fsigned-char
1703 Let the type @code{char} be signed, like @code{signed char}.
1705 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1706 the negative form of @option{-funsigned-char}. Likewise, the option
1707 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1709 @item -fsigned-bitfields
1710 @itemx -funsigned-bitfields
1711 @itemx -fno-signed-bitfields
1712 @itemx -fno-unsigned-bitfields
1713 @opindex fsigned-bitfields
1714 @opindex funsigned-bitfields
1715 @opindex fno-signed-bitfields
1716 @opindex fno-unsigned-bitfields
1717 These options control whether a bit-field is signed or unsigned, when the
1718 declaration does not use either @code{signed} or @code{unsigned}. By
1719 default, such a bit-field is signed, because this is consistent: the
1720 basic integer types such as @code{int} are signed types.
1723 @node C++ Dialect Options
1724 @section Options Controlling C++ Dialect
1726 @cindex compiler options, C++
1727 @cindex C++ options, command line
1728 @cindex options, C++
1729 This section describes the command-line options that are only meaningful
1730 for C++ programs; but you can also use most of the GNU compiler options
1731 regardless of what language your program is in. For example, you
1732 might compile a file @code{firstClass.C} like this:
1735 g++ -g -frepo -O -c firstClass.C
1739 In this example, only @option{-frepo} is an option meant
1740 only for C++ programs; you can use the other options with any
1741 language supported by GCC@.
1743 Here is a list of options that are @emph{only} for compiling C++ programs:
1747 @item -fabi-version=@var{n}
1748 @opindex fabi-version
1749 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1750 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1751 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1752 the version that conforms most closely to the C++ ABI specification.
1753 Therefore, the ABI obtained using version 0 will change as ABI bugs
1756 The default is version 2.
1758 @item -fno-access-control
1759 @opindex fno-access-control
1760 Turn off all access checking. This switch is mainly useful for working
1761 around bugs in the access control code.
1765 Check that the pointer returned by @code{operator new} is non-null
1766 before attempting to modify the storage allocated. This check is
1767 normally unnecessary because the C++ standard specifies that
1768 @code{operator new} will only return @code{0} if it is declared
1769 @samp{throw()}, in which case the compiler will always check the
1770 return value even without this option. In all other cases, when
1771 @code{operator new} has a non-empty exception specification, memory
1772 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1773 @samp{new (nothrow)}.
1775 @item -fconserve-space
1776 @opindex fconserve-space
1777 Put uninitialized or runtime-initialized global variables into the
1778 common segment, as C does. This saves space in the executable at the
1779 cost of not diagnosing duplicate definitions. If you compile with this
1780 flag and your program mysteriously crashes after @code{main()} has
1781 completed, you may have an object that is being destroyed twice because
1782 two definitions were merged.
1784 This option is no longer useful on most targets, now that support has
1785 been added for putting variables into BSS without making them common.
1787 @item -ffriend-injection
1788 @opindex ffriend-injection
1789 Inject friend functions into the enclosing namespace, so that they are
1790 visible outside the scope of the class in which they are declared.
1791 Friend functions were documented to work this way in the old Annotated
1792 C++ Reference Manual, and versions of G++ before 4.1 always worked
1793 that way. However, in ISO C++ a friend function which is not declared
1794 in an enclosing scope can only be found using argument dependent
1795 lookup. This option causes friends to be injected as they were in
1798 This option is for compatibility, and may be removed in a future
1801 @item -fno-elide-constructors
1802 @opindex fno-elide-constructors
1803 The C++ standard allows an implementation to omit creating a temporary
1804 which is only used to initialize another object of the same type.
1805 Specifying this option disables that optimization, and forces G++ to
1806 call the copy constructor in all cases.
1808 @item -fno-enforce-eh-specs
1809 @opindex fno-enforce-eh-specs
1810 Don't generate code to check for violation of exception specifications
1811 at runtime. This option violates the C++ standard, but may be useful
1812 for reducing code size in production builds, much like defining
1813 @samp{NDEBUG}. This does not give user code permission to throw
1814 exceptions in violation of the exception specifications; the compiler
1815 will still optimize based on the specifications, so throwing an
1816 unexpected exception will result in undefined behavior.
1819 @itemx -fno-for-scope
1821 @opindex fno-for-scope
1822 If @option{-ffor-scope} is specified, the scope of variables declared in
1823 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1824 as specified by the C++ standard.
1825 If @option{-fno-for-scope} is specified, the scope of variables declared in
1826 a @i{for-init-statement} extends to the end of the enclosing scope,
1827 as was the case in old versions of G++, and other (traditional)
1828 implementations of C++.
1830 The default if neither flag is given to follow the standard,
1831 but to allow and give a warning for old-style code that would
1832 otherwise be invalid, or have different behavior.
1834 @item -fno-gnu-keywords
1835 @opindex fno-gnu-keywords
1836 Do not recognize @code{typeof} as a keyword, so that code can use this
1837 word as an identifier. You can use the keyword @code{__typeof__} instead.
1838 @option{-ansi} implies @option{-fno-gnu-keywords}.
1840 @item -fno-implicit-templates
1841 @opindex fno-implicit-templates
1842 Never emit code for non-inline templates which are instantiated
1843 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1844 @xref{Template Instantiation}, for more information.
1846 @item -fno-implicit-inline-templates
1847 @opindex fno-implicit-inline-templates
1848 Don't emit code for implicit instantiations of inline templates, either.
1849 The default is to handle inlines differently so that compiles with and
1850 without optimization will need the same set of explicit instantiations.
1852 @item -fno-implement-inlines
1853 @opindex fno-implement-inlines
1854 To save space, do not emit out-of-line copies of inline functions
1855 controlled by @samp{#pragma implementation}. This will cause linker
1856 errors if these functions are not inlined everywhere they are called.
1858 @item -fms-extensions
1859 @opindex fms-extensions
1860 Disable pedantic warnings about constructs used in MFC, such as implicit
1861 int and getting a pointer to member function via non-standard syntax.
1863 @item -fno-nonansi-builtins
1864 @opindex fno-nonansi-builtins
1865 Disable built-in declarations of functions that are not mandated by
1866 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1867 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1869 @item -fno-operator-names
1870 @opindex fno-operator-names
1871 Do not treat the operator name keywords @code{and}, @code{bitand},
1872 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1873 synonyms as keywords.
1875 @item -fno-optional-diags
1876 @opindex fno-optional-diags
1877 Disable diagnostics that the standard says a compiler does not need to
1878 issue. Currently, the only such diagnostic issued by G++ is the one for
1879 a name having multiple meanings within a class.
1882 @opindex fpermissive
1883 Downgrade some diagnostics about nonconformant code from errors to
1884 warnings. Thus, using @option{-fpermissive} will allow some
1885 nonconforming code to compile.
1887 @item -fno-pretty-templates
1888 @opindex fno-pretty-templates
1889 When an error message refers to a specialization of a function
1890 template, the compiler will normally print the signature of the
1891 template followed by the template arguments and any typedefs or
1892 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1893 rather than @code{void f(int)}) so that it's clear which template is
1894 involved. When an error message refers to a specialization of a class
1895 template, the compiler will omit any template arguments which match
1896 the default template arguments for that template. If either of these
1897 behaviors make it harder to understand the error message rather than
1898 easier, using @option{-fno-pretty-templates} will disable them.
1902 Enable automatic template instantiation at link time. This option also
1903 implies @option{-fno-implicit-templates}. @xref{Template
1904 Instantiation}, for more information.
1908 Disable generation of information about every class with virtual
1909 functions for use by the C++ runtime type identification features
1910 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1911 of the language, you can save some space by using this flag. Note that
1912 exception handling uses the same information, but it will generate it as
1913 needed. The @samp{dynamic_cast} operator can still be used for casts that
1914 do not require runtime type information, i.e.@: casts to @code{void *} or to
1915 unambiguous base classes.
1919 Emit statistics about front-end processing at the end of the compilation.
1920 This information is generally only useful to the G++ development team.
1922 @item -ftemplate-depth-@var{n}
1923 @opindex ftemplate-depth
1924 Set the maximum instantiation depth for template classes to @var{n}.
1925 A limit on the template instantiation depth is needed to detect
1926 endless recursions during template class instantiation. ANSI/ISO C++
1927 conforming programs must not rely on a maximum depth greater than 17
1928 (changed to 1024 in C++0x).
1930 @item -fno-threadsafe-statics
1931 @opindex fno-threadsafe-statics
1932 Do not emit the extra code to use the routines specified in the C++
1933 ABI for thread-safe initialization of local statics. You can use this
1934 option to reduce code size slightly in code that doesn't need to be
1937 @item -fuse-cxa-atexit
1938 @opindex fuse-cxa-atexit
1939 Register destructors for objects with static storage duration with the
1940 @code{__cxa_atexit} function rather than the @code{atexit} function.
1941 This option is required for fully standards-compliant handling of static
1942 destructors, but will only work if your C library supports
1943 @code{__cxa_atexit}.
1945 @item -fno-use-cxa-get-exception-ptr
1946 @opindex fno-use-cxa-get-exception-ptr
1947 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1948 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1949 if the runtime routine is not available.
1951 @item -fvisibility-inlines-hidden
1952 @opindex fvisibility-inlines-hidden
1953 This switch declares that the user does not attempt to compare
1954 pointers to inline methods where the addresses of the two functions
1955 were taken in different shared objects.
1957 The effect of this is that GCC may, effectively, mark inline methods with
1958 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1959 appear in the export table of a DSO and do not require a PLT indirection
1960 when used within the DSO@. Enabling this option can have a dramatic effect
1961 on load and link times of a DSO as it massively reduces the size of the
1962 dynamic export table when the library makes heavy use of templates.
1964 The behavior of this switch is not quite the same as marking the
1965 methods as hidden directly, because it does not affect static variables
1966 local to the function or cause the compiler to deduce that
1967 the function is defined in only one shared object.
1969 You may mark a method as having a visibility explicitly to negate the
1970 effect of the switch for that method. For example, if you do want to
1971 compare pointers to a particular inline method, you might mark it as
1972 having default visibility. Marking the enclosing class with explicit
1973 visibility will have no effect.
1975 Explicitly instantiated inline methods are unaffected by this option
1976 as their linkage might otherwise cross a shared library boundary.
1977 @xref{Template Instantiation}.
1979 @item -fvisibility-ms-compat
1980 @opindex fvisibility-ms-compat
1981 This flag attempts to use visibility settings to make GCC's C++
1982 linkage model compatible with that of Microsoft Visual Studio.
1984 The flag makes these changes to GCC's linkage model:
1988 It sets the default visibility to @code{hidden}, like
1989 @option{-fvisibility=hidden}.
1992 Types, but not their members, are not hidden by default.
1995 The One Definition Rule is relaxed for types without explicit
1996 visibility specifications which are defined in more than one different
1997 shared object: those declarations are permitted if they would have
1998 been permitted when this option was not used.
2001 In new code it is better to use @option{-fvisibility=hidden} and
2002 export those classes which are intended to be externally visible.
2003 Unfortunately it is possible for code to rely, perhaps accidentally,
2004 on the Visual Studio behavior.
2006 Among the consequences of these changes are that static data members
2007 of the same type with the same name but defined in different shared
2008 objects will be different, so changing one will not change the other;
2009 and that pointers to function members defined in different shared
2010 objects may not compare equal. When this flag is given, it is a
2011 violation of the ODR to define types with the same name differently.
2015 Do not use weak symbol support, even if it is provided by the linker.
2016 By default, G++ will use weak symbols if they are available. This
2017 option exists only for testing, and should not be used by end-users;
2018 it will result in inferior code and has no benefits. This option may
2019 be removed in a future release of G++.
2023 Do not search for header files in the standard directories specific to
2024 C++, but do still search the other standard directories. (This option
2025 is used when building the C++ library.)
2028 In addition, these optimization, warning, and code generation options
2029 have meanings only for C++ programs:
2032 @item -fno-default-inline
2033 @opindex fno-default-inline
2034 Do not assume @samp{inline} for functions defined inside a class scope.
2035 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2036 functions will have linkage like inline functions; they just won't be
2039 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2042 Warn when G++ generates code that is probably not compatible with the
2043 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2044 all such cases, there are probably some cases that are not warned about,
2045 even though G++ is generating incompatible code. There may also be
2046 cases where warnings are emitted even though the code that is generated
2049 You should rewrite your code to avoid these warnings if you are
2050 concerned about the fact that code generated by G++ may not be binary
2051 compatible with code generated by other compilers.
2053 The known incompatibilities at this point include:
2058 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2059 pack data into the same byte as a base class. For example:
2062 struct A @{ virtual void f(); int f1 : 1; @};
2063 struct B : public A @{ int f2 : 1; @};
2067 In this case, G++ will place @code{B::f2} into the same byte
2068 as@code{A::f1}; other compilers will not. You can avoid this problem
2069 by explicitly padding @code{A} so that its size is a multiple of the
2070 byte size on your platform; that will cause G++ and other compilers to
2071 layout @code{B} identically.
2074 Incorrect handling of tail-padding for virtual bases. G++ does not use
2075 tail padding when laying out virtual bases. For example:
2078 struct A @{ virtual void f(); char c1; @};
2079 struct B @{ B(); char c2; @};
2080 struct C : public A, public virtual B @{@};
2084 In this case, G++ will not place @code{B} into the tail-padding for
2085 @code{A}; other compilers will. You can avoid this problem by
2086 explicitly padding @code{A} so that its size is a multiple of its
2087 alignment (ignoring virtual base classes); that will cause G++ and other
2088 compilers to layout @code{C} identically.
2091 Incorrect handling of bit-fields with declared widths greater than that
2092 of their underlying types, when the bit-fields appear in a union. For
2096 union U @{ int i : 4096; @};
2100 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2101 union too small by the number of bits in an @code{int}.
2104 Empty classes can be placed at incorrect offsets. For example:
2114 struct C : public B, public A @{@};
2118 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2119 it should be placed at offset zero. G++ mistakenly believes that the
2120 @code{A} data member of @code{B} is already at offset zero.
2123 Names of template functions whose types involve @code{typename} or
2124 template template parameters can be mangled incorrectly.
2127 template <typename Q>
2128 void f(typename Q::X) @{@}
2130 template <template <typename> class Q>
2131 void f(typename Q<int>::X) @{@}
2135 Instantiations of these templates may be mangled incorrectly.
2139 It also warns psABI related changes. The known psABI changes at this
2145 For SYSV/x86-64, when passing union with long double, it is changed to
2146 pass in memory as specified in psABI. For example:
2156 @code{union U} will always be passed in memory.
2160 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2161 @opindex Wctor-dtor-privacy
2162 @opindex Wno-ctor-dtor-privacy
2163 Warn when a class seems unusable because all the constructors or
2164 destructors in that class are private, and it has neither friends nor
2165 public static member functions.
2167 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2168 @opindex Wnon-virtual-dtor
2169 @opindex Wno-non-virtual-dtor
2170 Warn when a class has virtual functions and accessible non-virtual
2171 destructor, in which case it would be possible but unsafe to delete
2172 an instance of a derived class through a pointer to the base class.
2173 This warning is also enabled if -Weffc++ is specified.
2175 @item -Wreorder @r{(C++ and Objective-C++ only)}
2177 @opindex Wno-reorder
2178 @cindex reordering, warning
2179 @cindex warning for reordering of member initializers
2180 Warn when the order of member initializers given in the code does not
2181 match the order in which they must be executed. For instance:
2187 A(): j (0), i (1) @{ @}
2191 The compiler will rearrange the member initializers for @samp{i}
2192 and @samp{j} to match the declaration order of the members, emitting
2193 a warning to that effect. This warning is enabled by @option{-Wall}.
2196 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2199 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2202 Warn about violations of the following style guidelines from Scott Meyers'
2203 @cite{Effective C++} book:
2207 Item 11: Define a copy constructor and an assignment operator for classes
2208 with dynamically allocated memory.
2211 Item 12: Prefer initialization to assignment in constructors.
2214 Item 14: Make destructors virtual in base classes.
2217 Item 15: Have @code{operator=} return a reference to @code{*this}.
2220 Item 23: Don't try to return a reference when you must return an object.
2224 Also warn about violations of the following style guidelines from
2225 Scott Meyers' @cite{More Effective C++} book:
2229 Item 6: Distinguish between prefix and postfix forms of increment and
2230 decrement operators.
2233 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2237 When selecting this option, be aware that the standard library
2238 headers do not obey all of these guidelines; use @samp{grep -v}
2239 to filter out those warnings.
2241 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2242 @opindex Wstrict-null-sentinel
2243 @opindex Wno-strict-null-sentinel
2244 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2245 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2246 to @code{__null}. Although it is a null pointer constant not a null pointer,
2247 it is guaranteed to be of the same size as a pointer. But this use is
2248 not portable across different compilers.
2250 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2251 @opindex Wno-non-template-friend
2252 @opindex Wnon-template-friend
2253 Disable warnings when non-templatized friend functions are declared
2254 within a template. Since the advent of explicit template specification
2255 support in G++, if the name of the friend is an unqualified-id (i.e.,
2256 @samp{friend foo(int)}), the C++ language specification demands that the
2257 friend declare or define an ordinary, nontemplate function. (Section
2258 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2259 could be interpreted as a particular specialization of a templatized
2260 function. Because this non-conforming behavior is no longer the default
2261 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2262 check existing code for potential trouble spots and is on by default.
2263 This new compiler behavior can be turned off with
2264 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2265 but disables the helpful warning.
2267 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2268 @opindex Wold-style-cast
2269 @opindex Wno-old-style-cast
2270 Warn if an old-style (C-style) cast to a non-void type is used within
2271 a C++ program. The new-style casts (@samp{dynamic_cast},
2272 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2273 less vulnerable to unintended effects and much easier to search for.
2275 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2276 @opindex Woverloaded-virtual
2277 @opindex Wno-overloaded-virtual
2278 @cindex overloaded virtual fn, warning
2279 @cindex warning for overloaded virtual fn
2280 Warn when a function declaration hides virtual functions from a
2281 base class. For example, in:
2288 struct B: public A @{
2293 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2301 will fail to compile.
2303 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2304 @opindex Wno-pmf-conversions
2305 @opindex Wpmf-conversions
2306 Disable the diagnostic for converting a bound pointer to member function
2309 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2310 @opindex Wsign-promo
2311 @opindex Wno-sign-promo
2312 Warn when overload resolution chooses a promotion from unsigned or
2313 enumerated type to a signed type, over a conversion to an unsigned type of
2314 the same size. Previous versions of G++ would try to preserve
2315 unsignedness, but the standard mandates the current behavior.
2320 A& operator = (int);
2330 In this example, G++ will synthesize a default @samp{A& operator =
2331 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2334 @node Objective-C and Objective-C++ Dialect Options
2335 @section Options Controlling Objective-C and Objective-C++ Dialects
2337 @cindex compiler options, Objective-C and Objective-C++
2338 @cindex Objective-C and Objective-C++ options, command line
2339 @cindex options, Objective-C and Objective-C++
2340 (NOTE: This manual does not describe the Objective-C and Objective-C++
2341 languages themselves. See @xref{Standards,,Language Standards
2342 Supported by GCC}, for references.)
2344 This section describes the command-line options that are only meaningful
2345 for Objective-C and Objective-C++ programs, but you can also use most of
2346 the language-independent GNU compiler options.
2347 For example, you might compile a file @code{some_class.m} like this:
2350 gcc -g -fgnu-runtime -O -c some_class.m
2354 In this example, @option{-fgnu-runtime} is an option meant only for
2355 Objective-C and Objective-C++ programs; you can use the other options with
2356 any language supported by GCC@.
2358 Note that since Objective-C is an extension of the C language, Objective-C
2359 compilations may also use options specific to the C front-end (e.g.,
2360 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2361 C++-specific options (e.g., @option{-Wabi}).
2363 Here is a list of options that are @emph{only} for compiling Objective-C
2364 and Objective-C++ programs:
2367 @item -fconstant-string-class=@var{class-name}
2368 @opindex fconstant-string-class
2369 Use @var{class-name} as the name of the class to instantiate for each
2370 literal string specified with the syntax @code{@@"@dots{}"}. The default
2371 class name is @code{NXConstantString} if the GNU runtime is being used, and
2372 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2373 @option{-fconstant-cfstrings} option, if also present, will override the
2374 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2375 to be laid out as constant CoreFoundation strings.
2378 @opindex fgnu-runtime
2379 Generate object code compatible with the standard GNU Objective-C
2380 runtime. This is the default for most types of systems.
2382 @item -fnext-runtime
2383 @opindex fnext-runtime
2384 Generate output compatible with the NeXT runtime. This is the default
2385 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2386 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2389 @item -fno-nil-receivers
2390 @opindex fno-nil-receivers
2391 Assume that all Objective-C message dispatches (e.g.,
2392 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2393 is not @code{nil}. This allows for more efficient entry points in the runtime
2394 to be used. Currently, this option is only available in conjunction with
2395 the NeXT runtime on Mac OS X 10.3 and later.
2397 @item -fobjc-call-cxx-cdtors
2398 @opindex fobjc-call-cxx-cdtors
2399 For each Objective-C class, check if any of its instance variables is a
2400 C++ object with a non-trivial default constructor. If so, synthesize a
2401 special @code{- (id) .cxx_construct} instance method that will run
2402 non-trivial default constructors on any such instance variables, in order,
2403 and then return @code{self}. Similarly, check if any instance variable
2404 is a C++ object with a non-trivial destructor, and if so, synthesize a
2405 special @code{- (void) .cxx_destruct} method that will run
2406 all such default destructors, in reverse order.
2408 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2409 thusly generated will only operate on instance variables declared in the
2410 current Objective-C class, and not those inherited from superclasses. It
2411 is the responsibility of the Objective-C runtime to invoke all such methods
2412 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2413 will be invoked by the runtime immediately after a new object
2414 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2415 be invoked immediately before the runtime deallocates an object instance.
2417 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2418 support for invoking the @code{- (id) .cxx_construct} and
2419 @code{- (void) .cxx_destruct} methods.
2421 @item -fobjc-direct-dispatch
2422 @opindex fobjc-direct-dispatch
2423 Allow fast jumps to the message dispatcher. On Darwin this is
2424 accomplished via the comm page.
2426 @item -fobjc-exceptions
2427 @opindex fobjc-exceptions
2428 Enable syntactic support for structured exception handling in Objective-C,
2429 similar to what is offered by C++ and Java. This option is
2430 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2439 @@catch (AnObjCClass *exc) @{
2446 @@catch (AnotherClass *exc) @{
2449 @@catch (id allOthers) @{
2459 The @code{@@throw} statement may appear anywhere in an Objective-C or
2460 Objective-C++ program; when used inside of a @code{@@catch} block, the
2461 @code{@@throw} may appear without an argument (as shown above), in which case
2462 the object caught by the @code{@@catch} will be rethrown.
2464 Note that only (pointers to) Objective-C objects may be thrown and
2465 caught using this scheme. When an object is thrown, it will be caught
2466 by the nearest @code{@@catch} clause capable of handling objects of that type,
2467 analogously to how @code{catch} blocks work in C++ and Java. A
2468 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2469 any and all Objective-C exceptions not caught by previous @code{@@catch}
2472 The @code{@@finally} clause, if present, will be executed upon exit from the
2473 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2474 regardless of whether any exceptions are thrown, caught or rethrown
2475 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2476 of the @code{finally} clause in Java.
2478 There are several caveats to using the new exception mechanism:
2482 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2483 idioms provided by the @code{NSException} class, the new
2484 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2485 systems, due to additional functionality needed in the (NeXT) Objective-C
2489 As mentioned above, the new exceptions do not support handling
2490 types other than Objective-C objects. Furthermore, when used from
2491 Objective-C++, the Objective-C exception model does not interoperate with C++
2492 exceptions at this time. This means you cannot @code{@@throw} an exception
2493 from Objective-C and @code{catch} it in C++, or vice versa
2494 (i.e., @code{throw @dots{} @@catch}).
2497 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2498 blocks for thread-safe execution:
2501 @@synchronized (ObjCClass *guard) @{
2506 Upon entering the @code{@@synchronized} block, a thread of execution shall
2507 first check whether a lock has been placed on the corresponding @code{guard}
2508 object by another thread. If it has, the current thread shall wait until
2509 the other thread relinquishes its lock. Once @code{guard} becomes available,
2510 the current thread will place its own lock on it, execute the code contained in
2511 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2512 making @code{guard} available to other threads).
2514 Unlike Java, Objective-C does not allow for entire methods to be marked
2515 @code{@@synchronized}. Note that throwing exceptions out of
2516 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2517 to be unlocked properly.
2521 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2523 @item -freplace-objc-classes
2524 @opindex freplace-objc-classes
2525 Emit a special marker instructing @command{ld(1)} not to statically link in
2526 the resulting object file, and allow @command{dyld(1)} to load it in at
2527 run time instead. This is used in conjunction with the Fix-and-Continue
2528 debugging mode, where the object file in question may be recompiled and
2529 dynamically reloaded in the course of program execution, without the need
2530 to restart the program itself. Currently, Fix-and-Continue functionality
2531 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2536 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2537 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2538 compile time) with static class references that get initialized at load time,
2539 which improves run-time performance. Specifying the @option{-fzero-link} flag
2540 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2541 to be retained. This is useful in Zero-Link debugging mode, since it allows
2542 for individual class implementations to be modified during program execution.
2546 Dump interface declarations for all classes seen in the source file to a
2547 file named @file{@var{sourcename}.decl}.
2549 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2550 @opindex Wassign-intercept
2551 @opindex Wno-assign-intercept
2552 Warn whenever an Objective-C assignment is being intercepted by the
2555 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2556 @opindex Wno-protocol
2558 If a class is declared to implement a protocol, a warning is issued for
2559 every method in the protocol that is not implemented by the class. The
2560 default behavior is to issue a warning for every method not explicitly
2561 implemented in the class, even if a method implementation is inherited
2562 from the superclass. If you use the @option{-Wno-protocol} option, then
2563 methods inherited from the superclass are considered to be implemented,
2564 and no warning is issued for them.
2566 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2568 @opindex Wno-selector
2569 Warn if multiple methods of different types for the same selector are
2570 found during compilation. The check is performed on the list of methods
2571 in the final stage of compilation. Additionally, a check is performed
2572 for each selector appearing in a @code{@@selector(@dots{})}
2573 expression, and a corresponding method for that selector has been found
2574 during compilation. Because these checks scan the method table only at
2575 the end of compilation, these warnings are not produced if the final
2576 stage of compilation is not reached, for example because an error is
2577 found during compilation, or because the @option{-fsyntax-only} option is
2580 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2581 @opindex Wstrict-selector-match
2582 @opindex Wno-strict-selector-match
2583 Warn if multiple methods with differing argument and/or return types are
2584 found for a given selector when attempting to send a message using this
2585 selector to a receiver of type @code{id} or @code{Class}. When this flag
2586 is off (which is the default behavior), the compiler will omit such warnings
2587 if any differences found are confined to types which share the same size
2590 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2591 @opindex Wundeclared-selector
2592 @opindex Wno-undeclared-selector
2593 Warn if a @code{@@selector(@dots{})} expression referring to an
2594 undeclared selector is found. A selector is considered undeclared if no
2595 method with that name has been declared before the
2596 @code{@@selector(@dots{})} expression, either explicitly in an
2597 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2598 an @code{@@implementation} section. This option always performs its
2599 checks as soon as a @code{@@selector(@dots{})} expression is found,
2600 while @option{-Wselector} only performs its checks in the final stage of
2601 compilation. This also enforces the coding style convention
2602 that methods and selectors must be declared before being used.
2604 @item -print-objc-runtime-info
2605 @opindex print-objc-runtime-info
2606 Generate C header describing the largest structure that is passed by
2611 @node Language Independent Options
2612 @section Options to Control Diagnostic Messages Formatting
2613 @cindex options to control diagnostics formatting
2614 @cindex diagnostic messages
2615 @cindex message formatting
2617 Traditionally, diagnostic messages have been formatted irrespective of
2618 the output device's aspect (e.g.@: its width, @dots{}). The options described
2619 below can be used to control the diagnostic messages formatting
2620 algorithm, e.g.@: how many characters per line, how often source location
2621 information should be reported. Right now, only the C++ front end can
2622 honor these options. However it is expected, in the near future, that
2623 the remaining front ends would be able to digest them correctly.
2626 @item -fmessage-length=@var{n}
2627 @opindex fmessage-length
2628 Try to format error messages so that they fit on lines of about @var{n}
2629 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2630 the front ends supported by GCC@. If @var{n} is zero, then no
2631 line-wrapping will be done; each error message will appear on a single
2634 @opindex fdiagnostics-show-location
2635 @item -fdiagnostics-show-location=once
2636 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2637 reporter to emit @emph{once} source location information; that is, in
2638 case the message is too long to fit on a single physical line and has to
2639 be wrapped, the source location won't be emitted (as prefix) again,
2640 over and over, in subsequent continuation lines. This is the default
2643 @item -fdiagnostics-show-location=every-line
2644 Only meaningful in line-wrapping mode. Instructs the diagnostic
2645 messages reporter to emit the same source location information (as
2646 prefix) for physical lines that result from the process of breaking
2647 a message which is too long to fit on a single line.
2649 @item -fdiagnostics-show-option
2650 @opindex fdiagnostics-show-option
2651 This option instructs the diagnostic machinery to add text to each
2652 diagnostic emitted, which indicates which command line option directly
2653 controls that diagnostic, when such an option is known to the
2654 diagnostic machinery.
2656 @item -Wcoverage-mismatch
2657 @opindex Wcoverage-mismatch
2658 Warn if feedback profiles do not match when using the
2659 @option{-fprofile-use} option.
2660 If a source file was changed between @option{-fprofile-gen} and
2661 @option{-fprofile-use}, the files with the profile feedback can fail
2662 to match the source file and GCC can not use the profile feedback
2663 information. By default, GCC emits an error message in this case.
2664 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2665 error. GCC does not use appropriate feedback profiles, so using this
2666 option can result in poorly optimized code. This option is useful
2667 only in the case of very minor changes such as bug fixes to an
2672 @node Warning Options
2673 @section Options to Request or Suppress Warnings
2674 @cindex options to control warnings
2675 @cindex warning messages
2676 @cindex messages, warning
2677 @cindex suppressing warnings
2679 Warnings are diagnostic messages that report constructions which
2680 are not inherently erroneous but which are risky or suggest there
2681 may have been an error.
2683 The following language-independent options do not enable specific
2684 warnings but control the kinds of diagnostics produced by GCC.
2687 @cindex syntax checking
2689 @opindex fsyntax-only
2690 Check the code for syntax errors, but don't do anything beyond that.
2694 Inhibit all warning messages.
2699 Make all warnings into errors.
2704 Make the specified warning into an error. The specifier for a warning
2705 is appended, for example @option{-Werror=switch} turns the warnings
2706 controlled by @option{-Wswitch} into errors. This switch takes a
2707 negative form, to be used to negate @option{-Werror} for specific
2708 warnings, for example @option{-Wno-error=switch} makes
2709 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2710 is in effect. You can use the @option{-fdiagnostics-show-option}
2711 option to have each controllable warning amended with the option which
2712 controls it, to determine what to use with this option.
2714 Note that specifying @option{-Werror=}@var{foo} automatically implies
2715 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2718 @item -Wfatal-errors
2719 @opindex Wfatal-errors
2720 @opindex Wno-fatal-errors
2721 This option causes the compiler to abort compilation on the first error
2722 occurred rather than trying to keep going and printing further error
2727 You can request many specific warnings with options beginning
2728 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2729 implicit declarations. Each of these specific warning options also
2730 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2731 example, @option{-Wno-implicit}. This manual lists only one of the
2732 two forms, whichever is not the default. For further,
2733 language-specific options also refer to @ref{C++ Dialect Options} and
2734 @ref{Objective-C and Objective-C++ Dialect Options}.
2739 Issue all the warnings demanded by strict ISO C and ISO C++;
2740 reject all programs that use forbidden extensions, and some other
2741 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2742 version of the ISO C standard specified by any @option{-std} option used.
2744 Valid ISO C and ISO C++ programs should compile properly with or without
2745 this option (though a rare few will require @option{-ansi} or a
2746 @option{-std} option specifying the required version of ISO C)@. However,
2747 without this option, certain GNU extensions and traditional C and C++
2748 features are supported as well. With this option, they are rejected.
2750 @option{-pedantic} does not cause warning messages for use of the
2751 alternate keywords whose names begin and end with @samp{__}. Pedantic
2752 warnings are also disabled in the expression that follows
2753 @code{__extension__}. However, only system header files should use
2754 these escape routes; application programs should avoid them.
2755 @xref{Alternate Keywords}.
2757 Some users try to use @option{-pedantic} to check programs for strict ISO
2758 C conformance. They soon find that it does not do quite what they want:
2759 it finds some non-ISO practices, but not all---only those for which
2760 ISO C @emph{requires} a diagnostic, and some others for which
2761 diagnostics have been added.
2763 A feature to report any failure to conform to ISO C might be useful in
2764 some instances, but would require considerable additional work and would
2765 be quite different from @option{-pedantic}. We don't have plans to
2766 support such a feature in the near future.
2768 Where the standard specified with @option{-std} represents a GNU
2769 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2770 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2771 extended dialect is based. Warnings from @option{-pedantic} are given
2772 where they are required by the base standard. (It would not make sense
2773 for such warnings to be given only for features not in the specified GNU
2774 C dialect, since by definition the GNU dialects of C include all
2775 features the compiler supports with the given option, and there would be
2776 nothing to warn about.)
2778 @item -pedantic-errors
2779 @opindex pedantic-errors
2780 Like @option{-pedantic}, except that errors are produced rather than
2786 This enables all the warnings about constructions that some users
2787 consider questionable, and that are easy to avoid (or modify to
2788 prevent the warning), even in conjunction with macros. This also
2789 enables some language-specific warnings described in @ref{C++ Dialect
2790 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2792 @option{-Wall} turns on the following warning flags:
2794 @gccoptlist{-Waddress @gol
2795 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2797 -Wchar-subscripts @gol
2798 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2800 -Wimplicit-function-declaration @gol
2803 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2804 -Wmissing-braces @gol
2810 -Wsequence-point @gol
2811 -Wsign-compare @r{(only in C++)} @gol
2812 -Wstrict-aliasing @gol
2813 -Wstrict-overflow=1 @gol
2816 -Wuninitialized @gol
2817 -Wunknown-pragmas @gol
2818 -Wunused-function @gol
2821 -Wunused-variable @gol
2822 -Wvolatile-register-var @gol
2825 Note that some warning flags are not implied by @option{-Wall}. Some of
2826 them warn about constructions that users generally do not consider
2827 questionable, but which occasionally you might wish to check for;
2828 others warn about constructions that are necessary or hard to avoid in
2829 some cases, and there is no simple way to modify the code to suppress
2830 the warning. Some of them are enabled by @option{-Wextra} but many of
2831 them must be enabled individually.
2837 This enables some extra warning flags that are not enabled by
2838 @option{-Wall}. (This option used to be called @option{-W}. The older
2839 name is still supported, but the newer name is more descriptive.)
2841 @gccoptlist{-Wclobbered @gol
2843 -Wignored-qualifiers @gol
2844 -Wmissing-field-initializers @gol
2845 -Wmissing-parameter-type @r{(C only)} @gol
2846 -Wold-style-declaration @r{(C only)} @gol
2847 -Woverride-init @gol
2850 -Wuninitialized @gol
2851 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2854 The option @option{-Wextra} also prints warning messages for the
2860 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2861 @samp{>}, or @samp{>=}.
2864 (C++ only) An enumerator and a non-enumerator both appear in a
2865 conditional expression.
2868 (C++ only) Ambiguous virtual bases.
2871 (C++ only) Subscripting an array which has been declared @samp{register}.
2874 (C++ only) Taking the address of a variable which has been declared
2878 (C++ only) A base class is not initialized in a derived class' copy
2883 @item -Wchar-subscripts
2884 @opindex Wchar-subscripts
2885 @opindex Wno-char-subscripts
2886 Warn if an array subscript has type @code{char}. This is a common cause
2887 of error, as programmers often forget that this type is signed on some
2889 This warning is enabled by @option{-Wall}.
2893 @opindex Wno-comment
2894 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2895 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2896 This warning is enabled by @option{-Wall}.
2901 @opindex ffreestanding
2902 @opindex fno-builtin
2903 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2904 the arguments supplied have types appropriate to the format string
2905 specified, and that the conversions specified in the format string make
2906 sense. This includes standard functions, and others specified by format
2907 attributes (@pxref{Function Attributes}), in the @code{printf},
2908 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2909 not in the C standard) families (or other target-specific families).
2910 Which functions are checked without format attributes having been
2911 specified depends on the standard version selected, and such checks of
2912 functions without the attribute specified are disabled by
2913 @option{-ffreestanding} or @option{-fno-builtin}.
2915 The formats are checked against the format features supported by GNU
2916 libc version 2.2. These include all ISO C90 and C99 features, as well
2917 as features from the Single Unix Specification and some BSD and GNU
2918 extensions. Other library implementations may not support all these
2919 features; GCC does not support warning about features that go beyond a
2920 particular library's limitations. However, if @option{-pedantic} is used
2921 with @option{-Wformat}, warnings will be given about format features not
2922 in the selected standard version (but not for @code{strfmon} formats,
2923 since those are not in any version of the C standard). @xref{C Dialect
2924 Options,,Options Controlling C Dialect}.
2926 Since @option{-Wformat} also checks for null format arguments for
2927 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2929 @option{-Wformat} is included in @option{-Wall}. For more control over some
2930 aspects of format checking, the options @option{-Wformat-y2k},
2931 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2932 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2933 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2936 @opindex Wformat-y2k
2937 @opindex Wno-format-y2k
2938 If @option{-Wformat} is specified, also warn about @code{strftime}
2939 formats which may yield only a two-digit year.
2941 @item -Wno-format-contains-nul
2942 @opindex Wno-format-contains-nul
2943 @opindex Wformat-contains-nul
2944 If @option{-Wformat} is specified, do not warn about format strings that
2947 @item -Wno-format-extra-args
2948 @opindex Wno-format-extra-args
2949 @opindex Wformat-extra-args
2950 If @option{-Wformat} is specified, do not warn about excess arguments to a
2951 @code{printf} or @code{scanf} format function. The C standard specifies
2952 that such arguments are ignored.
2954 Where the unused arguments lie between used arguments that are
2955 specified with @samp{$} operand number specifications, normally
2956 warnings are still given, since the implementation could not know what
2957 type to pass to @code{va_arg} to skip the unused arguments. However,
2958 in the case of @code{scanf} formats, this option will suppress the
2959 warning if the unused arguments are all pointers, since the Single
2960 Unix Specification says that such unused arguments are allowed.
2962 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2963 @opindex Wno-format-zero-length
2964 @opindex Wformat-zero-length
2965 If @option{-Wformat} is specified, do not warn about zero-length formats.
2966 The C standard specifies that zero-length formats are allowed.
2968 @item -Wformat-nonliteral
2969 @opindex Wformat-nonliteral
2970 @opindex Wno-format-nonliteral
2971 If @option{-Wformat} is specified, also warn if the format string is not a
2972 string literal and so cannot be checked, unless the format function
2973 takes its format arguments as a @code{va_list}.
2975 @item -Wformat-security
2976 @opindex Wformat-security
2977 @opindex Wno-format-security
2978 If @option{-Wformat} is specified, also warn about uses of format
2979 functions that represent possible security problems. At present, this
2980 warns about calls to @code{printf} and @code{scanf} functions where the
2981 format string is not a string literal and there are no format arguments,
2982 as in @code{printf (foo);}. This may be a security hole if the format
2983 string came from untrusted input and contains @samp{%n}. (This is
2984 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2985 in future warnings may be added to @option{-Wformat-security} that are not
2986 included in @option{-Wformat-nonliteral}.)
2990 @opindex Wno-format=2
2991 Enable @option{-Wformat} plus format checks not included in
2992 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2993 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2995 @item -Wnonnull @r{(C and Objective-C only)}
2997 @opindex Wno-nonnull
2998 Warn about passing a null pointer for arguments marked as
2999 requiring a non-null value by the @code{nonnull} function attribute.
3001 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3002 can be disabled with the @option{-Wno-nonnull} option.
3004 @item -Wjump-misses-init @r{(C, Objective-C only)}
3005 @opindex Wjump-misses-init
3006 @opindex Wno-jump-misses-init
3007 Warn if a @code{goto} statement or a @code{switch} statement jumps
3008 forward across the initialization of a variable, or jumps backward to a
3009 label after the variable has been initialized. This only warns about
3010 variables which are initialized when they are declared. This warning is
3011 only supported for C and Objective C; in C++ this sort of branch is an
3014 @option{-Wjump-misses-init} is included in @option{-Wall} and
3015 @option{-Wc++-compat}. It can be disabled with the
3016 @option{-Wno-jump-misses-init} option.
3018 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3020 @opindex Wno-init-self
3021 Warn about uninitialized variables which are initialized with themselves.
3022 Note this option can only be used with the @option{-Wuninitialized} option.
3024 For example, GCC will warn about @code{i} being uninitialized in the
3025 following snippet only when @option{-Winit-self} has been specified:
3036 @item -Wimplicit-int @r{(C and Objective-C only)}
3037 @opindex Wimplicit-int
3038 @opindex Wno-implicit-int
3039 Warn when a declaration does not specify a type.
3040 This warning is enabled by @option{-Wall}.
3042 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3043 @opindex Wimplicit-function-declaration
3044 @opindex Wno-implicit-function-declaration
3045 Give a warning whenever a function is used before being declared. In
3046 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3047 enabled by default and it is made into an error by
3048 @option{-pedantic-errors}. This warning is also enabled by
3053 @opindex Wno-implicit
3054 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3055 This warning is enabled by @option{-Wall}.
3057 @item -Wignored-qualifiers @r{(C and C++ only)}
3058 @opindex Wignored-qualifiers
3059 @opindex Wno-ignored-qualifiers
3060 Warn if the return type of a function has a type qualifier
3061 such as @code{const}. For ISO C such a type qualifier has no effect,
3062 since the value returned by a function is not an lvalue.
3063 For C++, the warning is only emitted for scalar types or @code{void}.
3064 ISO C prohibits qualified @code{void} return types on function
3065 definitions, so such return types always receive a warning
3066 even without this option.
3068 This warning is also enabled by @option{-Wextra}.
3073 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3074 a function with external linkage, returning int, taking either zero
3075 arguments, two, or three arguments of appropriate types. This warning
3076 is enabled by default in C++ and is enabled by either @option{-Wall}
3077 or @option{-pedantic}.
3079 @item -Wmissing-braces
3080 @opindex Wmissing-braces
3081 @opindex Wno-missing-braces
3082 Warn if an aggregate or union initializer is not fully bracketed. In
3083 the following example, the initializer for @samp{a} is not fully
3084 bracketed, but that for @samp{b} is fully bracketed.
3087 int a[2][2] = @{ 0, 1, 2, 3 @};
3088 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3091 This warning is enabled by @option{-Wall}.
3093 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3094 @opindex Wmissing-include-dirs
3095 @opindex Wno-missing-include-dirs
3096 Warn if a user-supplied include directory does not exist.
3099 @opindex Wparentheses
3100 @opindex Wno-parentheses
3101 Warn if parentheses are omitted in certain contexts, such
3102 as when there is an assignment in a context where a truth value
3103 is expected, or when operators are nested whose precedence people
3104 often get confused about.
3106 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3107 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3108 interpretation from that of ordinary mathematical notation.
3110 Also warn about constructions where there may be confusion to which
3111 @code{if} statement an @code{else} branch belongs. Here is an example of
3126 In C/C++, every @code{else} branch belongs to the innermost possible
3127 @code{if} statement, which in this example is @code{if (b)}. This is
3128 often not what the programmer expected, as illustrated in the above
3129 example by indentation the programmer chose. When there is the
3130 potential for this confusion, GCC will issue a warning when this flag
3131 is specified. To eliminate the warning, add explicit braces around
3132 the innermost @code{if} statement so there is no way the @code{else}
3133 could belong to the enclosing @code{if}. The resulting code would
3150 This warning is enabled by @option{-Wall}.
3152 @item -Wsequence-point
3153 @opindex Wsequence-point
3154 @opindex Wno-sequence-point
3155 Warn about code that may have undefined semantics because of violations
3156 of sequence point rules in the C and C++ standards.
3158 The C and C++ standards defines the order in which expressions in a C/C++
3159 program are evaluated in terms of @dfn{sequence points}, which represent
3160 a partial ordering between the execution of parts of the program: those
3161 executed before the sequence point, and those executed after it. These
3162 occur after the evaluation of a full expression (one which is not part
3163 of a larger expression), after the evaluation of the first operand of a
3164 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3165 function is called (but after the evaluation of its arguments and the
3166 expression denoting the called function), and in certain other places.
3167 Other than as expressed by the sequence point rules, the order of
3168 evaluation of subexpressions of an expression is not specified. All
3169 these rules describe only a partial order rather than a total order,
3170 since, for example, if two functions are called within one expression
3171 with no sequence point between them, the order in which the functions
3172 are called is not specified. However, the standards committee have
3173 ruled that function calls do not overlap.
3175 It is not specified when between sequence points modifications to the
3176 values of objects take effect. Programs whose behavior depends on this
3177 have undefined behavior; the C and C++ standards specify that ``Between
3178 the previous and next sequence point an object shall have its stored
3179 value modified at most once by the evaluation of an expression.
3180 Furthermore, the prior value shall be read only to determine the value
3181 to be stored.''. If a program breaks these rules, the results on any
3182 particular implementation are entirely unpredictable.
3184 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3185 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3186 diagnosed by this option, and it may give an occasional false positive
3187 result, but in general it has been found fairly effective at detecting
3188 this sort of problem in programs.
3190 The standard is worded confusingly, therefore there is some debate
3191 over the precise meaning of the sequence point rules in subtle cases.
3192 Links to discussions of the problem, including proposed formal
3193 definitions, may be found on the GCC readings page, at
3194 @w{@uref{http://gcc.gnu.org/readings.html}}.
3196 This warning is enabled by @option{-Wall} for C and C++.
3199 @opindex Wreturn-type
3200 @opindex Wno-return-type
3201 Warn whenever a function is defined with a return-type that defaults
3202 to @code{int}. Also warn about any @code{return} statement with no
3203 return-value in a function whose return-type is not @code{void}
3204 (falling off the end of the function body is considered returning
3205 without a value), and about a @code{return} statement with an
3206 expression in a function whose return-type is @code{void}.
3208 For C++, a function without return type always produces a diagnostic
3209 message, even when @option{-Wno-return-type} is specified. The only
3210 exceptions are @samp{main} and functions defined in system headers.
3212 This warning is enabled by @option{-Wall}.
3217 Warn whenever a @code{switch} statement has an index of enumerated type
3218 and lacks a @code{case} for one or more of the named codes of that
3219 enumeration. (The presence of a @code{default} label prevents this
3220 warning.) @code{case} labels outside the enumeration range also
3221 provoke warnings when this option is used (even if there is a
3222 @code{default} label).
3223 This warning is enabled by @option{-Wall}.
3225 @item -Wswitch-default
3226 @opindex Wswitch-default
3227 @opindex Wno-switch-default
3228 Warn whenever a @code{switch} statement does not have a @code{default}
3232 @opindex Wswitch-enum
3233 @opindex Wno-switch-enum
3234 Warn whenever a @code{switch} statement has an index of enumerated type
3235 and lacks a @code{case} for one or more of the named codes of that
3236 enumeration. @code{case} labels outside the enumeration range also
3237 provoke warnings when this option is used. The only difference
3238 between @option{-Wswitch} and this option is that this option gives a
3239 warning about an omitted enumeration code even if there is a
3240 @code{default} label.
3242 @item -Wsync-nand @r{(C and C++ only)}
3244 @opindex Wno-sync-nand
3245 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3246 built-in functions are used. These functions changed semantics in GCC 4.4.
3250 @opindex Wno-trigraphs
3251 Warn if any trigraphs are encountered that might change the meaning of
3252 the program (trigraphs within comments are not warned about).
3253 This warning is enabled by @option{-Wall}.
3255 @item -Wunused-function
3256 @opindex Wunused-function
3257 @opindex Wno-unused-function
3258 Warn whenever a static function is declared but not defined or a
3259 non-inline static function is unused.
3260 This warning is enabled by @option{-Wall}.
3262 @item -Wunused-label
3263 @opindex Wunused-label
3264 @opindex Wno-unused-label
3265 Warn whenever a label is declared but not used.
3266 This warning is enabled by @option{-Wall}.
3268 To suppress this warning use the @samp{unused} attribute
3269 (@pxref{Variable Attributes}).
3271 @item -Wunused-parameter
3272 @opindex Wunused-parameter
3273 @opindex Wno-unused-parameter
3274 Warn whenever a function parameter is unused aside from its declaration.
3276 To suppress this warning use the @samp{unused} attribute
3277 (@pxref{Variable Attributes}).
3279 @item -Wno-unused-result
3280 @opindex Wunused-result
3281 @opindex Wno-unused-result
3282 Do not warn if a caller of a function marked with attribute
3283 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3284 its return value. The default is @option{-Wunused-result}.
3286 @item -Wunused-variable
3287 @opindex Wunused-variable
3288 @opindex Wno-unused-variable
3289 Warn whenever a local variable or non-constant static variable is unused
3290 aside from its declaration.
3291 This warning is enabled by @option{-Wall}.
3293 To suppress this warning use the @samp{unused} attribute
3294 (@pxref{Variable Attributes}).
3296 @item -Wunused-value
3297 @opindex Wunused-value
3298 @opindex Wno-unused-value
3299 Warn whenever a statement computes a result that is explicitly not
3300 used. To suppress this warning cast the unused expression to
3301 @samp{void}. This includes an expression-statement or the left-hand
3302 side of a comma expression that contains no side effects. For example,
3303 an expression such as @samp{x[i,j]} will cause a warning, while
3304 @samp{x[(void)i,j]} will not.
3306 This warning is enabled by @option{-Wall}.
3311 All the above @option{-Wunused} options combined.
3313 In order to get a warning about an unused function parameter, you must
3314 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3315 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3317 @item -Wuninitialized
3318 @opindex Wuninitialized
3319 @opindex Wno-uninitialized
3320 Warn if an automatic variable is used without first being initialized
3321 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3322 warn if a non-static reference or non-static @samp{const} member
3323 appears in a class without constructors.
3325 If you want to warn about code which uses the uninitialized value of the
3326 variable in its own initializer, use the @option{-Winit-self} option.
3328 These warnings occur for individual uninitialized or clobbered
3329 elements of structure, union or array variables as well as for
3330 variables which are uninitialized or clobbered as a whole. They do
3331 not occur for variables or elements declared @code{volatile}. Because
3332 these warnings depend on optimization, the exact variables or elements
3333 for which there are warnings will depend on the precise optimization
3334 options and version of GCC used.
3336 Note that there may be no warning about a variable that is used only
3337 to compute a value that itself is never used, because such
3338 computations may be deleted by data flow analysis before the warnings
3341 These warnings are made optional because GCC is not smart
3342 enough to see all the reasons why the code might be correct
3343 despite appearing to have an error. Here is one example of how
3364 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3365 always initialized, but GCC doesn't know this. Here is
3366 another common case:
3371 if (change_y) save_y = y, y = new_y;
3373 if (change_y) y = save_y;
3378 This has no bug because @code{save_y} is used only if it is set.
3380 @cindex @code{longjmp} warnings
3381 This option also warns when a non-volatile automatic variable might be
3382 changed by a call to @code{longjmp}. These warnings as well are possible
3383 only in optimizing compilation.
3385 The compiler sees only the calls to @code{setjmp}. It cannot know
3386 where @code{longjmp} will be called; in fact, a signal handler could
3387 call it at any point in the code. As a result, you may get a warning
3388 even when there is in fact no problem because @code{longjmp} cannot
3389 in fact be called at the place which would cause a problem.
3391 Some spurious warnings can be avoided if you declare all the functions
3392 you use that never return as @code{noreturn}. @xref{Function
3395 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3397 @item -Wunknown-pragmas
3398 @opindex Wunknown-pragmas
3399 @opindex Wno-unknown-pragmas
3400 @cindex warning for unknown pragmas
3401 @cindex unknown pragmas, warning
3402 @cindex pragmas, warning of unknown
3403 Warn when a #pragma directive is encountered which is not understood by
3404 GCC@. If this command line option is used, warnings will even be issued
3405 for unknown pragmas in system header files. This is not the case if
3406 the warnings were only enabled by the @option{-Wall} command line option.
3409 @opindex Wno-pragmas
3411 Do not warn about misuses of pragmas, such as incorrect parameters,
3412 invalid syntax, or conflicts between pragmas. See also
3413 @samp{-Wunknown-pragmas}.
3415 @item -Wstrict-aliasing
3416 @opindex Wstrict-aliasing
3417 @opindex Wno-strict-aliasing
3418 This option is only active when @option{-fstrict-aliasing} is active.
3419 It warns about code which might break the strict aliasing rules that the
3420 compiler is using for optimization. The warning does not catch all
3421 cases, but does attempt to catch the more common pitfalls. It is
3422 included in @option{-Wall}.
3423 It is equivalent to @option{-Wstrict-aliasing=3}
3425 @item -Wstrict-aliasing=n
3426 @opindex Wstrict-aliasing=n
3427 @opindex Wno-strict-aliasing=n
3428 This option is only active when @option{-fstrict-aliasing} is active.
3429 It warns about code which might break the strict aliasing rules that the
3430 compiler is using for optimization.
3431 Higher levels correspond to higher accuracy (fewer false positives).
3432 Higher levels also correspond to more effort, similar to the way -O works.
3433 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3436 Level 1: Most aggressive, quick, least accurate.
3437 Possibly useful when higher levels
3438 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3439 false negatives. However, it has many false positives.
3440 Warns for all pointer conversions between possibly incompatible types,
3441 even if never dereferenced. Runs in the frontend only.
3443 Level 2: Aggressive, quick, not too precise.
3444 May still have many false positives (not as many as level 1 though),
3445 and few false negatives (but possibly more than level 1).
3446 Unlike level 1, it only warns when an address is taken. Warns about
3447 incomplete types. Runs in the frontend only.
3449 Level 3 (default for @option{-Wstrict-aliasing}):
3450 Should have very few false positives and few false
3451 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3452 Takes care of the common punn+dereference pattern in the frontend:
3453 @code{*(int*)&some_float}.
3454 If optimization is enabled, it also runs in the backend, where it deals
3455 with multiple statement cases using flow-sensitive points-to information.
3456 Only warns when the converted pointer is dereferenced.
3457 Does not warn about incomplete types.
3459 @item -Wstrict-overflow
3460 @itemx -Wstrict-overflow=@var{n}
3461 @opindex Wstrict-overflow
3462 @opindex Wno-strict-overflow
3463 This option is only active when @option{-fstrict-overflow} is active.
3464 It warns about cases where the compiler optimizes based on the
3465 assumption that signed overflow does not occur. Note that it does not
3466 warn about all cases where the code might overflow: it only warns
3467 about cases where the compiler implements some optimization. Thus
3468 this warning depends on the optimization level.
3470 An optimization which assumes that signed overflow does not occur is
3471 perfectly safe if the values of the variables involved are such that
3472 overflow never does, in fact, occur. Therefore this warning can
3473 easily give a false positive: a warning about code which is not
3474 actually a problem. To help focus on important issues, several
3475 warning levels are defined. No warnings are issued for the use of
3476 undefined signed overflow when estimating how many iterations a loop
3477 will require, in particular when determining whether a loop will be
3481 @item -Wstrict-overflow=1
3482 Warn about cases which are both questionable and easy to avoid. For
3483 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3484 compiler will simplify this to @code{1}. This level of
3485 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3486 are not, and must be explicitly requested.
3488 @item -Wstrict-overflow=2
3489 Also warn about other cases where a comparison is simplified to a
3490 constant. For example: @code{abs (x) >= 0}. This can only be
3491 simplified when @option{-fstrict-overflow} is in effect, because
3492 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3493 zero. @option{-Wstrict-overflow} (with no level) is the same as
3494 @option{-Wstrict-overflow=2}.
3496 @item -Wstrict-overflow=3
3497 Also warn about other cases where a comparison is simplified. For
3498 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3500 @item -Wstrict-overflow=4
3501 Also warn about other simplifications not covered by the above cases.
3502 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3504 @item -Wstrict-overflow=5
3505 Also warn about cases where the compiler reduces the magnitude of a
3506 constant involved in a comparison. For example: @code{x + 2 > y} will
3507 be simplified to @code{x + 1 >= y}. This is reported only at the
3508 highest warning level because this simplification applies to many
3509 comparisons, so this warning level will give a very large number of
3513 @item -Warray-bounds
3514 @opindex Wno-array-bounds
3515 @opindex Warray-bounds
3516 This option is only active when @option{-ftree-vrp} is active
3517 (default for -O2 and above). It warns about subscripts to arrays
3518 that are always out of bounds. This warning is enabled by @option{-Wall}.
3520 @item -Wno-div-by-zero
3521 @opindex Wno-div-by-zero
3522 @opindex Wdiv-by-zero
3523 Do not warn about compile-time integer division by zero. Floating point
3524 division by zero is not warned about, as it can be a legitimate way of
3525 obtaining infinities and NaNs.
3527 @item -Wsystem-headers
3528 @opindex Wsystem-headers
3529 @opindex Wno-system-headers
3530 @cindex warnings from system headers
3531 @cindex system headers, warnings from
3532 Print warning messages for constructs found in system header files.
3533 Warnings from system headers are normally suppressed, on the assumption
3534 that they usually do not indicate real problems and would only make the
3535 compiler output harder to read. Using this command line option tells
3536 GCC to emit warnings from system headers as if they occurred in user
3537 code. However, note that using @option{-Wall} in conjunction with this
3538 option will @emph{not} warn about unknown pragmas in system
3539 headers---for that, @option{-Wunknown-pragmas} must also be used.
3542 @opindex Wfloat-equal
3543 @opindex Wno-float-equal
3544 Warn if floating point values are used in equality comparisons.
3546 The idea behind this is that sometimes it is convenient (for the
3547 programmer) to consider floating-point values as approximations to
3548 infinitely precise real numbers. If you are doing this, then you need
3549 to compute (by analyzing the code, or in some other way) the maximum or
3550 likely maximum error that the computation introduces, and allow for it
3551 when performing comparisons (and when producing output, but that's a
3552 different problem). In particular, instead of testing for equality, you
3553 would check to see whether the two values have ranges that overlap; and
3554 this is done with the relational operators, so equality comparisons are
3557 @item -Wtraditional @r{(C and Objective-C only)}
3558 @opindex Wtraditional
3559 @opindex Wno-traditional
3560 Warn about certain constructs that behave differently in traditional and
3561 ISO C@. Also warn about ISO C constructs that have no traditional C
3562 equivalent, and/or problematic constructs which should be avoided.
3566 Macro parameters that appear within string literals in the macro body.
3567 In traditional C macro replacement takes place within string literals,
3568 but does not in ISO C@.
3571 In traditional C, some preprocessor directives did not exist.
3572 Traditional preprocessors would only consider a line to be a directive
3573 if the @samp{#} appeared in column 1 on the line. Therefore
3574 @option{-Wtraditional} warns about directives that traditional C
3575 understands but would ignore because the @samp{#} does not appear as the
3576 first character on the line. It also suggests you hide directives like
3577 @samp{#pragma} not understood by traditional C by indenting them. Some
3578 traditional implementations would not recognize @samp{#elif}, so it
3579 suggests avoiding it altogether.
3582 A function-like macro that appears without arguments.
3585 The unary plus operator.
3588 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3589 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3590 constants.) Note, these suffixes appear in macros defined in the system
3591 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3592 Use of these macros in user code might normally lead to spurious
3593 warnings, however GCC's integrated preprocessor has enough context to
3594 avoid warning in these cases.
3597 A function declared external in one block and then used after the end of
3601 A @code{switch} statement has an operand of type @code{long}.
3604 A non-@code{static} function declaration follows a @code{static} one.
3605 This construct is not accepted by some traditional C compilers.
3608 The ISO type of an integer constant has a different width or
3609 signedness from its traditional type. This warning is only issued if
3610 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3611 typically represent bit patterns, are not warned about.
3614 Usage of ISO string concatenation is detected.
3617 Initialization of automatic aggregates.
3620 Identifier conflicts with labels. Traditional C lacks a separate
3621 namespace for labels.
3624 Initialization of unions. If the initializer is zero, the warning is
3625 omitted. This is done under the assumption that the zero initializer in
3626 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3627 initializer warnings and relies on default initialization to zero in the
3631 Conversions by prototypes between fixed/floating point values and vice
3632 versa. The absence of these prototypes when compiling with traditional
3633 C would cause serious problems. This is a subset of the possible
3634 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3637 Use of ISO C style function definitions. This warning intentionally is
3638 @emph{not} issued for prototype declarations or variadic functions
3639 because these ISO C features will appear in your code when using
3640 libiberty's traditional C compatibility macros, @code{PARAMS} and
3641 @code{VPARAMS}. This warning is also bypassed for nested functions
3642 because that feature is already a GCC extension and thus not relevant to
3643 traditional C compatibility.
3646 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3647 @opindex Wtraditional-conversion
3648 @opindex Wno-traditional-conversion
3649 Warn if a prototype causes a type conversion that is different from what
3650 would happen to the same argument in the absence of a prototype. This
3651 includes conversions of fixed point to floating and vice versa, and
3652 conversions changing the width or signedness of a fixed point argument
3653 except when the same as the default promotion.
3655 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3656 @opindex Wdeclaration-after-statement
3657 @opindex Wno-declaration-after-statement
3658 Warn when a declaration is found after a statement in a block. This
3659 construct, known from C++, was introduced with ISO C99 and is by default
3660 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3661 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3666 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3668 @item -Wno-endif-labels
3669 @opindex Wno-endif-labels
3670 @opindex Wendif-labels
3671 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3676 Warn whenever a local variable shadows another local variable, parameter or
3677 global variable or whenever a built-in function is shadowed.
3679 @item -Wlarger-than=@var{len}
3680 @opindex Wlarger-than=@var{len}
3681 @opindex Wlarger-than-@var{len}
3682 Warn whenever an object of larger than @var{len} bytes is defined.
3684 @item -Wframe-larger-than=@var{len}
3685 @opindex Wframe-larger-than
3686 Warn if the size of a function frame is larger than @var{len} bytes.
3687 The computation done to determine the stack frame size is approximate
3688 and not conservative.
3689 The actual requirements may be somewhat greater than @var{len}
3690 even if you do not get a warning. In addition, any space allocated
3691 via @code{alloca}, variable-length arrays, or related constructs
3692 is not included by the compiler when determining
3693 whether or not to issue a warning.
3695 @item -Wunsafe-loop-optimizations
3696 @opindex Wunsafe-loop-optimizations
3697 @opindex Wno-unsafe-loop-optimizations
3698 Warn if the loop cannot be optimized because the compiler could not
3699 assume anything on the bounds of the loop indices. With
3700 @option{-funsafe-loop-optimizations} warn if the compiler made
3703 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3704 @opindex Wno-pedantic-ms-format
3705 @opindex Wpedantic-ms-format
3706 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3707 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3708 depending on the MS runtime, when you are using the options @option{-Wformat}
3709 and @option{-pedantic} without gnu-extensions.
3711 @item -Wpointer-arith
3712 @opindex Wpointer-arith
3713 @opindex Wno-pointer-arith
3714 Warn about anything that depends on the ``size of'' a function type or
3715 of @code{void}. GNU C assigns these types a size of 1, for
3716 convenience in calculations with @code{void *} pointers and pointers
3717 to functions. In C++, warn also when an arithmetic operation involves
3718 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3721 @opindex Wtype-limits
3722 @opindex Wno-type-limits
3723 Warn if a comparison is always true or always false due to the limited
3724 range of the data type, but do not warn for constant expressions. For
3725 example, warn if an unsigned variable is compared against zero with
3726 @samp{<} or @samp{>=}. This warning is also enabled by
3729 @item -Wbad-function-cast @r{(C and Objective-C only)}
3730 @opindex Wbad-function-cast
3731 @opindex Wno-bad-function-cast
3732 Warn whenever a function call is cast to a non-matching type.
3733 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3735 @item -Wc++-compat @r{(C and Objective-C only)}
3736 Warn about ISO C constructs that are outside of the common subset of
3737 ISO C and ISO C++, e.g.@: request for implicit conversion from
3738 @code{void *} to a pointer to non-@code{void} type.
3740 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3741 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3742 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3743 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3747 @opindex Wno-cast-qual
3748 Warn whenever a pointer is cast so as to remove a type qualifier from
3749 the target type. For example, warn if a @code{const char *} is cast
3750 to an ordinary @code{char *}.
3752 Also warn when making a cast which introduces a type qualifier in an
3753 unsafe way. For example, casting @code{char **} to @code{const char **}
3754 is unsafe, as in this example:
3757 /* p is char ** value. */
3758 const char **q = (const char **) p;
3759 /* Assignment of readonly string to const char * is OK. */
3761 /* Now char** pointer points to read-only memory. */
3766 @opindex Wcast-align
3767 @opindex Wno-cast-align
3768 Warn whenever a pointer is cast such that the required alignment of the
3769 target is increased. For example, warn if a @code{char *} is cast to
3770 an @code{int *} on machines where integers can only be accessed at
3771 two- or four-byte boundaries.
3773 @item -Wwrite-strings
3774 @opindex Wwrite-strings
3775 @opindex Wno-write-strings
3776 When compiling C, give string constants the type @code{const
3777 char[@var{length}]} so that copying the address of one into a
3778 non-@code{const} @code{char *} pointer will get a warning. These
3779 warnings will help you find at compile time code that can try to write
3780 into a string constant, but only if you have been very careful about
3781 using @code{const} in declarations and prototypes. Otherwise, it will
3782 just be a nuisance. This is why we did not make @option{-Wall} request
3785 When compiling C++, warn about the deprecated conversion from string
3786 literals to @code{char *}. This warning is enabled by default for C++
3791 @opindex Wno-clobbered
3792 Warn for variables that might be changed by @samp{longjmp} or
3793 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3796 @opindex Wconversion
3797 @opindex Wno-conversion
3798 Warn for implicit conversions that may alter a value. This includes
3799 conversions between real and integer, like @code{abs (x)} when
3800 @code{x} is @code{double}; conversions between signed and unsigned,
3801 like @code{unsigned ui = -1}; and conversions to smaller types, like
3802 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3803 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3804 changed by the conversion like in @code{abs (2.0)}. Warnings about
3805 conversions between signed and unsigned integers can be disabled by
3806 using @option{-Wno-sign-conversion}.
3808 For C++, also warn for conversions between @code{NULL} and non-pointer
3809 types; confusing overload resolution for user-defined conversions; and
3810 conversions that will never use a type conversion operator:
3811 conversions to @code{void}, the same type, a base class or a reference
3812 to them. Warnings about conversions between signed and unsigned
3813 integers are disabled by default in C++ unless
3814 @option{-Wsign-conversion} is explicitly enabled.
3817 @opindex Wempty-body
3818 @opindex Wno-empty-body
3819 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3820 while} statement. This warning is also enabled by @option{-Wextra}.
3822 @item -Wenum-compare
3823 @opindex Wenum-compare
3824 @opindex Wno-enum-compare
3825 Warn about a comparison between values of different enum types. In C++
3826 this warning is enabled by default. In C this warning is enabled by
3829 @item -Wsign-compare
3830 @opindex Wsign-compare
3831 @opindex Wno-sign-compare
3832 @cindex warning for comparison of signed and unsigned values
3833 @cindex comparison of signed and unsigned values, warning
3834 @cindex signed and unsigned values, comparison warning
3835 Warn when a comparison between signed and unsigned values could produce
3836 an incorrect result when the signed value is converted to unsigned.
3837 This warning is also enabled by @option{-Wextra}; to get the other warnings
3838 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3840 @item -Wsign-conversion
3841 @opindex Wsign-conversion
3842 @opindex Wno-sign-conversion
3843 Warn for implicit conversions that may change the sign of an integer
3844 value, like assigning a signed integer expression to an unsigned
3845 integer variable. An explicit cast silences the warning. In C, this
3846 option is enabled also by @option{-Wconversion}.
3850 @opindex Wno-address
3851 Warn about suspicious uses of memory addresses. These include using
3852 the address of a function in a conditional expression, such as
3853 @code{void func(void); if (func)}, and comparisons against the memory
3854 address of a string literal, such as @code{if (x == "abc")}. Such
3855 uses typically indicate a programmer error: the address of a function
3856 always evaluates to true, so their use in a conditional usually
3857 indicate that the programmer forgot the parentheses in a function
3858 call; and comparisons against string literals result in unspecified
3859 behavior and are not portable in C, so they usually indicate that the
3860 programmer intended to use @code{strcmp}. This warning is enabled by
3864 @opindex Wlogical-op
3865 @opindex Wno-logical-op
3866 Warn about suspicious uses of logical operators in expressions.
3867 This includes using logical operators in contexts where a
3868 bit-wise operator is likely to be expected.
3870 @item -Waggregate-return
3871 @opindex Waggregate-return
3872 @opindex Wno-aggregate-return
3873 Warn if any functions that return structures or unions are defined or
3874 called. (In languages where you can return an array, this also elicits
3877 @item -Wno-attributes
3878 @opindex Wno-attributes
3879 @opindex Wattributes
3880 Do not warn if an unexpected @code{__attribute__} is used, such as
3881 unrecognized attributes, function attributes applied to variables,
3882 etc. This will not stop errors for incorrect use of supported
3885 @item -Wno-builtin-macro-redefined
3886 @opindex Wno-builtin-macro-redefined
3887 @opindex Wbuiltin-macro-redefined
3888 Do not warn if certain built-in macros are redefined. This suppresses
3889 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3890 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3892 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3893 @opindex Wstrict-prototypes
3894 @opindex Wno-strict-prototypes
3895 Warn if a function is declared or defined without specifying the
3896 argument types. (An old-style function definition is permitted without
3897 a warning if preceded by a declaration which specifies the argument
3900 @item -Wold-style-declaration @r{(C and Objective-C only)}
3901 @opindex Wold-style-declaration
3902 @opindex Wno-old-style-declaration
3903 Warn for obsolescent usages, according to the C Standard, in a
3904 declaration. For example, warn if storage-class specifiers like
3905 @code{static} are not the first things in a declaration. This warning
3906 is also enabled by @option{-Wextra}.
3908 @item -Wold-style-definition @r{(C and Objective-C only)}
3909 @opindex Wold-style-definition
3910 @opindex Wno-old-style-definition
3911 Warn if an old-style function definition is used. A warning is given
3912 even if there is a previous prototype.
3914 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3915 @opindex Wmissing-parameter-type
3916 @opindex Wno-missing-parameter-type
3917 A function parameter is declared without a type specifier in K&R-style
3924 This warning is also enabled by @option{-Wextra}.
3926 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3927 @opindex Wmissing-prototypes
3928 @opindex Wno-missing-prototypes
3929 Warn if a global function is defined without a previous prototype
3930 declaration. This warning is issued even if the definition itself
3931 provides a prototype. The aim is to detect global functions that fail
3932 to be declared in header files.
3934 @item -Wmissing-declarations
3935 @opindex Wmissing-declarations
3936 @opindex Wno-missing-declarations
3937 Warn if a global function is defined without a previous declaration.
3938 Do so even if the definition itself provides a prototype.
3939 Use this option to detect global functions that are not declared in
3940 header files. In C++, no warnings are issued for function templates,
3941 or for inline functions, or for functions in anonymous namespaces.
3943 @item -Wmissing-field-initializers
3944 @opindex Wmissing-field-initializers
3945 @opindex Wno-missing-field-initializers
3949 Warn if a structure's initializer has some fields missing. For
3950 example, the following code would cause such a warning, because
3951 @code{x.h} is implicitly zero:
3954 struct s @{ int f, g, h; @};
3955 struct s x = @{ 3, 4 @};
3958 This option does not warn about designated initializers, so the following
3959 modification would not trigger a warning:
3962 struct s @{ int f, g, h; @};
3963 struct s x = @{ .f = 3, .g = 4 @};
3966 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3967 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3969 @item -Wmissing-noreturn
3970 @opindex Wmissing-noreturn
3971 @opindex Wno-missing-noreturn
3972 Warn about functions which might be candidates for attribute @code{noreturn}.
3973 Note these are only possible candidates, not absolute ones. Care should
3974 be taken to manually verify functions actually do not ever return before
3975 adding the @code{noreturn} attribute, otherwise subtle code generation
3976 bugs could be introduced. You will not get a warning for @code{main} in
3977 hosted C environments.
3979 @item -Wmissing-format-attribute
3980 @opindex Wmissing-format-attribute
3981 @opindex Wno-missing-format-attribute
3984 Warn about function pointers which might be candidates for @code{format}
3985 attributes. Note these are only possible candidates, not absolute ones.
3986 GCC will guess that function pointers with @code{format} attributes that
3987 are used in assignment, initialization, parameter passing or return
3988 statements should have a corresponding @code{format} attribute in the
3989 resulting type. I.e.@: the left-hand side of the assignment or
3990 initialization, the type of the parameter variable, or the return type
3991 of the containing function respectively should also have a @code{format}
3992 attribute to avoid the warning.
3994 GCC will also warn about function definitions which might be
3995 candidates for @code{format} attributes. Again, these are only
3996 possible candidates. GCC will guess that @code{format} attributes
3997 might be appropriate for any function that calls a function like
3998 @code{vprintf} or @code{vscanf}, but this might not always be the
3999 case, and some functions for which @code{format} attributes are
4000 appropriate may not be detected.
4002 @item -Wno-multichar
4003 @opindex Wno-multichar
4005 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4006 Usually they indicate a typo in the user's code, as they have
4007 implementation-defined values, and should not be used in portable code.
4009 @item -Wnormalized=<none|id|nfc|nfkc>
4010 @opindex Wnormalized=
4013 @cindex character set, input normalization
4014 In ISO C and ISO C++, two identifiers are different if they are
4015 different sequences of characters. However, sometimes when characters
4016 outside the basic ASCII character set are used, you can have two
4017 different character sequences that look the same. To avoid confusion,
4018 the ISO 10646 standard sets out some @dfn{normalization rules} which
4019 when applied ensure that two sequences that look the same are turned into
4020 the same sequence. GCC can warn you if you are using identifiers which
4021 have not been normalized; this option controls that warning.
4023 There are four levels of warning that GCC supports. The default is
4024 @option{-Wnormalized=nfc}, which warns about any identifier which is
4025 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4026 recommended form for most uses.
4028 Unfortunately, there are some characters which ISO C and ISO C++ allow
4029 in identifiers that when turned into NFC aren't allowable as
4030 identifiers. That is, there's no way to use these symbols in portable
4031 ISO C or C++ and have all your identifiers in NFC@.
4032 @option{-Wnormalized=id} suppresses the warning for these characters.
4033 It is hoped that future versions of the standards involved will correct
4034 this, which is why this option is not the default.
4036 You can switch the warning off for all characters by writing
4037 @option{-Wnormalized=none}. You would only want to do this if you
4038 were using some other normalization scheme (like ``D''), because
4039 otherwise you can easily create bugs that are literally impossible to see.
4041 Some characters in ISO 10646 have distinct meanings but look identical
4042 in some fonts or display methodologies, especially once formatting has
4043 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4044 LETTER N'', will display just like a regular @code{n} which has been
4045 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4046 normalization scheme to convert all these into a standard form as
4047 well, and GCC will warn if your code is not in NFKC if you use
4048 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4049 about every identifier that contains the letter O because it might be
4050 confused with the digit 0, and so is not the default, but may be
4051 useful as a local coding convention if the programming environment is
4052 unable to be fixed to display these characters distinctly.
4054 @item -Wno-deprecated
4055 @opindex Wno-deprecated
4056 @opindex Wdeprecated
4057 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4059 @item -Wno-deprecated-declarations
4060 @opindex Wno-deprecated-declarations
4061 @opindex Wdeprecated-declarations
4062 Do not warn about uses of functions (@pxref{Function Attributes}),
4063 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4064 Attributes}) marked as deprecated by using the @code{deprecated}
4068 @opindex Wno-overflow
4070 Do not warn about compile-time overflow in constant expressions.
4072 @item -Woverride-init @r{(C and Objective-C only)}
4073 @opindex Woverride-init
4074 @opindex Wno-override-init
4078 Warn if an initialized field without side effects is overridden when
4079 using designated initializers (@pxref{Designated Inits, , Designated
4082 This warning is included in @option{-Wextra}. To get other
4083 @option{-Wextra} warnings without this one, use @samp{-Wextra
4084 -Wno-override-init}.
4089 Warn if a structure is given the packed attribute, but the packed
4090 attribute has no effect on the layout or size of the structure.
4091 Such structures may be mis-aligned for little benefit. For
4092 instance, in this code, the variable @code{f.x} in @code{struct bar}
4093 will be misaligned even though @code{struct bar} does not itself
4094 have the packed attribute:
4101 @} __attribute__((packed));
4109 @item -Wpacked-bitfield-compat
4110 @opindex Wpacked-bitfield-compat
4111 @opindex Wno-packed-bitfield-compat
4112 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4113 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4114 the change can lead to differences in the structure layout. GCC
4115 informs you when the offset of such a field has changed in GCC 4.4.
4116 For example there is no longer a 4-bit padding between field @code{a}
4117 and @code{b} in this structure:
4124 @} __attribute__ ((packed));
4127 This warning is enabled by default. Use
4128 @option{-Wno-packed-bitfield-compat} to disable this warning.
4133 Warn if padding is included in a structure, either to align an element
4134 of the structure or to align the whole structure. Sometimes when this
4135 happens it is possible to rearrange the fields of the structure to
4136 reduce the padding and so make the structure smaller.
4138 @item -Wredundant-decls
4139 @opindex Wredundant-decls
4140 @opindex Wno-redundant-decls
4141 Warn if anything is declared more than once in the same scope, even in
4142 cases where multiple declaration is valid and changes nothing.
4144 @item -Wnested-externs @r{(C and Objective-C only)}
4145 @opindex Wnested-externs
4146 @opindex Wno-nested-externs
4147 Warn if an @code{extern} declaration is encountered within a function.
4149 @item -Wunreachable-code
4150 @opindex Wunreachable-code
4151 @opindex Wno-unreachable-code
4152 Warn if the compiler detects that code will never be executed.
4154 This option is intended to warn when the compiler detects that at
4155 least a whole line of source code will never be executed, because
4156 some condition is never satisfied or because it is after a
4157 procedure that never returns.
4159 It is possible for this option to produce a warning even though there
4160 are circumstances under which part of the affected line can be executed,
4161 so care should be taken when removing apparently-unreachable code.
4163 For instance, when a function is inlined, a warning may mean that the
4164 line is unreachable in only one inlined copy of the function.
4166 This option is not made part of @option{-Wall} because in a debugging
4167 version of a program there is often substantial code which checks
4168 correct functioning of the program and is, hopefully, unreachable
4169 because the program does work. Another common use of unreachable
4170 code is to provide behavior which is selectable at compile-time.
4175 Warn if a function can not be inlined and it was declared as inline.
4176 Even with this option, the compiler will not warn about failures to
4177 inline functions declared in system headers.
4179 The compiler uses a variety of heuristics to determine whether or not
4180 to inline a function. For example, the compiler takes into account
4181 the size of the function being inlined and the amount of inlining
4182 that has already been done in the current function. Therefore,
4183 seemingly insignificant changes in the source program can cause the
4184 warnings produced by @option{-Winline} to appear or disappear.
4186 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4187 @opindex Wno-invalid-offsetof
4188 @opindex Winvalid-offsetof
4189 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4190 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4191 to a non-POD type is undefined. In existing C++ implementations,
4192 however, @samp{offsetof} typically gives meaningful results even when
4193 applied to certain kinds of non-POD types. (Such as a simple
4194 @samp{struct} that fails to be a POD type only by virtue of having a
4195 constructor.) This flag is for users who are aware that they are
4196 writing nonportable code and who have deliberately chosen to ignore the
4199 The restrictions on @samp{offsetof} may be relaxed in a future version
4200 of the C++ standard.
4202 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4203 @opindex Wno-int-to-pointer-cast
4204 @opindex Wint-to-pointer-cast
4205 Suppress warnings from casts to pointer type of an integer of a
4208 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4209 @opindex Wno-pointer-to-int-cast
4210 @opindex Wpointer-to-int-cast
4211 Suppress warnings from casts from a pointer to an integer type of a
4215 @opindex Winvalid-pch
4216 @opindex Wno-invalid-pch
4217 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4218 the search path but can't be used.
4222 @opindex Wno-long-long
4223 Warn if @samp{long long} type is used. This is enabled by either
4224 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4225 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4227 @item -Wvariadic-macros
4228 @opindex Wvariadic-macros
4229 @opindex Wno-variadic-macros
4230 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4231 alternate syntax when in pedantic ISO C99 mode. This is default.
4232 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4237 Warn if variable length array is used in the code.
4238 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4239 the variable length array.
4241 @item -Wvolatile-register-var
4242 @opindex Wvolatile-register-var
4243 @opindex Wno-volatile-register-var
4244 Warn if a register variable is declared volatile. The volatile
4245 modifier does not inhibit all optimizations that may eliminate reads
4246 and/or writes to register variables. This warning is enabled by
4249 @item -Wdisabled-optimization
4250 @opindex Wdisabled-optimization
4251 @opindex Wno-disabled-optimization
4252 Warn if a requested optimization pass is disabled. This warning does
4253 not generally indicate that there is anything wrong with your code; it
4254 merely indicates that GCC's optimizers were unable to handle the code
4255 effectively. Often, the problem is that your code is too big or too
4256 complex; GCC will refuse to optimize programs when the optimization
4257 itself is likely to take inordinate amounts of time.
4259 @item -Wpointer-sign @r{(C and Objective-C only)}
4260 @opindex Wpointer-sign
4261 @opindex Wno-pointer-sign
4262 Warn for pointer argument passing or assignment with different signedness.
4263 This option is only supported for C and Objective-C@. It is implied by
4264 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4265 @option{-Wno-pointer-sign}.
4267 @item -Wstack-protector
4268 @opindex Wstack-protector
4269 @opindex Wno-stack-protector
4270 This option is only active when @option{-fstack-protector} is active. It
4271 warns about functions that will not be protected against stack smashing.
4274 @opindex Wno-mudflap
4275 Suppress warnings about constructs that cannot be instrumented by
4278 @item -Woverlength-strings
4279 @opindex Woverlength-strings
4280 @opindex Wno-overlength-strings
4281 Warn about string constants which are longer than the ``minimum
4282 maximum'' length specified in the C standard. Modern compilers
4283 generally allow string constants which are much longer than the
4284 standard's minimum limit, but very portable programs should avoid
4285 using longer strings.
4287 The limit applies @emph{after} string constant concatenation, and does
4288 not count the trailing NUL@. In C89, the limit was 509 characters; in
4289 C99, it was raised to 4095. C++98 does not specify a normative
4290 minimum maximum, so we do not diagnose overlength strings in C++@.
4292 This option is implied by @option{-pedantic}, and can be disabled with
4293 @option{-Wno-overlength-strings}.
4295 @item -Wunsuffixed-float-constants
4296 @opindex Wunsuffixed-float-constants
4298 GCC will issue a warning for any floating constant that does not have
4299 a suffix. When used together with @option{-Wsystem-headers} it will
4300 warn about such constants in system header files. This can be useful
4301 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4302 from the decimal floating-point extension to C99.
4305 @node Debugging Options
4306 @section Options for Debugging Your Program or GCC
4307 @cindex options, debugging
4308 @cindex debugging information options
4310 GCC has various special options that are used for debugging
4311 either your program or GCC:
4316 Produce debugging information in the operating system's native format
4317 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4320 On most systems that use stabs format, @option{-g} enables use of extra
4321 debugging information that only GDB can use; this extra information
4322 makes debugging work better in GDB but will probably make other debuggers
4324 refuse to read the program. If you want to control for certain whether
4325 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4326 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4328 GCC allows you to use @option{-g} with
4329 @option{-O}. The shortcuts taken by optimized code may occasionally
4330 produce surprising results: some variables you declared may not exist
4331 at all; flow of control may briefly move where you did not expect it;
4332 some statements may not be executed because they compute constant
4333 results or their values were already at hand; some statements may
4334 execute in different places because they were moved out of loops.
4336 Nevertheless it proves possible to debug optimized output. This makes
4337 it reasonable to use the optimizer for programs that might have bugs.
4339 The following options are useful when GCC is generated with the
4340 capability for more than one debugging format.
4344 Produce debugging information for use by GDB@. This means to use the
4345 most expressive format available (DWARF 2, stabs, or the native format
4346 if neither of those are supported), including GDB extensions if at all
4351 Produce debugging information in stabs format (if that is supported),
4352 without GDB extensions. This is the format used by DBX on most BSD
4353 systems. On MIPS, Alpha and System V Release 4 systems this option
4354 produces stabs debugging output which is not understood by DBX or SDB@.
4355 On System V Release 4 systems this option requires the GNU assembler.
4357 @item -feliminate-unused-debug-symbols
4358 @opindex feliminate-unused-debug-symbols
4359 Produce debugging information in stabs format (if that is supported),
4360 for only symbols that are actually used.
4362 @item -femit-class-debug-always
4363 Instead of emitting debugging information for a C++ class in only one
4364 object file, emit it in all object files using the class. This option
4365 should be used only with debuggers that are unable to handle the way GCC
4366 normally emits debugging information for classes because using this
4367 option will increase the size of debugging information by as much as a
4372 Produce debugging information in stabs format (if that is supported),
4373 using GNU extensions understood only by the GNU debugger (GDB)@. The
4374 use of these extensions is likely to make other debuggers crash or
4375 refuse to read the program.
4379 Produce debugging information in COFF format (if that is supported).
4380 This is the format used by SDB on most System V systems prior to
4385 Produce debugging information in XCOFF format (if that is supported).
4386 This is the format used by the DBX debugger on IBM RS/6000 systems.
4390 Produce debugging information in XCOFF format (if that is supported),
4391 using GNU extensions understood only by the GNU debugger (GDB)@. The
4392 use of these extensions is likely to make other debuggers crash or
4393 refuse to read the program, and may cause assemblers other than the GNU
4394 assembler (GAS) to fail with an error.
4396 @item -gdwarf-@var{version}
4397 @opindex gdwarf-@var{version}
4398 Produce debugging information in DWARF format (if that is
4399 supported). This is the format used by DBX on IRIX 6. The value
4400 of @var{version} may be either 2 or 3; the default version is 2.
4402 Note that with DWARF version 2 some ports require, and will always
4403 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4407 Produce debugging information in VMS debug format (if that is
4408 supported). This is the format used by DEBUG on VMS systems.
4411 @itemx -ggdb@var{level}
4412 @itemx -gstabs@var{level}
4413 @itemx -gcoff@var{level}
4414 @itemx -gxcoff@var{level}
4415 @itemx -gvms@var{level}
4416 Request debugging information and also use @var{level} to specify how
4417 much information. The default level is 2.
4419 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4422 Level 1 produces minimal information, enough for making backtraces in
4423 parts of the program that you don't plan to debug. This includes
4424 descriptions of functions and external variables, but no information
4425 about local variables and no line numbers.
4427 Level 3 includes extra information, such as all the macro definitions
4428 present in the program. Some debuggers support macro expansion when
4429 you use @option{-g3}.
4431 @option{-gdwarf-2} does not accept a concatenated debug level, because
4432 GCC used to support an option @option{-gdwarf} that meant to generate
4433 debug information in version 1 of the DWARF format (which is very
4434 different from version 2), and it would have been too confusing. That
4435 debug format is long obsolete, but the option cannot be changed now.
4436 Instead use an additional @option{-g@var{level}} option to change the
4437 debug level for DWARF.
4441 Turn off generation of debug info, if leaving out this option would have
4442 generated it, or turn it on at level 2 otherwise. The position of this
4443 argument in the command line does not matter, it takes effect after all
4444 other options are processed, and it does so only once, no matter how
4445 many times it is given. This is mainly intended to be used with
4446 @option{-fcompare-debug}.
4448 @item -fdump-final-insns=@var{file}
4449 @opindex fdump-final-insns=
4450 Dump the final internal representation (RTL) to @var{file}.
4452 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4453 @opindex fcompare-debug
4454 @opindex fno-compare-debug
4455 If no error occurs during compilation, run the compiler a second time,
4456 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4457 passed to the second compilation. Dump the final internal
4458 representation in both compilations, and print an error if they differ.
4460 If the equal sign is omitted, the default @option{-gtoggle} is used.
4462 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4463 and nonzero, implicitly enables @option{-fcompare-debug}. If
4464 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4465 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4468 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4469 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4470 of the final representation and the second compilation, preventing even
4471 @env{GCC_COMPARE_DEBUG} from taking effect.
4473 To verify full coverage during @option{-fcompare-debug} testing, set
4474 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4475 which GCC will reject as an invalid option in any actual compilation
4476 (rather than preprocessing, assembly or linking). To get just a
4477 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4478 not overridden} will do.
4480 @item -fcompare-debug-second
4481 @opindex fcompare-debug-second
4482 This option is implicitly passed to the compiler for the second
4483 compilation requested by @option{-fcompare-debug}, along with options to
4484 silence warnings, and omitting other options that would cause
4485 side-effect compiler outputs to files or to the standard output. Dump
4486 files and preserved temporary files are renamed so as to contain the
4487 @code{.gk} additional extension during the second compilation, to avoid
4488 overwriting those generated by the first.
4490 When this option is passed to the compiler driver, it causes the
4491 @emph{first} compilation to be skipped, which makes it useful for little
4492 other than debugging the compiler proper.
4494 @item -feliminate-dwarf2-dups
4495 @opindex feliminate-dwarf2-dups
4496 Compress DWARF2 debugging information by eliminating duplicated
4497 information about each symbol. This option only makes sense when
4498 generating DWARF2 debugging information with @option{-gdwarf-2}.
4500 @item -femit-struct-debug-baseonly
4501 Emit debug information for struct-like types
4502 only when the base name of the compilation source file
4503 matches the base name of file in which the struct was defined.
4505 This option substantially reduces the size of debugging information,
4506 but at significant potential loss in type information to the debugger.
4507 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4508 See @option{-femit-struct-debug-detailed} for more detailed control.
4510 This option works only with DWARF 2.
4512 @item -femit-struct-debug-reduced
4513 Emit debug information for struct-like types
4514 only when the base name of the compilation source file
4515 matches the base name of file in which the type was defined,
4516 unless the struct is a template or defined in a system header.
4518 This option significantly reduces the size of debugging information,
4519 with some potential loss in type information to the debugger.
4520 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4521 See @option{-femit-struct-debug-detailed} for more detailed control.
4523 This option works only with DWARF 2.
4525 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4526 Specify the struct-like types
4527 for which the compiler will generate debug information.
4528 The intent is to reduce duplicate struct debug information
4529 between different object files within the same program.
4531 This option is a detailed version of
4532 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4533 which will serve for most needs.
4535 A specification has the syntax
4536 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4538 The optional first word limits the specification to
4539 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4540 A struct type is used directly when it is the type of a variable, member.
4541 Indirect uses arise through pointers to structs.
4542 That is, when use of an incomplete struct would be legal, the use is indirect.
4544 @samp{struct one direct; struct two * indirect;}.
4546 The optional second word limits the specification to
4547 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4548 Generic structs are a bit complicated to explain.
4549 For C++, these are non-explicit specializations of template classes,
4550 or non-template classes within the above.
4551 Other programming languages have generics,
4552 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4554 The third word specifies the source files for those
4555 structs for which the compiler will emit debug information.
4556 The values @samp{none} and @samp{any} have the normal meaning.
4557 The value @samp{base} means that
4558 the base of name of the file in which the type declaration appears
4559 must match the base of the name of the main compilation file.
4560 In practice, this means that
4561 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4562 but types declared in other header will not.
4563 The value @samp{sys} means those types satisfying @samp{base}
4564 or declared in system or compiler headers.
4566 You may need to experiment to determine the best settings for your application.
4568 The default is @samp{-femit-struct-debug-detailed=all}.
4570 This option works only with DWARF 2.
4572 @item -fno-merge-debug-strings
4573 @opindex fmerge-debug-strings
4574 @opindex fno-merge-debug-strings
4575 Direct the linker to not merge together strings in the debugging
4576 information which are identical in different object files. Merging is
4577 not supported by all assemblers or linkers. Merging decreases the size
4578 of the debug information in the output file at the cost of increasing
4579 link processing time. Merging is enabled by default.
4581 @item -fdebug-prefix-map=@var{old}=@var{new}
4582 @opindex fdebug-prefix-map
4583 When compiling files in directory @file{@var{old}}, record debugging
4584 information describing them as in @file{@var{new}} instead.
4586 @item -fno-dwarf2-cfi-asm
4587 @opindex fdwarf2-cfi-asm
4588 @opindex fno-dwarf2-cfi-asm
4589 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4590 instead of using GAS @code{.cfi_*} directives.
4592 @cindex @command{prof}
4595 Generate extra code to write profile information suitable for the
4596 analysis program @command{prof}. You must use this option when compiling
4597 the source files you want data about, and you must also use it when
4600 @cindex @command{gprof}
4603 Generate extra code to write profile information suitable for the
4604 analysis program @command{gprof}. You must use this option when compiling
4605 the source files you want data about, and you must also use it when
4610 Makes the compiler print out each function name as it is compiled, and
4611 print some statistics about each pass when it finishes.
4614 @opindex ftime-report
4615 Makes the compiler print some statistics about the time consumed by each
4616 pass when it finishes.
4619 @opindex fmem-report
4620 Makes the compiler print some statistics about permanent memory
4621 allocation when it finishes.
4623 @item -fpre-ipa-mem-report
4624 @opindex fpre-ipa-mem-report
4625 @item -fpost-ipa-mem-report
4626 @opindex fpost-ipa-mem-report
4627 Makes the compiler print some statistics about permanent memory
4628 allocation before or after interprocedural optimization.
4630 @item -fprofile-arcs
4631 @opindex fprofile-arcs
4632 Add code so that program flow @dfn{arcs} are instrumented. During
4633 execution the program records how many times each branch and call is
4634 executed and how many times it is taken or returns. When the compiled
4635 program exits it saves this data to a file called
4636 @file{@var{auxname}.gcda} for each source file. The data may be used for
4637 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4638 test coverage analysis (@option{-ftest-coverage}). Each object file's
4639 @var{auxname} is generated from the name of the output file, if
4640 explicitly specified and it is not the final executable, otherwise it is
4641 the basename of the source file. In both cases any suffix is removed
4642 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4643 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4644 @xref{Cross-profiling}.
4646 @cindex @command{gcov}
4650 This option is used to compile and link code instrumented for coverage
4651 analysis. The option is a synonym for @option{-fprofile-arcs}
4652 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4653 linking). See the documentation for those options for more details.
4658 Compile the source files with @option{-fprofile-arcs} plus optimization
4659 and code generation options. For test coverage analysis, use the
4660 additional @option{-ftest-coverage} option. You do not need to profile
4661 every source file in a program.
4664 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4665 (the latter implies the former).
4668 Run the program on a representative workload to generate the arc profile
4669 information. This may be repeated any number of times. You can run
4670 concurrent instances of your program, and provided that the file system
4671 supports locking, the data files will be correctly updated. Also
4672 @code{fork} calls are detected and correctly handled (double counting
4676 For profile-directed optimizations, compile the source files again with
4677 the same optimization and code generation options plus
4678 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4679 Control Optimization}).
4682 For test coverage analysis, use @command{gcov} to produce human readable
4683 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4684 @command{gcov} documentation for further information.
4688 With @option{-fprofile-arcs}, for each function of your program GCC
4689 creates a program flow graph, then finds a spanning tree for the graph.
4690 Only arcs that are not on the spanning tree have to be instrumented: the
4691 compiler adds code to count the number of times that these arcs are
4692 executed. When an arc is the only exit or only entrance to a block, the
4693 instrumentation code can be added to the block; otherwise, a new basic
4694 block must be created to hold the instrumentation code.
4697 @item -ftest-coverage
4698 @opindex ftest-coverage
4699 Produce a notes file that the @command{gcov} code-coverage utility
4700 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4701 show program coverage. Each source file's note file is called
4702 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4703 above for a description of @var{auxname} and instructions on how to
4704 generate test coverage data. Coverage data will match the source files
4705 more closely, if you do not optimize.
4707 @item -fdbg-cnt-list
4708 @opindex fdbg-cnt-list
4709 Print the name and the counter upperbound for all debug counters.
4711 @item -fdbg-cnt=@var{counter-value-list}
4713 Set the internal debug counter upperbound. @var{counter-value-list}
4714 is a comma-separated list of @var{name}:@var{value} pairs
4715 which sets the upperbound of each debug counter @var{name} to @var{value}.
4716 All debug counters have the initial upperbound of @var{UINT_MAX},
4717 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4718 e.g. With -fdbg-cnt=dce:10,tail_call:0
4719 dbg_cnt(dce) will return true only for first 10 invocations
4720 and dbg_cnt(tail_call) will return false always.
4722 @item -d@var{letters}
4723 @itemx -fdump-rtl-@var{pass}
4725 Says to make debugging dumps during compilation at times specified by
4726 @var{letters}. This is used for debugging the RTL-based passes of the
4727 compiler. The file names for most of the dumps are made by appending
4728 a pass number and a word to the @var{dumpname}, and the files are
4729 created in the directory of the output file. @var{dumpname} is
4730 generated from the name of the output file, if explicitly specified
4731 and it is not an executable, otherwise it is the basename of the
4732 source file. These switches may have different effects when
4733 @option{-E} is used for preprocessing.
4735 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4736 @option{-d} option @var{letters}. Here are the possible
4737 letters for use in @var{pass} and @var{letters}, and their meanings:
4741 @item -fdump-rtl-alignments
4742 @opindex fdump-rtl-alignments
4743 Dump after branch alignments have been computed.
4745 @item -fdump-rtl-asmcons
4746 @opindex fdump-rtl-asmcons
4747 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4749 @item -fdump-rtl-auto_inc_dec
4750 @opindex fdump-rtl-auto_inc_dec
4751 Dump after auto-inc-dec discovery. This pass is only run on
4752 architectures that have auto inc or auto dec instructions.
4754 @item -fdump-rtl-barriers
4755 @opindex fdump-rtl-barriers
4756 Dump after cleaning up the barrier instructions.
4758 @item -fdump-rtl-bbpart
4759 @opindex fdump-rtl-bbpart
4760 Dump after partitioning hot and cold basic blocks.
4762 @item -fdump-rtl-bbro
4763 @opindex fdump-rtl-bbro
4764 Dump after block reordering.
4766 @item -fdump-rtl-btl1
4767 @itemx -fdump-rtl-btl2
4768 @opindex fdump-rtl-btl2
4769 @opindex fdump-rtl-btl2
4770 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4771 after the two branch
4772 target load optimization passes.
4774 @item -fdump-rtl-bypass
4775 @opindex fdump-rtl-bypass
4776 Dump after jump bypassing and control flow optimizations.
4778 @item -fdump-rtl-combine
4779 @opindex fdump-rtl-combine
4780 Dump after the RTL instruction combination pass.
4782 @item -fdump-rtl-compgotos
4783 @opindex fdump-rtl-compgotos
4784 Dump after duplicating the computed gotos.
4786 @item -fdump-rtl-ce1
4787 @itemx -fdump-rtl-ce2
4788 @itemx -fdump-rtl-ce3
4789 @opindex fdump-rtl-ce1
4790 @opindex fdump-rtl-ce2
4791 @opindex fdump-rtl-ce3
4792 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4793 @option{-fdump-rtl-ce3} enable dumping after the three
4794 if conversion passes.
4796 @itemx -fdump-rtl-cprop_hardreg
4797 @opindex fdump-rtl-cprop_hardreg
4798 Dump after hard register copy propagation.
4800 @itemx -fdump-rtl-csa
4801 @opindex fdump-rtl-csa
4802 Dump after combining stack adjustments.
4804 @item -fdump-rtl-cse1
4805 @itemx -fdump-rtl-cse2
4806 @opindex fdump-rtl-cse1
4807 @opindex fdump-rtl-cse2
4808 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4809 the two common sub-expression elimination passes.
4811 @itemx -fdump-rtl-dce
4812 @opindex fdump-rtl-dce
4813 Dump after the standalone dead code elimination passes.
4815 @itemx -fdump-rtl-dbr
4816 @opindex fdump-rtl-dbr
4817 Dump after delayed branch scheduling.
4819 @item -fdump-rtl-dce1
4820 @itemx -fdump-rtl-dce2
4821 @opindex fdump-rtl-dce1
4822 @opindex fdump-rtl-dce2
4823 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4824 the two dead store elimination passes.
4827 @opindex fdump-rtl-eh
4828 Dump after finalization of EH handling code.
4830 @item -fdump-rtl-eh_ranges
4831 @opindex fdump-rtl-eh_ranges
4832 Dump after conversion of EH handling range regions.
4834 @item -fdump-rtl-expand
4835 @opindex fdump-rtl-expand
4836 Dump after RTL generation.
4838 @item -fdump-rtl-fwprop1
4839 @itemx -fdump-rtl-fwprop2
4840 @opindex fdump-rtl-fwprop1
4841 @opindex fdump-rtl-fwprop2
4842 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4843 dumping after the two forward propagation passes.
4845 @item -fdump-rtl-gcse1
4846 @itemx -fdump-rtl-gcse2
4847 @opindex fdump-rtl-gcse1
4848 @opindex fdump-rtl-gcse2
4849 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4850 after global common subexpression elimination.
4852 @item -fdump-rtl-init-regs
4853 @opindex fdump-rtl-init-regs
4854 Dump after the initialization of the registers.
4856 @item -fdump-rtl-initvals
4857 @opindex fdump-rtl-initvals
4858 Dump after the computation of the initial value sets.
4860 @itemx -fdump-rtl-into_cfglayout
4861 @opindex fdump-rtl-into_cfglayout
4862 Dump after converting to cfglayout mode.
4864 @item -fdump-rtl-ira
4865 @opindex fdump-rtl-ira
4866 Dump after iterated register allocation.
4868 @item -fdump-rtl-jump
4869 @opindex fdump-rtl-jump
4870 Dump after the second jump optimization.
4872 @item -fdump-rtl-loop2
4873 @opindex fdump-rtl-loop2
4874 @option{-fdump-rtl-loop2} enables dumping after the rtl
4875 loop optimization passes.
4877 @item -fdump-rtl-mach
4878 @opindex fdump-rtl-mach
4879 Dump after performing the machine dependent reorganization pass, if that
4882 @item -fdump-rtl-mode_sw
4883 @opindex fdump-rtl-mode_sw
4884 Dump after removing redundant mode switches.
4886 @item -fdump-rtl-rnreg
4887 @opindex fdump-rtl-rnreg
4888 Dump after register renumbering.
4890 @itemx -fdump-rtl-outof_cfglayout
4891 @opindex fdump-rtl-outof_cfglayout
4892 Dump after converting from cfglayout mode.
4894 @item -fdump-rtl-peephole2
4895 @opindex fdump-rtl-peephole2
4896 Dump after the peephole pass.
4898 @item -fdump-rtl-postreload
4899 @opindex fdump-rtl-postreload
4900 Dump after post-reload optimizations.
4902 @itemx -fdump-rtl-pro_and_epilogue
4903 @opindex fdump-rtl-pro_and_epilogue
4904 Dump after generating the function pro and epilogues.
4906 @item -fdump-rtl-regmove
4907 @opindex fdump-rtl-regmove
4908 Dump after the register move pass.
4910 @item -fdump-rtl-sched1
4911 @itemx -fdump-rtl-sched2
4912 @opindex fdump-rtl-sched1
4913 @opindex fdump-rtl-sched2
4914 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4915 after the basic block scheduling passes.
4917 @item -fdump-rtl-see
4918 @opindex fdump-rtl-see
4919 Dump after sign extension elimination.
4921 @item -fdump-rtl-seqabstr
4922 @opindex fdump-rtl-seqabstr
4923 Dump after common sequence discovery.
4925 @item -fdump-rtl-shorten
4926 @opindex fdump-rtl-shorten
4927 Dump after shortening branches.
4929 @item -fdump-rtl-sibling
4930 @opindex fdump-rtl-sibling
4931 Dump after sibling call optimizations.
4933 @item -fdump-rtl-split1
4934 @itemx -fdump-rtl-split2
4935 @itemx -fdump-rtl-split3
4936 @itemx -fdump-rtl-split4
4937 @itemx -fdump-rtl-split5
4938 @opindex fdump-rtl-split1
4939 @opindex fdump-rtl-split2
4940 @opindex fdump-rtl-split3
4941 @opindex fdump-rtl-split4
4942 @opindex fdump-rtl-split5
4943 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4944 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4945 @option{-fdump-rtl-split5} enable dumping after five rounds of
4946 instruction splitting.
4948 @item -fdump-rtl-sms
4949 @opindex fdump-rtl-sms
4950 Dump after modulo scheduling. This pass is only run on some
4953 @item -fdump-rtl-stack
4954 @opindex fdump-rtl-stack
4955 Dump after conversion from GCC's "flat register file" registers to the
4956 x87's stack-like registers. This pass is only run on x86 variants.
4958 @item -fdump-rtl-subreg1
4959 @itemx -fdump-rtl-subreg2
4960 @opindex fdump-rtl-subreg1
4961 @opindex fdump-rtl-subreg2
4962 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4963 the two subreg expansion passes.
4965 @item -fdump-rtl-unshare
4966 @opindex fdump-rtl-unshare
4967 Dump after all rtl has been unshared.
4969 @item -fdump-rtl-vartrack
4970 @opindex fdump-rtl-vartrack
4971 Dump after variable tracking.
4973 @item -fdump-rtl-vregs
4974 @opindex fdump-rtl-vregs
4975 Dump after converting virtual registers to hard registers.
4977 @item -fdump-rtl-web
4978 @opindex fdump-rtl-web
4979 Dump after live range splitting.
4981 @item -fdump-rtl-regclass
4982 @itemx -fdump-rtl-subregs_of_mode_init
4983 @itemx -fdump-rtl-subregs_of_mode_finish
4984 @itemx -fdump-rtl-dfinit
4985 @itemx -fdump-rtl-dfinish
4986 @opindex fdump-rtl-regclass
4987 @opindex fdump-rtl-subregs_of_mode_init
4988 @opindex fdump-rtl-subregs_of_mode_finish
4989 @opindex fdump-rtl-dfinit
4990 @opindex fdump-rtl-dfinish
4991 These dumps are defined but always produce empty files.
4993 @item -fdump-rtl-all
4994 @opindex fdump-rtl-all
4995 Produce all the dumps listed above.
4999 Annotate the assembler output with miscellaneous debugging information.
5003 Dump all macro definitions, at the end of preprocessing, in addition to
5008 Produce a core dump whenever an error occurs.
5012 Print statistics on memory usage, at the end of the run, to
5017 Annotate the assembler output with a comment indicating which
5018 pattern and alternative was used. The length of each instruction is
5023 Dump the RTL in the assembler output as a comment before each instruction.
5024 Also turns on @option{-dp} annotation.
5028 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5029 dump a representation of the control flow graph suitable for viewing with VCG
5030 to @file{@var{file}.@var{pass}.vcg}.
5034 Just generate RTL for a function instead of compiling it. Usually used
5035 with @option{-fdump-rtl-expand}.
5039 Dump debugging information during parsing, to standard error.
5043 @opindex fdump-noaddr
5044 When doing debugging dumps, suppress address output. This makes it more
5045 feasible to use diff on debugging dumps for compiler invocations with
5046 different compiler binaries and/or different
5047 text / bss / data / heap / stack / dso start locations.
5049 @item -fdump-unnumbered
5050 @opindex fdump-unnumbered
5051 When doing debugging dumps, suppress instruction numbers and address output.
5052 This makes it more feasible to use diff on debugging dumps for compiler
5053 invocations with different options, in particular with and without
5056 @item -fdump-unnumbered-links
5057 @opindex fdump-unnumbered-links
5058 When doing debugging dumps (see @option{-d} option above), suppress
5059 instruction numbers for the links to the previous and next instructions
5062 @item -fdump-translation-unit @r{(C++ only)}
5063 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5064 @opindex fdump-translation-unit
5065 Dump a representation of the tree structure for the entire translation
5066 unit to a file. The file name is made by appending @file{.tu} to the
5067 source file name, and the file is created in the same directory as the
5068 output file. If the @samp{-@var{options}} form is used, @var{options}
5069 controls the details of the dump as described for the
5070 @option{-fdump-tree} options.
5072 @item -fdump-class-hierarchy @r{(C++ only)}
5073 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5074 @opindex fdump-class-hierarchy
5075 Dump a representation of each class's hierarchy and virtual function
5076 table layout to a file. The file name is made by appending
5077 @file{.class} to the source file name, and the file is created in the
5078 same directory as the output file. If the @samp{-@var{options}} form
5079 is used, @var{options} controls the details of the dump as described
5080 for the @option{-fdump-tree} options.
5082 @item -fdump-ipa-@var{switch}
5084 Control the dumping at various stages of inter-procedural analysis
5085 language tree to a file. The file name is generated by appending a
5086 switch specific suffix to the source file name, and the file is created
5087 in the same directory as the output file. The following dumps are
5092 Enables all inter-procedural analysis dumps.
5095 Dumps information about call-graph optimization, unused function removal,
5096 and inlining decisions.
5099 Dump after function inlining.
5103 @item -fdump-statistics-@var{option}
5104 @opindex fdump-statistics
5105 Enable and control dumping of pass statistics in a separate file. The
5106 file name is generated by appending a suffix ending in
5107 @samp{.statistics} to the source file name, and the file is created in
5108 the same directory as the output file. If the @samp{-@var{option}}
5109 form is used, @samp{-stats} will cause counters to be summed over the
5110 whole compilation unit while @samp{-details} will dump every event as
5111 the passes generate them. The default with no option is to sum
5112 counters for each function compiled.
5114 @item -fdump-tree-@var{switch}
5115 @itemx -fdump-tree-@var{switch}-@var{options}
5117 Control the dumping at various stages of processing the intermediate
5118 language tree to a file. The file name is generated by appending a
5119 switch specific suffix to the source file name, and the file is
5120 created in the same directory as the output file. If the
5121 @samp{-@var{options}} form is used, @var{options} is a list of
5122 @samp{-} separated options that control the details of the dump. Not
5123 all options are applicable to all dumps, those which are not
5124 meaningful will be ignored. The following options are available
5128 Print the address of each node. Usually this is not meaningful as it
5129 changes according to the environment and source file. Its primary use
5130 is for tying up a dump file with a debug environment.
5132 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5133 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5134 use working backward from mangled names in the assembly file.
5136 Inhibit dumping of members of a scope or body of a function merely
5137 because that scope has been reached. Only dump such items when they
5138 are directly reachable by some other path. When dumping pretty-printed
5139 trees, this option inhibits dumping the bodies of control structures.
5141 Print a raw representation of the tree. By default, trees are
5142 pretty-printed into a C-like representation.
5144 Enable more detailed dumps (not honored by every dump option).
5146 Enable dumping various statistics about the pass (not honored by every dump
5149 Enable showing basic block boundaries (disabled in raw dumps).
5151 Enable showing virtual operands for every statement.
5153 Enable showing line numbers for statements.
5155 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5157 Enable showing the tree dump for each statement.
5159 Enable showing the EH region number holding each statement.
5161 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5162 and @option{lineno}.
5165 The following tree dumps are possible:
5169 @opindex fdump-tree-original
5170 Dump before any tree based optimization, to @file{@var{file}.original}.
5173 @opindex fdump-tree-optimized
5174 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5177 @opindex fdump-tree-gimple
5178 Dump each function before and after the gimplification pass to a file. The
5179 file name is made by appending @file{.gimple} to the source file name.
5182 @opindex fdump-tree-cfg
5183 Dump the control flow graph of each function to a file. The file name is
5184 made by appending @file{.cfg} to the source file name.
5187 @opindex fdump-tree-vcg
5188 Dump the control flow graph of each function to a file in VCG format. The
5189 file name is made by appending @file{.vcg} to the source file name. Note
5190 that if the file contains more than one function, the generated file cannot
5191 be used directly by VCG@. You will need to cut and paste each function's
5192 graph into its own separate file first.
5195 @opindex fdump-tree-ch
5196 Dump each function after copying loop headers. The file name is made by
5197 appending @file{.ch} to the source file name.
5200 @opindex fdump-tree-ssa
5201 Dump SSA related information to a file. The file name is made by appending
5202 @file{.ssa} to the source file name.
5205 @opindex fdump-tree-alias
5206 Dump aliasing information for each function. The file name is made by
5207 appending @file{.alias} to the source file name.
5210 @opindex fdump-tree-ccp
5211 Dump each function after CCP@. The file name is made by appending
5212 @file{.ccp} to the source file name.
5215 @opindex fdump-tree-storeccp
5216 Dump each function after STORE-CCP@. The file name is made by appending
5217 @file{.storeccp} to the source file name.
5220 @opindex fdump-tree-pre
5221 Dump trees after partial redundancy elimination. The file name is made
5222 by appending @file{.pre} to the source file name.
5225 @opindex fdump-tree-fre
5226 Dump trees after full redundancy elimination. The file name is made
5227 by appending @file{.fre} to the source file name.
5230 @opindex fdump-tree-copyprop
5231 Dump trees after copy propagation. The file name is made
5232 by appending @file{.copyprop} to the source file name.
5234 @item store_copyprop
5235 @opindex fdump-tree-store_copyprop
5236 Dump trees after store copy-propagation. The file name is made
5237 by appending @file{.store_copyprop} to the source file name.
5240 @opindex fdump-tree-dce
5241 Dump each function after dead code elimination. The file name is made by
5242 appending @file{.dce} to the source file name.
5245 @opindex fdump-tree-mudflap
5246 Dump each function after adding mudflap instrumentation. The file name is
5247 made by appending @file{.mudflap} to the source file name.
5250 @opindex fdump-tree-sra
5251 Dump each function after performing scalar replacement of aggregates. The
5252 file name is made by appending @file{.sra} to the source file name.
5255 @opindex fdump-tree-sink
5256 Dump each function after performing code sinking. The file name is made
5257 by appending @file{.sink} to the source file name.
5260 @opindex fdump-tree-dom
5261 Dump each function after applying dominator tree optimizations. The file
5262 name is made by appending @file{.dom} to the source file name.
5265 @opindex fdump-tree-dse
5266 Dump each function after applying dead store elimination. The file
5267 name is made by appending @file{.dse} to the source file name.
5270 @opindex fdump-tree-phiopt
5271 Dump each function after optimizing PHI nodes into straightline code. The file
5272 name is made by appending @file{.phiopt} to the source file name.
5275 @opindex fdump-tree-forwprop
5276 Dump each function after forward propagating single use variables. The file
5277 name is made by appending @file{.forwprop} to the source file name.
5280 @opindex fdump-tree-copyrename
5281 Dump each function after applying the copy rename optimization. The file
5282 name is made by appending @file{.copyrename} to the source file name.
5285 @opindex fdump-tree-nrv
5286 Dump each function after applying the named return value optimization on
5287 generic trees. The file name is made by appending @file{.nrv} to the source
5291 @opindex fdump-tree-vect
5292 Dump each function after applying vectorization of loops. The file name is
5293 made by appending @file{.vect} to the source file name.
5296 @opindex fdump-tree-vrp
5297 Dump each function after Value Range Propagation (VRP). The file name
5298 is made by appending @file{.vrp} to the source file name.
5301 @opindex fdump-tree-all
5302 Enable all the available tree dumps with the flags provided in this option.
5305 @item -ftree-vectorizer-verbose=@var{n}
5306 @opindex ftree-vectorizer-verbose
5307 This option controls the amount of debugging output the vectorizer prints.
5308 This information is written to standard error, unless
5309 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5310 in which case it is output to the usual dump listing file, @file{.vect}.
5311 For @var{n}=0 no diagnostic information is reported.
5312 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5313 and the total number of loops that got vectorized.
5314 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5315 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5316 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5317 level that @option{-fdump-tree-vect-stats} uses.
5318 Higher verbosity levels mean either more information dumped for each
5319 reported loop, or same amount of information reported for more loops:
5320 If @var{n}=3, alignment related information is added to the reports.
5321 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5322 memory access-patterns) is added to the reports.
5323 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5324 that did not pass the first analysis phase (i.e., may not be countable, or
5325 may have complicated control-flow).
5326 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5327 For @var{n}=7, all the information the vectorizer generates during its
5328 analysis and transformation is reported. This is the same verbosity level
5329 that @option{-fdump-tree-vect-details} uses.
5331 @item -frandom-seed=@var{string}
5332 @opindex frandom-seed
5333 This option provides a seed that GCC uses when it would otherwise use
5334 random numbers. It is used to generate certain symbol names
5335 that have to be different in every compiled file. It is also used to
5336 place unique stamps in coverage data files and the object files that
5337 produce them. You can use the @option{-frandom-seed} option to produce
5338 reproducibly identical object files.
5340 The @var{string} should be different for every file you compile.
5342 @item -fsched-verbose=@var{n}
5343 @opindex fsched-verbose
5344 On targets that use instruction scheduling, this option controls the
5345 amount of debugging output the scheduler prints. This information is
5346 written to standard error, unless @option{-fdump-rtl-sched1} or
5347 @option{-fdump-rtl-sched2} is specified, in which case it is output
5348 to the usual dump listing file, @file{.sched} or @file{.sched2}
5349 respectively. However for @var{n} greater than nine, the output is
5350 always printed to standard error.
5352 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5353 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5354 For @var{n} greater than one, it also output basic block probabilities,
5355 detailed ready list information and unit/insn info. For @var{n} greater
5356 than two, it includes RTL at abort point, control-flow and regions info.
5357 And for @var{n} over four, @option{-fsched-verbose} also includes
5361 @itemx -save-temps=cwd
5363 Store the usual ``temporary'' intermediate files permanently; place them
5364 in the current directory and name them based on the source file. Thus,
5365 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5366 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5367 preprocessed @file{foo.i} output file even though the compiler now
5368 normally uses an integrated preprocessor.
5370 When used in combination with the @option{-x} command line option,
5371 @option{-save-temps} is sensible enough to avoid over writing an
5372 input source file with the same extension as an intermediate file.
5373 The corresponding intermediate file may be obtained by renaming the
5374 source file before using @option{-save-temps}.
5376 If you invoke GCC in parallel, compiling several different source
5377 files that share a common base name in different subdirectories or the
5378 same source file compiled for multiple output destinations, it is
5379 likely that the different parallel compilers will interfere with each
5380 other, and overwrite the temporary files. For instance:
5383 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5384 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5387 may result in @file{foo.i} and @file{foo.o} being written to
5388 simultaneously by both compilers.
5390 @item -save-temps=obj
5391 @opindex save-temps=obj
5392 Store the usual ``temporary'' intermediate files permanently. If the
5393 @option{-o} option is used, the temporary files are based on the
5394 object file. If the @option{-o} option is not used, the
5395 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5400 gcc -save-temps=obj -c foo.c
5401 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5402 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5405 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5406 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5407 @file{dir2/yfoobar.o}.
5409 @item -time@r{[}=@var{file}@r{]}
5411 Report the CPU time taken by each subprocess in the compilation
5412 sequence. For C source files, this is the compiler proper and assembler
5413 (plus the linker if linking is done).
5415 Without the specification of an output file, the output looks like this:
5422 The first number on each line is the ``user time'', that is time spent
5423 executing the program itself. The second number is ``system time'',
5424 time spent executing operating system routines on behalf of the program.
5425 Both numbers are in seconds.
5427 With the specification of an output file, the output is appended to the
5428 named file, and it looks like this:
5431 0.12 0.01 cc1 @var{options}
5432 0.00 0.01 as @var{options}
5435 The ``user time'' and the ``system time'' are moved before the program
5436 name, and the options passed to the program are displayed, so that one
5437 can later tell what file was being compiled, and with which options.
5439 @item -fvar-tracking
5440 @opindex fvar-tracking
5441 Run variable tracking pass. It computes where variables are stored at each
5442 position in code. Better debugging information is then generated
5443 (if the debugging information format supports this information).
5445 It is enabled by default when compiling with optimization (@option{-Os},
5446 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5447 the debug info format supports it.
5449 @item -print-file-name=@var{library}
5450 @opindex print-file-name
5451 Print the full absolute name of the library file @var{library} that
5452 would be used when linking---and don't do anything else. With this
5453 option, GCC does not compile or link anything; it just prints the
5456 @item -print-multi-directory
5457 @opindex print-multi-directory
5458 Print the directory name corresponding to the multilib selected by any
5459 other switches present in the command line. This directory is supposed
5460 to exist in @env{GCC_EXEC_PREFIX}.
5462 @item -print-multi-lib
5463 @opindex print-multi-lib
5464 Print the mapping from multilib directory names to compiler switches
5465 that enable them. The directory name is separated from the switches by
5466 @samp{;}, and each switch starts with an @samp{@@} instead of the
5467 @samp{-}, without spaces between multiple switches. This is supposed to
5468 ease shell-processing.
5470 @item -print-prog-name=@var{program}
5471 @opindex print-prog-name
5472 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5474 @item -print-libgcc-file-name
5475 @opindex print-libgcc-file-name
5476 Same as @option{-print-file-name=libgcc.a}.
5478 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5479 but you do want to link with @file{libgcc.a}. You can do
5482 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5485 @item -print-search-dirs
5486 @opindex print-search-dirs
5487 Print the name of the configured installation directory and a list of
5488 program and library directories @command{gcc} will search---and don't do anything else.
5490 This is useful when @command{gcc} prints the error message
5491 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5492 To resolve this you either need to put @file{cpp0} and the other compiler
5493 components where @command{gcc} expects to find them, or you can set the environment
5494 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5495 Don't forget the trailing @samp{/}.
5496 @xref{Environment Variables}.
5498 @item -print-sysroot
5499 @opindex print-sysroot
5500 Print the target sysroot directory that will be used during
5501 compilation. This is the target sysroot specified either at configure
5502 time or using the @option{--sysroot} option, possibly with an extra
5503 suffix that depends on compilation options. If no target sysroot is
5504 specified, the option prints nothing.
5506 @item -print-sysroot-headers-suffix
5507 @opindex print-sysroot-headers-suffix
5508 Print the suffix added to the target sysroot when searching for
5509 headers, or give an error if the compiler is not configured with such
5510 a suffix---and don't do anything else.
5513 @opindex dumpmachine
5514 Print the compiler's target machine (for example,
5515 @samp{i686-pc-linux-gnu})---and don't do anything else.
5518 @opindex dumpversion
5519 Print the compiler version (for example, @samp{3.0})---and don't do
5524 Print the compiler's built-in specs---and don't do anything else. (This
5525 is used when GCC itself is being built.) @xref{Spec Files}.
5527 @item -feliminate-unused-debug-types
5528 @opindex feliminate-unused-debug-types
5529 Normally, when producing DWARF2 output, GCC will emit debugging
5530 information for all types declared in a compilation
5531 unit, regardless of whether or not they are actually used
5532 in that compilation unit. Sometimes this is useful, such as
5533 if, in the debugger, you want to cast a value to a type that is
5534 not actually used in your program (but is declared). More often,
5535 however, this results in a significant amount of wasted space.
5536 With this option, GCC will avoid producing debug symbol output
5537 for types that are nowhere used in the source file being compiled.
5540 @node Optimize Options
5541 @section Options That Control Optimization
5542 @cindex optimize options
5543 @cindex options, optimization
5545 These options control various sorts of optimizations.
5547 Without any optimization option, the compiler's goal is to reduce the
5548 cost of compilation and to make debugging produce the expected
5549 results. Statements are independent: if you stop the program with a
5550 breakpoint between statements, you can then assign a new value to any
5551 variable or change the program counter to any other statement in the
5552 function and get exactly the results you would expect from the source
5555 Turning on optimization flags makes the compiler attempt to improve
5556 the performance and/or code size at the expense of compilation time
5557 and possibly the ability to debug the program.
5559 The compiler performs optimization based on the knowledge it has of the
5560 program. Compiling multiple files at once to a single output file mode allows
5561 the compiler to use information gained from all of the files when compiling
5564 Not all optimizations are controlled directly by a flag. Only
5565 optimizations that have a flag are listed in this section.
5567 Depending on the target and how GCC was configured, a slightly different
5568 set of optimizations may be enabled at each @option{-O} level than
5569 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5570 to find out the exact set of optimizations that are enabled at each level.
5571 @xref{Overall Options}, for examples.
5578 Optimize. Optimizing compilation takes somewhat more time, and a lot
5579 more memory for a large function.
5581 With @option{-O}, the compiler tries to reduce code size and execution
5582 time, without performing any optimizations that take a great deal of
5585 @option{-O} turns on the following optimization flags:
5588 -fcprop-registers @gol
5591 -fdelayed-branch @gol
5593 -fguess-branch-probability @gol
5594 -fif-conversion2 @gol
5595 -fif-conversion @gol
5596 -finline-small-functions @gol
5597 -fipa-pure-const @gol
5598 -fipa-reference @gol
5600 -fsplit-wide-types @gol
5601 -ftree-builtin-call-dce @gol
5604 -ftree-copyrename @gol
5606 -ftree-dominator-opts @gol
5608 -ftree-forwprop @gol
5616 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5617 where doing so does not interfere with debugging.
5621 Optimize even more. GCC performs nearly all supported optimizations
5622 that do not involve a space-speed tradeoff.
5623 As compared to @option{-O}, this option increases both compilation time
5624 and the performance of the generated code.
5626 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5627 also turns on the following optimization flags:
5628 @gccoptlist{-fthread-jumps @gol
5629 -falign-functions -falign-jumps @gol
5630 -falign-loops -falign-labels @gol
5633 -fcse-follow-jumps -fcse-skip-blocks @gol
5634 -fdelete-null-pointer-checks @gol
5635 -fexpensive-optimizations @gol
5636 -fgcse -fgcse-lm @gol
5637 -findirect-inlining @gol
5638 -foptimize-sibling-calls @gol
5641 -freorder-blocks -freorder-functions @gol
5642 -frerun-cse-after-loop @gol
5643 -fsched-interblock -fsched-spec @gol
5644 -fschedule-insns -fschedule-insns2 @gol
5645 -fstrict-aliasing -fstrict-overflow @gol
5646 -ftree-switch-conversion @gol
5650 Please note the warning under @option{-fgcse} about
5651 invoking @option{-O2} on programs that use computed gotos.
5655 Optimize yet more. @option{-O3} turns on all optimizations specified
5656 by @option{-O2} and also turns on the @option{-finline-functions},
5657 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5658 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5662 Reduce compilation time and make debugging produce the expected
5663 results. This is the default.
5667 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5668 do not typically increase code size. It also performs further
5669 optimizations designed to reduce code size.
5671 @option{-Os} disables the following optimization flags:
5672 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5673 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5674 -fprefetch-loop-arrays -ftree-vect-loop-version}
5676 If you use multiple @option{-O} options, with or without level numbers,
5677 the last such option is the one that is effective.
5680 Options of the form @option{-f@var{flag}} specify machine-independent
5681 flags. Most flags have both positive and negative forms; the negative
5682 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5683 below, only one of the forms is listed---the one you typically will
5684 use. You can figure out the other form by either removing @samp{no-}
5687 The following options control specific optimizations. They are either
5688 activated by @option{-O} options or are related to ones that are. You
5689 can use the following flags in the rare cases when ``fine-tuning'' of
5690 optimizations to be performed is desired.
5693 @item -fno-default-inline
5694 @opindex fno-default-inline
5695 Do not make member functions inline by default merely because they are
5696 defined inside the class scope (C++ only). Otherwise, when you specify
5697 @w{@option{-O}}, member functions defined inside class scope are compiled
5698 inline by default; i.e., you don't need to add @samp{inline} in front of
5699 the member function name.
5701 @item -fno-defer-pop
5702 @opindex fno-defer-pop
5703 Always pop the arguments to each function call as soon as that function
5704 returns. For machines which must pop arguments after a function call,
5705 the compiler normally lets arguments accumulate on the stack for several
5706 function calls and pops them all at once.
5708 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5710 @item -fforward-propagate
5711 @opindex fforward-propagate
5712 Perform a forward propagation pass on RTL@. The pass tries to combine two
5713 instructions and checks if the result can be simplified. If loop unrolling
5714 is active, two passes are performed and the second is scheduled after
5717 This option is enabled by default at optimization levels @option{-O},
5718 @option{-O2}, @option{-O3}, @option{-Os}.
5720 @item -fomit-frame-pointer
5721 @opindex fomit-frame-pointer
5722 Don't keep the frame pointer in a register for functions that
5723 don't need one. This avoids the instructions to save, set up and
5724 restore frame pointers; it also makes an extra register available
5725 in many functions. @strong{It also makes debugging impossible on
5728 On some machines, such as the VAX, this flag has no effect, because
5729 the standard calling sequence automatically handles the frame pointer
5730 and nothing is saved by pretending it doesn't exist. The
5731 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5732 whether a target machine supports this flag. @xref{Registers,,Register
5733 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5735 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5737 @item -foptimize-sibling-calls
5738 @opindex foptimize-sibling-calls
5739 Optimize sibling and tail recursive calls.
5741 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5745 Don't pay attention to the @code{inline} keyword. Normally this option
5746 is used to keep the compiler from expanding any functions inline.
5747 Note that if you are not optimizing, no functions can be expanded inline.
5749 @item -finline-small-functions
5750 @opindex finline-small-functions
5751 Integrate functions into their callers when their body is smaller than expected
5752 function call code (so overall size of program gets smaller). The compiler
5753 heuristically decides which functions are simple enough to be worth integrating
5756 Enabled at level @option{-O2}.
5758 @item -findirect-inlining
5759 @opindex findirect-inlining
5760 Inline also indirect calls that are discovered to be known at compile
5761 time thanks to previous inlining. This option has any effect only
5762 when inlining itself is turned on by the @option{-finline-functions}
5763 or @option{-finline-small-functions} options.
5765 Enabled at level @option{-O2}.
5767 @item -finline-functions
5768 @opindex finline-functions
5769 Integrate all simple functions into their callers. The compiler
5770 heuristically decides which functions are simple enough to be worth
5771 integrating in this way.
5773 If all calls to a given function are integrated, and the function is
5774 declared @code{static}, then the function is normally not output as
5775 assembler code in its own right.
5777 Enabled at level @option{-O3}.
5779 @item -finline-functions-called-once
5780 @opindex finline-functions-called-once
5781 Consider all @code{static} functions called once for inlining into their
5782 caller even if they are not marked @code{inline}. If a call to a given
5783 function is integrated, then the function is not output as assembler code
5786 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5788 @item -fearly-inlining
5789 @opindex fearly-inlining
5790 Inline functions marked by @code{always_inline} and functions whose body seems
5791 smaller than the function call overhead early before doing
5792 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5793 makes profiling significantly cheaper and usually inlining faster on programs
5794 having large chains of nested wrapper functions.
5798 @item -finline-limit=@var{n}
5799 @opindex finline-limit
5800 By default, GCC limits the size of functions that can be inlined. This flag
5801 allows coarse control of this limit. @var{n} is the size of functions that
5802 can be inlined in number of pseudo instructions.
5804 Inlining is actually controlled by a number of parameters, which may be
5805 specified individually by using @option{--param @var{name}=@var{value}}.
5806 The @option{-finline-limit=@var{n}} option sets some of these parameters
5810 @item max-inline-insns-single
5811 is set to @var{n}/2.
5812 @item max-inline-insns-auto
5813 is set to @var{n}/2.
5816 See below for a documentation of the individual
5817 parameters controlling inlining and for the defaults of these parameters.
5819 @emph{Note:} there may be no value to @option{-finline-limit} that results
5820 in default behavior.
5822 @emph{Note:} pseudo instruction represents, in this particular context, an
5823 abstract measurement of function's size. In no way does it represent a count
5824 of assembly instructions and as such its exact meaning might change from one
5825 release to an another.
5827 @item -fkeep-inline-functions
5828 @opindex fkeep-inline-functions
5829 In C, emit @code{static} functions that are declared @code{inline}
5830 into the object file, even if the function has been inlined into all
5831 of its callers. This switch does not affect functions using the
5832 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5833 inline functions into the object file.
5835 @item -fkeep-static-consts
5836 @opindex fkeep-static-consts
5837 Emit variables declared @code{static const} when optimization isn't turned
5838 on, even if the variables aren't referenced.
5840 GCC enables this option by default. If you want to force the compiler to
5841 check if the variable was referenced, regardless of whether or not
5842 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5844 @item -fmerge-constants
5845 @opindex fmerge-constants
5846 Attempt to merge identical constants (string constants and floating point
5847 constants) across compilation units.
5849 This option is the default for optimized compilation if the assembler and
5850 linker support it. Use @option{-fno-merge-constants} to inhibit this
5853 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5855 @item -fmerge-all-constants
5856 @opindex fmerge-all-constants
5857 Attempt to merge identical constants and identical variables.
5859 This option implies @option{-fmerge-constants}. In addition to
5860 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5861 arrays or initialized constant variables with integral or floating point
5862 types. Languages like C or C++ require each variable, including multiple
5863 instances of the same variable in recursive calls, to have distinct locations,
5864 so using this option will result in non-conforming
5867 @item -fmodulo-sched
5868 @opindex fmodulo-sched
5869 Perform swing modulo scheduling immediately before the first scheduling
5870 pass. This pass looks at innermost loops and reorders their
5871 instructions by overlapping different iterations.
5873 @item -fmodulo-sched-allow-regmoves
5874 @opindex fmodulo-sched-allow-regmoves
5875 Perform more aggressive SMS based modulo scheduling with register moves
5876 allowed. By setting this flag certain anti-dependences edges will be
5877 deleted which will trigger the generation of reg-moves based on the
5878 life-range analysis. This option is effective only with
5879 @option{-fmodulo-sched} enabled.
5881 @item -fno-branch-count-reg
5882 @opindex fno-branch-count-reg
5883 Do not use ``decrement and branch'' instructions on a count register,
5884 but instead generate a sequence of instructions that decrement a
5885 register, compare it against zero, then branch based upon the result.
5886 This option is only meaningful on architectures that support such
5887 instructions, which include x86, PowerPC, IA-64 and S/390.
5889 The default is @option{-fbranch-count-reg}.
5891 @item -fno-function-cse
5892 @opindex fno-function-cse
5893 Do not put function addresses in registers; make each instruction that
5894 calls a constant function contain the function's address explicitly.
5896 This option results in less efficient code, but some strange hacks
5897 that alter the assembler output may be confused by the optimizations
5898 performed when this option is not used.
5900 The default is @option{-ffunction-cse}
5902 @item -fno-zero-initialized-in-bss
5903 @opindex fno-zero-initialized-in-bss
5904 If the target supports a BSS section, GCC by default puts variables that
5905 are initialized to zero into BSS@. This can save space in the resulting
5908 This option turns off this behavior because some programs explicitly
5909 rely on variables going to the data section. E.g., so that the
5910 resulting executable can find the beginning of that section and/or make
5911 assumptions based on that.
5913 The default is @option{-fzero-initialized-in-bss}.
5915 @item -fmudflap -fmudflapth -fmudflapir
5919 @cindex bounds checking
5921 For front-ends that support it (C and C++), instrument all risky
5922 pointer/array dereferencing operations, some standard library
5923 string/heap functions, and some other associated constructs with
5924 range/validity tests. Modules so instrumented should be immune to
5925 buffer overflows, invalid heap use, and some other classes of C/C++
5926 programming errors. The instrumentation relies on a separate runtime
5927 library (@file{libmudflap}), which will be linked into a program if
5928 @option{-fmudflap} is given at link time. Run-time behavior of the
5929 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5930 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5933 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5934 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5935 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5936 instrumentation should ignore pointer reads. This produces less
5937 instrumentation (and therefore faster execution) and still provides
5938 some protection against outright memory corrupting writes, but allows
5939 erroneously read data to propagate within a program.
5941 @item -fthread-jumps
5942 @opindex fthread-jumps
5943 Perform optimizations where we check to see if a jump branches to a
5944 location where another comparison subsumed by the first is found. If
5945 so, the first branch is redirected to either the destination of the
5946 second branch or a point immediately following it, depending on whether
5947 the condition is known to be true or false.
5949 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5951 @item -fsplit-wide-types
5952 @opindex fsplit-wide-types
5953 When using a type that occupies multiple registers, such as @code{long
5954 long} on a 32-bit system, split the registers apart and allocate them
5955 independently. This normally generates better code for those types,
5956 but may make debugging more difficult.
5958 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5961 @item -fcse-follow-jumps
5962 @opindex fcse-follow-jumps
5963 In common subexpression elimination (CSE), scan through jump instructions
5964 when the target of the jump is not reached by any other path. For
5965 example, when CSE encounters an @code{if} statement with an
5966 @code{else} clause, CSE will follow the jump when the condition
5969 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5971 @item -fcse-skip-blocks
5972 @opindex fcse-skip-blocks
5973 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5974 follow jumps which conditionally skip over blocks. When CSE
5975 encounters a simple @code{if} statement with no else clause,
5976 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5977 body of the @code{if}.
5979 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5981 @item -frerun-cse-after-loop
5982 @opindex frerun-cse-after-loop
5983 Re-run common subexpression elimination after loop optimizations has been
5986 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5990 Perform a global common subexpression elimination pass.
5991 This pass also performs global constant and copy propagation.
5993 @emph{Note:} When compiling a program using computed gotos, a GCC
5994 extension, you may get better runtime performance if you disable
5995 the global common subexpression elimination pass by adding
5996 @option{-fno-gcse} to the command line.
5998 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6002 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6003 attempt to move loads which are only killed by stores into themselves. This
6004 allows a loop containing a load/store sequence to be changed to a load outside
6005 the loop, and a copy/store within the loop.
6007 Enabled by default when gcse is enabled.
6011 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6012 global common subexpression elimination. This pass will attempt to move
6013 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6014 loops containing a load/store sequence can be changed to a load before
6015 the loop and a store after the loop.
6017 Not enabled at any optimization level.
6021 When @option{-fgcse-las} is enabled, the global common subexpression
6022 elimination pass eliminates redundant loads that come after stores to the
6023 same memory location (both partial and full redundancies).
6025 Not enabled at any optimization level.
6027 @item -fgcse-after-reload
6028 @opindex fgcse-after-reload
6029 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6030 pass is performed after reload. The purpose of this pass is to cleanup
6033 @item -funsafe-loop-optimizations
6034 @opindex funsafe-loop-optimizations
6035 If given, the loop optimizer will assume that loop indices do not
6036 overflow, and that the loops with nontrivial exit condition are not
6037 infinite. This enables a wider range of loop optimizations even if
6038 the loop optimizer itself cannot prove that these assumptions are valid.
6039 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6040 if it finds this kind of loop.
6042 @item -fcrossjumping
6043 @opindex fcrossjumping
6044 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6045 resulting code may or may not perform better than without cross-jumping.
6047 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6049 @item -fauto-inc-dec
6050 @opindex fauto-inc-dec
6051 Combine increments or decrements of addresses with memory accesses.
6052 This pass is always skipped on architectures that do not have
6053 instructions to support this. Enabled by default at @option{-O} and
6054 higher on architectures that support this.
6058 Perform dead code elimination (DCE) on RTL@.
6059 Enabled by default at @option{-O} and higher.
6063 Perform dead store elimination (DSE) on RTL@.
6064 Enabled by default at @option{-O} and higher.
6066 @item -fif-conversion
6067 @opindex fif-conversion
6068 Attempt to transform conditional jumps into branch-less equivalents. This
6069 include use of conditional moves, min, max, set flags and abs instructions, and
6070 some tricks doable by standard arithmetics. The use of conditional execution
6071 on chips where it is available is controlled by @code{if-conversion2}.
6073 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6075 @item -fif-conversion2
6076 @opindex fif-conversion2
6077 Use conditional execution (where available) to transform conditional jumps into
6078 branch-less equivalents.
6080 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6082 @item -fdelete-null-pointer-checks
6083 @opindex fdelete-null-pointer-checks
6084 Assume that programs cannot safely dereference null pointers, and that
6085 no code or data element resides there. This enables simple constant
6086 folding optimizations at all optimization levels. In addition, other
6087 optimization passes in GCC use this flag to control global dataflow
6088 analyses that eliminate useless checks for null pointers; these assume
6089 that if a pointer is checked after it has already been dereferenced,
6092 Note however that in some environments this assumption is not true.
6093 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6094 for programs which depend on that behavior.
6096 Some targets, especially embedded ones, disable this option at all levels.
6097 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6098 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6099 are enabled independently at different optimization levels.
6101 @item -fexpensive-optimizations
6102 @opindex fexpensive-optimizations
6103 Perform a number of minor optimizations that are relatively expensive.
6105 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6107 @item -foptimize-register-move
6109 @opindex foptimize-register-move
6111 Attempt to reassign register numbers in move instructions and as
6112 operands of other simple instructions in order to maximize the amount of
6113 register tying. This is especially helpful on machines with two-operand
6116 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6119 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6121 @item -fira-algorithm=@var{algorithm}
6122 Use specified coloring algorithm for the integrated register
6123 allocator. The @var{algorithm} argument should be @code{priority} or
6124 @code{CB}. The first algorithm specifies Chow's priority coloring,
6125 the second one specifies Chaitin-Briggs coloring. The second
6126 algorithm can be unimplemented for some architectures. If it is
6127 implemented, it is the default because Chaitin-Briggs coloring as a
6128 rule generates a better code.
6130 @item -fira-region=@var{region}
6131 Use specified regions for the integrated register allocator. The
6132 @var{region} argument should be one of @code{all}, @code{mixed}, or
6133 @code{one}. The first value means using all loops as register
6134 allocation regions, the second value which is the default means using
6135 all loops except for loops with small register pressure as the
6136 regions, and third one means using all function as a single region.
6137 The first value can give best result for machines with small size and
6138 irregular register set, the third one results in faster and generates
6139 decent code and the smallest size code, and the default value usually
6140 give the best results in most cases and for most architectures.
6142 @item -fira-coalesce
6143 @opindex fira-coalesce
6144 Do optimistic register coalescing. This option might be profitable for
6145 architectures with big regular register files.
6147 @item -fno-ira-share-save-slots
6148 @opindex fno-ira-share-save-slots
6149 Switch off sharing stack slots used for saving call used hard
6150 registers living through a call. Each hard register will get a
6151 separate stack slot and as a result function stack frame will be
6154 @item -fno-ira-share-spill-slots
6155 @opindex fno-ira-share-spill-slots
6156 Switch off sharing stack slots allocated for pseudo-registers. Each
6157 pseudo-register which did not get a hard register will get a separate
6158 stack slot and as a result function stack frame will be bigger.
6160 @item -fira-verbose=@var{n}
6161 @opindex fira-verbose
6162 Set up how verbose dump file for the integrated register allocator
6163 will be. Default value is 5. If the value is greater or equal to 10,
6164 the dump file will be stderr as if the value were @var{n} minus 10.
6166 @item -fdelayed-branch
6167 @opindex fdelayed-branch
6168 If supported for the target machine, attempt to reorder instructions
6169 to exploit instruction slots available after delayed branch
6172 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6174 @item -fschedule-insns
6175 @opindex fschedule-insns
6176 If supported for the target machine, attempt to reorder instructions to
6177 eliminate execution stalls due to required data being unavailable. This
6178 helps machines that have slow floating point or memory load instructions
6179 by allowing other instructions to be issued until the result of the load
6180 or floating point instruction is required.
6182 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6184 @item -fschedule-insns2
6185 @opindex fschedule-insns2
6186 Similar to @option{-fschedule-insns}, but requests an additional pass of
6187 instruction scheduling after register allocation has been done. This is
6188 especially useful on machines with a relatively small number of
6189 registers and where memory load instructions take more than one cycle.
6191 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6193 @item -fno-sched-interblock
6194 @opindex fno-sched-interblock
6195 Don't schedule instructions across basic blocks. This is normally
6196 enabled by default when scheduling before register allocation, i.e.@:
6197 with @option{-fschedule-insns} or at @option{-O2} or higher.
6199 @item -fno-sched-spec
6200 @opindex fno-sched-spec
6201 Don't allow speculative motion of non-load instructions. This is normally
6202 enabled by default when scheduling before register allocation, i.e.@:
6203 with @option{-fschedule-insns} or at @option{-O2} or higher.
6205 @item -fsched-spec-load
6206 @opindex fsched-spec-load
6207 Allow speculative motion of some load instructions. This only makes
6208 sense when scheduling before register allocation, i.e.@: with
6209 @option{-fschedule-insns} or at @option{-O2} or higher.
6211 @item -fsched-spec-load-dangerous
6212 @opindex fsched-spec-load-dangerous
6213 Allow speculative motion of more load instructions. This only makes
6214 sense when scheduling before register allocation, i.e.@: with
6215 @option{-fschedule-insns} or at @option{-O2} or higher.
6217 @item -fsched-stalled-insns
6218 @itemx -fsched-stalled-insns=@var{n}
6219 @opindex fsched-stalled-insns
6220 Define how many insns (if any) can be moved prematurely from the queue
6221 of stalled insns into the ready list, during the second scheduling pass.
6222 @option{-fno-sched-stalled-insns} means that no insns will be moved
6223 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6224 on how many queued insns can be moved prematurely.
6225 @option{-fsched-stalled-insns} without a value is equivalent to
6226 @option{-fsched-stalled-insns=1}.
6228 @item -fsched-stalled-insns-dep
6229 @itemx -fsched-stalled-insns-dep=@var{n}
6230 @opindex fsched-stalled-insns-dep
6231 Define how many insn groups (cycles) will be examined for a dependency
6232 on a stalled insn that is candidate for premature removal from the queue
6233 of stalled insns. This has an effect only during the second scheduling pass,
6234 and only if @option{-fsched-stalled-insns} is used.
6235 @option{-fno-sched-stalled-insns-dep} is equivalent to
6236 @option{-fsched-stalled-insns-dep=0}.
6237 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6238 @option{-fsched-stalled-insns-dep=1}.
6240 @item -fsched2-use-superblocks
6241 @opindex fsched2-use-superblocks
6242 When scheduling after register allocation, do use superblock scheduling
6243 algorithm. Superblock scheduling allows motion across basic block boundaries
6244 resulting on faster schedules. This option is experimental, as not all machine
6245 descriptions used by GCC model the CPU closely enough to avoid unreliable
6246 results from the algorithm.
6248 This only makes sense when scheduling after register allocation, i.e.@: with
6249 @option{-fschedule-insns2} or at @option{-O2} or higher.
6251 @item -fsched-group-heuristic
6252 @opindex fsched-group-heuristic
6253 Enable the group heuristic in the scheduler. This heuristic favors
6254 the instruction that belongs to a schedule group. This is enabled
6255 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6256 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6258 @item -fsched-critical-path-heuristic
6259 @opindex fsched-critical-path-heuristic
6260 Enable the critical-path heuristic in the scheduler. This heuristic favors
6261 instructions on the critical path. This is enabled by default when
6262 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6263 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6265 @item -fsched-spec-insn-heuristic
6266 @opindex fsched-spec-insn-heuristic
6267 Enable the speculative instruction heuristic in the scheduler. This
6268 heuristic favors speculative instructions with greater dependency weakness.
6269 This is enabled by default when scheduling is enabled, i.e.@:
6270 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6271 or at @option{-O2} or higher.
6273 @item -fsched-reg-pressure-heuristic
6274 @opindex fsched-reg-pressure-heuristic
6275 Enable the register pressure heuristic in the scheduler. This heuristic
6276 favors the instruction with smaller contribution to register pressure.
6277 This only makes sense when scheduling before register allocation, i.e.@:
6278 with @option{-fschedule-insns} or at @option{-O2} or higher.
6280 @item -fsched-rank-heuristic
6281 @opindex fsched-rank-heuristic
6282 Enable the rank heuristic in the scheduler. This heuristic favors
6283 the instruction belonging to a basic block with greater size or frequency.
6284 This is enabled by default when scheduling is enabled, i.e.@:
6285 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6286 at @option{-O2} or higher.
6288 @item -fsched-last-insn-heuristic
6289 @opindex fsched-last-insn-heuristic
6290 Enable the last-instruction heuristic in the scheduler. This heuristic
6291 favors the instruction that is less dependent on the last instruction
6292 scheduled. This is enabled by default when scheduling is enabled,
6293 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6294 at @option{-O2} or higher.
6296 @item -fsched-dep-count-heuristic
6297 @opindex fsched-dep-count-heuristic
6298 Enable the dependent-count heuristic in the scheduler. This heuristic
6299 favors the instruction that has more instructions depending on it.
6300 This is enabled by default when scheduling is enabled, i.e.@:
6301 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6302 at @option{-O2} or higher.
6304 @item -fsched2-use-traces
6305 @opindex fsched2-use-traces
6306 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6307 allocation and additionally perform code duplication in order to increase the
6308 size of superblocks using tracer pass. See @option{-ftracer} for details on
6311 This mode should produce faster but significantly longer programs. Also
6312 without @option{-fbranch-probabilities} the traces constructed may not
6313 match the reality and hurt the performance. This only makes
6314 sense when scheduling after register allocation, i.e.@: with
6315 @option{-fschedule-insns2} or at @option{-O2} or higher.
6317 @item -freschedule-modulo-scheduled-loops
6318 @opindex freschedule-modulo-scheduled-loops
6319 The modulo scheduling comes before the traditional scheduling, if a loop
6320 was modulo scheduled we may want to prevent the later scheduling passes
6321 from changing its schedule, we use this option to control that.
6323 @item -fselective-scheduling
6324 @opindex fselective-scheduling
6325 Schedule instructions using selective scheduling algorithm. Selective
6326 scheduling runs instead of the first scheduler pass.
6328 @item -fselective-scheduling2
6329 @opindex fselective-scheduling2
6330 Schedule instructions using selective scheduling algorithm. Selective
6331 scheduling runs instead of the second scheduler pass.
6333 @item -fsel-sched-pipelining
6334 @opindex fsel-sched-pipelining
6335 Enable software pipelining of innermost loops during selective scheduling.
6336 This option has no effect until one of @option{-fselective-scheduling} or
6337 @option{-fselective-scheduling2} is turned on.
6339 @item -fsel-sched-pipelining-outer-loops
6340 @opindex fsel-sched-pipelining-outer-loops
6341 When pipelining loops during selective scheduling, also pipeline outer loops.
6342 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6344 @item -fcaller-saves
6345 @opindex fcaller-saves
6346 Enable values to be allocated in registers that will be clobbered by
6347 function calls, by emitting extra instructions to save and restore the
6348 registers around such calls. Such allocation is done only when it
6349 seems to result in better code than would otherwise be produced.
6351 This option is always enabled by default on certain machines, usually
6352 those which have no call-preserved registers to use instead.
6354 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6356 @item -fconserve-stack
6357 @opindex fconserve-stack
6358 Attempt to minimize stack usage. The compiler will attempt to use less
6359 stack space, even if that makes the program slower. This option
6360 implies setting the @option{large-stack-frame} parameter to 100
6361 and the @option{large-stack-frame-growth} parameter to 400.
6363 @item -ftree-reassoc
6364 @opindex ftree-reassoc
6365 Perform reassociation on trees. This flag is enabled by default
6366 at @option{-O} and higher.
6370 Perform partial redundancy elimination (PRE) on trees. This flag is
6371 enabled by default at @option{-O2} and @option{-O3}.
6373 @item -ftree-forwprop
6374 @opindex ftree-forwprop
6375 Perform forward propagation on trees. This flag is enabled by default
6376 at @option{-O} and higher.
6380 Perform full redundancy elimination (FRE) on trees. The difference
6381 between FRE and PRE is that FRE only considers expressions
6382 that are computed on all paths leading to the redundant computation.
6383 This analysis is faster than PRE, though it exposes fewer redundancies.
6384 This flag is enabled by default at @option{-O} and higher.
6386 @item -ftree-phiprop
6387 @opindex ftree-phiprop
6388 Perform hoisting of loads from conditional pointers on trees. This
6389 pass is enabled by default at @option{-O} and higher.
6391 @item -ftree-copy-prop
6392 @opindex ftree-copy-prop
6393 Perform copy propagation on trees. This pass eliminates unnecessary
6394 copy operations. This flag is enabled by default at @option{-O} and
6397 @item -fipa-pure-const
6398 @opindex fipa-pure-const
6399 Discover which functions are pure or constant.
6400 Enabled by default at @option{-O} and higher.
6402 @item -fipa-reference
6403 @opindex fipa-reference
6404 Discover which static variables do not escape cannot escape the
6406 Enabled by default at @option{-O} and higher.
6408 @item -fipa-struct-reorg
6409 @opindex fipa-struct-reorg
6410 Perform structure reorganization optimization, that change C-like structures
6411 layout in order to better utilize spatial locality. This transformation is
6412 affective for programs containing arrays of structures. Available in two
6413 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6414 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6415 to provide the safety of this transformation. It works only in whole program
6416 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6417 enabled. Structures considered @samp{cold} by this transformation are not
6418 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6420 With this flag, the program debug info reflects a new structure layout.
6424 Perform interprocedural pointer analysis. This option is experimental
6425 and does not affect generated code.
6429 Perform interprocedural constant propagation.
6430 This optimization analyzes the program to determine when values passed
6431 to functions are constants and then optimizes accordingly.
6432 This optimization can substantially increase performance
6433 if the application has constants passed to functions.
6434 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6436 @item -fipa-cp-clone
6437 @opindex fipa-cp-clone
6438 Perform function cloning to make interprocedural constant propagation stronger.
6439 When enabled, interprocedural constant propagation will perform function cloning
6440 when externally visible function can be called with constant arguments.
6441 Because this optimization can create multiple copies of functions,
6442 it may significantly increase code size
6443 (see @option{--param ipcp-unit-growth=@var{value}}).
6444 This flag is enabled by default at @option{-O3}.
6446 @item -fipa-matrix-reorg
6447 @opindex fipa-matrix-reorg
6448 Perform matrix flattening and transposing.
6449 Matrix flattening tries to replace an @math{m}-dimensional matrix
6450 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6451 This reduces the level of indirection needed for accessing the elements
6452 of the matrix. The second optimization is matrix transposing that
6453 attempts to change the order of the matrix's dimensions in order to
6454 improve cache locality.
6455 Both optimizations need the @option{-fwhole-program} flag.
6456 Transposing is enabled only if profiling information is available.
6460 Perform forward store motion on trees. This flag is
6461 enabled by default at @option{-O} and higher.
6465 Perform sparse conditional constant propagation (CCP) on trees. This
6466 pass only operates on local scalar variables and is enabled by default
6467 at @option{-O} and higher.
6469 @item -ftree-switch-conversion
6470 Perform conversion of simple initializations in a switch to
6471 initializations from a scalar array. This flag is enabled by default
6472 at @option{-O2} and higher.
6476 Perform dead code elimination (DCE) on trees. This flag is enabled by
6477 default at @option{-O} and higher.
6479 @item -ftree-builtin-call-dce
6480 @opindex ftree-builtin-call-dce
6481 Perform conditional dead code elimination (DCE) for calls to builtin functions
6482 that may set @code{errno} but are otherwise side-effect free. This flag is
6483 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6486 @item -ftree-dominator-opts
6487 @opindex ftree-dominator-opts
6488 Perform a variety of simple scalar cleanups (constant/copy
6489 propagation, redundancy elimination, range propagation and expression
6490 simplification) based on a dominator tree traversal. This also
6491 performs jump threading (to reduce jumps to jumps). This flag is
6492 enabled by default at @option{-O} and higher.
6496 Perform dead store elimination (DSE) on trees. A dead store is a store into
6497 a memory location which will later be overwritten by another store without
6498 any intervening loads. In this case the earlier store can be deleted. This
6499 flag is enabled by default at @option{-O} and higher.
6503 Perform loop header copying on trees. This is beneficial since it increases
6504 effectiveness of code motion optimizations. It also saves one jump. This flag
6505 is enabled by default at @option{-O} and higher. It is not enabled
6506 for @option{-Os}, since it usually increases code size.
6508 @item -ftree-loop-optimize
6509 @opindex ftree-loop-optimize
6510 Perform loop optimizations on trees. This flag is enabled by default
6511 at @option{-O} and higher.
6513 @item -ftree-loop-linear
6514 @opindex ftree-loop-linear
6515 Perform linear loop transformations on tree. This flag can improve cache
6516 performance and allow further loop optimizations to take place.
6518 @item -floop-interchange
6519 Perform loop interchange transformations on loops. Interchanging two
6520 nested loops switches the inner and outer loops. For example, given a
6525 A(J, I) = A(J, I) * C
6529 loop interchange will transform the loop as if the user had written:
6533 A(J, I) = A(J, I) * C
6537 which can be beneficial when @code{N} is larger than the caches,
6538 because in Fortran, the elements of an array are stored in memory
6539 contiguously by column, and the original loop iterates over rows,
6540 potentially creating at each access a cache miss. This optimization
6541 applies to all the languages supported by GCC and is not limited to
6542 Fortran. To use this code transformation, GCC has to be configured
6543 with @option{--with-ppl} and @option{--with-cloog} to enable the
6544 Graphite loop transformation infrastructure.
6546 @item -floop-strip-mine
6547 Perform loop strip mining transformations on loops. Strip mining
6548 splits a loop into two nested loops. The outer loop has strides
6549 equal to the strip size and the inner loop has strides of the
6550 original loop within a strip. For example, given a loop like:
6556 loop strip mining will transform the loop as if the user had written:
6559 DO I = II, min (II + 3, N)
6564 This optimization applies to all the languages supported by GCC and is
6565 not limited to Fortran. To use this code transformation, GCC has to
6566 be configured with @option{--with-ppl} and @option{--with-cloog} to
6567 enable the Graphite loop transformation infrastructure.
6570 Perform loop blocking transformations on loops. Blocking strip mines
6571 each loop in the loop nest such that the memory accesses of the
6572 element loops fit inside caches. For example, given a loop like:
6576 A(J, I) = B(I) + C(J)
6580 loop blocking will transform the loop as if the user had written:
6584 DO I = II, min (II + 63, N)
6585 DO J = JJ, min (JJ + 63, M)
6586 A(J, I) = B(I) + C(J)
6592 which can be beneficial when @code{M} is larger than the caches,
6593 because the innermost loop will iterate over a smaller amount of data
6594 that can be kept in the caches. This optimization applies to all the
6595 languages supported by GCC and is not limited to Fortran. To use this
6596 code transformation, GCC has to be configured with @option{--with-ppl}
6597 and @option{--with-cloog} to enable the Graphite loop transformation
6600 @item -fgraphite-identity
6601 @opindex fgraphite-identity
6602 Enable the identity transformation for graphite. For every SCoP we generate
6603 the polyhedral representation and transform it back to gimple. Using
6604 @option{-fgraphite-identity} we can check the costs or benefits of the
6605 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6606 are also performed by the code generator CLooG, like index splitting and
6607 dead code elimination in loops.
6609 @item -floop-parallelize-all
6610 Use the Graphite data dependence analysis to identify loops that can
6611 be parallelized. Parallelize all the loops that can be analyzed to
6612 not contain loop carried dependences without checking that it is
6613 profitable to parallelize the loops.
6615 @item -fcheck-data-deps
6616 @opindex fcheck-data-deps
6617 Compare the results of several data dependence analyzers. This option
6618 is used for debugging the data dependence analyzers.
6620 @item -ftree-loop-distribution
6621 Perform loop distribution. This flag can improve cache performance on
6622 big loop bodies and allow further loop optimizations, like
6623 parallelization or vectorization, to take place. For example, the loop
6640 @item -ftree-loop-im
6641 @opindex ftree-loop-im
6642 Perform loop invariant motion on trees. This pass moves only invariants that
6643 would be hard to handle at RTL level (function calls, operations that expand to
6644 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6645 operands of conditions that are invariant out of the loop, so that we can use
6646 just trivial invariantness analysis in loop unswitching. The pass also includes
6649 @item -ftree-loop-ivcanon
6650 @opindex ftree-loop-ivcanon
6651 Create a canonical counter for number of iterations in the loop for that
6652 determining number of iterations requires complicated analysis. Later
6653 optimizations then may determine the number easily. Useful especially
6654 in connection with unrolling.
6658 Perform induction variable optimizations (strength reduction, induction
6659 variable merging and induction variable elimination) on trees.
6661 @item -ftree-parallelize-loops=n
6662 @opindex ftree-parallelize-loops
6663 Parallelize loops, i.e., split their iteration space to run in n threads.
6664 This is only possible for loops whose iterations are independent
6665 and can be arbitrarily reordered. The optimization is only
6666 profitable on multiprocessor machines, for loops that are CPU-intensive,
6667 rather than constrained e.g.@: by memory bandwidth. This option
6668 implies @option{-pthread}, and thus is only supported on targets
6669 that have support for @option{-pthread}.
6673 Perform function-local points-to analysis on trees. This flag is
6674 enabled by default at @option{-O} and higher.
6678 Perform scalar replacement of aggregates. This pass replaces structure
6679 references with scalars to prevent committing structures to memory too
6680 early. This flag is enabled by default at @option{-O} and higher.
6682 @item -ftree-copyrename
6683 @opindex ftree-copyrename
6684 Perform copy renaming on trees. This pass attempts to rename compiler
6685 temporaries to other variables at copy locations, usually resulting in
6686 variable names which more closely resemble the original variables. This flag
6687 is enabled by default at @option{-O} and higher.
6691 Perform temporary expression replacement during the SSA->normal phase. Single
6692 use/single def temporaries are replaced at their use location with their
6693 defining expression. This results in non-GIMPLE code, but gives the expanders
6694 much more complex trees to work on resulting in better RTL generation. This is
6695 enabled by default at @option{-O} and higher.
6697 @item -ftree-vectorize
6698 @opindex ftree-vectorize
6699 Perform loop vectorization on trees. This flag is enabled by default at
6702 @item -ftree-vect-loop-version
6703 @opindex ftree-vect-loop-version
6704 Perform loop versioning when doing loop vectorization on trees. When a loop
6705 appears to be vectorizable except that data alignment or data dependence cannot
6706 be determined at compile time then vectorized and non-vectorized versions of
6707 the loop are generated along with runtime checks for alignment or dependence
6708 to control which version is executed. This option is enabled by default
6709 except at level @option{-Os} where it is disabled.
6711 @item -fvect-cost-model
6712 @opindex fvect-cost-model
6713 Enable cost model for vectorization.
6717 Perform Value Range Propagation on trees. This is similar to the
6718 constant propagation pass, but instead of values, ranges of values are
6719 propagated. This allows the optimizers to remove unnecessary range
6720 checks like array bound checks and null pointer checks. This is
6721 enabled by default at @option{-O2} and higher. Null pointer check
6722 elimination is only done if @option{-fdelete-null-pointer-checks} is
6727 Perform tail duplication to enlarge superblock size. This transformation
6728 simplifies the control flow of the function allowing other optimizations to do
6731 @item -funroll-loops
6732 @opindex funroll-loops
6733 Unroll loops whose number of iterations can be determined at compile
6734 time or upon entry to the loop. @option{-funroll-loops} implies
6735 @option{-frerun-cse-after-loop}. This option makes code larger,
6736 and may or may not make it run faster.
6738 @item -funroll-all-loops
6739 @opindex funroll-all-loops
6740 Unroll all loops, even if their number of iterations is uncertain when
6741 the loop is entered. This usually makes programs run more slowly.
6742 @option{-funroll-all-loops} implies the same options as
6743 @option{-funroll-loops},
6745 @item -fsplit-ivs-in-unroller
6746 @opindex fsplit-ivs-in-unroller
6747 Enables expressing of values of induction variables in later iterations
6748 of the unrolled loop using the value in the first iteration. This breaks
6749 long dependency chains, thus improving efficiency of the scheduling passes.
6751 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6752 same effect. However in cases the loop body is more complicated than
6753 a single basic block, this is not reliable. It also does not work at all
6754 on some of the architectures due to restrictions in the CSE pass.
6756 This optimization is enabled by default.
6758 @item -fvariable-expansion-in-unroller
6759 @opindex fvariable-expansion-in-unroller
6760 With this option, the compiler will create multiple copies of some
6761 local variables when unrolling a loop which can result in superior code.
6763 @item -fpredictive-commoning
6764 @opindex fpredictive-commoning
6765 Perform predictive commoning optimization, i.e., reusing computations
6766 (especially memory loads and stores) performed in previous
6767 iterations of loops.
6769 This option is enabled at level @option{-O3}.
6771 @item -fprefetch-loop-arrays
6772 @opindex fprefetch-loop-arrays
6773 If supported by the target machine, generate instructions to prefetch
6774 memory to improve the performance of loops that access large arrays.
6776 This option may generate better or worse code; results are highly
6777 dependent on the structure of loops within the source code.
6779 Disabled at level @option{-Os}.
6782 @itemx -fno-peephole2
6783 @opindex fno-peephole
6784 @opindex fno-peephole2
6785 Disable any machine-specific peephole optimizations. The difference
6786 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6787 are implemented in the compiler; some targets use one, some use the
6788 other, a few use both.
6790 @option{-fpeephole} is enabled by default.
6791 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6793 @item -fno-guess-branch-probability
6794 @opindex fno-guess-branch-probability
6795 Do not guess branch probabilities using heuristics.
6797 GCC will use heuristics to guess branch probabilities if they are
6798 not provided by profiling feedback (@option{-fprofile-arcs}). These
6799 heuristics are based on the control flow graph. If some branch probabilities
6800 are specified by @samp{__builtin_expect}, then the heuristics will be
6801 used to guess branch probabilities for the rest of the control flow graph,
6802 taking the @samp{__builtin_expect} info into account. The interactions
6803 between the heuristics and @samp{__builtin_expect} can be complex, and in
6804 some cases, it may be useful to disable the heuristics so that the effects
6805 of @samp{__builtin_expect} are easier to understand.
6807 The default is @option{-fguess-branch-probability} at levels
6808 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6810 @item -freorder-blocks
6811 @opindex freorder-blocks
6812 Reorder basic blocks in the compiled function in order to reduce number of
6813 taken branches and improve code locality.
6815 Enabled at levels @option{-O2}, @option{-O3}.
6817 @item -freorder-blocks-and-partition
6818 @opindex freorder-blocks-and-partition
6819 In addition to reordering basic blocks in the compiled function, in order
6820 to reduce number of taken branches, partitions hot and cold basic blocks
6821 into separate sections of the assembly and .o files, to improve
6822 paging and cache locality performance.
6824 This optimization is automatically turned off in the presence of
6825 exception handling, for linkonce sections, for functions with a user-defined
6826 section attribute and on any architecture that does not support named
6829 @item -freorder-functions
6830 @opindex freorder-functions
6831 Reorder functions in the object file in order to
6832 improve code locality. This is implemented by using special
6833 subsections @code{.text.hot} for most frequently executed functions and
6834 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6835 the linker so object file format must support named sections and linker must
6836 place them in a reasonable way.
6838 Also profile feedback must be available in to make this option effective. See
6839 @option{-fprofile-arcs} for details.
6841 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6843 @item -fstrict-aliasing
6844 @opindex fstrict-aliasing
6845 Allows the compiler to assume the strictest aliasing rules applicable to
6846 the language being compiled. For C (and C++), this activates
6847 optimizations based on the type of expressions. In particular, an
6848 object of one type is assumed never to reside at the same address as an
6849 object of a different type, unless the types are almost the same. For
6850 example, an @code{unsigned int} can alias an @code{int}, but not a
6851 @code{void*} or a @code{double}. A character type may alias any other
6854 @anchor{Type-punning}Pay special attention to code like this:
6867 The practice of reading from a different union member than the one most
6868 recently written to (called ``type-punning'') is common. Even with
6869 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6870 is accessed through the union type. So, the code above will work as
6871 expected. @xref{Structures unions enumerations and bit-fields
6872 implementation}. However, this code might not:
6883 Similarly, access by taking the address, casting the resulting pointer
6884 and dereferencing the result has undefined behavior, even if the cast
6885 uses a union type, e.g.:
6889 return ((union a_union *) &d)->i;
6893 The @option{-fstrict-aliasing} option is enabled at levels
6894 @option{-O2}, @option{-O3}, @option{-Os}.
6896 @item -fstrict-overflow
6897 @opindex fstrict-overflow
6898 Allow the compiler to assume strict signed overflow rules, depending
6899 on the language being compiled. For C (and C++) this means that
6900 overflow when doing arithmetic with signed numbers is undefined, which
6901 means that the compiler may assume that it will not happen. This
6902 permits various optimizations. For example, the compiler will assume
6903 that an expression like @code{i + 10 > i} will always be true for
6904 signed @code{i}. This assumption is only valid if signed overflow is
6905 undefined, as the expression is false if @code{i + 10} overflows when
6906 using twos complement arithmetic. When this option is in effect any
6907 attempt to determine whether an operation on signed numbers will
6908 overflow must be written carefully to not actually involve overflow.
6910 This option also allows the compiler to assume strict pointer
6911 semantics: given a pointer to an object, if adding an offset to that
6912 pointer does not produce a pointer to the same object, the addition is
6913 undefined. This permits the compiler to conclude that @code{p + u >
6914 p} is always true for a pointer @code{p} and unsigned integer
6915 @code{u}. This assumption is only valid because pointer wraparound is
6916 undefined, as the expression is false if @code{p + u} overflows using
6917 twos complement arithmetic.
6919 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6920 that integer signed overflow is fully defined: it wraps. When
6921 @option{-fwrapv} is used, there is no difference between
6922 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6923 integers. With @option{-fwrapv} certain types of overflow are
6924 permitted. For example, if the compiler gets an overflow when doing
6925 arithmetic on constants, the overflowed value can still be used with
6926 @option{-fwrapv}, but not otherwise.
6928 The @option{-fstrict-overflow} option is enabled at levels
6929 @option{-O2}, @option{-O3}, @option{-Os}.
6931 @item -falign-functions
6932 @itemx -falign-functions=@var{n}
6933 @opindex falign-functions
6934 Align the start of functions to the next power-of-two greater than
6935 @var{n}, skipping up to @var{n} bytes. For instance,
6936 @option{-falign-functions=32} aligns functions to the next 32-byte
6937 boundary, but @option{-falign-functions=24} would align to the next
6938 32-byte boundary only if this can be done by skipping 23 bytes or less.
6940 @option{-fno-align-functions} and @option{-falign-functions=1} are
6941 equivalent and mean that functions will not be aligned.
6943 Some assemblers only support this flag when @var{n} is a power of two;
6944 in that case, it is rounded up.
6946 If @var{n} is not specified or is zero, use a machine-dependent default.
6948 Enabled at levels @option{-O2}, @option{-O3}.
6950 @item -falign-labels
6951 @itemx -falign-labels=@var{n}
6952 @opindex falign-labels
6953 Align all branch targets to a power-of-two boundary, skipping up to
6954 @var{n} bytes like @option{-falign-functions}. This option can easily
6955 make code slower, because it must insert dummy operations for when the
6956 branch target is reached in the usual flow of the code.
6958 @option{-fno-align-labels} and @option{-falign-labels=1} are
6959 equivalent and mean that labels will not be aligned.
6961 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6962 are greater than this value, then their values are used instead.
6964 If @var{n} is not specified or is zero, use a machine-dependent default
6965 which is very likely to be @samp{1}, meaning no alignment.
6967 Enabled at levels @option{-O2}, @option{-O3}.
6970 @itemx -falign-loops=@var{n}
6971 @opindex falign-loops
6972 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6973 like @option{-falign-functions}. The hope is that the loop will be
6974 executed many times, which will make up for any execution of the dummy
6977 @option{-fno-align-loops} and @option{-falign-loops=1} are
6978 equivalent and mean that loops will not be aligned.
6980 If @var{n} is not specified or is zero, use a machine-dependent default.
6982 Enabled at levels @option{-O2}, @option{-O3}.
6985 @itemx -falign-jumps=@var{n}
6986 @opindex falign-jumps
6987 Align branch targets to a power-of-two boundary, for branch targets
6988 where the targets can only be reached by jumping, skipping up to @var{n}
6989 bytes like @option{-falign-functions}. In this case, no dummy operations
6992 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6993 equivalent and mean that loops will not be aligned.
6995 If @var{n} is not specified or is zero, use a machine-dependent default.
6997 Enabled at levels @option{-O2}, @option{-O3}.
6999 @item -funit-at-a-time
7000 @opindex funit-at-a-time
7001 This option is left for compatibility reasons. @option{-funit-at-a-time}
7002 has no effect, while @option{-fno-unit-at-a-time} implies
7003 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7007 @item -fno-toplevel-reorder
7008 @opindex fno-toplevel-reorder
7009 Do not reorder top-level functions, variables, and @code{asm}
7010 statements. Output them in the same order that they appear in the
7011 input file. When this option is used, unreferenced static variables
7012 will not be removed. This option is intended to support existing code
7013 which relies on a particular ordering. For new code, it is better to
7016 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7017 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7022 Constructs webs as commonly used for register allocation purposes and assign
7023 each web individual pseudo register. This allows the register allocation pass
7024 to operate on pseudos directly, but also strengthens several other optimization
7025 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7026 however, make debugging impossible, since variables will no longer stay in a
7029 Enabled by default with @option{-funroll-loops}.
7031 @item -fwhole-program
7032 @opindex fwhole-program
7033 Assume that the current compilation unit represents the whole program being
7034 compiled. All public functions and variables with the exception of @code{main}
7035 and those merged by attribute @code{externally_visible} become static functions
7036 and in effect are optimized more aggressively by interprocedural optimizers.
7037 While this option is equivalent to proper use of the @code{static} keyword for
7038 programs consisting of a single file, in combination with option
7039 @option{--combine} this flag can be used to compile many smaller scale C
7040 programs since the functions and variables become local for the whole combined
7041 compilation unit, not for the single source file itself.
7043 This option implies @option{-fwhole-file} for Fortran programs.
7045 @item -fcprop-registers
7046 @opindex fcprop-registers
7047 After register allocation and post-register allocation instruction splitting,
7048 we perform a copy-propagation pass to try to reduce scheduling dependencies
7049 and occasionally eliminate the copy.
7051 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7053 @item -fprofile-correction
7054 @opindex fprofile-correction
7055 Profiles collected using an instrumented binary for multi-threaded programs may
7056 be inconsistent due to missed counter updates. When this option is specified,
7057 GCC will use heuristics to correct or smooth out such inconsistencies. By
7058 default, GCC will emit an error message when an inconsistent profile is detected.
7060 @item -fprofile-dir=@var{path}
7061 @opindex fprofile-dir
7063 Set the directory to search the profile data files in to @var{path}.
7064 This option affects only the profile data generated by
7065 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7066 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7067 and its related options.
7068 By default, GCC will use the current directory as @var{path}
7069 thus the profile data file will appear in the same directory as the object file.
7071 @item -fprofile-generate
7072 @itemx -fprofile-generate=@var{path}
7073 @opindex fprofile-generate
7075 Enable options usually used for instrumenting application to produce
7076 profile useful for later recompilation with profile feedback based
7077 optimization. You must use @option{-fprofile-generate} both when
7078 compiling and when linking your program.
7080 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7082 If @var{path} is specified, GCC will look at the @var{path} to find
7083 the profile feedback data files. See @option{-fprofile-dir}.
7086 @itemx -fprofile-use=@var{path}
7087 @opindex fprofile-use
7088 Enable profile feedback directed optimizations, and optimizations
7089 generally profitable only with profile feedback available.
7091 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7092 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7094 By default, GCC emits an error message if the feedback profiles do not
7095 match the source code. This error can be turned into a warning by using
7096 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7099 If @var{path} is specified, GCC will look at the @var{path} to find
7100 the profile feedback data files. See @option{-fprofile-dir}.
7103 The following options control compiler behavior regarding floating
7104 point arithmetic. These options trade off between speed and
7105 correctness. All must be specifically enabled.
7109 @opindex ffloat-store
7110 Do not store floating point variables in registers, and inhibit other
7111 options that might change whether a floating point value is taken from a
7114 @cindex floating point precision
7115 This option prevents undesirable excess precision on machines such as
7116 the 68000 where the floating registers (of the 68881) keep more
7117 precision than a @code{double} is supposed to have. Similarly for the
7118 x86 architecture. For most programs, the excess precision does only
7119 good, but a few programs rely on the precise definition of IEEE floating
7120 point. Use @option{-ffloat-store} for such programs, after modifying
7121 them to store all pertinent intermediate computations into variables.
7123 @item -fexcess-precision=@var{style}
7124 @opindex fexcess-precision
7125 This option allows further control over excess precision on machines
7126 where floating-point registers have more precision than the IEEE
7127 @code{float} and @code{double} types and the processor does not
7128 support operations rounding to those types. By default,
7129 @option{-fexcess-precision=fast} is in effect; this means that
7130 operations are carried out in the precision of the registers and that
7131 it is unpredictable when rounding to the types specified in the source
7132 code takes place. When compiling C, if
7133 @option{-fexcess-precision=standard} is specified then excess
7134 precision will follow the rules specified in ISO C99; in particular,
7135 both casts and assignments cause values to be rounded to their
7136 semantic types (whereas @option{-ffloat-store} only affects
7137 assignments). This option is enabled by default for C if a strict
7138 conformance option such as @option{-std=c99} is used.
7141 @option{-fexcess-precision=standard} is not implemented for languages
7142 other than C, and has no effect if
7143 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7144 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7145 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7146 semantics apply without excess precision, and in the latter, rounding
7151 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7152 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7153 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7155 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7157 This option is not turned on by any @option{-O} option since
7158 it can result in incorrect output for programs which depend on
7159 an exact implementation of IEEE or ISO rules/specifications for
7160 math functions. It may, however, yield faster code for programs
7161 that do not require the guarantees of these specifications.
7163 @item -fno-math-errno
7164 @opindex fno-math-errno
7165 Do not set ERRNO after calling math functions that are executed
7166 with a single instruction, e.g., sqrt. A program that relies on
7167 IEEE exceptions for math error handling may want to use this flag
7168 for speed while maintaining IEEE arithmetic compatibility.
7170 This option is not turned on by any @option{-O} option since
7171 it can result in incorrect output for programs which depend on
7172 an exact implementation of IEEE or ISO rules/specifications for
7173 math functions. It may, however, yield faster code for programs
7174 that do not require the guarantees of these specifications.
7176 The default is @option{-fmath-errno}.
7178 On Darwin systems, the math library never sets @code{errno}. There is
7179 therefore no reason for the compiler to consider the possibility that
7180 it might, and @option{-fno-math-errno} is the default.
7182 @item -funsafe-math-optimizations
7183 @opindex funsafe-math-optimizations
7185 Allow optimizations for floating-point arithmetic that (a) assume
7186 that arguments and results are valid and (b) may violate IEEE or
7187 ANSI standards. When used at link-time, it may include libraries
7188 or startup files that change the default FPU control word or other
7189 similar optimizations.
7191 This option is not turned on by any @option{-O} option since
7192 it can result in incorrect output for programs which depend on
7193 an exact implementation of IEEE or ISO rules/specifications for
7194 math functions. It may, however, yield faster code for programs
7195 that do not require the guarantees of these specifications.
7196 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7197 @option{-fassociative-math} and @option{-freciprocal-math}.
7199 The default is @option{-fno-unsafe-math-optimizations}.
7201 @item -fassociative-math
7202 @opindex fassociative-math
7204 Allow re-association of operands in series of floating-point operations.
7205 This violates the ISO C and C++ language standard by possibly changing
7206 computation result. NOTE: re-ordering may change the sign of zero as
7207 well as ignore NaNs and inhibit or create underflow or overflow (and
7208 thus cannot be used on a code which relies on rounding behavior like
7209 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7210 and thus may not be used when ordered comparisons are required.
7211 This option requires that both @option{-fno-signed-zeros} and
7212 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7213 much sense with @option{-frounding-math}.
7215 The default is @option{-fno-associative-math}.
7217 @item -freciprocal-math
7218 @opindex freciprocal-math
7220 Allow the reciprocal of a value to be used instead of dividing by
7221 the value if this enables optimizations. For example @code{x / y}
7222 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7223 is subject to common subexpression elimination. Note that this loses
7224 precision and increases the number of flops operating on the value.
7226 The default is @option{-fno-reciprocal-math}.
7228 @item -ffinite-math-only
7229 @opindex ffinite-math-only
7230 Allow optimizations for floating-point arithmetic that assume
7231 that arguments and results are not NaNs or +-Infs.
7233 This option is not turned on by any @option{-O} option since
7234 it can result in incorrect output for programs which depend on
7235 an exact implementation of IEEE or ISO rules/specifications for
7236 math functions. It may, however, yield faster code for programs
7237 that do not require the guarantees of these specifications.
7239 The default is @option{-fno-finite-math-only}.
7241 @item -fno-signed-zeros
7242 @opindex fno-signed-zeros
7243 Allow optimizations for floating point arithmetic that ignore the
7244 signedness of zero. IEEE arithmetic specifies the behavior of
7245 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7246 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7247 This option implies that the sign of a zero result isn't significant.
7249 The default is @option{-fsigned-zeros}.
7251 @item -fno-trapping-math
7252 @opindex fno-trapping-math
7253 Compile code assuming that floating-point operations cannot generate
7254 user-visible traps. These traps include division by zero, overflow,
7255 underflow, inexact result and invalid operation. This option requires
7256 that @option{-fno-signaling-nans} be in effect. Setting this option may
7257 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7259 This option should never be turned on by any @option{-O} option since
7260 it can result in incorrect output for programs which depend on
7261 an exact implementation of IEEE or ISO rules/specifications for
7264 The default is @option{-ftrapping-math}.
7266 @item -frounding-math
7267 @opindex frounding-math
7268 Disable transformations and optimizations that assume default floating
7269 point rounding behavior. This is round-to-zero for all floating point
7270 to integer conversions, and round-to-nearest for all other arithmetic
7271 truncations. This option should be specified for programs that change
7272 the FP rounding mode dynamically, or that may be executed with a
7273 non-default rounding mode. This option disables constant folding of
7274 floating point expressions at compile-time (which may be affected by
7275 rounding mode) and arithmetic transformations that are unsafe in the
7276 presence of sign-dependent rounding modes.
7278 The default is @option{-fno-rounding-math}.
7280 This option is experimental and does not currently guarantee to
7281 disable all GCC optimizations that are affected by rounding mode.
7282 Future versions of GCC may provide finer control of this setting
7283 using C99's @code{FENV_ACCESS} pragma. This command line option
7284 will be used to specify the default state for @code{FENV_ACCESS}.
7286 @item -fsignaling-nans
7287 @opindex fsignaling-nans
7288 Compile code assuming that IEEE signaling NaNs may generate user-visible
7289 traps during floating-point operations. Setting this option disables
7290 optimizations that may change the number of exceptions visible with
7291 signaling NaNs. This option implies @option{-ftrapping-math}.
7293 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7296 The default is @option{-fno-signaling-nans}.
7298 This option is experimental and does not currently guarantee to
7299 disable all GCC optimizations that affect signaling NaN behavior.
7301 @item -fsingle-precision-constant
7302 @opindex fsingle-precision-constant
7303 Treat floating point constant as single precision constant instead of
7304 implicitly converting it to double precision constant.
7306 @item -fcx-limited-range
7307 @opindex fcx-limited-range
7308 When enabled, this option states that a range reduction step is not
7309 needed when performing complex division. Also, there is no checking
7310 whether the result of a complex multiplication or division is @code{NaN
7311 + I*NaN}, with an attempt to rescue the situation in that case. The
7312 default is @option{-fno-cx-limited-range}, but is enabled by
7313 @option{-ffast-math}.
7315 This option controls the default setting of the ISO C99
7316 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7319 @item -fcx-fortran-rules
7320 @opindex fcx-fortran-rules
7321 Complex multiplication and division follow Fortran rules. Range
7322 reduction is done as part of complex division, but there is no checking
7323 whether the result of a complex multiplication or division is @code{NaN
7324 + I*NaN}, with an attempt to rescue the situation in that case.
7326 The default is @option{-fno-cx-fortran-rules}.
7330 The following options control optimizations that may improve
7331 performance, but are not enabled by any @option{-O} options. This
7332 section includes experimental options that may produce broken code.
7335 @item -fbranch-probabilities
7336 @opindex fbranch-probabilities
7337 After running a program compiled with @option{-fprofile-arcs}
7338 (@pxref{Debugging Options,, Options for Debugging Your Program or
7339 @command{gcc}}), you can compile it a second time using
7340 @option{-fbranch-probabilities}, to improve optimizations based on
7341 the number of times each branch was taken. When the program
7342 compiled with @option{-fprofile-arcs} exits it saves arc execution
7343 counts to a file called @file{@var{sourcename}.gcda} for each source
7344 file. The information in this data file is very dependent on the
7345 structure of the generated code, so you must use the same source code
7346 and the same optimization options for both compilations.
7348 With @option{-fbranch-probabilities}, GCC puts a
7349 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7350 These can be used to improve optimization. Currently, they are only
7351 used in one place: in @file{reorg.c}, instead of guessing which path a
7352 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7353 exactly determine which path is taken more often.
7355 @item -fprofile-values
7356 @opindex fprofile-values
7357 If combined with @option{-fprofile-arcs}, it adds code so that some
7358 data about values of expressions in the program is gathered.
7360 With @option{-fbranch-probabilities}, it reads back the data gathered
7361 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7362 notes to instructions for their later usage in optimizations.
7364 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7368 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7369 a code to gather information about values of expressions.
7371 With @option{-fbranch-probabilities}, it reads back the data gathered
7372 and actually performs the optimizations based on them.
7373 Currently the optimizations include specialization of division operation
7374 using the knowledge about the value of the denominator.
7376 @item -frename-registers
7377 @opindex frename-registers
7378 Attempt to avoid false dependencies in scheduled code by making use
7379 of registers left over after register allocation. This optimization
7380 will most benefit processors with lots of registers. Depending on the
7381 debug information format adopted by the target, however, it can
7382 make debugging impossible, since variables will no longer stay in
7383 a ``home register''.
7385 Enabled by default with @option{-funroll-loops}.
7389 Perform tail duplication to enlarge superblock size. This transformation
7390 simplifies the control flow of the function allowing other optimizations to do
7393 Enabled with @option{-fprofile-use}.
7395 @item -funroll-loops
7396 @opindex funroll-loops
7397 Unroll loops whose number of iterations can be determined at compile time or
7398 upon entry to the loop. @option{-funroll-loops} implies
7399 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7400 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7401 small constant number of iterations). This option makes code larger, and may
7402 or may not make it run faster.
7404 Enabled with @option{-fprofile-use}.
7406 @item -funroll-all-loops
7407 @opindex funroll-all-loops
7408 Unroll all loops, even if their number of iterations is uncertain when
7409 the loop is entered. This usually makes programs run more slowly.
7410 @option{-funroll-all-loops} implies the same options as
7411 @option{-funroll-loops}.
7414 @opindex fpeel-loops
7415 Peels the loops for that there is enough information that they do not
7416 roll much (from profile feedback). It also turns on complete loop peeling
7417 (i.e.@: complete removal of loops with small constant number of iterations).
7419 Enabled with @option{-fprofile-use}.
7421 @item -fmove-loop-invariants
7422 @opindex fmove-loop-invariants
7423 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7424 at level @option{-O1}
7426 @item -funswitch-loops
7427 @opindex funswitch-loops
7428 Move branches with loop invariant conditions out of the loop, with duplicates
7429 of the loop on both branches (modified according to result of the condition).
7431 @item -ffunction-sections
7432 @itemx -fdata-sections
7433 @opindex ffunction-sections
7434 @opindex fdata-sections
7435 Place each function or data item into its own section in the output
7436 file if the target supports arbitrary sections. The name of the
7437 function or the name of the data item determines the section's name
7440 Use these options on systems where the linker can perform optimizations
7441 to improve locality of reference in the instruction space. Most systems
7442 using the ELF object format and SPARC processors running Solaris 2 have
7443 linkers with such optimizations. AIX may have these optimizations in
7446 Only use these options when there are significant benefits from doing
7447 so. When you specify these options, the assembler and linker will
7448 create larger object and executable files and will also be slower.
7449 You will not be able to use @code{gprof} on all systems if you
7450 specify this option and you may have problems with debugging if
7451 you specify both this option and @option{-g}.
7453 @item -fbranch-target-load-optimize
7454 @opindex fbranch-target-load-optimize
7455 Perform branch target register load optimization before prologue / epilogue
7457 The use of target registers can typically be exposed only during reload,
7458 thus hoisting loads out of loops and doing inter-block scheduling needs
7459 a separate optimization pass.
7461 @item -fbranch-target-load-optimize2
7462 @opindex fbranch-target-load-optimize2
7463 Perform branch target register load optimization after prologue / epilogue
7466 @item -fbtr-bb-exclusive
7467 @opindex fbtr-bb-exclusive
7468 When performing branch target register load optimization, don't reuse
7469 branch target registers in within any basic block.
7471 @item -fstack-protector
7472 @opindex fstack-protector
7473 Emit extra code to check for buffer overflows, such as stack smashing
7474 attacks. This is done by adding a guard variable to functions with
7475 vulnerable objects. This includes functions that call alloca, and
7476 functions with buffers larger than 8 bytes. The guards are initialized
7477 when a function is entered and then checked when the function exits.
7478 If a guard check fails, an error message is printed and the program exits.
7480 @item -fstack-protector-all
7481 @opindex fstack-protector-all
7482 Like @option{-fstack-protector} except that all functions are protected.
7484 @item -fsection-anchors
7485 @opindex fsection-anchors
7486 Try to reduce the number of symbolic address calculations by using
7487 shared ``anchor'' symbols to address nearby objects. This transformation
7488 can help to reduce the number of GOT entries and GOT accesses on some
7491 For example, the implementation of the following function @code{foo}:
7495 int foo (void) @{ return a + b + c; @}
7498 would usually calculate the addresses of all three variables, but if you
7499 compile it with @option{-fsection-anchors}, it will access the variables
7500 from a common anchor point instead. The effect is similar to the
7501 following pseudocode (which isn't valid C):
7506 register int *xr = &x;
7507 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7511 Not all targets support this option.
7513 @item --param @var{name}=@var{value}
7515 In some places, GCC uses various constants to control the amount of
7516 optimization that is done. For example, GCC will not inline functions
7517 that contain more that a certain number of instructions. You can
7518 control some of these constants on the command-line using the
7519 @option{--param} option.
7521 The names of specific parameters, and the meaning of the values, are
7522 tied to the internals of the compiler, and are subject to change
7523 without notice in future releases.
7525 In each case, the @var{value} is an integer. The allowable choices for
7526 @var{name} are given in the following table:
7529 @item struct-reorg-cold-struct-ratio
7530 The threshold ratio (as a percentage) between a structure frequency
7531 and the frequency of the hottest structure in the program. This parameter
7532 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7533 We say that if the ratio of a structure frequency, calculated by profiling,
7534 to the hottest structure frequency in the program is less than this
7535 parameter, then structure reorganization is not applied to this structure.
7538 @item predictable-branch-cost-outcome
7539 When branch is predicted to be taken with probability lower than this threshold
7540 (in percent), then it is considered well predictable. The default is 10.
7542 @item max-crossjump-edges
7543 The maximum number of incoming edges to consider for crossjumping.
7544 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7545 the number of edges incoming to each block. Increasing values mean
7546 more aggressive optimization, making the compile time increase with
7547 probably small improvement in executable size.
7549 @item min-crossjump-insns
7550 The minimum number of instructions which must be matched at the end
7551 of two blocks before crossjumping will be performed on them. This
7552 value is ignored in the case where all instructions in the block being
7553 crossjumped from are matched. The default value is 5.
7555 @item max-grow-copy-bb-insns
7556 The maximum code size expansion factor when copying basic blocks
7557 instead of jumping. The expansion is relative to a jump instruction.
7558 The default value is 8.
7560 @item max-goto-duplication-insns
7561 The maximum number of instructions to duplicate to a block that jumps
7562 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7563 passes, GCC factors computed gotos early in the compilation process,
7564 and unfactors them as late as possible. Only computed jumps at the
7565 end of a basic blocks with no more than max-goto-duplication-insns are
7566 unfactored. The default value is 8.
7568 @item max-delay-slot-insn-search
7569 The maximum number of instructions to consider when looking for an
7570 instruction to fill a delay slot. If more than this arbitrary number of
7571 instructions is searched, the time savings from filling the delay slot
7572 will be minimal so stop searching. Increasing values mean more
7573 aggressive optimization, making the compile time increase with probably
7574 small improvement in executable run time.
7576 @item max-delay-slot-live-search
7577 When trying to fill delay slots, the maximum number of instructions to
7578 consider when searching for a block with valid live register
7579 information. Increasing this arbitrarily chosen value means more
7580 aggressive optimization, increasing the compile time. This parameter
7581 should be removed when the delay slot code is rewritten to maintain the
7584 @item max-gcse-memory
7585 The approximate maximum amount of memory that will be allocated in
7586 order to perform the global common subexpression elimination
7587 optimization. If more memory than specified is required, the
7588 optimization will not be done.
7590 @item max-pending-list-length
7591 The maximum number of pending dependencies scheduling will allow
7592 before flushing the current state and starting over. Large functions
7593 with few branches or calls can create excessively large lists which
7594 needlessly consume memory and resources.
7596 @item max-inline-insns-single
7597 Several parameters control the tree inliner used in gcc.
7598 This number sets the maximum number of instructions (counted in GCC's
7599 internal representation) in a single function that the tree inliner
7600 will consider for inlining. This only affects functions declared
7601 inline and methods implemented in a class declaration (C++).
7602 The default value is 300.
7604 @item max-inline-insns-auto
7605 When you use @option{-finline-functions} (included in @option{-O3}),
7606 a lot of functions that would otherwise not be considered for inlining
7607 by the compiler will be investigated. To those functions, a different
7608 (more restrictive) limit compared to functions declared inline can
7610 The default value is 60.
7612 @item large-function-insns
7613 The limit specifying really large functions. For functions larger than this
7614 limit after inlining, inlining is constrained by
7615 @option{--param large-function-growth}. This parameter is useful primarily
7616 to avoid extreme compilation time caused by non-linear algorithms used by the
7618 The default value is 2700.
7620 @item large-function-growth
7621 Specifies maximal growth of large function caused by inlining in percents.
7622 The default value is 100 which limits large function growth to 2.0 times
7625 @item large-unit-insns
7626 The limit specifying large translation unit. Growth caused by inlining of
7627 units larger than this limit is limited by @option{--param inline-unit-growth}.
7628 For small units this might be too tight (consider unit consisting of function A
7629 that is inline and B that just calls A three time. If B is small relative to
7630 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7631 large units consisting of small inlineable functions however the overall unit
7632 growth limit is needed to avoid exponential explosion of code size. Thus for
7633 smaller units, the size is increased to @option{--param large-unit-insns}
7634 before applying @option{--param inline-unit-growth}. The default is 10000
7636 @item inline-unit-growth
7637 Specifies maximal overall growth of the compilation unit caused by inlining.
7638 The default value is 30 which limits unit growth to 1.3 times the original
7641 @item ipcp-unit-growth
7642 Specifies maximal overall growth of the compilation unit caused by
7643 interprocedural constant propagation. The default value is 10 which limits
7644 unit growth to 1.1 times the original size.
7646 @item large-stack-frame
7647 The limit specifying large stack frames. While inlining the algorithm is trying
7648 to not grow past this limit too much. Default value is 256 bytes.
7650 @item large-stack-frame-growth
7651 Specifies maximal growth of large stack frames caused by inlining in percents.
7652 The default value is 1000 which limits large stack frame growth to 11 times
7655 @item max-inline-insns-recursive
7656 @itemx max-inline-insns-recursive-auto
7657 Specifies maximum number of instructions out-of-line copy of self recursive inline
7658 function can grow into by performing recursive inlining.
7660 For functions declared inline @option{--param max-inline-insns-recursive} is
7661 taken into account. For function not declared inline, recursive inlining
7662 happens only when @option{-finline-functions} (included in @option{-O3}) is
7663 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7664 default value is 450.
7666 @item max-inline-recursive-depth
7667 @itemx max-inline-recursive-depth-auto
7668 Specifies maximum recursion depth used by the recursive inlining.
7670 For functions declared inline @option{--param max-inline-recursive-depth} is
7671 taken into account. For function not declared inline, recursive inlining
7672 happens only when @option{-finline-functions} (included in @option{-O3}) is
7673 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7676 @item min-inline-recursive-probability
7677 Recursive inlining is profitable only for function having deep recursion
7678 in average and can hurt for function having little recursion depth by
7679 increasing the prologue size or complexity of function body to other
7682 When profile feedback is available (see @option{-fprofile-generate}) the actual
7683 recursion depth can be guessed from probability that function will recurse via
7684 given call expression. This parameter limits inlining only to call expression
7685 whose probability exceeds given threshold (in percents). The default value is
7688 @item early-inlining-insns
7689 Specify growth that early inliner can make. In effect it increases amount of
7690 inlining for code having large abstraction penalty. The default value is 12.
7692 @item max-early-inliner-iterations
7693 @itemx max-early-inliner-iterations
7694 Limit of iterations of early inliner. This basically bounds number of nested
7695 indirect calls early inliner can resolve. Deeper chains are still handled by
7698 @item min-vect-loop-bound
7699 The minimum number of iterations under which a loop will not get vectorized
7700 when @option{-ftree-vectorize} is used. The number of iterations after
7701 vectorization needs to be greater than the value specified by this option
7702 to allow vectorization. The default value is 0.
7704 @item max-unrolled-insns
7705 The maximum number of instructions that a loop should have if that loop
7706 is unrolled, and if the loop is unrolled, it determines how many times
7707 the loop code is unrolled.
7709 @item max-average-unrolled-insns
7710 The maximum number of instructions biased by probabilities of their execution
7711 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7712 it determines how many times the loop code is unrolled.
7714 @item max-unroll-times
7715 The maximum number of unrollings of a single loop.
7717 @item max-peeled-insns
7718 The maximum number of instructions that a loop should have if that loop
7719 is peeled, and if the loop is peeled, it determines how many times
7720 the loop code is peeled.
7722 @item max-peel-times
7723 The maximum number of peelings of a single loop.
7725 @item max-completely-peeled-insns
7726 The maximum number of insns of a completely peeled loop.
7728 @item max-completely-peel-times
7729 The maximum number of iterations of a loop to be suitable for complete peeling.
7731 @item max-unswitch-insns
7732 The maximum number of insns of an unswitched loop.
7734 @item max-unswitch-level
7735 The maximum number of branches unswitched in a single loop.
7738 The minimum cost of an expensive expression in the loop invariant motion.
7740 @item iv-consider-all-candidates-bound
7741 Bound on number of candidates for induction variables below that
7742 all candidates are considered for each use in induction variable
7743 optimizations. Only the most relevant candidates are considered
7744 if there are more candidates, to avoid quadratic time complexity.
7746 @item iv-max-considered-uses
7747 The induction variable optimizations give up on loops that contain more
7748 induction variable uses.
7750 @item iv-always-prune-cand-set-bound
7751 If number of candidates in the set is smaller than this value,
7752 we always try to remove unnecessary ivs from the set during its
7753 optimization when a new iv is added to the set.
7755 @item scev-max-expr-size
7756 Bound on size of expressions used in the scalar evolutions analyzer.
7757 Large expressions slow the analyzer.
7759 @item omega-max-vars
7760 The maximum number of variables in an Omega constraint system.
7761 The default value is 128.
7763 @item omega-max-geqs
7764 The maximum number of inequalities in an Omega constraint system.
7765 The default value is 256.
7768 The maximum number of equalities in an Omega constraint system.
7769 The default value is 128.
7771 @item omega-max-wild-cards
7772 The maximum number of wildcard variables that the Omega solver will
7773 be able to insert. The default value is 18.
7775 @item omega-hash-table-size
7776 The size of the hash table in the Omega solver. The default value is
7779 @item omega-max-keys
7780 The maximal number of keys used by the Omega solver. The default
7783 @item omega-eliminate-redundant-constraints
7784 When set to 1, use expensive methods to eliminate all redundant
7785 constraints. The default value is 0.
7787 @item vect-max-version-for-alignment-checks
7788 The maximum number of runtime checks that can be performed when
7789 doing loop versioning for alignment in the vectorizer. See option
7790 ftree-vect-loop-version for more information.
7792 @item vect-max-version-for-alias-checks
7793 The maximum number of runtime checks that can be performed when
7794 doing loop versioning for alias in the vectorizer. See option
7795 ftree-vect-loop-version for more information.
7797 @item max-iterations-to-track
7799 The maximum number of iterations of a loop the brute force algorithm
7800 for analysis of # of iterations of the loop tries to evaluate.
7802 @item hot-bb-count-fraction
7803 Select fraction of the maximal count of repetitions of basic block in program
7804 given basic block needs to have to be considered hot.
7806 @item hot-bb-frequency-fraction
7807 Select fraction of the maximal frequency of executions of basic block in
7808 function given basic block needs to have to be considered hot
7810 @item max-predicted-iterations
7811 The maximum number of loop iterations we predict statically. This is useful
7812 in cases where function contain single loop with known bound and other loop
7813 with unknown. We predict the known number of iterations correctly, while
7814 the unknown number of iterations average to roughly 10. This means that the
7815 loop without bounds would appear artificially cold relative to the other one.
7817 @item align-threshold
7819 Select fraction of the maximal frequency of executions of basic block in
7820 function given basic block will get aligned.
7822 @item align-loop-iterations
7824 A loop expected to iterate at lest the selected number of iterations will get
7827 @item tracer-dynamic-coverage
7828 @itemx tracer-dynamic-coverage-feedback
7830 This value is used to limit superblock formation once the given percentage of
7831 executed instructions is covered. This limits unnecessary code size
7834 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7835 feedback is available. The real profiles (as opposed to statically estimated
7836 ones) are much less balanced allowing the threshold to be larger value.
7838 @item tracer-max-code-growth
7839 Stop tail duplication once code growth has reached given percentage. This is
7840 rather hokey argument, as most of the duplicates will be eliminated later in
7841 cross jumping, so it may be set to much higher values than is the desired code
7844 @item tracer-min-branch-ratio
7846 Stop reverse growth when the reverse probability of best edge is less than this
7847 threshold (in percent).
7849 @item tracer-min-branch-ratio
7850 @itemx tracer-min-branch-ratio-feedback
7852 Stop forward growth if the best edge do have probability lower than this
7855 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7856 compilation for profile feedback and one for compilation without. The value
7857 for compilation with profile feedback needs to be more conservative (higher) in
7858 order to make tracer effective.
7860 @item max-cse-path-length
7862 Maximum number of basic blocks on path that cse considers. The default is 10.
7865 The maximum instructions CSE process before flushing. The default is 1000.
7867 @item ggc-min-expand
7869 GCC uses a garbage collector to manage its own memory allocation. This
7870 parameter specifies the minimum percentage by which the garbage
7871 collector's heap should be allowed to expand between collections.
7872 Tuning this may improve compilation speed; it has no effect on code
7875 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7876 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7877 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7878 GCC is not able to calculate RAM on a particular platform, the lower
7879 bound of 30% is used. Setting this parameter and
7880 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7881 every opportunity. This is extremely slow, but can be useful for
7884 @item ggc-min-heapsize
7886 Minimum size of the garbage collector's heap before it begins bothering
7887 to collect garbage. The first collection occurs after the heap expands
7888 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7889 tuning this may improve compilation speed, and has no effect on code
7892 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7893 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7894 with a lower bound of 4096 (four megabytes) and an upper bound of
7895 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7896 particular platform, the lower bound is used. Setting this parameter
7897 very large effectively disables garbage collection. Setting this
7898 parameter and @option{ggc-min-expand} to zero causes a full collection
7899 to occur at every opportunity.
7901 @item max-reload-search-insns
7902 The maximum number of instruction reload should look backward for equivalent
7903 register. Increasing values mean more aggressive optimization, making the
7904 compile time increase with probably slightly better performance. The default
7907 @item max-cselib-memory-locations
7908 The maximum number of memory locations cselib should take into account.
7909 Increasing values mean more aggressive optimization, making the compile time
7910 increase with probably slightly better performance. The default value is 500.
7912 @item reorder-blocks-duplicate
7913 @itemx reorder-blocks-duplicate-feedback
7915 Used by basic block reordering pass to decide whether to use unconditional
7916 branch or duplicate the code on its destination. Code is duplicated when its
7917 estimated size is smaller than this value multiplied by the estimated size of
7918 unconditional jump in the hot spots of the program.
7920 The @option{reorder-block-duplicate-feedback} is used only when profile
7921 feedback is available and may be set to higher values than
7922 @option{reorder-block-duplicate} since information about the hot spots is more
7925 @item max-sched-ready-insns
7926 The maximum number of instructions ready to be issued the scheduler should
7927 consider at any given time during the first scheduling pass. Increasing
7928 values mean more thorough searches, making the compilation time increase
7929 with probably little benefit. The default value is 100.
7931 @item max-sched-region-blocks
7932 The maximum number of blocks in a region to be considered for
7933 interblock scheduling. The default value is 10.
7935 @item max-pipeline-region-blocks
7936 The maximum number of blocks in a region to be considered for
7937 pipelining in the selective scheduler. The default value is 15.
7939 @item max-sched-region-insns
7940 The maximum number of insns in a region to be considered for
7941 interblock scheduling. The default value is 100.
7943 @item max-pipeline-region-insns
7944 The maximum number of insns in a region to be considered for
7945 pipelining in the selective scheduler. The default value is 200.
7948 The minimum probability (in percents) of reaching a source block
7949 for interblock speculative scheduling. The default value is 40.
7951 @item max-sched-extend-regions-iters
7952 The maximum number of iterations through CFG to extend regions.
7953 0 - disable region extension,
7954 N - do at most N iterations.
7955 The default value is 0.
7957 @item max-sched-insn-conflict-delay
7958 The maximum conflict delay for an insn to be considered for speculative motion.
7959 The default value is 3.
7961 @item sched-spec-prob-cutoff
7962 The minimal probability of speculation success (in percents), so that
7963 speculative insn will be scheduled.
7964 The default value is 40.
7966 @item sched-mem-true-dep-cost
7967 Minimal distance (in CPU cycles) between store and load targeting same
7968 memory locations. The default value is 1.
7970 @item selsched-max-lookahead
7971 The maximum size of the lookahead window of selective scheduling. It is a
7972 depth of search for available instructions.
7973 The default value is 50.
7975 @item selsched-max-sched-times
7976 The maximum number of times that an instruction will be scheduled during
7977 selective scheduling. This is the limit on the number of iterations
7978 through which the instruction may be pipelined. The default value is 2.
7980 @item selsched-max-insns-to-rename
7981 The maximum number of best instructions in the ready list that are considered
7982 for renaming in the selective scheduler. The default value is 2.
7984 @item max-last-value-rtl
7985 The maximum size measured as number of RTLs that can be recorded in an expression
7986 in combiner for a pseudo register as last known value of that register. The default
7989 @item integer-share-limit
7990 Small integer constants can use a shared data structure, reducing the
7991 compiler's memory usage and increasing its speed. This sets the maximum
7992 value of a shared integer constant. The default value is 256.
7994 @item min-virtual-mappings
7995 Specifies the minimum number of virtual mappings in the incremental
7996 SSA updater that should be registered to trigger the virtual mappings
7997 heuristic defined by virtual-mappings-ratio. The default value is
8000 @item virtual-mappings-ratio
8001 If the number of virtual mappings is virtual-mappings-ratio bigger
8002 than the number of virtual symbols to be updated, then the incremental
8003 SSA updater switches to a full update for those symbols. The default
8006 @item ssp-buffer-size
8007 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8008 protection when @option{-fstack-protection} is used.
8010 @item max-jump-thread-duplication-stmts
8011 Maximum number of statements allowed in a block that needs to be
8012 duplicated when threading jumps.
8014 @item max-fields-for-field-sensitive
8015 Maximum number of fields in a structure we will treat in
8016 a field sensitive manner during pointer analysis. The default is zero
8017 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8019 @item prefetch-latency
8020 Estimate on average number of instructions that are executed before
8021 prefetch finishes. The distance we prefetch ahead is proportional
8022 to this constant. Increasing this number may also lead to less
8023 streams being prefetched (see @option{simultaneous-prefetches}).
8025 @item simultaneous-prefetches
8026 Maximum number of prefetches that can run at the same time.
8028 @item l1-cache-line-size
8029 The size of cache line in L1 cache, in bytes.
8032 The size of L1 cache, in kilobytes.
8035 The size of L2 cache, in kilobytes.
8037 @item min-insn-to-prefetch-ratio
8038 The minimum ratio between the number of instructions and the
8039 number of prefetches to enable prefetching in a loop with an
8042 @item prefetch-min-insn-to-mem-ratio
8043 The minimum ratio between the number of instructions and the
8044 number of memory references to enable prefetching in a loop.
8046 @item use-canonical-types
8047 Whether the compiler should use the ``canonical'' type system. By
8048 default, this should always be 1, which uses a more efficient internal
8049 mechanism for comparing types in C++ and Objective-C++. However, if
8050 bugs in the canonical type system are causing compilation failures,
8051 set this value to 0 to disable canonical types.
8053 @item switch-conversion-max-branch-ratio
8054 Switch initialization conversion will refuse to create arrays that are
8055 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8056 branches in the switch.
8058 @item max-partial-antic-length
8059 Maximum length of the partial antic set computed during the tree
8060 partial redundancy elimination optimization (@option{-ftree-pre}) when
8061 optimizing at @option{-O3} and above. For some sorts of source code
8062 the enhanced partial redundancy elimination optimization can run away,
8063 consuming all of the memory available on the host machine. This
8064 parameter sets a limit on the length of the sets that are computed,
8065 which prevents the runaway behavior. Setting a value of 0 for
8066 this parameter will allow an unlimited set length.
8068 @item sccvn-max-scc-size
8069 Maximum size of a strongly connected component (SCC) during SCCVN
8070 processing. If this limit is hit, SCCVN processing for the whole
8071 function will not be done and optimizations depending on it will
8072 be disabled. The default maximum SCC size is 10000.
8074 @item ira-max-loops-num
8075 IRA uses a regional register allocation by default. If a function
8076 contains loops more than number given by the parameter, only at most
8077 given number of the most frequently executed loops will form regions
8078 for the regional register allocation. The default value of the
8081 @item ira-max-conflict-table-size
8082 Although IRA uses a sophisticated algorithm of compression conflict
8083 table, the table can be still big for huge functions. If the conflict
8084 table for a function could be more than size in MB given by the
8085 parameter, the conflict table is not built and faster, simpler, and
8086 lower quality register allocation algorithm will be used. The
8087 algorithm do not use pseudo-register conflicts. The default value of
8088 the parameter is 2000.
8090 @item loop-invariant-max-bbs-in-loop
8091 Loop invariant motion can be very expensive, both in compile time and
8092 in amount of needed compile time memory, with very large loops. Loops
8093 with more basic blocks than this parameter won't have loop invariant
8094 motion optimization performed on them. The default value of the
8095 parameter is 1000 for -O1 and 10000 for -O2 and above.
8100 @node Preprocessor Options
8101 @section Options Controlling the Preprocessor
8102 @cindex preprocessor options
8103 @cindex options, preprocessor
8105 These options control the C preprocessor, which is run on each C source
8106 file before actual compilation.
8108 If you use the @option{-E} option, nothing is done except preprocessing.
8109 Some of these options make sense only together with @option{-E} because
8110 they cause the preprocessor output to be unsuitable for actual
8114 @item -Wp,@var{option}
8116 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8117 and pass @var{option} directly through to the preprocessor. If
8118 @var{option} contains commas, it is split into multiple options at the
8119 commas. However, many options are modified, translated or interpreted
8120 by the compiler driver before being passed to the preprocessor, and
8121 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8122 interface is undocumented and subject to change, so whenever possible
8123 you should avoid using @option{-Wp} and let the driver handle the
8126 @item -Xpreprocessor @var{option}
8127 @opindex Xpreprocessor
8128 Pass @var{option} as an option to the preprocessor. You can use this to
8129 supply system-specific preprocessor options which GCC does not know how to
8132 If you want to pass an option that takes an argument, you must use
8133 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8136 @include cppopts.texi
8138 @node Assembler Options
8139 @section Passing Options to the Assembler
8141 @c prevent bad page break with this line
8142 You can pass options to the assembler.
8145 @item -Wa,@var{option}
8147 Pass @var{option} as an option to the assembler. If @var{option}
8148 contains commas, it is split into multiple options at the commas.
8150 @item -Xassembler @var{option}
8152 Pass @var{option} as an option to the assembler. You can use this to
8153 supply system-specific assembler options which GCC does not know how to
8156 If you want to pass an option that takes an argument, you must use
8157 @option{-Xassembler} twice, once for the option and once for the argument.
8162 @section Options for Linking
8163 @cindex link options
8164 @cindex options, linking
8166 These options come into play when the compiler links object files into
8167 an executable output file. They are meaningless if the compiler is
8168 not doing a link step.
8172 @item @var{object-file-name}
8173 A file name that does not end in a special recognized suffix is
8174 considered to name an object file or library. (Object files are
8175 distinguished from libraries by the linker according to the file
8176 contents.) If linking is done, these object files are used as input
8185 If any of these options is used, then the linker is not run, and
8186 object file names should not be used as arguments. @xref{Overall
8190 @item -l@var{library}
8191 @itemx -l @var{library}
8193 Search the library named @var{library} when linking. (The second
8194 alternative with the library as a separate argument is only for
8195 POSIX compliance and is not recommended.)
8197 It makes a difference where in the command you write this option; the
8198 linker searches and processes libraries and object files in the order they
8199 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8200 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8201 to functions in @samp{z}, those functions may not be loaded.
8203 The linker searches a standard list of directories for the library,
8204 which is actually a file named @file{lib@var{library}.a}. The linker
8205 then uses this file as if it had been specified precisely by name.
8207 The directories searched include several standard system directories
8208 plus any that you specify with @option{-L}.
8210 Normally the files found this way are library files---archive files
8211 whose members are object files. The linker handles an archive file by
8212 scanning through it for members which define symbols that have so far
8213 been referenced but not defined. But if the file that is found is an
8214 ordinary object file, it is linked in the usual fashion. The only
8215 difference between using an @option{-l} option and specifying a file name
8216 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8217 and searches several directories.
8221 You need this special case of the @option{-l} option in order to
8222 link an Objective-C or Objective-C++ program.
8225 @opindex nostartfiles
8226 Do not use the standard system startup files when linking.
8227 The standard system libraries are used normally, unless @option{-nostdlib}
8228 or @option{-nodefaultlibs} is used.
8230 @item -nodefaultlibs
8231 @opindex nodefaultlibs
8232 Do not use the standard system libraries when linking.
8233 Only the libraries you specify will be passed to the linker, options
8234 specifying linkage of the system libraries, such as @code{-static-libgcc}
8235 or @code{-shared-libgcc}, will be ignored.
8236 The standard startup files are used normally, unless @option{-nostartfiles}
8237 is used. The compiler may generate calls to @code{memcmp},
8238 @code{memset}, @code{memcpy} and @code{memmove}.
8239 These entries are usually resolved by entries in
8240 libc. These entry points should be supplied through some other
8241 mechanism when this option is specified.
8245 Do not use the standard system startup files or libraries when linking.
8246 No startup files and only the libraries you specify will be passed to
8247 the linker, options specifying linkage of the system libraries, such as
8248 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8249 The compiler may generate calls to @code{memcmp}, @code{memset},
8250 @code{memcpy} and @code{memmove}.
8251 These entries are usually resolved by entries in
8252 libc. These entry points should be supplied through some other
8253 mechanism when this option is specified.
8255 @cindex @option{-lgcc}, use with @option{-nostdlib}
8256 @cindex @option{-nostdlib} and unresolved references
8257 @cindex unresolved references and @option{-nostdlib}
8258 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8259 @cindex @option{-nodefaultlibs} and unresolved references
8260 @cindex unresolved references and @option{-nodefaultlibs}
8261 One of the standard libraries bypassed by @option{-nostdlib} and
8262 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8263 that GCC uses to overcome shortcomings of particular machines, or special
8264 needs for some languages.
8265 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8266 Collection (GCC) Internals},
8267 for more discussion of @file{libgcc.a}.)
8268 In most cases, you need @file{libgcc.a} even when you want to avoid
8269 other standard libraries. In other words, when you specify @option{-nostdlib}
8270 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8271 This ensures that you have no unresolved references to internal GCC
8272 library subroutines. (For example, @samp{__main}, used to ensure C++
8273 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8274 GNU Compiler Collection (GCC) Internals}.)
8278 Produce a position independent executable on targets which support it.
8279 For predictable results, you must also specify the same set of options
8280 that were used to generate code (@option{-fpie}, @option{-fPIE},
8281 or model suboptions) when you specify this option.
8285 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8286 that support it. This instructs the linker to add all symbols, not
8287 only used ones, to the dynamic symbol table. This option is needed
8288 for some uses of @code{dlopen} or to allow obtaining backtraces
8289 from within a program.
8293 Remove all symbol table and relocation information from the executable.
8297 On systems that support dynamic linking, this prevents linking with the shared
8298 libraries. On other systems, this option has no effect.
8302 Produce a shared object which can then be linked with other objects to
8303 form an executable. Not all systems support this option. For predictable
8304 results, you must also specify the same set of options that were used to
8305 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8306 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8307 needs to build supplementary stub code for constructors to work. On
8308 multi-libbed systems, @samp{gcc -shared} must select the correct support
8309 libraries to link against. Failing to supply the correct flags may lead
8310 to subtle defects. Supplying them in cases where they are not necessary
8313 @item -shared-libgcc
8314 @itemx -static-libgcc
8315 @opindex shared-libgcc
8316 @opindex static-libgcc
8317 On systems that provide @file{libgcc} as a shared library, these options
8318 force the use of either the shared or static version respectively.
8319 If no shared version of @file{libgcc} was built when the compiler was
8320 configured, these options have no effect.
8322 There are several situations in which an application should use the
8323 shared @file{libgcc} instead of the static version. The most common
8324 of these is when the application wishes to throw and catch exceptions
8325 across different shared libraries. In that case, each of the libraries
8326 as well as the application itself should use the shared @file{libgcc}.
8328 Therefore, the G++ and GCJ drivers automatically add
8329 @option{-shared-libgcc} whenever you build a shared library or a main
8330 executable, because C++ and Java programs typically use exceptions, so
8331 this is the right thing to do.
8333 If, instead, you use the GCC driver to create shared libraries, you may
8334 find that they will not always be linked with the shared @file{libgcc}.
8335 If GCC finds, at its configuration time, that you have a non-GNU linker
8336 or a GNU linker that does not support option @option{--eh-frame-hdr},
8337 it will link the shared version of @file{libgcc} into shared libraries
8338 by default. Otherwise, it will take advantage of the linker and optimize
8339 away the linking with the shared version of @file{libgcc}, linking with
8340 the static version of libgcc by default. This allows exceptions to
8341 propagate through such shared libraries, without incurring relocation
8342 costs at library load time.
8344 However, if a library or main executable is supposed to throw or catch
8345 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8346 for the languages used in the program, or using the option
8347 @option{-shared-libgcc}, such that it is linked with the shared
8350 @item -static-libstdc++
8351 When the @command{g++} program is used to link a C++ program, it will
8352 normally automatically link against @option{libstdc++}. If
8353 @file{libstdc++} is available as a shared library, and the
8354 @option{-static} option is not used, then this will link against the
8355 shared version of @file{libstdc++}. That is normally fine. However, it
8356 is sometimes useful to freeze the version of @file{libstdc++} used by
8357 the program without going all the way to a fully static link. The
8358 @option{-static-libstdc++} option directs the @command{g++} driver to
8359 link @file{libstdc++} statically, without necessarily linking other
8360 libraries statically.
8364 Bind references to global symbols when building a shared object. Warn
8365 about any unresolved references (unless overridden by the link editor
8366 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8369 @item -T @var{script}
8371 @cindex linker script
8372 Use @var{script} as the linker script. This option is supported by most
8373 systems using the GNU linker. On some targets, such as bare-board
8374 targets without an operating system, the @option{-T} option may be required
8375 when linking to avoid references to undefined symbols.
8377 @item -Xlinker @var{option}
8379 Pass @var{option} as an option to the linker. You can use this to
8380 supply system-specific linker options which GCC does not know how to
8383 If you want to pass an option that takes a separate argument, you must use
8384 @option{-Xlinker} twice, once for the option and once for the argument.
8385 For example, to pass @option{-assert definitions}, you must write
8386 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8387 @option{-Xlinker "-assert definitions"}, because this passes the entire
8388 string as a single argument, which is not what the linker expects.
8390 When using the GNU linker, it is usually more convenient to pass
8391 arguments to linker options using the @option{@var{option}=@var{value}}
8392 syntax than as separate arguments. For example, you can specify
8393 @samp{-Xlinker -Map=output.map} rather than
8394 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8395 this syntax for command-line options.
8397 @item -Wl,@var{option}
8399 Pass @var{option} as an option to the linker. If @var{option} contains
8400 commas, it is split into multiple options at the commas. You can use this
8401 syntax to pass an argument to the option.
8402 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8403 linker. When using the GNU linker, you can also get the same effect with
8404 @samp{-Wl,-Map=output.map}.
8406 @item -u @var{symbol}
8408 Pretend the symbol @var{symbol} is undefined, to force linking of
8409 library modules to define it. You can use @option{-u} multiple times with
8410 different symbols to force loading of additional library modules.
8413 @node Directory Options
8414 @section Options for Directory Search
8415 @cindex directory options
8416 @cindex options, directory search
8419 These options specify directories to search for header files, for
8420 libraries and for parts of the compiler:
8425 Add the directory @var{dir} to the head of the list of directories to be
8426 searched for header files. This can be used to override a system header
8427 file, substituting your own version, since these directories are
8428 searched before the system header file directories. However, you should
8429 not use this option to add directories that contain vendor-supplied
8430 system header files (use @option{-isystem} for that). If you use more than
8431 one @option{-I} option, the directories are scanned in left-to-right
8432 order; the standard system directories come after.
8434 If a standard system include directory, or a directory specified with
8435 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8436 option will be ignored. The directory will still be searched but as a
8437 system directory at its normal position in the system include chain.
8438 This is to ensure that GCC's procedure to fix buggy system headers and
8439 the ordering for the include_next directive are not inadvertently changed.
8440 If you really need to change the search order for system directories,
8441 use the @option{-nostdinc} and/or @option{-isystem} options.
8443 @item -iquote@var{dir}
8445 Add the directory @var{dir} to the head of the list of directories to
8446 be searched for header files only for the case of @samp{#include
8447 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8448 otherwise just like @option{-I}.
8452 Add directory @var{dir} to the list of directories to be searched
8455 @item -B@var{prefix}
8457 This option specifies where to find the executables, libraries,
8458 include files, and data files of the compiler itself.
8460 The compiler driver program runs one or more of the subprograms
8461 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8462 @var{prefix} as a prefix for each program it tries to run, both with and
8463 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8465 For each subprogram to be run, the compiler driver first tries the
8466 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8467 was not specified, the driver tries two standard prefixes, which are
8468 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8469 those results in a file name that is found, the unmodified program
8470 name is searched for using the directories specified in your
8471 @env{PATH} environment variable.
8473 The compiler will check to see if the path provided by the @option{-B}
8474 refers to a directory, and if necessary it will add a directory
8475 separator character at the end of the path.
8477 @option{-B} prefixes that effectively specify directory names also apply
8478 to libraries in the linker, because the compiler translates these
8479 options into @option{-L} options for the linker. They also apply to
8480 includes files in the preprocessor, because the compiler translates these
8481 options into @option{-isystem} options for the preprocessor. In this case,
8482 the compiler appends @samp{include} to the prefix.
8484 The run-time support file @file{libgcc.a} can also be searched for using
8485 the @option{-B} prefix, if needed. If it is not found there, the two
8486 standard prefixes above are tried, and that is all. The file is left
8487 out of the link if it is not found by those means.
8489 Another way to specify a prefix much like the @option{-B} prefix is to use
8490 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8493 As a special kludge, if the path provided by @option{-B} is
8494 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8495 9, then it will be replaced by @file{[dir/]include}. This is to help
8496 with boot-strapping the compiler.
8498 @item -specs=@var{file}
8500 Process @var{file} after the compiler reads in the standard @file{specs}
8501 file, in order to override the defaults that the @file{gcc} driver
8502 program uses when determining what switches to pass to @file{cc1},
8503 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8504 @option{-specs=@var{file}} can be specified on the command line, and they
8505 are processed in order, from left to right.
8507 @item --sysroot=@var{dir}
8509 Use @var{dir} as the logical root directory for headers and libraries.
8510 For example, if the compiler would normally search for headers in
8511 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8512 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8514 If you use both this option and the @option{-isysroot} option, then
8515 the @option{--sysroot} option will apply to libraries, but the
8516 @option{-isysroot} option will apply to header files.
8518 The GNU linker (beginning with version 2.16) has the necessary support
8519 for this option. If your linker does not support this option, the
8520 header file aspect of @option{--sysroot} will still work, but the
8521 library aspect will not.
8525 This option has been deprecated. Please use @option{-iquote} instead for
8526 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8527 Any directories you specify with @option{-I} options before the @option{-I-}
8528 option are searched only for the case of @samp{#include "@var{file}"};
8529 they are not searched for @samp{#include <@var{file}>}.
8531 If additional directories are specified with @option{-I} options after
8532 the @option{-I-}, these directories are searched for all @samp{#include}
8533 directives. (Ordinarily @emph{all} @option{-I} directories are used
8536 In addition, the @option{-I-} option inhibits the use of the current
8537 directory (where the current input file came from) as the first search
8538 directory for @samp{#include "@var{file}"}. There is no way to
8539 override this effect of @option{-I-}. With @option{-I.} you can specify
8540 searching the directory which was current when the compiler was
8541 invoked. That is not exactly the same as what the preprocessor does
8542 by default, but it is often satisfactory.
8544 @option{-I-} does not inhibit the use of the standard system directories
8545 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8552 @section Specifying subprocesses and the switches to pass to them
8555 @command{gcc} is a driver program. It performs its job by invoking a
8556 sequence of other programs to do the work of compiling, assembling and
8557 linking. GCC interprets its command-line parameters and uses these to
8558 deduce which programs it should invoke, and which command-line options
8559 it ought to place on their command lines. This behavior is controlled
8560 by @dfn{spec strings}. In most cases there is one spec string for each
8561 program that GCC can invoke, but a few programs have multiple spec
8562 strings to control their behavior. The spec strings built into GCC can
8563 be overridden by using the @option{-specs=} command-line switch to specify
8566 @dfn{Spec files} are plaintext files that are used to construct spec
8567 strings. They consist of a sequence of directives separated by blank
8568 lines. The type of directive is determined by the first non-whitespace
8569 character on the line and it can be one of the following:
8572 @item %@var{command}
8573 Issues a @var{command} to the spec file processor. The commands that can
8577 @item %include <@var{file}>
8579 Search for @var{file} and insert its text at the current point in the
8582 @item %include_noerr <@var{file}>
8583 @cindex %include_noerr
8584 Just like @samp{%include}, but do not generate an error message if the include
8585 file cannot be found.
8587 @item %rename @var{old_name} @var{new_name}
8589 Rename the spec string @var{old_name} to @var{new_name}.
8593 @item *[@var{spec_name}]:
8594 This tells the compiler to create, override or delete the named spec
8595 string. All lines after this directive up to the next directive or
8596 blank line are considered to be the text for the spec string. If this
8597 results in an empty string then the spec will be deleted. (Or, if the
8598 spec did not exist, then nothing will happened.) Otherwise, if the spec
8599 does not currently exist a new spec will be created. If the spec does
8600 exist then its contents will be overridden by the text of this
8601 directive, unless the first character of that text is the @samp{+}
8602 character, in which case the text will be appended to the spec.
8604 @item [@var{suffix}]:
8605 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8606 and up to the next directive or blank line are considered to make up the
8607 spec string for the indicated suffix. When the compiler encounters an
8608 input file with the named suffix, it will processes the spec string in
8609 order to work out how to compile that file. For example:
8616 This says that any input file whose name ends in @samp{.ZZ} should be
8617 passed to the program @samp{z-compile}, which should be invoked with the
8618 command-line switch @option{-input} and with the result of performing the
8619 @samp{%i} substitution. (See below.)
8621 As an alternative to providing a spec string, the text that follows a
8622 suffix directive can be one of the following:
8625 @item @@@var{language}
8626 This says that the suffix is an alias for a known @var{language}. This is
8627 similar to using the @option{-x} command-line switch to GCC to specify a
8628 language explicitly. For example:
8635 Says that .ZZ files are, in fact, C++ source files.
8638 This causes an error messages saying:
8641 @var{name} compiler not installed on this system.
8645 GCC already has an extensive list of suffixes built into it.
8646 This directive will add an entry to the end of the list of suffixes, but
8647 since the list is searched from the end backwards, it is effectively
8648 possible to override earlier entries using this technique.
8652 GCC has the following spec strings built into it. Spec files can
8653 override these strings or create their own. Note that individual
8654 targets can also add their own spec strings to this list.
8657 asm Options to pass to the assembler
8658 asm_final Options to pass to the assembler post-processor
8659 cpp Options to pass to the C preprocessor
8660 cc1 Options to pass to the C compiler
8661 cc1plus Options to pass to the C++ compiler
8662 endfile Object files to include at the end of the link
8663 link Options to pass to the linker
8664 lib Libraries to include on the command line to the linker
8665 libgcc Decides which GCC support library to pass to the linker
8666 linker Sets the name of the linker
8667 predefines Defines to be passed to the C preprocessor
8668 signed_char Defines to pass to CPP to say whether @code{char} is signed
8670 startfile Object files to include at the start of the link
8673 Here is a small example of a spec file:
8679 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8682 This example renames the spec called @samp{lib} to @samp{old_lib} and
8683 then overrides the previous definition of @samp{lib} with a new one.
8684 The new definition adds in some extra command-line options before
8685 including the text of the old definition.
8687 @dfn{Spec strings} are a list of command-line options to be passed to their
8688 corresponding program. In addition, the spec strings can contain
8689 @samp{%}-prefixed sequences to substitute variable text or to
8690 conditionally insert text into the command line. Using these constructs
8691 it is possible to generate quite complex command lines.
8693 Here is a table of all defined @samp{%}-sequences for spec
8694 strings. Note that spaces are not generated automatically around the
8695 results of expanding these sequences. Therefore you can concatenate them
8696 together or combine them with constant text in a single argument.
8700 Substitute one @samp{%} into the program name or argument.
8703 Substitute the name of the input file being processed.
8706 Substitute the basename of the input file being processed.
8707 This is the substring up to (and not including) the last period
8708 and not including the directory.
8711 This is the same as @samp{%b}, but include the file suffix (text after
8715 Marks the argument containing or following the @samp{%d} as a
8716 temporary file name, so that that file will be deleted if GCC exits
8717 successfully. Unlike @samp{%g}, this contributes no text to the
8720 @item %g@var{suffix}
8721 Substitute a file name that has suffix @var{suffix} and is chosen
8722 once per compilation, and mark the argument in the same way as
8723 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8724 name is now chosen in a way that is hard to predict even when previously
8725 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8726 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8727 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8728 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8729 was simply substituted with a file name chosen once per compilation,
8730 without regard to any appended suffix (which was therefore treated
8731 just like ordinary text), making such attacks more likely to succeed.
8733 @item %u@var{suffix}
8734 Like @samp{%g}, but generates a new temporary file name even if
8735 @samp{%u@var{suffix}} was already seen.
8737 @item %U@var{suffix}
8738 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8739 new one if there is no such last file name. In the absence of any
8740 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8741 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8742 would involve the generation of two distinct file names, one
8743 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8744 simply substituted with a file name chosen for the previous @samp{%u},
8745 without regard to any appended suffix.
8747 @item %j@var{suffix}
8748 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8749 writable, and if save-temps is off; otherwise, substitute the name
8750 of a temporary file, just like @samp{%u}. This temporary file is not
8751 meant for communication between processes, but rather as a junk
8754 @item %|@var{suffix}
8755 @itemx %m@var{suffix}
8756 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8757 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8758 all. These are the two most common ways to instruct a program that it
8759 should read from standard input or write to standard output. If you
8760 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8761 construct: see for example @file{f/lang-specs.h}.
8763 @item %.@var{SUFFIX}
8764 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8765 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8766 terminated by the next space or %.
8769 Marks the argument containing or following the @samp{%w} as the
8770 designated output file of this compilation. This puts the argument
8771 into the sequence of arguments that @samp{%o} will substitute later.
8774 Substitutes the names of all the output files, with spaces
8775 automatically placed around them. You should write spaces
8776 around the @samp{%o} as well or the results are undefined.
8777 @samp{%o} is for use in the specs for running the linker.
8778 Input files whose names have no recognized suffix are not compiled
8779 at all, but they are included among the output files, so they will
8783 Substitutes the suffix for object files. Note that this is
8784 handled specially when it immediately follows @samp{%g, %u, or %U},
8785 because of the need for those to form complete file names. The
8786 handling is such that @samp{%O} is treated exactly as if it had already
8787 been substituted, except that @samp{%g, %u, and %U} do not currently
8788 support additional @var{suffix} characters following @samp{%O} as they would
8789 following, for example, @samp{.o}.
8792 Substitutes the standard macro predefinitions for the
8793 current target machine. Use this when running @code{cpp}.
8796 Like @samp{%p}, but puts @samp{__} before and after the name of each
8797 predefined macro, except for macros that start with @samp{__} or with
8798 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8802 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8803 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8804 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8805 and @option{-imultilib} as necessary.
8808 Current argument is the name of a library or startup file of some sort.
8809 Search for that file in a standard list of directories and substitute
8810 the full name found.
8813 Print @var{str} as an error message. @var{str} is terminated by a newline.
8814 Use this when inconsistent options are detected.
8817 Substitute the contents of spec string @var{name} at this point.
8820 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8822 @item %x@{@var{option}@}
8823 Accumulate an option for @samp{%X}.
8826 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8830 Output the accumulated assembler options specified by @option{-Wa}.
8833 Output the accumulated preprocessor options specified by @option{-Wp}.
8836 Process the @code{asm} spec. This is used to compute the
8837 switches to be passed to the assembler.
8840 Process the @code{asm_final} spec. This is a spec string for
8841 passing switches to an assembler post-processor, if such a program is
8845 Process the @code{link} spec. This is the spec for computing the
8846 command line passed to the linker. Typically it will make use of the
8847 @samp{%L %G %S %D and %E} sequences.
8850 Dump out a @option{-L} option for each directory that GCC believes might
8851 contain startup files. If the target supports multilibs then the
8852 current multilib directory will be prepended to each of these paths.
8855 Process the @code{lib} spec. This is a spec string for deciding which
8856 libraries should be included on the command line to the linker.
8859 Process the @code{libgcc} spec. This is a spec string for deciding
8860 which GCC support library should be included on the command line to the linker.
8863 Process the @code{startfile} spec. This is a spec for deciding which
8864 object files should be the first ones passed to the linker. Typically
8865 this might be a file named @file{crt0.o}.
8868 Process the @code{endfile} spec. This is a spec string that specifies
8869 the last object files that will be passed to the linker.
8872 Process the @code{cpp} spec. This is used to construct the arguments
8873 to be passed to the C preprocessor.
8876 Process the @code{cc1} spec. This is used to construct the options to be
8877 passed to the actual C compiler (@samp{cc1}).
8880 Process the @code{cc1plus} spec. This is used to construct the options to be
8881 passed to the actual C++ compiler (@samp{cc1plus}).
8884 Substitute the variable part of a matched option. See below.
8885 Note that each comma in the substituted string is replaced by
8889 Remove all occurrences of @code{-S} from the command line. Note---this
8890 command is position dependent. @samp{%} commands in the spec string
8891 before this one will see @code{-S}, @samp{%} commands in the spec string
8892 after this one will not.
8894 @item %:@var{function}(@var{args})
8895 Call the named function @var{function}, passing it @var{args}.
8896 @var{args} is first processed as a nested spec string, then split
8897 into an argument vector in the usual fashion. The function returns
8898 a string which is processed as if it had appeared literally as part
8899 of the current spec.
8901 The following built-in spec functions are provided:
8905 The @code{getenv} spec function takes two arguments: an environment
8906 variable name and a string. If the environment variable is not
8907 defined, a fatal error is issued. Otherwise, the return value is the
8908 value of the environment variable concatenated with the string. For
8909 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8912 %:getenv(TOPDIR /include)
8915 expands to @file{/path/to/top/include}.
8917 @item @code{if-exists}
8918 The @code{if-exists} spec function takes one argument, an absolute
8919 pathname to a file. If the file exists, @code{if-exists} returns the
8920 pathname. Here is a small example of its usage:
8924 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8927 @item @code{if-exists-else}
8928 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8929 spec function, except that it takes two arguments. The first argument is
8930 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8931 returns the pathname. If it does not exist, it returns the second argument.
8932 This way, @code{if-exists-else} can be used to select one file or another,
8933 based on the existence of the first. Here is a small example of its usage:
8937 crt0%O%s %:if-exists(crti%O%s) \
8938 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8941 @item @code{replace-outfile}
8942 The @code{replace-outfile} spec function takes two arguments. It looks for the
8943 first argument in the outfiles array and replaces it with the second argument. Here
8944 is a small example of its usage:
8947 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8950 @item @code{print-asm-header}
8951 The @code{print-asm-header} function takes no arguments and simply
8952 prints a banner like:
8958 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8961 It is used to separate compiler options from assembler options
8962 in the @option{--target-help} output.
8966 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8967 If that switch was not specified, this substitutes nothing. Note that
8968 the leading dash is omitted when specifying this option, and it is
8969 automatically inserted if the substitution is performed. Thus the spec
8970 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8971 and would output the command line option @option{-foo}.
8973 @item %W@{@code{S}@}
8974 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8977 @item %@{@code{S}*@}
8978 Substitutes all the switches specified to GCC whose names start
8979 with @code{-S}, but which also take an argument. This is used for
8980 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8981 GCC considers @option{-o foo} as being
8982 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8983 text, including the space. Thus two arguments would be generated.
8985 @item %@{@code{S}*&@code{T}*@}
8986 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8987 (the order of @code{S} and @code{T} in the spec is not significant).
8988 There can be any number of ampersand-separated variables; for each the
8989 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8991 @item %@{@code{S}:@code{X}@}
8992 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8994 @item %@{!@code{S}:@code{X}@}
8995 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8997 @item %@{@code{S}*:@code{X}@}
8998 Substitutes @code{X} if one or more switches whose names start with
8999 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9000 once, no matter how many such switches appeared. However, if @code{%*}
9001 appears somewhere in @code{X}, then @code{X} will be substituted once
9002 for each matching switch, with the @code{%*} replaced by the part of
9003 that switch that matched the @code{*}.
9005 @item %@{.@code{S}:@code{X}@}
9006 Substitutes @code{X}, if processing a file with suffix @code{S}.
9008 @item %@{!.@code{S}:@code{X}@}
9009 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9011 @item %@{,@code{S}:@code{X}@}
9012 Substitutes @code{X}, if processing a file for language @code{S}.
9014 @item %@{!,@code{S}:@code{X}@}
9015 Substitutes @code{X}, if not processing a file for language @code{S}.
9017 @item %@{@code{S}|@code{P}:@code{X}@}
9018 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9019 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9020 @code{*} sequences as well, although they have a stronger binding than
9021 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9022 alternatives must be starred, and only the first matching alternative
9025 For example, a spec string like this:
9028 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9031 will output the following command-line options from the following input
9032 command-line options:
9037 -d fred.c -foo -baz -boggle
9038 -d jim.d -bar -baz -boggle
9041 @item %@{S:X; T:Y; :D@}
9043 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9044 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9045 be as many clauses as you need. This may be combined with @code{.},
9046 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9051 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9052 construct may contain other nested @samp{%} constructs or spaces, or
9053 even newlines. They are processed as usual, as described above.
9054 Trailing white space in @code{X} is ignored. White space may also
9055 appear anywhere on the left side of the colon in these constructs,
9056 except between @code{.} or @code{*} and the corresponding word.
9058 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9059 handled specifically in these constructs. If another value of
9060 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9061 @option{-W} switch is found later in the command line, the earlier
9062 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9063 just one letter, which passes all matching options.
9065 The character @samp{|} at the beginning of the predicate text is used to
9066 indicate that a command should be piped to the following command, but
9067 only if @option{-pipe} is specified.
9069 It is built into GCC which switches take arguments and which do not.
9070 (You might think it would be useful to generalize this to allow each
9071 compiler's spec to say which switches take arguments. But this cannot
9072 be done in a consistent fashion. GCC cannot even decide which input
9073 files have been specified without knowing which switches take arguments,
9074 and it must know which input files to compile in order to tell which
9077 GCC also knows implicitly that arguments starting in @option{-l} are to be
9078 treated as compiler output files, and passed to the linker in their
9079 proper position among the other output files.
9081 @c man begin OPTIONS
9083 @node Target Options
9084 @section Specifying Target Machine and Compiler Version
9085 @cindex target options
9086 @cindex cross compiling
9087 @cindex specifying machine version
9088 @cindex specifying compiler version and target machine
9089 @cindex compiler version, specifying
9090 @cindex target machine, specifying
9092 The usual way to run GCC is to run the executable called @file{gcc}, or
9093 @file{<machine>-gcc} when cross-compiling, or
9094 @file{<machine>-gcc-<version>} to run a version other than the one that
9095 was installed last. Sometimes this is inconvenient, so GCC provides
9096 options that will switch to another cross-compiler or version.
9099 @item -b @var{machine}
9101 The argument @var{machine} specifies the target machine for compilation.
9103 The value to use for @var{machine} is the same as was specified as the
9104 machine type when configuring GCC as a cross-compiler. For
9105 example, if a cross-compiler was configured with @samp{configure
9106 arm-elf}, meaning to compile for an arm processor with elf binaries,
9107 then you would specify @option{-b arm-elf} to run that cross compiler.
9108 Because there are other options beginning with @option{-b}, the
9109 configuration must contain a hyphen, or @option{-b} alone should be one
9110 argument followed by the configuration in the next argument.
9112 @item -V @var{version}
9114 The argument @var{version} specifies which version of GCC to run.
9115 This is useful when multiple versions are installed. For example,
9116 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9119 The @option{-V} and @option{-b} options work by running the
9120 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9121 use them if you can just run that directly.
9123 @node Submodel Options
9124 @section Hardware Models and Configurations
9125 @cindex submodel options
9126 @cindex specifying hardware config
9127 @cindex hardware models and configurations, specifying
9128 @cindex machine dependent options
9130 Earlier we discussed the standard option @option{-b} which chooses among
9131 different installed compilers for completely different target
9132 machines, such as VAX vs.@: 68000 vs.@: 80386.
9134 In addition, each of these target machine types can have its own
9135 special options, starting with @samp{-m}, to choose among various
9136 hardware models or configurations---for example, 68010 vs 68020,
9137 floating coprocessor or none. A single installed version of the
9138 compiler can compile for any model or configuration, according to the
9141 Some configurations of the compiler also support additional special
9142 options, usually for compatibility with other compilers on the same
9145 @c This list is ordered alphanumerically by subsection name.
9146 @c It should be the same order and spelling as these options are listed
9147 @c in Machine Dependent Options
9153 * Blackfin Options::
9157 * DEC Alpha Options::
9158 * DEC Alpha/VMS Options::
9161 * GNU/Linux Options::
9164 * i386 and x86-64 Options::
9165 * i386 and x86-64 Windows Options::
9167 * IA-64/VMS Options::
9178 * picoChip Options::
9180 * RS/6000 and PowerPC Options::
9181 * S/390 and zSeries Options::
9186 * System V Options::
9191 * Xstormy16 Options::
9197 @subsection ARC Options
9200 These options are defined for ARC implementations:
9205 Compile code for little endian mode. This is the default.
9209 Compile code for big endian mode.
9212 @opindex mmangle-cpu
9213 Prepend the name of the cpu to all public symbol names.
9214 In multiple-processor systems, there are many ARC variants with different
9215 instruction and register set characteristics. This flag prevents code
9216 compiled for one cpu to be linked with code compiled for another.
9217 No facility exists for handling variants that are ``almost identical''.
9218 This is an all or nothing option.
9220 @item -mcpu=@var{cpu}
9222 Compile code for ARC variant @var{cpu}.
9223 Which variants are supported depend on the configuration.
9224 All variants support @option{-mcpu=base}, this is the default.
9226 @item -mtext=@var{text-section}
9227 @itemx -mdata=@var{data-section}
9228 @itemx -mrodata=@var{readonly-data-section}
9232 Put functions, data, and readonly data in @var{text-section},
9233 @var{data-section}, and @var{readonly-data-section} respectively
9234 by default. This can be overridden with the @code{section} attribute.
9235 @xref{Variable Attributes}.
9237 @item -mfix-cortex-m3-ldrd
9238 @opindex mfix-cortex-m3-ldrd
9239 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9240 with overlapping destination and base registers are used. This option avoids
9241 generating these instructions. This option is enabled by default when
9242 @option{-mcpu=cortex-m3} is specified.
9247 @subsection ARM Options
9250 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9254 @item -mabi=@var{name}
9256 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9257 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9260 @opindex mapcs-frame
9261 Generate a stack frame that is compliant with the ARM Procedure Call
9262 Standard for all functions, even if this is not strictly necessary for
9263 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9264 with this option will cause the stack frames not to be generated for
9265 leaf functions. The default is @option{-mno-apcs-frame}.
9269 This is a synonym for @option{-mapcs-frame}.
9272 @c not currently implemented
9273 @item -mapcs-stack-check
9274 @opindex mapcs-stack-check
9275 Generate code to check the amount of stack space available upon entry to
9276 every function (that actually uses some stack space). If there is
9277 insufficient space available then either the function
9278 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9279 called, depending upon the amount of stack space required. The run time
9280 system is required to provide these functions. The default is
9281 @option{-mno-apcs-stack-check}, since this produces smaller code.
9283 @c not currently implemented
9285 @opindex mapcs-float
9286 Pass floating point arguments using the float point registers. This is
9287 one of the variants of the APCS@. This option is recommended if the
9288 target hardware has a floating point unit or if a lot of floating point
9289 arithmetic is going to be performed by the code. The default is
9290 @option{-mno-apcs-float}, since integer only code is slightly increased in
9291 size if @option{-mapcs-float} is used.
9293 @c not currently implemented
9294 @item -mapcs-reentrant
9295 @opindex mapcs-reentrant
9296 Generate reentrant, position independent code. The default is
9297 @option{-mno-apcs-reentrant}.
9300 @item -mthumb-interwork
9301 @opindex mthumb-interwork
9302 Generate code which supports calling between the ARM and Thumb
9303 instruction sets. Without this option the two instruction sets cannot
9304 be reliably used inside one program. The default is
9305 @option{-mno-thumb-interwork}, since slightly larger code is generated
9306 when @option{-mthumb-interwork} is specified.
9308 @item -mno-sched-prolog
9309 @opindex mno-sched-prolog
9310 Prevent the reordering of instructions in the function prolog, or the
9311 merging of those instruction with the instructions in the function's
9312 body. This means that all functions will start with a recognizable set
9313 of instructions (or in fact one of a choice from a small set of
9314 different function prologues), and this information can be used to
9315 locate the start if functions inside an executable piece of code. The
9316 default is @option{-msched-prolog}.
9318 @item -mfloat-abi=@var{name}
9320 Specifies which floating-point ABI to use. Permissible values
9321 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9323 Specifying @samp{soft} causes GCC to generate output containing
9324 library calls for floating-point operations.
9325 @samp{softfp} allows the generation of code using hardware floating-point
9326 instructions, but still uses the soft-float calling conventions.
9327 @samp{hard} allows generation of floating-point instructions
9328 and uses FPU-specific calling conventions.
9330 The default depends on the specific target configuration. Note that
9331 the hard-float and soft-float ABIs are not link-compatible; you must
9332 compile your entire program with the same ABI, and link with a
9333 compatible set of libraries.
9336 @opindex mhard-float
9337 Equivalent to @option{-mfloat-abi=hard}.
9340 @opindex msoft-float
9341 Equivalent to @option{-mfloat-abi=soft}.
9343 @item -mlittle-endian
9344 @opindex mlittle-endian
9345 Generate code for a processor running in little-endian mode. This is
9346 the default for all standard configurations.
9349 @opindex mbig-endian
9350 Generate code for a processor running in big-endian mode; the default is
9351 to compile code for a little-endian processor.
9353 @item -mwords-little-endian
9354 @opindex mwords-little-endian
9355 This option only applies when generating code for big-endian processors.
9356 Generate code for a little-endian word order but a big-endian byte
9357 order. That is, a byte order of the form @samp{32107654}. Note: this
9358 option should only be used if you require compatibility with code for
9359 big-endian ARM processors generated by versions of the compiler prior to
9362 @item -mcpu=@var{name}
9364 This specifies the name of the target ARM processor. GCC uses this name
9365 to determine what kind of instructions it can emit when generating
9366 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9367 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9368 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9369 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9370 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9372 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9373 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9374 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9375 @samp{strongarm1110},
9376 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9377 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9378 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9379 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9380 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9381 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9382 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9383 @samp{cortex-a8}, @samp{cortex-a9},
9384 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9387 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9389 @item -mtune=@var{name}
9391 This option is very similar to the @option{-mcpu=} option, except that
9392 instead of specifying the actual target processor type, and hence
9393 restricting which instructions can be used, it specifies that GCC should
9394 tune the performance of the code as if the target were of the type
9395 specified in this option, but still choosing the instructions that it
9396 will generate based on the cpu specified by a @option{-mcpu=} option.
9397 For some ARM implementations better performance can be obtained by using
9400 @item -march=@var{name}
9402 This specifies the name of the target ARM architecture. GCC uses this
9403 name to determine what kind of instructions it can emit when generating
9404 assembly code. This option can be used in conjunction with or instead
9405 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9406 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9407 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9408 @samp{armv6}, @samp{armv6j},
9409 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9410 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9411 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9413 @item -mfpu=@var{name}
9414 @itemx -mfpe=@var{number}
9415 @itemx -mfp=@var{number}
9419 This specifies what floating point hardware (or hardware emulation) is
9420 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9421 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9422 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9423 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9424 with older versions of GCC@.
9426 If @option{-msoft-float} is specified this specifies the format of
9427 floating point values.
9429 @item -mfp16-format=@var{name}
9430 @opindex mfp16-format
9431 Specify the format of the @code{__fp16} half-precision floating-point type.
9432 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9433 the default is @samp{none}, in which case the @code{__fp16} type is not
9434 defined. @xref{Half-Precision}, for more information.
9436 @item -mstructure-size-boundary=@var{n}
9437 @opindex mstructure-size-boundary
9438 The size of all structures and unions will be rounded up to a multiple
9439 of the number of bits set by this option. Permissible values are 8, 32
9440 and 64. The default value varies for different toolchains. For the COFF
9441 targeted toolchain the default value is 8. A value of 64 is only allowed
9442 if the underlying ABI supports it.
9444 Specifying the larger number can produce faster, more efficient code, but
9445 can also increase the size of the program. Different values are potentially
9446 incompatible. Code compiled with one value cannot necessarily expect to
9447 work with code or libraries compiled with another value, if they exchange
9448 information using structures or unions.
9450 @item -mabort-on-noreturn
9451 @opindex mabort-on-noreturn
9452 Generate a call to the function @code{abort} at the end of a
9453 @code{noreturn} function. It will be executed if the function tries to
9457 @itemx -mno-long-calls
9458 @opindex mlong-calls
9459 @opindex mno-long-calls
9460 Tells the compiler to perform function calls by first loading the
9461 address of the function into a register and then performing a subroutine
9462 call on this register. This switch is needed if the target function
9463 will lie outside of the 64 megabyte addressing range of the offset based
9464 version of subroutine call instruction.
9466 Even if this switch is enabled, not all function calls will be turned
9467 into long calls. The heuristic is that static functions, functions
9468 which have the @samp{short-call} attribute, functions that are inside
9469 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9470 definitions have already been compiled within the current compilation
9471 unit, will not be turned into long calls. The exception to this rule is
9472 that weak function definitions, functions with the @samp{long-call}
9473 attribute or the @samp{section} attribute, and functions that are within
9474 the scope of a @samp{#pragma long_calls} directive, will always be
9475 turned into long calls.
9477 This feature is not enabled by default. Specifying
9478 @option{-mno-long-calls} will restore the default behavior, as will
9479 placing the function calls within the scope of a @samp{#pragma
9480 long_calls_off} directive. Note these switches have no effect on how
9481 the compiler generates code to handle function calls via function
9484 @item -msingle-pic-base
9485 @opindex msingle-pic-base
9486 Treat the register used for PIC addressing as read-only, rather than
9487 loading it in the prologue for each function. The run-time system is
9488 responsible for initializing this register with an appropriate value
9489 before execution begins.
9491 @item -mpic-register=@var{reg}
9492 @opindex mpic-register
9493 Specify the register to be used for PIC addressing. The default is R10
9494 unless stack-checking is enabled, when R9 is used.
9496 @item -mcirrus-fix-invalid-insns
9497 @opindex mcirrus-fix-invalid-insns
9498 @opindex mno-cirrus-fix-invalid-insns
9499 Insert NOPs into the instruction stream to in order to work around
9500 problems with invalid Maverick instruction combinations. This option
9501 is only valid if the @option{-mcpu=ep9312} option has been used to
9502 enable generation of instructions for the Cirrus Maverick floating
9503 point co-processor. This option is not enabled by default, since the
9504 problem is only present in older Maverick implementations. The default
9505 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9508 @item -mpoke-function-name
9509 @opindex mpoke-function-name
9510 Write the name of each function into the text section, directly
9511 preceding the function prologue. The generated code is similar to this:
9515 .ascii "arm_poke_function_name", 0
9518 .word 0xff000000 + (t1 - t0)
9519 arm_poke_function_name
9521 stmfd sp!, @{fp, ip, lr, pc@}
9525 When performing a stack backtrace, code can inspect the value of
9526 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9527 location @code{pc - 12} and the top 8 bits are set, then we know that
9528 there is a function name embedded immediately preceding this location
9529 and has length @code{((pc[-3]) & 0xff000000)}.
9533 Generate code for the Thumb instruction set. The default is to
9534 use the 32-bit ARM instruction set.
9535 This option automatically enables either 16-bit Thumb-1 or
9536 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9537 and @option{-march=@var{name}} options. This option is not passed to the
9538 assembler. If you want to force assembler files to be interpreted as Thumb code,
9539 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9540 option directly to the assembler by prefixing it with @option{-Wa}.
9543 @opindex mtpcs-frame
9544 Generate a stack frame that is compliant with the Thumb Procedure Call
9545 Standard for all non-leaf functions. (A leaf function is one that does
9546 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9548 @item -mtpcs-leaf-frame
9549 @opindex mtpcs-leaf-frame
9550 Generate a stack frame that is compliant with the Thumb Procedure Call
9551 Standard for all leaf functions. (A leaf function is one that does
9552 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9554 @item -mcallee-super-interworking
9555 @opindex mcallee-super-interworking
9556 Gives all externally visible functions in the file being compiled an ARM
9557 instruction set header which switches to Thumb mode before executing the
9558 rest of the function. This allows these functions to be called from
9559 non-interworking code. This option is not valid in AAPCS configurations
9560 because interworking is enabled by default.
9562 @item -mcaller-super-interworking
9563 @opindex mcaller-super-interworking
9564 Allows calls via function pointers (including virtual functions) to
9565 execute correctly regardless of whether the target code has been
9566 compiled for interworking or not. There is a small overhead in the cost
9567 of executing a function pointer if this option is enabled. This option
9568 is not valid in AAPCS configurations because interworking is enabled
9571 @item -mtp=@var{name}
9573 Specify the access model for the thread local storage pointer. The valid
9574 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9575 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9576 (supported in the arm6k architecture), and @option{auto}, which uses the
9577 best available method for the selected processor. The default setting is
9580 @item -mword-relocations
9581 @opindex mword-relocations
9582 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9583 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9584 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9590 @subsection AVR Options
9593 These options are defined for AVR implementations:
9596 @item -mmcu=@var{mcu}
9598 Specify ATMEL AVR instruction set or MCU type.
9600 Instruction set avr1 is for the minimal AVR core, not supported by the C
9601 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9602 attiny11, attiny12, attiny15, attiny28).
9604 Instruction set avr2 (default) is for the classic AVR core with up to
9605 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9606 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9607 at90c8534, at90s8535).
9609 Instruction set avr3 is for the classic AVR core with up to 128K program
9610 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9612 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9613 memory space (MCU types: atmega8, atmega83, atmega85).
9615 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9616 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9617 atmega64, atmega128, at43usb355, at94k).
9621 Output instruction sizes to the asm file.
9623 @item -minit-stack=@var{N}
9624 @opindex minit-stack
9625 Specify the initial stack address, which may be a symbol or numeric value,
9626 @samp{__stack} is the default.
9628 @item -mno-interrupts
9629 @opindex mno-interrupts
9630 Generated code is not compatible with hardware interrupts.
9631 Code size will be smaller.
9633 @item -mcall-prologues
9634 @opindex mcall-prologues
9635 Functions prologues/epilogues expanded as call to appropriate
9636 subroutines. Code size will be smaller.
9639 @opindex mtiny-stack
9640 Change only the low 8 bits of the stack pointer.
9644 Assume int to be 8 bit integer. This affects the sizes of all types: A
9645 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9646 and long long will be 4 bytes. Please note that this option does not
9647 comply to the C standards, but it will provide you with smaller code
9651 @node Blackfin Options
9652 @subsection Blackfin Options
9653 @cindex Blackfin Options
9656 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9658 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9659 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9660 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9661 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9662 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9663 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9665 The optional @var{sirevision} specifies the silicon revision of the target
9666 Blackfin processor. Any workarounds available for the targeted silicon revision
9667 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9668 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9669 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9670 hexadecimal digits representing the major and minor numbers in the silicon
9671 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9672 is not defined. If @var{sirevision} is @samp{any}, the
9673 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9674 If this optional @var{sirevision} is not used, GCC assumes the latest known
9675 silicon revision of the targeted Blackfin processor.
9677 Support for @samp{bf561} is incomplete. For @samp{bf561},
9678 Only the processor macro is defined.
9679 Without this option, @samp{bf532} is used as the processor by default.
9680 The corresponding predefined processor macros for @var{cpu} is to
9681 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9682 provided by libgloss to be linked in if @option{-msim} is not given.
9686 Specifies that the program will be run on the simulator. This causes
9687 the simulator BSP provided by libgloss to be linked in. This option
9688 has effect only for @samp{bfin-elf} toolchain.
9689 Certain other options, such as @option{-mid-shared-library} and
9690 @option{-mfdpic}, imply @option{-msim}.
9692 @item -momit-leaf-frame-pointer
9693 @opindex momit-leaf-frame-pointer
9694 Don't keep the frame pointer in a register for leaf functions. This
9695 avoids the instructions to save, set up and restore frame pointers and
9696 makes an extra register available in leaf functions. The option
9697 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9698 which might make debugging harder.
9700 @item -mspecld-anomaly
9701 @opindex mspecld-anomaly
9702 When enabled, the compiler will ensure that the generated code does not
9703 contain speculative loads after jump instructions. If this option is used,
9704 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9706 @item -mno-specld-anomaly
9707 @opindex mno-specld-anomaly
9708 Don't generate extra code to prevent speculative loads from occurring.
9710 @item -mcsync-anomaly
9711 @opindex mcsync-anomaly
9712 When enabled, the compiler will ensure that the generated code does not
9713 contain CSYNC or SSYNC instructions too soon after conditional branches.
9714 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9716 @item -mno-csync-anomaly
9717 @opindex mno-csync-anomaly
9718 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9719 occurring too soon after a conditional branch.
9723 When enabled, the compiler is free to take advantage of the knowledge that
9724 the entire program fits into the low 64k of memory.
9727 @opindex mno-low-64k
9728 Assume that the program is arbitrarily large. This is the default.
9730 @item -mstack-check-l1
9731 @opindex mstack-check-l1
9732 Do stack checking using information placed into L1 scratchpad memory by the
9735 @item -mid-shared-library
9736 @opindex mid-shared-library
9737 Generate code that supports shared libraries via the library ID method.
9738 This allows for execute in place and shared libraries in an environment
9739 without virtual memory management. This option implies @option{-fPIC}.
9740 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9742 @item -mno-id-shared-library
9743 @opindex mno-id-shared-library
9744 Generate code that doesn't assume ID based shared libraries are being used.
9745 This is the default.
9747 @item -mleaf-id-shared-library
9748 @opindex mleaf-id-shared-library
9749 Generate code that supports shared libraries via the library ID method,
9750 but assumes that this library or executable won't link against any other
9751 ID shared libraries. That allows the compiler to use faster code for jumps
9754 @item -mno-leaf-id-shared-library
9755 @opindex mno-leaf-id-shared-library
9756 Do not assume that the code being compiled won't link against any ID shared
9757 libraries. Slower code will be generated for jump and call insns.
9759 @item -mshared-library-id=n
9760 @opindex mshared-library-id
9761 Specified the identification number of the ID based shared library being
9762 compiled. Specifying a value of 0 will generate more compact code, specifying
9763 other values will force the allocation of that number to the current
9764 library but is no more space or time efficient than omitting this option.
9768 Generate code that allows the data segment to be located in a different
9769 area of memory from the text segment. This allows for execute in place in
9770 an environment without virtual memory management by eliminating relocations
9771 against the text section.
9774 @opindex mno-sep-data
9775 Generate code that assumes that the data segment follows the text segment.
9776 This is the default.
9779 @itemx -mno-long-calls
9780 @opindex mlong-calls
9781 @opindex mno-long-calls
9782 Tells the compiler to perform function calls by first loading the
9783 address of the function into a register and then performing a subroutine
9784 call on this register. This switch is needed if the target function
9785 will lie outside of the 24 bit addressing range of the offset based
9786 version of subroutine call instruction.
9788 This feature is not enabled by default. Specifying
9789 @option{-mno-long-calls} will restore the default behavior. Note these
9790 switches have no effect on how the compiler generates code to handle
9791 function calls via function pointers.
9795 Link with the fast floating-point library. This library relaxes some of
9796 the IEEE floating-point standard's rules for checking inputs against
9797 Not-a-Number (NAN), in the interest of performance.
9800 @opindex minline-plt
9801 Enable inlining of PLT entries in function calls to functions that are
9802 not known to bind locally. It has no effect without @option{-mfdpic}.
9806 Build standalone application for multicore Blackfin processor. Proper
9807 start files and link scripts will be used to support multicore.
9808 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9809 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9810 @option{-mcorea} or @option{-mcoreb}. If it's used without
9811 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9812 programming model is used. In this model, the main function of Core B
9813 should be named as coreb_main. If it's used with @option{-mcorea} or
9814 @option{-mcoreb}, one application per core programming model is used.
9815 If this option is not used, single core application programming
9820 Build standalone application for Core A of BF561 when using
9821 one application per core programming model. Proper start files
9822 and link scripts will be used to support Core A. This option
9823 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9827 Build standalone application for Core B of BF561 when using
9828 one application per core programming model. Proper start files
9829 and link scripts will be used to support Core B. This option
9830 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9831 should be used instead of main. It must be used with
9832 @option{-mmulticore}.
9836 Build standalone application for SDRAM. Proper start files and
9837 link scripts will be used to put the application into SDRAM.
9838 Loader should initialize SDRAM before loading the application
9839 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9843 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9844 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9845 are enabled; for standalone applications the default is off.
9849 @subsection CRIS Options
9850 @cindex CRIS Options
9852 These options are defined specifically for the CRIS ports.
9855 @item -march=@var{architecture-type}
9856 @itemx -mcpu=@var{architecture-type}
9859 Generate code for the specified architecture. The choices for
9860 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9861 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9862 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9865 @item -mtune=@var{architecture-type}
9867 Tune to @var{architecture-type} everything applicable about the generated
9868 code, except for the ABI and the set of available instructions. The
9869 choices for @var{architecture-type} are the same as for
9870 @option{-march=@var{architecture-type}}.
9872 @item -mmax-stack-frame=@var{n}
9873 @opindex mmax-stack-frame
9874 Warn when the stack frame of a function exceeds @var{n} bytes.
9880 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9881 @option{-march=v3} and @option{-march=v8} respectively.
9883 @item -mmul-bug-workaround
9884 @itemx -mno-mul-bug-workaround
9885 @opindex mmul-bug-workaround
9886 @opindex mno-mul-bug-workaround
9887 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9888 models where it applies. This option is active by default.
9892 Enable CRIS-specific verbose debug-related information in the assembly
9893 code. This option also has the effect to turn off the @samp{#NO_APP}
9894 formatted-code indicator to the assembler at the beginning of the
9899 Do not use condition-code results from previous instruction; always emit
9900 compare and test instructions before use of condition codes.
9902 @item -mno-side-effects
9903 @opindex mno-side-effects
9904 Do not emit instructions with side-effects in addressing modes other than
9908 @itemx -mno-stack-align
9910 @itemx -mno-data-align
9911 @itemx -mconst-align
9912 @itemx -mno-const-align
9913 @opindex mstack-align
9914 @opindex mno-stack-align
9915 @opindex mdata-align
9916 @opindex mno-data-align
9917 @opindex mconst-align
9918 @opindex mno-const-align
9919 These options (no-options) arranges (eliminate arrangements) for the
9920 stack-frame, individual data and constants to be aligned for the maximum
9921 single data access size for the chosen CPU model. The default is to
9922 arrange for 32-bit alignment. ABI details such as structure layout are
9923 not affected by these options.
9931 Similar to the stack- data- and const-align options above, these options
9932 arrange for stack-frame, writable data and constants to all be 32-bit,
9933 16-bit or 8-bit aligned. The default is 32-bit alignment.
9935 @item -mno-prologue-epilogue
9936 @itemx -mprologue-epilogue
9937 @opindex mno-prologue-epilogue
9938 @opindex mprologue-epilogue
9939 With @option{-mno-prologue-epilogue}, the normal function prologue and
9940 epilogue that sets up the stack-frame are omitted and no return
9941 instructions or return sequences are generated in the code. Use this
9942 option only together with visual inspection of the compiled code: no
9943 warnings or errors are generated when call-saved registers must be saved,
9944 or storage for local variable needs to be allocated.
9950 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9951 instruction sequences that load addresses for functions from the PLT part
9952 of the GOT rather than (traditional on other architectures) calls to the
9953 PLT@. The default is @option{-mgotplt}.
9957 Legacy no-op option only recognized with the cris-axis-elf and
9958 cris-axis-linux-gnu targets.
9962 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9966 This option, recognized for the cris-axis-elf arranges
9967 to link with input-output functions from a simulator library. Code,
9968 initialized data and zero-initialized data are allocated consecutively.
9972 Like @option{-sim}, but pass linker options to locate initialized data at
9973 0x40000000 and zero-initialized data at 0x80000000.
9977 @subsection CRX Options
9980 These options are defined specifically for the CRX ports.
9986 Enable the use of multiply-accumulate instructions. Disabled by default.
9990 Push instructions will be used to pass outgoing arguments when functions
9991 are called. Enabled by default.
9994 @node Darwin Options
9995 @subsection Darwin Options
9996 @cindex Darwin options
9998 These options are defined for all architectures running the Darwin operating
10001 FSF GCC on Darwin does not create ``fat'' object files; it will create
10002 an object file for the single architecture that it was built to
10003 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10004 @option{-arch} options are used; it does so by running the compiler or
10005 linker multiple times and joining the results together with
10008 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10009 @samp{i686}) is determined by the flags that specify the ISA
10010 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10011 @option{-force_cpusubtype_ALL} option can be used to override this.
10013 The Darwin tools vary in their behavior when presented with an ISA
10014 mismatch. The assembler, @file{as}, will only permit instructions to
10015 be used that are valid for the subtype of the file it is generating,
10016 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10017 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10018 and print an error if asked to create a shared library with a less
10019 restrictive subtype than its input files (for instance, trying to put
10020 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10021 for executables, @file{ld}, will quietly give the executable the most
10022 restrictive subtype of any of its input files.
10027 Add the framework directory @var{dir} to the head of the list of
10028 directories to be searched for header files. These directories are
10029 interleaved with those specified by @option{-I} options and are
10030 scanned in a left-to-right order.
10032 A framework directory is a directory with frameworks in it. A
10033 framework is a directory with a @samp{"Headers"} and/or
10034 @samp{"PrivateHeaders"} directory contained directly in it that ends
10035 in @samp{".framework"}. The name of a framework is the name of this
10036 directory excluding the @samp{".framework"}. Headers associated with
10037 the framework are found in one of those two directories, with
10038 @samp{"Headers"} being searched first. A subframework is a framework
10039 directory that is in a framework's @samp{"Frameworks"} directory.
10040 Includes of subframework headers can only appear in a header of a
10041 framework that contains the subframework, or in a sibling subframework
10042 header. Two subframeworks are siblings if they occur in the same
10043 framework. A subframework should not have the same name as a
10044 framework, a warning will be issued if this is violated. Currently a
10045 subframework cannot have subframeworks, in the future, the mechanism
10046 may be extended to support this. The standard frameworks can be found
10047 in @samp{"/System/Library/Frameworks"} and
10048 @samp{"/Library/Frameworks"}. An example include looks like
10049 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10050 the name of the framework and header.h is found in the
10051 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10053 @item -iframework@var{dir}
10054 @opindex iframework
10055 Like @option{-F} except the directory is a treated as a system
10056 directory. The main difference between this @option{-iframework} and
10057 @option{-F} is that with @option{-iframework} the compiler does not
10058 warn about constructs contained within header files found via
10059 @var{dir}. This option is valid only for the C family of languages.
10063 Emit debugging information for symbols that are used. For STABS
10064 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10065 This is by default ON@.
10069 Emit debugging information for all symbols and types.
10071 @item -mmacosx-version-min=@var{version}
10072 The earliest version of MacOS X that this executable will run on
10073 is @var{version}. Typical values of @var{version} include @code{10.1},
10074 @code{10.2}, and @code{10.3.9}.
10076 If the compiler was built to use the system's headers by default,
10077 then the default for this option is the system version on which the
10078 compiler is running, otherwise the default is to make choices which
10079 are compatible with as many systems and code bases as possible.
10083 Enable kernel development mode. The @option{-mkernel} option sets
10084 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10085 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10086 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10087 applicable. This mode also sets @option{-mno-altivec},
10088 @option{-msoft-float}, @option{-fno-builtin} and
10089 @option{-mlong-branch} for PowerPC targets.
10091 @item -mone-byte-bool
10092 @opindex mone-byte-bool
10093 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10094 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10095 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10096 option has no effect on x86.
10098 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10099 to generate code that is not binary compatible with code generated
10100 without that switch. Using this switch may require recompiling all
10101 other modules in a program, including system libraries. Use this
10102 switch to conform to a non-default data model.
10104 @item -mfix-and-continue
10105 @itemx -ffix-and-continue
10106 @itemx -findirect-data
10107 @opindex mfix-and-continue
10108 @opindex ffix-and-continue
10109 @opindex findirect-data
10110 Generate code suitable for fast turn around development. Needed to
10111 enable gdb to dynamically load @code{.o} files into already running
10112 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10113 are provided for backwards compatibility.
10117 Loads all members of static archive libraries.
10118 See man ld(1) for more information.
10120 @item -arch_errors_fatal
10121 @opindex arch_errors_fatal
10122 Cause the errors having to do with files that have the wrong architecture
10125 @item -bind_at_load
10126 @opindex bind_at_load
10127 Causes the output file to be marked such that the dynamic linker will
10128 bind all undefined references when the file is loaded or launched.
10132 Produce a Mach-o bundle format file.
10133 See man ld(1) for more information.
10135 @item -bundle_loader @var{executable}
10136 @opindex bundle_loader
10137 This option specifies the @var{executable} that will be loading the build
10138 output file being linked. See man ld(1) for more information.
10141 @opindex dynamiclib
10142 When passed this option, GCC will produce a dynamic library instead of
10143 an executable when linking, using the Darwin @file{libtool} command.
10145 @item -force_cpusubtype_ALL
10146 @opindex force_cpusubtype_ALL
10147 This causes GCC's output file to have the @var{ALL} subtype, instead of
10148 one controlled by the @option{-mcpu} or @option{-march} option.
10150 @item -allowable_client @var{client_name}
10151 @itemx -client_name
10152 @itemx -compatibility_version
10153 @itemx -current_version
10155 @itemx -dependency-file
10157 @itemx -dylinker_install_name
10159 @itemx -exported_symbols_list
10161 @itemx -flat_namespace
10162 @itemx -force_flat_namespace
10163 @itemx -headerpad_max_install_names
10166 @itemx -install_name
10167 @itemx -keep_private_externs
10168 @itemx -multi_module
10169 @itemx -multiply_defined
10170 @itemx -multiply_defined_unused
10172 @itemx -no_dead_strip_inits_and_terms
10173 @itemx -nofixprebinding
10174 @itemx -nomultidefs
10176 @itemx -noseglinkedit
10177 @itemx -pagezero_size
10179 @itemx -prebind_all_twolevel_modules
10180 @itemx -private_bundle
10181 @itemx -read_only_relocs
10183 @itemx -sectobjectsymbols
10187 @itemx -sectobjectsymbols
10190 @itemx -segs_read_only_addr
10191 @itemx -segs_read_write_addr
10192 @itemx -seg_addr_table
10193 @itemx -seg_addr_table_filename
10194 @itemx -seglinkedit
10196 @itemx -segs_read_only_addr
10197 @itemx -segs_read_write_addr
10198 @itemx -single_module
10200 @itemx -sub_library
10201 @itemx -sub_umbrella
10202 @itemx -twolevel_namespace
10205 @itemx -unexported_symbols_list
10206 @itemx -weak_reference_mismatches
10207 @itemx -whatsloaded
10208 @opindex allowable_client
10209 @opindex client_name
10210 @opindex compatibility_version
10211 @opindex current_version
10212 @opindex dead_strip
10213 @opindex dependency-file
10214 @opindex dylib_file
10215 @opindex dylinker_install_name
10217 @opindex exported_symbols_list
10219 @opindex flat_namespace
10220 @opindex force_flat_namespace
10221 @opindex headerpad_max_install_names
10222 @opindex image_base
10224 @opindex install_name
10225 @opindex keep_private_externs
10226 @opindex multi_module
10227 @opindex multiply_defined
10228 @opindex multiply_defined_unused
10229 @opindex noall_load
10230 @opindex no_dead_strip_inits_and_terms
10231 @opindex nofixprebinding
10232 @opindex nomultidefs
10234 @opindex noseglinkedit
10235 @opindex pagezero_size
10237 @opindex prebind_all_twolevel_modules
10238 @opindex private_bundle
10239 @opindex read_only_relocs
10241 @opindex sectobjectsymbols
10244 @opindex sectcreate
10245 @opindex sectobjectsymbols
10248 @opindex segs_read_only_addr
10249 @opindex segs_read_write_addr
10250 @opindex seg_addr_table
10251 @opindex seg_addr_table_filename
10252 @opindex seglinkedit
10254 @opindex segs_read_only_addr
10255 @opindex segs_read_write_addr
10256 @opindex single_module
10258 @opindex sub_library
10259 @opindex sub_umbrella
10260 @opindex twolevel_namespace
10263 @opindex unexported_symbols_list
10264 @opindex weak_reference_mismatches
10265 @opindex whatsloaded
10266 These options are passed to the Darwin linker. The Darwin linker man page
10267 describes them in detail.
10270 @node DEC Alpha Options
10271 @subsection DEC Alpha Options
10273 These @samp{-m} options are defined for the DEC Alpha implementations:
10276 @item -mno-soft-float
10277 @itemx -msoft-float
10278 @opindex mno-soft-float
10279 @opindex msoft-float
10280 Use (do not use) the hardware floating-point instructions for
10281 floating-point operations. When @option{-msoft-float} is specified,
10282 functions in @file{libgcc.a} will be used to perform floating-point
10283 operations. Unless they are replaced by routines that emulate the
10284 floating-point operations, or compiled in such a way as to call such
10285 emulations routines, these routines will issue floating-point
10286 operations. If you are compiling for an Alpha without floating-point
10287 operations, you must ensure that the library is built so as not to call
10290 Note that Alpha implementations without floating-point operations are
10291 required to have floating-point registers.
10294 @itemx -mno-fp-regs
10296 @opindex mno-fp-regs
10297 Generate code that uses (does not use) the floating-point register set.
10298 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10299 register set is not used, floating point operands are passed in integer
10300 registers as if they were integers and floating-point results are passed
10301 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10302 so any function with a floating-point argument or return value called by code
10303 compiled with @option{-mno-fp-regs} must also be compiled with that
10306 A typical use of this option is building a kernel that does not use,
10307 and hence need not save and restore, any floating-point registers.
10311 The Alpha architecture implements floating-point hardware optimized for
10312 maximum performance. It is mostly compliant with the IEEE floating
10313 point standard. However, for full compliance, software assistance is
10314 required. This option generates code fully IEEE compliant code
10315 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10316 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10317 defined during compilation. The resulting code is less efficient but is
10318 able to correctly support denormalized numbers and exceptional IEEE
10319 values such as not-a-number and plus/minus infinity. Other Alpha
10320 compilers call this option @option{-ieee_with_no_inexact}.
10322 @item -mieee-with-inexact
10323 @opindex mieee-with-inexact
10324 This is like @option{-mieee} except the generated code also maintains
10325 the IEEE @var{inexact-flag}. Turning on this option causes the
10326 generated code to implement fully-compliant IEEE math. In addition to
10327 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10328 macro. On some Alpha implementations the resulting code may execute
10329 significantly slower than the code generated by default. Since there is
10330 very little code that depends on the @var{inexact-flag}, you should
10331 normally not specify this option. Other Alpha compilers call this
10332 option @option{-ieee_with_inexact}.
10334 @item -mfp-trap-mode=@var{trap-mode}
10335 @opindex mfp-trap-mode
10336 This option controls what floating-point related traps are enabled.
10337 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10338 The trap mode can be set to one of four values:
10342 This is the default (normal) setting. The only traps that are enabled
10343 are the ones that cannot be disabled in software (e.g., division by zero
10347 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10351 Like @samp{u}, but the instructions are marked to be safe for software
10352 completion (see Alpha architecture manual for details).
10355 Like @samp{su}, but inexact traps are enabled as well.
10358 @item -mfp-rounding-mode=@var{rounding-mode}
10359 @opindex mfp-rounding-mode
10360 Selects the IEEE rounding mode. Other Alpha compilers call this option
10361 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10366 Normal IEEE rounding mode. Floating point numbers are rounded towards
10367 the nearest machine number or towards the even machine number in case
10371 Round towards minus infinity.
10374 Chopped rounding mode. Floating point numbers are rounded towards zero.
10377 Dynamic rounding mode. A field in the floating point control register
10378 (@var{fpcr}, see Alpha architecture reference manual) controls the
10379 rounding mode in effect. The C library initializes this register for
10380 rounding towards plus infinity. Thus, unless your program modifies the
10381 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10384 @item -mtrap-precision=@var{trap-precision}
10385 @opindex mtrap-precision
10386 In the Alpha architecture, floating point traps are imprecise. This
10387 means without software assistance it is impossible to recover from a
10388 floating trap and program execution normally needs to be terminated.
10389 GCC can generate code that can assist operating system trap handlers
10390 in determining the exact location that caused a floating point trap.
10391 Depending on the requirements of an application, different levels of
10392 precisions can be selected:
10396 Program precision. This option is the default and means a trap handler
10397 can only identify which program caused a floating point exception.
10400 Function precision. The trap handler can determine the function that
10401 caused a floating point exception.
10404 Instruction precision. The trap handler can determine the exact
10405 instruction that caused a floating point exception.
10408 Other Alpha compilers provide the equivalent options called
10409 @option{-scope_safe} and @option{-resumption_safe}.
10411 @item -mieee-conformant
10412 @opindex mieee-conformant
10413 This option marks the generated code as IEEE conformant. You must not
10414 use this option unless you also specify @option{-mtrap-precision=i} and either
10415 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10416 is to emit the line @samp{.eflag 48} in the function prologue of the
10417 generated assembly file. Under DEC Unix, this has the effect that
10418 IEEE-conformant math library routines will be linked in.
10420 @item -mbuild-constants
10421 @opindex mbuild-constants
10422 Normally GCC examines a 32- or 64-bit integer constant to
10423 see if it can construct it from smaller constants in two or three
10424 instructions. If it cannot, it will output the constant as a literal and
10425 generate code to load it from the data segment at runtime.
10427 Use this option to require GCC to construct @emph{all} integer constants
10428 using code, even if it takes more instructions (the maximum is six).
10430 You would typically use this option to build a shared library dynamic
10431 loader. Itself a shared library, it must relocate itself in memory
10432 before it can find the variables and constants in its own data segment.
10438 Select whether to generate code to be assembled by the vendor-supplied
10439 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10457 Indicate whether GCC should generate code to use the optional BWX,
10458 CIX, FIX and MAX instruction sets. The default is to use the instruction
10459 sets supported by the CPU type specified via @option{-mcpu=} option or that
10460 of the CPU on which GCC was built if none was specified.
10463 @itemx -mfloat-ieee
10464 @opindex mfloat-vax
10465 @opindex mfloat-ieee
10466 Generate code that uses (does not use) VAX F and G floating point
10467 arithmetic instead of IEEE single and double precision.
10469 @item -mexplicit-relocs
10470 @itemx -mno-explicit-relocs
10471 @opindex mexplicit-relocs
10472 @opindex mno-explicit-relocs
10473 Older Alpha assemblers provided no way to generate symbol relocations
10474 except via assembler macros. Use of these macros does not allow
10475 optimal instruction scheduling. GNU binutils as of version 2.12
10476 supports a new syntax that allows the compiler to explicitly mark
10477 which relocations should apply to which instructions. This option
10478 is mostly useful for debugging, as GCC detects the capabilities of
10479 the assembler when it is built and sets the default accordingly.
10482 @itemx -mlarge-data
10483 @opindex msmall-data
10484 @opindex mlarge-data
10485 When @option{-mexplicit-relocs} is in effect, static data is
10486 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10487 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10488 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10489 16-bit relocations off of the @code{$gp} register. This limits the
10490 size of the small data area to 64KB, but allows the variables to be
10491 directly accessed via a single instruction.
10493 The default is @option{-mlarge-data}. With this option the data area
10494 is limited to just below 2GB@. Programs that require more than 2GB of
10495 data must use @code{malloc} or @code{mmap} to allocate the data in the
10496 heap instead of in the program's data segment.
10498 When generating code for shared libraries, @option{-fpic} implies
10499 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10502 @itemx -mlarge-text
10503 @opindex msmall-text
10504 @opindex mlarge-text
10505 When @option{-msmall-text} is used, the compiler assumes that the
10506 code of the entire program (or shared library) fits in 4MB, and is
10507 thus reachable with a branch instruction. When @option{-msmall-data}
10508 is used, the compiler can assume that all local symbols share the
10509 same @code{$gp} value, and thus reduce the number of instructions
10510 required for a function call from 4 to 1.
10512 The default is @option{-mlarge-text}.
10514 @item -mcpu=@var{cpu_type}
10516 Set the instruction set and instruction scheduling parameters for
10517 machine type @var{cpu_type}. You can specify either the @samp{EV}
10518 style name or the corresponding chip number. GCC supports scheduling
10519 parameters for the EV4, EV5 and EV6 family of processors and will
10520 choose the default values for the instruction set from the processor
10521 you specify. If you do not specify a processor type, GCC will default
10522 to the processor on which the compiler was built.
10524 Supported values for @var{cpu_type} are
10530 Schedules as an EV4 and has no instruction set extensions.
10534 Schedules as an EV5 and has no instruction set extensions.
10538 Schedules as an EV5 and supports the BWX extension.
10543 Schedules as an EV5 and supports the BWX and MAX extensions.
10547 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10551 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10554 Native Linux/GNU toolchains also support the value @samp{native},
10555 which selects the best architecture option for the host processor.
10556 @option{-mcpu=native} has no effect if GCC does not recognize
10559 @item -mtune=@var{cpu_type}
10561 Set only the instruction scheduling parameters for machine type
10562 @var{cpu_type}. The instruction set is not changed.
10564 Native Linux/GNU toolchains also support the value @samp{native},
10565 which selects the best architecture option for the host processor.
10566 @option{-mtune=native} has no effect if GCC does not recognize
10569 @item -mmemory-latency=@var{time}
10570 @opindex mmemory-latency
10571 Sets the latency the scheduler should assume for typical memory
10572 references as seen by the application. This number is highly
10573 dependent on the memory access patterns used by the application
10574 and the size of the external cache on the machine.
10576 Valid options for @var{time} are
10580 A decimal number representing clock cycles.
10586 The compiler contains estimates of the number of clock cycles for
10587 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10588 (also called Dcache, Scache, and Bcache), as well as to main memory.
10589 Note that L3 is only valid for EV5.
10594 @node DEC Alpha/VMS Options
10595 @subsection DEC Alpha/VMS Options
10597 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10600 @item -mvms-return-codes
10601 @opindex mvms-return-codes
10602 Return VMS condition codes from main. The default is to return POSIX
10603 style condition (e.g.@: error) codes.
10605 @item -mdebug-main=@var{prefix}
10606 @opindex mdebug-main=@var{prefix}
10607 Flag the first routine whose name starts with @var{prefix} as the main
10608 routine for the debugger.
10612 Default to 64bit memory allocation routines.
10616 @subsection FR30 Options
10617 @cindex FR30 Options
10619 These options are defined specifically for the FR30 port.
10623 @item -msmall-model
10624 @opindex msmall-model
10625 Use the small address space model. This can produce smaller code, but
10626 it does assume that all symbolic values and addresses will fit into a
10631 Assume that run-time support has been provided and so there is no need
10632 to include the simulator library (@file{libsim.a}) on the linker
10638 @subsection FRV Options
10639 @cindex FRV Options
10645 Only use the first 32 general purpose registers.
10650 Use all 64 general purpose registers.
10655 Use only the first 32 floating point registers.
10660 Use all 64 floating point registers
10663 @opindex mhard-float
10665 Use hardware instructions for floating point operations.
10668 @opindex msoft-float
10670 Use library routines for floating point operations.
10675 Dynamically allocate condition code registers.
10680 Do not try to dynamically allocate condition code registers, only
10681 use @code{icc0} and @code{fcc0}.
10686 Change ABI to use double word insns.
10691 Do not use double word instructions.
10696 Use floating point double instructions.
10699 @opindex mno-double
10701 Do not use floating point double instructions.
10706 Use media instructions.
10711 Do not use media instructions.
10716 Use multiply and add/subtract instructions.
10719 @opindex mno-muladd
10721 Do not use multiply and add/subtract instructions.
10726 Select the FDPIC ABI, that uses function descriptors to represent
10727 pointers to functions. Without any PIC/PIE-related options, it
10728 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10729 assumes GOT entries and small data are within a 12-bit range from the
10730 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10731 are computed with 32 bits.
10732 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10735 @opindex minline-plt
10737 Enable inlining of PLT entries in function calls to functions that are
10738 not known to bind locally. It has no effect without @option{-mfdpic}.
10739 It's enabled by default if optimizing for speed and compiling for
10740 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10741 optimization option such as @option{-O3} or above is present in the
10747 Assume a large TLS segment when generating thread-local code.
10752 Do not assume a large TLS segment when generating thread-local code.
10757 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10758 that is known to be in read-only sections. It's enabled by default,
10759 except for @option{-fpic} or @option{-fpie}: even though it may help
10760 make the global offset table smaller, it trades 1 instruction for 4.
10761 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10762 one of which may be shared by multiple symbols, and it avoids the need
10763 for a GOT entry for the referenced symbol, so it's more likely to be a
10764 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10766 @item -multilib-library-pic
10767 @opindex multilib-library-pic
10769 Link with the (library, not FD) pic libraries. It's implied by
10770 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10771 @option{-fpic} without @option{-mfdpic}. You should never have to use
10775 @opindex mlinked-fp
10777 Follow the EABI requirement of always creating a frame pointer whenever
10778 a stack frame is allocated. This option is enabled by default and can
10779 be disabled with @option{-mno-linked-fp}.
10782 @opindex mlong-calls
10784 Use indirect addressing to call functions outside the current
10785 compilation unit. This allows the functions to be placed anywhere
10786 within the 32-bit address space.
10788 @item -malign-labels
10789 @opindex malign-labels
10791 Try to align labels to an 8-byte boundary by inserting nops into the
10792 previous packet. This option only has an effect when VLIW packing
10793 is enabled. It doesn't create new packets; it merely adds nops to
10796 @item -mlibrary-pic
10797 @opindex mlibrary-pic
10799 Generate position-independent EABI code.
10804 Use only the first four media accumulator registers.
10809 Use all eight media accumulator registers.
10814 Pack VLIW instructions.
10819 Do not pack VLIW instructions.
10822 @opindex mno-eflags
10824 Do not mark ABI switches in e_flags.
10827 @opindex mcond-move
10829 Enable the use of conditional-move instructions (default).
10831 This switch is mainly for debugging the compiler and will likely be removed
10832 in a future version.
10834 @item -mno-cond-move
10835 @opindex mno-cond-move
10837 Disable the use of conditional-move instructions.
10839 This switch is mainly for debugging the compiler and will likely be removed
10840 in a future version.
10845 Enable the use of conditional set instructions (default).
10847 This switch is mainly for debugging the compiler and will likely be removed
10848 in a future version.
10853 Disable the use of conditional set instructions.
10855 This switch is mainly for debugging the compiler and will likely be removed
10856 in a future version.
10859 @opindex mcond-exec
10861 Enable the use of conditional execution (default).
10863 This switch is mainly for debugging the compiler and will likely be removed
10864 in a future version.
10866 @item -mno-cond-exec
10867 @opindex mno-cond-exec
10869 Disable the use of conditional execution.
10871 This switch is mainly for debugging the compiler and will likely be removed
10872 in a future version.
10874 @item -mvliw-branch
10875 @opindex mvliw-branch
10877 Run a pass to pack branches into VLIW instructions (default).
10879 This switch is mainly for debugging the compiler and will likely be removed
10880 in a future version.
10882 @item -mno-vliw-branch
10883 @opindex mno-vliw-branch
10885 Do not run a pass to pack branches into VLIW instructions.
10887 This switch is mainly for debugging the compiler and will likely be removed
10888 in a future version.
10890 @item -mmulti-cond-exec
10891 @opindex mmulti-cond-exec
10893 Enable optimization of @code{&&} and @code{||} in conditional execution
10896 This switch is mainly for debugging the compiler and will likely be removed
10897 in a future version.
10899 @item -mno-multi-cond-exec
10900 @opindex mno-multi-cond-exec
10902 Disable optimization of @code{&&} and @code{||} in conditional execution.
10904 This switch is mainly for debugging the compiler and will likely be removed
10905 in a future version.
10907 @item -mnested-cond-exec
10908 @opindex mnested-cond-exec
10910 Enable nested conditional execution optimizations (default).
10912 This switch is mainly for debugging the compiler and will likely be removed
10913 in a future version.
10915 @item -mno-nested-cond-exec
10916 @opindex mno-nested-cond-exec
10918 Disable nested conditional execution optimizations.
10920 This switch is mainly for debugging the compiler and will likely be removed
10921 in a future version.
10923 @item -moptimize-membar
10924 @opindex moptimize-membar
10926 This switch removes redundant @code{membar} instructions from the
10927 compiler generated code. It is enabled by default.
10929 @item -mno-optimize-membar
10930 @opindex mno-optimize-membar
10932 This switch disables the automatic removal of redundant @code{membar}
10933 instructions from the generated code.
10935 @item -mtomcat-stats
10936 @opindex mtomcat-stats
10938 Cause gas to print out tomcat statistics.
10940 @item -mcpu=@var{cpu}
10943 Select the processor type for which to generate code. Possible values are
10944 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10945 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10949 @node GNU/Linux Options
10950 @subsection GNU/Linux Options
10952 These @samp{-m} options are defined for GNU/Linux targets:
10957 Use the GNU C library instead of uClibc. This is the default except
10958 on @samp{*-*-linux-*uclibc*} targets.
10962 Use uClibc instead of the GNU C library. This is the default on
10963 @samp{*-*-linux-*uclibc*} targets.
10966 @node H8/300 Options
10967 @subsection H8/300 Options
10969 These @samp{-m} options are defined for the H8/300 implementations:
10974 Shorten some address references at link time, when possible; uses the
10975 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10976 ld, Using ld}, for a fuller description.
10980 Generate code for the H8/300H@.
10984 Generate code for the H8S@.
10988 Generate code for the H8S and H8/300H in the normal mode. This switch
10989 must be used either with @option{-mh} or @option{-ms}.
10993 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10997 Make @code{int} data 32 bits by default.
11000 @opindex malign-300
11001 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11002 The default for the H8/300H and H8S is to align longs and floats on 4
11004 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11005 This option has no effect on the H8/300.
11009 @subsection HPPA Options
11010 @cindex HPPA Options
11012 These @samp{-m} options are defined for the HPPA family of computers:
11015 @item -march=@var{architecture-type}
11017 Generate code for the specified architecture. The choices for
11018 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11019 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11020 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11021 architecture option for your machine. Code compiled for lower numbered
11022 architectures will run on higher numbered architectures, but not the
11025 @item -mpa-risc-1-0
11026 @itemx -mpa-risc-1-1
11027 @itemx -mpa-risc-2-0
11028 @opindex mpa-risc-1-0
11029 @opindex mpa-risc-1-1
11030 @opindex mpa-risc-2-0
11031 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11034 @opindex mbig-switch
11035 Generate code suitable for big switch tables. Use this option only if
11036 the assembler/linker complain about out of range branches within a switch
11039 @item -mjump-in-delay
11040 @opindex mjump-in-delay
11041 Fill delay slots of function calls with unconditional jump instructions
11042 by modifying the return pointer for the function call to be the target
11043 of the conditional jump.
11045 @item -mdisable-fpregs
11046 @opindex mdisable-fpregs
11047 Prevent floating point registers from being used in any manner. This is
11048 necessary for compiling kernels which perform lazy context switching of
11049 floating point registers. If you use this option and attempt to perform
11050 floating point operations, the compiler will abort.
11052 @item -mdisable-indexing
11053 @opindex mdisable-indexing
11054 Prevent the compiler from using indexing address modes. This avoids some
11055 rather obscure problems when compiling MIG generated code under MACH@.
11057 @item -mno-space-regs
11058 @opindex mno-space-regs
11059 Generate code that assumes the target has no space registers. This allows
11060 GCC to generate faster indirect calls and use unscaled index address modes.
11062 Such code is suitable for level 0 PA systems and kernels.
11064 @item -mfast-indirect-calls
11065 @opindex mfast-indirect-calls
11066 Generate code that assumes calls never cross space boundaries. This
11067 allows GCC to emit code which performs faster indirect calls.
11069 This option will not work in the presence of shared libraries or nested
11072 @item -mfixed-range=@var{register-range}
11073 @opindex mfixed-range
11074 Generate code treating the given register range as fixed registers.
11075 A fixed register is one that the register allocator can not use. This is
11076 useful when compiling kernel code. A register range is specified as
11077 two registers separated by a dash. Multiple register ranges can be
11078 specified separated by a comma.
11080 @item -mlong-load-store
11081 @opindex mlong-load-store
11082 Generate 3-instruction load and store sequences as sometimes required by
11083 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11086 @item -mportable-runtime
11087 @opindex mportable-runtime
11088 Use the portable calling conventions proposed by HP for ELF systems.
11092 Enable the use of assembler directives only GAS understands.
11094 @item -mschedule=@var{cpu-type}
11096 Schedule code according to the constraints for the machine type
11097 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11098 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11099 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11100 proper scheduling option for your machine. The default scheduling is
11104 @opindex mlinker-opt
11105 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11106 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11107 linkers in which they give bogus error messages when linking some programs.
11110 @opindex msoft-float
11111 Generate output containing library calls for floating point.
11112 @strong{Warning:} the requisite libraries are not available for all HPPA
11113 targets. Normally the facilities of the machine's usual C compiler are
11114 used, but this cannot be done directly in cross-compilation. You must make
11115 your own arrangements to provide suitable library functions for
11118 @option{-msoft-float} changes the calling convention in the output file;
11119 therefore, it is only useful if you compile @emph{all} of a program with
11120 this option. In particular, you need to compile @file{libgcc.a}, the
11121 library that comes with GCC, with @option{-msoft-float} in order for
11126 Generate the predefine, @code{_SIO}, for server IO@. The default is
11127 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11128 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11129 options are available under HP-UX and HI-UX@.
11133 Use GNU ld specific options. This passes @option{-shared} to ld when
11134 building a shared library. It is the default when GCC is configured,
11135 explicitly or implicitly, with the GNU linker. This option does not
11136 have any affect on which ld is called, it only changes what parameters
11137 are passed to that ld. The ld that is called is determined by the
11138 @option{--with-ld} configure option, GCC's program search path, and
11139 finally by the user's @env{PATH}. The linker used by GCC can be printed
11140 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11141 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11145 Use HP ld specific options. This passes @option{-b} to ld when building
11146 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11147 links. It is the default when GCC is configured, explicitly or
11148 implicitly, with the HP linker. This option does not have any affect on
11149 which ld is called, it only changes what parameters are passed to that
11150 ld. The ld that is called is determined by the @option{--with-ld}
11151 configure option, GCC's program search path, and finally by the user's
11152 @env{PATH}. The linker used by GCC can be printed using @samp{which
11153 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11154 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11157 @opindex mno-long-calls
11158 Generate code that uses long call sequences. This ensures that a call
11159 is always able to reach linker generated stubs. The default is to generate
11160 long calls only when the distance from the call site to the beginning
11161 of the function or translation unit, as the case may be, exceeds a
11162 predefined limit set by the branch type being used. The limits for
11163 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11164 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11167 Distances are measured from the beginning of functions when using the
11168 @option{-ffunction-sections} option, or when using the @option{-mgas}
11169 and @option{-mno-portable-runtime} options together under HP-UX with
11172 It is normally not desirable to use this option as it will degrade
11173 performance. However, it may be useful in large applications,
11174 particularly when partial linking is used to build the application.
11176 The types of long calls used depends on the capabilities of the
11177 assembler and linker, and the type of code being generated. The
11178 impact on systems that support long absolute calls, and long pic
11179 symbol-difference or pc-relative calls should be relatively small.
11180 However, an indirect call is used on 32-bit ELF systems in pic code
11181 and it is quite long.
11183 @item -munix=@var{unix-std}
11185 Generate compiler predefines and select a startfile for the specified
11186 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11187 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11188 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11189 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11190 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11193 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11194 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11195 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11196 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11197 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11198 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11200 It is @emph{important} to note that this option changes the interfaces
11201 for various library routines. It also affects the operational behavior
11202 of the C library. Thus, @emph{extreme} care is needed in using this
11205 Library code that is intended to operate with more than one UNIX
11206 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11207 as appropriate. Most GNU software doesn't provide this capability.
11211 Suppress the generation of link options to search libdld.sl when the
11212 @option{-static} option is specified on HP-UX 10 and later.
11216 The HP-UX implementation of setlocale in libc has a dependency on
11217 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11218 when the @option{-static} option is specified, special link options
11219 are needed to resolve this dependency.
11221 On HP-UX 10 and later, the GCC driver adds the necessary options to
11222 link with libdld.sl when the @option{-static} option is specified.
11223 This causes the resulting binary to be dynamic. On the 64-bit port,
11224 the linkers generate dynamic binaries by default in any case. The
11225 @option{-nolibdld} option can be used to prevent the GCC driver from
11226 adding these link options.
11230 Add support for multithreading with the @dfn{dce thread} library
11231 under HP-UX@. This option sets flags for both the preprocessor and
11235 @node i386 and x86-64 Options
11236 @subsection Intel 386 and AMD x86-64 Options
11237 @cindex i386 Options
11238 @cindex x86-64 Options
11239 @cindex Intel 386 Options
11240 @cindex AMD x86-64 Options
11242 These @samp{-m} options are defined for the i386 and x86-64 family of
11246 @item -mtune=@var{cpu-type}
11248 Tune to @var{cpu-type} everything applicable about the generated code, except
11249 for the ABI and the set of available instructions. The choices for
11250 @var{cpu-type} are:
11253 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11254 If you know the CPU on which your code will run, then you should use
11255 the corresponding @option{-mtune} option instead of
11256 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11257 of your application will have, then you should use this option.
11259 As new processors are deployed in the marketplace, the behavior of this
11260 option will change. Therefore, if you upgrade to a newer version of
11261 GCC, the code generated option will change to reflect the processors
11262 that were most common when that version of GCC was released.
11264 There is no @option{-march=generic} option because @option{-march}
11265 indicates the instruction set the compiler can use, and there is no
11266 generic instruction set applicable to all processors. In contrast,
11267 @option{-mtune} indicates the processor (or, in this case, collection of
11268 processors) for which the code is optimized.
11270 This selects the CPU to tune for at compilation time by determining
11271 the processor type of the compiling machine. Using @option{-mtune=native}
11272 will produce code optimized for the local machine under the constraints
11273 of the selected instruction set. Using @option{-march=native} will
11274 enable all instruction subsets supported by the local machine (hence
11275 the result might not run on different machines).
11277 Original Intel's i386 CPU@.
11279 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11280 @item i586, pentium
11281 Intel Pentium CPU with no MMX support.
11283 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11285 Intel PentiumPro CPU@.
11287 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11288 instruction set will be used, so the code will run on all i686 family chips.
11290 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11291 @item pentium3, pentium3m
11292 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11295 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11296 support. Used by Centrino notebooks.
11297 @item pentium4, pentium4m
11298 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11300 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11303 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11304 SSE2 and SSE3 instruction set support.
11306 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11307 instruction set support.
11309 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11310 instruction set support.
11312 AMD K6 CPU with MMX instruction set support.
11314 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11315 @item athlon, athlon-tbird
11316 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11318 @item athlon-4, athlon-xp, athlon-mp
11319 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11320 instruction set support.
11321 @item k8, opteron, athlon64, athlon-fx
11322 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11323 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11324 @item k8-sse3, opteron-sse3, athlon64-sse3
11325 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11326 @item amdfam10, barcelona
11327 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11328 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11329 instruction set extensions.)
11331 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11334 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11335 instruction set support.
11337 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11338 implemented for this chip.)
11340 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11341 implemented for this chip.)
11343 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11346 While picking a specific @var{cpu-type} will schedule things appropriately
11347 for that particular chip, the compiler will not generate any code that
11348 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11351 @item -march=@var{cpu-type}
11353 Generate instructions for the machine type @var{cpu-type}. The choices
11354 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11355 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11357 @item -mcpu=@var{cpu-type}
11359 A deprecated synonym for @option{-mtune}.
11361 @item -mfpmath=@var{unit}
11363 Generate floating point arithmetics for selected unit @var{unit}. The choices
11364 for @var{unit} are:
11368 Use the standard 387 floating point coprocessor present majority of chips and
11369 emulated otherwise. Code compiled with this option will run almost everywhere.
11370 The temporary results are computed in 80bit precision instead of precision
11371 specified by the type resulting in slightly different results compared to most
11372 of other chips. See @option{-ffloat-store} for more detailed description.
11374 This is the default choice for i386 compiler.
11377 Use scalar floating point instructions present in the SSE instruction set.
11378 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11379 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11380 instruction set supports only single precision arithmetics, thus the double and
11381 extended precision arithmetics is still done using 387. Later version, present
11382 only in Pentium4 and the future AMD x86-64 chips supports double precision
11385 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11386 or @option{-msse2} switches to enable SSE extensions and make this option
11387 effective. For the x86-64 compiler, these extensions are enabled by default.
11389 The resulting code should be considerably faster in the majority of cases and avoid
11390 the numerical instability problems of 387 code, but may break some existing
11391 code that expects temporaries to be 80bit.
11393 This is the default choice for the x86-64 compiler.
11398 Attempt to utilize both instruction sets at once. This effectively double the
11399 amount of available registers and on chips with separate execution units for
11400 387 and SSE the execution resources too. Use this option with care, as it is
11401 still experimental, because the GCC register allocator does not model separate
11402 functional units well resulting in instable performance.
11405 @item -masm=@var{dialect}
11406 @opindex masm=@var{dialect}
11407 Output asm instructions using selected @var{dialect}. Supported
11408 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11409 not support @samp{intel}.
11412 @itemx -mno-ieee-fp
11414 @opindex mno-ieee-fp
11415 Control whether or not the compiler uses IEEE floating point
11416 comparisons. These handle correctly the case where the result of a
11417 comparison is unordered.
11420 @opindex msoft-float
11421 Generate output containing library calls for floating point.
11422 @strong{Warning:} the requisite libraries are not part of GCC@.
11423 Normally the facilities of the machine's usual C compiler are used, but
11424 this can't be done directly in cross-compilation. You must make your
11425 own arrangements to provide suitable library functions for
11428 On machines where a function returns floating point results in the 80387
11429 register stack, some floating point opcodes may be emitted even if
11430 @option{-msoft-float} is used.
11432 @item -mno-fp-ret-in-387
11433 @opindex mno-fp-ret-in-387
11434 Do not use the FPU registers for return values of functions.
11436 The usual calling convention has functions return values of types
11437 @code{float} and @code{double} in an FPU register, even if there
11438 is no FPU@. The idea is that the operating system should emulate
11441 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11442 in ordinary CPU registers instead.
11444 @item -mno-fancy-math-387
11445 @opindex mno-fancy-math-387
11446 Some 387 emulators do not support the @code{sin}, @code{cos} and
11447 @code{sqrt} instructions for the 387. Specify this option to avoid
11448 generating those instructions. This option is the default on FreeBSD,
11449 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11450 indicates that the target cpu will always have an FPU and so the
11451 instruction will not need emulation. As of revision 2.6.1, these
11452 instructions are not generated unless you also use the
11453 @option{-funsafe-math-optimizations} switch.
11455 @item -malign-double
11456 @itemx -mno-align-double
11457 @opindex malign-double
11458 @opindex mno-align-double
11459 Control whether GCC aligns @code{double}, @code{long double}, and
11460 @code{long long} variables on a two word boundary or a one word
11461 boundary. Aligning @code{double} variables on a two word boundary will
11462 produce code that runs somewhat faster on a @samp{Pentium} at the
11463 expense of more memory.
11465 On x86-64, @option{-malign-double} is enabled by default.
11467 @strong{Warning:} if you use the @option{-malign-double} switch,
11468 structures containing the above types will be aligned differently than
11469 the published application binary interface specifications for the 386
11470 and will not be binary compatible with structures in code compiled
11471 without that switch.
11473 @item -m96bit-long-double
11474 @itemx -m128bit-long-double
11475 @opindex m96bit-long-double
11476 @opindex m128bit-long-double
11477 These switches control the size of @code{long double} type. The i386
11478 application binary interface specifies the size to be 96 bits,
11479 so @option{-m96bit-long-double} is the default in 32 bit mode.
11481 Modern architectures (Pentium and newer) would prefer @code{long double}
11482 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11483 conforming to the ABI, this would not be possible. So specifying a
11484 @option{-m128bit-long-double} will align @code{long double}
11485 to a 16 byte boundary by padding the @code{long double} with an additional
11488 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11489 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11491 Notice that neither of these options enable any extra precision over the x87
11492 standard of 80 bits for a @code{long double}.
11494 @strong{Warning:} if you override the default value for your target ABI, the
11495 structures and arrays containing @code{long double} variables will change
11496 their size as well as function calling convention for function taking
11497 @code{long double} will be modified. Hence they will not be binary
11498 compatible with arrays or structures in code compiled without that switch.
11500 @item -mlarge-data-threshold=@var{number}
11501 @opindex mlarge-data-threshold=@var{number}
11502 When @option{-mcmodel=medium} is specified, the data greater than
11503 @var{threshold} are placed in large data section. This value must be the
11504 same across all object linked into the binary and defaults to 65535.
11508 Use a different function-calling convention, in which functions that
11509 take a fixed number of arguments return with the @code{ret} @var{num}
11510 instruction, which pops their arguments while returning. This saves one
11511 instruction in the caller since there is no need to pop the arguments
11514 You can specify that an individual function is called with this calling
11515 sequence with the function attribute @samp{stdcall}. You can also
11516 override the @option{-mrtd} option by using the function attribute
11517 @samp{cdecl}. @xref{Function Attributes}.
11519 @strong{Warning:} this calling convention is incompatible with the one
11520 normally used on Unix, so you cannot use it if you need to call
11521 libraries compiled with the Unix compiler.
11523 Also, you must provide function prototypes for all functions that
11524 take variable numbers of arguments (including @code{printf});
11525 otherwise incorrect code will be generated for calls to those
11528 In addition, seriously incorrect code will result if you call a
11529 function with too many arguments. (Normally, extra arguments are
11530 harmlessly ignored.)
11532 @item -mregparm=@var{num}
11534 Control how many registers are used to pass integer arguments. By
11535 default, no registers are used to pass arguments, and at most 3
11536 registers can be used. You can control this behavior for a specific
11537 function by using the function attribute @samp{regparm}.
11538 @xref{Function Attributes}.
11540 @strong{Warning:} if you use this switch, and
11541 @var{num} is nonzero, then you must build all modules with the same
11542 value, including any libraries. This includes the system libraries and
11546 @opindex msseregparm
11547 Use SSE register passing conventions for float and double arguments
11548 and return values. You can control this behavior for a specific
11549 function by using the function attribute @samp{sseregparm}.
11550 @xref{Function Attributes}.
11552 @strong{Warning:} if you use this switch then you must build all
11553 modules with the same value, including any libraries. This includes
11554 the system libraries and startup modules.
11563 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11564 is specified, the significands of results of floating-point operations are
11565 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11566 significands of results of floating-point operations to 53 bits (double
11567 precision) and @option{-mpc80} rounds the significands of results of
11568 floating-point operations to 64 bits (extended double precision), which is
11569 the default. When this option is used, floating-point operations in higher
11570 precisions are not available to the programmer without setting the FPU
11571 control word explicitly.
11573 Setting the rounding of floating-point operations to less than the default
11574 80 bits can speed some programs by 2% or more. Note that some mathematical
11575 libraries assume that extended precision (80 bit) floating-point operations
11576 are enabled by default; routines in such libraries could suffer significant
11577 loss of accuracy, typically through so-called "catastrophic cancellation",
11578 when this option is used to set the precision to less than extended precision.
11580 @item -mstackrealign
11581 @opindex mstackrealign
11582 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11583 option will generate an alternate prologue and epilogue that realigns the
11584 runtime stack if necessary. This supports mixing legacy codes that keep
11585 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11586 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11587 applicable to individual functions.
11589 @item -mpreferred-stack-boundary=@var{num}
11590 @opindex mpreferred-stack-boundary
11591 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11592 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11593 the default is 4 (16 bytes or 128 bits).
11595 @item -mincoming-stack-boundary=@var{num}
11596 @opindex mincoming-stack-boundary
11597 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11598 boundary. If @option{-mincoming-stack-boundary} is not specified,
11599 the one specified by @option{-mpreferred-stack-boundary} will be used.
11601 On Pentium and PentiumPro, @code{double} and @code{long double} values
11602 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11603 suffer significant run time performance penalties. On Pentium III, the
11604 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11605 properly if it is not 16 byte aligned.
11607 To ensure proper alignment of this values on the stack, the stack boundary
11608 must be as aligned as that required by any value stored on the stack.
11609 Further, every function must be generated such that it keeps the stack
11610 aligned. Thus calling a function compiled with a higher preferred
11611 stack boundary from a function compiled with a lower preferred stack
11612 boundary will most likely misalign the stack. It is recommended that
11613 libraries that use callbacks always use the default setting.
11615 This extra alignment does consume extra stack space, and generally
11616 increases code size. Code that is sensitive to stack space usage, such
11617 as embedded systems and operating system kernels, may want to reduce the
11618 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11656 These switches enable or disable the use of instructions in the MMX,
11657 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, ABM or
11658 3DNow!@: extended instruction sets.
11659 These extensions are also available as built-in functions: see
11660 @ref{X86 Built-in Functions}, for details of the functions enabled and
11661 disabled by these switches.
11663 To have SSE/SSE2 instructions generated automatically from floating-point
11664 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11666 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11667 generates new AVX instructions or AVX equivalence for all SSEx instructions
11670 These options will enable GCC to use these extended instructions in
11671 generated code, even without @option{-mfpmath=sse}. Applications which
11672 perform runtime CPU detection must compile separate files for each
11673 supported architecture, using the appropriate flags. In particular,
11674 the file containing the CPU detection code should be compiled without
11679 This option instructs GCC to emit a @code{cld} instruction in the prologue
11680 of functions that use string instructions. String instructions depend on
11681 the DF flag to select between autoincrement or autodecrement mode. While the
11682 ABI specifies the DF flag to be cleared on function entry, some operating
11683 systems violate this specification by not clearing the DF flag in their
11684 exception dispatchers. The exception handler can be invoked with the DF flag
11685 set which leads to wrong direction mode, when string instructions are used.
11686 This option can be enabled by default on 32-bit x86 targets by configuring
11687 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11688 instructions can be suppressed with the @option{-mno-cld} compiler option
11693 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11694 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11695 data types. This is useful for high resolution counters that could be updated
11696 by multiple processors (or cores). This instruction is generated as part of
11697 atomic built-in functions: see @ref{Atomic Builtins} for details.
11701 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11702 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11703 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11704 SAHF are load and store instructions, respectively, for certain status flags.
11705 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11706 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11710 This option will enable GCC to use movbe instruction to implement
11711 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
11715 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
11716 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
11717 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
11721 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11722 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11723 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11724 variants) for single precision floating point arguments. These instructions
11725 are generated only when @option{-funsafe-math-optimizations} is enabled
11726 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11727 Note that while the throughput of the sequence is higher than the throughput
11728 of the non-reciprocal instruction, the precision of the sequence can be
11729 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11731 @item -mveclibabi=@var{type}
11732 @opindex mveclibabi
11733 Specifies the ABI type to use for vectorizing intrinsics using an
11734 external library. Supported types are @code{svml} for the Intel short
11735 vector math library and @code{acml} for the AMD math core library style
11736 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11737 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11738 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11739 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11740 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11741 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11742 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11743 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11744 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11745 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11746 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11747 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11748 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11749 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11750 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11751 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11752 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11753 compatible library will have to be specified at link time.
11755 @item -mabi=@var{name}
11757 Generate code for the specified calling convention. Permissible values
11758 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11759 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11760 ABI when targeting Windows. On all other systems, the default is the
11761 SYSV ABI. You can control this behavior for a specific function by
11762 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11763 @xref{Function Attributes}.
11766 @itemx -mno-push-args
11767 @opindex mpush-args
11768 @opindex mno-push-args
11769 Use PUSH operations to store outgoing parameters. This method is shorter
11770 and usually equally fast as method using SUB/MOV operations and is enabled
11771 by default. In some cases disabling it may improve performance because of
11772 improved scheduling and reduced dependencies.
11774 @item -maccumulate-outgoing-args
11775 @opindex maccumulate-outgoing-args
11776 If enabled, the maximum amount of space required for outgoing arguments will be
11777 computed in the function prologue. This is faster on most modern CPUs
11778 because of reduced dependencies, improved scheduling and reduced stack usage
11779 when preferred stack boundary is not equal to 2. The drawback is a notable
11780 increase in code size. This switch implies @option{-mno-push-args}.
11784 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11785 on thread-safe exception handling must compile and link all code with the
11786 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11787 @option{-D_MT}; when linking, it links in a special thread helper library
11788 @option{-lmingwthrd} which cleans up per thread exception handling data.
11790 @item -mno-align-stringops
11791 @opindex mno-align-stringops
11792 Do not align destination of inlined string operations. This switch reduces
11793 code size and improves performance in case the destination is already aligned,
11794 but GCC doesn't know about it.
11796 @item -minline-all-stringops
11797 @opindex minline-all-stringops
11798 By default GCC inlines string operations only when destination is known to be
11799 aligned at least to 4 byte boundary. This enables more inlining, increase code
11800 size, but may improve performance of code that depends on fast memcpy, strlen
11801 and memset for short lengths.
11803 @item -minline-stringops-dynamically
11804 @opindex minline-stringops-dynamically
11805 For string operation of unknown size, inline runtime checks so for small
11806 blocks inline code is used, while for large blocks library call is used.
11808 @item -mstringop-strategy=@var{alg}
11809 @opindex mstringop-strategy=@var{alg}
11810 Overwrite internal decision heuristic about particular algorithm to inline
11811 string operation with. The allowed values are @code{rep_byte},
11812 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11813 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11814 expanding inline loop, @code{libcall} for always expanding library call.
11816 @item -momit-leaf-frame-pointer
11817 @opindex momit-leaf-frame-pointer
11818 Don't keep the frame pointer in a register for leaf functions. This
11819 avoids the instructions to save, set up and restore frame pointers and
11820 makes an extra register available in leaf functions. The option
11821 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11822 which might make debugging harder.
11824 @item -mtls-direct-seg-refs
11825 @itemx -mno-tls-direct-seg-refs
11826 @opindex mtls-direct-seg-refs
11827 Controls whether TLS variables may be accessed with offsets from the
11828 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11829 or whether the thread base pointer must be added. Whether or not this
11830 is legal depends on the operating system, and whether it maps the
11831 segment to cover the entire TLS area.
11833 For systems that use GNU libc, the default is on.
11836 @itemx -mno-sse2avx
11838 Specify that the assembler should encode SSE instructions with VEX
11839 prefix. The option @option{-mavx} turns this on by default.
11842 These @samp{-m} switches are supported in addition to the above
11843 on AMD x86-64 processors in 64-bit environments.
11850 Generate code for a 32-bit or 64-bit environment.
11851 The 32-bit environment sets int, long and pointer to 32 bits and
11852 generates code that runs on any i386 system.
11853 The 64-bit environment sets int to 32 bits and long and pointer
11854 to 64 bits and generates code for AMD's x86-64 architecture. For
11855 darwin only the -m64 option turns off the @option{-fno-pic} and
11856 @option{-mdynamic-no-pic} options.
11858 @item -mno-red-zone
11859 @opindex mno-red-zone
11860 Do not use a so called red zone for x86-64 code. The red zone is mandated
11861 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11862 stack pointer that will not be modified by signal or interrupt handlers
11863 and therefore can be used for temporary data without adjusting the stack
11864 pointer. The flag @option{-mno-red-zone} disables this red zone.
11866 @item -mcmodel=small
11867 @opindex mcmodel=small
11868 Generate code for the small code model: the program and its symbols must
11869 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11870 Programs can be statically or dynamically linked. This is the default
11873 @item -mcmodel=kernel
11874 @opindex mcmodel=kernel
11875 Generate code for the kernel code model. The kernel runs in the
11876 negative 2 GB of the address space.
11877 This model has to be used for Linux kernel code.
11879 @item -mcmodel=medium
11880 @opindex mcmodel=medium
11881 Generate code for the medium model: The program is linked in the lower 2
11882 GB of the address space. Small symbols are also placed there. Symbols
11883 with sizes larger than @option{-mlarge-data-threshold} are put into
11884 large data or bss sections and can be located above 2GB. Programs can
11885 be statically or dynamically linked.
11887 @item -mcmodel=large
11888 @opindex mcmodel=large
11889 Generate code for the large model: This model makes no assumptions
11890 about addresses and sizes of sections.
11893 @node IA-64 Options
11894 @subsection IA-64 Options
11895 @cindex IA-64 Options
11897 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11901 @opindex mbig-endian
11902 Generate code for a big endian target. This is the default for HP-UX@.
11904 @item -mlittle-endian
11905 @opindex mlittle-endian
11906 Generate code for a little endian target. This is the default for AIX5
11912 @opindex mno-gnu-as
11913 Generate (or don't) code for the GNU assembler. This is the default.
11914 @c Also, this is the default if the configure option @option{--with-gnu-as}
11920 @opindex mno-gnu-ld
11921 Generate (or don't) code for the GNU linker. This is the default.
11922 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11927 Generate code that does not use a global pointer register. The result
11928 is not position independent code, and violates the IA-64 ABI@.
11930 @item -mvolatile-asm-stop
11931 @itemx -mno-volatile-asm-stop
11932 @opindex mvolatile-asm-stop
11933 @opindex mno-volatile-asm-stop
11934 Generate (or don't) a stop bit immediately before and after volatile asm
11937 @item -mregister-names
11938 @itemx -mno-register-names
11939 @opindex mregister-names
11940 @opindex mno-register-names
11941 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11942 the stacked registers. This may make assembler output more readable.
11948 Disable (or enable) optimizations that use the small data section. This may
11949 be useful for working around optimizer bugs.
11951 @item -mconstant-gp
11952 @opindex mconstant-gp
11953 Generate code that uses a single constant global pointer value. This is
11954 useful when compiling kernel code.
11958 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11959 This is useful when compiling firmware code.
11961 @item -minline-float-divide-min-latency
11962 @opindex minline-float-divide-min-latency
11963 Generate code for inline divides of floating point values
11964 using the minimum latency algorithm.
11966 @item -minline-float-divide-max-throughput
11967 @opindex minline-float-divide-max-throughput
11968 Generate code for inline divides of floating point values
11969 using the maximum throughput algorithm.
11971 @item -mno-inline-float-divide
11972 @opindex mno-inline-float-divide
11973 Do not generate inline code for divides of floating point values.
11975 @item -minline-int-divide-min-latency
11976 @opindex minline-int-divide-min-latency
11977 Generate code for inline divides of integer values
11978 using the minimum latency algorithm.
11980 @item -minline-int-divide-max-throughput
11981 @opindex minline-int-divide-max-throughput
11982 Generate code for inline divides of integer values
11983 using the maximum throughput algorithm.
11985 @item -mno-inline-int-divide
11986 @opindex mno-inline-int-divide
11987 Do not generate inline code for divides of integer values.
11989 @item -minline-sqrt-min-latency
11990 @opindex minline-sqrt-min-latency
11991 Generate code for inline square roots
11992 using the minimum latency algorithm.
11994 @item -minline-sqrt-max-throughput
11995 @opindex minline-sqrt-max-throughput
11996 Generate code for inline square roots
11997 using the maximum throughput algorithm.
11999 @item -mno-inline-sqrt
12000 @opindex mno-inline-sqrt
12001 Do not generate inline code for sqrt.
12004 @itemx -mno-fused-madd
12005 @opindex mfused-madd
12006 @opindex mno-fused-madd
12007 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12008 instructions. The default is to use these instructions.
12010 @item -mno-dwarf2-asm
12011 @itemx -mdwarf2-asm
12012 @opindex mno-dwarf2-asm
12013 @opindex mdwarf2-asm
12014 Don't (or do) generate assembler code for the DWARF2 line number debugging
12015 info. This may be useful when not using the GNU assembler.
12017 @item -mearly-stop-bits
12018 @itemx -mno-early-stop-bits
12019 @opindex mearly-stop-bits
12020 @opindex mno-early-stop-bits
12021 Allow stop bits to be placed earlier than immediately preceding the
12022 instruction that triggered the stop bit. This can improve instruction
12023 scheduling, but does not always do so.
12025 @item -mfixed-range=@var{register-range}
12026 @opindex mfixed-range
12027 Generate code treating the given register range as fixed registers.
12028 A fixed register is one that the register allocator can not use. This is
12029 useful when compiling kernel code. A register range is specified as
12030 two registers separated by a dash. Multiple register ranges can be
12031 specified separated by a comma.
12033 @item -mtls-size=@var{tls-size}
12035 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12038 @item -mtune=@var{cpu-type}
12040 Tune the instruction scheduling for a particular CPU, Valid values are
12041 itanium, itanium1, merced, itanium2, and mckinley.
12047 Generate code for a 32-bit or 64-bit environment.
12048 The 32-bit environment sets int, long and pointer to 32 bits.
12049 The 64-bit environment sets int to 32 bits and long and pointer
12050 to 64 bits. These are HP-UX specific flags.
12052 @item -mno-sched-br-data-spec
12053 @itemx -msched-br-data-spec
12054 @opindex mno-sched-br-data-spec
12055 @opindex msched-br-data-spec
12056 (Dis/En)able data speculative scheduling before reload.
12057 This will result in generation of the ld.a instructions and
12058 the corresponding check instructions (ld.c / chk.a).
12059 The default is 'disable'.
12061 @item -msched-ar-data-spec
12062 @itemx -mno-sched-ar-data-spec
12063 @opindex msched-ar-data-spec
12064 @opindex mno-sched-ar-data-spec
12065 (En/Dis)able data speculative scheduling after reload.
12066 This will result in generation of the ld.a instructions and
12067 the corresponding check instructions (ld.c / chk.a).
12068 The default is 'enable'.
12070 @item -mno-sched-control-spec
12071 @itemx -msched-control-spec
12072 @opindex mno-sched-control-spec
12073 @opindex msched-control-spec
12074 (Dis/En)able control speculative scheduling. This feature is
12075 available only during region scheduling (i.e.@: before reload).
12076 This will result in generation of the ld.s instructions and
12077 the corresponding check instructions chk.s .
12078 The default is 'disable'.
12080 @item -msched-br-in-data-spec
12081 @itemx -mno-sched-br-in-data-spec
12082 @opindex msched-br-in-data-spec
12083 @opindex mno-sched-br-in-data-spec
12084 (En/Dis)able speculative scheduling of the instructions that
12085 are dependent on the data speculative loads before reload.
12086 This is effective only with @option{-msched-br-data-spec} enabled.
12087 The default is 'enable'.
12089 @item -msched-ar-in-data-spec
12090 @itemx -mno-sched-ar-in-data-spec
12091 @opindex msched-ar-in-data-spec
12092 @opindex mno-sched-ar-in-data-spec
12093 (En/Dis)able speculative scheduling of the instructions that
12094 are dependent on the data speculative loads after reload.
12095 This is effective only with @option{-msched-ar-data-spec} enabled.
12096 The default is 'enable'.
12098 @item -msched-in-control-spec
12099 @itemx -mno-sched-in-control-spec
12100 @opindex msched-in-control-spec
12101 @opindex mno-sched-in-control-spec
12102 (En/Dis)able speculative scheduling of the instructions that
12103 are dependent on the control speculative loads.
12104 This is effective only with @option{-msched-control-spec} enabled.
12105 The default is 'enable'.
12107 @item -mno-sched-prefer-non-data-spec-insns
12108 @itemx -msched-prefer-non-data-spec-insns
12109 @opindex mno-sched-prefer-non-data-spec-insns
12110 @opindex msched-prefer-non-data-spec-insns
12111 If enabled, data speculative instructions will be chosen for schedule
12112 only if there are no other choices at the moment. This will make
12113 the use of the data speculation much more conservative.
12114 The default is 'disable'.
12116 @item -mno-sched-prefer-non-control-spec-insns
12117 @itemx -msched-prefer-non-control-spec-insns
12118 @opindex mno-sched-prefer-non-control-spec-insns
12119 @opindex msched-prefer-non-control-spec-insns
12120 If enabled, control speculative instructions will be chosen for schedule
12121 only if there are no other choices at the moment. This will make
12122 the use of the control speculation much more conservative.
12123 The default is 'disable'.
12125 @item -mno-sched-count-spec-in-critical-path
12126 @itemx -msched-count-spec-in-critical-path
12127 @opindex mno-sched-count-spec-in-critical-path
12128 @opindex msched-count-spec-in-critical-path
12129 If enabled, speculative dependencies will be considered during
12130 computation of the instructions priorities. This will make the use of the
12131 speculation a bit more conservative.
12132 The default is 'disable'.
12134 @item -msched-spec-ldc
12135 @opindex msched-spec-ldc
12136 Use a simple data speculation check. This option is on by default.
12138 @item -msched-control-spec-ldc
12139 @opindex msched-spec-ldc
12140 Use a simple check for control speculation. This option is on by default.
12142 @item -msched-stop-bits-after-every-cycle
12143 @opindex msched-stop-bits-after-every-cycle
12144 Place a stop bit after every cycle when scheduling. This option is on
12147 @item -msched-fp-mem-deps-zero-cost
12148 @opindex msched-fp-mem-deps-zero-cost
12149 Assume that floating-point stores and loads are not likely to cause a conflict
12150 when placed into the same instruction group. This option is disabled by
12153 @item -msel-sched-dont-check-control-spec
12154 @opindex msel-sched-dont-check-control-spec
12155 Generate checks for control speculation in selective scheduling.
12156 This flag is disabled by default.
12158 @item -msched-max-memory-insns=@var{max-insns}
12159 @opindex msched-max-memory-insns
12160 Limit on the number of memory insns per instruction group, giving lower
12161 priority to subsequent memory insns attempting to schedule in the same
12162 instruction group. Frequently useful to prevent cache bank conflicts.
12163 The default value is 1.
12165 @item -msched-max-memory-insns-hard-limit
12166 @opindex msched-max-memory-insns-hard-limit
12167 Disallow more than `msched-max-memory-insns' in instruction group.
12168 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12169 when limit is reached but may still schedule memory operations.
12173 @node IA-64/VMS Options
12174 @subsection IA-64/VMS Options
12176 These @samp{-m} options are defined for the IA-64/VMS implementations:
12179 @item -mvms-return-codes
12180 @opindex mvms-return-codes
12181 Return VMS condition codes from main. The default is to return POSIX
12182 style condition (e.g.@ error) codes.
12184 @item -mdebug-main=@var{prefix}
12185 @opindex mdebug-main=@var{prefix}
12186 Flag the first routine whose name starts with @var{prefix} as the main
12187 routine for the debugger.
12191 Default to 64bit memory allocation routines.
12195 @subsection M32C Options
12196 @cindex M32C options
12199 @item -mcpu=@var{name}
12201 Select the CPU for which code is generated. @var{name} may be one of
12202 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12203 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12204 the M32C/80 series.
12208 Specifies that the program will be run on the simulator. This causes
12209 an alternate runtime library to be linked in which supports, for
12210 example, file I/O@. You must not use this option when generating
12211 programs that will run on real hardware; you must provide your own
12212 runtime library for whatever I/O functions are needed.
12214 @item -memregs=@var{number}
12216 Specifies the number of memory-based pseudo-registers GCC will use
12217 during code generation. These pseudo-registers will be used like real
12218 registers, so there is a tradeoff between GCC's ability to fit the
12219 code into available registers, and the performance penalty of using
12220 memory instead of registers. Note that all modules in a program must
12221 be compiled with the same value for this option. Because of that, you
12222 must not use this option with the default runtime libraries gcc
12227 @node M32R/D Options
12228 @subsection M32R/D Options
12229 @cindex M32R/D options
12231 These @option{-m} options are defined for Renesas M32R/D architectures:
12236 Generate code for the M32R/2@.
12240 Generate code for the M32R/X@.
12244 Generate code for the M32R@. This is the default.
12246 @item -mmodel=small
12247 @opindex mmodel=small
12248 Assume all objects live in the lower 16MB of memory (so that their addresses
12249 can be loaded with the @code{ld24} instruction), and assume all subroutines
12250 are reachable with the @code{bl} instruction.
12251 This is the default.
12253 The addressability of a particular object can be set with the
12254 @code{model} attribute.
12256 @item -mmodel=medium
12257 @opindex mmodel=medium
12258 Assume objects may be anywhere in the 32-bit address space (the compiler
12259 will generate @code{seth/add3} instructions to load their addresses), and
12260 assume all subroutines are reachable with the @code{bl} instruction.
12262 @item -mmodel=large
12263 @opindex mmodel=large
12264 Assume objects may be anywhere in the 32-bit address space (the compiler
12265 will generate @code{seth/add3} instructions to load their addresses), and
12266 assume subroutines may not be reachable with the @code{bl} instruction
12267 (the compiler will generate the much slower @code{seth/add3/jl}
12268 instruction sequence).
12271 @opindex msdata=none
12272 Disable use of the small data area. Variables will be put into
12273 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12274 @code{section} attribute has been specified).
12275 This is the default.
12277 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12278 Objects may be explicitly put in the small data area with the
12279 @code{section} attribute using one of these sections.
12281 @item -msdata=sdata
12282 @opindex msdata=sdata
12283 Put small global and static data in the small data area, but do not
12284 generate special code to reference them.
12287 @opindex msdata=use
12288 Put small global and static data in the small data area, and generate
12289 special instructions to reference them.
12293 @cindex smaller data references
12294 Put global and static objects less than or equal to @var{num} bytes
12295 into the small data or bss sections instead of the normal data or bss
12296 sections. The default value of @var{num} is 8.
12297 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12298 for this option to have any effect.
12300 All modules should be compiled with the same @option{-G @var{num}} value.
12301 Compiling with different values of @var{num} may or may not work; if it
12302 doesn't the linker will give an error message---incorrect code will not be
12307 Makes the M32R specific code in the compiler display some statistics
12308 that might help in debugging programs.
12310 @item -malign-loops
12311 @opindex malign-loops
12312 Align all loops to a 32-byte boundary.
12314 @item -mno-align-loops
12315 @opindex mno-align-loops
12316 Do not enforce a 32-byte alignment for loops. This is the default.
12318 @item -missue-rate=@var{number}
12319 @opindex missue-rate=@var{number}
12320 Issue @var{number} instructions per cycle. @var{number} can only be 1
12323 @item -mbranch-cost=@var{number}
12324 @opindex mbranch-cost=@var{number}
12325 @var{number} can only be 1 or 2. If it is 1 then branches will be
12326 preferred over conditional code, if it is 2, then the opposite will
12329 @item -mflush-trap=@var{number}
12330 @opindex mflush-trap=@var{number}
12331 Specifies the trap number to use to flush the cache. The default is
12332 12. Valid numbers are between 0 and 15 inclusive.
12334 @item -mno-flush-trap
12335 @opindex mno-flush-trap
12336 Specifies that the cache cannot be flushed by using a trap.
12338 @item -mflush-func=@var{name}
12339 @opindex mflush-func=@var{name}
12340 Specifies the name of the operating system function to call to flush
12341 the cache. The default is @emph{_flush_cache}, but a function call
12342 will only be used if a trap is not available.
12344 @item -mno-flush-func
12345 @opindex mno-flush-func
12346 Indicates that there is no OS function for flushing the cache.
12350 @node M680x0 Options
12351 @subsection M680x0 Options
12352 @cindex M680x0 options
12354 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12355 The default settings depend on which architecture was selected when
12356 the compiler was configured; the defaults for the most common choices
12360 @item -march=@var{arch}
12362 Generate code for a specific M680x0 or ColdFire instruction set
12363 architecture. Permissible values of @var{arch} for M680x0
12364 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12365 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12366 architectures are selected according to Freescale's ISA classification
12367 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12368 @samp{isab} and @samp{isac}.
12370 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12371 code for a ColdFire target. The @var{arch} in this macro is one of the
12372 @option{-march} arguments given above.
12374 When used together, @option{-march} and @option{-mtune} select code
12375 that runs on a family of similar processors but that is optimized
12376 for a particular microarchitecture.
12378 @item -mcpu=@var{cpu}
12380 Generate code for a specific M680x0 or ColdFire processor.
12381 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12382 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12383 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12384 below, which also classifies the CPUs into families:
12386 @multitable @columnfractions 0.20 0.80
12387 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12388 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12389 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12390 @item @samp{5206e} @tab @samp{5206e}
12391 @item @samp{5208} @tab @samp{5207} @samp{5208}
12392 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12393 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12394 @item @samp{5216} @tab @samp{5214} @samp{5216}
12395 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12396 @item @samp{5225} @tab @samp{5224} @samp{5225}
12397 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12398 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12399 @item @samp{5249} @tab @samp{5249}
12400 @item @samp{5250} @tab @samp{5250}
12401 @item @samp{5271} @tab @samp{5270} @samp{5271}
12402 @item @samp{5272} @tab @samp{5272}
12403 @item @samp{5275} @tab @samp{5274} @samp{5275}
12404 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12405 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12406 @item @samp{5307} @tab @samp{5307}
12407 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12408 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12409 @item @samp{5407} @tab @samp{5407}
12410 @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}
12413 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12414 @var{arch} is compatible with @var{cpu}. Other combinations of
12415 @option{-mcpu} and @option{-march} are rejected.
12417 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12418 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12419 where the value of @var{family} is given by the table above.
12421 @item -mtune=@var{tune}
12423 Tune the code for a particular microarchitecture, within the
12424 constraints set by @option{-march} and @option{-mcpu}.
12425 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12426 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12427 and @samp{cpu32}. The ColdFire microarchitectures
12428 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12430 You can also use @option{-mtune=68020-40} for code that needs
12431 to run relatively well on 68020, 68030 and 68040 targets.
12432 @option{-mtune=68020-60} is similar but includes 68060 targets
12433 as well. These two options select the same tuning decisions as
12434 @option{-m68020-40} and @option{-m68020-60} respectively.
12436 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12437 when tuning for 680x0 architecture @var{arch}. It also defines
12438 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12439 option is used. If gcc is tuning for a range of architectures,
12440 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12441 it defines the macros for every architecture in the range.
12443 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12444 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12445 of the arguments given above.
12451 Generate output for a 68000. This is the default
12452 when the compiler is configured for 68000-based systems.
12453 It is equivalent to @option{-march=68000}.
12455 Use this option for microcontrollers with a 68000 or EC000 core,
12456 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12460 Generate output for a 68010. This is the default
12461 when the compiler is configured for 68010-based systems.
12462 It is equivalent to @option{-march=68010}.
12468 Generate output for a 68020. This is the default
12469 when the compiler is configured for 68020-based systems.
12470 It is equivalent to @option{-march=68020}.
12474 Generate output for a 68030. This is the default when the compiler is
12475 configured for 68030-based systems. It is equivalent to
12476 @option{-march=68030}.
12480 Generate output for a 68040. This is the default when the compiler is
12481 configured for 68040-based systems. It is equivalent to
12482 @option{-march=68040}.
12484 This option inhibits the use of 68881/68882 instructions that have to be
12485 emulated by software on the 68040. Use this option if your 68040 does not
12486 have code to emulate those instructions.
12490 Generate output for a 68060. This is the default when the compiler is
12491 configured for 68060-based systems. It is equivalent to
12492 @option{-march=68060}.
12494 This option inhibits the use of 68020 and 68881/68882 instructions that
12495 have to be emulated by software on the 68060. Use this option if your 68060
12496 does not have code to emulate those instructions.
12500 Generate output for a CPU32. This is the default
12501 when the compiler is configured for CPU32-based systems.
12502 It is equivalent to @option{-march=cpu32}.
12504 Use this option for microcontrollers with a
12505 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12506 68336, 68340, 68341, 68349 and 68360.
12510 Generate output for a 520X ColdFire CPU@. This is the default
12511 when the compiler is configured for 520X-based systems.
12512 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12513 in favor of that option.
12515 Use this option for microcontroller with a 5200 core, including
12516 the MCF5202, MCF5203, MCF5204 and MCF5206.
12520 Generate output for a 5206e ColdFire CPU@. The option is now
12521 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12525 Generate output for a member of the ColdFire 528X family.
12526 The option is now deprecated in favor of the equivalent
12527 @option{-mcpu=528x}.
12531 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12532 in favor of the equivalent @option{-mcpu=5307}.
12536 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12537 in favor of the equivalent @option{-mcpu=5407}.
12541 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12542 This includes use of hardware floating point instructions.
12543 The option is equivalent to @option{-mcpu=547x}, and is now
12544 deprecated in favor of that option.
12548 Generate output for a 68040, without using any of the new instructions.
12549 This results in code which can run relatively efficiently on either a
12550 68020/68881 or a 68030 or a 68040. The generated code does use the
12551 68881 instructions that are emulated on the 68040.
12553 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12557 Generate output for a 68060, without using any of the new instructions.
12558 This results in code which can run relatively efficiently on either a
12559 68020/68881 or a 68030 or a 68040. The generated code does use the
12560 68881 instructions that are emulated on the 68060.
12562 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12566 @opindex mhard-float
12568 Generate floating-point instructions. This is the default for 68020
12569 and above, and for ColdFire devices that have an FPU@. It defines the
12570 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12571 on ColdFire targets.
12574 @opindex msoft-float
12575 Do not generate floating-point instructions; use library calls instead.
12576 This is the default for 68000, 68010, and 68832 targets. It is also
12577 the default for ColdFire devices that have no FPU.
12583 Generate (do not generate) ColdFire hardware divide and remainder
12584 instructions. If @option{-march} is used without @option{-mcpu},
12585 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12586 architectures. Otherwise, the default is taken from the target CPU
12587 (either the default CPU, or the one specified by @option{-mcpu}). For
12588 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12589 @option{-mcpu=5206e}.
12591 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12595 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12596 Additionally, parameters passed on the stack are also aligned to a
12597 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12601 Do not consider type @code{int} to be 16 bits wide. This is the default.
12604 @itemx -mno-bitfield
12605 @opindex mnobitfield
12606 @opindex mno-bitfield
12607 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12608 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12612 Do use the bit-field instructions. The @option{-m68020} option implies
12613 @option{-mbitfield}. This is the default if you use a configuration
12614 designed for a 68020.
12618 Use a different function-calling convention, in which functions
12619 that take a fixed number of arguments return with the @code{rtd}
12620 instruction, which pops their arguments while returning. This
12621 saves one instruction in the caller since there is no need to pop
12622 the arguments there.
12624 This calling convention is incompatible with the one normally
12625 used on Unix, so you cannot use it if you need to call libraries
12626 compiled with the Unix compiler.
12628 Also, you must provide function prototypes for all functions that
12629 take variable numbers of arguments (including @code{printf});
12630 otherwise incorrect code will be generated for calls to those
12633 In addition, seriously incorrect code will result if you call a
12634 function with too many arguments. (Normally, extra arguments are
12635 harmlessly ignored.)
12637 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12638 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12642 Do not use the calling conventions selected by @option{-mrtd}.
12643 This is the default.
12646 @itemx -mno-align-int
12647 @opindex malign-int
12648 @opindex mno-align-int
12649 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12650 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12651 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12652 Aligning variables on 32-bit boundaries produces code that runs somewhat
12653 faster on processors with 32-bit busses at the expense of more memory.
12655 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12656 align structures containing the above types differently than
12657 most published application binary interface specifications for the m68k.
12661 Use the pc-relative addressing mode of the 68000 directly, instead of
12662 using a global offset table. At present, this option implies @option{-fpic},
12663 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12664 not presently supported with @option{-mpcrel}, though this could be supported for
12665 68020 and higher processors.
12667 @item -mno-strict-align
12668 @itemx -mstrict-align
12669 @opindex mno-strict-align
12670 @opindex mstrict-align
12671 Do not (do) assume that unaligned memory references will be handled by
12675 Generate code that allows the data segment to be located in a different
12676 area of memory from the text segment. This allows for execute in place in
12677 an environment without virtual memory management. This option implies
12680 @item -mno-sep-data
12681 Generate code that assumes that the data segment follows the text segment.
12682 This is the default.
12684 @item -mid-shared-library
12685 Generate code that supports shared libraries via the library ID method.
12686 This allows for execute in place and shared libraries in an environment
12687 without virtual memory management. This option implies @option{-fPIC}.
12689 @item -mno-id-shared-library
12690 Generate code that doesn't assume ID based shared libraries are being used.
12691 This is the default.
12693 @item -mshared-library-id=n
12694 Specified the identification number of the ID based shared library being
12695 compiled. Specifying a value of 0 will generate more compact code, specifying
12696 other values will force the allocation of that number to the current
12697 library but is no more space or time efficient than omitting this option.
12703 When generating position-independent code for ColdFire, generate code
12704 that works if the GOT has more than 8192 entries. This code is
12705 larger and slower than code generated without this option. On M680x0
12706 processors, this option is not needed; @option{-fPIC} suffices.
12708 GCC normally uses a single instruction to load values from the GOT@.
12709 While this is relatively efficient, it only works if the GOT
12710 is smaller than about 64k. Anything larger causes the linker
12711 to report an error such as:
12713 @cindex relocation truncated to fit (ColdFire)
12715 relocation truncated to fit: R_68K_GOT16O foobar
12718 If this happens, you should recompile your code with @option{-mxgot}.
12719 It should then work with very large GOTs. However, code generated with
12720 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12721 the value of a global symbol.
12723 Note that some linkers, including newer versions of the GNU linker,
12724 can create multiple GOTs and sort GOT entries. If you have such a linker,
12725 you should only need to use @option{-mxgot} when compiling a single
12726 object file that accesses more than 8192 GOT entries. Very few do.
12728 These options have no effect unless GCC is generating
12729 position-independent code.
12733 @node M68hc1x Options
12734 @subsection M68hc1x Options
12735 @cindex M68hc1x options
12737 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12738 microcontrollers. The default values for these options depends on
12739 which style of microcontroller was selected when the compiler was configured;
12740 the defaults for the most common choices are given below.
12747 Generate output for a 68HC11. This is the default
12748 when the compiler is configured for 68HC11-based systems.
12754 Generate output for a 68HC12. This is the default
12755 when the compiler is configured for 68HC12-based systems.
12761 Generate output for a 68HCS12.
12763 @item -mauto-incdec
12764 @opindex mauto-incdec
12765 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12772 Enable the use of 68HC12 min and max instructions.
12775 @itemx -mno-long-calls
12776 @opindex mlong-calls
12777 @opindex mno-long-calls
12778 Treat all calls as being far away (near). If calls are assumed to be
12779 far away, the compiler will use the @code{call} instruction to
12780 call a function and the @code{rtc} instruction for returning.
12784 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12786 @item -msoft-reg-count=@var{count}
12787 @opindex msoft-reg-count
12788 Specify the number of pseudo-soft registers which are used for the
12789 code generation. The maximum number is 32. Using more pseudo-soft
12790 register may or may not result in better code depending on the program.
12791 The default is 4 for 68HC11 and 2 for 68HC12.
12795 @node MCore Options
12796 @subsection MCore Options
12797 @cindex MCore options
12799 These are the @samp{-m} options defined for the Motorola M*Core
12805 @itemx -mno-hardlit
12807 @opindex mno-hardlit
12808 Inline constants into the code stream if it can be done in two
12809 instructions or less.
12815 Use the divide instruction. (Enabled by default).
12817 @item -mrelax-immediate
12818 @itemx -mno-relax-immediate
12819 @opindex mrelax-immediate
12820 @opindex mno-relax-immediate
12821 Allow arbitrary sized immediates in bit operations.
12823 @item -mwide-bitfields
12824 @itemx -mno-wide-bitfields
12825 @opindex mwide-bitfields
12826 @opindex mno-wide-bitfields
12827 Always treat bit-fields as int-sized.
12829 @item -m4byte-functions
12830 @itemx -mno-4byte-functions
12831 @opindex m4byte-functions
12832 @opindex mno-4byte-functions
12833 Force all functions to be aligned to a four byte boundary.
12835 @item -mcallgraph-data
12836 @itemx -mno-callgraph-data
12837 @opindex mcallgraph-data
12838 @opindex mno-callgraph-data
12839 Emit callgraph information.
12842 @itemx -mno-slow-bytes
12843 @opindex mslow-bytes
12844 @opindex mno-slow-bytes
12845 Prefer word access when reading byte quantities.
12847 @item -mlittle-endian
12848 @itemx -mbig-endian
12849 @opindex mlittle-endian
12850 @opindex mbig-endian
12851 Generate code for a little endian target.
12857 Generate code for the 210 processor.
12861 Assume that run-time support has been provided and so omit the
12862 simulator library (@file{libsim.a)} from the linker command line.
12864 @item -mstack-increment=@var{size}
12865 @opindex mstack-increment
12866 Set the maximum amount for a single stack increment operation. Large
12867 values can increase the speed of programs which contain functions
12868 that need a large amount of stack space, but they can also trigger a
12869 segmentation fault if the stack is extended too much. The default
12875 @subsection MeP Options
12876 @cindex MeP options
12882 Enables the @code{abs} instruction, which is the absolute difference
12883 between two registers.
12887 Enables all the optional instructions - average, multiply, divide, bit
12888 operations, leading zero, absolute difference, min/max, clip, and
12894 Enables the @code{ave} instruction, which computes the average of two
12897 @item -mbased=@var{n}
12899 Variables of size @var{n} bytes or smaller will be placed in the
12900 @code{.based} section by default. Based variables use the @code{$tp}
12901 register as a base register, and there is a 128 byte limit to the
12902 @code{.based} section.
12906 Enables the bit operation instructions - bit test (@code{btstm}), set
12907 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
12908 test-and-set (@code{tas}).
12910 @item -mc=@var{name}
12912 Selects which section constant data will be placed in. @var{name} may
12913 be @code{tiny}, @code{near}, or @code{far}.
12917 Enables the @code{clip} instruction. Note that @code{-mclip} is not
12918 useful unless you also provide @code{-mminmax}.
12920 @item -mconfig=@var{name}
12922 Selects one of the build-in core configurations. Each MeP chip has
12923 one or more modules in it; each module has a core CPU and a variety of
12924 coprocessors, optional instructions, and peripherals. The
12925 @code{MeP-Integrator} tool, not part of GCC, provides these
12926 configurations through this option; using this option is the same as
12927 using all the corresponding command line options. The default
12928 configuration is @code{default}.
12932 Enables the coprocessor instructions. By default, this is a 32-bit
12933 coprocessor. Note that the coprocessor is normally enabled via the
12934 @code{-mconfig=} option.
12938 Enables the 32-bit coprocessor's instructions.
12942 Enables the 64-bit coprocessor's instructions.
12946 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
12950 Causes constant variables to be placed in the @code{.near} section.
12954 Enables the @code{div} and @code{divu} instructions.
12958 Generate big-endian code.
12962 Generate little-endian code.
12964 @item -mio-volatile
12965 @opindex mio-volatile
12966 Tells the compiler that any variable marked with the @code{io}
12967 attribute is to be considered volatile.
12971 Causes variables to be assigned to the @code{.far} section by default.
12975 Enables the @code{leadz} (leading zero) instruction.
12979 Causes variables to be assigned to the @code{.near} section by default.
12983 Enables the @code{min} and @code{max} instructions.
12987 Enables the multiplication and multiply-accumulate instructions.
12991 Disables all the optional instructions enabled by @code{-mall-opts}.
12995 Enables the @code{repeat} and @code{erepeat} instructions, used for
12996 low-overhead looping.
13000 Causes all variables to default to the @code{.tiny} section. Note
13001 that there is a 65536 byte limit to this section. Accesses to these
13002 variables use the @code{%gp} base register.
13006 Enables the saturation instructions. Note that the compiler does not
13007 currently generate these itself, but this option is included for
13008 compatibility with other tools, like @code{as}.
13012 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13016 Link the simulator runtime libraries.
13020 Link the simulator runtime libraries, excluding built-in support
13021 for reset and exception vectors and tables.
13025 Causes all functions to default to the @code{.far} section. Without
13026 this option, functions default to the @code{.near} section.
13028 @item -mtiny=@var{n}
13030 Variables that are @var{n} bytes or smaller will be allocated to the
13031 @code{.tiny} section. These variables use the @code{$gp} base
13032 register. The default for this option is 4, but note that there's a
13033 65536 byte limit to the @code{.tiny} section.
13038 @subsection MIPS Options
13039 @cindex MIPS options
13045 Generate big-endian code.
13049 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13052 @item -march=@var{arch}
13054 Generate code that will run on @var{arch}, which can be the name of a
13055 generic MIPS ISA, or the name of a particular processor.
13057 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13058 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13059 The processor names are:
13060 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13061 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13062 @samp{5kc}, @samp{5kf},
13064 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13065 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13066 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13067 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13068 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13069 @samp{loongson2e}, @samp{loongson2f},
13073 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13074 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13075 @samp{rm7000}, @samp{rm9000},
13076 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13079 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13080 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13082 The special value @samp{from-abi} selects the
13083 most compatible architecture for the selected ABI (that is,
13084 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13086 Native Linux/GNU toolchains also support the value @samp{native},
13087 which selects the best architecture option for the host processor.
13088 @option{-march=native} has no effect if GCC does not recognize
13091 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13092 (for example, @samp{-march=r2k}). Prefixes are optional, and
13093 @samp{vr} may be written @samp{r}.
13095 Names of the form @samp{@var{n}f2_1} refer to processors with
13096 FPUs clocked at half the rate of the core, names of the form
13097 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13098 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13099 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13100 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13101 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13102 accepted as synonyms for @samp{@var{n}f1_1}.
13104 GCC defines two macros based on the value of this option. The first
13105 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13106 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13107 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13108 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13109 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13111 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13112 above. In other words, it will have the full prefix and will not
13113 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13114 the macro names the resolved architecture (either @samp{"mips1"} or
13115 @samp{"mips3"}). It names the default architecture when no
13116 @option{-march} option is given.
13118 @item -mtune=@var{arch}
13120 Optimize for @var{arch}. Among other things, this option controls
13121 the way instructions are scheduled, and the perceived cost of arithmetic
13122 operations. The list of @var{arch} values is the same as for
13125 When this option is not used, GCC will optimize for the processor
13126 specified by @option{-march}. By using @option{-march} and
13127 @option{-mtune} together, it is possible to generate code that will
13128 run on a family of processors, but optimize the code for one
13129 particular member of that family.
13131 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13132 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13133 @samp{-march} ones described above.
13137 Equivalent to @samp{-march=mips1}.
13141 Equivalent to @samp{-march=mips2}.
13145 Equivalent to @samp{-march=mips3}.
13149 Equivalent to @samp{-march=mips4}.
13153 Equivalent to @samp{-march=mips32}.
13157 Equivalent to @samp{-march=mips32r2}.
13161 Equivalent to @samp{-march=mips64}.
13165 Equivalent to @samp{-march=mips64r2}.
13170 @opindex mno-mips16
13171 Generate (do not generate) MIPS16 code. If GCC is targetting a
13172 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13174 MIPS16 code generation can also be controlled on a per-function basis
13175 by means of @code{mips16} and @code{nomips16} attributes.
13176 @xref{Function Attributes}, for more information.
13178 @item -mflip-mips16
13179 @opindex mflip-mips16
13180 Generate MIPS16 code on alternating functions. This option is provided
13181 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13182 not intended for ordinary use in compiling user code.
13184 @item -minterlink-mips16
13185 @itemx -mno-interlink-mips16
13186 @opindex minterlink-mips16
13187 @opindex mno-interlink-mips16
13188 Require (do not require) that non-MIPS16 code be link-compatible with
13191 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13192 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13193 therefore disables direct jumps unless GCC knows that the target of the
13194 jump is not MIPS16.
13206 Generate code for the given ABI@.
13208 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13209 generates 64-bit code when you select a 64-bit architecture, but you
13210 can use @option{-mgp32} to get 32-bit code instead.
13212 For information about the O64 ABI, see
13213 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13215 GCC supports a variant of the o32 ABI in which floating-point registers
13216 are 64 rather than 32 bits wide. You can select this combination with
13217 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13218 and @samp{mfhc1} instructions and is therefore only supported for
13219 MIPS32R2 processors.
13221 The register assignments for arguments and return values remain the
13222 same, but each scalar value is passed in a single 64-bit register
13223 rather than a pair of 32-bit registers. For example, scalar
13224 floating-point values are returned in @samp{$f0} only, not a
13225 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13226 remains the same, but all 64 bits are saved.
13229 @itemx -mno-abicalls
13231 @opindex mno-abicalls
13232 Generate (do not generate) code that is suitable for SVR4-style
13233 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13238 Generate (do not generate) code that is fully position-independent,
13239 and that can therefore be linked into shared libraries. This option
13240 only affects @option{-mabicalls}.
13242 All @option{-mabicalls} code has traditionally been position-independent,
13243 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13244 as an extension, the GNU toolchain allows executables to use absolute
13245 accesses for locally-binding symbols. It can also use shorter GP
13246 initialization sequences and generate direct calls to locally-defined
13247 functions. This mode is selected by @option{-mno-shared}.
13249 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13250 objects that can only be linked by the GNU linker. However, the option
13251 does not affect the ABI of the final executable; it only affects the ABI
13252 of relocatable objects. Using @option{-mno-shared} will generally make
13253 executables both smaller and quicker.
13255 @option{-mshared} is the default.
13261 Assume (do not assume) that the static and dynamic linkers
13262 support PLTs and copy relocations. This option only affects
13263 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13264 has no effect without @samp{-msym32}.
13266 You can make @option{-mplt} the default by configuring
13267 GCC with @option{--with-mips-plt}. The default is
13268 @option{-mno-plt} otherwise.
13274 Lift (do not lift) the usual restrictions on the size of the global
13277 GCC normally uses a single instruction to load values from the GOT@.
13278 While this is relatively efficient, it will only work if the GOT
13279 is smaller than about 64k. Anything larger will cause the linker
13280 to report an error such as:
13282 @cindex relocation truncated to fit (MIPS)
13284 relocation truncated to fit: R_MIPS_GOT16 foobar
13287 If this happens, you should recompile your code with @option{-mxgot}.
13288 It should then work with very large GOTs, although it will also be
13289 less efficient, since it will take three instructions to fetch the
13290 value of a global symbol.
13292 Note that some linkers can create multiple GOTs. If you have such a
13293 linker, you should only need to use @option{-mxgot} when a single object
13294 file accesses more than 64k's worth of GOT entries. Very few do.
13296 These options have no effect unless GCC is generating position
13301 Assume that general-purpose registers are 32 bits wide.
13305 Assume that general-purpose registers are 64 bits wide.
13309 Assume that floating-point registers are 32 bits wide.
13313 Assume that floating-point registers are 64 bits wide.
13316 @opindex mhard-float
13317 Use floating-point coprocessor instructions.
13320 @opindex msoft-float
13321 Do not use floating-point coprocessor instructions. Implement
13322 floating-point calculations using library calls instead.
13324 @item -msingle-float
13325 @opindex msingle-float
13326 Assume that the floating-point coprocessor only supports single-precision
13329 @item -mdouble-float
13330 @opindex mdouble-float
13331 Assume that the floating-point coprocessor supports double-precision
13332 operations. This is the default.
13338 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13339 implement atomic memory built-in functions. When neither option is
13340 specified, GCC will use the instructions if the target architecture
13343 @option{-mllsc} is useful if the runtime environment can emulate the
13344 instructions and @option{-mno-llsc} can be useful when compiling for
13345 nonstandard ISAs. You can make either option the default by
13346 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13347 respectively. @option{--with-llsc} is the default for some
13348 configurations; see the installation documentation for details.
13354 Use (do not use) revision 1 of the MIPS DSP ASE@.
13355 @xref{MIPS DSP Built-in Functions}. This option defines the
13356 preprocessor macro @samp{__mips_dsp}. It also defines
13357 @samp{__mips_dsp_rev} to 1.
13363 Use (do not use) revision 2 of the MIPS DSP ASE@.
13364 @xref{MIPS DSP Built-in Functions}. This option defines the
13365 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13366 It also defines @samp{__mips_dsp_rev} to 2.
13369 @itemx -mno-smartmips
13370 @opindex msmartmips
13371 @opindex mno-smartmips
13372 Use (do not use) the MIPS SmartMIPS ASE.
13374 @item -mpaired-single
13375 @itemx -mno-paired-single
13376 @opindex mpaired-single
13377 @opindex mno-paired-single
13378 Use (do not use) paired-single floating-point instructions.
13379 @xref{MIPS Paired-Single Support}. This option requires
13380 hardware floating-point support to be enabled.
13386 Use (do not use) MIPS Digital Media Extension instructions.
13387 This option can only be used when generating 64-bit code and requires
13388 hardware floating-point support to be enabled.
13393 @opindex mno-mips3d
13394 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13395 The option @option{-mips3d} implies @option{-mpaired-single}.
13401 Use (do not use) MT Multithreading instructions.
13405 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13406 an explanation of the default and the way that the pointer size is
13411 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13413 The default size of @code{int}s, @code{long}s and pointers depends on
13414 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13415 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13416 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13417 or the same size as integer registers, whichever is smaller.
13423 Assume (do not assume) that all symbols have 32-bit values, regardless
13424 of the selected ABI@. This option is useful in combination with
13425 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13426 to generate shorter and faster references to symbolic addresses.
13430 Put definitions of externally-visible data in a small data section
13431 if that data is no bigger than @var{num} bytes. GCC can then access
13432 the data more efficiently; see @option{-mgpopt} for details.
13434 The default @option{-G} option depends on the configuration.
13436 @item -mlocal-sdata
13437 @itemx -mno-local-sdata
13438 @opindex mlocal-sdata
13439 @opindex mno-local-sdata
13440 Extend (do not extend) the @option{-G} behavior to local data too,
13441 such as to static variables in C@. @option{-mlocal-sdata} is the
13442 default for all configurations.
13444 If the linker complains that an application is using too much small data,
13445 you might want to try rebuilding the less performance-critical parts with
13446 @option{-mno-local-sdata}. You might also want to build large
13447 libraries with @option{-mno-local-sdata}, so that the libraries leave
13448 more room for the main program.
13450 @item -mextern-sdata
13451 @itemx -mno-extern-sdata
13452 @opindex mextern-sdata
13453 @opindex mno-extern-sdata
13454 Assume (do not assume) that externally-defined data will be in
13455 a small data section if that data is within the @option{-G} limit.
13456 @option{-mextern-sdata} is the default for all configurations.
13458 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13459 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13460 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13461 is placed in a small data section. If @var{Var} is defined by another
13462 module, you must either compile that module with a high-enough
13463 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13464 definition. If @var{Var} is common, you must link the application
13465 with a high-enough @option{-G} setting.
13467 The easiest way of satisfying these restrictions is to compile
13468 and link every module with the same @option{-G} option. However,
13469 you may wish to build a library that supports several different
13470 small data limits. You can do this by compiling the library with
13471 the highest supported @option{-G} setting and additionally using
13472 @option{-mno-extern-sdata} to stop the library from making assumptions
13473 about externally-defined data.
13479 Use (do not use) GP-relative accesses for symbols that are known to be
13480 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13481 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13484 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13485 might not hold the value of @code{_gp}. For example, if the code is
13486 part of a library that might be used in a boot monitor, programs that
13487 call boot monitor routines will pass an unknown value in @code{$gp}.
13488 (In such situations, the boot monitor itself would usually be compiled
13489 with @option{-G0}.)
13491 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13492 @option{-mno-extern-sdata}.
13494 @item -membedded-data
13495 @itemx -mno-embedded-data
13496 @opindex membedded-data
13497 @opindex mno-embedded-data
13498 Allocate variables to the read-only data section first if possible, then
13499 next in the small data section if possible, otherwise in data. This gives
13500 slightly slower code than the default, but reduces the amount of RAM required
13501 when executing, and thus may be preferred for some embedded systems.
13503 @item -muninit-const-in-rodata
13504 @itemx -mno-uninit-const-in-rodata
13505 @opindex muninit-const-in-rodata
13506 @opindex mno-uninit-const-in-rodata
13507 Put uninitialized @code{const} variables in the read-only data section.
13508 This option is only meaningful in conjunction with @option{-membedded-data}.
13510 @item -mcode-readable=@var{setting}
13511 @opindex mcode-readable
13512 Specify whether GCC may generate code that reads from executable sections.
13513 There are three possible settings:
13516 @item -mcode-readable=yes
13517 Instructions may freely access executable sections. This is the
13520 @item -mcode-readable=pcrel
13521 MIPS16 PC-relative load instructions can access executable sections,
13522 but other instructions must not do so. This option is useful on 4KSc
13523 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13524 It is also useful on processors that can be configured to have a dual
13525 instruction/data SRAM interface and that, like the M4K, automatically
13526 redirect PC-relative loads to the instruction RAM.
13528 @item -mcode-readable=no
13529 Instructions must not access executable sections. This option can be
13530 useful on targets that are configured to have a dual instruction/data
13531 SRAM interface but that (unlike the M4K) do not automatically redirect
13532 PC-relative loads to the instruction RAM.
13535 @item -msplit-addresses
13536 @itemx -mno-split-addresses
13537 @opindex msplit-addresses
13538 @opindex mno-split-addresses
13539 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13540 relocation operators. This option has been superseded by
13541 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13543 @item -mexplicit-relocs
13544 @itemx -mno-explicit-relocs
13545 @opindex mexplicit-relocs
13546 @opindex mno-explicit-relocs
13547 Use (do not use) assembler relocation operators when dealing with symbolic
13548 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13549 is to use assembler macros instead.
13551 @option{-mexplicit-relocs} is the default if GCC was configured
13552 to use an assembler that supports relocation operators.
13554 @item -mcheck-zero-division
13555 @itemx -mno-check-zero-division
13556 @opindex mcheck-zero-division
13557 @opindex mno-check-zero-division
13558 Trap (do not trap) on integer division by zero.
13560 The default is @option{-mcheck-zero-division}.
13562 @item -mdivide-traps
13563 @itemx -mdivide-breaks
13564 @opindex mdivide-traps
13565 @opindex mdivide-breaks
13566 MIPS systems check for division by zero by generating either a
13567 conditional trap or a break instruction. Using traps results in
13568 smaller code, but is only supported on MIPS II and later. Also, some
13569 versions of the Linux kernel have a bug that prevents trap from
13570 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13571 allow conditional traps on architectures that support them and
13572 @option{-mdivide-breaks} to force the use of breaks.
13574 The default is usually @option{-mdivide-traps}, but this can be
13575 overridden at configure time using @option{--with-divide=breaks}.
13576 Divide-by-zero checks can be completely disabled using
13577 @option{-mno-check-zero-division}.
13582 @opindex mno-memcpy
13583 Force (do not force) the use of @code{memcpy()} for non-trivial block
13584 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13585 most constant-sized copies.
13588 @itemx -mno-long-calls
13589 @opindex mlong-calls
13590 @opindex mno-long-calls
13591 Disable (do not disable) use of the @code{jal} instruction. Calling
13592 functions using @code{jal} is more efficient but requires the caller
13593 and callee to be in the same 256 megabyte segment.
13595 This option has no effect on abicalls code. The default is
13596 @option{-mno-long-calls}.
13602 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13603 instructions, as provided by the R4650 ISA@.
13606 @itemx -mno-fused-madd
13607 @opindex mfused-madd
13608 @opindex mno-fused-madd
13609 Enable (disable) use of the floating point multiply-accumulate
13610 instructions, when they are available. The default is
13611 @option{-mfused-madd}.
13613 When multiply-accumulate instructions are used, the intermediate
13614 product is calculated to infinite precision and is not subject to
13615 the FCSR Flush to Zero bit. This may be undesirable in some
13620 Tell the MIPS assembler to not run its preprocessor over user
13621 assembler files (with a @samp{.s} suffix) when assembling them.
13624 @itemx -mno-fix-r4000
13625 @opindex mfix-r4000
13626 @opindex mno-fix-r4000
13627 Work around certain R4000 CPU errata:
13630 A double-word or a variable shift may give an incorrect result if executed
13631 immediately after starting an integer division.
13633 A double-word or a variable shift may give an incorrect result if executed
13634 while an integer multiplication is in progress.
13636 An integer division may give an incorrect result if started in a delay slot
13637 of a taken branch or a jump.
13641 @itemx -mno-fix-r4400
13642 @opindex mfix-r4400
13643 @opindex mno-fix-r4400
13644 Work around certain R4400 CPU errata:
13647 A double-word or a variable shift may give an incorrect result if executed
13648 immediately after starting an integer division.
13652 @itemx -mno-fix-r10000
13653 @opindex mfix-r10000
13654 @opindex mno-fix-r10000
13655 Work around certain R10000 errata:
13658 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13659 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13662 This option can only be used if the target architecture supports
13663 branch-likely instructions. @option{-mfix-r10000} is the default when
13664 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13668 @itemx -mno-fix-vr4120
13669 @opindex mfix-vr4120
13670 Work around certain VR4120 errata:
13673 @code{dmultu} does not always produce the correct result.
13675 @code{div} and @code{ddiv} do not always produce the correct result if one
13676 of the operands is negative.
13678 The workarounds for the division errata rely on special functions in
13679 @file{libgcc.a}. At present, these functions are only provided by
13680 the @code{mips64vr*-elf} configurations.
13682 Other VR4120 errata require a nop to be inserted between certain pairs of
13683 instructions. These errata are handled by the assembler, not by GCC itself.
13686 @opindex mfix-vr4130
13687 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13688 workarounds are implemented by the assembler rather than by GCC,
13689 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13690 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13691 instructions are available instead.
13694 @itemx -mno-fix-sb1
13696 Work around certain SB-1 CPU core errata.
13697 (This flag currently works around the SB-1 revision 2
13698 ``F1'' and ``F2'' floating point errata.)
13700 @item -mr10k-cache-barrier=@var{setting}
13701 @opindex mr10k-cache-barrier
13702 Specify whether GCC should insert cache barriers to avoid the
13703 side-effects of speculation on R10K processors.
13705 In common with many processors, the R10K tries to predict the outcome
13706 of a conditional branch and speculatively executes instructions from
13707 the ``taken'' branch. It later aborts these instructions if the
13708 predicted outcome was wrong. However, on the R10K, even aborted
13709 instructions can have side effects.
13711 This problem only affects kernel stores and, depending on the system,
13712 kernel loads. As an example, a speculatively-executed store may load
13713 the target memory into cache and mark the cache line as dirty, even if
13714 the store itself is later aborted. If a DMA operation writes to the
13715 same area of memory before the ``dirty'' line is flushed, the cached
13716 data will overwrite the DMA-ed data. See the R10K processor manual
13717 for a full description, including other potential problems.
13719 One workaround is to insert cache barrier instructions before every memory
13720 access that might be speculatively executed and that might have side
13721 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13722 controls GCC's implementation of this workaround. It assumes that
13723 aborted accesses to any byte in the following regions will not have
13728 the memory occupied by the current function's stack frame;
13731 the memory occupied by an incoming stack argument;
13734 the memory occupied by an object with a link-time-constant address.
13737 It is the kernel's responsibility to ensure that speculative
13738 accesses to these regions are indeed safe.
13740 If the input program contains a function declaration such as:
13746 then the implementation of @code{foo} must allow @code{j foo} and
13747 @code{jal foo} to be executed speculatively. GCC honors this
13748 restriction for functions it compiles itself. It expects non-GCC
13749 functions (such as hand-written assembly code) to do the same.
13751 The option has three forms:
13754 @item -mr10k-cache-barrier=load-store
13755 Insert a cache barrier before a load or store that might be
13756 speculatively executed and that might have side effects even
13759 @item -mr10k-cache-barrier=store
13760 Insert a cache barrier before a store that might be speculatively
13761 executed and that might have side effects even if aborted.
13763 @item -mr10k-cache-barrier=none
13764 Disable the insertion of cache barriers. This is the default setting.
13767 @item -mflush-func=@var{func}
13768 @itemx -mno-flush-func
13769 @opindex mflush-func
13770 Specifies the function to call to flush the I and D caches, or to not
13771 call any such function. If called, the function must take the same
13772 arguments as the common @code{_flush_func()}, that is, the address of the
13773 memory range for which the cache is being flushed, the size of the
13774 memory range, and the number 3 (to flush both caches). The default
13775 depends on the target GCC was configured for, but commonly is either
13776 @samp{_flush_func} or @samp{__cpu_flush}.
13778 @item mbranch-cost=@var{num}
13779 @opindex mbranch-cost
13780 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13781 This cost is only a heuristic and is not guaranteed to produce
13782 consistent results across releases. A zero cost redundantly selects
13783 the default, which is based on the @option{-mtune} setting.
13785 @item -mbranch-likely
13786 @itemx -mno-branch-likely
13787 @opindex mbranch-likely
13788 @opindex mno-branch-likely
13789 Enable or disable use of Branch Likely instructions, regardless of the
13790 default for the selected architecture. By default, Branch Likely
13791 instructions may be generated if they are supported by the selected
13792 architecture. An exception is for the MIPS32 and MIPS64 architectures
13793 and processors which implement those architectures; for those, Branch
13794 Likely instructions will not be generated by default because the MIPS32
13795 and MIPS64 architectures specifically deprecate their use.
13797 @item -mfp-exceptions
13798 @itemx -mno-fp-exceptions
13799 @opindex mfp-exceptions
13800 Specifies whether FP exceptions are enabled. This affects how we schedule
13801 FP instructions for some processors. The default is that FP exceptions are
13804 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13805 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13808 @item -mvr4130-align
13809 @itemx -mno-vr4130-align
13810 @opindex mvr4130-align
13811 The VR4130 pipeline is two-way superscalar, but can only issue two
13812 instructions together if the first one is 8-byte aligned. When this
13813 option is enabled, GCC will align pairs of instructions that it
13814 thinks should execute in parallel.
13816 This option only has an effect when optimizing for the VR4130.
13817 It normally makes code faster, but at the expense of making it bigger.
13818 It is enabled by default at optimization level @option{-O3}.
13823 Enable (disable) generation of @code{synci} instructions on
13824 architectures that support it. The @code{synci} instructions (if
13825 enabled) will be generated when @code{__builtin___clear_cache()} is
13828 This option defaults to @code{-mno-synci}, but the default can be
13829 overridden by configuring with @code{--with-synci}.
13831 When compiling code for single processor systems, it is generally safe
13832 to use @code{synci}. However, on many multi-core (SMP) systems, it
13833 will not invalidate the instruction caches on all cores and may lead
13834 to undefined behavior.
13838 @subsection MMIX Options
13839 @cindex MMIX Options
13841 These options are defined for the MMIX:
13845 @itemx -mno-libfuncs
13847 @opindex mno-libfuncs
13848 Specify that intrinsic library functions are being compiled, passing all
13849 values in registers, no matter the size.
13852 @itemx -mno-epsilon
13854 @opindex mno-epsilon
13855 Generate floating-point comparison instructions that compare with respect
13856 to the @code{rE} epsilon register.
13858 @item -mabi=mmixware
13860 @opindex mabi=mmixware
13862 Generate code that passes function parameters and return values that (in
13863 the called function) are seen as registers @code{$0} and up, as opposed to
13864 the GNU ABI which uses global registers @code{$231} and up.
13866 @item -mzero-extend
13867 @itemx -mno-zero-extend
13868 @opindex mzero-extend
13869 @opindex mno-zero-extend
13870 When reading data from memory in sizes shorter than 64 bits, use (do not
13871 use) zero-extending load instructions by default, rather than
13872 sign-extending ones.
13875 @itemx -mno-knuthdiv
13877 @opindex mno-knuthdiv
13878 Make the result of a division yielding a remainder have the same sign as
13879 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13880 remainder follows the sign of the dividend. Both methods are
13881 arithmetically valid, the latter being almost exclusively used.
13883 @item -mtoplevel-symbols
13884 @itemx -mno-toplevel-symbols
13885 @opindex mtoplevel-symbols
13886 @opindex mno-toplevel-symbols
13887 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13888 code can be used with the @code{PREFIX} assembly directive.
13892 Generate an executable in the ELF format, rather than the default
13893 @samp{mmo} format used by the @command{mmix} simulator.
13895 @item -mbranch-predict
13896 @itemx -mno-branch-predict
13897 @opindex mbranch-predict
13898 @opindex mno-branch-predict
13899 Use (do not use) the probable-branch instructions, when static branch
13900 prediction indicates a probable branch.
13902 @item -mbase-addresses
13903 @itemx -mno-base-addresses
13904 @opindex mbase-addresses
13905 @opindex mno-base-addresses
13906 Generate (do not generate) code that uses @emph{base addresses}. Using a
13907 base address automatically generates a request (handled by the assembler
13908 and the linker) for a constant to be set up in a global register. The
13909 register is used for one or more base address requests within the range 0
13910 to 255 from the value held in the register. The generally leads to short
13911 and fast code, but the number of different data items that can be
13912 addressed is limited. This means that a program that uses lots of static
13913 data may require @option{-mno-base-addresses}.
13915 @item -msingle-exit
13916 @itemx -mno-single-exit
13917 @opindex msingle-exit
13918 @opindex mno-single-exit
13919 Force (do not force) generated code to have a single exit point in each
13923 @node MN10300 Options
13924 @subsection MN10300 Options
13925 @cindex MN10300 options
13927 These @option{-m} options are defined for Matsushita MN10300 architectures:
13932 Generate code to avoid bugs in the multiply instructions for the MN10300
13933 processors. This is the default.
13935 @item -mno-mult-bug
13936 @opindex mno-mult-bug
13937 Do not generate code to avoid bugs in the multiply instructions for the
13938 MN10300 processors.
13942 Generate code which uses features specific to the AM33 processor.
13946 Do not generate code which uses features specific to the AM33 processor. This
13949 @item -mreturn-pointer-on-d0
13950 @opindex mreturn-pointer-on-d0
13951 When generating a function which returns a pointer, return the pointer
13952 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13953 only in a0, and attempts to call such functions without a prototype
13954 would result in errors. Note that this option is on by default; use
13955 @option{-mno-return-pointer-on-d0} to disable it.
13959 Do not link in the C run-time initialization object file.
13963 Indicate to the linker that it should perform a relaxation optimization pass
13964 to shorten branches, calls and absolute memory addresses. This option only
13965 has an effect when used on the command line for the final link step.
13967 This option makes symbolic debugging impossible.
13970 @node PDP-11 Options
13971 @subsection PDP-11 Options
13972 @cindex PDP-11 Options
13974 These options are defined for the PDP-11:
13979 Use hardware FPP floating point. This is the default. (FIS floating
13980 point on the PDP-11/40 is not supported.)
13983 @opindex msoft-float
13984 Do not use hardware floating point.
13988 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13992 Return floating-point results in memory. This is the default.
13996 Generate code for a PDP-11/40.
14000 Generate code for a PDP-11/45. This is the default.
14004 Generate code for a PDP-11/10.
14006 @item -mbcopy-builtin
14007 @opindex mbcopy-builtin
14008 Use inline @code{movmemhi} patterns for copying memory. This is the
14013 Do not use inline @code{movmemhi} patterns for copying memory.
14019 Use 16-bit @code{int}. This is the default.
14025 Use 32-bit @code{int}.
14028 @itemx -mno-float32
14030 @opindex mno-float32
14031 Use 64-bit @code{float}. This is the default.
14034 @itemx -mno-float64
14036 @opindex mno-float64
14037 Use 32-bit @code{float}.
14041 Use @code{abshi2} pattern. This is the default.
14045 Do not use @code{abshi2} pattern.
14047 @item -mbranch-expensive
14048 @opindex mbranch-expensive
14049 Pretend that branches are expensive. This is for experimenting with
14050 code generation only.
14052 @item -mbranch-cheap
14053 @opindex mbranch-cheap
14054 Do not pretend that branches are expensive. This is the default.
14058 Generate code for a system with split I&D@.
14062 Generate code for a system without split I&D@. This is the default.
14066 Use Unix assembler syntax. This is the default when configured for
14067 @samp{pdp11-*-bsd}.
14071 Use DEC assembler syntax. This is the default when configured for any
14072 PDP-11 target other than @samp{pdp11-*-bsd}.
14075 @node picoChip Options
14076 @subsection picoChip Options
14077 @cindex picoChip options
14079 These @samp{-m} options are defined for picoChip implementations:
14083 @item -mae=@var{ae_type}
14085 Set the instruction set, register set, and instruction scheduling
14086 parameters for array element type @var{ae_type}. Supported values
14087 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14089 @option{-mae=ANY} selects a completely generic AE type. Code
14090 generated with this option will run on any of the other AE types. The
14091 code will not be as efficient as it would be if compiled for a specific
14092 AE type, and some types of operation (e.g., multiplication) will not
14093 work properly on all types of AE.
14095 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14096 for compiled code, and is the default.
14098 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14099 option may suffer from poor performance of byte (char) manipulation,
14100 since the DSP AE does not provide hardware support for byte load/stores.
14102 @item -msymbol-as-address
14103 Enable the compiler to directly use a symbol name as an address in a
14104 load/store instruction, without first loading it into a
14105 register. Typically, the use of this option will generate larger
14106 programs, which run faster than when the option isn't used. However, the
14107 results vary from program to program, so it is left as a user option,
14108 rather than being permanently enabled.
14110 @item -mno-inefficient-warnings
14111 Disables warnings about the generation of inefficient code. These
14112 warnings can be generated, for example, when compiling code which
14113 performs byte-level memory operations on the MAC AE type. The MAC AE has
14114 no hardware support for byte-level memory operations, so all byte
14115 load/stores must be synthesized from word load/store operations. This is
14116 inefficient and a warning will be generated indicating to the programmer
14117 that they should rewrite the code to avoid byte operations, or to target
14118 an AE type which has the necessary hardware support. This option enables
14119 the warning to be turned off.
14123 @node PowerPC Options
14124 @subsection PowerPC Options
14125 @cindex PowerPC options
14127 These are listed under @xref{RS/6000 and PowerPC Options}.
14129 @node RS/6000 and PowerPC Options
14130 @subsection IBM RS/6000 and PowerPC Options
14131 @cindex RS/6000 and PowerPC Options
14132 @cindex IBM RS/6000 and PowerPC Options
14134 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14141 @itemx -mno-powerpc
14142 @itemx -mpowerpc-gpopt
14143 @itemx -mno-powerpc-gpopt
14144 @itemx -mpowerpc-gfxopt
14145 @itemx -mno-powerpc-gfxopt
14147 @itemx -mno-powerpc64
14151 @itemx -mno-popcntb
14153 @itemx -mno-popcntd
14161 @itemx -mno-hard-dfp
14165 @opindex mno-power2
14167 @opindex mno-powerpc
14168 @opindex mpowerpc-gpopt
14169 @opindex mno-powerpc-gpopt
14170 @opindex mpowerpc-gfxopt
14171 @opindex mno-powerpc-gfxopt
14172 @opindex mpowerpc64
14173 @opindex mno-powerpc64
14177 @opindex mno-popcntb
14179 @opindex mno-popcntd
14185 @opindex mno-mfpgpr
14187 @opindex mno-hard-dfp
14188 GCC supports two related instruction set architectures for the
14189 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14190 instructions supported by the @samp{rios} chip set used in the original
14191 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14192 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14193 the IBM 4xx, 6xx, and follow-on microprocessors.
14195 Neither architecture is a subset of the other. However there is a
14196 large common subset of instructions supported by both. An MQ
14197 register is included in processors supporting the POWER architecture.
14199 You use these options to specify which instructions are available on the
14200 processor you are using. The default value of these options is
14201 determined when configuring GCC@. Specifying the
14202 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14203 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14204 rather than the options listed above.
14206 The @option{-mpower} option allows GCC to generate instructions that
14207 are found only in the POWER architecture and to use the MQ register.
14208 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14209 to generate instructions that are present in the POWER2 architecture but
14210 not the original POWER architecture.
14212 The @option{-mpowerpc} option allows GCC to generate instructions that
14213 are found only in the 32-bit subset of the PowerPC architecture.
14214 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14215 GCC to use the optional PowerPC architecture instructions in the
14216 General Purpose group, including floating-point square root. Specifying
14217 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14218 use the optional PowerPC architecture instructions in the Graphics
14219 group, including floating-point select.
14221 The @option{-mmfcrf} option allows GCC to generate the move from
14222 condition register field instruction implemented on the POWER4
14223 processor and other processors that support the PowerPC V2.01
14225 The @option{-mpopcntb} option allows GCC to generate the popcount and
14226 double precision FP reciprocal estimate instruction implemented on the
14227 POWER5 processor and other processors that support the PowerPC V2.02
14229 The @option{-mpopcntd} option allows GCC to generate the popcount
14230 instruction implemented on the POWER7 processor and other processors
14231 that support the PowerPC V2.06 architecture.
14232 The @option{-mfprnd} option allows GCC to generate the FP round to
14233 integer instructions implemented on the POWER5+ processor and other
14234 processors that support the PowerPC V2.03 architecture.
14235 The @option{-mcmpb} option allows GCC to generate the compare bytes
14236 instruction implemented on the POWER6 processor and other processors
14237 that support the PowerPC V2.05 architecture.
14238 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14239 general purpose register instructions implemented on the POWER6X
14240 processor and other processors that support the extended PowerPC V2.05
14242 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14243 point instructions implemented on some POWER processors.
14245 The @option{-mpowerpc64} option allows GCC to generate the additional
14246 64-bit instructions that are found in the full PowerPC64 architecture
14247 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14248 @option{-mno-powerpc64}.
14250 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14251 will use only the instructions in the common subset of both
14252 architectures plus some special AIX common-mode calls, and will not use
14253 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14254 permits GCC to use any instruction from either architecture and to
14255 allow use of the MQ register; specify this for the Motorola MPC601.
14257 @item -mnew-mnemonics
14258 @itemx -mold-mnemonics
14259 @opindex mnew-mnemonics
14260 @opindex mold-mnemonics
14261 Select which mnemonics to use in the generated assembler code. With
14262 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14263 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14264 assembler mnemonics defined for the POWER architecture. Instructions
14265 defined in only one architecture have only one mnemonic; GCC uses that
14266 mnemonic irrespective of which of these options is specified.
14268 GCC defaults to the mnemonics appropriate for the architecture in
14269 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14270 value of these option. Unless you are building a cross-compiler, you
14271 should normally not specify either @option{-mnew-mnemonics} or
14272 @option{-mold-mnemonics}, but should instead accept the default.
14274 @item -mcpu=@var{cpu_type}
14276 Set architecture type, register usage, choice of mnemonics, and
14277 instruction scheduling parameters for machine type @var{cpu_type}.
14278 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14279 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14280 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
14281 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
14282 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14283 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
14284 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14285 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14286 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
14287 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14288 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14290 @option{-mcpu=common} selects a completely generic processor. Code
14291 generated under this option will run on any POWER or PowerPC processor.
14292 GCC will use only the instructions in the common subset of both
14293 architectures, and will not use the MQ register. GCC assumes a generic
14294 processor model for scheduling purposes.
14296 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14297 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14298 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14299 types, with an appropriate, generic processor model assumed for
14300 scheduling purposes.
14302 The other options specify a specific processor. Code generated under
14303 those options will run best on that processor, and may not run at all on
14306 The @option{-mcpu} options automatically enable or disable the
14309 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14310 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14311 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14312 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14314 The particular options set for any particular CPU will vary between
14315 compiler versions, depending on what setting seems to produce optimal
14316 code for that CPU; it doesn't necessarily reflect the actual hardware's
14317 capabilities. If you wish to set an individual option to a particular
14318 value, you may specify it after the @option{-mcpu} option, like
14319 @samp{-mcpu=970 -mno-altivec}.
14321 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14322 not enabled or disabled by the @option{-mcpu} option at present because
14323 AIX does not have full support for these options. You may still
14324 enable or disable them individually if you're sure it'll work in your
14327 @item -mtune=@var{cpu_type}
14329 Set the instruction scheduling parameters for machine type
14330 @var{cpu_type}, but do not set the architecture type, register usage, or
14331 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14332 values for @var{cpu_type} are used for @option{-mtune} as for
14333 @option{-mcpu}. If both are specified, the code generated will use the
14334 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14335 scheduling parameters set by @option{-mtune}.
14341 Generate code to compute division as reciprocal estimate and iterative
14342 refinement, creating opportunities for increased throughput. This
14343 feature requires: optional PowerPC Graphics instruction set for single
14344 precision and FRE instruction for double precision, assuming divides
14345 cannot generate user-visible traps, and the domain values not include
14346 Infinities, denormals or zero denominator.
14349 @itemx -mno-altivec
14351 @opindex mno-altivec
14352 Generate code that uses (does not use) AltiVec instructions, and also
14353 enable the use of built-in functions that allow more direct access to
14354 the AltiVec instruction set. You may also need to set
14355 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14361 @opindex mno-vrsave
14362 Generate VRSAVE instructions when generating AltiVec code.
14364 @item -mgen-cell-microcode
14365 @opindex mgen-cell-microcode
14366 Generate Cell microcode instructions
14368 @item -mwarn-cell-microcode
14369 @opindex mwarn-cell-microcode
14370 Warning when a Cell microcode instruction is going to emitted. An example
14371 of a Cell microcode instruction is a variable shift.
14374 @opindex msecure-plt
14375 Generate code that allows ld and ld.so to build executables and shared
14376 libraries with non-exec .plt and .got sections. This is a PowerPC
14377 32-bit SYSV ABI option.
14381 Generate code that uses a BSS .plt section that ld.so fills in, and
14382 requires .plt and .got sections that are both writable and executable.
14383 This is a PowerPC 32-bit SYSV ABI option.
14389 This switch enables or disables the generation of ISEL instructions.
14391 @item -misel=@var{yes/no}
14392 This switch has been deprecated. Use @option{-misel} and
14393 @option{-mno-isel} instead.
14399 This switch enables or disables the generation of SPE simd
14405 @opindex mno-paired
14406 This switch enables or disables the generation of PAIRED simd
14409 @item -mspe=@var{yes/no}
14410 This option has been deprecated. Use @option{-mspe} and
14411 @option{-mno-spe} instead.
14417 Generate code that uses (does not use) vector/scalar (VSX)
14418 instructions, and also enable the use of built-in functions that allow
14419 more direct access to the VSX instruction set.
14421 @item -mfloat-gprs=@var{yes/single/double/no}
14422 @itemx -mfloat-gprs
14423 @opindex mfloat-gprs
14424 This switch enables or disables the generation of floating point
14425 operations on the general purpose registers for architectures that
14428 The argument @var{yes} or @var{single} enables the use of
14429 single-precision floating point operations.
14431 The argument @var{double} enables the use of single and
14432 double-precision floating point operations.
14434 The argument @var{no} disables floating point operations on the
14435 general purpose registers.
14437 This option is currently only available on the MPC854x.
14443 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14444 targets (including GNU/Linux). The 32-bit environment sets int, long
14445 and pointer to 32 bits and generates code that runs on any PowerPC
14446 variant. The 64-bit environment sets int to 32 bits and long and
14447 pointer to 64 bits, and generates code for PowerPC64, as for
14448 @option{-mpowerpc64}.
14451 @itemx -mno-fp-in-toc
14452 @itemx -mno-sum-in-toc
14453 @itemx -mminimal-toc
14455 @opindex mno-fp-in-toc
14456 @opindex mno-sum-in-toc
14457 @opindex mminimal-toc
14458 Modify generation of the TOC (Table Of Contents), which is created for
14459 every executable file. The @option{-mfull-toc} option is selected by
14460 default. In that case, GCC will allocate at least one TOC entry for
14461 each unique non-automatic variable reference in your program. GCC
14462 will also place floating-point constants in the TOC@. However, only
14463 16,384 entries are available in the TOC@.
14465 If you receive a linker error message that saying you have overflowed
14466 the available TOC space, you can reduce the amount of TOC space used
14467 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14468 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14469 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14470 generate code to calculate the sum of an address and a constant at
14471 run-time instead of putting that sum into the TOC@. You may specify one
14472 or both of these options. Each causes GCC to produce very slightly
14473 slower and larger code at the expense of conserving TOC space.
14475 If you still run out of space in the TOC even when you specify both of
14476 these options, specify @option{-mminimal-toc} instead. This option causes
14477 GCC to make only one TOC entry for every file. When you specify this
14478 option, GCC will produce code that is slower and larger but which
14479 uses extremely little TOC space. You may wish to use this option
14480 only on files that contain less frequently executed code.
14486 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14487 @code{long} type, and the infrastructure needed to support them.
14488 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14489 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14490 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14493 @itemx -mno-xl-compat
14494 @opindex mxl-compat
14495 @opindex mno-xl-compat
14496 Produce code that conforms more closely to IBM XL compiler semantics
14497 when using AIX-compatible ABI@. Pass floating-point arguments to
14498 prototyped functions beyond the register save area (RSA) on the stack
14499 in addition to argument FPRs. Do not assume that most significant
14500 double in 128-bit long double value is properly rounded when comparing
14501 values and converting to double. Use XL symbol names for long double
14504 The AIX calling convention was extended but not initially documented to
14505 handle an obscure K&R C case of calling a function that takes the
14506 address of its arguments with fewer arguments than declared. IBM XL
14507 compilers access floating point arguments which do not fit in the
14508 RSA from the stack when a subroutine is compiled without
14509 optimization. Because always storing floating-point arguments on the
14510 stack is inefficient and rarely needed, this option is not enabled by
14511 default and only is necessary when calling subroutines compiled by IBM
14512 XL compilers without optimization.
14516 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14517 application written to use message passing with special startup code to
14518 enable the application to run. The system must have PE installed in the
14519 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14520 must be overridden with the @option{-specs=} option to specify the
14521 appropriate directory location. The Parallel Environment does not
14522 support threads, so the @option{-mpe} option and the @option{-pthread}
14523 option are incompatible.
14525 @item -malign-natural
14526 @itemx -malign-power
14527 @opindex malign-natural
14528 @opindex malign-power
14529 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14530 @option{-malign-natural} overrides the ABI-defined alignment of larger
14531 types, such as floating-point doubles, on their natural size-based boundary.
14532 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14533 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14535 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14539 @itemx -mhard-float
14540 @opindex msoft-float
14541 @opindex mhard-float
14542 Generate code that does not use (uses) the floating-point register set.
14543 Software floating point emulation is provided if you use the
14544 @option{-msoft-float} option, and pass the option to GCC when linking.
14546 @item -msingle-float
14547 @itemx -mdouble-float
14548 @opindex msingle-float
14549 @opindex mdouble-float
14550 Generate code for single or double-precision floating point operations.
14551 @option{-mdouble-float} implies @option{-msingle-float}.
14554 @opindex msimple-fpu
14555 Do not generate sqrt and div instructions for hardware floating point unit.
14559 Specify type of floating point unit. Valid values are @var{sp_lite}
14560 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14561 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14562 and @var{dp_full} (equivalent to -mdouble-float).
14565 @opindex mxilinx-fpu
14566 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14569 @itemx -mno-multiple
14571 @opindex mno-multiple
14572 Generate code that uses (does not use) the load multiple word
14573 instructions and the store multiple word instructions. These
14574 instructions are generated by default on POWER systems, and not
14575 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14576 endian PowerPC systems, since those instructions do not work when the
14577 processor is in little endian mode. The exceptions are PPC740 and
14578 PPC750 which permit the instructions usage in little endian mode.
14583 @opindex mno-string
14584 Generate code that uses (does not use) the load string instructions
14585 and the store string word instructions to save multiple registers and
14586 do small block moves. These instructions are generated by default on
14587 POWER systems, and not generated on PowerPC systems. Do not use
14588 @option{-mstring} on little endian PowerPC systems, since those
14589 instructions do not work when the processor is in little endian mode.
14590 The exceptions are PPC740 and PPC750 which permit the instructions
14591 usage in little endian mode.
14596 @opindex mno-update
14597 Generate code that uses (does not use) the load or store instructions
14598 that update the base register to the address of the calculated memory
14599 location. These instructions are generated by default. If you use
14600 @option{-mno-update}, there is a small window between the time that the
14601 stack pointer is updated and the address of the previous frame is
14602 stored, which means code that walks the stack frame across interrupts or
14603 signals may get corrupted data.
14605 @item -mavoid-indexed-addresses
14606 @item -mno-avoid-indexed-addresses
14607 @opindex mavoid-indexed-addresses
14608 @opindex mno-avoid-indexed-addresses
14609 Generate code that tries to avoid (not avoid) the use of indexed load
14610 or store instructions. These instructions can incur a performance
14611 penalty on Power6 processors in certain situations, such as when
14612 stepping through large arrays that cross a 16M boundary. This option
14613 is enabled by default when targetting Power6 and disabled otherwise.
14616 @itemx -mno-fused-madd
14617 @opindex mfused-madd
14618 @opindex mno-fused-madd
14619 Generate code that uses (does not use) the floating point multiply and
14620 accumulate instructions. These instructions are generated by default if
14621 hardware floating is used.
14627 Generate code that uses (does not use) the half-word multiply and
14628 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14629 These instructions are generated by default when targetting those
14636 Generate code that uses (does not use) the string-search @samp{dlmzb}
14637 instruction on the IBM 405, 440 and 464 processors. This instruction is
14638 generated by default when targetting those processors.
14640 @item -mno-bit-align
14642 @opindex mno-bit-align
14643 @opindex mbit-align
14644 On System V.4 and embedded PowerPC systems do not (do) force structures
14645 and unions that contain bit-fields to be aligned to the base type of the
14648 For example, by default a structure containing nothing but 8
14649 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14650 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14651 the structure would be aligned to a 1 byte boundary and be one byte in
14654 @item -mno-strict-align
14655 @itemx -mstrict-align
14656 @opindex mno-strict-align
14657 @opindex mstrict-align
14658 On System V.4 and embedded PowerPC systems do not (do) assume that
14659 unaligned memory references will be handled by the system.
14661 @item -mrelocatable
14662 @itemx -mno-relocatable
14663 @opindex mrelocatable
14664 @opindex mno-relocatable
14665 On embedded PowerPC systems generate code that allows (does not allow)
14666 the program to be relocated to a different address at runtime. If you
14667 use @option{-mrelocatable} on any module, all objects linked together must
14668 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14670 @item -mrelocatable-lib
14671 @itemx -mno-relocatable-lib
14672 @opindex mrelocatable-lib
14673 @opindex mno-relocatable-lib
14674 On embedded PowerPC systems generate code that allows (does not allow)
14675 the program to be relocated to a different address at runtime. Modules
14676 compiled with @option{-mrelocatable-lib} can be linked with either modules
14677 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14678 with modules compiled with the @option{-mrelocatable} options.
14684 On System V.4 and embedded PowerPC systems do not (do) assume that
14685 register 2 contains a pointer to a global area pointing to the addresses
14686 used in the program.
14689 @itemx -mlittle-endian
14691 @opindex mlittle-endian
14692 On System V.4 and embedded PowerPC systems compile code for the
14693 processor in little endian mode. The @option{-mlittle-endian} option is
14694 the same as @option{-mlittle}.
14697 @itemx -mbig-endian
14699 @opindex mbig-endian
14700 On System V.4 and embedded PowerPC systems compile code for the
14701 processor in big endian mode. The @option{-mbig-endian} option is
14702 the same as @option{-mbig}.
14704 @item -mdynamic-no-pic
14705 @opindex mdynamic-no-pic
14706 On Darwin and Mac OS X systems, compile code so that it is not
14707 relocatable, but that its external references are relocatable. The
14708 resulting code is suitable for applications, but not shared
14711 @item -mprioritize-restricted-insns=@var{priority}
14712 @opindex mprioritize-restricted-insns
14713 This option controls the priority that is assigned to
14714 dispatch-slot restricted instructions during the second scheduling
14715 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14716 @var{no/highest/second-highest} priority to dispatch slot restricted
14719 @item -msched-costly-dep=@var{dependence_type}
14720 @opindex msched-costly-dep
14721 This option controls which dependences are considered costly
14722 by the target during instruction scheduling. The argument
14723 @var{dependence_type} takes one of the following values:
14724 @var{no}: no dependence is costly,
14725 @var{all}: all dependences are costly,
14726 @var{true_store_to_load}: a true dependence from store to load is costly,
14727 @var{store_to_load}: any dependence from store to load is costly,
14728 @var{number}: any dependence which latency >= @var{number} is costly.
14730 @item -minsert-sched-nops=@var{scheme}
14731 @opindex minsert-sched-nops
14732 This option controls which nop insertion scheme will be used during
14733 the second scheduling pass. The argument @var{scheme} takes one of the
14735 @var{no}: Don't insert nops.
14736 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14737 according to the scheduler's grouping.
14738 @var{regroup_exact}: Insert nops to force costly dependent insns into
14739 separate groups. Insert exactly as many nops as needed to force an insn
14740 to a new group, according to the estimated processor grouping.
14741 @var{number}: Insert nops to force costly dependent insns into
14742 separate groups. Insert @var{number} nops to force an insn to a new group.
14745 @opindex mcall-sysv
14746 On System V.4 and embedded PowerPC systems compile code using calling
14747 conventions that adheres to the March 1995 draft of the System V
14748 Application Binary Interface, PowerPC processor supplement. This is the
14749 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14751 @item -mcall-sysv-eabi
14753 @opindex mcall-sysv-eabi
14754 @opindex mcall-eabi
14755 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14757 @item -mcall-sysv-noeabi
14758 @opindex mcall-sysv-noeabi
14759 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14761 @item -mcall-aixdesc
14763 On System V.4 and embedded PowerPC systems compile code for the AIX
14767 @opindex mcall-linux
14768 On System V.4 and embedded PowerPC systems compile code for the
14769 Linux-based GNU system.
14773 On System V.4 and embedded PowerPC systems compile code for the
14774 Hurd-based GNU system.
14776 @item -mcall-freebsd
14777 @opindex mcall-freebsd
14778 On System V.4 and embedded PowerPC systems compile code for the
14779 FreeBSD operating system.
14781 @item -mcall-netbsd
14782 @opindex mcall-netbsd
14783 On System V.4 and embedded PowerPC systems compile code for the
14784 NetBSD operating system.
14786 @item -mcall-openbsd
14787 @opindex mcall-netbsd
14788 On System V.4 and embedded PowerPC systems compile code for the
14789 OpenBSD operating system.
14791 @item -maix-struct-return
14792 @opindex maix-struct-return
14793 Return all structures in memory (as specified by the AIX ABI)@.
14795 @item -msvr4-struct-return
14796 @opindex msvr4-struct-return
14797 Return structures smaller than 8 bytes in registers (as specified by the
14800 @item -mabi=@var{abi-type}
14802 Extend the current ABI with a particular extension, or remove such extension.
14803 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14804 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14808 Extend the current ABI with SPE ABI extensions. This does not change
14809 the default ABI, instead it adds the SPE ABI extensions to the current
14813 @opindex mabi=no-spe
14814 Disable Booke SPE ABI extensions for the current ABI@.
14816 @item -mabi=ibmlongdouble
14817 @opindex mabi=ibmlongdouble
14818 Change the current ABI to use IBM extended precision long double.
14819 This is a PowerPC 32-bit SYSV ABI option.
14821 @item -mabi=ieeelongdouble
14822 @opindex mabi=ieeelongdouble
14823 Change the current ABI to use IEEE extended precision long double.
14824 This is a PowerPC 32-bit Linux ABI option.
14827 @itemx -mno-prototype
14828 @opindex mprototype
14829 @opindex mno-prototype
14830 On System V.4 and embedded PowerPC systems assume that all calls to
14831 variable argument functions are properly prototyped. Otherwise, the
14832 compiler must insert an instruction before every non prototyped call to
14833 set or clear bit 6 of the condition code register (@var{CR}) to
14834 indicate whether floating point values were passed in the floating point
14835 registers in case the function takes a variable arguments. With
14836 @option{-mprototype}, only calls to prototyped variable argument functions
14837 will set or clear the bit.
14841 On embedded PowerPC systems, assume that the startup module is called
14842 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14843 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14848 On embedded PowerPC systems, assume that the startup module is called
14849 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14854 On embedded PowerPC systems, assume that the startup module is called
14855 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14858 @item -myellowknife
14859 @opindex myellowknife
14860 On embedded PowerPC systems, assume that the startup module is called
14861 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14866 On System V.4 and embedded PowerPC systems, specify that you are
14867 compiling for a VxWorks system.
14871 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14872 header to indicate that @samp{eabi} extended relocations are used.
14878 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14879 Embedded Applications Binary Interface (eabi) which is a set of
14880 modifications to the System V.4 specifications. Selecting @option{-meabi}
14881 means that the stack is aligned to an 8 byte boundary, a function
14882 @code{__eabi} is called to from @code{main} to set up the eabi
14883 environment, and the @option{-msdata} option can use both @code{r2} and
14884 @code{r13} to point to two separate small data areas. Selecting
14885 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14886 do not call an initialization function from @code{main}, and the
14887 @option{-msdata} option will only use @code{r13} to point to a single
14888 small data area. The @option{-meabi} option is on by default if you
14889 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14892 @opindex msdata=eabi
14893 On System V.4 and embedded PowerPC systems, put small initialized
14894 @code{const} global and static data in the @samp{.sdata2} section, which
14895 is pointed to by register @code{r2}. Put small initialized
14896 non-@code{const} global and static data in the @samp{.sdata} section,
14897 which is pointed to by register @code{r13}. Put small uninitialized
14898 global and static data in the @samp{.sbss} section, which is adjacent to
14899 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14900 incompatible with the @option{-mrelocatable} option. The
14901 @option{-msdata=eabi} option also sets the @option{-memb} option.
14904 @opindex msdata=sysv
14905 On System V.4 and embedded PowerPC systems, put small global and static
14906 data in the @samp{.sdata} section, which is pointed to by register
14907 @code{r13}. Put small uninitialized global and static data in the
14908 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14909 The @option{-msdata=sysv} option is incompatible with the
14910 @option{-mrelocatable} option.
14912 @item -msdata=default
14914 @opindex msdata=default
14916 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14917 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14918 same as @option{-msdata=sysv}.
14921 @opindex msdata=data
14922 On System V.4 and embedded PowerPC systems, put small global
14923 data in the @samp{.sdata} section. Put small uninitialized global
14924 data in the @samp{.sbss} section. Do not use register @code{r13}
14925 to address small data however. This is the default behavior unless
14926 other @option{-msdata} options are used.
14930 @opindex msdata=none
14932 On embedded PowerPC systems, put all initialized global and static data
14933 in the @samp{.data} section, and all uninitialized data in the
14934 @samp{.bss} section.
14938 @cindex smaller data references (PowerPC)
14939 @cindex .sdata/.sdata2 references (PowerPC)
14940 On embedded PowerPC systems, put global and static items less than or
14941 equal to @var{num} bytes into the small data or bss sections instead of
14942 the normal data or bss section. By default, @var{num} is 8. The
14943 @option{-G @var{num}} switch is also passed to the linker.
14944 All modules should be compiled with the same @option{-G @var{num}} value.
14947 @itemx -mno-regnames
14949 @opindex mno-regnames
14950 On System V.4 and embedded PowerPC systems do (do not) emit register
14951 names in the assembly language output using symbolic forms.
14954 @itemx -mno-longcall
14956 @opindex mno-longcall
14957 By default assume that all calls are far away so that a longer more
14958 expensive calling sequence is required. This is required for calls
14959 further than 32 megabytes (33,554,432 bytes) from the current location.
14960 A short call will be generated if the compiler knows
14961 the call cannot be that far away. This setting can be overridden by
14962 the @code{shortcall} function attribute, or by @code{#pragma
14965 Some linkers are capable of detecting out-of-range calls and generating
14966 glue code on the fly. On these systems, long calls are unnecessary and
14967 generate slower code. As of this writing, the AIX linker can do this,
14968 as can the GNU linker for PowerPC/64. It is planned to add this feature
14969 to the GNU linker for 32-bit PowerPC systems as well.
14971 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14972 callee, L42'', plus a ``branch island'' (glue code). The two target
14973 addresses represent the callee and the ``branch island''. The
14974 Darwin/PPC linker will prefer the first address and generate a ``bl
14975 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14976 otherwise, the linker will generate ``bl L42'' to call the ``branch
14977 island''. The ``branch island'' is appended to the body of the
14978 calling function; it computes the full 32-bit address of the callee
14981 On Mach-O (Darwin) systems, this option directs the compiler emit to
14982 the glue for every direct call, and the Darwin linker decides whether
14983 to use or discard it.
14985 In the future, we may cause GCC to ignore all longcall specifications
14986 when the linker is known to generate glue.
14988 @item -mtls-markers
14989 @itemx -mno-tls-markers
14990 @opindex mtls-markers
14991 @opindex mno-tls-markers
14992 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
14993 specifying the function argument. The relocation allows ld to
14994 reliably associate function call with argument setup instructions for
14995 TLS optimization, which in turn allows gcc to better schedule the
15000 Adds support for multithreading with the @dfn{pthreads} library.
15001 This option sets flags for both the preprocessor and linker.
15005 @node S/390 and zSeries Options
15006 @subsection S/390 and zSeries Options
15007 @cindex S/390 and zSeries Options
15009 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15013 @itemx -msoft-float
15014 @opindex mhard-float
15015 @opindex msoft-float
15016 Use (do not use) the hardware floating-point instructions and registers
15017 for floating-point operations. When @option{-msoft-float} is specified,
15018 functions in @file{libgcc.a} will be used to perform floating-point
15019 operations. When @option{-mhard-float} is specified, the compiler
15020 generates IEEE floating-point instructions. This is the default.
15023 @itemx -mno-hard-dfp
15025 @opindex mno-hard-dfp
15026 Use (do not use) the hardware decimal-floating-point instructions for
15027 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15028 specified, functions in @file{libgcc.a} will be used to perform
15029 decimal-floating-point operations. When @option{-mhard-dfp} is
15030 specified, the compiler generates decimal-floating-point hardware
15031 instructions. This is the default for @option{-march=z9-ec} or higher.
15033 @item -mlong-double-64
15034 @itemx -mlong-double-128
15035 @opindex mlong-double-64
15036 @opindex mlong-double-128
15037 These switches control the size of @code{long double} type. A size
15038 of 64bit makes the @code{long double} type equivalent to the @code{double}
15039 type. This is the default.
15042 @itemx -mno-backchain
15043 @opindex mbackchain
15044 @opindex mno-backchain
15045 Store (do not store) the address of the caller's frame as backchain pointer
15046 into the callee's stack frame.
15047 A backchain may be needed to allow debugging using tools that do not understand
15048 DWARF-2 call frame information.
15049 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15050 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15051 the backchain is placed into the topmost word of the 96/160 byte register
15054 In general, code compiled with @option{-mbackchain} is call-compatible with
15055 code compiled with @option{-mmo-backchain}; however, use of the backchain
15056 for debugging purposes usually requires that the whole binary is built with
15057 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15058 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15059 to build a linux kernel use @option{-msoft-float}.
15061 The default is to not maintain the backchain.
15063 @item -mpacked-stack
15064 @itemx -mno-packed-stack
15065 @opindex mpacked-stack
15066 @opindex mno-packed-stack
15067 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15068 specified, the compiler uses the all fields of the 96/160 byte register save
15069 area only for their default purpose; unused fields still take up stack space.
15070 When @option{-mpacked-stack} is specified, register save slots are densely
15071 packed at the top of the register save area; unused space is reused for other
15072 purposes, allowing for more efficient use of the available stack space.
15073 However, when @option{-mbackchain} is also in effect, the topmost word of
15074 the save area is always used to store the backchain, and the return address
15075 register is always saved two words below the backchain.
15077 As long as the stack frame backchain is not used, code generated with
15078 @option{-mpacked-stack} is call-compatible with code generated with
15079 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15080 S/390 or zSeries generated code that uses the stack frame backchain at run
15081 time, not just for debugging purposes. Such code is not call-compatible
15082 with code compiled with @option{-mpacked-stack}. Also, note that the
15083 combination of @option{-mbackchain},
15084 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15085 to build a linux kernel use @option{-msoft-float}.
15087 The default is to not use the packed stack layout.
15090 @itemx -mno-small-exec
15091 @opindex msmall-exec
15092 @opindex mno-small-exec
15093 Generate (or do not generate) code using the @code{bras} instruction
15094 to do subroutine calls.
15095 This only works reliably if the total executable size does not
15096 exceed 64k. The default is to use the @code{basr} instruction instead,
15097 which does not have this limitation.
15103 When @option{-m31} is specified, generate code compliant to the
15104 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15105 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15106 particular to generate 64-bit instructions. For the @samp{s390}
15107 targets, the default is @option{-m31}, while the @samp{s390x}
15108 targets default to @option{-m64}.
15114 When @option{-mzarch} is specified, generate code using the
15115 instructions available on z/Architecture.
15116 When @option{-mesa} is specified, generate code using the
15117 instructions available on ESA/390. Note that @option{-mesa} is
15118 not possible with @option{-m64}.
15119 When generating code compliant to the GNU/Linux for S/390 ABI,
15120 the default is @option{-mesa}. When generating code compliant
15121 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15127 Generate (or do not generate) code using the @code{mvcle} instruction
15128 to perform block moves. When @option{-mno-mvcle} is specified,
15129 use a @code{mvc} loop instead. This is the default unless optimizing for
15136 Print (or do not print) additional debug information when compiling.
15137 The default is to not print debug information.
15139 @item -march=@var{cpu-type}
15141 Generate code that will run on @var{cpu-type}, which is the name of a system
15142 representing a certain processor type. Possible values for
15143 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15144 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15145 When generating code using the instructions available on z/Architecture,
15146 the default is @option{-march=z900}. Otherwise, the default is
15147 @option{-march=g5}.
15149 @item -mtune=@var{cpu-type}
15151 Tune to @var{cpu-type} everything applicable about the generated code,
15152 except for the ABI and the set of available instructions.
15153 The list of @var{cpu-type} values is the same as for @option{-march}.
15154 The default is the value used for @option{-march}.
15157 @itemx -mno-tpf-trace
15158 @opindex mtpf-trace
15159 @opindex mno-tpf-trace
15160 Generate code that adds (does not add) in TPF OS specific branches to trace
15161 routines in the operating system. This option is off by default, even
15162 when compiling for the TPF OS@.
15165 @itemx -mno-fused-madd
15166 @opindex mfused-madd
15167 @opindex mno-fused-madd
15168 Generate code that uses (does not use) the floating point multiply and
15169 accumulate instructions. These instructions are generated by default if
15170 hardware floating point is used.
15172 @item -mwarn-framesize=@var{framesize}
15173 @opindex mwarn-framesize
15174 Emit a warning if the current function exceeds the given frame size. Because
15175 this is a compile time check it doesn't need to be a real problem when the program
15176 runs. It is intended to identify functions which most probably cause
15177 a stack overflow. It is useful to be used in an environment with limited stack
15178 size e.g.@: the linux kernel.
15180 @item -mwarn-dynamicstack
15181 @opindex mwarn-dynamicstack
15182 Emit a warning if the function calls alloca or uses dynamically
15183 sized arrays. This is generally a bad idea with a limited stack size.
15185 @item -mstack-guard=@var{stack-guard}
15186 @itemx -mstack-size=@var{stack-size}
15187 @opindex mstack-guard
15188 @opindex mstack-size
15189 If these options are provided the s390 back end emits additional instructions in
15190 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15191 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15192 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15193 the frame size of the compiled function is chosen.
15194 These options are intended to be used to help debugging stack overflow problems.
15195 The additionally emitted code causes only little overhead and hence can also be
15196 used in production like systems without greater performance degradation. The given
15197 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15198 @var{stack-guard} without exceeding 64k.
15199 In order to be efficient the extra code makes the assumption that the stack starts
15200 at an address aligned to the value given by @var{stack-size}.
15201 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15204 @node Score Options
15205 @subsection Score Options
15206 @cindex Score Options
15208 These options are defined for Score implementations:
15213 Compile code for big endian mode. This is the default.
15217 Compile code for little endian mode.
15221 Disable generate bcnz instruction.
15225 Enable generate unaligned load and store instruction.
15229 Enable the use of multiply-accumulate instructions. Disabled by default.
15233 Specify the SCORE5 as the target architecture.
15237 Specify the SCORE5U of the target architecture.
15241 Specify the SCORE7 as the target architecture. This is the default.
15245 Specify the SCORE7D as the target architecture.
15249 @subsection SH Options
15251 These @samp{-m} options are defined for the SH implementations:
15256 Generate code for the SH1.
15260 Generate code for the SH2.
15263 Generate code for the SH2e.
15267 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15268 that the floating-point unit is not used.
15270 @item -m2a-single-only
15271 @opindex m2a-single-only
15272 Generate code for the SH2a-FPU, in such a way that no double-precision
15273 floating point operations are used.
15276 @opindex m2a-single
15277 Generate code for the SH2a-FPU assuming the floating-point unit is in
15278 single-precision mode by default.
15282 Generate code for the SH2a-FPU assuming the floating-point unit is in
15283 double-precision mode by default.
15287 Generate code for the SH3.
15291 Generate code for the SH3e.
15295 Generate code for the SH4 without a floating-point unit.
15297 @item -m4-single-only
15298 @opindex m4-single-only
15299 Generate code for the SH4 with a floating-point unit that only
15300 supports single-precision arithmetic.
15304 Generate code for the SH4 assuming the floating-point unit is in
15305 single-precision mode by default.
15309 Generate code for the SH4.
15313 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15314 floating-point unit is not used.
15316 @item -m4a-single-only
15317 @opindex m4a-single-only
15318 Generate code for the SH4a, in such a way that no double-precision
15319 floating point operations are used.
15322 @opindex m4a-single
15323 Generate code for the SH4a assuming the floating-point unit is in
15324 single-precision mode by default.
15328 Generate code for the SH4a.
15332 Same as @option{-m4a-nofpu}, except that it implicitly passes
15333 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15334 instructions at the moment.
15338 Compile code for the processor in big endian mode.
15342 Compile code for the processor in little endian mode.
15346 Align doubles at 64-bit boundaries. Note that this changes the calling
15347 conventions, and thus some functions from the standard C library will
15348 not work unless you recompile it first with @option{-mdalign}.
15352 Shorten some address references at link time, when possible; uses the
15353 linker option @option{-relax}.
15357 Use 32-bit offsets in @code{switch} tables. The default is to use
15362 Enable the use of bit manipulation instructions on SH2A.
15366 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15367 alignment constraints.
15371 Comply with the calling conventions defined by Renesas.
15375 Comply with the calling conventions defined by Renesas.
15379 Comply with the calling conventions defined for GCC before the Renesas
15380 conventions were available. This option is the default for all
15381 targets of the SH toolchain except for @samp{sh-symbianelf}.
15384 @opindex mnomacsave
15385 Mark the @code{MAC} register as call-clobbered, even if
15386 @option{-mhitachi} is given.
15390 Increase IEEE-compliance of floating-point code.
15391 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15392 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15393 comparisons of NANs / infinities incurs extra overhead in every
15394 floating point comparison, therefore the default is set to
15395 @option{-ffinite-math-only}.
15397 @item -minline-ic_invalidate
15398 @opindex minline-ic_invalidate
15399 Inline code to invalidate instruction cache entries after setting up
15400 nested function trampolines.
15401 This option has no effect if -musermode is in effect and the selected
15402 code generation option (e.g. -m4) does not allow the use of the icbi
15404 If the selected code generation option does not allow the use of the icbi
15405 instruction, and -musermode is not in effect, the inlined code will
15406 manipulate the instruction cache address array directly with an associative
15407 write. This not only requires privileged mode, but it will also
15408 fail if the cache line had been mapped via the TLB and has become unmapped.
15412 Dump instruction size and location in the assembly code.
15415 @opindex mpadstruct
15416 This option is deprecated. It pads structures to multiple of 4 bytes,
15417 which is incompatible with the SH ABI@.
15421 Optimize for space instead of speed. Implied by @option{-Os}.
15424 @opindex mprefergot
15425 When generating position-independent code, emit function calls using
15426 the Global Offset Table instead of the Procedure Linkage Table.
15430 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15431 if the inlined code would not work in user mode.
15432 This is the default when the target is @code{sh-*-linux*}.
15434 @item -multcost=@var{number}
15435 @opindex multcost=@var{number}
15436 Set the cost to assume for a multiply insn.
15438 @item -mdiv=@var{strategy}
15439 @opindex mdiv=@var{strategy}
15440 Set the division strategy to use for SHmedia code. @var{strategy} must be
15441 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15442 inv:call2, inv:fp .
15443 "fp" performs the operation in floating point. This has a very high latency,
15444 but needs only a few instructions, so it might be a good choice if
15445 your code has enough easily exploitable ILP to allow the compiler to
15446 schedule the floating point instructions together with other instructions.
15447 Division by zero causes a floating point exception.
15448 "inv" uses integer operations to calculate the inverse of the divisor,
15449 and then multiplies the dividend with the inverse. This strategy allows
15450 cse and hoisting of the inverse calculation. Division by zero calculates
15451 an unspecified result, but does not trap.
15452 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15453 have been found, or if the entire operation has been hoisted to the same
15454 place, the last stages of the inverse calculation are intertwined with the
15455 final multiply to reduce the overall latency, at the expense of using a few
15456 more instructions, and thus offering fewer scheduling opportunities with
15458 "call" calls a library function that usually implements the inv:minlat
15460 This gives high code density for m5-*media-nofpu compilations.
15461 "call2" uses a different entry point of the same library function, where it
15462 assumes that a pointer to a lookup table has already been set up, which
15463 exposes the pointer load to cse / code hoisting optimizations.
15464 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15465 code generation, but if the code stays unoptimized, revert to the "call",
15466 "call2", or "fp" strategies, respectively. Note that the
15467 potentially-trapping side effect of division by zero is carried by a
15468 separate instruction, so it is possible that all the integer instructions
15469 are hoisted out, but the marker for the side effect stays where it is.
15470 A recombination to fp operations or a call is not possible in that case.
15471 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15472 that the inverse calculation was nor separated from the multiply, they speed
15473 up division where the dividend fits into 20 bits (plus sign where applicable),
15474 by inserting a test to skip a number of operations in this case; this test
15475 slows down the case of larger dividends. inv20u assumes the case of a such
15476 a small dividend to be unlikely, and inv20l assumes it to be likely.
15478 @item -mdivsi3_libfunc=@var{name}
15479 @opindex mdivsi3_libfunc=@var{name}
15480 Set the name of the library function used for 32 bit signed division to
15481 @var{name}. This only affect the name used in the call and inv:call
15482 division strategies, and the compiler will still expect the same
15483 sets of input/output/clobbered registers as if this option was not present.
15485 @item -mfixed-range=@var{register-range}
15486 @opindex mfixed-range
15487 Generate code treating the given register range as fixed registers.
15488 A fixed register is one that the register allocator can not use. This is
15489 useful when compiling kernel code. A register range is specified as
15490 two registers separated by a dash. Multiple register ranges can be
15491 specified separated by a comma.
15493 @item -madjust-unroll
15494 @opindex madjust-unroll
15495 Throttle unrolling to avoid thrashing target registers.
15496 This option only has an effect if the gcc code base supports the
15497 TARGET_ADJUST_UNROLL_MAX target hook.
15499 @item -mindexed-addressing
15500 @opindex mindexed-addressing
15501 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15502 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15503 semantics for the indexed addressing mode. The architecture allows the
15504 implementation of processors with 64 bit MMU, which the OS could use to
15505 get 32 bit addressing, but since no current hardware implementation supports
15506 this or any other way to make the indexed addressing mode safe to use in
15507 the 32 bit ABI, the default is -mno-indexed-addressing.
15509 @item -mgettrcost=@var{number}
15510 @opindex mgettrcost=@var{number}
15511 Set the cost assumed for the gettr instruction to @var{number}.
15512 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15516 Assume pt* instructions won't trap. This will generally generate better
15517 scheduled code, but is unsafe on current hardware. The current architecture
15518 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15519 This has the unintentional effect of making it unsafe to schedule ptabs /
15520 ptrel before a branch, or hoist it out of a loop. For example,
15521 __do_global_ctors, a part of libgcc that runs constructors at program
15522 startup, calls functions in a list which is delimited by @minus{}1. With the
15523 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15524 That means that all the constructors will be run a bit quicker, but when
15525 the loop comes to the end of the list, the program crashes because ptabs
15526 loads @minus{}1 into a target register. Since this option is unsafe for any
15527 hardware implementing the current architecture specification, the default
15528 is -mno-pt-fixed. Unless the user specifies a specific cost with
15529 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15530 this deters register allocation using target registers for storing
15533 @item -minvalid-symbols
15534 @opindex minvalid-symbols
15535 Assume symbols might be invalid. Ordinary function symbols generated by
15536 the compiler will always be valid to load with movi/shori/ptabs or
15537 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15538 to generate symbols that will cause ptabs / ptrel to trap.
15539 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15540 It will then prevent cross-basic-block cse, hoisting and most scheduling
15541 of symbol loads. The default is @option{-mno-invalid-symbols}.
15544 @node SPARC Options
15545 @subsection SPARC Options
15546 @cindex SPARC options
15548 These @samp{-m} options are supported on the SPARC:
15551 @item -mno-app-regs
15553 @opindex mno-app-regs
15555 Specify @option{-mapp-regs} to generate output using the global registers
15556 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15559 To be fully SVR4 ABI compliant at the cost of some performance loss,
15560 specify @option{-mno-app-regs}. You should compile libraries and system
15561 software with this option.
15564 @itemx -mhard-float
15566 @opindex mhard-float
15567 Generate output containing floating point instructions. This is the
15571 @itemx -msoft-float
15573 @opindex msoft-float
15574 Generate output containing library calls for floating point.
15575 @strong{Warning:} the requisite libraries are not available for all SPARC
15576 targets. Normally the facilities of the machine's usual C compiler are
15577 used, but this cannot be done directly in cross-compilation. You must make
15578 your own arrangements to provide suitable library functions for
15579 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15580 @samp{sparclite-*-*} do provide software floating point support.
15582 @option{-msoft-float} changes the calling convention in the output file;
15583 therefore, it is only useful if you compile @emph{all} of a program with
15584 this option. In particular, you need to compile @file{libgcc.a}, the
15585 library that comes with GCC, with @option{-msoft-float} in order for
15588 @item -mhard-quad-float
15589 @opindex mhard-quad-float
15590 Generate output containing quad-word (long double) floating point
15593 @item -msoft-quad-float
15594 @opindex msoft-quad-float
15595 Generate output containing library calls for quad-word (long double)
15596 floating point instructions. The functions called are those specified
15597 in the SPARC ABI@. This is the default.
15599 As of this writing, there are no SPARC implementations that have hardware
15600 support for the quad-word floating point instructions. They all invoke
15601 a trap handler for one of these instructions, and then the trap handler
15602 emulates the effect of the instruction. Because of the trap handler overhead,
15603 this is much slower than calling the ABI library routines. Thus the
15604 @option{-msoft-quad-float} option is the default.
15606 @item -mno-unaligned-doubles
15607 @itemx -munaligned-doubles
15608 @opindex mno-unaligned-doubles
15609 @opindex munaligned-doubles
15610 Assume that doubles have 8 byte alignment. This is the default.
15612 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15613 alignment only if they are contained in another type, or if they have an
15614 absolute address. Otherwise, it assumes they have 4 byte alignment.
15615 Specifying this option avoids some rare compatibility problems with code
15616 generated by other compilers. It is not the default because it results
15617 in a performance loss, especially for floating point code.
15619 @item -mno-faster-structs
15620 @itemx -mfaster-structs
15621 @opindex mno-faster-structs
15622 @opindex mfaster-structs
15623 With @option{-mfaster-structs}, the compiler assumes that structures
15624 should have 8 byte alignment. This enables the use of pairs of
15625 @code{ldd} and @code{std} instructions for copies in structure
15626 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15627 However, the use of this changed alignment directly violates the SPARC
15628 ABI@. Thus, it's intended only for use on targets where the developer
15629 acknowledges that their resulting code will not be directly in line with
15630 the rules of the ABI@.
15632 @item -mimpure-text
15633 @opindex mimpure-text
15634 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15635 the compiler to not pass @option{-z text} to the linker when linking a
15636 shared object. Using this option, you can link position-dependent
15637 code into a shared object.
15639 @option{-mimpure-text} suppresses the ``relocations remain against
15640 allocatable but non-writable sections'' linker error message.
15641 However, the necessary relocations will trigger copy-on-write, and the
15642 shared object is not actually shared across processes. Instead of
15643 using @option{-mimpure-text}, you should compile all source code with
15644 @option{-fpic} or @option{-fPIC}.
15646 This option is only available on SunOS and Solaris.
15648 @item -mcpu=@var{cpu_type}
15650 Set the instruction set, register set, and instruction scheduling parameters
15651 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15652 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15653 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15654 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15655 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15657 Default instruction scheduling parameters are used for values that select
15658 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15659 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15661 Here is a list of each supported architecture and their supported
15666 v8: supersparc, hypersparc
15667 sparclite: f930, f934, sparclite86x
15669 v9: ultrasparc, ultrasparc3, niagara, niagara2
15672 By default (unless configured otherwise), GCC generates code for the V7
15673 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15674 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15675 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15676 SPARCStation 1, 2, IPX etc.
15678 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15679 architecture. The only difference from V7 code is that the compiler emits
15680 the integer multiply and integer divide instructions which exist in SPARC-V8
15681 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15682 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15685 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15686 the SPARC architecture. This adds the integer multiply, integer divide step
15687 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15688 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15689 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15690 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15691 MB86934 chip, which is the more recent SPARClite with FPU@.
15693 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15694 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15695 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15696 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15697 optimizes it for the TEMIC SPARClet chip.
15699 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15700 architecture. This adds 64-bit integer and floating-point move instructions,
15701 3 additional floating-point condition code registers and conditional move
15702 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15703 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15704 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15705 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15706 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15707 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15708 additionally optimizes it for Sun UltraSPARC T2 chips.
15710 @item -mtune=@var{cpu_type}
15712 Set the instruction scheduling parameters for machine type
15713 @var{cpu_type}, but do not set the instruction set or register set that the
15714 option @option{-mcpu=@var{cpu_type}} would.
15716 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15717 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15718 that select a particular cpu implementation. Those are @samp{cypress},
15719 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15720 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15721 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15726 @opindex mno-v8plus
15727 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15728 difference from the V8 ABI is that the global and out registers are
15729 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15730 mode for all SPARC-V9 processors.
15736 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15737 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15740 These @samp{-m} options are supported in addition to the above
15741 on SPARC-V9 processors in 64-bit environments:
15744 @item -mlittle-endian
15745 @opindex mlittle-endian
15746 Generate code for a processor running in little-endian mode. It is only
15747 available for a few configurations and most notably not on Solaris and Linux.
15753 Generate code for a 32-bit or 64-bit environment.
15754 The 32-bit environment sets int, long and pointer to 32 bits.
15755 The 64-bit environment sets int to 32 bits and long and pointer
15758 @item -mcmodel=medlow
15759 @opindex mcmodel=medlow
15760 Generate code for the Medium/Low code model: 64-bit addresses, programs
15761 must be linked in the low 32 bits of memory. Programs can be statically
15762 or dynamically linked.
15764 @item -mcmodel=medmid
15765 @opindex mcmodel=medmid
15766 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15767 must be linked in the low 44 bits of memory, the text and data segments must
15768 be less than 2GB in size and the data segment must be located within 2GB of
15771 @item -mcmodel=medany
15772 @opindex mcmodel=medany
15773 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15774 may be linked anywhere in memory, the text and data segments must be less
15775 than 2GB in size and the data segment must be located within 2GB of the
15778 @item -mcmodel=embmedany
15779 @opindex mcmodel=embmedany
15780 Generate code for the Medium/Anywhere code model for embedded systems:
15781 64-bit addresses, the text and data segments must be less than 2GB in
15782 size, both starting anywhere in memory (determined at link time). The
15783 global register %g4 points to the base of the data segment. Programs
15784 are statically linked and PIC is not supported.
15787 @itemx -mno-stack-bias
15788 @opindex mstack-bias
15789 @opindex mno-stack-bias
15790 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15791 frame pointer if present, are offset by @minus{}2047 which must be added back
15792 when making stack frame references. This is the default in 64-bit mode.
15793 Otherwise, assume no such offset is present.
15796 These switches are supported in addition to the above on Solaris:
15801 Add support for multithreading using the Solaris threads library. This
15802 option sets flags for both the preprocessor and linker. This option does
15803 not affect the thread safety of object code produced by the compiler or
15804 that of libraries supplied with it.
15808 Add support for multithreading using the POSIX threads library. This
15809 option sets flags for both the preprocessor and linker. This option does
15810 not affect the thread safety of object code produced by the compiler or
15811 that of libraries supplied with it.
15815 This is a synonym for @option{-pthreads}.
15819 @subsection SPU Options
15820 @cindex SPU options
15822 These @samp{-m} options are supported on the SPU:
15826 @itemx -merror-reloc
15827 @opindex mwarn-reloc
15828 @opindex merror-reloc
15830 The loader for SPU does not handle dynamic relocations. By default, GCC
15831 will give an error when it generates code that requires a dynamic
15832 relocation. @option{-mno-error-reloc} disables the error,
15833 @option{-mwarn-reloc} will generate a warning instead.
15836 @itemx -munsafe-dma
15838 @opindex munsafe-dma
15840 Instructions which initiate or test completion of DMA must not be
15841 reordered with respect to loads and stores of the memory which is being
15842 accessed. Users typically address this problem using the volatile
15843 keyword, but that can lead to inefficient code in places where the
15844 memory is known to not change. Rather than mark the memory as volatile
15845 we treat the DMA instructions as potentially effecting all memory. With
15846 @option{-munsafe-dma} users must use the volatile keyword to protect
15849 @item -mbranch-hints
15850 @opindex mbranch-hints
15852 By default, GCC will generate a branch hint instruction to avoid
15853 pipeline stalls for always taken or probably taken branches. A hint
15854 will not be generated closer than 8 instructions away from its branch.
15855 There is little reason to disable them, except for debugging purposes,
15856 or to make an object a little bit smaller.
15860 @opindex msmall-mem
15861 @opindex mlarge-mem
15863 By default, GCC generates code assuming that addresses are never larger
15864 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15865 a full 32 bit address.
15870 By default, GCC links against startup code that assumes the SPU-style
15871 main function interface (which has an unconventional parameter list).
15872 With @option{-mstdmain}, GCC will link your program against startup
15873 code that assumes a C99-style interface to @code{main}, including a
15874 local copy of @code{argv} strings.
15876 @item -mfixed-range=@var{register-range}
15877 @opindex mfixed-range
15878 Generate code treating the given register range as fixed registers.
15879 A fixed register is one that the register allocator can not use. This is
15880 useful when compiling kernel code. A register range is specified as
15881 two registers separated by a dash. Multiple register ranges can be
15882 specified separated by a comma.
15885 @itemx -mdual-nops=@var{n}
15886 @opindex mdual-nops
15887 By default, GCC will insert nops to increase dual issue when it expects
15888 it to increase performance. @var{n} can be a value from 0 to 10. A
15889 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15890 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15892 @item -mhint-max-nops=@var{n}
15893 @opindex mhint-max-nops
15894 Maximum number of nops to insert for a branch hint. A branch hint must
15895 be at least 8 instructions away from the branch it is effecting. GCC
15896 will insert up to @var{n} nops to enforce this, otherwise it will not
15897 generate the branch hint.
15899 @item -mhint-max-distance=@var{n}
15900 @opindex mhint-max-distance
15901 The encoding of the branch hint instruction limits the hint to be within
15902 256 instructions of the branch it is effecting. By default, GCC makes
15903 sure it is within 125.
15906 @opindex msafe-hints
15907 Work around a hardware bug which causes the SPU to stall indefinitely.
15908 By default, GCC will insert the @code{hbrp} instruction to make sure
15909 this stall won't happen.
15913 @node System V Options
15914 @subsection Options for System V
15916 These additional options are available on System V Release 4 for
15917 compatibility with other compilers on those systems:
15922 Create a shared object.
15923 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15927 Identify the versions of each tool used by the compiler, in a
15928 @code{.ident} assembler directive in the output.
15932 Refrain from adding @code{.ident} directives to the output file (this is
15935 @item -YP,@var{dirs}
15937 Search the directories @var{dirs}, and no others, for libraries
15938 specified with @option{-l}.
15940 @item -Ym,@var{dir}
15942 Look in the directory @var{dir} to find the M4 preprocessor.
15943 The assembler uses this option.
15944 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15945 @c the generic assembler that comes with Solaris takes just -Ym.
15949 @subsection V850 Options
15950 @cindex V850 Options
15952 These @samp{-m} options are defined for V850 implementations:
15956 @itemx -mno-long-calls
15957 @opindex mlong-calls
15958 @opindex mno-long-calls
15959 Treat all calls as being far away (near). If calls are assumed to be
15960 far away, the compiler will always load the functions address up into a
15961 register, and call indirect through the pointer.
15967 Do not optimize (do optimize) basic blocks that use the same index
15968 pointer 4 or more times to copy pointer into the @code{ep} register, and
15969 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15970 option is on by default if you optimize.
15972 @item -mno-prolog-function
15973 @itemx -mprolog-function
15974 @opindex mno-prolog-function
15975 @opindex mprolog-function
15976 Do not use (do use) external functions to save and restore registers
15977 at the prologue and epilogue of a function. The external functions
15978 are slower, but use less code space if more than one function saves
15979 the same number of registers. The @option{-mprolog-function} option
15980 is on by default if you optimize.
15984 Try to make the code as small as possible. At present, this just turns
15985 on the @option{-mep} and @option{-mprolog-function} options.
15987 @item -mtda=@var{n}
15989 Put static or global variables whose size is @var{n} bytes or less into
15990 the tiny data area that register @code{ep} points to. The tiny data
15991 area can hold up to 256 bytes in total (128 bytes for byte references).
15993 @item -msda=@var{n}
15995 Put static or global variables whose size is @var{n} bytes or less into
15996 the small data area that register @code{gp} points to. The small data
15997 area can hold up to 64 kilobytes.
15999 @item -mzda=@var{n}
16001 Put static or global variables whose size is @var{n} bytes or less into
16002 the first 32 kilobytes of memory.
16006 Specify that the target processor is the V850.
16009 @opindex mbig-switch
16010 Generate code suitable for big switch tables. Use this option only if
16011 the assembler/linker complain about out of range branches within a switch
16016 This option will cause r2 and r5 to be used in the code generated by
16017 the compiler. This setting is the default.
16019 @item -mno-app-regs
16020 @opindex mno-app-regs
16021 This option will cause r2 and r5 to be treated as fixed registers.
16025 Specify that the target processor is the V850E1. The preprocessor
16026 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16027 this option is used.
16031 Specify that the target processor is the V850E@. The preprocessor
16032 constant @samp{__v850e__} will be defined if this option is used.
16034 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16035 are defined then a default target processor will be chosen and the
16036 relevant @samp{__v850*__} preprocessor constant will be defined.
16038 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16039 defined, regardless of which processor variant is the target.
16041 @item -mdisable-callt
16042 @opindex mdisable-callt
16043 This option will suppress generation of the CALLT instruction for the
16044 v850e and v850e1 flavors of the v850 architecture. The default is
16045 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16050 @subsection VAX Options
16051 @cindex VAX options
16053 These @samp{-m} options are defined for the VAX:
16058 Do not output certain jump instructions (@code{aobleq} and so on)
16059 that the Unix assembler for the VAX cannot handle across long
16064 Do output those jump instructions, on the assumption that you
16065 will assemble with the GNU assembler.
16069 Output code for g-format floating point numbers instead of d-format.
16072 @node VxWorks Options
16073 @subsection VxWorks Options
16074 @cindex VxWorks Options
16076 The options in this section are defined for all VxWorks targets.
16077 Options specific to the target hardware are listed with the other
16078 options for that target.
16083 GCC can generate code for both VxWorks kernels and real time processes
16084 (RTPs). This option switches from the former to the latter. It also
16085 defines the preprocessor macro @code{__RTP__}.
16088 @opindex non-static
16089 Link an RTP executable against shared libraries rather than static
16090 libraries. The options @option{-static} and @option{-shared} can
16091 also be used for RTPs (@pxref{Link Options}); @option{-static}
16098 These options are passed down to the linker. They are defined for
16099 compatibility with Diab.
16102 @opindex Xbind-lazy
16103 Enable lazy binding of function calls. This option is equivalent to
16104 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16108 Disable lazy binding of function calls. This option is the default and
16109 is defined for compatibility with Diab.
16112 @node x86-64 Options
16113 @subsection x86-64 Options
16114 @cindex x86-64 options
16116 These are listed under @xref{i386 and x86-64 Options}.
16118 @node i386 and x86-64 Windows Options
16119 @subsection i386 and x86-64 Windows Options
16120 @cindex i386 and x86-64 Windows Options
16122 These additional options are available for Windows targets:
16127 This option is available for Cygwin and MinGW targets. It
16128 specifies that a console application is to be generated, by
16129 instructing the linker to set the PE header subsystem type
16130 required for console applications.
16131 This is the default behavior for Cygwin and MinGW targets.
16135 This option is available for Cygwin targets. It specifies that
16136 the Cygwin internal interface is to be used for predefined
16137 preprocessor macros, C runtime libraries and related linker
16138 paths and options. For Cygwin targets this is the default behavior.
16139 This option is deprecated and will be removed in a future release.
16142 @opindex mno-cygwin
16143 This option is available for Cygwin targets. It specifies that
16144 the MinGW internal interface is to be used instead of Cygwin's, by
16145 setting MinGW-related predefined macros and linker paths and default
16147 This option is deprecated and will be removed in a future release.
16151 This option is available for Cygwin and MinGW targets. It
16152 specifies that a DLL - a dynamic link library - is to be
16153 generated, enabling the selection of the required runtime
16154 startup object and entry point.
16156 @item -mnop-fun-dllimport
16157 @opindex mnop-fun-dllimport
16158 This option is available for Cygwin and MinGW targets. It
16159 specifies that the dllimport attribute should be ignored.
16163 This option is available for MinGW targets. It specifies
16164 that MinGW-specific thread support is to be used.
16168 This option is available for mingw-w64 targets. It specifies
16169 that the UNICODE macro is getting pre-defined and that the
16170 unicode capable runtime startup code is choosen.
16174 This option is available for Cygwin and MinGW targets. It
16175 specifies that the typical Windows pre-defined macros are to
16176 be set in the pre-processor, but does not influence the choice
16177 of runtime library/startup code.
16181 This option is available for Cygwin and MinGW targets. It
16182 specifies that a GUI application is to be generated by
16183 instructing the linker to set the PE header subsystem type
16186 @item -mpe-aligned-commons
16187 @opindex mpe-aligned-commons
16188 This option is available for Cygwin and MinGW targets. It
16189 specifies that the GNU extension to the PE file format that
16190 permits the correct alignment of COMMON variables should be
16191 used when generating code. It will be enabled by default if
16192 GCC detects that the target assembler found during configuration
16193 supports the feature.
16196 See also under @ref{i386 and x86-64 Options} for standard options.
16198 @node Xstormy16 Options
16199 @subsection Xstormy16 Options
16200 @cindex Xstormy16 Options
16202 These options are defined for Xstormy16:
16207 Choose startup files and linker script suitable for the simulator.
16210 @node Xtensa Options
16211 @subsection Xtensa Options
16212 @cindex Xtensa Options
16214 These options are supported for Xtensa targets:
16218 @itemx -mno-const16
16220 @opindex mno-const16
16221 Enable or disable use of @code{CONST16} instructions for loading
16222 constant values. The @code{CONST16} instruction is currently not a
16223 standard option from Tensilica. When enabled, @code{CONST16}
16224 instructions are always used in place of the standard @code{L32R}
16225 instructions. The use of @code{CONST16} is enabled by default only if
16226 the @code{L32R} instruction is not available.
16229 @itemx -mno-fused-madd
16230 @opindex mfused-madd
16231 @opindex mno-fused-madd
16232 Enable or disable use of fused multiply/add and multiply/subtract
16233 instructions in the floating-point option. This has no effect if the
16234 floating-point option is not also enabled. Disabling fused multiply/add
16235 and multiply/subtract instructions forces the compiler to use separate
16236 instructions for the multiply and add/subtract operations. This may be
16237 desirable in some cases where strict IEEE 754-compliant results are
16238 required: the fused multiply add/subtract instructions do not round the
16239 intermediate result, thereby producing results with @emph{more} bits of
16240 precision than specified by the IEEE standard. Disabling fused multiply
16241 add/subtract instructions also ensures that the program output is not
16242 sensitive to the compiler's ability to combine multiply and add/subtract
16245 @item -mserialize-volatile
16246 @itemx -mno-serialize-volatile
16247 @opindex mserialize-volatile
16248 @opindex mno-serialize-volatile
16249 When this option is enabled, GCC inserts @code{MEMW} instructions before
16250 @code{volatile} memory references to guarantee sequential consistency.
16251 The default is @option{-mserialize-volatile}. Use
16252 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16254 @item -mtext-section-literals
16255 @itemx -mno-text-section-literals
16256 @opindex mtext-section-literals
16257 @opindex mno-text-section-literals
16258 Control the treatment of literal pools. The default is
16259 @option{-mno-text-section-literals}, which places literals in a separate
16260 section in the output file. This allows the literal pool to be placed
16261 in a data RAM/ROM, and it also allows the linker to combine literal
16262 pools from separate object files to remove redundant literals and
16263 improve code size. With @option{-mtext-section-literals}, the literals
16264 are interspersed in the text section in order to keep them as close as
16265 possible to their references. This may be necessary for large assembly
16268 @item -mtarget-align
16269 @itemx -mno-target-align
16270 @opindex mtarget-align
16271 @opindex mno-target-align
16272 When this option is enabled, GCC instructs the assembler to
16273 automatically align instructions to reduce branch penalties at the
16274 expense of some code density. The assembler attempts to widen density
16275 instructions to align branch targets and the instructions following call
16276 instructions. If there are not enough preceding safe density
16277 instructions to align a target, no widening will be performed. The
16278 default is @option{-mtarget-align}. These options do not affect the
16279 treatment of auto-aligned instructions like @code{LOOP}, which the
16280 assembler will always align, either by widening density instructions or
16281 by inserting no-op instructions.
16284 @itemx -mno-longcalls
16285 @opindex mlongcalls
16286 @opindex mno-longcalls
16287 When this option is enabled, GCC instructs the assembler to translate
16288 direct calls to indirect calls unless it can determine that the target
16289 of a direct call is in the range allowed by the call instruction. This
16290 translation typically occurs for calls to functions in other source
16291 files. Specifically, the assembler translates a direct @code{CALL}
16292 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16293 The default is @option{-mno-longcalls}. This option should be used in
16294 programs where the call target can potentially be out of range. This
16295 option is implemented in the assembler, not the compiler, so the
16296 assembly code generated by GCC will still show direct call
16297 instructions---look at the disassembled object code to see the actual
16298 instructions. Note that the assembler will use an indirect call for
16299 every cross-file call, not just those that really will be out of range.
16302 @node zSeries Options
16303 @subsection zSeries Options
16304 @cindex zSeries options
16306 These are listed under @xref{S/390 and zSeries Options}.
16308 @node Code Gen Options
16309 @section Options for Code Generation Conventions
16310 @cindex code generation conventions
16311 @cindex options, code generation
16312 @cindex run-time options
16314 These machine-independent options control the interface conventions
16315 used in code generation.
16317 Most of them have both positive and negative forms; the negative form
16318 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16319 one of the forms is listed---the one which is not the default. You
16320 can figure out the other form by either removing @samp{no-} or adding
16324 @item -fbounds-check
16325 @opindex fbounds-check
16326 For front-ends that support it, generate additional code to check that
16327 indices used to access arrays are within the declared range. This is
16328 currently only supported by the Java and Fortran front-ends, where
16329 this option defaults to true and false respectively.
16333 This option generates traps for signed overflow on addition, subtraction,
16334 multiplication operations.
16338 This option instructs the compiler to assume that signed arithmetic
16339 overflow of addition, subtraction and multiplication wraps around
16340 using twos-complement representation. This flag enables some optimizations
16341 and disables others. This option is enabled by default for the Java
16342 front-end, as required by the Java language specification.
16345 @opindex fexceptions
16346 Enable exception handling. Generates extra code needed to propagate
16347 exceptions. For some targets, this implies GCC will generate frame
16348 unwind information for all functions, which can produce significant data
16349 size overhead, although it does not affect execution. If you do not
16350 specify this option, GCC will enable it by default for languages like
16351 C++ which normally require exception handling, and disable it for
16352 languages like C that do not normally require it. However, you may need
16353 to enable this option when compiling C code that needs to interoperate
16354 properly with exception handlers written in C++. You may also wish to
16355 disable this option if you are compiling older C++ programs that don't
16356 use exception handling.
16358 @item -fnon-call-exceptions
16359 @opindex fnon-call-exceptions
16360 Generate code that allows trapping instructions to throw exceptions.
16361 Note that this requires platform-specific runtime support that does
16362 not exist everywhere. Moreover, it only allows @emph{trapping}
16363 instructions to throw exceptions, i.e.@: memory references or floating
16364 point instructions. It does not allow exceptions to be thrown from
16365 arbitrary signal handlers such as @code{SIGALRM}.
16367 @item -funwind-tables
16368 @opindex funwind-tables
16369 Similar to @option{-fexceptions}, except that it will just generate any needed
16370 static data, but will not affect the generated code in any other way.
16371 You will normally not enable this option; instead, a language processor
16372 that needs this handling would enable it on your behalf.
16374 @item -fasynchronous-unwind-tables
16375 @opindex fasynchronous-unwind-tables
16376 Generate unwind table in dwarf2 format, if supported by target machine. The
16377 table is exact at each instruction boundary, so it can be used for stack
16378 unwinding from asynchronous events (such as debugger or garbage collector).
16380 @item -fpcc-struct-return
16381 @opindex fpcc-struct-return
16382 Return ``short'' @code{struct} and @code{union} values in memory like
16383 longer ones, rather than in registers. This convention is less
16384 efficient, but it has the advantage of allowing intercallability between
16385 GCC-compiled files and files compiled with other compilers, particularly
16386 the Portable C Compiler (pcc).
16388 The precise convention for returning structures in memory depends
16389 on the target configuration macros.
16391 Short structures and unions are those whose size and alignment match
16392 that of some integer type.
16394 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16395 switch is not binary compatible with code compiled with the
16396 @option{-freg-struct-return} switch.
16397 Use it to conform to a non-default application binary interface.
16399 @item -freg-struct-return
16400 @opindex freg-struct-return
16401 Return @code{struct} and @code{union} values in registers when possible.
16402 This is more efficient for small structures than
16403 @option{-fpcc-struct-return}.
16405 If you specify neither @option{-fpcc-struct-return} nor
16406 @option{-freg-struct-return}, GCC defaults to whichever convention is
16407 standard for the target. If there is no standard convention, GCC
16408 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16409 the principal compiler. In those cases, we can choose the standard, and
16410 we chose the more efficient register return alternative.
16412 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16413 switch is not binary compatible with code compiled with the
16414 @option{-fpcc-struct-return} switch.
16415 Use it to conform to a non-default application binary interface.
16417 @item -fshort-enums
16418 @opindex fshort-enums
16419 Allocate to an @code{enum} type only as many bytes as it needs for the
16420 declared range of possible values. Specifically, the @code{enum} type
16421 will be equivalent to the smallest integer type which has enough room.
16423 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16424 code that is not binary compatible with code generated without that switch.
16425 Use it to conform to a non-default application binary interface.
16427 @item -fshort-double
16428 @opindex fshort-double
16429 Use the same size for @code{double} as for @code{float}.
16431 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16432 code that is not binary compatible with code generated without that switch.
16433 Use it to conform to a non-default application binary interface.
16435 @item -fshort-wchar
16436 @opindex fshort-wchar
16437 Override the underlying type for @samp{wchar_t} to be @samp{short
16438 unsigned int} instead of the default for the target. This option is
16439 useful for building programs to run under WINE@.
16441 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16442 code that is not binary compatible with code generated without that switch.
16443 Use it to conform to a non-default application binary interface.
16446 @opindex fno-common
16447 In C code, controls the placement of uninitialized global variables.
16448 Unix C compilers have traditionally permitted multiple definitions of
16449 such variables in different compilation units by placing the variables
16451 This is the behavior specified by @option{-fcommon}, and is the default
16452 for GCC on most targets.
16453 On the other hand, this behavior is not required by ISO C, and on some
16454 targets may carry a speed or code size penalty on variable references.
16455 The @option{-fno-common} option specifies that the compiler should place
16456 uninitialized global variables in the data section of the object file,
16457 rather than generating them as common blocks.
16458 This has the effect that if the same variable is declared
16459 (without @code{extern}) in two different compilations,
16460 you will get a multiple-definition error when you link them.
16461 In this case, you must compile with @option{-fcommon} instead.
16462 Compiling with @option{-fno-common} is useful on targets for which
16463 it provides better performance, or if you wish to verify that the
16464 program will work on other systems which always treat uninitialized
16465 variable declarations this way.
16469 Ignore the @samp{#ident} directive.
16471 @item -finhibit-size-directive
16472 @opindex finhibit-size-directive
16473 Don't output a @code{.size} assembler directive, or anything else that
16474 would cause trouble if the function is split in the middle, and the
16475 two halves are placed at locations far apart in memory. This option is
16476 used when compiling @file{crtstuff.c}; you should not need to use it
16479 @item -fverbose-asm
16480 @opindex fverbose-asm
16481 Put extra commentary information in the generated assembly code to
16482 make it more readable. This option is generally only of use to those
16483 who actually need to read the generated assembly code (perhaps while
16484 debugging the compiler itself).
16486 @option{-fno-verbose-asm}, the default, causes the
16487 extra information to be omitted and is useful when comparing two assembler
16490 @item -frecord-gcc-switches
16491 @opindex frecord-gcc-switches
16492 This switch causes the command line that was used to invoke the
16493 compiler to be recorded into the object file that is being created.
16494 This switch is only implemented on some targets and the exact format
16495 of the recording is target and binary file format dependent, but it
16496 usually takes the form of a section containing ASCII text. This
16497 switch is related to the @option{-fverbose-asm} switch, but that
16498 switch only records information in the assembler output file as
16499 comments, so it never reaches the object file.
16503 @cindex global offset table
16505 Generate position-independent code (PIC) suitable for use in a shared
16506 library, if supported for the target machine. Such code accesses all
16507 constant addresses through a global offset table (GOT)@. The dynamic
16508 loader resolves the GOT entries when the program starts (the dynamic
16509 loader is not part of GCC; it is part of the operating system). If
16510 the GOT size for the linked executable exceeds a machine-specific
16511 maximum size, you get an error message from the linker indicating that
16512 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16513 instead. (These maximums are 8k on the SPARC and 32k
16514 on the m68k and RS/6000. The 386 has no such limit.)
16516 Position-independent code requires special support, and therefore works
16517 only on certain machines. For the 386, GCC supports PIC for System V
16518 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16519 position-independent.
16521 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16526 If supported for the target machine, emit position-independent code,
16527 suitable for dynamic linking and avoiding any limit on the size of the
16528 global offset table. This option makes a difference on the m68k,
16529 PowerPC and SPARC@.
16531 Position-independent code requires special support, and therefore works
16532 only on certain machines.
16534 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16541 These options are similar to @option{-fpic} and @option{-fPIC}, but
16542 generated position independent code can be only linked into executables.
16543 Usually these options are used when @option{-pie} GCC option will be
16544 used during linking.
16546 @option{-fpie} and @option{-fPIE} both define the macros
16547 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16548 for @option{-fpie} and 2 for @option{-fPIE}.
16550 @item -fno-jump-tables
16551 @opindex fno-jump-tables
16552 Do not use jump tables for switch statements even where it would be
16553 more efficient than other code generation strategies. This option is
16554 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16555 building code which forms part of a dynamic linker and cannot
16556 reference the address of a jump table. On some targets, jump tables
16557 do not require a GOT and this option is not needed.
16559 @item -ffixed-@var{reg}
16561 Treat the register named @var{reg} as a fixed register; generated code
16562 should never refer to it (except perhaps as a stack pointer, frame
16563 pointer or in some other fixed role).
16565 @var{reg} must be the name of a register. The register names accepted
16566 are machine-specific and are defined in the @code{REGISTER_NAMES}
16567 macro in the machine description macro file.
16569 This flag does not have a negative form, because it specifies a
16572 @item -fcall-used-@var{reg}
16573 @opindex fcall-used
16574 Treat the register named @var{reg} as an allocable register that is
16575 clobbered by function calls. It may be allocated for temporaries or
16576 variables that do not live across a call. Functions compiled this way
16577 will not save and restore the register @var{reg}.
16579 It is an error to used this flag with the frame pointer or stack pointer.
16580 Use of this flag for other registers that have fixed pervasive roles in
16581 the machine's execution model will produce disastrous results.
16583 This flag does not have a negative form, because it specifies a
16586 @item -fcall-saved-@var{reg}
16587 @opindex fcall-saved
16588 Treat the register named @var{reg} as an allocable register saved by
16589 functions. It may be allocated even for temporaries or variables that
16590 live across a call. Functions compiled this way will save and restore
16591 the register @var{reg} if they use it.
16593 It is an error to used this flag with the frame pointer or stack pointer.
16594 Use of this flag for other registers that have fixed pervasive roles in
16595 the machine's execution model will produce disastrous results.
16597 A different sort of disaster will result from the use of this flag for
16598 a register in which function values may be returned.
16600 This flag does not have a negative form, because it specifies a
16603 @item -fpack-struct[=@var{n}]
16604 @opindex fpack-struct
16605 Without a value specified, pack all structure members together without
16606 holes. When a value is specified (which must be a small power of two), pack
16607 structure members according to this value, representing the maximum
16608 alignment (that is, objects with default alignment requirements larger than
16609 this will be output potentially unaligned at the next fitting location.
16611 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16612 code that is not binary compatible with code generated without that switch.
16613 Additionally, it makes the code suboptimal.
16614 Use it to conform to a non-default application binary interface.
16616 @item -finstrument-functions
16617 @opindex finstrument-functions
16618 Generate instrumentation calls for entry and exit to functions. Just
16619 after function entry and just before function exit, the following
16620 profiling functions will be called with the address of the current
16621 function and its call site. (On some platforms,
16622 @code{__builtin_return_address} does not work beyond the current
16623 function, so the call site information may not be available to the
16624 profiling functions otherwise.)
16627 void __cyg_profile_func_enter (void *this_fn,
16629 void __cyg_profile_func_exit (void *this_fn,
16633 The first argument is the address of the start of the current function,
16634 which may be looked up exactly in the symbol table.
16636 This instrumentation is also done for functions expanded inline in other
16637 functions. The profiling calls will indicate where, conceptually, the
16638 inline function is entered and exited. This means that addressable
16639 versions of such functions must be available. If all your uses of a
16640 function are expanded inline, this may mean an additional expansion of
16641 code size. If you use @samp{extern inline} in your C code, an
16642 addressable version of such functions must be provided. (This is
16643 normally the case anyways, but if you get lucky and the optimizer always
16644 expands the functions inline, you might have gotten away without
16645 providing static copies.)
16647 A function may be given the attribute @code{no_instrument_function}, in
16648 which case this instrumentation will not be done. This can be used, for
16649 example, for the profiling functions listed above, high-priority
16650 interrupt routines, and any functions from which the profiling functions
16651 cannot safely be called (perhaps signal handlers, if the profiling
16652 routines generate output or allocate memory).
16654 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16655 @opindex finstrument-functions-exclude-file-list
16657 Set the list of functions that are excluded from instrumentation (see
16658 the description of @code{-finstrument-functions}). If the file that
16659 contains a function definition matches with one of @var{file}, then
16660 that function is not instrumented. The match is done on substrings:
16661 if the @var{file} parameter is a substring of the file name, it is
16662 considered to be a match.
16665 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16666 will exclude any inline function defined in files whose pathnames
16667 contain @code{/bits/stl} or @code{include/sys}.
16669 If, for some reason, you want to include letter @code{','} in one of
16670 @var{sym}, write @code{'\,'}. For example,
16671 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16672 (note the single quote surrounding the option).
16674 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16675 @opindex finstrument-functions-exclude-function-list
16677 This is similar to @code{-finstrument-functions-exclude-file-list},
16678 but this option sets the list of function names to be excluded from
16679 instrumentation. The function name to be matched is its user-visible
16680 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16681 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16682 match is done on substrings: if the @var{sym} parameter is a substring
16683 of the function name, it is considered to be a match. For C99 and C++
16684 extended identifiers, the function name must be given in UTF-8, not
16685 using universal character names.
16687 @item -fstack-check
16688 @opindex fstack-check
16689 Generate code to verify that you do not go beyond the boundary of the
16690 stack. You should specify this flag if you are running in an
16691 environment with multiple threads, but only rarely need to specify it in
16692 a single-threaded environment since stack overflow is automatically
16693 detected on nearly all systems if there is only one stack.
16695 Note that this switch does not actually cause checking to be done; the
16696 operating system or the language runtime must do that. The switch causes
16697 generation of code to ensure that they see the stack being extended.
16699 You can additionally specify a string parameter: @code{no} means no
16700 checking, @code{generic} means force the use of old-style checking,
16701 @code{specific} means use the best checking method and is equivalent
16702 to bare @option{-fstack-check}.
16704 Old-style checking is a generic mechanism that requires no specific
16705 target support in the compiler but comes with the following drawbacks:
16709 Modified allocation strategy for large objects: they will always be
16710 allocated dynamically if their size exceeds a fixed threshold.
16713 Fixed limit on the size of the static frame of functions: when it is
16714 topped by a particular function, stack checking is not reliable and
16715 a warning is issued by the compiler.
16718 Inefficiency: because of both the modified allocation strategy and the
16719 generic implementation, the performances of the code are hampered.
16722 Note that old-style stack checking is also the fallback method for
16723 @code{specific} if no target support has been added in the compiler.
16725 @item -fstack-limit-register=@var{reg}
16726 @itemx -fstack-limit-symbol=@var{sym}
16727 @itemx -fno-stack-limit
16728 @opindex fstack-limit-register
16729 @opindex fstack-limit-symbol
16730 @opindex fno-stack-limit
16731 Generate code to ensure that the stack does not grow beyond a certain value,
16732 either the value of a register or the address of a symbol. If the stack
16733 would grow beyond the value, a signal is raised. For most targets,
16734 the signal is raised before the stack overruns the boundary, so
16735 it is possible to catch the signal without taking special precautions.
16737 For instance, if the stack starts at absolute address @samp{0x80000000}
16738 and grows downwards, you can use the flags
16739 @option{-fstack-limit-symbol=__stack_limit} and
16740 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16741 of 128KB@. Note that this may only work with the GNU linker.
16743 @cindex aliasing of parameters
16744 @cindex parameters, aliased
16745 @item -fargument-alias
16746 @itemx -fargument-noalias
16747 @itemx -fargument-noalias-global
16748 @itemx -fargument-noalias-anything
16749 @opindex fargument-alias
16750 @opindex fargument-noalias
16751 @opindex fargument-noalias-global
16752 @opindex fargument-noalias-anything
16753 Specify the possible relationships among parameters and between
16754 parameters and global data.
16756 @option{-fargument-alias} specifies that arguments (parameters) may
16757 alias each other and may alias global storage.@*
16758 @option{-fargument-noalias} specifies that arguments do not alias
16759 each other, but may alias global storage.@*
16760 @option{-fargument-noalias-global} specifies that arguments do not
16761 alias each other and do not alias global storage.
16762 @option{-fargument-noalias-anything} specifies that arguments do not
16763 alias any other storage.
16765 Each language will automatically use whatever option is required by
16766 the language standard. You should not need to use these options yourself.
16768 @item -fleading-underscore
16769 @opindex fleading-underscore
16770 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16771 change the way C symbols are represented in the object file. One use
16772 is to help link with legacy assembly code.
16774 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16775 generate code that is not binary compatible with code generated without that
16776 switch. Use it to conform to a non-default application binary interface.
16777 Not all targets provide complete support for this switch.
16779 @item -ftls-model=@var{model}
16780 @opindex ftls-model
16781 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16782 The @var{model} argument should be one of @code{global-dynamic},
16783 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16785 The default without @option{-fpic} is @code{initial-exec}; with
16786 @option{-fpic} the default is @code{global-dynamic}.
16788 @item -fvisibility=@var{default|internal|hidden|protected}
16789 @opindex fvisibility
16790 Set the default ELF image symbol visibility to the specified option---all
16791 symbols will be marked with this unless overridden within the code.
16792 Using this feature can very substantially improve linking and
16793 load times of shared object libraries, produce more optimized
16794 code, provide near-perfect API export and prevent symbol clashes.
16795 It is @strong{strongly} recommended that you use this in any shared objects
16798 Despite the nomenclature, @code{default} always means public ie;
16799 available to be linked against from outside the shared object.
16800 @code{protected} and @code{internal} are pretty useless in real-world
16801 usage so the only other commonly used option will be @code{hidden}.
16802 The default if @option{-fvisibility} isn't specified is
16803 @code{default}, i.e., make every
16804 symbol public---this causes the same behavior as previous versions of
16807 A good explanation of the benefits offered by ensuring ELF
16808 symbols have the correct visibility is given by ``How To Write
16809 Shared Libraries'' by Ulrich Drepper (which can be found at
16810 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16811 solution made possible by this option to marking things hidden when
16812 the default is public is to make the default hidden and mark things
16813 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16814 and @code{__attribute__ ((visibility("default")))} instead of
16815 @code{__declspec(dllexport)} you get almost identical semantics with
16816 identical syntax. This is a great boon to those working with
16817 cross-platform projects.
16819 For those adding visibility support to existing code, you may find
16820 @samp{#pragma GCC visibility} of use. This works by you enclosing
16821 the declarations you wish to set visibility for with (for example)
16822 @samp{#pragma GCC visibility push(hidden)} and
16823 @samp{#pragma GCC visibility pop}.
16824 Bear in mind that symbol visibility should be viewed @strong{as
16825 part of the API interface contract} and thus all new code should
16826 always specify visibility when it is not the default ie; declarations
16827 only for use within the local DSO should @strong{always} be marked explicitly
16828 as hidden as so to avoid PLT indirection overheads---making this
16829 abundantly clear also aids readability and self-documentation of the code.
16830 Note that due to ISO C++ specification requirements, operator new and
16831 operator delete must always be of default visibility.
16833 Be aware that headers from outside your project, in particular system
16834 headers and headers from any other library you use, may not be
16835 expecting to be compiled with visibility other than the default. You
16836 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16837 before including any such headers.
16839 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16840 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16841 no modifications. However, this means that calls to @samp{extern}
16842 functions with no explicit visibility will use the PLT, so it is more
16843 effective to use @samp{__attribute ((visibility))} and/or
16844 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16845 declarations should be treated as hidden.
16847 Note that @samp{-fvisibility} does affect C++ vague linkage
16848 entities. This means that, for instance, an exception class that will
16849 be thrown between DSOs must be explicitly marked with default
16850 visibility so that the @samp{type_info} nodes will be unified between
16853 An overview of these techniques, their benefits and how to use them
16854 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16860 @node Environment Variables
16861 @section Environment Variables Affecting GCC
16862 @cindex environment variables
16864 @c man begin ENVIRONMENT
16865 This section describes several environment variables that affect how GCC
16866 operates. Some of them work by specifying directories or prefixes to use
16867 when searching for various kinds of files. Some are used to specify other
16868 aspects of the compilation environment.
16870 Note that you can also specify places to search using options such as
16871 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16872 take precedence over places specified using environment variables, which
16873 in turn take precedence over those specified by the configuration of GCC@.
16874 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16875 GNU Compiler Collection (GCC) Internals}.
16880 @c @itemx LC_COLLATE
16882 @c @itemx LC_MONETARY
16883 @c @itemx LC_NUMERIC
16888 @c @findex LC_COLLATE
16889 @findex LC_MESSAGES
16890 @c @findex LC_MONETARY
16891 @c @findex LC_NUMERIC
16895 These environment variables control the way that GCC uses
16896 localization information that allow GCC to work with different
16897 national conventions. GCC inspects the locale categories
16898 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16899 so. These locale categories can be set to any value supported by your
16900 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16901 Kingdom encoded in UTF-8.
16903 The @env{LC_CTYPE} environment variable specifies character
16904 classification. GCC uses it to determine the character boundaries in
16905 a string; this is needed for some multibyte encodings that contain quote
16906 and escape characters that would otherwise be interpreted as a string
16909 The @env{LC_MESSAGES} environment variable specifies the language to
16910 use in diagnostic messages.
16912 If the @env{LC_ALL} environment variable is set, it overrides the value
16913 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16914 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16915 environment variable. If none of these variables are set, GCC
16916 defaults to traditional C English behavior.
16920 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16921 files. GCC uses temporary files to hold the output of one stage of
16922 compilation which is to be used as input to the next stage: for example,
16923 the output of the preprocessor, which is the input to the compiler
16926 @item GCC_EXEC_PREFIX
16927 @findex GCC_EXEC_PREFIX
16928 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16929 names of the subprograms executed by the compiler. No slash is added
16930 when this prefix is combined with the name of a subprogram, but you can
16931 specify a prefix that ends with a slash if you wish.
16933 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16934 an appropriate prefix to use based on the pathname it was invoked with.
16936 If GCC cannot find the subprogram using the specified prefix, it
16937 tries looking in the usual places for the subprogram.
16939 The default value of @env{GCC_EXEC_PREFIX} is
16940 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16941 the installed compiler. In many cases @var{prefix} is the value
16942 of @code{prefix} when you ran the @file{configure} script.
16944 Other prefixes specified with @option{-B} take precedence over this prefix.
16946 This prefix is also used for finding files such as @file{crt0.o} that are
16949 In addition, the prefix is used in an unusual way in finding the
16950 directories to search for header files. For each of the standard
16951 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16952 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16953 replacing that beginning with the specified prefix to produce an
16954 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16955 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16956 These alternate directories are searched first; the standard directories
16957 come next. If a standard directory begins with the configured
16958 @var{prefix} then the value of @var{prefix} is replaced by
16959 @env{GCC_EXEC_PREFIX} when looking for header files.
16961 @item COMPILER_PATH
16962 @findex COMPILER_PATH
16963 The value of @env{COMPILER_PATH} is a colon-separated list of
16964 directories, much like @env{PATH}. GCC tries the directories thus
16965 specified when searching for subprograms, if it can't find the
16966 subprograms using @env{GCC_EXEC_PREFIX}.
16969 @findex LIBRARY_PATH
16970 The value of @env{LIBRARY_PATH} is a colon-separated list of
16971 directories, much like @env{PATH}. When configured as a native compiler,
16972 GCC tries the directories thus specified when searching for special
16973 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16974 using GCC also uses these directories when searching for ordinary
16975 libraries for the @option{-l} option (but directories specified with
16976 @option{-L} come first).
16980 @cindex locale definition
16981 This variable is used to pass locale information to the compiler. One way in
16982 which this information is used is to determine the character set to be used
16983 when character literals, string literals and comments are parsed in C and C++.
16984 When the compiler is configured to allow multibyte characters,
16985 the following values for @env{LANG} are recognized:
16989 Recognize JIS characters.
16991 Recognize SJIS characters.
16993 Recognize EUCJP characters.
16996 If @env{LANG} is not defined, or if it has some other value, then the
16997 compiler will use mblen and mbtowc as defined by the default locale to
16998 recognize and translate multibyte characters.
17002 Some additional environments variables affect the behavior of the
17005 @include cppenv.texi
17009 @node Precompiled Headers
17010 @section Using Precompiled Headers
17011 @cindex precompiled headers
17012 @cindex speed of compilation
17014 Often large projects have many header files that are included in every
17015 source file. The time the compiler takes to process these header files
17016 over and over again can account for nearly all of the time required to
17017 build the project. To make builds faster, GCC allows users to
17018 `precompile' a header file; then, if builds can use the precompiled
17019 header file they will be much faster.
17021 To create a precompiled header file, simply compile it as you would any
17022 other file, if necessary using the @option{-x} option to make the driver
17023 treat it as a C or C++ header file. You will probably want to use a
17024 tool like @command{make} to keep the precompiled header up-to-date when
17025 the headers it contains change.
17027 A precompiled header file will be searched for when @code{#include} is
17028 seen in the compilation. As it searches for the included file
17029 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17030 compiler looks for a precompiled header in each directory just before it
17031 looks for the include file in that directory. The name searched for is
17032 the name specified in the @code{#include} with @samp{.gch} appended. If
17033 the precompiled header file can't be used, it is ignored.
17035 For instance, if you have @code{#include "all.h"}, and you have
17036 @file{all.h.gch} in the same directory as @file{all.h}, then the
17037 precompiled header file will be used if possible, and the original
17038 header will be used otherwise.
17040 Alternatively, you might decide to put the precompiled header file in a
17041 directory and use @option{-I} to ensure that directory is searched
17042 before (or instead of) the directory containing the original header.
17043 Then, if you want to check that the precompiled header file is always
17044 used, you can put a file of the same name as the original header in this
17045 directory containing an @code{#error} command.
17047 This also works with @option{-include}. So yet another way to use
17048 precompiled headers, good for projects not designed with precompiled
17049 header files in mind, is to simply take most of the header files used by
17050 a project, include them from another header file, precompile that header
17051 file, and @option{-include} the precompiled header. If the header files
17052 have guards against multiple inclusion, they will be skipped because
17053 they've already been included (in the precompiled header).
17055 If you need to precompile the same header file for different
17056 languages, targets, or compiler options, you can instead make a
17057 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17058 header in the directory, perhaps using @option{-o}. It doesn't matter
17059 what you call the files in the directory, every precompiled header in
17060 the directory will be considered. The first precompiled header
17061 encountered in the directory that is valid for this compilation will
17062 be used; they're searched in no particular order.
17064 There are many other possibilities, limited only by your imagination,
17065 good sense, and the constraints of your build system.
17067 A precompiled header file can be used only when these conditions apply:
17071 Only one precompiled header can be used in a particular compilation.
17074 A precompiled header can't be used once the first C token is seen. You
17075 can have preprocessor directives before a precompiled header; you can
17076 even include a precompiled header from inside another header, so long as
17077 there are no C tokens before the @code{#include}.
17080 The precompiled header file must be produced for the same language as
17081 the current compilation. You can't use a C precompiled header for a C++
17085 The precompiled header file must have been produced by the same compiler
17086 binary as the current compilation is using.
17089 Any macros defined before the precompiled header is included must
17090 either be defined in the same way as when the precompiled header was
17091 generated, or must not affect the precompiled header, which usually
17092 means that they don't appear in the precompiled header at all.
17094 The @option{-D} option is one way to define a macro before a
17095 precompiled header is included; using a @code{#define} can also do it.
17096 There are also some options that define macros implicitly, like
17097 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17100 @item If debugging information is output when using the precompiled
17101 header, using @option{-g} or similar, the same kind of debugging information
17102 must have been output when building the precompiled header. However,
17103 a precompiled header built using @option{-g} can be used in a compilation
17104 when no debugging information is being output.
17106 @item The same @option{-m} options must generally be used when building
17107 and using the precompiled header. @xref{Submodel Options},
17108 for any cases where this rule is relaxed.
17110 @item Each of the following options must be the same when building and using
17111 the precompiled header:
17113 @gccoptlist{-fexceptions}
17116 Some other command-line options starting with @option{-f},
17117 @option{-p}, or @option{-O} must be defined in the same way as when
17118 the precompiled header was generated. At present, it's not clear
17119 which options are safe to change and which are not; the safest choice
17120 is to use exactly the same options when generating and using the
17121 precompiled header. The following are known to be safe:
17123 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17124 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17125 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17130 For all of these except the last, the compiler will automatically
17131 ignore the precompiled header if the conditions aren't met. If you
17132 find an option combination that doesn't work and doesn't cause the
17133 precompiled header to be ignored, please consider filing a bug report,
17136 If you do use differing options when generating and using the
17137 precompiled header, the actual behavior will be a mixture of the
17138 behavior for the options. For instance, if you use @option{-g} to
17139 generate the precompiled header but not when using it, you may or may
17140 not get debugging information for routines in the precompiled header.