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 -fenable-icf-debug @gol
311 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
312 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
313 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
314 -ftest-coverage -ftime-report -fvar-tracking @gol
315 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
316 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
317 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
318 -gvms -gxcoff -gxcoff+ @gol
319 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
320 -fdebug-prefix-map=@var{old}=@var{new} @gol
321 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
322 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
323 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
324 -print-multi-directory -print-multi-lib @gol
325 -print-prog-name=@var{program} -print-search-dirs -Q @gol
326 -print-sysroot -print-sysroot-headers-suffix @gol
327 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
329 @item Optimization Options
330 @xref{Optimize Options,,Options that Control Optimization}.
332 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
333 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
334 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
335 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
336 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
337 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
338 -fdata-sections -fdce -fdce @gol
339 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
340 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
341 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
342 -fforward-propagate -ffunction-sections @gol
343 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
344 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
345 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
346 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
347 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
348 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
349 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
350 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
351 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
352 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
353 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
354 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
355 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
356 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
357 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
358 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
359 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
360 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
361 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
362 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
363 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
364 -fprofile-generate=@var{path} @gol
365 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
366 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
367 -freorder-blocks-and-partition -freorder-functions @gol
368 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
369 -frounding-math -fsched2-use-superblocks @gol
370 -fsched2-use-traces -fsched-pressure @gol
371 -fsched-spec-load -fsched-spec-load-dangerous @gol
372 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
373 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
374 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
375 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
376 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
377 -fselective-scheduling -fselective-scheduling2 @gol
378 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
379 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
380 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
381 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
382 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
383 -ftree-copyrename -ftree-dce @gol
384 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
385 -ftree-phiprop -ftree-loop-distribution @gol
386 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
387 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
388 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
389 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
390 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
391 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
392 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
393 -fwhole-program -fwhopr -fwpa -use-linker-plugin @gol
394 --param @var{name}=@var{value}
395 -O -O0 -O1 -O2 -O3 -Os}
397 @item Preprocessor Options
398 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
399 @gccoptlist{-A@var{question}=@var{answer} @gol
400 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
401 -C -dD -dI -dM -dN @gol
402 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
403 -idirafter @var{dir} @gol
404 -include @var{file} -imacros @var{file} @gol
405 -iprefix @var{file} -iwithprefix @var{dir} @gol
406 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
407 -imultilib @var{dir} -isysroot @var{dir} @gol
408 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
409 -P -fworking-directory -remap @gol
410 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
411 -Xpreprocessor @var{option}}
413 @item Assembler Option
414 @xref{Assembler Options,,Passing Options to the Assembler}.
415 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
418 @xref{Link Options,,Options for Linking}.
419 @gccoptlist{@var{object-file-name} -l@var{library} @gol
420 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
421 -s -static -static-libgcc -static-libstdc++ -shared @gol
422 -shared-libgcc -symbolic @gol
423 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
426 @item Directory Options
427 @xref{Directory Options,,Options for Directory Search}.
428 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
429 -specs=@var{file} -I- --sysroot=@var{dir}}
432 @c I wrote this xref this way to avoid overfull hbox. -- rms
433 @xref{Target Options}.
434 @gccoptlist{-V @var{version} -b @var{machine}}
436 @item Machine Dependent Options
437 @xref{Submodel Options,,Hardware Models and Configurations}.
438 @c This list is ordered alphanumerically by subsection name.
439 @c Try and put the significant identifier (CPU or system) first,
440 @c so users have a clue at guessing where the ones they want will be.
443 @gccoptlist{-EB -EL @gol
444 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
445 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
448 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
449 -mabi=@var{name} @gol
450 -mapcs-stack-check -mno-apcs-stack-check @gol
451 -mapcs-float -mno-apcs-float @gol
452 -mapcs-reentrant -mno-apcs-reentrant @gol
453 -msched-prolog -mno-sched-prolog @gol
454 -mlittle-endian -mbig-endian -mwords-little-endian @gol
455 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
456 -mfp16-format=@var{name}
457 -mthumb-interwork -mno-thumb-interwork @gol
458 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
459 -mstructure-size-boundary=@var{n} @gol
460 -mabort-on-noreturn @gol
461 -mlong-calls -mno-long-calls @gol
462 -msingle-pic-base -mno-single-pic-base @gol
463 -mpic-register=@var{reg} @gol
464 -mnop-fun-dllimport @gol
465 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
466 -mpoke-function-name @gol
468 -mtpcs-frame -mtpcs-leaf-frame @gol
469 -mcaller-super-interworking -mcallee-super-interworking @gol
471 -mword-relocations @gol
472 -mfix-cortex-m3-ldrd}
475 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
476 -mcall-prologues -mtiny-stack -mint8}
478 @emph{Blackfin Options}
479 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
480 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
481 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
482 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
483 -mno-id-shared-library -mshared-library-id=@var{n} @gol
484 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
485 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
486 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
490 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
491 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
492 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
493 -mstack-align -mdata-align -mconst-align @gol
494 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
495 -melf -maout -melinux -mlinux -sim -sim2 @gol
496 -mmul-bug-workaround -mno-mul-bug-workaround}
499 @gccoptlist{-mmac -mpush-args}
501 @emph{Darwin Options}
502 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
503 -arch_only -bind_at_load -bundle -bundle_loader @gol
504 -client_name -compatibility_version -current_version @gol
506 -dependency-file -dylib_file -dylinker_install_name @gol
507 -dynamic -dynamiclib -exported_symbols_list @gol
508 -filelist -flat_namespace -force_cpusubtype_ALL @gol
509 -force_flat_namespace -headerpad_max_install_names @gol
511 -image_base -init -install_name -keep_private_externs @gol
512 -multi_module -multiply_defined -multiply_defined_unused @gol
513 -noall_load -no_dead_strip_inits_and_terms @gol
514 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
515 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
516 -private_bundle -read_only_relocs -sectalign @gol
517 -sectobjectsymbols -whyload -seg1addr @gol
518 -sectcreate -sectobjectsymbols -sectorder @gol
519 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
520 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
521 -segprot -segs_read_only_addr -segs_read_write_addr @gol
522 -single_module -static -sub_library -sub_umbrella @gol
523 -twolevel_namespace -umbrella -undefined @gol
524 -unexported_symbols_list -weak_reference_mismatches @gol
525 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
526 -mkernel -mone-byte-bool}
528 @emph{DEC Alpha Options}
529 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
530 -mieee -mieee-with-inexact -mieee-conformant @gol
531 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
532 -mtrap-precision=@var{mode} -mbuild-constants @gol
533 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
534 -mbwx -mmax -mfix -mcix @gol
535 -mfloat-vax -mfloat-ieee @gol
536 -mexplicit-relocs -msmall-data -mlarge-data @gol
537 -msmall-text -mlarge-text @gol
538 -mmemory-latency=@var{time}}
540 @emph{DEC Alpha/VMS Options}
541 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
544 @gccoptlist{-msmall-model -mno-lsim}
547 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
548 -mhard-float -msoft-float @gol
549 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
550 -mdouble -mno-double @gol
551 -mmedia -mno-media -mmuladd -mno-muladd @gol
552 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
553 -mlinked-fp -mlong-calls -malign-labels @gol
554 -mlibrary-pic -macc-4 -macc-8 @gol
555 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
556 -moptimize-membar -mno-optimize-membar @gol
557 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
558 -mvliw-branch -mno-vliw-branch @gol
559 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
560 -mno-nested-cond-exec -mtomcat-stats @gol
564 @emph{GNU/Linux Options}
565 @gccoptlist{-muclibc}
567 @emph{H8/300 Options}
568 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
571 @gccoptlist{-march=@var{architecture-type} @gol
572 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
573 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
574 -mfixed-range=@var{register-range} @gol
575 -mjump-in-delay -mlinker-opt -mlong-calls @gol
576 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
577 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
578 -mno-jump-in-delay -mno-long-load-store @gol
579 -mno-portable-runtime -mno-soft-float @gol
580 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
581 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
582 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
583 -munix=@var{unix-std} -nolibdld -static -threads}
585 @emph{i386 and x86-64 Options}
586 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
587 -mfpmath=@var{unit} @gol
588 -masm=@var{dialect} -mno-fancy-math-387 @gol
589 -mno-fp-ret-in-387 -msoft-float @gol
590 -mno-wide-multiply -mrtd -malign-double @gol
591 -mpreferred-stack-boundary=@var{num}
592 -mincoming-stack-boundary=@var{num}
593 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
594 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
596 -msse4a -m3dnow -mpopcnt -mabm -mfma4 @gol
597 -mthreads -mno-align-stringops -minline-all-stringops @gol
598 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
599 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
600 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
601 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
602 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
603 -mcmodel=@var{code-model} -mabi=@var{name} @gol
604 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
608 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
609 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
610 -mconstant-gp -mauto-pic -mfused-madd @gol
611 -minline-float-divide-min-latency @gol
612 -minline-float-divide-max-throughput @gol
613 -mno-inline-float-divide @gol
614 -minline-int-divide-min-latency @gol
615 -minline-int-divide-max-throughput @gol
616 -mno-inline-int-divide @gol
617 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
618 -mno-inline-sqrt @gol
619 -mdwarf2-asm -mearly-stop-bits @gol
620 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
621 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
622 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
623 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
624 -msched-spec-ldc -msched-spec-control-ldc @gol
625 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
626 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
627 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
628 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
630 @emph{IA-64/VMS Options}
631 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
633 @emph{M32R/D Options}
634 @gccoptlist{-m32r2 -m32rx -m32r @gol
636 -malign-loops -mno-align-loops @gol
637 -missue-rate=@var{number} @gol
638 -mbranch-cost=@var{number} @gol
639 -mmodel=@var{code-size-model-type} @gol
640 -msdata=@var{sdata-type} @gol
641 -mno-flush-func -mflush-func=@var{name} @gol
642 -mno-flush-trap -mflush-trap=@var{number} @gol
646 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
648 @emph{M680x0 Options}
649 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
650 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
651 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
652 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
653 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
654 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
655 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
656 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
659 @emph{M68hc1x Options}
660 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
661 -mauto-incdec -minmax -mlong-calls -mshort @gol
662 -msoft-reg-count=@var{count}}
665 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
666 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
667 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
668 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
669 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
672 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
673 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
674 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
675 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
679 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
680 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
681 -mips64 -mips64r2 @gol
682 -mips16 -mno-mips16 -mflip-mips16 @gol
683 -minterlink-mips16 -mno-interlink-mips16 @gol
684 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
685 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
686 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
687 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
688 -mfpu=@var{fpu-type} @gol
689 -msmartmips -mno-smartmips @gol
690 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
691 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
692 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
693 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
694 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
695 -membedded-data -mno-embedded-data @gol
696 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
697 -mcode-readable=@var{setting} @gol
698 -msplit-addresses -mno-split-addresses @gol
699 -mexplicit-relocs -mno-explicit-relocs @gol
700 -mcheck-zero-division -mno-check-zero-division @gol
701 -mdivide-traps -mdivide-breaks @gol
702 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
703 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
704 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
705 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
706 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
707 -mflush-func=@var{func} -mno-flush-func @gol
708 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
709 -mfp-exceptions -mno-fp-exceptions @gol
710 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
711 -mrelax-pic-calls -mno-relax-pic-calls}
714 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
715 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
716 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
717 -mno-base-addresses -msingle-exit -mno-single-exit}
719 @emph{MN10300 Options}
720 @gccoptlist{-mmult-bug -mno-mult-bug @gol
721 -mam33 -mno-am33 @gol
722 -mam33-2 -mno-am33-2 @gol
723 -mreturn-pointer-on-d0 @gol
726 @emph{PDP-11 Options}
727 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
728 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
729 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
730 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
731 -mbranch-expensive -mbranch-cheap @gol
732 -msplit -mno-split -munix-asm -mdec-asm}
734 @emph{picoChip Options}
735 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
736 -msymbol-as-address -mno-inefficient-warnings}
738 @emph{PowerPC Options}
739 See RS/6000 and PowerPC Options.
741 @emph{RS/6000 and PowerPC Options}
742 @gccoptlist{-mcpu=@var{cpu-type} @gol
743 -mtune=@var{cpu-type} @gol
744 -mpower -mno-power -mpower2 -mno-power2 @gol
745 -mpowerpc -mpowerpc64 -mno-powerpc @gol
746 -maltivec -mno-altivec @gol
747 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
748 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
749 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
750 -mfprnd -mno-fprnd @gol
751 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
752 -mnew-mnemonics -mold-mnemonics @gol
753 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
754 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
755 -malign-power -malign-natural @gol
756 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
757 -msingle-float -mdouble-float -msimple-fpu @gol
758 -mstring -mno-string -mupdate -mno-update @gol
759 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
760 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
761 -mstrict-align -mno-strict-align -mrelocatable @gol
762 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
763 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
764 -mdynamic-no-pic -maltivec -mswdiv @gol
765 -mprioritize-restricted-insns=@var{priority} @gol
766 -msched-costly-dep=@var{dependence_type} @gol
767 -minsert-sched-nops=@var{scheme} @gol
768 -mcall-sysv -mcall-netbsd @gol
769 -maix-struct-return -msvr4-struct-return @gol
770 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
771 -misel -mno-isel @gol
772 -misel=yes -misel=no @gol
774 -mspe=yes -mspe=no @gol
776 -mgen-cell-microcode -mwarn-cell-microcode @gol
777 -mvrsave -mno-vrsave @gol
778 -mmulhw -mno-mulhw @gol
779 -mdlmzb -mno-dlmzb @gol
780 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
781 -mprototype -mno-prototype @gol
782 -msim -mmvme -mads -myellowknife -memb -msdata @gol
783 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
785 @emph{S/390 and zSeries Options}
786 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
787 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
788 -mlong-double-64 -mlong-double-128 @gol
789 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
790 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
791 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
792 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
793 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
796 @gccoptlist{-meb -mel @gol
800 -mscore5 -mscore5u -mscore7 -mscore7d}
803 @gccoptlist{-m1 -m2 -m2e @gol
804 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
806 -m4-nofpu -m4-single-only -m4-single -m4 @gol
807 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
808 -m5-64media -m5-64media-nofpu @gol
809 -m5-32media -m5-32media-nofpu @gol
810 -m5-compact -m5-compact-nofpu @gol
811 -mb -ml -mdalign -mrelax @gol
812 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
813 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
814 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
815 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
816 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
820 @gccoptlist{-mcpu=@var{cpu-type} @gol
821 -mtune=@var{cpu-type} @gol
822 -mcmodel=@var{code-model} @gol
823 -m32 -m64 -mapp-regs -mno-app-regs @gol
824 -mfaster-structs -mno-faster-structs @gol
825 -mfpu -mno-fpu -mhard-float -msoft-float @gol
826 -mhard-quad-float -msoft-quad-float @gol
827 -mimpure-text -mno-impure-text -mlittle-endian @gol
828 -mstack-bias -mno-stack-bias @gol
829 -munaligned-doubles -mno-unaligned-doubles @gol
830 -mv8plus -mno-v8plus -mvis -mno-vis
831 -threads -pthreads -pthread}
834 @gccoptlist{-mwarn-reloc -merror-reloc @gol
835 -msafe-dma -munsafe-dma @gol
837 -msmall-mem -mlarge-mem -mstdmain @gol
838 -mfixed-range=@var{register-range}}
840 @emph{System V Options}
841 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
844 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
845 -mprolog-function -mno-prolog-function -mspace @gol
846 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
847 -mapp-regs -mno-app-regs @gol
848 -mdisable-callt -mno-disable-callt @gol
854 @gccoptlist{-mg -mgnu -munix}
856 @emph{VxWorks Options}
857 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
858 -Xbind-lazy -Xbind-now}
860 @emph{x86-64 Options}
861 See i386 and x86-64 Options.
863 @emph{i386 and x86-64 Windows Options}
864 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
865 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
867 @emph{Xstormy16 Options}
870 @emph{Xtensa Options}
871 @gccoptlist{-mconst16 -mno-const16 @gol
872 -mfused-madd -mno-fused-madd @gol
873 -mserialize-volatile -mno-serialize-volatile @gol
874 -mtext-section-literals -mno-text-section-literals @gol
875 -mtarget-align -mno-target-align @gol
876 -mlongcalls -mno-longcalls}
878 @emph{zSeries Options}
879 See S/390 and zSeries Options.
881 @item Code Generation Options
882 @xref{Code Gen Options,,Options for Code Generation Conventions}.
883 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
884 -ffixed-@var{reg} -fexceptions @gol
885 -fnon-call-exceptions -funwind-tables @gol
886 -fasynchronous-unwind-tables @gol
887 -finhibit-size-directive -finstrument-functions @gol
888 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
889 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
890 -fno-common -fno-ident @gol
891 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
892 -fno-jump-tables @gol
893 -frecord-gcc-switches @gol
894 -freg-struct-return -fshort-enums @gol
895 -fshort-double -fshort-wchar @gol
896 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
897 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
898 -fno-stack-limit -fargument-alias -fargument-noalias @gol
899 -fargument-noalias-global -fargument-noalias-anything @gol
900 -fleading-underscore -ftls-model=@var{model} @gol
901 -ftrapv -fwrapv -fbounds-check @gol
906 * Overall Options:: Controlling the kind of output:
907 an executable, object files, assembler files,
908 or preprocessed source.
909 * C Dialect Options:: Controlling the variant of C language compiled.
910 * C++ Dialect Options:: Variations on C++.
911 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
913 * Language Independent Options:: Controlling how diagnostics should be
915 * Warning Options:: How picky should the compiler be?
916 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
917 * Optimize Options:: How much optimization?
918 * Preprocessor Options:: Controlling header files and macro definitions.
919 Also, getting dependency information for Make.
920 * Assembler Options:: Passing options to the assembler.
921 * Link Options:: Specifying libraries and so on.
922 * Directory Options:: Where to find header files and libraries.
923 Where to find the compiler executable files.
924 * Spec Files:: How to pass switches to sub-processes.
925 * Target Options:: Running a cross-compiler, or an old version of GCC.
928 @node Overall Options
929 @section Options Controlling the Kind of Output
931 Compilation can involve up to four stages: preprocessing, compilation
932 proper, assembly and linking, always in that order. GCC is capable of
933 preprocessing and compiling several files either into several
934 assembler input files, or into one assembler input file; then each
935 assembler input file produces an object file, and linking combines all
936 the object files (those newly compiled, and those specified as input)
937 into an executable file.
939 @cindex file name suffix
940 For any given input file, the file name suffix determines what kind of
945 C source code which must be preprocessed.
948 C source code which should not be preprocessed.
951 C++ source code which should not be preprocessed.
954 Objective-C source code. Note that you must link with the @file{libobjc}
955 library to make an Objective-C program work.
958 Objective-C source code which should not be preprocessed.
962 Objective-C++ source code. Note that you must link with the @file{libobjc}
963 library to make an Objective-C++ program work. Note that @samp{.M} refers
964 to a literal capital M@.
967 Objective-C++ source code which should not be preprocessed.
970 C, C++, Objective-C or Objective-C++ header file to be turned into a
975 @itemx @var{file}.cxx
976 @itemx @var{file}.cpp
977 @itemx @var{file}.CPP
978 @itemx @var{file}.c++
980 C++ source code which must be preprocessed. Note that in @samp{.cxx},
981 the last two letters must both be literally @samp{x}. Likewise,
982 @samp{.C} refers to a literal capital C@.
986 Objective-C++ source code which must be preprocessed.
989 Objective-C++ source code which should not be preprocessed.
994 @itemx @var{file}.hxx
995 @itemx @var{file}.hpp
996 @itemx @var{file}.HPP
997 @itemx @var{file}.h++
998 @itemx @var{file}.tcc
999 C++ header file to be turned into a precompiled header.
1002 @itemx @var{file}.for
1003 @itemx @var{file}.ftn
1004 Fixed form Fortran source code which should not be preprocessed.
1007 @itemx @var{file}.FOR
1008 @itemx @var{file}.fpp
1009 @itemx @var{file}.FPP
1010 @itemx @var{file}.FTN
1011 Fixed form Fortran source code which must be preprocessed (with the traditional
1014 @item @var{file}.f90
1015 @itemx @var{file}.f95
1016 @itemx @var{file}.f03
1017 @itemx @var{file}.f08
1018 Free form Fortran source code which should not be preprocessed.
1020 @item @var{file}.F90
1021 @itemx @var{file}.F95
1022 @itemx @var{file}.F03
1023 @itemx @var{file}.F08
1024 Free form Fortran source code which must be preprocessed (with the
1025 traditional preprocessor).
1027 @c FIXME: Descriptions of Java file types.
1033 @item @var{file}.ads
1034 Ada source code file which contains a library unit declaration (a
1035 declaration of a package, subprogram, or generic, or a generic
1036 instantiation), or a library unit renaming declaration (a package,
1037 generic, or subprogram renaming declaration). Such files are also
1040 @item @var{file}.adb
1041 Ada source code file containing a library unit body (a subprogram or
1042 package body). Such files are also called @dfn{bodies}.
1044 @c GCC also knows about some suffixes for languages not yet included:
1055 @itemx @var{file}.sx
1056 Assembler code which must be preprocessed.
1059 An object file to be fed straight into linking.
1060 Any file name with no recognized suffix is treated this way.
1064 You can specify the input language explicitly with the @option{-x} option:
1067 @item -x @var{language}
1068 Specify explicitly the @var{language} for the following input files
1069 (rather than letting the compiler choose a default based on the file
1070 name suffix). This option applies to all following input files until
1071 the next @option{-x} option. Possible values for @var{language} are:
1073 c c-header c-cpp-output
1074 c++ c++-header c++-cpp-output
1075 objective-c objective-c-header objective-c-cpp-output
1076 objective-c++ objective-c++-header objective-c++-cpp-output
1077 assembler assembler-with-cpp
1079 f77 f77-cpp-input f95 f95-cpp-input
1084 Turn off any specification of a language, so that subsequent files are
1085 handled according to their file name suffixes (as they are if @option{-x}
1086 has not been used at all).
1088 @item -pass-exit-codes
1089 @opindex pass-exit-codes
1090 Normally the @command{gcc} program will exit with the code of 1 if any
1091 phase of the compiler returns a non-success return code. If you specify
1092 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1093 numerically highest error produced by any phase that returned an error
1094 indication. The C, C++, and Fortran frontends return 4, if an internal
1095 compiler error is encountered.
1098 If you only want some of the stages of compilation, you can use
1099 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1100 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1101 @command{gcc} is to stop. Note that some combinations (for example,
1102 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1107 Compile or assemble the source files, but do not link. The linking
1108 stage simply is not done. The ultimate output is in the form of an
1109 object file for each source file.
1111 By default, the object file name for a source file is made by replacing
1112 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1114 Unrecognized input files, not requiring compilation or assembly, are
1119 Stop after the stage of compilation proper; do not assemble. The output
1120 is in the form of an assembler code file for each non-assembler input
1123 By default, the assembler file name for a source file is made by
1124 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1126 Input files that don't require compilation are ignored.
1130 Stop after the preprocessing stage; do not run the compiler proper. The
1131 output is in the form of preprocessed source code, which is sent to the
1134 Input files which don't require preprocessing are ignored.
1136 @cindex output file option
1139 Place output in file @var{file}. This applies regardless to whatever
1140 sort of output is being produced, whether it be an executable file,
1141 an object file, an assembler file or preprocessed C code.
1143 If @option{-o} is not specified, the default is to put an executable
1144 file in @file{a.out}, the object file for
1145 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1146 assembler file in @file{@var{source}.s}, a precompiled header file in
1147 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1152 Print (on standard error output) the commands executed to run the stages
1153 of compilation. Also print the version number of the compiler driver
1154 program and of the preprocessor and the compiler proper.
1158 Like @option{-v} except the commands are not executed and all command
1159 arguments are quoted. This is useful for shell scripts to capture the
1160 driver-generated command lines.
1164 Use pipes rather than temporary files for communication between the
1165 various stages of compilation. This fails to work on some systems where
1166 the assembler is unable to read from a pipe; but the GNU assembler has
1171 If you are compiling multiple source files, this option tells the driver
1172 to pass all the source files to the compiler at once (for those
1173 languages for which the compiler can handle this). This will allow
1174 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1175 language for which this is supported is C@. If you pass source files for
1176 multiple languages to the driver, using this option, the driver will invoke
1177 the compiler(s) that support IMA once each, passing each compiler all the
1178 source files appropriate for it. For those languages that do not support
1179 IMA this option will be ignored, and the compiler will be invoked once for
1180 each source file in that language. If you use this option in conjunction
1181 with @option{-save-temps}, the compiler will generate multiple
1183 (one for each source file), but only one (combined) @file{.o} or
1188 Print (on the standard output) a description of the command line options
1189 understood by @command{gcc}. If the @option{-v} option is also specified
1190 then @option{--help} will also be passed on to the various processes
1191 invoked by @command{gcc}, so that they can display the command line options
1192 they accept. If the @option{-Wextra} option has also been specified
1193 (prior to the @option{--help} option), then command line options which
1194 have no documentation associated with them will also be displayed.
1197 @opindex target-help
1198 Print (on the standard output) a description of target-specific command
1199 line options for each tool. For some targets extra target-specific
1200 information may also be printed.
1202 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1203 Print (on the standard output) a description of the command line
1204 options understood by the compiler that fit into all specified classes
1205 and qualifiers. These are the supported classes:
1208 @item @samp{optimizers}
1209 This will display all of the optimization options supported by the
1212 @item @samp{warnings}
1213 This will display all of the options controlling warning messages
1214 produced by the compiler.
1217 This will display target-specific options. Unlike the
1218 @option{--target-help} option however, target-specific options of the
1219 linker and assembler will not be displayed. This is because those
1220 tools do not currently support the extended @option{--help=} syntax.
1223 This will display the values recognized by the @option{--param}
1226 @item @var{language}
1227 This will display the options supported for @var{language}, where
1228 @var{language} is the name of one of the languages supported in this
1232 This will display the options that are common to all languages.
1235 These are the supported qualifiers:
1238 @item @samp{undocumented}
1239 Display only those options which are undocumented.
1242 Display options which take an argument that appears after an equal
1243 sign in the same continuous piece of text, such as:
1244 @samp{--help=target}.
1246 @item @samp{separate}
1247 Display options which take an argument that appears as a separate word
1248 following the original option, such as: @samp{-o output-file}.
1251 Thus for example to display all the undocumented target-specific
1252 switches supported by the compiler the following can be used:
1255 --help=target,undocumented
1258 The sense of a qualifier can be inverted by prefixing it with the
1259 @samp{^} character, so for example to display all binary warning
1260 options (i.e., ones that are either on or off and that do not take an
1261 argument), which have a description the following can be used:
1264 --help=warnings,^joined,^undocumented
1267 The argument to @option{--help=} should not consist solely of inverted
1270 Combining several classes is possible, although this usually
1271 restricts the output by so much that there is nothing to display. One
1272 case where it does work however is when one of the classes is
1273 @var{target}. So for example to display all the target-specific
1274 optimization options the following can be used:
1277 --help=target,optimizers
1280 The @option{--help=} option can be repeated on the command line. Each
1281 successive use will display its requested class of options, skipping
1282 those that have already been displayed.
1284 If the @option{-Q} option appears on the command line before the
1285 @option{--help=} option, then the descriptive text displayed by
1286 @option{--help=} is changed. Instead of describing the displayed
1287 options, an indication is given as to whether the option is enabled,
1288 disabled or set to a specific value (assuming that the compiler
1289 knows this at the point where the @option{--help=} option is used).
1291 Here is a truncated example from the ARM port of @command{gcc}:
1294 % gcc -Q -mabi=2 --help=target -c
1295 The following options are target specific:
1297 -mabort-on-noreturn [disabled]
1301 The output is sensitive to the effects of previous command line
1302 options, so for example it is possible to find out which optimizations
1303 are enabled at @option{-O2} by using:
1306 -Q -O2 --help=optimizers
1309 Alternatively you can discover which binary optimizations are enabled
1310 by @option{-O3} by using:
1313 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1314 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1315 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1318 @item -no-canonical-prefixes
1319 @opindex no-canonical-prefixes
1320 Do not expand any symbolic links, resolve references to @samp{/../}
1321 or @samp{/./}, or make the path absolute when generating a relative
1326 Display the version number and copyrights of the invoked GCC@.
1330 Invoke all subcommands under a wrapper program. It takes a single
1331 comma separated list as an argument, which will be used to invoke
1335 gcc -c t.c -wrapper gdb,--args
1338 This will invoke all subprograms of gcc under "gdb --args",
1339 thus cc1 invocation will be "gdb --args cc1 ...".
1341 @item -fplugin=@var{name}.so
1342 Load the plugin code in file @var{name}.so, assumed to be a
1343 shared object to be dlopen'd by the compiler. The base name of
1344 the shared object file is used to identify the plugin for the
1345 purposes of argument parsing (See
1346 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1347 Each plugin should define the callback functions specified in the
1350 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1351 Define an argument called @var{key} with a value of @var{value}
1352 for the plugin called @var{name}.
1354 @include @value{srcdir}/../libiberty/at-file.texi
1358 @section Compiling C++ Programs
1360 @cindex suffixes for C++ source
1361 @cindex C++ source file suffixes
1362 C++ source files conventionally use one of the suffixes @samp{.C},
1363 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1364 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1365 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1366 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1367 files with these names and compiles them as C++ programs even if you
1368 call the compiler the same way as for compiling C programs (usually
1369 with the name @command{gcc}).
1373 However, the use of @command{gcc} does not add the C++ library.
1374 @command{g++} is a program that calls GCC and treats @samp{.c},
1375 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1376 files unless @option{-x} is used, and automatically specifies linking
1377 against the C++ library. This program is also useful when
1378 precompiling a C header file with a @samp{.h} extension for use in C++
1379 compilations. On many systems, @command{g++} is also installed with
1380 the name @command{c++}.
1382 @cindex invoking @command{g++}
1383 When you compile C++ programs, you may specify many of the same
1384 command-line options that you use for compiling programs in any
1385 language; or command-line options meaningful for C and related
1386 languages; or options that are meaningful only for C++ programs.
1387 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1388 explanations of options for languages related to C@.
1389 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1390 explanations of options that are meaningful only for C++ programs.
1392 @node C Dialect Options
1393 @section Options Controlling C Dialect
1394 @cindex dialect options
1395 @cindex language dialect options
1396 @cindex options, dialect
1398 The following options control the dialect of C (or languages derived
1399 from C, such as C++, Objective-C and Objective-C++) that the compiler
1403 @cindex ANSI support
1407 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1408 equivalent to @samp{-std=c++98}.
1410 This turns off certain features of GCC that are incompatible with ISO
1411 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1412 such as the @code{asm} and @code{typeof} keywords, and
1413 predefined macros such as @code{unix} and @code{vax} that identify the
1414 type of system you are using. It also enables the undesirable and
1415 rarely used ISO trigraph feature. For the C compiler,
1416 it disables recognition of C++ style @samp{//} comments as well as
1417 the @code{inline} keyword.
1419 The alternate keywords @code{__asm__}, @code{__extension__},
1420 @code{__inline__} and @code{__typeof__} continue to work despite
1421 @option{-ansi}. You would not want to use them in an ISO C program, of
1422 course, but it is useful to put them in header files that might be included
1423 in compilations done with @option{-ansi}. Alternate predefined macros
1424 such as @code{__unix__} and @code{__vax__} are also available, with or
1425 without @option{-ansi}.
1427 The @option{-ansi} option does not cause non-ISO programs to be
1428 rejected gratuitously. For that, @option{-pedantic} is required in
1429 addition to @option{-ansi}. @xref{Warning Options}.
1431 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1432 option is used. Some header files may notice this macro and refrain
1433 from declaring certain functions or defining certain macros that the
1434 ISO standard doesn't call for; this is to avoid interfering with any
1435 programs that might use these names for other things.
1437 Functions that would normally be built in but do not have semantics
1438 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1439 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1440 built-in functions provided by GCC}, for details of the functions
1445 Determine the language standard. @xref{Standards,,Language Standards
1446 Supported by GCC}, for details of these standard versions. This option
1447 is currently only supported when compiling C or C++.
1449 The compiler can accept several base standards, such as @samp{c89} or
1450 @samp{c++98}, and GNU dialects of those standards, such as
1451 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1452 compiler will accept all programs following that standard and those
1453 using GNU extensions that do not contradict it. For example,
1454 @samp{-std=c89} turns off certain features of GCC that are
1455 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1456 keywords, but not other GNU extensions that do not have a meaning in
1457 ISO C90, such as omitting the middle term of a @code{?:}
1458 expression. On the other hand, by specifying a GNU dialect of a
1459 standard, all features the compiler support are enabled, even when
1460 those features change the meaning of the base standard and some
1461 strict-conforming programs may be rejected. The particular standard
1462 is used by @option{-pedantic} to identify which features are GNU
1463 extensions given that version of the standard. For example
1464 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1465 comments, while @samp{-std=gnu99 -pedantic} would not.
1467 A value for this option must be provided; possible values are
1472 Support all ISO C90 programs (certain GNU extensions that conflict
1473 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1475 @item iso9899:199409
1476 ISO C90 as modified in amendment 1.
1482 ISO C99. Note that this standard is not yet fully supported; see
1483 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1484 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1487 GNU dialect of ISO C90 (including some C99 features). This
1488 is the default for C code.
1492 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1493 this will become the default. The name @samp{gnu9x} is deprecated.
1496 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1500 GNU dialect of @option{-std=c++98}. This is the default for
1504 The working draft of the upcoming ISO C++0x standard. This option
1505 enables experimental features that are likely to be included in
1506 C++0x. The working draft is constantly changing, and any feature that is
1507 enabled by this flag may be removed from future versions of GCC if it is
1508 not part of the C++0x standard.
1511 GNU dialect of @option{-std=c++0x}. This option enables
1512 experimental features that may be removed in future versions of GCC.
1515 @item -fgnu89-inline
1516 @opindex fgnu89-inline
1517 The option @option{-fgnu89-inline} tells GCC to use the traditional
1518 GNU semantics for @code{inline} functions when in C99 mode.
1519 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1520 is accepted and ignored by GCC versions 4.1.3 up to but not including
1521 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1522 C99 mode. Using this option is roughly equivalent to adding the
1523 @code{gnu_inline} function attribute to all inline functions
1524 (@pxref{Function Attributes}).
1526 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1527 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1528 specifies the default behavior). This option was first supported in
1529 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1531 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1532 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1533 in effect for @code{inline} functions. @xref{Common Predefined
1534 Macros,,,cpp,The C Preprocessor}.
1536 @item -aux-info @var{filename}
1538 Output to the given filename prototyped declarations for all functions
1539 declared and/or defined in a translation unit, including those in header
1540 files. This option is silently ignored in any language other than C@.
1542 Besides declarations, the file indicates, in comments, the origin of
1543 each declaration (source file and line), whether the declaration was
1544 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1545 @samp{O} for old, respectively, in the first character after the line
1546 number and the colon), and whether it came from a declaration or a
1547 definition (@samp{C} or @samp{F}, respectively, in the following
1548 character). In the case of function definitions, a K&R-style list of
1549 arguments followed by their declarations is also provided, inside
1550 comments, after the declaration.
1554 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1555 keyword, so that code can use these words as identifiers. You can use
1556 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1557 instead. @option{-ansi} implies @option{-fno-asm}.
1559 In C++, this switch only affects the @code{typeof} keyword, since
1560 @code{asm} and @code{inline} are standard keywords. You may want to
1561 use the @option{-fno-gnu-keywords} flag instead, which has the same
1562 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1563 switch only affects the @code{asm} and @code{typeof} keywords, since
1564 @code{inline} is a standard keyword in ISO C99.
1567 @itemx -fno-builtin-@var{function}
1568 @opindex fno-builtin
1569 @cindex built-in functions
1570 Don't recognize built-in functions that do not begin with
1571 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1572 functions provided by GCC}, for details of the functions affected,
1573 including those which are not built-in functions when @option{-ansi} or
1574 @option{-std} options for strict ISO C conformance are used because they
1575 do not have an ISO standard meaning.
1577 GCC normally generates special code to handle certain built-in functions
1578 more efficiently; for instance, calls to @code{alloca} may become single
1579 instructions that adjust the stack directly, and calls to @code{memcpy}
1580 may become inline copy loops. The resulting code is often both smaller
1581 and faster, but since the function calls no longer appear as such, you
1582 cannot set a breakpoint on those calls, nor can you change the behavior
1583 of the functions by linking with a different library. In addition,
1584 when a function is recognized as a built-in function, GCC may use
1585 information about that function to warn about problems with calls to
1586 that function, or to generate more efficient code, even if the
1587 resulting code still contains calls to that function. For example,
1588 warnings are given with @option{-Wformat} for bad calls to
1589 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1590 known not to modify global memory.
1592 With the @option{-fno-builtin-@var{function}} option
1593 only the built-in function @var{function} is
1594 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1595 function is named that is not built-in in this version of GCC, this
1596 option is ignored. There is no corresponding
1597 @option{-fbuiltin-@var{function}} option; if you wish to enable
1598 built-in functions selectively when using @option{-fno-builtin} or
1599 @option{-ffreestanding}, you may define macros such as:
1602 #define abs(n) __builtin_abs ((n))
1603 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1608 @cindex hosted environment
1610 Assert that compilation takes place in a hosted environment. This implies
1611 @option{-fbuiltin}. A hosted environment is one in which the
1612 entire standard library is available, and in which @code{main} has a return
1613 type of @code{int}. Examples are nearly everything except a kernel.
1614 This is equivalent to @option{-fno-freestanding}.
1616 @item -ffreestanding
1617 @opindex ffreestanding
1618 @cindex hosted environment
1620 Assert that compilation takes place in a freestanding environment. This
1621 implies @option{-fno-builtin}. A freestanding environment
1622 is one in which the standard library may not exist, and program startup may
1623 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1624 This is equivalent to @option{-fno-hosted}.
1626 @xref{Standards,,Language Standards Supported by GCC}, for details of
1627 freestanding and hosted environments.
1631 @cindex openmp parallel
1632 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1633 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1634 compiler generates parallel code according to the OpenMP Application
1635 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1636 implies @option{-pthread}, and thus is only supported on targets that
1637 have support for @option{-pthread}.
1639 @item -fms-extensions
1640 @opindex fms-extensions
1641 Accept some non-standard constructs used in Microsoft header files.
1643 Some cases of unnamed fields in structures and unions are only
1644 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1645 fields within structs/unions}, for details.
1649 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1650 options for strict ISO C conformance) implies @option{-trigraphs}.
1652 @item -no-integrated-cpp
1653 @opindex no-integrated-cpp
1654 Performs a compilation in two passes: preprocessing and compiling. This
1655 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1656 @option{-B} option. The user supplied compilation step can then add in
1657 an additional preprocessing step after normal preprocessing but before
1658 compiling. The default is to use the integrated cpp (internal cpp)
1660 The semantics of this option will change if "cc1", "cc1plus", and
1661 "cc1obj" are merged.
1663 @cindex traditional C language
1664 @cindex C language, traditional
1666 @itemx -traditional-cpp
1667 @opindex traditional-cpp
1668 @opindex traditional
1669 Formerly, these options caused GCC to attempt to emulate a pre-standard
1670 C compiler. They are now only supported with the @option{-E} switch.
1671 The preprocessor continues to support a pre-standard mode. See the GNU
1672 CPP manual for details.
1674 @item -fcond-mismatch
1675 @opindex fcond-mismatch
1676 Allow conditional expressions with mismatched types in the second and
1677 third arguments. The value of such an expression is void. This option
1678 is not supported for C++.
1680 @item -flax-vector-conversions
1681 @opindex flax-vector-conversions
1682 Allow implicit conversions between vectors with differing numbers of
1683 elements and/or incompatible element types. This option should not be
1686 @item -funsigned-char
1687 @opindex funsigned-char
1688 Let the type @code{char} be unsigned, like @code{unsigned char}.
1690 Each kind of machine has a default for what @code{char} should
1691 be. It is either like @code{unsigned char} by default or like
1692 @code{signed char} by default.
1694 Ideally, a portable program should always use @code{signed char} or
1695 @code{unsigned char} when it depends on the signedness of an object.
1696 But many programs have been written to use plain @code{char} and
1697 expect it to be signed, or expect it to be unsigned, depending on the
1698 machines they were written for. This option, and its inverse, let you
1699 make such a program work with the opposite default.
1701 The type @code{char} is always a distinct type from each of
1702 @code{signed char} or @code{unsigned char}, even though its behavior
1703 is always just like one of those two.
1706 @opindex fsigned-char
1707 Let the type @code{char} be signed, like @code{signed char}.
1709 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1710 the negative form of @option{-funsigned-char}. Likewise, the option
1711 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1713 @item -fsigned-bitfields
1714 @itemx -funsigned-bitfields
1715 @itemx -fno-signed-bitfields
1716 @itemx -fno-unsigned-bitfields
1717 @opindex fsigned-bitfields
1718 @opindex funsigned-bitfields
1719 @opindex fno-signed-bitfields
1720 @opindex fno-unsigned-bitfields
1721 These options control whether a bit-field is signed or unsigned, when the
1722 declaration does not use either @code{signed} or @code{unsigned}. By
1723 default, such a bit-field is signed, because this is consistent: the
1724 basic integer types such as @code{int} are signed types.
1727 @node C++ Dialect Options
1728 @section Options Controlling C++ Dialect
1730 @cindex compiler options, C++
1731 @cindex C++ options, command line
1732 @cindex options, C++
1733 This section describes the command-line options that are only meaningful
1734 for C++ programs; but you can also use most of the GNU compiler options
1735 regardless of what language your program is in. For example, you
1736 might compile a file @code{firstClass.C} like this:
1739 g++ -g -frepo -O -c firstClass.C
1743 In this example, only @option{-frepo} is an option meant
1744 only for C++ programs; you can use the other options with any
1745 language supported by GCC@.
1747 Here is a list of options that are @emph{only} for compiling C++ programs:
1751 @item -fabi-version=@var{n}
1752 @opindex fabi-version
1753 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1754 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1755 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1756 the version that conforms most closely to the C++ ABI specification.
1757 Therefore, the ABI obtained using version 0 will change as ABI bugs
1760 The default is version 2.
1762 @item -fno-access-control
1763 @opindex fno-access-control
1764 Turn off all access checking. This switch is mainly useful for working
1765 around bugs in the access control code.
1769 Check that the pointer returned by @code{operator new} is non-null
1770 before attempting to modify the storage allocated. This check is
1771 normally unnecessary because the C++ standard specifies that
1772 @code{operator new} will only return @code{0} if it is declared
1773 @samp{throw()}, in which case the compiler will always check the
1774 return value even without this option. In all other cases, when
1775 @code{operator new} has a non-empty exception specification, memory
1776 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1777 @samp{new (nothrow)}.
1779 @item -fconserve-space
1780 @opindex fconserve-space
1781 Put uninitialized or runtime-initialized global variables into the
1782 common segment, as C does. This saves space in the executable at the
1783 cost of not diagnosing duplicate definitions. If you compile with this
1784 flag and your program mysteriously crashes after @code{main()} has
1785 completed, you may have an object that is being destroyed twice because
1786 two definitions were merged.
1788 This option is no longer useful on most targets, now that support has
1789 been added for putting variables into BSS without making them common.
1791 @item -fno-deduce-init-list
1792 @opindex fno-deduce-init-list
1793 Disable deduction of a template type parameter as
1794 std::initializer_list from a brace-enclosed initializer list, i.e.
1797 template <class T> auto forward(T t) -> decltype (realfn (t))
1804 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1808 This option is present because this deduction is an extension to the
1809 current specification in the C++0x working draft, and there was
1810 some concern about potential overload resolution problems.
1812 @item -ffriend-injection
1813 @opindex ffriend-injection
1814 Inject friend functions into the enclosing namespace, so that they are
1815 visible outside the scope of the class in which they are declared.
1816 Friend functions were documented to work this way in the old Annotated
1817 C++ Reference Manual, and versions of G++ before 4.1 always worked
1818 that way. However, in ISO C++ a friend function which is not declared
1819 in an enclosing scope can only be found using argument dependent
1820 lookup. This option causes friends to be injected as they were in
1823 This option is for compatibility, and may be removed in a future
1826 @item -fno-elide-constructors
1827 @opindex fno-elide-constructors
1828 The C++ standard allows an implementation to omit creating a temporary
1829 which is only used to initialize another object of the same type.
1830 Specifying this option disables that optimization, and forces G++ to
1831 call the copy constructor in all cases.
1833 @item -fno-enforce-eh-specs
1834 @opindex fno-enforce-eh-specs
1835 Don't generate code to check for violation of exception specifications
1836 at runtime. This option violates the C++ standard, but may be useful
1837 for reducing code size in production builds, much like defining
1838 @samp{NDEBUG}. This does not give user code permission to throw
1839 exceptions in violation of the exception specifications; the compiler
1840 will still optimize based on the specifications, so throwing an
1841 unexpected exception will result in undefined behavior.
1844 @itemx -fno-for-scope
1846 @opindex fno-for-scope
1847 If @option{-ffor-scope} is specified, the scope of variables declared in
1848 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1849 as specified by the C++ standard.
1850 If @option{-fno-for-scope} is specified, the scope of variables declared in
1851 a @i{for-init-statement} extends to the end of the enclosing scope,
1852 as was the case in old versions of G++, and other (traditional)
1853 implementations of C++.
1855 The default if neither flag is given to follow the standard,
1856 but to allow and give a warning for old-style code that would
1857 otherwise be invalid, or have different behavior.
1859 @item -fno-gnu-keywords
1860 @opindex fno-gnu-keywords
1861 Do not recognize @code{typeof} as a keyword, so that code can use this
1862 word as an identifier. You can use the keyword @code{__typeof__} instead.
1863 @option{-ansi} implies @option{-fno-gnu-keywords}.
1865 @item -fno-implicit-templates
1866 @opindex fno-implicit-templates
1867 Never emit code for non-inline templates which are instantiated
1868 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1869 @xref{Template Instantiation}, for more information.
1871 @item -fno-implicit-inline-templates
1872 @opindex fno-implicit-inline-templates
1873 Don't emit code for implicit instantiations of inline templates, either.
1874 The default is to handle inlines differently so that compiles with and
1875 without optimization will need the same set of explicit instantiations.
1877 @item -fno-implement-inlines
1878 @opindex fno-implement-inlines
1879 To save space, do not emit out-of-line copies of inline functions
1880 controlled by @samp{#pragma implementation}. This will cause linker
1881 errors if these functions are not inlined everywhere they are called.
1883 @item -fms-extensions
1884 @opindex fms-extensions
1885 Disable pedantic warnings about constructs used in MFC, such as implicit
1886 int and getting a pointer to member function via non-standard syntax.
1888 @item -fno-nonansi-builtins
1889 @opindex fno-nonansi-builtins
1890 Disable built-in declarations of functions that are not mandated by
1891 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1892 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1894 @item -fno-operator-names
1895 @opindex fno-operator-names
1896 Do not treat the operator name keywords @code{and}, @code{bitand},
1897 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1898 synonyms as keywords.
1900 @item -fno-optional-diags
1901 @opindex fno-optional-diags
1902 Disable diagnostics that the standard says a compiler does not need to
1903 issue. Currently, the only such diagnostic issued by G++ is the one for
1904 a name having multiple meanings within a class.
1907 @opindex fpermissive
1908 Downgrade some diagnostics about nonconformant code from errors to
1909 warnings. Thus, using @option{-fpermissive} will allow some
1910 nonconforming code to compile.
1912 @item -fno-pretty-templates
1913 @opindex fno-pretty-templates
1914 When an error message refers to a specialization of a function
1915 template, the compiler will normally print the signature of the
1916 template followed by the template arguments and any typedefs or
1917 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1918 rather than @code{void f(int)}) so that it's clear which template is
1919 involved. When an error message refers to a specialization of a class
1920 template, the compiler will omit any template arguments which match
1921 the default template arguments for that template. If either of these
1922 behaviors make it harder to understand the error message rather than
1923 easier, using @option{-fno-pretty-templates} will disable them.
1927 Enable automatic template instantiation at link time. This option also
1928 implies @option{-fno-implicit-templates}. @xref{Template
1929 Instantiation}, for more information.
1933 Disable generation of information about every class with virtual
1934 functions for use by the C++ runtime type identification features
1935 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1936 of the language, you can save some space by using this flag. Note that
1937 exception handling uses the same information, but it will generate it as
1938 needed. The @samp{dynamic_cast} operator can still be used for casts that
1939 do not require runtime type information, i.e.@: casts to @code{void *} or to
1940 unambiguous base classes.
1944 Emit statistics about front-end processing at the end of the compilation.
1945 This information is generally only useful to the G++ development team.
1947 @item -ftemplate-depth-@var{n}
1948 @opindex ftemplate-depth
1949 Set the maximum instantiation depth for template classes to @var{n}.
1950 A limit on the template instantiation depth is needed to detect
1951 endless recursions during template class instantiation. ANSI/ISO C++
1952 conforming programs must not rely on a maximum depth greater than 17
1953 (changed to 1024 in C++0x).
1955 @item -fno-threadsafe-statics
1956 @opindex fno-threadsafe-statics
1957 Do not emit the extra code to use the routines specified in the C++
1958 ABI for thread-safe initialization of local statics. You can use this
1959 option to reduce code size slightly in code that doesn't need to be
1962 @item -fuse-cxa-atexit
1963 @opindex fuse-cxa-atexit
1964 Register destructors for objects with static storage duration with the
1965 @code{__cxa_atexit} function rather than the @code{atexit} function.
1966 This option is required for fully standards-compliant handling of static
1967 destructors, but will only work if your C library supports
1968 @code{__cxa_atexit}.
1970 @item -fno-use-cxa-get-exception-ptr
1971 @opindex fno-use-cxa-get-exception-ptr
1972 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1973 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1974 if the runtime routine is not available.
1976 @item -fvisibility-inlines-hidden
1977 @opindex fvisibility-inlines-hidden
1978 This switch declares that the user does not attempt to compare
1979 pointers to inline methods where the addresses of the two functions
1980 were taken in different shared objects.
1982 The effect of this is that GCC may, effectively, mark inline methods with
1983 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1984 appear in the export table of a DSO and do not require a PLT indirection
1985 when used within the DSO@. Enabling this option can have a dramatic effect
1986 on load and link times of a DSO as it massively reduces the size of the
1987 dynamic export table when the library makes heavy use of templates.
1989 The behavior of this switch is not quite the same as marking the
1990 methods as hidden directly, because it does not affect static variables
1991 local to the function or cause the compiler to deduce that
1992 the function is defined in only one shared object.
1994 You may mark a method as having a visibility explicitly to negate the
1995 effect of the switch for that method. For example, if you do want to
1996 compare pointers to a particular inline method, you might mark it as
1997 having default visibility. Marking the enclosing class with explicit
1998 visibility will have no effect.
2000 Explicitly instantiated inline methods are unaffected by this option
2001 as their linkage might otherwise cross a shared library boundary.
2002 @xref{Template Instantiation}.
2004 @item -fvisibility-ms-compat
2005 @opindex fvisibility-ms-compat
2006 This flag attempts to use visibility settings to make GCC's C++
2007 linkage model compatible with that of Microsoft Visual Studio.
2009 The flag makes these changes to GCC's linkage model:
2013 It sets the default visibility to @code{hidden}, like
2014 @option{-fvisibility=hidden}.
2017 Types, but not their members, are not hidden by default.
2020 The One Definition Rule is relaxed for types without explicit
2021 visibility specifications which are defined in more than one different
2022 shared object: those declarations are permitted if they would have
2023 been permitted when this option was not used.
2026 In new code it is better to use @option{-fvisibility=hidden} and
2027 export those classes which are intended to be externally visible.
2028 Unfortunately it is possible for code to rely, perhaps accidentally,
2029 on the Visual Studio behavior.
2031 Among the consequences of these changes are that static data members
2032 of the same type with the same name but defined in different shared
2033 objects will be different, so changing one will not change the other;
2034 and that pointers to function members defined in different shared
2035 objects may not compare equal. When this flag is given, it is a
2036 violation of the ODR to define types with the same name differently.
2040 Do not use weak symbol support, even if it is provided by the linker.
2041 By default, G++ will use weak symbols if they are available. This
2042 option exists only for testing, and should not be used by end-users;
2043 it will result in inferior code and has no benefits. This option may
2044 be removed in a future release of G++.
2048 Do not search for header files in the standard directories specific to
2049 C++, but do still search the other standard directories. (This option
2050 is used when building the C++ library.)
2053 In addition, these optimization, warning, and code generation options
2054 have meanings only for C++ programs:
2057 @item -fno-default-inline
2058 @opindex fno-default-inline
2059 Do not assume @samp{inline} for functions defined inside a class scope.
2060 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2061 functions will have linkage like inline functions; they just won't be
2064 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2067 Warn when G++ generates code that is probably not compatible with the
2068 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2069 all such cases, there are probably some cases that are not warned about,
2070 even though G++ is generating incompatible code. There may also be
2071 cases where warnings are emitted even though the code that is generated
2074 You should rewrite your code to avoid these warnings if you are
2075 concerned about the fact that code generated by G++ may not be binary
2076 compatible with code generated by other compilers.
2078 The known incompatibilities at this point include:
2083 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2084 pack data into the same byte as a base class. For example:
2087 struct A @{ virtual void f(); int f1 : 1; @};
2088 struct B : public A @{ int f2 : 1; @};
2092 In this case, G++ will place @code{B::f2} into the same byte
2093 as@code{A::f1}; other compilers will not. You can avoid this problem
2094 by explicitly padding @code{A} so that its size is a multiple of the
2095 byte size on your platform; that will cause G++ and other compilers to
2096 layout @code{B} identically.
2099 Incorrect handling of tail-padding for virtual bases. G++ does not use
2100 tail padding when laying out virtual bases. For example:
2103 struct A @{ virtual void f(); char c1; @};
2104 struct B @{ B(); char c2; @};
2105 struct C : public A, public virtual B @{@};
2109 In this case, G++ will not place @code{B} into the tail-padding for
2110 @code{A}; other compilers will. You can avoid this problem by
2111 explicitly padding @code{A} so that its size is a multiple of its
2112 alignment (ignoring virtual base classes); that will cause G++ and other
2113 compilers to layout @code{C} identically.
2116 Incorrect handling of bit-fields with declared widths greater than that
2117 of their underlying types, when the bit-fields appear in a union. For
2121 union U @{ int i : 4096; @};
2125 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2126 union too small by the number of bits in an @code{int}.
2129 Empty classes can be placed at incorrect offsets. For example:
2139 struct C : public B, public A @{@};
2143 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2144 it should be placed at offset zero. G++ mistakenly believes that the
2145 @code{A} data member of @code{B} is already at offset zero.
2148 Names of template functions whose types involve @code{typename} or
2149 template template parameters can be mangled incorrectly.
2152 template <typename Q>
2153 void f(typename Q::X) @{@}
2155 template <template <typename> class Q>
2156 void f(typename Q<int>::X) @{@}
2160 Instantiations of these templates may be mangled incorrectly.
2164 It also warns psABI related changes. The known psABI changes at this
2170 For SYSV/x86-64, when passing union with long double, it is changed to
2171 pass in memory as specified in psABI. For example:
2181 @code{union U} will always be passed in memory.
2185 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2186 @opindex Wctor-dtor-privacy
2187 @opindex Wno-ctor-dtor-privacy
2188 Warn when a class seems unusable because all the constructors or
2189 destructors in that class are private, and it has neither friends nor
2190 public static member functions.
2192 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2193 @opindex Wnon-virtual-dtor
2194 @opindex Wno-non-virtual-dtor
2195 Warn when a class has virtual functions and accessible non-virtual
2196 destructor, in which case it would be possible but unsafe to delete
2197 an instance of a derived class through a pointer to the base class.
2198 This warning is also enabled if -Weffc++ is specified.
2200 @item -Wreorder @r{(C++ and Objective-C++ only)}
2202 @opindex Wno-reorder
2203 @cindex reordering, warning
2204 @cindex warning for reordering of member initializers
2205 Warn when the order of member initializers given in the code does not
2206 match the order in which they must be executed. For instance:
2212 A(): j (0), i (1) @{ @}
2216 The compiler will rearrange the member initializers for @samp{i}
2217 and @samp{j} to match the declaration order of the members, emitting
2218 a warning to that effect. This warning is enabled by @option{-Wall}.
2221 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2224 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2227 Warn about violations of the following style guidelines from Scott Meyers'
2228 @cite{Effective C++} book:
2232 Item 11: Define a copy constructor and an assignment operator for classes
2233 with dynamically allocated memory.
2236 Item 12: Prefer initialization to assignment in constructors.
2239 Item 14: Make destructors virtual in base classes.
2242 Item 15: Have @code{operator=} return a reference to @code{*this}.
2245 Item 23: Don't try to return a reference when you must return an object.
2249 Also warn about violations of the following style guidelines from
2250 Scott Meyers' @cite{More Effective C++} book:
2254 Item 6: Distinguish between prefix and postfix forms of increment and
2255 decrement operators.
2258 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2262 When selecting this option, be aware that the standard library
2263 headers do not obey all of these guidelines; use @samp{grep -v}
2264 to filter out those warnings.
2266 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2267 @opindex Wstrict-null-sentinel
2268 @opindex Wno-strict-null-sentinel
2269 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2270 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2271 to @code{__null}. Although it is a null pointer constant not a null pointer,
2272 it is guaranteed to be of the same size as a pointer. But this use is
2273 not portable across different compilers.
2275 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2276 @opindex Wno-non-template-friend
2277 @opindex Wnon-template-friend
2278 Disable warnings when non-templatized friend functions are declared
2279 within a template. Since the advent of explicit template specification
2280 support in G++, if the name of the friend is an unqualified-id (i.e.,
2281 @samp{friend foo(int)}), the C++ language specification demands that the
2282 friend declare or define an ordinary, nontemplate function. (Section
2283 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2284 could be interpreted as a particular specialization of a templatized
2285 function. Because this non-conforming behavior is no longer the default
2286 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2287 check existing code for potential trouble spots and is on by default.
2288 This new compiler behavior can be turned off with
2289 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2290 but disables the helpful warning.
2292 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2293 @opindex Wold-style-cast
2294 @opindex Wno-old-style-cast
2295 Warn if an old-style (C-style) cast to a non-void type is used within
2296 a C++ program. The new-style casts (@samp{dynamic_cast},
2297 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2298 less vulnerable to unintended effects and much easier to search for.
2300 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2301 @opindex Woverloaded-virtual
2302 @opindex Wno-overloaded-virtual
2303 @cindex overloaded virtual fn, warning
2304 @cindex warning for overloaded virtual fn
2305 Warn when a function declaration hides virtual functions from a
2306 base class. For example, in:
2313 struct B: public A @{
2318 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2326 will fail to compile.
2328 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2329 @opindex Wno-pmf-conversions
2330 @opindex Wpmf-conversions
2331 Disable the diagnostic for converting a bound pointer to member function
2334 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2335 @opindex Wsign-promo
2336 @opindex Wno-sign-promo
2337 Warn when overload resolution chooses a promotion from unsigned or
2338 enumerated type to a signed type, over a conversion to an unsigned type of
2339 the same size. Previous versions of G++ would try to preserve
2340 unsignedness, but the standard mandates the current behavior.
2345 A& operator = (int);
2355 In this example, G++ will synthesize a default @samp{A& operator =
2356 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2359 @node Objective-C and Objective-C++ Dialect Options
2360 @section Options Controlling Objective-C and Objective-C++ Dialects
2362 @cindex compiler options, Objective-C and Objective-C++
2363 @cindex Objective-C and Objective-C++ options, command line
2364 @cindex options, Objective-C and Objective-C++
2365 (NOTE: This manual does not describe the Objective-C and Objective-C++
2366 languages themselves. See @xref{Standards,,Language Standards
2367 Supported by GCC}, for references.)
2369 This section describes the command-line options that are only meaningful
2370 for Objective-C and Objective-C++ programs, but you can also use most of
2371 the language-independent GNU compiler options.
2372 For example, you might compile a file @code{some_class.m} like this:
2375 gcc -g -fgnu-runtime -O -c some_class.m
2379 In this example, @option{-fgnu-runtime} is an option meant only for
2380 Objective-C and Objective-C++ programs; you can use the other options with
2381 any language supported by GCC@.
2383 Note that since Objective-C is an extension of the C language, Objective-C
2384 compilations may also use options specific to the C front-end (e.g.,
2385 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2386 C++-specific options (e.g., @option{-Wabi}).
2388 Here is a list of options that are @emph{only} for compiling Objective-C
2389 and Objective-C++ programs:
2392 @item -fconstant-string-class=@var{class-name}
2393 @opindex fconstant-string-class
2394 Use @var{class-name} as the name of the class to instantiate for each
2395 literal string specified with the syntax @code{@@"@dots{}"}. The default
2396 class name is @code{NXConstantString} if the GNU runtime is being used, and
2397 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2398 @option{-fconstant-cfstrings} option, if also present, will override the
2399 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2400 to be laid out as constant CoreFoundation strings.
2403 @opindex fgnu-runtime
2404 Generate object code compatible with the standard GNU Objective-C
2405 runtime. This is the default for most types of systems.
2407 @item -fnext-runtime
2408 @opindex fnext-runtime
2409 Generate output compatible with the NeXT runtime. This is the default
2410 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2411 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2414 @item -fno-nil-receivers
2415 @opindex fno-nil-receivers
2416 Assume that all Objective-C message dispatches (e.g.,
2417 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2418 is not @code{nil}. This allows for more efficient entry points in the runtime
2419 to be used. Currently, this option is only available in conjunction with
2420 the NeXT runtime on Mac OS X 10.3 and later.
2422 @item -fobjc-call-cxx-cdtors
2423 @opindex fobjc-call-cxx-cdtors
2424 For each Objective-C class, check if any of its instance variables is a
2425 C++ object with a non-trivial default constructor. If so, synthesize a
2426 special @code{- (id) .cxx_construct} instance method that will run
2427 non-trivial default constructors on any such instance variables, in order,
2428 and then return @code{self}. Similarly, check if any instance variable
2429 is a C++ object with a non-trivial destructor, and if so, synthesize a
2430 special @code{- (void) .cxx_destruct} method that will run
2431 all such default destructors, in reverse order.
2433 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2434 thusly generated will only operate on instance variables declared in the
2435 current Objective-C class, and not those inherited from superclasses. It
2436 is the responsibility of the Objective-C runtime to invoke all such methods
2437 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2438 will be invoked by the runtime immediately after a new object
2439 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2440 be invoked immediately before the runtime deallocates an object instance.
2442 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2443 support for invoking the @code{- (id) .cxx_construct} and
2444 @code{- (void) .cxx_destruct} methods.
2446 @item -fobjc-direct-dispatch
2447 @opindex fobjc-direct-dispatch
2448 Allow fast jumps to the message dispatcher. On Darwin this is
2449 accomplished via the comm page.
2451 @item -fobjc-exceptions
2452 @opindex fobjc-exceptions
2453 Enable syntactic support for structured exception handling in Objective-C,
2454 similar to what is offered by C++ and Java. This option is
2455 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2464 @@catch (AnObjCClass *exc) @{
2471 @@catch (AnotherClass *exc) @{
2474 @@catch (id allOthers) @{
2484 The @code{@@throw} statement may appear anywhere in an Objective-C or
2485 Objective-C++ program; when used inside of a @code{@@catch} block, the
2486 @code{@@throw} may appear without an argument (as shown above), in which case
2487 the object caught by the @code{@@catch} will be rethrown.
2489 Note that only (pointers to) Objective-C objects may be thrown and
2490 caught using this scheme. When an object is thrown, it will be caught
2491 by the nearest @code{@@catch} clause capable of handling objects of that type,
2492 analogously to how @code{catch} blocks work in C++ and Java. A
2493 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2494 any and all Objective-C exceptions not caught by previous @code{@@catch}
2497 The @code{@@finally} clause, if present, will be executed upon exit from the
2498 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2499 regardless of whether any exceptions are thrown, caught or rethrown
2500 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2501 of the @code{finally} clause in Java.
2503 There are several caveats to using the new exception mechanism:
2507 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2508 idioms provided by the @code{NSException} class, the new
2509 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2510 systems, due to additional functionality needed in the (NeXT) Objective-C
2514 As mentioned above, the new exceptions do not support handling
2515 types other than Objective-C objects. Furthermore, when used from
2516 Objective-C++, the Objective-C exception model does not interoperate with C++
2517 exceptions at this time. This means you cannot @code{@@throw} an exception
2518 from Objective-C and @code{catch} it in C++, or vice versa
2519 (i.e., @code{throw @dots{} @@catch}).
2522 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2523 blocks for thread-safe execution:
2526 @@synchronized (ObjCClass *guard) @{
2531 Upon entering the @code{@@synchronized} block, a thread of execution shall
2532 first check whether a lock has been placed on the corresponding @code{guard}
2533 object by another thread. If it has, the current thread shall wait until
2534 the other thread relinquishes its lock. Once @code{guard} becomes available,
2535 the current thread will place its own lock on it, execute the code contained in
2536 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2537 making @code{guard} available to other threads).
2539 Unlike Java, Objective-C does not allow for entire methods to be marked
2540 @code{@@synchronized}. Note that throwing exceptions out of
2541 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2542 to be unlocked properly.
2546 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2548 @item -freplace-objc-classes
2549 @opindex freplace-objc-classes
2550 Emit a special marker instructing @command{ld(1)} not to statically link in
2551 the resulting object file, and allow @command{dyld(1)} to load it in at
2552 run time instead. This is used in conjunction with the Fix-and-Continue
2553 debugging mode, where the object file in question may be recompiled and
2554 dynamically reloaded in the course of program execution, without the need
2555 to restart the program itself. Currently, Fix-and-Continue functionality
2556 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2561 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2562 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2563 compile time) with static class references that get initialized at load time,
2564 which improves run-time performance. Specifying the @option{-fzero-link} flag
2565 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2566 to be retained. This is useful in Zero-Link debugging mode, since it allows
2567 for individual class implementations to be modified during program execution.
2571 Dump interface declarations for all classes seen in the source file to a
2572 file named @file{@var{sourcename}.decl}.
2574 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2575 @opindex Wassign-intercept
2576 @opindex Wno-assign-intercept
2577 Warn whenever an Objective-C assignment is being intercepted by the
2580 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2581 @opindex Wno-protocol
2583 If a class is declared to implement a protocol, a warning is issued for
2584 every method in the protocol that is not implemented by the class. The
2585 default behavior is to issue a warning for every method not explicitly
2586 implemented in the class, even if a method implementation is inherited
2587 from the superclass. If you use the @option{-Wno-protocol} option, then
2588 methods inherited from the superclass are considered to be implemented,
2589 and no warning is issued for them.
2591 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2593 @opindex Wno-selector
2594 Warn if multiple methods of different types for the same selector are
2595 found during compilation. The check is performed on the list of methods
2596 in the final stage of compilation. Additionally, a check is performed
2597 for each selector appearing in a @code{@@selector(@dots{})}
2598 expression, and a corresponding method for that selector has been found
2599 during compilation. Because these checks scan the method table only at
2600 the end of compilation, these warnings are not produced if the final
2601 stage of compilation is not reached, for example because an error is
2602 found during compilation, or because the @option{-fsyntax-only} option is
2605 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2606 @opindex Wstrict-selector-match
2607 @opindex Wno-strict-selector-match
2608 Warn if multiple methods with differing argument and/or return types are
2609 found for a given selector when attempting to send a message using this
2610 selector to a receiver of type @code{id} or @code{Class}. When this flag
2611 is off (which is the default behavior), the compiler will omit such warnings
2612 if any differences found are confined to types which share the same size
2615 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2616 @opindex Wundeclared-selector
2617 @opindex Wno-undeclared-selector
2618 Warn if a @code{@@selector(@dots{})} expression referring to an
2619 undeclared selector is found. A selector is considered undeclared if no
2620 method with that name has been declared before the
2621 @code{@@selector(@dots{})} expression, either explicitly in an
2622 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2623 an @code{@@implementation} section. This option always performs its
2624 checks as soon as a @code{@@selector(@dots{})} expression is found,
2625 while @option{-Wselector} only performs its checks in the final stage of
2626 compilation. This also enforces the coding style convention
2627 that methods and selectors must be declared before being used.
2629 @item -print-objc-runtime-info
2630 @opindex print-objc-runtime-info
2631 Generate C header describing the largest structure that is passed by
2636 @node Language Independent Options
2637 @section Options to Control Diagnostic Messages Formatting
2638 @cindex options to control diagnostics formatting
2639 @cindex diagnostic messages
2640 @cindex message formatting
2642 Traditionally, diagnostic messages have been formatted irrespective of
2643 the output device's aspect (e.g.@: its width, @dots{}). The options described
2644 below can be used to control the diagnostic messages formatting
2645 algorithm, e.g.@: how many characters per line, how often source location
2646 information should be reported. Right now, only the C++ front end can
2647 honor these options. However it is expected, in the near future, that
2648 the remaining front ends would be able to digest them correctly.
2651 @item -fmessage-length=@var{n}
2652 @opindex fmessage-length
2653 Try to format error messages so that they fit on lines of about @var{n}
2654 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2655 the front ends supported by GCC@. If @var{n} is zero, then no
2656 line-wrapping will be done; each error message will appear on a single
2659 @opindex fdiagnostics-show-location
2660 @item -fdiagnostics-show-location=once
2661 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2662 reporter to emit @emph{once} source location information; that is, in
2663 case the message is too long to fit on a single physical line and has to
2664 be wrapped, the source location won't be emitted (as prefix) again,
2665 over and over, in subsequent continuation lines. This is the default
2668 @item -fdiagnostics-show-location=every-line
2669 Only meaningful in line-wrapping mode. Instructs the diagnostic
2670 messages reporter to emit the same source location information (as
2671 prefix) for physical lines that result from the process of breaking
2672 a message which is too long to fit on a single line.
2674 @item -fdiagnostics-show-option
2675 @opindex fdiagnostics-show-option
2676 This option instructs the diagnostic machinery to add text to each
2677 diagnostic emitted, which indicates which command line option directly
2678 controls that diagnostic, when such an option is known to the
2679 diagnostic machinery.
2681 @item -Wcoverage-mismatch
2682 @opindex Wcoverage-mismatch
2683 Warn if feedback profiles do not match when using the
2684 @option{-fprofile-use} option.
2685 If a source file was changed between @option{-fprofile-gen} and
2686 @option{-fprofile-use}, the files with the profile feedback can fail
2687 to match the source file and GCC can not use the profile feedback
2688 information. By default, GCC emits an error message in this case.
2689 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2690 error. GCC does not use appropriate feedback profiles, so using this
2691 option can result in poorly optimized code. This option is useful
2692 only in the case of very minor changes such as bug fixes to an
2697 @node Warning Options
2698 @section Options to Request or Suppress Warnings
2699 @cindex options to control warnings
2700 @cindex warning messages
2701 @cindex messages, warning
2702 @cindex suppressing warnings
2704 Warnings are diagnostic messages that report constructions which
2705 are not inherently erroneous but which are risky or suggest there
2706 may have been an error.
2708 The following language-independent options do not enable specific
2709 warnings but control the kinds of diagnostics produced by GCC.
2712 @cindex syntax checking
2714 @opindex fsyntax-only
2715 Check the code for syntax errors, but don't do anything beyond that.
2719 Inhibit all warning messages.
2724 Make all warnings into errors.
2729 Make the specified warning into an error. The specifier for a warning
2730 is appended, for example @option{-Werror=switch} turns the warnings
2731 controlled by @option{-Wswitch} into errors. This switch takes a
2732 negative form, to be used to negate @option{-Werror} for specific
2733 warnings, for example @option{-Wno-error=switch} makes
2734 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2735 is in effect. You can use the @option{-fdiagnostics-show-option}
2736 option to have each controllable warning amended with the option which
2737 controls it, to determine what to use with this option.
2739 Note that specifying @option{-Werror=}@var{foo} automatically implies
2740 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2743 @item -Wfatal-errors
2744 @opindex Wfatal-errors
2745 @opindex Wno-fatal-errors
2746 This option causes the compiler to abort compilation on the first error
2747 occurred rather than trying to keep going and printing further error
2752 You can request many specific warnings with options beginning
2753 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2754 implicit declarations. Each of these specific warning options also
2755 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2756 example, @option{-Wno-implicit}. This manual lists only one of the
2757 two forms, whichever is not the default. For further,
2758 language-specific options also refer to @ref{C++ Dialect Options} and
2759 @ref{Objective-C and Objective-C++ Dialect Options}.
2764 Issue all the warnings demanded by strict ISO C and ISO C++;
2765 reject all programs that use forbidden extensions, and some other
2766 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2767 version of the ISO C standard specified by any @option{-std} option used.
2769 Valid ISO C and ISO C++ programs should compile properly with or without
2770 this option (though a rare few will require @option{-ansi} or a
2771 @option{-std} option specifying the required version of ISO C)@. However,
2772 without this option, certain GNU extensions and traditional C and C++
2773 features are supported as well. With this option, they are rejected.
2775 @option{-pedantic} does not cause warning messages for use of the
2776 alternate keywords whose names begin and end with @samp{__}. Pedantic
2777 warnings are also disabled in the expression that follows
2778 @code{__extension__}. However, only system header files should use
2779 these escape routes; application programs should avoid them.
2780 @xref{Alternate Keywords}.
2782 Some users try to use @option{-pedantic} to check programs for strict ISO
2783 C conformance. They soon find that it does not do quite what they want:
2784 it finds some non-ISO practices, but not all---only those for which
2785 ISO C @emph{requires} a diagnostic, and some others for which
2786 diagnostics have been added.
2788 A feature to report any failure to conform to ISO C might be useful in
2789 some instances, but would require considerable additional work and would
2790 be quite different from @option{-pedantic}. We don't have plans to
2791 support such a feature in the near future.
2793 Where the standard specified with @option{-std} represents a GNU
2794 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2795 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2796 extended dialect is based. Warnings from @option{-pedantic} are given
2797 where they are required by the base standard. (It would not make sense
2798 for such warnings to be given only for features not in the specified GNU
2799 C dialect, since by definition the GNU dialects of C include all
2800 features the compiler supports with the given option, and there would be
2801 nothing to warn about.)
2803 @item -pedantic-errors
2804 @opindex pedantic-errors
2805 Like @option{-pedantic}, except that errors are produced rather than
2811 This enables all the warnings about constructions that some users
2812 consider questionable, and that are easy to avoid (or modify to
2813 prevent the warning), even in conjunction with macros. This also
2814 enables some language-specific warnings described in @ref{C++ Dialect
2815 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2817 @option{-Wall} turns on the following warning flags:
2819 @gccoptlist{-Waddress @gol
2820 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2822 -Wchar-subscripts @gol
2823 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2825 -Wimplicit-function-declaration @gol
2828 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2829 -Wmissing-braces @gol
2835 -Wsequence-point @gol
2836 -Wsign-compare @r{(only in C++)} @gol
2837 -Wstrict-aliasing @gol
2838 -Wstrict-overflow=1 @gol
2841 -Wuninitialized @gol
2842 -Wunknown-pragmas @gol
2843 -Wunused-function @gol
2846 -Wunused-variable @gol
2847 -Wvolatile-register-var @gol
2850 Note that some warning flags are not implied by @option{-Wall}. Some of
2851 them warn about constructions that users generally do not consider
2852 questionable, but which occasionally you might wish to check for;
2853 others warn about constructions that are necessary or hard to avoid in
2854 some cases, and there is no simple way to modify the code to suppress
2855 the warning. Some of them are enabled by @option{-Wextra} but many of
2856 them must be enabled individually.
2862 This enables some extra warning flags that are not enabled by
2863 @option{-Wall}. (This option used to be called @option{-W}. The older
2864 name is still supported, but the newer name is more descriptive.)
2866 @gccoptlist{-Wclobbered @gol
2868 -Wignored-qualifiers @gol
2869 -Wmissing-field-initializers @gol
2870 -Wmissing-parameter-type @r{(C only)} @gol
2871 -Wold-style-declaration @r{(C only)} @gol
2872 -Woverride-init @gol
2875 -Wuninitialized @gol
2876 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2879 The option @option{-Wextra} also prints warning messages for the
2885 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2886 @samp{>}, or @samp{>=}.
2889 (C++ only) An enumerator and a non-enumerator both appear in a
2890 conditional expression.
2893 (C++ only) Ambiguous virtual bases.
2896 (C++ only) Subscripting an array which has been declared @samp{register}.
2899 (C++ only) Taking the address of a variable which has been declared
2903 (C++ only) A base class is not initialized in a derived class' copy
2908 @item -Wchar-subscripts
2909 @opindex Wchar-subscripts
2910 @opindex Wno-char-subscripts
2911 Warn if an array subscript has type @code{char}. This is a common cause
2912 of error, as programmers often forget that this type is signed on some
2914 This warning is enabled by @option{-Wall}.
2918 @opindex Wno-comment
2919 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2920 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2921 This warning is enabled by @option{-Wall}.
2926 @opindex ffreestanding
2927 @opindex fno-builtin
2928 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2929 the arguments supplied have types appropriate to the format string
2930 specified, and that the conversions specified in the format string make
2931 sense. This includes standard functions, and others specified by format
2932 attributes (@pxref{Function Attributes}), in the @code{printf},
2933 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2934 not in the C standard) families (or other target-specific families).
2935 Which functions are checked without format attributes having been
2936 specified depends on the standard version selected, and such checks of
2937 functions without the attribute specified are disabled by
2938 @option{-ffreestanding} or @option{-fno-builtin}.
2940 The formats are checked against the format features supported by GNU
2941 libc version 2.2. These include all ISO C90 and C99 features, as well
2942 as features from the Single Unix Specification and some BSD and GNU
2943 extensions. Other library implementations may not support all these
2944 features; GCC does not support warning about features that go beyond a
2945 particular library's limitations. However, if @option{-pedantic} is used
2946 with @option{-Wformat}, warnings will be given about format features not
2947 in the selected standard version (but not for @code{strfmon} formats,
2948 since those are not in any version of the C standard). @xref{C Dialect
2949 Options,,Options Controlling C Dialect}.
2951 Since @option{-Wformat} also checks for null format arguments for
2952 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2954 @option{-Wformat} is included in @option{-Wall}. For more control over some
2955 aspects of format checking, the options @option{-Wformat-y2k},
2956 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2957 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2958 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2961 @opindex Wformat-y2k
2962 @opindex Wno-format-y2k
2963 If @option{-Wformat} is specified, also warn about @code{strftime}
2964 formats which may yield only a two-digit year.
2966 @item -Wno-format-contains-nul
2967 @opindex Wno-format-contains-nul
2968 @opindex Wformat-contains-nul
2969 If @option{-Wformat} is specified, do not warn about format strings that
2972 @item -Wno-format-extra-args
2973 @opindex Wno-format-extra-args
2974 @opindex Wformat-extra-args
2975 If @option{-Wformat} is specified, do not warn about excess arguments to a
2976 @code{printf} or @code{scanf} format function. The C standard specifies
2977 that such arguments are ignored.
2979 Where the unused arguments lie between used arguments that are
2980 specified with @samp{$} operand number specifications, normally
2981 warnings are still given, since the implementation could not know what
2982 type to pass to @code{va_arg} to skip the unused arguments. However,
2983 in the case of @code{scanf} formats, this option will suppress the
2984 warning if the unused arguments are all pointers, since the Single
2985 Unix Specification says that such unused arguments are allowed.
2987 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2988 @opindex Wno-format-zero-length
2989 @opindex Wformat-zero-length
2990 If @option{-Wformat} is specified, do not warn about zero-length formats.
2991 The C standard specifies that zero-length formats are allowed.
2993 @item -Wformat-nonliteral
2994 @opindex Wformat-nonliteral
2995 @opindex Wno-format-nonliteral
2996 If @option{-Wformat} is specified, also warn if the format string is not a
2997 string literal and so cannot be checked, unless the format function
2998 takes its format arguments as a @code{va_list}.
3000 @item -Wformat-security
3001 @opindex Wformat-security
3002 @opindex Wno-format-security
3003 If @option{-Wformat} is specified, also warn about uses of format
3004 functions that represent possible security problems. At present, this
3005 warns about calls to @code{printf} and @code{scanf} functions where the
3006 format string is not a string literal and there are no format arguments,
3007 as in @code{printf (foo);}. This may be a security hole if the format
3008 string came from untrusted input and contains @samp{%n}. (This is
3009 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3010 in future warnings may be added to @option{-Wformat-security} that are not
3011 included in @option{-Wformat-nonliteral}.)
3015 @opindex Wno-format=2
3016 Enable @option{-Wformat} plus format checks not included in
3017 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3018 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3020 @item -Wnonnull @r{(C and Objective-C only)}
3022 @opindex Wno-nonnull
3023 Warn about passing a null pointer for arguments marked as
3024 requiring a non-null value by the @code{nonnull} function attribute.
3026 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3027 can be disabled with the @option{-Wno-nonnull} option.
3029 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3031 @opindex Wno-init-self
3032 Warn about uninitialized variables which are initialized with themselves.
3033 Note this option can only be used with the @option{-Wuninitialized} option.
3035 For example, GCC will warn about @code{i} being uninitialized in the
3036 following snippet only when @option{-Winit-self} has been specified:
3047 @item -Wimplicit-int @r{(C and Objective-C only)}
3048 @opindex Wimplicit-int
3049 @opindex Wno-implicit-int
3050 Warn when a declaration does not specify a type.
3051 This warning is enabled by @option{-Wall}.
3053 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3054 @opindex Wimplicit-function-declaration
3055 @opindex Wno-implicit-function-declaration
3056 Give a warning whenever a function is used before being declared. In
3057 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3058 enabled by default and it is made into an error by
3059 @option{-pedantic-errors}. This warning is also enabled by
3064 @opindex Wno-implicit
3065 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3066 This warning is enabled by @option{-Wall}.
3068 @item -Wignored-qualifiers @r{(C and C++ only)}
3069 @opindex Wignored-qualifiers
3070 @opindex Wno-ignored-qualifiers
3071 Warn if the return type of a function has a type qualifier
3072 such as @code{const}. For ISO C such a type qualifier has no effect,
3073 since the value returned by a function is not an lvalue.
3074 For C++, the warning is only emitted for scalar types or @code{void}.
3075 ISO C prohibits qualified @code{void} return types on function
3076 definitions, so such return types always receive a warning
3077 even without this option.
3079 This warning is also enabled by @option{-Wextra}.
3084 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3085 a function with external linkage, returning int, taking either zero
3086 arguments, two, or three arguments of appropriate types. This warning
3087 is enabled by default in C++ and is enabled by either @option{-Wall}
3088 or @option{-pedantic}.
3090 @item -Wmissing-braces
3091 @opindex Wmissing-braces
3092 @opindex Wno-missing-braces
3093 Warn if an aggregate or union initializer is not fully bracketed. In
3094 the following example, the initializer for @samp{a} is not fully
3095 bracketed, but that for @samp{b} is fully bracketed.
3098 int a[2][2] = @{ 0, 1, 2, 3 @};
3099 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3102 This warning is enabled by @option{-Wall}.
3104 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3105 @opindex Wmissing-include-dirs
3106 @opindex Wno-missing-include-dirs
3107 Warn if a user-supplied include directory does not exist.
3110 @opindex Wparentheses
3111 @opindex Wno-parentheses
3112 Warn if parentheses are omitted in certain contexts, such
3113 as when there is an assignment in a context where a truth value
3114 is expected, or when operators are nested whose precedence people
3115 often get confused about.
3117 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3118 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3119 interpretation from that of ordinary mathematical notation.
3121 Also warn about constructions where there may be confusion to which
3122 @code{if} statement an @code{else} branch belongs. Here is an example of
3137 In C/C++, every @code{else} branch belongs to the innermost possible
3138 @code{if} statement, which in this example is @code{if (b)}. This is
3139 often not what the programmer expected, as illustrated in the above
3140 example by indentation the programmer chose. When there is the
3141 potential for this confusion, GCC will issue a warning when this flag
3142 is specified. To eliminate the warning, add explicit braces around
3143 the innermost @code{if} statement so there is no way the @code{else}
3144 could belong to the enclosing @code{if}. The resulting code would
3161 This warning is enabled by @option{-Wall}.
3163 @item -Wsequence-point
3164 @opindex Wsequence-point
3165 @opindex Wno-sequence-point
3166 Warn about code that may have undefined semantics because of violations
3167 of sequence point rules in the C and C++ standards.
3169 The C and C++ standards defines the order in which expressions in a C/C++
3170 program are evaluated in terms of @dfn{sequence points}, which represent
3171 a partial ordering between the execution of parts of the program: those
3172 executed before the sequence point, and those executed after it. These
3173 occur after the evaluation of a full expression (one which is not part
3174 of a larger expression), after the evaluation of the first operand of a
3175 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3176 function is called (but after the evaluation of its arguments and the
3177 expression denoting the called function), and in certain other places.
3178 Other than as expressed by the sequence point rules, the order of
3179 evaluation of subexpressions of an expression is not specified. All
3180 these rules describe only a partial order rather than a total order,
3181 since, for example, if two functions are called within one expression
3182 with no sequence point between them, the order in which the functions
3183 are called is not specified. However, the standards committee have
3184 ruled that function calls do not overlap.
3186 It is not specified when between sequence points modifications to the
3187 values of objects take effect. Programs whose behavior depends on this
3188 have undefined behavior; the C and C++ standards specify that ``Between
3189 the previous and next sequence point an object shall have its stored
3190 value modified at most once by the evaluation of an expression.
3191 Furthermore, the prior value shall be read only to determine the value
3192 to be stored.''. If a program breaks these rules, the results on any
3193 particular implementation are entirely unpredictable.
3195 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3196 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3197 diagnosed by this option, and it may give an occasional false positive
3198 result, but in general it has been found fairly effective at detecting
3199 this sort of problem in programs.
3201 The standard is worded confusingly, therefore there is some debate
3202 over the precise meaning of the sequence point rules in subtle cases.
3203 Links to discussions of the problem, including proposed formal
3204 definitions, may be found on the GCC readings page, at
3205 @w{@uref{http://gcc.gnu.org/readings.html}}.
3207 This warning is enabled by @option{-Wall} for C and C++.
3210 @opindex Wreturn-type
3211 @opindex Wno-return-type
3212 Warn whenever a function is defined with a return-type that defaults
3213 to @code{int}. Also warn about any @code{return} statement with no
3214 return-value in a function whose return-type is not @code{void}
3215 (falling off the end of the function body is considered returning
3216 without a value), and about a @code{return} statement with an
3217 expression in a function whose return-type is @code{void}.
3219 For C++, a function without return type always produces a diagnostic
3220 message, even when @option{-Wno-return-type} is specified. The only
3221 exceptions are @samp{main} and functions defined in system headers.
3223 This warning is enabled by @option{-Wall}.
3228 Warn whenever a @code{switch} statement has an index of enumerated type
3229 and lacks a @code{case} for one or more of the named codes of that
3230 enumeration. (The presence of a @code{default} label prevents this
3231 warning.) @code{case} labels outside the enumeration range also
3232 provoke warnings when this option is used (even if there is a
3233 @code{default} label).
3234 This warning is enabled by @option{-Wall}.
3236 @item -Wswitch-default
3237 @opindex Wswitch-default
3238 @opindex Wno-switch-default
3239 Warn whenever a @code{switch} statement does not have a @code{default}
3243 @opindex Wswitch-enum
3244 @opindex Wno-switch-enum
3245 Warn whenever a @code{switch} statement has an index of enumerated type
3246 and lacks a @code{case} for one or more of the named codes of that
3247 enumeration. @code{case} labels outside the enumeration range also
3248 provoke warnings when this option is used. The only difference
3249 between @option{-Wswitch} and this option is that this option gives a
3250 warning about an omitted enumeration code even if there is a
3251 @code{default} label.
3253 @item -Wsync-nand @r{(C and C++ only)}
3255 @opindex Wno-sync-nand
3256 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3257 built-in functions are used. These functions changed semantics in GCC 4.4.
3261 @opindex Wno-trigraphs
3262 Warn if any trigraphs are encountered that might change the meaning of
3263 the program (trigraphs within comments are not warned about).
3264 This warning is enabled by @option{-Wall}.
3266 @item -Wunused-function
3267 @opindex Wunused-function
3268 @opindex Wno-unused-function
3269 Warn whenever a static function is declared but not defined or a
3270 non-inline static function is unused.
3271 This warning is enabled by @option{-Wall}.
3273 @item -Wunused-label
3274 @opindex Wunused-label
3275 @opindex Wno-unused-label
3276 Warn whenever a label is declared but not used.
3277 This warning is enabled by @option{-Wall}.
3279 To suppress this warning use the @samp{unused} attribute
3280 (@pxref{Variable Attributes}).
3282 @item -Wunused-parameter
3283 @opindex Wunused-parameter
3284 @opindex Wno-unused-parameter
3285 Warn whenever a function parameter is unused aside from its declaration.
3287 To suppress this warning use the @samp{unused} attribute
3288 (@pxref{Variable Attributes}).
3290 @item -Wno-unused-result
3291 @opindex Wunused-result
3292 @opindex Wno-unused-result
3293 Do not warn if a caller of a function marked with attribute
3294 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3295 its return value. The default is @option{-Wunused-result}.
3297 @item -Wunused-variable
3298 @opindex Wunused-variable
3299 @opindex Wno-unused-variable
3300 Warn whenever a local variable or non-constant static variable is unused
3301 aside from its declaration.
3302 This warning is enabled by @option{-Wall}.
3304 To suppress this warning use the @samp{unused} attribute
3305 (@pxref{Variable Attributes}).
3307 @item -Wunused-value
3308 @opindex Wunused-value
3309 @opindex Wno-unused-value
3310 Warn whenever a statement computes a result that is explicitly not
3311 used. To suppress this warning cast the unused expression to
3312 @samp{void}. This includes an expression-statement or the left-hand
3313 side of a comma expression that contains no side effects. For example,
3314 an expression such as @samp{x[i,j]} will cause a warning, while
3315 @samp{x[(void)i,j]} will not.
3317 This warning is enabled by @option{-Wall}.
3322 All the above @option{-Wunused} options combined.
3324 In order to get a warning about an unused function parameter, you must
3325 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3326 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3328 @item -Wuninitialized
3329 @opindex Wuninitialized
3330 @opindex Wno-uninitialized
3331 Warn if an automatic variable is used without first being initialized
3332 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3333 warn if a non-static reference or non-static @samp{const} member
3334 appears in a class without constructors.
3336 If you want to warn about code which uses the uninitialized value of the
3337 variable in its own initializer, use the @option{-Winit-self} option.
3339 These warnings occur for individual uninitialized or clobbered
3340 elements of structure, union or array variables as well as for
3341 variables which are uninitialized or clobbered as a whole. They do
3342 not occur for variables or elements declared @code{volatile}. Because
3343 these warnings depend on optimization, the exact variables or elements
3344 for which there are warnings will depend on the precise optimization
3345 options and version of GCC used.
3347 Note that there may be no warning about a variable that is used only
3348 to compute a value that itself is never used, because such
3349 computations may be deleted by data flow analysis before the warnings
3352 These warnings are made optional because GCC is not smart
3353 enough to see all the reasons why the code might be correct
3354 despite appearing to have an error. Here is one example of how
3375 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3376 always initialized, but GCC doesn't know this. Here is
3377 another common case:
3382 if (change_y) save_y = y, y = new_y;
3384 if (change_y) y = save_y;
3389 This has no bug because @code{save_y} is used only if it is set.
3391 @cindex @code{longjmp} warnings
3392 This option also warns when a non-volatile automatic variable might be
3393 changed by a call to @code{longjmp}. These warnings as well are possible
3394 only in optimizing compilation.
3396 The compiler sees only the calls to @code{setjmp}. It cannot know
3397 where @code{longjmp} will be called; in fact, a signal handler could
3398 call it at any point in the code. As a result, you may get a warning
3399 even when there is in fact no problem because @code{longjmp} cannot
3400 in fact be called at the place which would cause a problem.
3402 Some spurious warnings can be avoided if you declare all the functions
3403 you use that never return as @code{noreturn}. @xref{Function
3406 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3408 @item -Wunknown-pragmas
3409 @opindex Wunknown-pragmas
3410 @opindex Wno-unknown-pragmas
3411 @cindex warning for unknown pragmas
3412 @cindex unknown pragmas, warning
3413 @cindex pragmas, warning of unknown
3414 Warn when a #pragma directive is encountered which is not understood by
3415 GCC@. If this command line option is used, warnings will even be issued
3416 for unknown pragmas in system header files. This is not the case if
3417 the warnings were only enabled by the @option{-Wall} command line option.
3420 @opindex Wno-pragmas
3422 Do not warn about misuses of pragmas, such as incorrect parameters,
3423 invalid syntax, or conflicts between pragmas. See also
3424 @samp{-Wunknown-pragmas}.
3426 @item -Wstrict-aliasing
3427 @opindex Wstrict-aliasing
3428 @opindex Wno-strict-aliasing
3429 This option is only active when @option{-fstrict-aliasing} is active.
3430 It warns about code which might break the strict aliasing rules that the
3431 compiler is using for optimization. The warning does not catch all
3432 cases, but does attempt to catch the more common pitfalls. It is
3433 included in @option{-Wall}.
3434 It is equivalent to @option{-Wstrict-aliasing=3}
3436 @item -Wstrict-aliasing=n
3437 @opindex Wstrict-aliasing=n
3438 @opindex Wno-strict-aliasing=n
3439 This option is only active when @option{-fstrict-aliasing} is active.
3440 It warns about code which might break the strict aliasing rules that the
3441 compiler is using for optimization.
3442 Higher levels correspond to higher accuracy (fewer false positives).
3443 Higher levels also correspond to more effort, similar to the way -O works.
3444 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3447 Level 1: Most aggressive, quick, least accurate.
3448 Possibly useful when higher levels
3449 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3450 false negatives. However, it has many false positives.
3451 Warns for all pointer conversions between possibly incompatible types,
3452 even if never dereferenced. Runs in the frontend only.
3454 Level 2: Aggressive, quick, not too precise.
3455 May still have many false positives (not as many as level 1 though),
3456 and few false negatives (but possibly more than level 1).
3457 Unlike level 1, it only warns when an address is taken. Warns about
3458 incomplete types. Runs in the frontend only.
3460 Level 3 (default for @option{-Wstrict-aliasing}):
3461 Should have very few false positives and few false
3462 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3463 Takes care of the common punn+dereference pattern in the frontend:
3464 @code{*(int*)&some_float}.
3465 If optimization is enabled, it also runs in the backend, where it deals
3466 with multiple statement cases using flow-sensitive points-to information.
3467 Only warns when the converted pointer is dereferenced.
3468 Does not warn about incomplete types.
3470 @item -Wstrict-overflow
3471 @itemx -Wstrict-overflow=@var{n}
3472 @opindex Wstrict-overflow
3473 @opindex Wno-strict-overflow
3474 This option is only active when @option{-fstrict-overflow} is active.
3475 It warns about cases where the compiler optimizes based on the
3476 assumption that signed overflow does not occur. Note that it does not
3477 warn about all cases where the code might overflow: it only warns
3478 about cases where the compiler implements some optimization. Thus
3479 this warning depends on the optimization level.
3481 An optimization which assumes that signed overflow does not occur is
3482 perfectly safe if the values of the variables involved are such that
3483 overflow never does, in fact, occur. Therefore this warning can
3484 easily give a false positive: a warning about code which is not
3485 actually a problem. To help focus on important issues, several
3486 warning levels are defined. No warnings are issued for the use of
3487 undefined signed overflow when estimating how many iterations a loop
3488 will require, in particular when determining whether a loop will be
3492 @item -Wstrict-overflow=1
3493 Warn about cases which are both questionable and easy to avoid. For
3494 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3495 compiler will simplify this to @code{1}. This level of
3496 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3497 are not, and must be explicitly requested.
3499 @item -Wstrict-overflow=2
3500 Also warn about other cases where a comparison is simplified to a
3501 constant. For example: @code{abs (x) >= 0}. This can only be
3502 simplified when @option{-fstrict-overflow} is in effect, because
3503 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3504 zero. @option{-Wstrict-overflow} (with no level) is the same as
3505 @option{-Wstrict-overflow=2}.
3507 @item -Wstrict-overflow=3
3508 Also warn about other cases where a comparison is simplified. For
3509 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3511 @item -Wstrict-overflow=4
3512 Also warn about other simplifications not covered by the above cases.
3513 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3515 @item -Wstrict-overflow=5
3516 Also warn about cases where the compiler reduces the magnitude of a
3517 constant involved in a comparison. For example: @code{x + 2 > y} will
3518 be simplified to @code{x + 1 >= y}. This is reported only at the
3519 highest warning level because this simplification applies to many
3520 comparisons, so this warning level will give a very large number of
3524 @item -Warray-bounds
3525 @opindex Wno-array-bounds
3526 @opindex Warray-bounds
3527 This option is only active when @option{-ftree-vrp} is active
3528 (default for -O2 and above). It warns about subscripts to arrays
3529 that are always out of bounds. This warning is enabled by @option{-Wall}.
3531 @item -Wno-div-by-zero
3532 @opindex Wno-div-by-zero
3533 @opindex Wdiv-by-zero
3534 Do not warn about compile-time integer division by zero. Floating point
3535 division by zero is not warned about, as it can be a legitimate way of
3536 obtaining infinities and NaNs.
3538 @item -Wsystem-headers
3539 @opindex Wsystem-headers
3540 @opindex Wno-system-headers
3541 @cindex warnings from system headers
3542 @cindex system headers, warnings from
3543 Print warning messages for constructs found in system header files.
3544 Warnings from system headers are normally suppressed, on the assumption
3545 that they usually do not indicate real problems and would only make the
3546 compiler output harder to read. Using this command line option tells
3547 GCC to emit warnings from system headers as if they occurred in user
3548 code. However, note that using @option{-Wall} in conjunction with this
3549 option will @emph{not} warn about unknown pragmas in system
3550 headers---for that, @option{-Wunknown-pragmas} must also be used.
3553 @opindex Wfloat-equal
3554 @opindex Wno-float-equal
3555 Warn if floating point values are used in equality comparisons.
3557 The idea behind this is that sometimes it is convenient (for the
3558 programmer) to consider floating-point values as approximations to
3559 infinitely precise real numbers. If you are doing this, then you need
3560 to compute (by analyzing the code, or in some other way) the maximum or
3561 likely maximum error that the computation introduces, and allow for it
3562 when performing comparisons (and when producing output, but that's a
3563 different problem). In particular, instead of testing for equality, you
3564 would check to see whether the two values have ranges that overlap; and
3565 this is done with the relational operators, so equality comparisons are
3568 @item -Wtraditional @r{(C and Objective-C only)}
3569 @opindex Wtraditional
3570 @opindex Wno-traditional
3571 Warn about certain constructs that behave differently in traditional and
3572 ISO C@. Also warn about ISO C constructs that have no traditional C
3573 equivalent, and/or problematic constructs which should be avoided.
3577 Macro parameters that appear within string literals in the macro body.
3578 In traditional C macro replacement takes place within string literals,
3579 but does not in ISO C@.
3582 In traditional C, some preprocessor directives did not exist.
3583 Traditional preprocessors would only consider a line to be a directive
3584 if the @samp{#} appeared in column 1 on the line. Therefore
3585 @option{-Wtraditional} warns about directives that traditional C
3586 understands but would ignore because the @samp{#} does not appear as the
3587 first character on the line. It also suggests you hide directives like
3588 @samp{#pragma} not understood by traditional C by indenting them. Some
3589 traditional implementations would not recognize @samp{#elif}, so it
3590 suggests avoiding it altogether.
3593 A function-like macro that appears without arguments.
3596 The unary plus operator.
3599 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3600 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3601 constants.) Note, these suffixes appear in macros defined in the system
3602 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3603 Use of these macros in user code might normally lead to spurious
3604 warnings, however GCC's integrated preprocessor has enough context to
3605 avoid warning in these cases.
3608 A function declared external in one block and then used after the end of
3612 A @code{switch} statement has an operand of type @code{long}.
3615 A non-@code{static} function declaration follows a @code{static} one.
3616 This construct is not accepted by some traditional C compilers.
3619 The ISO type of an integer constant has a different width or
3620 signedness from its traditional type. This warning is only issued if
3621 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3622 typically represent bit patterns, are not warned about.
3625 Usage of ISO string concatenation is detected.
3628 Initialization of automatic aggregates.
3631 Identifier conflicts with labels. Traditional C lacks a separate
3632 namespace for labels.
3635 Initialization of unions. If the initializer is zero, the warning is
3636 omitted. This is done under the assumption that the zero initializer in
3637 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3638 initializer warnings and relies on default initialization to zero in the
3642 Conversions by prototypes between fixed/floating point values and vice
3643 versa. The absence of these prototypes when compiling with traditional
3644 C would cause serious problems. This is a subset of the possible
3645 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3648 Use of ISO C style function definitions. This warning intentionally is
3649 @emph{not} issued for prototype declarations or variadic functions
3650 because these ISO C features will appear in your code when using
3651 libiberty's traditional C compatibility macros, @code{PARAMS} and
3652 @code{VPARAMS}. This warning is also bypassed for nested functions
3653 because that feature is already a GCC extension and thus not relevant to
3654 traditional C compatibility.
3657 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3658 @opindex Wtraditional-conversion
3659 @opindex Wno-traditional-conversion
3660 Warn if a prototype causes a type conversion that is different from what
3661 would happen to the same argument in the absence of a prototype. This
3662 includes conversions of fixed point to floating and vice versa, and
3663 conversions changing the width or signedness of a fixed point argument
3664 except when the same as the default promotion.
3666 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3667 @opindex Wdeclaration-after-statement
3668 @opindex Wno-declaration-after-statement
3669 Warn when a declaration is found after a statement in a block. This
3670 construct, known from C++, was introduced with ISO C99 and is by default
3671 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3672 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3677 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3679 @item -Wno-endif-labels
3680 @opindex Wno-endif-labels
3681 @opindex Wendif-labels
3682 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3687 Warn whenever a local variable shadows another local variable, parameter or
3688 global variable or whenever a built-in function is shadowed.
3690 @item -Wlarger-than=@var{len}
3691 @opindex Wlarger-than=@var{len}
3692 @opindex Wlarger-than-@var{len}
3693 Warn whenever an object of larger than @var{len} bytes is defined.
3695 @item -Wframe-larger-than=@var{len}
3696 @opindex Wframe-larger-than
3697 Warn if the size of a function frame is larger than @var{len} bytes.
3698 The computation done to determine the stack frame size is approximate
3699 and not conservative.
3700 The actual requirements may be somewhat greater than @var{len}
3701 even if you do not get a warning. In addition, any space allocated
3702 via @code{alloca}, variable-length arrays, or related constructs
3703 is not included by the compiler when determining
3704 whether or not to issue a warning.
3706 @item -Wunsafe-loop-optimizations
3707 @opindex Wunsafe-loop-optimizations
3708 @opindex Wno-unsafe-loop-optimizations
3709 Warn if the loop cannot be optimized because the compiler could not
3710 assume anything on the bounds of the loop indices. With
3711 @option{-funsafe-loop-optimizations} warn if the compiler made
3714 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3715 @opindex Wno-pedantic-ms-format
3716 @opindex Wpedantic-ms-format
3717 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3718 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3719 depending on the MS runtime, when you are using the options @option{-Wformat}
3720 and @option{-pedantic} without gnu-extensions.
3722 @item -Wpointer-arith
3723 @opindex Wpointer-arith
3724 @opindex Wno-pointer-arith
3725 Warn about anything that depends on the ``size of'' a function type or
3726 of @code{void}. GNU C assigns these types a size of 1, for
3727 convenience in calculations with @code{void *} pointers and pointers
3728 to functions. In C++, warn also when an arithmetic operation involves
3729 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3732 @opindex Wtype-limits
3733 @opindex Wno-type-limits
3734 Warn if a comparison is always true or always false due to the limited
3735 range of the data type, but do not warn for constant expressions. For
3736 example, warn if an unsigned variable is compared against zero with
3737 @samp{<} or @samp{>=}. This warning is also enabled by
3740 @item -Wbad-function-cast @r{(C and Objective-C only)}
3741 @opindex Wbad-function-cast
3742 @opindex Wno-bad-function-cast
3743 Warn whenever a function call is cast to a non-matching type.
3744 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3746 @item -Wc++-compat @r{(C and Objective-C only)}
3747 Warn about ISO C constructs that are outside of the common subset of
3748 ISO C and ISO C++, e.g.@: request for implicit conversion from
3749 @code{void *} to a pointer to non-@code{void} type.
3751 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3752 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3753 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3754 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3758 @opindex Wno-cast-qual
3759 Warn whenever a pointer is cast so as to remove a type qualifier from
3760 the target type. For example, warn if a @code{const char *} is cast
3761 to an ordinary @code{char *}.
3763 Also warn when making a cast which introduces a type qualifier in an
3764 unsafe way. For example, casting @code{char **} to @code{const char **}
3765 is unsafe, as in this example:
3768 /* p is char ** value. */
3769 const char **q = (const char **) p;
3770 /* Assignment of readonly string to const char * is OK. */
3772 /* Now char** pointer points to read-only memory. */
3777 @opindex Wcast-align
3778 @opindex Wno-cast-align
3779 Warn whenever a pointer is cast such that the required alignment of the
3780 target is increased. For example, warn if a @code{char *} is cast to
3781 an @code{int *} on machines where integers can only be accessed at
3782 two- or four-byte boundaries.
3784 @item -Wwrite-strings
3785 @opindex Wwrite-strings
3786 @opindex Wno-write-strings
3787 When compiling C, give string constants the type @code{const
3788 char[@var{length}]} so that copying the address of one into a
3789 non-@code{const} @code{char *} pointer will get a warning. These
3790 warnings will help you find at compile time code that can try to write
3791 into a string constant, but only if you have been very careful about
3792 using @code{const} in declarations and prototypes. Otherwise, it will
3793 just be a nuisance. This is why we did not make @option{-Wall} request
3796 When compiling C++, warn about the deprecated conversion from string
3797 literals to @code{char *}. This warning is enabled by default for C++
3802 @opindex Wno-clobbered
3803 Warn for variables that might be changed by @samp{longjmp} or
3804 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3807 @opindex Wconversion
3808 @opindex Wno-conversion
3809 Warn for implicit conversions that may alter a value. This includes
3810 conversions between real and integer, like @code{abs (x)} when
3811 @code{x} is @code{double}; conversions between signed and unsigned,
3812 like @code{unsigned ui = -1}; and conversions to smaller types, like
3813 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3814 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3815 changed by the conversion like in @code{abs (2.0)}. Warnings about
3816 conversions between signed and unsigned integers can be disabled by
3817 using @option{-Wno-sign-conversion}.
3819 For C++, also warn for conversions between @code{NULL} and non-pointer
3820 types; confusing overload resolution for user-defined conversions; and
3821 conversions that will never use a type conversion operator:
3822 conversions to @code{void}, the same type, a base class or a reference
3823 to them. Warnings about conversions between signed and unsigned
3824 integers are disabled by default in C++ unless
3825 @option{-Wsign-conversion} is explicitly enabled.
3828 @opindex Wempty-body
3829 @opindex Wno-empty-body
3830 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3831 while} statement. This warning is also enabled by @option{-Wextra}.
3833 @item -Wenum-compare
3834 @opindex Wenum-compare
3835 @opindex Wno-enum-compare
3836 Warn about a comparison between values of different enum types. In C++
3837 this warning is enabled by default. In C this warning is enabled by
3840 @item -Wjump-misses-init @r{(C, Objective-C only)}
3841 @opindex Wjump-misses-init
3842 @opindex Wno-jump-misses-init
3843 Warn if a @code{goto} statement or a @code{switch} statement jumps
3844 forward across the initialization of a variable, or jumps backward to a
3845 label after the variable has been initialized. This only warns about
3846 variables which are initialized when they are declared. This warning is
3847 only supported for C and Objective C; in C++ this sort of branch is an
3850 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3851 can be disabled with the @option{-Wno-jump-misses-init} option.
3853 @item -Wsign-compare
3854 @opindex Wsign-compare
3855 @opindex Wno-sign-compare
3856 @cindex warning for comparison of signed and unsigned values
3857 @cindex comparison of signed and unsigned values, warning
3858 @cindex signed and unsigned values, comparison warning
3859 Warn when a comparison between signed and unsigned values could produce
3860 an incorrect result when the signed value is converted to unsigned.
3861 This warning is also enabled by @option{-Wextra}; to get the other warnings
3862 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3864 @item -Wsign-conversion
3865 @opindex Wsign-conversion
3866 @opindex Wno-sign-conversion
3867 Warn for implicit conversions that may change the sign of an integer
3868 value, like assigning a signed integer expression to an unsigned
3869 integer variable. An explicit cast silences the warning. In C, this
3870 option is enabled also by @option{-Wconversion}.
3874 @opindex Wno-address
3875 Warn about suspicious uses of memory addresses. These include using
3876 the address of a function in a conditional expression, such as
3877 @code{void func(void); if (func)}, and comparisons against the memory
3878 address of a string literal, such as @code{if (x == "abc")}. Such
3879 uses typically indicate a programmer error: the address of a function
3880 always evaluates to true, so their use in a conditional usually
3881 indicate that the programmer forgot the parentheses in a function
3882 call; and comparisons against string literals result in unspecified
3883 behavior and are not portable in C, so they usually indicate that the
3884 programmer intended to use @code{strcmp}. This warning is enabled by
3888 @opindex Wlogical-op
3889 @opindex Wno-logical-op
3890 Warn about suspicious uses of logical operators in expressions.
3891 This includes using logical operators in contexts where a
3892 bit-wise operator is likely to be expected.
3894 @item -Waggregate-return
3895 @opindex Waggregate-return
3896 @opindex Wno-aggregate-return
3897 Warn if any functions that return structures or unions are defined or
3898 called. (In languages where you can return an array, this also elicits
3901 @item -Wno-attributes
3902 @opindex Wno-attributes
3903 @opindex Wattributes
3904 Do not warn if an unexpected @code{__attribute__} is used, such as
3905 unrecognized attributes, function attributes applied to variables,
3906 etc. This will not stop errors for incorrect use of supported
3909 @item -Wno-builtin-macro-redefined
3910 @opindex Wno-builtin-macro-redefined
3911 @opindex Wbuiltin-macro-redefined
3912 Do not warn if certain built-in macros are redefined. This suppresses
3913 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3914 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3916 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3917 @opindex Wstrict-prototypes
3918 @opindex Wno-strict-prototypes
3919 Warn if a function is declared or defined without specifying the
3920 argument types. (An old-style function definition is permitted without
3921 a warning if preceded by a declaration which specifies the argument
3924 @item -Wold-style-declaration @r{(C and Objective-C only)}
3925 @opindex Wold-style-declaration
3926 @opindex Wno-old-style-declaration
3927 Warn for obsolescent usages, according to the C Standard, in a
3928 declaration. For example, warn if storage-class specifiers like
3929 @code{static} are not the first things in a declaration. This warning
3930 is also enabled by @option{-Wextra}.
3932 @item -Wold-style-definition @r{(C and Objective-C only)}
3933 @opindex Wold-style-definition
3934 @opindex Wno-old-style-definition
3935 Warn if an old-style function definition is used. A warning is given
3936 even if there is a previous prototype.
3938 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3939 @opindex Wmissing-parameter-type
3940 @opindex Wno-missing-parameter-type
3941 A function parameter is declared without a type specifier in K&R-style
3948 This warning is also enabled by @option{-Wextra}.
3950 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3951 @opindex Wmissing-prototypes
3952 @opindex Wno-missing-prototypes
3953 Warn if a global function is defined without a previous prototype
3954 declaration. This warning is issued even if the definition itself
3955 provides a prototype. The aim is to detect global functions that fail
3956 to be declared in header files.
3958 @item -Wmissing-declarations
3959 @opindex Wmissing-declarations
3960 @opindex Wno-missing-declarations
3961 Warn if a global function is defined without a previous declaration.
3962 Do so even if the definition itself provides a prototype.
3963 Use this option to detect global functions that are not declared in
3964 header files. In C++, no warnings are issued for function templates,
3965 or for inline functions, or for functions in anonymous namespaces.
3967 @item -Wmissing-field-initializers
3968 @opindex Wmissing-field-initializers
3969 @opindex Wno-missing-field-initializers
3973 Warn if a structure's initializer has some fields missing. For
3974 example, the following code would cause such a warning, because
3975 @code{x.h} is implicitly zero:
3978 struct s @{ int f, g, h; @};
3979 struct s x = @{ 3, 4 @};
3982 This option does not warn about designated initializers, so the following
3983 modification would not trigger a warning:
3986 struct s @{ int f, g, h; @};
3987 struct s x = @{ .f = 3, .g = 4 @};
3990 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3991 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3993 @item -Wmissing-noreturn
3994 @opindex Wmissing-noreturn
3995 @opindex Wno-missing-noreturn
3996 Warn about functions which might be candidates for attribute @code{noreturn}.
3997 Note these are only possible candidates, not absolute ones. Care should
3998 be taken to manually verify functions actually do not ever return before
3999 adding the @code{noreturn} attribute, otherwise subtle code generation
4000 bugs could be introduced. You will not get a warning for @code{main} in
4001 hosted C environments.
4003 @item -Wmissing-format-attribute
4004 @opindex Wmissing-format-attribute
4005 @opindex Wno-missing-format-attribute
4008 Warn about function pointers which might be candidates for @code{format}
4009 attributes. Note these are only possible candidates, not absolute ones.
4010 GCC will guess that function pointers with @code{format} attributes that
4011 are used in assignment, initialization, parameter passing or return
4012 statements should have a corresponding @code{format} attribute in the
4013 resulting type. I.e.@: the left-hand side of the assignment or
4014 initialization, the type of the parameter variable, or the return type
4015 of the containing function respectively should also have a @code{format}
4016 attribute to avoid the warning.
4018 GCC will also warn about function definitions which might be
4019 candidates for @code{format} attributes. Again, these are only
4020 possible candidates. GCC will guess that @code{format} attributes
4021 might be appropriate for any function that calls a function like
4022 @code{vprintf} or @code{vscanf}, but this might not always be the
4023 case, and some functions for which @code{format} attributes are
4024 appropriate may not be detected.
4026 @item -Wno-multichar
4027 @opindex Wno-multichar
4029 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4030 Usually they indicate a typo in the user's code, as they have
4031 implementation-defined values, and should not be used in portable code.
4033 @item -Wnormalized=<none|id|nfc|nfkc>
4034 @opindex Wnormalized=
4037 @cindex character set, input normalization
4038 In ISO C and ISO C++, two identifiers are different if they are
4039 different sequences of characters. However, sometimes when characters
4040 outside the basic ASCII character set are used, you can have two
4041 different character sequences that look the same. To avoid confusion,
4042 the ISO 10646 standard sets out some @dfn{normalization rules} which
4043 when applied ensure that two sequences that look the same are turned into
4044 the same sequence. GCC can warn you if you are using identifiers which
4045 have not been normalized; this option controls that warning.
4047 There are four levels of warning that GCC supports. The default is
4048 @option{-Wnormalized=nfc}, which warns about any identifier which is
4049 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4050 recommended form for most uses.
4052 Unfortunately, there are some characters which ISO C and ISO C++ allow
4053 in identifiers that when turned into NFC aren't allowable as
4054 identifiers. That is, there's no way to use these symbols in portable
4055 ISO C or C++ and have all your identifiers in NFC@.
4056 @option{-Wnormalized=id} suppresses the warning for these characters.
4057 It is hoped that future versions of the standards involved will correct
4058 this, which is why this option is not the default.
4060 You can switch the warning off for all characters by writing
4061 @option{-Wnormalized=none}. You would only want to do this if you
4062 were using some other normalization scheme (like ``D''), because
4063 otherwise you can easily create bugs that are literally impossible to see.
4065 Some characters in ISO 10646 have distinct meanings but look identical
4066 in some fonts or display methodologies, especially once formatting has
4067 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4068 LETTER N'', will display just like a regular @code{n} which has been
4069 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4070 normalization scheme to convert all these into a standard form as
4071 well, and GCC will warn if your code is not in NFKC if you use
4072 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4073 about every identifier that contains the letter O because it might be
4074 confused with the digit 0, and so is not the default, but may be
4075 useful as a local coding convention if the programming environment is
4076 unable to be fixed to display these characters distinctly.
4078 @item -Wno-deprecated
4079 @opindex Wno-deprecated
4080 @opindex Wdeprecated
4081 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4083 @item -Wno-deprecated-declarations
4084 @opindex Wno-deprecated-declarations
4085 @opindex Wdeprecated-declarations
4086 Do not warn about uses of functions (@pxref{Function Attributes}),
4087 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4088 Attributes}) marked as deprecated by using the @code{deprecated}
4092 @opindex Wno-overflow
4094 Do not warn about compile-time overflow in constant expressions.
4096 @item -Woverride-init @r{(C and Objective-C only)}
4097 @opindex Woverride-init
4098 @opindex Wno-override-init
4102 Warn if an initialized field without side effects is overridden when
4103 using designated initializers (@pxref{Designated Inits, , Designated
4106 This warning is included in @option{-Wextra}. To get other
4107 @option{-Wextra} warnings without this one, use @samp{-Wextra
4108 -Wno-override-init}.
4113 Warn if a structure is given the packed attribute, but the packed
4114 attribute has no effect on the layout or size of the structure.
4115 Such structures may be mis-aligned for little benefit. For
4116 instance, in this code, the variable @code{f.x} in @code{struct bar}
4117 will be misaligned even though @code{struct bar} does not itself
4118 have the packed attribute:
4125 @} __attribute__((packed));
4133 @item -Wpacked-bitfield-compat
4134 @opindex Wpacked-bitfield-compat
4135 @opindex Wno-packed-bitfield-compat
4136 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4137 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4138 the change can lead to differences in the structure layout. GCC
4139 informs you when the offset of such a field has changed in GCC 4.4.
4140 For example there is no longer a 4-bit padding between field @code{a}
4141 and @code{b} in this structure:
4148 @} __attribute__ ((packed));
4151 This warning is enabled by default. Use
4152 @option{-Wno-packed-bitfield-compat} to disable this warning.
4157 Warn if padding is included in a structure, either to align an element
4158 of the structure or to align the whole structure. Sometimes when this
4159 happens it is possible to rearrange the fields of the structure to
4160 reduce the padding and so make the structure smaller.
4162 @item -Wredundant-decls
4163 @opindex Wredundant-decls
4164 @opindex Wno-redundant-decls
4165 Warn if anything is declared more than once in the same scope, even in
4166 cases where multiple declaration is valid and changes nothing.
4168 @item -Wnested-externs @r{(C and Objective-C only)}
4169 @opindex Wnested-externs
4170 @opindex Wno-nested-externs
4171 Warn if an @code{extern} declaration is encountered within a function.
4173 @item -Wunreachable-code
4174 @opindex Wunreachable-code
4175 @opindex Wno-unreachable-code
4176 Warn if the compiler detects that code will never be executed.
4178 This option is intended to warn when the compiler detects that at
4179 least a whole line of source code will never be executed, because
4180 some condition is never satisfied or because it is after a
4181 procedure that never returns.
4183 It is possible for this option to produce a warning even though there
4184 are circumstances under which part of the affected line can be executed,
4185 so care should be taken when removing apparently-unreachable code.
4187 For instance, when a function is inlined, a warning may mean that the
4188 line is unreachable in only one inlined copy of the function.
4190 This option is not made part of @option{-Wall} because in a debugging
4191 version of a program there is often substantial code which checks
4192 correct functioning of the program and is, hopefully, unreachable
4193 because the program does work. Another common use of unreachable
4194 code is to provide behavior which is selectable at compile-time.
4199 Warn if a function can not be inlined and it was declared as inline.
4200 Even with this option, the compiler will not warn about failures to
4201 inline functions declared in system headers.
4203 The compiler uses a variety of heuristics to determine whether or not
4204 to inline a function. For example, the compiler takes into account
4205 the size of the function being inlined and the amount of inlining
4206 that has already been done in the current function. Therefore,
4207 seemingly insignificant changes in the source program can cause the
4208 warnings produced by @option{-Winline} to appear or disappear.
4210 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4211 @opindex Wno-invalid-offsetof
4212 @opindex Winvalid-offsetof
4213 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4214 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4215 to a non-POD type is undefined. In existing C++ implementations,
4216 however, @samp{offsetof} typically gives meaningful results even when
4217 applied to certain kinds of non-POD types. (Such as a simple
4218 @samp{struct} that fails to be a POD type only by virtue of having a
4219 constructor.) This flag is for users who are aware that they are
4220 writing nonportable code and who have deliberately chosen to ignore the
4223 The restrictions on @samp{offsetof} may be relaxed in a future version
4224 of the C++ standard.
4226 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4227 @opindex Wno-int-to-pointer-cast
4228 @opindex Wint-to-pointer-cast
4229 Suppress warnings from casts to pointer type of an integer of a
4232 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4233 @opindex Wno-pointer-to-int-cast
4234 @opindex Wpointer-to-int-cast
4235 Suppress warnings from casts from a pointer to an integer type of a
4239 @opindex Winvalid-pch
4240 @opindex Wno-invalid-pch
4241 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4242 the search path but can't be used.
4246 @opindex Wno-long-long
4247 Warn if @samp{long long} type is used. This is enabled by either
4248 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4249 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4251 @item -Wvariadic-macros
4252 @opindex Wvariadic-macros
4253 @opindex Wno-variadic-macros
4254 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4255 alternate syntax when in pedantic ISO C99 mode. This is default.
4256 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4261 Warn if variable length array is used in the code.
4262 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4263 the variable length array.
4265 @item -Wvolatile-register-var
4266 @opindex Wvolatile-register-var
4267 @opindex Wno-volatile-register-var
4268 Warn if a register variable is declared volatile. The volatile
4269 modifier does not inhibit all optimizations that may eliminate reads
4270 and/or writes to register variables. This warning is enabled by
4273 @item -Wdisabled-optimization
4274 @opindex Wdisabled-optimization
4275 @opindex Wno-disabled-optimization
4276 Warn if a requested optimization pass is disabled. This warning does
4277 not generally indicate that there is anything wrong with your code; it
4278 merely indicates that GCC's optimizers were unable to handle the code
4279 effectively. Often, the problem is that your code is too big or too
4280 complex; GCC will refuse to optimize programs when the optimization
4281 itself is likely to take inordinate amounts of time.
4283 @item -Wpointer-sign @r{(C and Objective-C only)}
4284 @opindex Wpointer-sign
4285 @opindex Wno-pointer-sign
4286 Warn for pointer argument passing or assignment with different signedness.
4287 This option is only supported for C and Objective-C@. It is implied by
4288 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4289 @option{-Wno-pointer-sign}.
4291 @item -Wstack-protector
4292 @opindex Wstack-protector
4293 @opindex Wno-stack-protector
4294 This option is only active when @option{-fstack-protector} is active. It
4295 warns about functions that will not be protected against stack smashing.
4298 @opindex Wno-mudflap
4299 Suppress warnings about constructs that cannot be instrumented by
4302 @item -Woverlength-strings
4303 @opindex Woverlength-strings
4304 @opindex Wno-overlength-strings
4305 Warn about string constants which are longer than the ``minimum
4306 maximum'' length specified in the C standard. Modern compilers
4307 generally allow string constants which are much longer than the
4308 standard's minimum limit, but very portable programs should avoid
4309 using longer strings.
4311 The limit applies @emph{after} string constant concatenation, and does
4312 not count the trailing NUL@. In C89, the limit was 509 characters; in
4313 C99, it was raised to 4095. C++98 does not specify a normative
4314 minimum maximum, so we do not diagnose overlength strings in C++@.
4316 This option is implied by @option{-pedantic}, and can be disabled with
4317 @option{-Wno-overlength-strings}.
4319 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4320 @opindex Wunsuffixed-float-constants
4322 GCC will issue a warning for any floating constant that does not have
4323 a suffix. When used together with @option{-Wsystem-headers} it will
4324 warn about such constants in system header files. This can be useful
4325 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4326 from the decimal floating-point extension to C99.
4329 @node Debugging Options
4330 @section Options for Debugging Your Program or GCC
4331 @cindex options, debugging
4332 @cindex debugging information options
4334 GCC has various special options that are used for debugging
4335 either your program or GCC:
4340 Produce debugging information in the operating system's native format
4341 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4344 On most systems that use stabs format, @option{-g} enables use of extra
4345 debugging information that only GDB can use; this extra information
4346 makes debugging work better in GDB but will probably make other debuggers
4348 refuse to read the program. If you want to control for certain whether
4349 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4350 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4352 GCC allows you to use @option{-g} with
4353 @option{-O}. The shortcuts taken by optimized code may occasionally
4354 produce surprising results: some variables you declared may not exist
4355 at all; flow of control may briefly move where you did not expect it;
4356 some statements may not be executed because they compute constant
4357 results or their values were already at hand; some statements may
4358 execute in different places because they were moved out of loops.
4360 Nevertheless it proves possible to debug optimized output. This makes
4361 it reasonable to use the optimizer for programs that might have bugs.
4363 The following options are useful when GCC is generated with the
4364 capability for more than one debugging format.
4368 Produce debugging information for use by GDB@. This means to use the
4369 most expressive format available (DWARF 2, stabs, or the native format
4370 if neither of those are supported), including GDB extensions if at all
4375 Produce debugging information in stabs format (if that is supported),
4376 without GDB extensions. This is the format used by DBX on most BSD
4377 systems. On MIPS, Alpha and System V Release 4 systems this option
4378 produces stabs debugging output which is not understood by DBX or SDB@.
4379 On System V Release 4 systems this option requires the GNU assembler.
4381 @item -feliminate-unused-debug-symbols
4382 @opindex feliminate-unused-debug-symbols
4383 Produce debugging information in stabs format (if that is supported),
4384 for only symbols that are actually used.
4386 @item -femit-class-debug-always
4387 Instead of emitting debugging information for a C++ class in only one
4388 object file, emit it in all object files using the class. This option
4389 should be used only with debuggers that are unable to handle the way GCC
4390 normally emits debugging information for classes because using this
4391 option will increase the size of debugging information by as much as a
4396 Produce debugging information in stabs format (if that is supported),
4397 using GNU extensions understood only by the GNU debugger (GDB)@. The
4398 use of these extensions is likely to make other debuggers crash or
4399 refuse to read the program.
4403 Produce debugging information in COFF format (if that is supported).
4404 This is the format used by SDB on most System V systems prior to
4409 Produce debugging information in XCOFF format (if that is supported).
4410 This is the format used by the DBX debugger on IBM RS/6000 systems.
4414 Produce debugging information in XCOFF format (if that is supported),
4415 using GNU extensions understood only by the GNU debugger (GDB)@. The
4416 use of these extensions is likely to make other debuggers crash or
4417 refuse to read the program, and may cause assemblers other than the GNU
4418 assembler (GAS) to fail with an error.
4420 @item -gdwarf-@var{version}
4421 @opindex gdwarf-@var{version}
4422 Produce debugging information in DWARF format (if that is
4423 supported). This is the format used by DBX on IRIX 6. The value
4424 of @var{version} may be either 2, 3 or 4; the default version is 2.
4426 Note that with DWARF version 2 some ports require, and will always
4427 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4429 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4430 for maximum benefit.
4432 @item -gstrict-dwarf
4433 @opindex gstrict-dwarf
4434 Disallow using extensions of later DWARF standard version than selected
4435 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4436 DWARF extensions from later standard versions is allowed.
4438 @item -gno-strict-dwarf
4439 @opindex gno-strict-dwarf
4440 Allow using extensions of later DWARF standard version than selected with
4441 @option{-gdwarf-@var{version}}.
4445 Produce debugging information in VMS debug format (if that is
4446 supported). This is the format used by DEBUG on VMS systems.
4449 @itemx -ggdb@var{level}
4450 @itemx -gstabs@var{level}
4451 @itemx -gcoff@var{level}
4452 @itemx -gxcoff@var{level}
4453 @itemx -gvms@var{level}
4454 Request debugging information and also use @var{level} to specify how
4455 much information. The default level is 2.
4457 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4460 Level 1 produces minimal information, enough for making backtraces in
4461 parts of the program that you don't plan to debug. This includes
4462 descriptions of functions and external variables, but no information
4463 about local variables and no line numbers.
4465 Level 3 includes extra information, such as all the macro definitions
4466 present in the program. Some debuggers support macro expansion when
4467 you use @option{-g3}.
4469 @option{-gdwarf-2} does not accept a concatenated debug level, because
4470 GCC used to support an option @option{-gdwarf} that meant to generate
4471 debug information in version 1 of the DWARF format (which is very
4472 different from version 2), and it would have been too confusing. That
4473 debug format is long obsolete, but the option cannot be changed now.
4474 Instead use an additional @option{-g@var{level}} option to change the
4475 debug level for DWARF.
4479 Turn off generation of debug info, if leaving out this option would have
4480 generated it, or turn it on at level 2 otherwise. The position of this
4481 argument in the command line does not matter, it takes effect after all
4482 other options are processed, and it does so only once, no matter how
4483 many times it is given. This is mainly intended to be used with
4484 @option{-fcompare-debug}.
4486 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4487 @opindex fdump-final-insns
4488 Dump the final internal representation (RTL) to @var{file}. If the
4489 optional argument is omitted (or if @var{file} is @code{.}), the name
4490 of the dump file will be determined by appending @code{.gkd} to the
4491 compilation output file name.
4493 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4494 @opindex fcompare-debug
4495 @opindex fno-compare-debug
4496 If no error occurs during compilation, run the compiler a second time,
4497 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4498 passed to the second compilation. Dump the final internal
4499 representation in both compilations, and print an error if they differ.
4501 If the equal sign is omitted, the default @option{-gtoggle} is used.
4503 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4504 and nonzero, implicitly enables @option{-fcompare-debug}. If
4505 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4506 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4509 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4510 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4511 of the final representation and the second compilation, preventing even
4512 @env{GCC_COMPARE_DEBUG} from taking effect.
4514 To verify full coverage during @option{-fcompare-debug} testing, set
4515 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4516 which GCC will reject as an invalid option in any actual compilation
4517 (rather than preprocessing, assembly or linking). To get just a
4518 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4519 not overridden} will do.
4521 @item -fcompare-debug-second
4522 @opindex fcompare-debug-second
4523 This option is implicitly passed to the compiler for the second
4524 compilation requested by @option{-fcompare-debug}, along with options to
4525 silence warnings, and omitting other options that would cause
4526 side-effect compiler outputs to files or to the standard output. Dump
4527 files and preserved temporary files are renamed so as to contain the
4528 @code{.gk} additional extension during the second compilation, to avoid
4529 overwriting those generated by the first.
4531 When this option is passed to the compiler driver, it causes the
4532 @emph{first} compilation to be skipped, which makes it useful for little
4533 other than debugging the compiler proper.
4535 @item -feliminate-dwarf2-dups
4536 @opindex feliminate-dwarf2-dups
4537 Compress DWARF2 debugging information by eliminating duplicated
4538 information about each symbol. This option only makes sense when
4539 generating DWARF2 debugging information with @option{-gdwarf-2}.
4541 @item -femit-struct-debug-baseonly
4542 Emit debug information for struct-like types
4543 only when the base name of the compilation source file
4544 matches the base name of file in which the struct was defined.
4546 This option substantially reduces the size of debugging information,
4547 but at significant potential loss in type information to the debugger.
4548 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4549 See @option{-femit-struct-debug-detailed} for more detailed control.
4551 This option works only with DWARF 2.
4553 @item -femit-struct-debug-reduced
4554 Emit debug information for struct-like types
4555 only when the base name of the compilation source file
4556 matches the base name of file in which the type was defined,
4557 unless the struct is a template or defined in a system header.
4559 This option significantly reduces the size of debugging information,
4560 with some potential loss in type information to the debugger.
4561 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4562 See @option{-femit-struct-debug-detailed} for more detailed control.
4564 This option works only with DWARF 2.
4566 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4567 Specify the struct-like types
4568 for which the compiler will generate debug information.
4569 The intent is to reduce duplicate struct debug information
4570 between different object files within the same program.
4572 This option is a detailed version of
4573 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4574 which will serve for most needs.
4576 A specification has the syntax
4577 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4579 The optional first word limits the specification to
4580 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4581 A struct type is used directly when it is the type of a variable, member.
4582 Indirect uses arise through pointers to structs.
4583 That is, when use of an incomplete struct would be legal, the use is indirect.
4585 @samp{struct one direct; struct two * indirect;}.
4587 The optional second word limits the specification to
4588 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4589 Generic structs are a bit complicated to explain.
4590 For C++, these are non-explicit specializations of template classes,
4591 or non-template classes within the above.
4592 Other programming languages have generics,
4593 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4595 The third word specifies the source files for those
4596 structs for which the compiler will emit debug information.
4597 The values @samp{none} and @samp{any} have the normal meaning.
4598 The value @samp{base} means that
4599 the base of name of the file in which the type declaration appears
4600 must match the base of the name of the main compilation file.
4601 In practice, this means that
4602 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4603 but types declared in other header will not.
4604 The value @samp{sys} means those types satisfying @samp{base}
4605 or declared in system or compiler headers.
4607 You may need to experiment to determine the best settings for your application.
4609 The default is @samp{-femit-struct-debug-detailed=all}.
4611 This option works only with DWARF 2.
4613 @item -fenable-icf-debug
4614 @opindex fenable-icf-debug
4615 Generate additional debug information to support identical code folding (ICF).
4616 This option only works with DWARF version 2 or higher.
4618 @item -fno-merge-debug-strings
4619 @opindex fmerge-debug-strings
4620 @opindex fno-merge-debug-strings
4621 Direct the linker to not merge together strings in the debugging
4622 information which are identical in different object files. Merging is
4623 not supported by all assemblers or linkers. Merging decreases the size
4624 of the debug information in the output file at the cost of increasing
4625 link processing time. Merging is enabled by default.
4627 @item -fdebug-prefix-map=@var{old}=@var{new}
4628 @opindex fdebug-prefix-map
4629 When compiling files in directory @file{@var{old}}, record debugging
4630 information describing them as in @file{@var{new}} instead.
4632 @item -fno-dwarf2-cfi-asm
4633 @opindex fdwarf2-cfi-asm
4634 @opindex fno-dwarf2-cfi-asm
4635 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4636 instead of using GAS @code{.cfi_*} directives.
4638 @cindex @command{prof}
4641 Generate extra code to write profile information suitable for the
4642 analysis program @command{prof}. You must use this option when compiling
4643 the source files you want data about, and you must also use it when
4646 @cindex @command{gprof}
4649 Generate extra code to write profile information suitable for the
4650 analysis program @command{gprof}. You must use this option when compiling
4651 the source files you want data about, and you must also use it when
4656 Makes the compiler print out each function name as it is compiled, and
4657 print some statistics about each pass when it finishes.
4660 @opindex ftime-report
4661 Makes the compiler print some statistics about the time consumed by each
4662 pass when it finishes.
4665 @opindex fmem-report
4666 Makes the compiler print some statistics about permanent memory
4667 allocation when it finishes.
4669 @item -fpre-ipa-mem-report
4670 @opindex fpre-ipa-mem-report
4671 @item -fpost-ipa-mem-report
4672 @opindex fpost-ipa-mem-report
4673 Makes the compiler print some statistics about permanent memory
4674 allocation before or after interprocedural optimization.
4676 @item -fprofile-arcs
4677 @opindex fprofile-arcs
4678 Add code so that program flow @dfn{arcs} are instrumented. During
4679 execution the program records how many times each branch and call is
4680 executed and how many times it is taken or returns. When the compiled
4681 program exits it saves this data to a file called
4682 @file{@var{auxname}.gcda} for each source file. The data may be used for
4683 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4684 test coverage analysis (@option{-ftest-coverage}). Each object file's
4685 @var{auxname} is generated from the name of the output file, if
4686 explicitly specified and it is not the final executable, otherwise it is
4687 the basename of the source file. In both cases any suffix is removed
4688 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4689 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4690 @xref{Cross-profiling}.
4692 @cindex @command{gcov}
4696 This option is used to compile and link code instrumented for coverage
4697 analysis. The option is a synonym for @option{-fprofile-arcs}
4698 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4699 linking). See the documentation for those options for more details.
4704 Compile the source files with @option{-fprofile-arcs} plus optimization
4705 and code generation options. For test coverage analysis, use the
4706 additional @option{-ftest-coverage} option. You do not need to profile
4707 every source file in a program.
4710 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4711 (the latter implies the former).
4714 Run the program on a representative workload to generate the arc profile
4715 information. This may be repeated any number of times. You can run
4716 concurrent instances of your program, and provided that the file system
4717 supports locking, the data files will be correctly updated. Also
4718 @code{fork} calls are detected and correctly handled (double counting
4722 For profile-directed optimizations, compile the source files again with
4723 the same optimization and code generation options plus
4724 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4725 Control Optimization}).
4728 For test coverage analysis, use @command{gcov} to produce human readable
4729 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4730 @command{gcov} documentation for further information.
4734 With @option{-fprofile-arcs}, for each function of your program GCC
4735 creates a program flow graph, then finds a spanning tree for the graph.
4736 Only arcs that are not on the spanning tree have to be instrumented: the
4737 compiler adds code to count the number of times that these arcs are
4738 executed. When an arc is the only exit or only entrance to a block, the
4739 instrumentation code can be added to the block; otherwise, a new basic
4740 block must be created to hold the instrumentation code.
4743 @item -ftest-coverage
4744 @opindex ftest-coverage
4745 Produce a notes file that the @command{gcov} code-coverage utility
4746 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4747 show program coverage. Each source file's note file is called
4748 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4749 above for a description of @var{auxname} and instructions on how to
4750 generate test coverage data. Coverage data will match the source files
4751 more closely, if you do not optimize.
4753 @item -fdbg-cnt-list
4754 @opindex fdbg-cnt-list
4755 Print the name and the counter upperbound for all debug counters.
4757 @item -fdbg-cnt=@var{counter-value-list}
4759 Set the internal debug counter upperbound. @var{counter-value-list}
4760 is a comma-separated list of @var{name}:@var{value} pairs
4761 which sets the upperbound of each debug counter @var{name} to @var{value}.
4762 All debug counters have the initial upperbound of @var{UINT_MAX},
4763 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4764 e.g. With -fdbg-cnt=dce:10,tail_call:0
4765 dbg_cnt(dce) will return true only for first 10 invocations
4766 and dbg_cnt(tail_call) will return false always.
4768 @item -d@var{letters}
4769 @itemx -fdump-rtl-@var{pass}
4771 Says to make debugging dumps during compilation at times specified by
4772 @var{letters}. This is used for debugging the RTL-based passes of the
4773 compiler. The file names for most of the dumps are made by appending
4774 a pass number and a word to the @var{dumpname}, and the files are
4775 created in the directory of the output file. @var{dumpname} is
4776 generated from the name of the output file, if explicitly specified
4777 and it is not an executable, otherwise it is the basename of the
4778 source file. These switches may have different effects when
4779 @option{-E} is used for preprocessing.
4781 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4782 @option{-d} option @var{letters}. Here are the possible
4783 letters for use in @var{pass} and @var{letters}, and their meanings:
4787 @item -fdump-rtl-alignments
4788 @opindex fdump-rtl-alignments
4789 Dump after branch alignments have been computed.
4791 @item -fdump-rtl-asmcons
4792 @opindex fdump-rtl-asmcons
4793 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4795 @item -fdump-rtl-auto_inc_dec
4796 @opindex fdump-rtl-auto_inc_dec
4797 Dump after auto-inc-dec discovery. This pass is only run on
4798 architectures that have auto inc or auto dec instructions.
4800 @item -fdump-rtl-barriers
4801 @opindex fdump-rtl-barriers
4802 Dump after cleaning up the barrier instructions.
4804 @item -fdump-rtl-bbpart
4805 @opindex fdump-rtl-bbpart
4806 Dump after partitioning hot and cold basic blocks.
4808 @item -fdump-rtl-bbro
4809 @opindex fdump-rtl-bbro
4810 Dump after block reordering.
4812 @item -fdump-rtl-btl1
4813 @itemx -fdump-rtl-btl2
4814 @opindex fdump-rtl-btl2
4815 @opindex fdump-rtl-btl2
4816 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4817 after the two branch
4818 target load optimization passes.
4820 @item -fdump-rtl-bypass
4821 @opindex fdump-rtl-bypass
4822 Dump after jump bypassing and control flow optimizations.
4824 @item -fdump-rtl-combine
4825 @opindex fdump-rtl-combine
4826 Dump after the RTL instruction combination pass.
4828 @item -fdump-rtl-compgotos
4829 @opindex fdump-rtl-compgotos
4830 Dump after duplicating the computed gotos.
4832 @item -fdump-rtl-ce1
4833 @itemx -fdump-rtl-ce2
4834 @itemx -fdump-rtl-ce3
4835 @opindex fdump-rtl-ce1
4836 @opindex fdump-rtl-ce2
4837 @opindex fdump-rtl-ce3
4838 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4839 @option{-fdump-rtl-ce3} enable dumping after the three
4840 if conversion passes.
4842 @itemx -fdump-rtl-cprop_hardreg
4843 @opindex fdump-rtl-cprop_hardreg
4844 Dump after hard register copy propagation.
4846 @itemx -fdump-rtl-csa
4847 @opindex fdump-rtl-csa
4848 Dump after combining stack adjustments.
4850 @item -fdump-rtl-cse1
4851 @itemx -fdump-rtl-cse2
4852 @opindex fdump-rtl-cse1
4853 @opindex fdump-rtl-cse2
4854 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4855 the two common sub-expression elimination passes.
4857 @itemx -fdump-rtl-dce
4858 @opindex fdump-rtl-dce
4859 Dump after the standalone dead code elimination passes.
4861 @itemx -fdump-rtl-dbr
4862 @opindex fdump-rtl-dbr
4863 Dump after delayed branch scheduling.
4865 @item -fdump-rtl-dce1
4866 @itemx -fdump-rtl-dce2
4867 @opindex fdump-rtl-dce1
4868 @opindex fdump-rtl-dce2
4869 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4870 the two dead store elimination passes.
4873 @opindex fdump-rtl-eh
4874 Dump after finalization of EH handling code.
4876 @item -fdump-rtl-eh_ranges
4877 @opindex fdump-rtl-eh_ranges
4878 Dump after conversion of EH handling range regions.
4880 @item -fdump-rtl-expand
4881 @opindex fdump-rtl-expand
4882 Dump after RTL generation.
4884 @item -fdump-rtl-fwprop1
4885 @itemx -fdump-rtl-fwprop2
4886 @opindex fdump-rtl-fwprop1
4887 @opindex fdump-rtl-fwprop2
4888 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4889 dumping after the two forward propagation passes.
4891 @item -fdump-rtl-gcse1
4892 @itemx -fdump-rtl-gcse2
4893 @opindex fdump-rtl-gcse1
4894 @opindex fdump-rtl-gcse2
4895 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4896 after global common subexpression elimination.
4898 @item -fdump-rtl-init-regs
4899 @opindex fdump-rtl-init-regs
4900 Dump after the initialization of the registers.
4902 @item -fdump-rtl-initvals
4903 @opindex fdump-rtl-initvals
4904 Dump after the computation of the initial value sets.
4906 @itemx -fdump-rtl-into_cfglayout
4907 @opindex fdump-rtl-into_cfglayout
4908 Dump after converting to cfglayout mode.
4910 @item -fdump-rtl-ira
4911 @opindex fdump-rtl-ira
4912 Dump after iterated register allocation.
4914 @item -fdump-rtl-jump
4915 @opindex fdump-rtl-jump
4916 Dump after the second jump optimization.
4918 @item -fdump-rtl-loop2
4919 @opindex fdump-rtl-loop2
4920 @option{-fdump-rtl-loop2} enables dumping after the rtl
4921 loop optimization passes.
4923 @item -fdump-rtl-mach
4924 @opindex fdump-rtl-mach
4925 Dump after performing the machine dependent reorganization pass, if that
4928 @item -fdump-rtl-mode_sw
4929 @opindex fdump-rtl-mode_sw
4930 Dump after removing redundant mode switches.
4932 @item -fdump-rtl-rnreg
4933 @opindex fdump-rtl-rnreg
4934 Dump after register renumbering.
4936 @itemx -fdump-rtl-outof_cfglayout
4937 @opindex fdump-rtl-outof_cfglayout
4938 Dump after converting from cfglayout mode.
4940 @item -fdump-rtl-peephole2
4941 @opindex fdump-rtl-peephole2
4942 Dump after the peephole pass.
4944 @item -fdump-rtl-postreload
4945 @opindex fdump-rtl-postreload
4946 Dump after post-reload optimizations.
4948 @itemx -fdump-rtl-pro_and_epilogue
4949 @opindex fdump-rtl-pro_and_epilogue
4950 Dump after generating the function pro and epilogues.
4952 @item -fdump-rtl-regmove
4953 @opindex fdump-rtl-regmove
4954 Dump after the register move pass.
4956 @item -fdump-rtl-sched1
4957 @itemx -fdump-rtl-sched2
4958 @opindex fdump-rtl-sched1
4959 @opindex fdump-rtl-sched2
4960 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4961 after the basic block scheduling passes.
4963 @item -fdump-rtl-see
4964 @opindex fdump-rtl-see
4965 Dump after sign extension elimination.
4967 @item -fdump-rtl-seqabstr
4968 @opindex fdump-rtl-seqabstr
4969 Dump after common sequence discovery.
4971 @item -fdump-rtl-shorten
4972 @opindex fdump-rtl-shorten
4973 Dump after shortening branches.
4975 @item -fdump-rtl-sibling
4976 @opindex fdump-rtl-sibling
4977 Dump after sibling call optimizations.
4979 @item -fdump-rtl-split1
4980 @itemx -fdump-rtl-split2
4981 @itemx -fdump-rtl-split3
4982 @itemx -fdump-rtl-split4
4983 @itemx -fdump-rtl-split5
4984 @opindex fdump-rtl-split1
4985 @opindex fdump-rtl-split2
4986 @opindex fdump-rtl-split3
4987 @opindex fdump-rtl-split4
4988 @opindex fdump-rtl-split5
4989 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4990 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4991 @option{-fdump-rtl-split5} enable dumping after five rounds of
4992 instruction splitting.
4994 @item -fdump-rtl-sms
4995 @opindex fdump-rtl-sms
4996 Dump after modulo scheduling. This pass is only run on some
4999 @item -fdump-rtl-stack
5000 @opindex fdump-rtl-stack
5001 Dump after conversion from GCC's "flat register file" registers to the
5002 x87's stack-like registers. This pass is only run on x86 variants.
5004 @item -fdump-rtl-subreg1
5005 @itemx -fdump-rtl-subreg2
5006 @opindex fdump-rtl-subreg1
5007 @opindex fdump-rtl-subreg2
5008 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5009 the two subreg expansion passes.
5011 @item -fdump-rtl-unshare
5012 @opindex fdump-rtl-unshare
5013 Dump after all rtl has been unshared.
5015 @item -fdump-rtl-vartrack
5016 @opindex fdump-rtl-vartrack
5017 Dump after variable tracking.
5019 @item -fdump-rtl-vregs
5020 @opindex fdump-rtl-vregs
5021 Dump after converting virtual registers to hard registers.
5023 @item -fdump-rtl-web
5024 @opindex fdump-rtl-web
5025 Dump after live range splitting.
5027 @item -fdump-rtl-regclass
5028 @itemx -fdump-rtl-subregs_of_mode_init
5029 @itemx -fdump-rtl-subregs_of_mode_finish
5030 @itemx -fdump-rtl-dfinit
5031 @itemx -fdump-rtl-dfinish
5032 @opindex fdump-rtl-regclass
5033 @opindex fdump-rtl-subregs_of_mode_init
5034 @opindex fdump-rtl-subregs_of_mode_finish
5035 @opindex fdump-rtl-dfinit
5036 @opindex fdump-rtl-dfinish
5037 These dumps are defined but always produce empty files.
5039 @item -fdump-rtl-all
5040 @opindex fdump-rtl-all
5041 Produce all the dumps listed above.
5045 Annotate the assembler output with miscellaneous debugging information.
5049 Dump all macro definitions, at the end of preprocessing, in addition to
5054 Produce a core dump whenever an error occurs.
5058 Print statistics on memory usage, at the end of the run, to
5063 Annotate the assembler output with a comment indicating which
5064 pattern and alternative was used. The length of each instruction is
5069 Dump the RTL in the assembler output as a comment before each instruction.
5070 Also turns on @option{-dp} annotation.
5074 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5075 dump a representation of the control flow graph suitable for viewing with VCG
5076 to @file{@var{file}.@var{pass}.vcg}.
5080 Just generate RTL for a function instead of compiling it. Usually used
5081 with @option{-fdump-rtl-expand}.
5085 Dump debugging information during parsing, to standard error.
5089 @opindex fdump-noaddr
5090 When doing debugging dumps, suppress address output. This makes it more
5091 feasible to use diff on debugging dumps for compiler invocations with
5092 different compiler binaries and/or different
5093 text / bss / data / heap / stack / dso start locations.
5095 @item -fdump-unnumbered
5096 @opindex fdump-unnumbered
5097 When doing debugging dumps, suppress instruction numbers and address output.
5098 This makes it more feasible to use diff on debugging dumps for compiler
5099 invocations with different options, in particular with and without
5102 @item -fdump-unnumbered-links
5103 @opindex fdump-unnumbered-links
5104 When doing debugging dumps (see @option{-d} option above), suppress
5105 instruction numbers for the links to the previous and next instructions
5108 @item -fdump-translation-unit @r{(C++ only)}
5109 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5110 @opindex fdump-translation-unit
5111 Dump a representation of the tree structure for the entire translation
5112 unit to a file. The file name is made by appending @file{.tu} to the
5113 source file name, and the file is created in the same directory as the
5114 output file. If the @samp{-@var{options}} form is used, @var{options}
5115 controls the details of the dump as described for the
5116 @option{-fdump-tree} options.
5118 @item -fdump-class-hierarchy @r{(C++ only)}
5119 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5120 @opindex fdump-class-hierarchy
5121 Dump a representation of each class's hierarchy and virtual function
5122 table layout to a file. The file name is made by appending
5123 @file{.class} to the source file name, and the file is created in the
5124 same directory as the output file. If the @samp{-@var{options}} form
5125 is used, @var{options} controls the details of the dump as described
5126 for the @option{-fdump-tree} options.
5128 @item -fdump-ipa-@var{switch}
5130 Control the dumping at various stages of inter-procedural analysis
5131 language tree to a file. The file name is generated by appending a
5132 switch specific suffix to the source file name, and the file is created
5133 in the same directory as the output file. The following dumps are
5138 Enables all inter-procedural analysis dumps.
5141 Dumps information about call-graph optimization, unused function removal,
5142 and inlining decisions.
5145 Dump after function inlining.
5149 @item -fdump-statistics-@var{option}
5150 @opindex fdump-statistics
5151 Enable and control dumping of pass statistics in a separate file. The
5152 file name is generated by appending a suffix ending in
5153 @samp{.statistics} to the source file name, and the file is created in
5154 the same directory as the output file. If the @samp{-@var{option}}
5155 form is used, @samp{-stats} will cause counters to be summed over the
5156 whole compilation unit while @samp{-details} will dump every event as
5157 the passes generate them. The default with no option is to sum
5158 counters for each function compiled.
5160 @item -fdump-tree-@var{switch}
5161 @itemx -fdump-tree-@var{switch}-@var{options}
5163 Control the dumping at various stages of processing the intermediate
5164 language tree to a file. The file name is generated by appending a
5165 switch specific suffix to the source file name, and the file is
5166 created in the same directory as the output file. If the
5167 @samp{-@var{options}} form is used, @var{options} is a list of
5168 @samp{-} separated options that control the details of the dump. Not
5169 all options are applicable to all dumps, those which are not
5170 meaningful will be ignored. The following options are available
5174 Print the address of each node. Usually this is not meaningful as it
5175 changes according to the environment and source file. Its primary use
5176 is for tying up a dump file with a debug environment.
5178 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5179 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5180 use working backward from mangled names in the assembly file.
5182 Inhibit dumping of members of a scope or body of a function merely
5183 because that scope has been reached. Only dump such items when they
5184 are directly reachable by some other path. When dumping pretty-printed
5185 trees, this option inhibits dumping the bodies of control structures.
5187 Print a raw representation of the tree. By default, trees are
5188 pretty-printed into a C-like representation.
5190 Enable more detailed dumps (not honored by every dump option).
5192 Enable dumping various statistics about the pass (not honored by every dump
5195 Enable showing basic block boundaries (disabled in raw dumps).
5197 Enable showing virtual operands for every statement.
5199 Enable showing line numbers for statements.
5201 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5203 Enable showing the tree dump for each statement.
5205 Enable showing the EH region number holding each statement.
5207 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5208 and @option{lineno}.
5211 The following tree dumps are possible:
5215 @opindex fdump-tree-original
5216 Dump before any tree based optimization, to @file{@var{file}.original}.
5219 @opindex fdump-tree-optimized
5220 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5223 @opindex fdump-tree-gimple
5224 Dump each function before and after the gimplification pass to a file. The
5225 file name is made by appending @file{.gimple} to the source file name.
5228 @opindex fdump-tree-cfg
5229 Dump the control flow graph of each function to a file. The file name is
5230 made by appending @file{.cfg} to the source file name.
5233 @opindex fdump-tree-vcg
5234 Dump the control flow graph of each function to a file in VCG format. The
5235 file name is made by appending @file{.vcg} to the source file name. Note
5236 that if the file contains more than one function, the generated file cannot
5237 be used directly by VCG@. You will need to cut and paste each function's
5238 graph into its own separate file first.
5241 @opindex fdump-tree-ch
5242 Dump each function after copying loop headers. The file name is made by
5243 appending @file{.ch} to the source file name.
5246 @opindex fdump-tree-ssa
5247 Dump SSA related information to a file. The file name is made by appending
5248 @file{.ssa} to the source file name.
5251 @opindex fdump-tree-alias
5252 Dump aliasing information for each function. The file name is made by
5253 appending @file{.alias} to the source file name.
5256 @opindex fdump-tree-ccp
5257 Dump each function after CCP@. The file name is made by appending
5258 @file{.ccp} to the source file name.
5261 @opindex fdump-tree-storeccp
5262 Dump each function after STORE-CCP@. The file name is made by appending
5263 @file{.storeccp} to the source file name.
5266 @opindex fdump-tree-pre
5267 Dump trees after partial redundancy elimination. The file name is made
5268 by appending @file{.pre} to the source file name.
5271 @opindex fdump-tree-fre
5272 Dump trees after full redundancy elimination. The file name is made
5273 by appending @file{.fre} to the source file name.
5276 @opindex fdump-tree-copyprop
5277 Dump trees after copy propagation. The file name is made
5278 by appending @file{.copyprop} to the source file name.
5280 @item store_copyprop
5281 @opindex fdump-tree-store_copyprop
5282 Dump trees after store copy-propagation. The file name is made
5283 by appending @file{.store_copyprop} to the source file name.
5286 @opindex fdump-tree-dce
5287 Dump each function after dead code elimination. The file name is made by
5288 appending @file{.dce} to the source file name.
5291 @opindex fdump-tree-mudflap
5292 Dump each function after adding mudflap instrumentation. The file name is
5293 made by appending @file{.mudflap} to the source file name.
5296 @opindex fdump-tree-sra
5297 Dump each function after performing scalar replacement of aggregates. The
5298 file name is made by appending @file{.sra} to the source file name.
5301 @opindex fdump-tree-sink
5302 Dump each function after performing code sinking. The file name is made
5303 by appending @file{.sink} to the source file name.
5306 @opindex fdump-tree-dom
5307 Dump each function after applying dominator tree optimizations. The file
5308 name is made by appending @file{.dom} to the source file name.
5311 @opindex fdump-tree-dse
5312 Dump each function after applying dead store elimination. The file
5313 name is made by appending @file{.dse} to the source file name.
5316 @opindex fdump-tree-phiopt
5317 Dump each function after optimizing PHI nodes into straightline code. The file
5318 name is made by appending @file{.phiopt} to the source file name.
5321 @opindex fdump-tree-forwprop
5322 Dump each function after forward propagating single use variables. The file
5323 name is made by appending @file{.forwprop} to the source file name.
5326 @opindex fdump-tree-copyrename
5327 Dump each function after applying the copy rename optimization. The file
5328 name is made by appending @file{.copyrename} to the source file name.
5331 @opindex fdump-tree-nrv
5332 Dump each function after applying the named return value optimization on
5333 generic trees. The file name is made by appending @file{.nrv} to the source
5337 @opindex fdump-tree-vect
5338 Dump each function after applying vectorization of loops. The file name is
5339 made by appending @file{.vect} to the source file name.
5342 @opindex fdump-tree-vrp
5343 Dump each function after Value Range Propagation (VRP). The file name
5344 is made by appending @file{.vrp} to the source file name.
5347 @opindex fdump-tree-all
5348 Enable all the available tree dumps with the flags provided in this option.
5351 @item -ftree-vectorizer-verbose=@var{n}
5352 @opindex ftree-vectorizer-verbose
5353 This option controls the amount of debugging output the vectorizer prints.
5354 This information is written to standard error, unless
5355 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5356 in which case it is output to the usual dump listing file, @file{.vect}.
5357 For @var{n}=0 no diagnostic information is reported.
5358 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5359 and the total number of loops that got vectorized.
5360 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5361 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5362 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5363 level that @option{-fdump-tree-vect-stats} uses.
5364 Higher verbosity levels mean either more information dumped for each
5365 reported loop, or same amount of information reported for more loops:
5366 If @var{n}=3, alignment related information is added to the reports.
5367 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5368 memory access-patterns) is added to the reports.
5369 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5370 that did not pass the first analysis phase (i.e., may not be countable, or
5371 may have complicated control-flow).
5372 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5373 For @var{n}=7, all the information the vectorizer generates during its
5374 analysis and transformation is reported. This is the same verbosity level
5375 that @option{-fdump-tree-vect-details} uses.
5377 @item -frandom-seed=@var{string}
5378 @opindex frandom-seed
5379 This option provides a seed that GCC uses when it would otherwise use
5380 random numbers. It is used to generate certain symbol names
5381 that have to be different in every compiled file. It is also used to
5382 place unique stamps in coverage data files and the object files that
5383 produce them. You can use the @option{-frandom-seed} option to produce
5384 reproducibly identical object files.
5386 The @var{string} should be different for every file you compile.
5388 @item -fsched-verbose=@var{n}
5389 @opindex fsched-verbose
5390 On targets that use instruction scheduling, this option controls the
5391 amount of debugging output the scheduler prints. This information is
5392 written to standard error, unless @option{-fdump-rtl-sched1} or
5393 @option{-fdump-rtl-sched2} is specified, in which case it is output
5394 to the usual dump listing file, @file{.sched} or @file{.sched2}
5395 respectively. However for @var{n} greater than nine, the output is
5396 always printed to standard error.
5398 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5399 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5400 For @var{n} greater than one, it also output basic block probabilities,
5401 detailed ready list information and unit/insn info. For @var{n} greater
5402 than two, it includes RTL at abort point, control-flow and regions info.
5403 And for @var{n} over four, @option{-fsched-verbose} also includes
5407 @itemx -save-temps=cwd
5409 Store the usual ``temporary'' intermediate files permanently; place them
5410 in the current directory and name them based on the source file. Thus,
5411 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5412 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5413 preprocessed @file{foo.i} output file even though the compiler now
5414 normally uses an integrated preprocessor.
5416 When used in combination with the @option{-x} command line option,
5417 @option{-save-temps} is sensible enough to avoid over writing an
5418 input source file with the same extension as an intermediate file.
5419 The corresponding intermediate file may be obtained by renaming the
5420 source file before using @option{-save-temps}.
5422 If you invoke GCC in parallel, compiling several different source
5423 files that share a common base name in different subdirectories or the
5424 same source file compiled for multiple output destinations, it is
5425 likely that the different parallel compilers will interfere with each
5426 other, and overwrite the temporary files. For instance:
5429 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5430 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5433 may result in @file{foo.i} and @file{foo.o} being written to
5434 simultaneously by both compilers.
5436 @item -save-temps=obj
5437 @opindex save-temps=obj
5438 Store the usual ``temporary'' intermediate files permanently. If the
5439 @option{-o} option is used, the temporary files are based on the
5440 object file. If the @option{-o} option is not used, the
5441 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5446 gcc -save-temps=obj -c foo.c
5447 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5448 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5451 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5452 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5453 @file{dir2/yfoobar.o}.
5455 @item -time@r{[}=@var{file}@r{]}
5457 Report the CPU time taken by each subprocess in the compilation
5458 sequence. For C source files, this is the compiler proper and assembler
5459 (plus the linker if linking is done).
5461 Without the specification of an output file, the output looks like this:
5468 The first number on each line is the ``user time'', that is time spent
5469 executing the program itself. The second number is ``system time'',
5470 time spent executing operating system routines on behalf of the program.
5471 Both numbers are in seconds.
5473 With the specification of an output file, the output is appended to the
5474 named file, and it looks like this:
5477 0.12 0.01 cc1 @var{options}
5478 0.00 0.01 as @var{options}
5481 The ``user time'' and the ``system time'' are moved before the program
5482 name, and the options passed to the program are displayed, so that one
5483 can later tell what file was being compiled, and with which options.
5485 @item -fvar-tracking
5486 @opindex fvar-tracking
5487 Run variable tracking pass. It computes where variables are stored at each
5488 position in code. Better debugging information is then generated
5489 (if the debugging information format supports this information).
5491 It is enabled by default when compiling with optimization (@option{-Os},
5492 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5493 the debug info format supports it.
5495 @item -fvar-tracking-assignments
5496 @opindex fvar-tracking-assignments
5497 @opindex fno-var-tracking-assignments
5498 Annotate assignments to user variables early in the compilation and
5499 attempt to carry the annotations over throughout the compilation all the
5500 way to the end, in an attempt to improve debug information while
5501 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5503 It can be enabled even if var-tracking is disabled, in which case
5504 annotations will be created and maintained, but discarded at the end.
5506 @item -fvar-tracking-assignments-toggle
5507 @opindex fvar-tracking-assignments-toggle
5508 @opindex fno-var-tracking-assignments-toggle
5509 Toggle @option{-fvar-tracking-assignments}, in the same way that
5510 @option{-gtoggle} toggles @option{-g}.
5512 @item -print-file-name=@var{library}
5513 @opindex print-file-name
5514 Print the full absolute name of the library file @var{library} that
5515 would be used when linking---and don't do anything else. With this
5516 option, GCC does not compile or link anything; it just prints the
5519 @item -print-multi-directory
5520 @opindex print-multi-directory
5521 Print the directory name corresponding to the multilib selected by any
5522 other switches present in the command line. This directory is supposed
5523 to exist in @env{GCC_EXEC_PREFIX}.
5525 @item -print-multi-lib
5526 @opindex print-multi-lib
5527 Print the mapping from multilib directory names to compiler switches
5528 that enable them. The directory name is separated from the switches by
5529 @samp{;}, and each switch starts with an @samp{@@} instead of the
5530 @samp{-}, without spaces between multiple switches. This is supposed to
5531 ease shell-processing.
5533 @item -print-prog-name=@var{program}
5534 @opindex print-prog-name
5535 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5537 @item -print-libgcc-file-name
5538 @opindex print-libgcc-file-name
5539 Same as @option{-print-file-name=libgcc.a}.
5541 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5542 but you do want to link with @file{libgcc.a}. You can do
5545 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5548 @item -print-search-dirs
5549 @opindex print-search-dirs
5550 Print the name of the configured installation directory and a list of
5551 program and library directories @command{gcc} will search---and don't do anything else.
5553 This is useful when @command{gcc} prints the error message
5554 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5555 To resolve this you either need to put @file{cpp0} and the other compiler
5556 components where @command{gcc} expects to find them, or you can set the environment
5557 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5558 Don't forget the trailing @samp{/}.
5559 @xref{Environment Variables}.
5561 @item -print-sysroot
5562 @opindex print-sysroot
5563 Print the target sysroot directory that will be used during
5564 compilation. This is the target sysroot specified either at configure
5565 time or using the @option{--sysroot} option, possibly with an extra
5566 suffix that depends on compilation options. If no target sysroot is
5567 specified, the option prints nothing.
5569 @item -print-sysroot-headers-suffix
5570 @opindex print-sysroot-headers-suffix
5571 Print the suffix added to the target sysroot when searching for
5572 headers, or give an error if the compiler is not configured with such
5573 a suffix---and don't do anything else.
5576 @opindex dumpmachine
5577 Print the compiler's target machine (for example,
5578 @samp{i686-pc-linux-gnu})---and don't do anything else.
5581 @opindex dumpversion
5582 Print the compiler version (for example, @samp{3.0})---and don't do
5587 Print the compiler's built-in specs---and don't do anything else. (This
5588 is used when GCC itself is being built.) @xref{Spec Files}.
5590 @item -feliminate-unused-debug-types
5591 @opindex feliminate-unused-debug-types
5592 Normally, when producing DWARF2 output, GCC will emit debugging
5593 information for all types declared in a compilation
5594 unit, regardless of whether or not they are actually used
5595 in that compilation unit. Sometimes this is useful, such as
5596 if, in the debugger, you want to cast a value to a type that is
5597 not actually used in your program (but is declared). More often,
5598 however, this results in a significant amount of wasted space.
5599 With this option, GCC will avoid producing debug symbol output
5600 for types that are nowhere used in the source file being compiled.
5603 @node Optimize Options
5604 @section Options That Control Optimization
5605 @cindex optimize options
5606 @cindex options, optimization
5608 These options control various sorts of optimizations.
5610 Without any optimization option, the compiler's goal is to reduce the
5611 cost of compilation and to make debugging produce the expected
5612 results. Statements are independent: if you stop the program with a
5613 breakpoint between statements, you can then assign a new value to any
5614 variable or change the program counter to any other statement in the
5615 function and get exactly the results you would expect from the source
5618 Turning on optimization flags makes the compiler attempt to improve
5619 the performance and/or code size at the expense of compilation time
5620 and possibly the ability to debug the program.
5622 The compiler performs optimization based on the knowledge it has of the
5623 program. Compiling multiple files at once to a single output file mode allows
5624 the compiler to use information gained from all of the files when compiling
5627 Not all optimizations are controlled directly by a flag. Only
5628 optimizations that have a flag are listed in this section.
5630 Depending on the target and how GCC was configured, a slightly different
5631 set of optimizations may be enabled at each @option{-O} level than
5632 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5633 to find out the exact set of optimizations that are enabled at each level.
5634 @xref{Overall Options}, for examples.
5641 Optimize. Optimizing compilation takes somewhat more time, and a lot
5642 more memory for a large function.
5644 With @option{-O}, the compiler tries to reduce code size and execution
5645 time, without performing any optimizations that take a great deal of
5648 @option{-O} turns on the following optimization flags:
5651 -fcprop-registers @gol
5654 -fdelayed-branch @gol
5656 -fguess-branch-probability @gol
5657 -fif-conversion2 @gol
5658 -fif-conversion @gol
5659 -fipa-pure-const @gol
5660 -fipa-reference @gol
5662 -fsplit-wide-types @gol
5663 -ftree-builtin-call-dce @gol
5666 -ftree-copyrename @gol
5668 -ftree-dominator-opts @gol
5670 -ftree-forwprop @gol
5678 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5679 where doing so does not interfere with debugging.
5683 Optimize even more. GCC performs nearly all supported optimizations
5684 that do not involve a space-speed tradeoff.
5685 As compared to @option{-O}, this option increases both compilation time
5686 and the performance of the generated code.
5688 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5689 also turns on the following optimization flags:
5690 @gccoptlist{-fthread-jumps @gol
5691 -falign-functions -falign-jumps @gol
5692 -falign-loops -falign-labels @gol
5695 -fcse-follow-jumps -fcse-skip-blocks @gol
5696 -fdelete-null-pointer-checks @gol
5697 -fexpensive-optimizations @gol
5698 -fgcse -fgcse-lm @gol
5699 -finline-small-functions @gol
5700 -findirect-inlining @gol
5702 -foptimize-sibling-calls @gol
5705 -freorder-blocks -freorder-functions @gol
5706 -frerun-cse-after-loop @gol
5707 -fsched-interblock -fsched-spec @gol
5708 -fschedule-insns -fschedule-insns2 @gol
5709 -fstrict-aliasing -fstrict-overflow @gol
5710 -ftree-switch-conversion @gol
5714 Please note the warning under @option{-fgcse} about
5715 invoking @option{-O2} on programs that use computed gotos.
5719 Optimize yet more. @option{-O3} turns on all optimizations specified
5720 by @option{-O2} and also turns on the @option{-finline-functions},
5721 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5722 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5726 Reduce compilation time and make debugging produce the expected
5727 results. This is the default.
5731 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5732 do not typically increase code size. It also performs further
5733 optimizations designed to reduce code size.
5735 @option{-Os} disables the following optimization flags:
5736 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5737 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5738 -fprefetch-loop-arrays -ftree-vect-loop-version}
5740 If you use multiple @option{-O} options, with or without level numbers,
5741 the last such option is the one that is effective.
5744 Options of the form @option{-f@var{flag}} specify machine-independent
5745 flags. Most flags have both positive and negative forms; the negative
5746 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5747 below, only one of the forms is listed---the one you typically will
5748 use. You can figure out the other form by either removing @samp{no-}
5751 The following options control specific optimizations. They are either
5752 activated by @option{-O} options or are related to ones that are. You
5753 can use the following flags in the rare cases when ``fine-tuning'' of
5754 optimizations to be performed is desired.
5757 @item -fno-default-inline
5758 @opindex fno-default-inline
5759 Do not make member functions inline by default merely because they are
5760 defined inside the class scope (C++ only). Otherwise, when you specify
5761 @w{@option{-O}}, member functions defined inside class scope are compiled
5762 inline by default; i.e., you don't need to add @samp{inline} in front of
5763 the member function name.
5765 @item -fno-defer-pop
5766 @opindex fno-defer-pop
5767 Always pop the arguments to each function call as soon as that function
5768 returns. For machines which must pop arguments after a function call,
5769 the compiler normally lets arguments accumulate on the stack for several
5770 function calls and pops them all at once.
5772 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5774 @item -fforward-propagate
5775 @opindex fforward-propagate
5776 Perform a forward propagation pass on RTL@. The pass tries to combine two
5777 instructions and checks if the result can be simplified. If loop unrolling
5778 is active, two passes are performed and the second is scheduled after
5781 This option is enabled by default at optimization levels @option{-O},
5782 @option{-O2}, @option{-O3}, @option{-Os}.
5784 @item -fomit-frame-pointer
5785 @opindex fomit-frame-pointer
5786 Don't keep the frame pointer in a register for functions that
5787 don't need one. This avoids the instructions to save, set up and
5788 restore frame pointers; it also makes an extra register available
5789 in many functions. @strong{It also makes debugging impossible on
5792 On some machines, such as the VAX, this flag has no effect, because
5793 the standard calling sequence automatically handles the frame pointer
5794 and nothing is saved by pretending it doesn't exist. The
5795 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5796 whether a target machine supports this flag. @xref{Registers,,Register
5797 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5799 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5801 @item -foptimize-sibling-calls
5802 @opindex foptimize-sibling-calls
5803 Optimize sibling and tail recursive calls.
5805 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5809 Don't pay attention to the @code{inline} keyword. Normally this option
5810 is used to keep the compiler from expanding any functions inline.
5811 Note that if you are not optimizing, no functions can be expanded inline.
5813 @item -finline-small-functions
5814 @opindex finline-small-functions
5815 Integrate functions into their callers when their body is smaller than expected
5816 function call code (so overall size of program gets smaller). The compiler
5817 heuristically decides which functions are simple enough to be worth integrating
5820 Enabled at level @option{-O2}.
5822 @item -findirect-inlining
5823 @opindex findirect-inlining
5824 Inline also indirect calls that are discovered to be known at compile
5825 time thanks to previous inlining. This option has any effect only
5826 when inlining itself is turned on by the @option{-finline-functions}
5827 or @option{-finline-small-functions} options.
5829 Enabled at level @option{-O2}.
5831 @item -finline-functions
5832 @opindex finline-functions
5833 Integrate all simple functions into their callers. The compiler
5834 heuristically decides which functions are simple enough to be worth
5835 integrating in this way.
5837 If all calls to a given function are integrated, and the function is
5838 declared @code{static}, then the function is normally not output as
5839 assembler code in its own right.
5841 Enabled at level @option{-O3}.
5843 @item -finline-functions-called-once
5844 @opindex finline-functions-called-once
5845 Consider all @code{static} functions called once for inlining into their
5846 caller even if they are not marked @code{inline}. If a call to a given
5847 function is integrated, then the function is not output as assembler code
5850 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5852 @item -fearly-inlining
5853 @opindex fearly-inlining
5854 Inline functions marked by @code{always_inline} and functions whose body seems
5855 smaller than the function call overhead early before doing
5856 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5857 makes profiling significantly cheaper and usually inlining faster on programs
5858 having large chains of nested wrapper functions.
5864 Perform interprocedural scalar replacement of aggregates, removal of
5865 unused parameters and replacement of parameters passed by reference
5866 by parameters passed by value.
5868 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5870 @item -finline-limit=@var{n}
5871 @opindex finline-limit
5872 By default, GCC limits the size of functions that can be inlined. This flag
5873 allows coarse control of this limit. @var{n} is the size of functions that
5874 can be inlined in number of pseudo instructions.
5876 Inlining is actually controlled by a number of parameters, which may be
5877 specified individually by using @option{--param @var{name}=@var{value}}.
5878 The @option{-finline-limit=@var{n}} option sets some of these parameters
5882 @item max-inline-insns-single
5883 is set to @var{n}/2.
5884 @item max-inline-insns-auto
5885 is set to @var{n}/2.
5888 See below for a documentation of the individual
5889 parameters controlling inlining and for the defaults of these parameters.
5891 @emph{Note:} there may be no value to @option{-finline-limit} that results
5892 in default behavior.
5894 @emph{Note:} pseudo instruction represents, in this particular context, an
5895 abstract measurement of function's size. In no way does it represent a count
5896 of assembly instructions and as such its exact meaning might change from one
5897 release to an another.
5899 @item -fkeep-inline-functions
5900 @opindex fkeep-inline-functions
5901 In C, emit @code{static} functions that are declared @code{inline}
5902 into the object file, even if the function has been inlined into all
5903 of its callers. This switch does not affect functions using the
5904 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5905 inline functions into the object file.
5907 @item -fkeep-static-consts
5908 @opindex fkeep-static-consts
5909 Emit variables declared @code{static const} when optimization isn't turned
5910 on, even if the variables aren't referenced.
5912 GCC enables this option by default. If you want to force the compiler to
5913 check if the variable was referenced, regardless of whether or not
5914 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5916 @item -fmerge-constants
5917 @opindex fmerge-constants
5918 Attempt to merge identical constants (string constants and floating point
5919 constants) across compilation units.
5921 This option is the default for optimized compilation if the assembler and
5922 linker support it. Use @option{-fno-merge-constants} to inhibit this
5925 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5927 @item -fmerge-all-constants
5928 @opindex fmerge-all-constants
5929 Attempt to merge identical constants and identical variables.
5931 This option implies @option{-fmerge-constants}. In addition to
5932 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5933 arrays or initialized constant variables with integral or floating point
5934 types. Languages like C or C++ require each variable, including multiple
5935 instances of the same variable in recursive calls, to have distinct locations,
5936 so using this option will result in non-conforming
5939 @item -fmodulo-sched
5940 @opindex fmodulo-sched
5941 Perform swing modulo scheduling immediately before the first scheduling
5942 pass. This pass looks at innermost loops and reorders their
5943 instructions by overlapping different iterations.
5945 @item -fmodulo-sched-allow-regmoves
5946 @opindex fmodulo-sched-allow-regmoves
5947 Perform more aggressive SMS based modulo scheduling with register moves
5948 allowed. By setting this flag certain anti-dependences edges will be
5949 deleted which will trigger the generation of reg-moves based on the
5950 life-range analysis. This option is effective only with
5951 @option{-fmodulo-sched} enabled.
5953 @item -fno-branch-count-reg
5954 @opindex fno-branch-count-reg
5955 Do not use ``decrement and branch'' instructions on a count register,
5956 but instead generate a sequence of instructions that decrement a
5957 register, compare it against zero, then branch based upon the result.
5958 This option is only meaningful on architectures that support such
5959 instructions, which include x86, PowerPC, IA-64 and S/390.
5961 The default is @option{-fbranch-count-reg}.
5963 @item -fno-function-cse
5964 @opindex fno-function-cse
5965 Do not put function addresses in registers; make each instruction that
5966 calls a constant function contain the function's address explicitly.
5968 This option results in less efficient code, but some strange hacks
5969 that alter the assembler output may be confused by the optimizations
5970 performed when this option is not used.
5972 The default is @option{-ffunction-cse}
5974 @item -fno-zero-initialized-in-bss
5975 @opindex fno-zero-initialized-in-bss
5976 If the target supports a BSS section, GCC by default puts variables that
5977 are initialized to zero into BSS@. This can save space in the resulting
5980 This option turns off this behavior because some programs explicitly
5981 rely on variables going to the data section. E.g., so that the
5982 resulting executable can find the beginning of that section and/or make
5983 assumptions based on that.
5985 The default is @option{-fzero-initialized-in-bss}.
5987 @item -fmudflap -fmudflapth -fmudflapir
5991 @cindex bounds checking
5993 For front-ends that support it (C and C++), instrument all risky
5994 pointer/array dereferencing operations, some standard library
5995 string/heap functions, and some other associated constructs with
5996 range/validity tests. Modules so instrumented should be immune to
5997 buffer overflows, invalid heap use, and some other classes of C/C++
5998 programming errors. The instrumentation relies on a separate runtime
5999 library (@file{libmudflap}), which will be linked into a program if
6000 @option{-fmudflap} is given at link time. Run-time behavior of the
6001 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6002 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6005 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6006 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6007 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6008 instrumentation should ignore pointer reads. This produces less
6009 instrumentation (and therefore faster execution) and still provides
6010 some protection against outright memory corrupting writes, but allows
6011 erroneously read data to propagate within a program.
6013 @item -fthread-jumps
6014 @opindex fthread-jumps
6015 Perform optimizations where we check to see if a jump branches to a
6016 location where another comparison subsumed by the first is found. If
6017 so, the first branch is redirected to either the destination of the
6018 second branch or a point immediately following it, depending on whether
6019 the condition is known to be true or false.
6021 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6023 @item -fsplit-wide-types
6024 @opindex fsplit-wide-types
6025 When using a type that occupies multiple registers, such as @code{long
6026 long} on a 32-bit system, split the registers apart and allocate them
6027 independently. This normally generates better code for those types,
6028 but may make debugging more difficult.
6030 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6033 @item -fcse-follow-jumps
6034 @opindex fcse-follow-jumps
6035 In common subexpression elimination (CSE), scan through jump instructions
6036 when the target of the jump is not reached by any other path. For
6037 example, when CSE encounters an @code{if} statement with an
6038 @code{else} clause, CSE will follow the jump when the condition
6041 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6043 @item -fcse-skip-blocks
6044 @opindex fcse-skip-blocks
6045 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6046 follow jumps which conditionally skip over blocks. When CSE
6047 encounters a simple @code{if} statement with no else clause,
6048 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6049 body of the @code{if}.
6051 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6053 @item -frerun-cse-after-loop
6054 @opindex frerun-cse-after-loop
6055 Re-run common subexpression elimination after loop optimizations has been
6058 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6062 Perform a global common subexpression elimination pass.
6063 This pass also performs global constant and copy propagation.
6065 @emph{Note:} When compiling a program using computed gotos, a GCC
6066 extension, you may get better runtime performance if you disable
6067 the global common subexpression elimination pass by adding
6068 @option{-fno-gcse} to the command line.
6070 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6074 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6075 attempt to move loads which are only killed by stores into themselves. This
6076 allows a loop containing a load/store sequence to be changed to a load outside
6077 the loop, and a copy/store within the loop.
6079 Enabled by default when gcse is enabled.
6083 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6084 global common subexpression elimination. This pass will attempt to move
6085 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6086 loops containing a load/store sequence can be changed to a load before
6087 the loop and a store after the loop.
6089 Not enabled at any optimization level.
6093 When @option{-fgcse-las} is enabled, the global common subexpression
6094 elimination pass eliminates redundant loads that come after stores to the
6095 same memory location (both partial and full redundancies).
6097 Not enabled at any optimization level.
6099 @item -fgcse-after-reload
6100 @opindex fgcse-after-reload
6101 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6102 pass is performed after reload. The purpose of this pass is to cleanup
6105 @item -funsafe-loop-optimizations
6106 @opindex funsafe-loop-optimizations
6107 If given, the loop optimizer will assume that loop indices do not
6108 overflow, and that the loops with nontrivial exit condition are not
6109 infinite. This enables a wider range of loop optimizations even if
6110 the loop optimizer itself cannot prove that these assumptions are valid.
6111 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6112 if it finds this kind of loop.
6114 @item -fcrossjumping
6115 @opindex fcrossjumping
6116 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6117 resulting code may or may not perform better than without cross-jumping.
6119 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6121 @item -fauto-inc-dec
6122 @opindex fauto-inc-dec
6123 Combine increments or decrements of addresses with memory accesses.
6124 This pass is always skipped on architectures that do not have
6125 instructions to support this. Enabled by default at @option{-O} and
6126 higher on architectures that support this.
6130 Perform dead code elimination (DCE) on RTL@.
6131 Enabled by default at @option{-O} and higher.
6135 Perform dead store elimination (DSE) on RTL@.
6136 Enabled by default at @option{-O} and higher.
6138 @item -fif-conversion
6139 @opindex fif-conversion
6140 Attempt to transform conditional jumps into branch-less equivalents. This
6141 include use of conditional moves, min, max, set flags and abs instructions, and
6142 some tricks doable by standard arithmetics. The use of conditional execution
6143 on chips where it is available is controlled by @code{if-conversion2}.
6145 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6147 @item -fif-conversion2
6148 @opindex fif-conversion2
6149 Use conditional execution (where available) to transform conditional jumps into
6150 branch-less equivalents.
6152 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6154 @item -fdelete-null-pointer-checks
6155 @opindex fdelete-null-pointer-checks
6156 Assume that programs cannot safely dereference null pointers, and that
6157 no code or data element resides there. This enables simple constant
6158 folding optimizations at all optimization levels. In addition, other
6159 optimization passes in GCC use this flag to control global dataflow
6160 analyses that eliminate useless checks for null pointers; these assume
6161 that if a pointer is checked after it has already been dereferenced,
6164 Note however that in some environments this assumption is not true.
6165 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6166 for programs which depend on that behavior.
6168 Some targets, especially embedded ones, disable this option at all levels.
6169 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6170 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6171 are enabled independently at different optimization levels.
6173 @item -fexpensive-optimizations
6174 @opindex fexpensive-optimizations
6175 Perform a number of minor optimizations that are relatively expensive.
6177 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6179 @item -foptimize-register-move
6181 @opindex foptimize-register-move
6183 Attempt to reassign register numbers in move instructions and as
6184 operands of other simple instructions in order to maximize the amount of
6185 register tying. This is especially helpful on machines with two-operand
6188 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6191 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6193 @item -fira-algorithm=@var{algorithm}
6194 Use specified coloring algorithm for the integrated register
6195 allocator. The @var{algorithm} argument should be @code{priority} or
6196 @code{CB}. The first algorithm specifies Chow's priority coloring,
6197 the second one specifies Chaitin-Briggs coloring. The second
6198 algorithm can be unimplemented for some architectures. If it is
6199 implemented, it is the default because Chaitin-Briggs coloring as a
6200 rule generates a better code.
6202 @item -fira-region=@var{region}
6203 Use specified regions for the integrated register allocator. The
6204 @var{region} argument should be one of @code{all}, @code{mixed}, or
6205 @code{one}. The first value means using all loops as register
6206 allocation regions, the second value which is the default means using
6207 all loops except for loops with small register pressure as the
6208 regions, and third one means using all function as a single region.
6209 The first value can give best result for machines with small size and
6210 irregular register set, the third one results in faster and generates
6211 decent code and the smallest size code, and the default value usually
6212 give the best results in most cases and for most architectures.
6214 @item -fira-coalesce
6215 @opindex fira-coalesce
6216 Do optimistic register coalescing. This option might be profitable for
6217 architectures with big regular register files.
6219 @item -fno-ira-share-save-slots
6220 @opindex fno-ira-share-save-slots
6221 Switch off sharing stack slots used for saving call used hard
6222 registers living through a call. Each hard register will get a
6223 separate stack slot and as a result function stack frame will be
6226 @item -fno-ira-share-spill-slots
6227 @opindex fno-ira-share-spill-slots
6228 Switch off sharing stack slots allocated for pseudo-registers. Each
6229 pseudo-register which did not get a hard register will get a separate
6230 stack slot and as a result function stack frame will be bigger.
6232 @item -fira-verbose=@var{n}
6233 @opindex fira-verbose
6234 Set up how verbose dump file for the integrated register allocator
6235 will be. Default value is 5. If the value is greater or equal to 10,
6236 the dump file will be stderr as if the value were @var{n} minus 10.
6238 @item -fdelayed-branch
6239 @opindex fdelayed-branch
6240 If supported for the target machine, attempt to reorder instructions
6241 to exploit instruction slots available after delayed branch
6244 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6246 @item -fschedule-insns
6247 @opindex fschedule-insns
6248 If supported for the target machine, attempt to reorder instructions to
6249 eliminate execution stalls due to required data being unavailable. This
6250 helps machines that have slow floating point or memory load instructions
6251 by allowing other instructions to be issued until the result of the load
6252 or floating point instruction is required.
6254 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6256 @item -fschedule-insns2
6257 @opindex fschedule-insns2
6258 Similar to @option{-fschedule-insns}, but requests an additional pass of
6259 instruction scheduling after register allocation has been done. This is
6260 especially useful on machines with a relatively small number of
6261 registers and where memory load instructions take more than one cycle.
6263 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6265 @item -fno-sched-interblock
6266 @opindex fno-sched-interblock
6267 Don't schedule instructions across basic blocks. This is normally
6268 enabled by default when scheduling before register allocation, i.e.@:
6269 with @option{-fschedule-insns} or at @option{-O2} or higher.
6271 @item -fno-sched-spec
6272 @opindex fno-sched-spec
6273 Don't allow speculative motion of non-load instructions. This is normally
6274 enabled by default when scheduling before register allocation, i.e.@:
6275 with @option{-fschedule-insns} or at @option{-O2} or higher.
6277 @item -fsched-pressure
6278 @opindex fsched-pressure
6279 Enable register pressure sensitive insn scheduling before the register
6280 allocation. This only makes sense when scheduling before register
6281 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6282 @option{-O2} or higher. Usage of this option can improve the
6283 generated code and decrease its size by preventing register pressure
6284 increase above the number of available hard registers and as a
6285 consequence register spills in the register allocation.
6287 @item -fsched-spec-load
6288 @opindex fsched-spec-load
6289 Allow speculative motion of some load instructions. This only makes
6290 sense when scheduling before register allocation, i.e.@: with
6291 @option{-fschedule-insns} or at @option{-O2} or higher.
6293 @item -fsched-spec-load-dangerous
6294 @opindex fsched-spec-load-dangerous
6295 Allow speculative motion of more load instructions. This only makes
6296 sense when scheduling before register allocation, i.e.@: with
6297 @option{-fschedule-insns} or at @option{-O2} or higher.
6299 @item -fsched-stalled-insns
6300 @itemx -fsched-stalled-insns=@var{n}
6301 @opindex fsched-stalled-insns
6302 Define how many insns (if any) can be moved prematurely from the queue
6303 of stalled insns into the ready list, during the second scheduling pass.
6304 @option{-fno-sched-stalled-insns} means that no insns will be moved
6305 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6306 on how many queued insns can be moved prematurely.
6307 @option{-fsched-stalled-insns} without a value is equivalent to
6308 @option{-fsched-stalled-insns=1}.
6310 @item -fsched-stalled-insns-dep
6311 @itemx -fsched-stalled-insns-dep=@var{n}
6312 @opindex fsched-stalled-insns-dep
6313 Define how many insn groups (cycles) will be examined for a dependency
6314 on a stalled insn that is candidate for premature removal from the queue
6315 of stalled insns. This has an effect only during the second scheduling pass,
6316 and only if @option{-fsched-stalled-insns} is used.
6317 @option{-fno-sched-stalled-insns-dep} is equivalent to
6318 @option{-fsched-stalled-insns-dep=0}.
6319 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6320 @option{-fsched-stalled-insns-dep=1}.
6322 @item -fsched2-use-superblocks
6323 @opindex fsched2-use-superblocks
6324 When scheduling after register allocation, do use superblock scheduling
6325 algorithm. Superblock scheduling allows motion across basic block boundaries
6326 resulting on faster schedules. This option is experimental, as not all machine
6327 descriptions used by GCC model the CPU closely enough to avoid unreliable
6328 results from the algorithm.
6330 This only makes sense when scheduling after register allocation, i.e.@: with
6331 @option{-fschedule-insns2} or at @option{-O2} or higher.
6333 @item -fsched-group-heuristic
6334 @opindex fsched-group-heuristic
6335 Enable the group heuristic in the scheduler. This heuristic favors
6336 the instruction that belongs to a schedule group. This is enabled
6337 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6338 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6340 @item -fsched-critical-path-heuristic
6341 @opindex fsched-critical-path-heuristic
6342 Enable the critical-path heuristic in the scheduler. This heuristic favors
6343 instructions on the critical path. This is enabled by default when
6344 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6345 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6347 @item -fsched-spec-insn-heuristic
6348 @opindex fsched-spec-insn-heuristic
6349 Enable the speculative instruction heuristic in the scheduler. This
6350 heuristic favors speculative instructions with greater dependency weakness.
6351 This is enabled by default when scheduling is enabled, i.e.@:
6352 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6353 or at @option{-O2} or higher.
6355 @item -fsched-rank-heuristic
6356 @opindex fsched-rank-heuristic
6357 Enable the rank heuristic in the scheduler. This heuristic favors
6358 the instruction belonging to a basic block with greater size or frequency.
6359 This is enabled by default when scheduling is enabled, i.e.@:
6360 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6361 at @option{-O2} or higher.
6363 @item -fsched-last-insn-heuristic
6364 @opindex fsched-last-insn-heuristic
6365 Enable the last-instruction heuristic in the scheduler. This heuristic
6366 favors the instruction that is less dependent on the last instruction
6367 scheduled. This is enabled by default when scheduling is enabled,
6368 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6369 at @option{-O2} or higher.
6371 @item -fsched-dep-count-heuristic
6372 @opindex fsched-dep-count-heuristic
6373 Enable the dependent-count heuristic in the scheduler. This heuristic
6374 favors the instruction that has more instructions depending on it.
6375 This is enabled by default when scheduling is enabled, i.e.@:
6376 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6377 at @option{-O2} or higher.
6379 @item -fsched2-use-traces
6380 @opindex fsched2-use-traces
6381 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6382 allocation and additionally perform code duplication in order to increase the
6383 size of superblocks using tracer pass. See @option{-ftracer} for details on
6386 This mode should produce faster but significantly longer programs. Also
6387 without @option{-fbranch-probabilities} the traces constructed may not
6388 match the reality and hurt the performance. This only makes
6389 sense when scheduling after register allocation, i.e.@: with
6390 @option{-fschedule-insns2} or at @option{-O2} or higher.
6392 @item -freschedule-modulo-scheduled-loops
6393 @opindex freschedule-modulo-scheduled-loops
6394 The modulo scheduling comes before the traditional scheduling, if a loop
6395 was modulo scheduled we may want to prevent the later scheduling passes
6396 from changing its schedule, we use this option to control that.
6398 @item -fselective-scheduling
6399 @opindex fselective-scheduling
6400 Schedule instructions using selective scheduling algorithm. Selective
6401 scheduling runs instead of the first scheduler pass.
6403 @item -fselective-scheduling2
6404 @opindex fselective-scheduling2
6405 Schedule instructions using selective scheduling algorithm. Selective
6406 scheduling runs instead of the second scheduler pass.
6408 @item -fsel-sched-pipelining
6409 @opindex fsel-sched-pipelining
6410 Enable software pipelining of innermost loops during selective scheduling.
6411 This option has no effect until one of @option{-fselective-scheduling} or
6412 @option{-fselective-scheduling2} is turned on.
6414 @item -fsel-sched-pipelining-outer-loops
6415 @opindex fsel-sched-pipelining-outer-loops
6416 When pipelining loops during selective scheduling, also pipeline outer loops.
6417 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6419 @item -fcaller-saves
6420 @opindex fcaller-saves
6421 Enable values to be allocated in registers that will be clobbered by
6422 function calls, by emitting extra instructions to save and restore the
6423 registers around such calls. Such allocation is done only when it
6424 seems to result in better code than would otherwise be produced.
6426 This option is always enabled by default on certain machines, usually
6427 those which have no call-preserved registers to use instead.
6429 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6431 @item -fconserve-stack
6432 @opindex fconserve-stack
6433 Attempt to minimize stack usage. The compiler will attempt to use less
6434 stack space, even if that makes the program slower. This option
6435 implies setting the @option{large-stack-frame} parameter to 100
6436 and the @option{large-stack-frame-growth} parameter to 400.
6438 @item -ftree-reassoc
6439 @opindex ftree-reassoc
6440 Perform reassociation on trees. This flag is enabled by default
6441 at @option{-O} and higher.
6445 Perform partial redundancy elimination (PRE) on trees. This flag is
6446 enabled by default at @option{-O2} and @option{-O3}.
6448 @item -ftree-forwprop
6449 @opindex ftree-forwprop
6450 Perform forward propagation on trees. This flag is enabled by default
6451 at @option{-O} and higher.
6455 Perform full redundancy elimination (FRE) on trees. The difference
6456 between FRE and PRE is that FRE only considers expressions
6457 that are computed on all paths leading to the redundant computation.
6458 This analysis is faster than PRE, though it exposes fewer redundancies.
6459 This flag is enabled by default at @option{-O} and higher.
6461 @item -ftree-phiprop
6462 @opindex ftree-phiprop
6463 Perform hoisting of loads from conditional pointers on trees. This
6464 pass is enabled by default at @option{-O} and higher.
6466 @item -ftree-copy-prop
6467 @opindex ftree-copy-prop
6468 Perform copy propagation on trees. This pass eliminates unnecessary
6469 copy operations. This flag is enabled by default at @option{-O} and
6472 @item -fipa-pure-const
6473 @opindex fipa-pure-const
6474 Discover which functions are pure or constant.
6475 Enabled by default at @option{-O} and higher.
6477 @item -fipa-reference
6478 @opindex fipa-reference
6479 Discover which static variables do not escape cannot escape the
6481 Enabled by default at @option{-O} and higher.
6483 @item -fipa-struct-reorg
6484 @opindex fipa-struct-reorg
6485 Perform structure reorganization optimization, that change C-like structures
6486 layout in order to better utilize spatial locality. This transformation is
6487 affective for programs containing arrays of structures. Available in two
6488 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6489 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6490 to provide the safety of this transformation. It works only in whole program
6491 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6492 enabled. Structures considered @samp{cold} by this transformation are not
6493 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6495 With this flag, the program debug info reflects a new structure layout.
6499 Perform interprocedural pointer analysis. This option is experimental
6500 and does not affect generated code.
6504 Perform interprocedural constant propagation.
6505 This optimization analyzes the program to determine when values passed
6506 to functions are constants and then optimizes accordingly.
6507 This optimization can substantially increase performance
6508 if the application has constants passed to functions.
6509 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6511 @item -fipa-cp-clone
6512 @opindex fipa-cp-clone
6513 Perform function cloning to make interprocedural constant propagation stronger.
6514 When enabled, interprocedural constant propagation will perform function cloning
6515 when externally visible function can be called with constant arguments.
6516 Because this optimization can create multiple copies of functions,
6517 it may significantly increase code size
6518 (see @option{--param ipcp-unit-growth=@var{value}}).
6519 This flag is enabled by default at @option{-O3}.
6521 @item -fipa-matrix-reorg
6522 @opindex fipa-matrix-reorg
6523 Perform matrix flattening and transposing.
6524 Matrix flattening tries to replace an @math{m}-dimensional matrix
6525 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6526 This reduces the level of indirection needed for accessing the elements
6527 of the matrix. The second optimization is matrix transposing that
6528 attempts to change the order of the matrix's dimensions in order to
6529 improve cache locality.
6530 Both optimizations need the @option{-fwhole-program} flag.
6531 Transposing is enabled only if profiling information is available.
6535 Perform forward store motion on trees. This flag is
6536 enabled by default at @option{-O} and higher.
6540 Perform sparse conditional constant propagation (CCP) on trees. This
6541 pass only operates on local scalar variables and is enabled by default
6542 at @option{-O} and higher.
6544 @item -ftree-switch-conversion
6545 Perform conversion of simple initializations in a switch to
6546 initializations from a scalar array. This flag is enabled by default
6547 at @option{-O2} and higher.
6551 Perform dead code elimination (DCE) on trees. This flag is enabled by
6552 default at @option{-O} and higher.
6554 @item -ftree-builtin-call-dce
6555 @opindex ftree-builtin-call-dce
6556 Perform conditional dead code elimination (DCE) for calls to builtin functions
6557 that may set @code{errno} but are otherwise side-effect free. This flag is
6558 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6561 @item -ftree-dominator-opts
6562 @opindex ftree-dominator-opts
6563 Perform a variety of simple scalar cleanups (constant/copy
6564 propagation, redundancy elimination, range propagation and expression
6565 simplification) based on a dominator tree traversal. This also
6566 performs jump threading (to reduce jumps to jumps). This flag is
6567 enabled by default at @option{-O} and higher.
6571 Perform dead store elimination (DSE) on trees. A dead store is a store into
6572 a memory location which will later be overwritten by another store without
6573 any intervening loads. In this case the earlier store can be deleted. This
6574 flag is enabled by default at @option{-O} and higher.
6578 Perform loop header copying on trees. This is beneficial since it increases
6579 effectiveness of code motion optimizations. It also saves one jump. This flag
6580 is enabled by default at @option{-O} and higher. It is not enabled
6581 for @option{-Os}, since it usually increases code size.
6583 @item -ftree-loop-optimize
6584 @opindex ftree-loop-optimize
6585 Perform loop optimizations on trees. This flag is enabled by default
6586 at @option{-O} and higher.
6588 @item -ftree-loop-linear
6589 @opindex ftree-loop-linear
6590 Perform linear loop transformations on tree. This flag can improve cache
6591 performance and allow further loop optimizations to take place.
6593 @item -floop-interchange
6594 Perform loop interchange transformations on loops. Interchanging two
6595 nested loops switches the inner and outer loops. For example, given a
6600 A(J, I) = A(J, I) * C
6604 loop interchange will transform the loop as if the user had written:
6608 A(J, I) = A(J, I) * C
6612 which can be beneficial when @code{N} is larger than the caches,
6613 because in Fortran, the elements of an array are stored in memory
6614 contiguously by column, and the original loop iterates over rows,
6615 potentially creating at each access a cache miss. This optimization
6616 applies to all the languages supported by GCC and is not limited to
6617 Fortran. To use this code transformation, GCC has to be configured
6618 with @option{--with-ppl} and @option{--with-cloog} to enable the
6619 Graphite loop transformation infrastructure.
6621 @item -floop-strip-mine
6622 Perform loop strip mining transformations on loops. Strip mining
6623 splits a loop into two nested loops. The outer loop has strides
6624 equal to the strip size and the inner loop has strides of the
6625 original loop within a strip. For example, given a loop like:
6631 loop strip mining will transform the loop as if the user had written:
6634 DO I = II, min (II + 3, N)
6639 This optimization applies to all the languages supported by GCC and is
6640 not limited to Fortran. To use this code transformation, GCC has to
6641 be configured with @option{--with-ppl} and @option{--with-cloog} to
6642 enable the Graphite loop transformation infrastructure.
6645 Perform loop blocking transformations on loops. Blocking strip mines
6646 each loop in the loop nest such that the memory accesses of the
6647 element loops fit inside caches. For example, given a loop like:
6651 A(J, I) = B(I) + C(J)
6655 loop blocking will transform the loop as if the user had written:
6659 DO I = II, min (II + 63, N)
6660 DO J = JJ, min (JJ + 63, M)
6661 A(J, I) = B(I) + C(J)
6667 which can be beneficial when @code{M} is larger than the caches,
6668 because the innermost loop will iterate over a smaller amount of data
6669 that can be kept in the caches. This optimization applies to all the
6670 languages supported by GCC and is not limited to Fortran. To use this
6671 code transformation, GCC has to be configured with @option{--with-ppl}
6672 and @option{--with-cloog} to enable the Graphite loop transformation
6675 @item -fgraphite-identity
6676 @opindex fgraphite-identity
6677 Enable the identity transformation for graphite. For every SCoP we generate
6678 the polyhedral representation and transform it back to gimple. Using
6679 @option{-fgraphite-identity} we can check the costs or benefits of the
6680 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6681 are also performed by the code generator CLooG, like index splitting and
6682 dead code elimination in loops.
6684 @item -floop-parallelize-all
6685 Use the Graphite data dependence analysis to identify loops that can
6686 be parallelized. Parallelize all the loops that can be analyzed to
6687 not contain loop carried dependences without checking that it is
6688 profitable to parallelize the loops.
6690 @item -fcheck-data-deps
6691 @opindex fcheck-data-deps
6692 Compare the results of several data dependence analyzers. This option
6693 is used for debugging the data dependence analyzers.
6695 @item -ftree-loop-distribution
6696 Perform loop distribution. This flag can improve cache performance on
6697 big loop bodies and allow further loop optimizations, like
6698 parallelization or vectorization, to take place. For example, the loop
6715 @item -ftree-loop-im
6716 @opindex ftree-loop-im
6717 Perform loop invariant motion on trees. This pass moves only invariants that
6718 would be hard to handle at RTL level (function calls, operations that expand to
6719 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6720 operands of conditions that are invariant out of the loop, so that we can use
6721 just trivial invariantness analysis in loop unswitching. The pass also includes
6724 @item -ftree-loop-ivcanon
6725 @opindex ftree-loop-ivcanon
6726 Create a canonical counter for number of iterations in the loop for that
6727 determining number of iterations requires complicated analysis. Later
6728 optimizations then may determine the number easily. Useful especially
6729 in connection with unrolling.
6733 Perform induction variable optimizations (strength reduction, induction
6734 variable merging and induction variable elimination) on trees.
6736 @item -ftree-parallelize-loops=n
6737 @opindex ftree-parallelize-loops
6738 Parallelize loops, i.e., split their iteration space to run in n threads.
6739 This is only possible for loops whose iterations are independent
6740 and can be arbitrarily reordered. The optimization is only
6741 profitable on multiprocessor machines, for loops that are CPU-intensive,
6742 rather than constrained e.g.@: by memory bandwidth. This option
6743 implies @option{-pthread}, and thus is only supported on targets
6744 that have support for @option{-pthread}.
6748 Perform function-local points-to analysis on trees. This flag is
6749 enabled by default at @option{-O} and higher.
6753 Perform scalar replacement of aggregates. This pass replaces structure
6754 references with scalars to prevent committing structures to memory too
6755 early. This flag is enabled by default at @option{-O} and higher.
6757 @item -ftree-copyrename
6758 @opindex ftree-copyrename
6759 Perform copy renaming on trees. This pass attempts to rename compiler
6760 temporaries to other variables at copy locations, usually resulting in
6761 variable names which more closely resemble the original variables. This flag
6762 is enabled by default at @option{-O} and higher.
6766 Perform temporary expression replacement during the SSA->normal phase. Single
6767 use/single def temporaries are replaced at their use location with their
6768 defining expression. This results in non-GIMPLE code, but gives the expanders
6769 much more complex trees to work on resulting in better RTL generation. This is
6770 enabled by default at @option{-O} and higher.
6772 @item -ftree-vectorize
6773 @opindex ftree-vectorize
6774 Perform loop vectorization on trees. This flag is enabled by default at
6777 @item -ftree-vect-loop-version
6778 @opindex ftree-vect-loop-version
6779 Perform loop versioning when doing loop vectorization on trees. When a loop
6780 appears to be vectorizable except that data alignment or data dependence cannot
6781 be determined at compile time then vectorized and non-vectorized versions of
6782 the loop are generated along with runtime checks for alignment or dependence
6783 to control which version is executed. This option is enabled by default
6784 except at level @option{-Os} where it is disabled.
6786 @item -fvect-cost-model
6787 @opindex fvect-cost-model
6788 Enable cost model for vectorization.
6792 Perform Value Range Propagation on trees. This is similar to the
6793 constant propagation pass, but instead of values, ranges of values are
6794 propagated. This allows the optimizers to remove unnecessary range
6795 checks like array bound checks and null pointer checks. This is
6796 enabled by default at @option{-O2} and higher. Null pointer check
6797 elimination is only done if @option{-fdelete-null-pointer-checks} is
6802 Perform tail duplication to enlarge superblock size. This transformation
6803 simplifies the control flow of the function allowing other optimizations to do
6806 @item -funroll-loops
6807 @opindex funroll-loops
6808 Unroll loops whose number of iterations can be determined at compile
6809 time or upon entry to the loop. @option{-funroll-loops} implies
6810 @option{-frerun-cse-after-loop}. This option makes code larger,
6811 and may or may not make it run faster.
6813 @item -funroll-all-loops
6814 @opindex funroll-all-loops
6815 Unroll all loops, even if their number of iterations is uncertain when
6816 the loop is entered. This usually makes programs run more slowly.
6817 @option{-funroll-all-loops} implies the same options as
6818 @option{-funroll-loops},
6820 @item -fsplit-ivs-in-unroller
6821 @opindex fsplit-ivs-in-unroller
6822 Enables expressing of values of induction variables in later iterations
6823 of the unrolled loop using the value in the first iteration. This breaks
6824 long dependency chains, thus improving efficiency of the scheduling passes.
6826 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6827 same effect. However in cases the loop body is more complicated than
6828 a single basic block, this is not reliable. It also does not work at all
6829 on some of the architectures due to restrictions in the CSE pass.
6831 This optimization is enabled by default.
6833 @item -fvariable-expansion-in-unroller
6834 @opindex fvariable-expansion-in-unroller
6835 With this option, the compiler will create multiple copies of some
6836 local variables when unrolling a loop which can result in superior code.
6838 @item -fpredictive-commoning
6839 @opindex fpredictive-commoning
6840 Perform predictive commoning optimization, i.e., reusing computations
6841 (especially memory loads and stores) performed in previous
6842 iterations of loops.
6844 This option is enabled at level @option{-O3}.
6846 @item -fprefetch-loop-arrays
6847 @opindex fprefetch-loop-arrays
6848 If supported by the target machine, generate instructions to prefetch
6849 memory to improve the performance of loops that access large arrays.
6851 This option may generate better or worse code; results are highly
6852 dependent on the structure of loops within the source code.
6854 Disabled at level @option{-Os}.
6857 @itemx -fno-peephole2
6858 @opindex fno-peephole
6859 @opindex fno-peephole2
6860 Disable any machine-specific peephole optimizations. The difference
6861 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6862 are implemented in the compiler; some targets use one, some use the
6863 other, a few use both.
6865 @option{-fpeephole} is enabled by default.
6866 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6868 @item -fno-guess-branch-probability
6869 @opindex fno-guess-branch-probability
6870 Do not guess branch probabilities using heuristics.
6872 GCC will use heuristics to guess branch probabilities if they are
6873 not provided by profiling feedback (@option{-fprofile-arcs}). These
6874 heuristics are based on the control flow graph. If some branch probabilities
6875 are specified by @samp{__builtin_expect}, then the heuristics will be
6876 used to guess branch probabilities for the rest of the control flow graph,
6877 taking the @samp{__builtin_expect} info into account. The interactions
6878 between the heuristics and @samp{__builtin_expect} can be complex, and in
6879 some cases, it may be useful to disable the heuristics so that the effects
6880 of @samp{__builtin_expect} are easier to understand.
6882 The default is @option{-fguess-branch-probability} at levels
6883 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6885 @item -freorder-blocks
6886 @opindex freorder-blocks
6887 Reorder basic blocks in the compiled function in order to reduce number of
6888 taken branches and improve code locality.
6890 Enabled at levels @option{-O2}, @option{-O3}.
6892 @item -freorder-blocks-and-partition
6893 @opindex freorder-blocks-and-partition
6894 In addition to reordering basic blocks in the compiled function, in order
6895 to reduce number of taken branches, partitions hot and cold basic blocks
6896 into separate sections of the assembly and .o files, to improve
6897 paging and cache locality performance.
6899 This optimization is automatically turned off in the presence of
6900 exception handling, for linkonce sections, for functions with a user-defined
6901 section attribute and on any architecture that does not support named
6904 @item -freorder-functions
6905 @opindex freorder-functions
6906 Reorder functions in the object file in order to
6907 improve code locality. This is implemented by using special
6908 subsections @code{.text.hot} for most frequently executed functions and
6909 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6910 the linker so object file format must support named sections and linker must
6911 place them in a reasonable way.
6913 Also profile feedback must be available in to make this option effective. See
6914 @option{-fprofile-arcs} for details.
6916 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6918 @item -fstrict-aliasing
6919 @opindex fstrict-aliasing
6920 Allow the compiler to assume the strictest aliasing rules applicable to
6921 the language being compiled. For C (and C++), this activates
6922 optimizations based on the type of expressions. In particular, an
6923 object of one type is assumed never to reside at the same address as an
6924 object of a different type, unless the types are almost the same. For
6925 example, an @code{unsigned int} can alias an @code{int}, but not a
6926 @code{void*} or a @code{double}. A character type may alias any other
6929 @anchor{Type-punning}Pay special attention to code like this:
6942 The practice of reading from a different union member than the one most
6943 recently written to (called ``type-punning'') is common. Even with
6944 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6945 is accessed through the union type. So, the code above will work as
6946 expected. @xref{Structures unions enumerations and bit-fields
6947 implementation}. However, this code might not:
6958 Similarly, access by taking the address, casting the resulting pointer
6959 and dereferencing the result has undefined behavior, even if the cast
6960 uses a union type, e.g.:
6964 return ((union a_union *) &d)->i;
6968 The @option{-fstrict-aliasing} option is enabled at levels
6969 @option{-O2}, @option{-O3}, @option{-Os}.
6971 @item -fstrict-overflow
6972 @opindex fstrict-overflow
6973 Allow the compiler to assume strict signed overflow rules, depending
6974 on the language being compiled. For C (and C++) this means that
6975 overflow when doing arithmetic with signed numbers is undefined, which
6976 means that the compiler may assume that it will not happen. This
6977 permits various optimizations. For example, the compiler will assume
6978 that an expression like @code{i + 10 > i} will always be true for
6979 signed @code{i}. This assumption is only valid if signed overflow is
6980 undefined, as the expression is false if @code{i + 10} overflows when
6981 using twos complement arithmetic. When this option is in effect any
6982 attempt to determine whether an operation on signed numbers will
6983 overflow must be written carefully to not actually involve overflow.
6985 This option also allows the compiler to assume strict pointer
6986 semantics: given a pointer to an object, if adding an offset to that
6987 pointer does not produce a pointer to the same object, the addition is
6988 undefined. This permits the compiler to conclude that @code{p + u >
6989 p} is always true for a pointer @code{p} and unsigned integer
6990 @code{u}. This assumption is only valid because pointer wraparound is
6991 undefined, as the expression is false if @code{p + u} overflows using
6992 twos complement arithmetic.
6994 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6995 that integer signed overflow is fully defined: it wraps. When
6996 @option{-fwrapv} is used, there is no difference between
6997 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6998 integers. With @option{-fwrapv} certain types of overflow are
6999 permitted. For example, if the compiler gets an overflow when doing
7000 arithmetic on constants, the overflowed value can still be used with
7001 @option{-fwrapv}, but not otherwise.
7003 The @option{-fstrict-overflow} option is enabled at levels
7004 @option{-O2}, @option{-O3}, @option{-Os}.
7006 @item -falign-functions
7007 @itemx -falign-functions=@var{n}
7008 @opindex falign-functions
7009 Align the start of functions to the next power-of-two greater than
7010 @var{n}, skipping up to @var{n} bytes. For instance,
7011 @option{-falign-functions=32} aligns functions to the next 32-byte
7012 boundary, but @option{-falign-functions=24} would align to the next
7013 32-byte boundary only if this can be done by skipping 23 bytes or less.
7015 @option{-fno-align-functions} and @option{-falign-functions=1} are
7016 equivalent and mean that functions will not be aligned.
7018 Some assemblers only support this flag when @var{n} is a power of two;
7019 in that case, it is rounded up.
7021 If @var{n} is not specified or is zero, use a machine-dependent default.
7023 Enabled at levels @option{-O2}, @option{-O3}.
7025 @item -falign-labels
7026 @itemx -falign-labels=@var{n}
7027 @opindex falign-labels
7028 Align all branch targets to a power-of-two boundary, skipping up to
7029 @var{n} bytes like @option{-falign-functions}. This option can easily
7030 make code slower, because it must insert dummy operations for when the
7031 branch target is reached in the usual flow of the code.
7033 @option{-fno-align-labels} and @option{-falign-labels=1} are
7034 equivalent and mean that labels will not be aligned.
7036 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7037 are greater than this value, then their values are used instead.
7039 If @var{n} is not specified or is zero, use a machine-dependent default
7040 which is very likely to be @samp{1}, meaning no alignment.
7042 Enabled at levels @option{-O2}, @option{-O3}.
7045 @itemx -falign-loops=@var{n}
7046 @opindex falign-loops
7047 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7048 like @option{-falign-functions}. The hope is that the loop will be
7049 executed many times, which will make up for any execution of the dummy
7052 @option{-fno-align-loops} and @option{-falign-loops=1} are
7053 equivalent and mean that loops will not be aligned.
7055 If @var{n} is not specified or is zero, use a machine-dependent default.
7057 Enabled at levels @option{-O2}, @option{-O3}.
7060 @itemx -falign-jumps=@var{n}
7061 @opindex falign-jumps
7062 Align branch targets to a power-of-two boundary, for branch targets
7063 where the targets can only be reached by jumping, skipping up to @var{n}
7064 bytes like @option{-falign-functions}. In this case, no dummy operations
7067 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7068 equivalent and mean that loops will not be aligned.
7070 If @var{n} is not specified or is zero, use a machine-dependent default.
7072 Enabled at levels @option{-O2}, @option{-O3}.
7074 @item -funit-at-a-time
7075 @opindex funit-at-a-time
7076 This option is left for compatibility reasons. @option{-funit-at-a-time}
7077 has no effect, while @option{-fno-unit-at-a-time} implies
7078 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7082 @item -fno-toplevel-reorder
7083 @opindex fno-toplevel-reorder
7084 Do not reorder top-level functions, variables, and @code{asm}
7085 statements. Output them in the same order that they appear in the
7086 input file. When this option is used, unreferenced static variables
7087 will not be removed. This option is intended to support existing code
7088 which relies on a particular ordering. For new code, it is better to
7091 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7092 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7097 Constructs webs as commonly used for register allocation purposes and assign
7098 each web individual pseudo register. This allows the register allocation pass
7099 to operate on pseudos directly, but also strengthens several other optimization
7100 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7101 however, make debugging impossible, since variables will no longer stay in a
7104 Enabled by default with @option{-funroll-loops}.
7106 @item -fwhole-program
7107 @opindex fwhole-program
7108 Assume that the current compilation unit represents the whole program being
7109 compiled. All public functions and variables with the exception of @code{main}
7110 and those merged by attribute @code{externally_visible} become static functions
7111 and in effect are optimized more aggressively by interprocedural optimizers.
7112 While this option is equivalent to proper use of the @code{static} keyword for
7113 programs consisting of a single file, in combination with option
7114 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7115 compile many smaller scale programs since the functions and variables become
7116 local for the whole combined compilation unit, not for the single source file
7119 This option implies @option{-fwhole-file} for Fortran programs.
7123 This option runs the standard link-time optimizer. When invoked
7124 with source code, it generates GIMPLE (one of GCC's internal
7125 representations) and writes it to special ELF sections in the object
7126 file. When the object files are linked together, all the function
7127 bodies are read from these ELF sections and instantiated as if they
7128 had been part of the same translation unit.
7130 To use the link-timer optimizer, @option{-flto} needs to be specified at
7131 compile time and during the final link. For example,
7134 gcc -c -O2 -flto foo.c
7135 gcc -c -O2 -flto bar.c
7136 gcc -o myprog -flto -O2 foo.o bar.o
7139 The first two invocations to GCC will save a bytecode representation
7140 of GIMPLE into special ELF sections inside @file{foo.o} and
7141 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7142 @file{foo.o} and @file{bar.o}, merge the two files into a single
7143 internal image, and compile the result as usual. Since both
7144 @file{foo.o} and @file{bar.o} are merged into a single image, this
7145 causes all the inter-procedural analyses and optimizations in GCC to
7146 work across the two files as if they were a single one. This means,
7147 for example, that the inliner will be able to inline functions in
7148 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7150 Another (simpler) way to enable link-time optimization is,
7153 gcc -o myprog -flto -O2 foo.c bar.c
7156 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7157 merge them together into a single GIMPLE representation and optimize
7158 them as usual to produce @file{myprog}.
7160 The only important thing to keep in mind is that to enable link-time
7161 optimizations the @option{-flto} flag needs to be passed to both the
7162 compile and the link commands.
7164 Note that when a file is compiled with @option{-flto}, the generated
7165 object file will be larger than a regular object file because it will
7166 contain GIMPLE bytecodes and the usual final code. This means that
7167 object files with LTO information can be linked as a normal object
7168 file. So, in the previous example, if the final link is done with
7171 gcc -o myprog foo.o bar.o
7174 The only difference will be that no inter-procedural optimizations
7175 will be applied to produce @file{myprog}. The two object files
7176 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7179 Additionally, the optimization flags used to compile individual files
7180 are not necessarily related to those used at link-time. For instance,
7183 gcc -c -O0 -flto foo.c
7184 gcc -c -O0 -flto bar.c
7185 gcc -o myprog -flto -O3 foo.o bar.o
7188 This will produce individual object files with unoptimized assembler
7189 code, but the resulting binary @file{myprog} will be optimized at
7190 @option{-O3}. Now, if the final binary is generated without
7191 @option{-flto}, then @file{myprog} will not be optimized.
7193 When producing the final binary with @option{-flto}, GCC will only
7194 apply link-time optimizations to those files that contain bytecode.
7195 Therefore, you can mix and match object files and libraries with
7196 GIMPLE bytecodes and final object code. GCC will automatically select
7197 which files to optimize in LTO mode and which files to link without
7200 There are some code generation flags that GCC will preserve when
7201 generating bytecodes, as they need to be used during the final link
7202 stage. Currently, the following options are saved into the GIMPLE
7203 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7204 @option{-m} target flags.
7206 At link time, these options are read-in and reapplied. Note that the
7207 current implementation makes no attempt at recognizing conflicting
7208 values for these options. If two or more files have a conflicting
7209 value (e.g., one file is compiled with @option{-fPIC} and another
7210 isn't), the compiler will simply use the last value read from the
7211 bytecode files. It is recommended, then, that all the files
7212 participating in the same link be compiled with the same options.
7214 Another feature of LTO is that it is possible to apply interprocedural
7215 optimizations on files written in different languages. This requires
7216 some support in the language front end. Currently, the C, C++ and
7217 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7218 something like this should work
7223 gfortran -c -flto baz.f90
7224 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7227 Notice that the final link is done with @command{g++} to get the C++
7228 runtime libraries and @option{-lgfortran} is added to get the Fortran
7229 runtime libraries. In general, when mixing languages in LTO mode, you
7230 should use the same link command used when mixing languages in a
7231 regular (non-LTO) compilation. This means that if your build process
7232 was mixing languages before, all you need to add is @option{-flto} to
7233 all the compile and link commands.
7235 If object files containing GIMPLE bytecode are stored in a library
7236 archive, say @file{libfoo.a}, it is possible to extract and use them
7237 in an LTO link if you are using @command{gold} as the linker (which,
7238 in turn requires GCC to be configured with @option{--enable-gold}).
7239 To enable this feature, use the flag @option{-use-linker-plugin} at
7243 gcc -o myprog -O2 -flto -use-linker-plugin a.o b.o -lfoo
7246 With the linker plugin enabled, @command{gold} will extract the needed
7247 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7248 to make them part of the aggregated GIMPLE image to be optimized.
7250 If you are not using @command{gold} and/or do not specify
7251 @option{-use-linker-plugin} then the objects inside @file{libfoo.a}
7252 will be extracted and linked as usual, but they will not participate
7253 in the LTO optimization process.
7255 Link time optimizations do not require the presence of the whole
7256 program to operate. If the program does not require any symbols to
7257 be exported, it is possible to combine @option{-flto} and
7258 @option{-fwhopr} with @option{-fwhole-program} to allow the
7259 interprocedural optimizers to use more aggressive assumptions which
7260 may lead to improved optimization opportunities.
7262 Regarding portability: the current implementation of LTO makes no
7263 attempt at generating bytecode that can be ported between different
7264 types of hosts. The bytecode files are versioned and there is a
7265 strict version check, so bytecode files generated in one version of
7266 GCC will not work with an older/newer version of GCC.
7268 This option is disabled by default.
7272 This option is identical in functionality to @option{-flto} but it
7273 differs in how the final link stage is executed. Instead of loading
7274 all the function bodies in memory, the callgraph is analyzed and
7275 optimization decisions are made (whole program analysis or WPA). Once
7276 optimization decisions are made, the callgraph is partitioned and the
7277 different sections are compiled separately (local transformations or
7278 LTRANS)@. This process allows optimizations on very large programs
7279 that otherwise would not fit in memory. This option enables
7280 @option{-fwpa} and @option{-fltrans} automatically.
7282 Disabled by default.
7286 This is an internal option used by GCC when compiling with
7287 @option{-fwhopr}. You should never need to use it.
7289 This option runs the link-time optimizer in the whole-program-analysis
7290 (WPA) mode, which reads in summary information from all inputs and
7291 performs a whole-program analysis based on summary information only.
7292 It generates object files for subsequent runs of the link-time
7293 optimizer where individual object files are optimized using both
7294 summary information from the WPA mode and the actual function bodies.
7295 It then drives the LTRANS phase.
7297 Disabled by default.
7301 This is an internal option used by GCC when compiling with
7302 @option{-fwhopr}. You should never need to use it.
7304 This option runs the link-time optimizer in the local-transformation (LTRANS)
7305 mode, which reads in output from a previous run of the LTO in WPA mode.
7306 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7308 Disabled by default.
7310 @item -fltrans-output-list=@var{file}
7311 @opindex fltrans-output-list
7312 This is an internal option used by GCC when compiling with
7313 @option{-fwhopr}. You should never need to use it.
7315 This option specifies a file to which the names of LTRANS output files are
7316 written. This option is only meaningful in conjunction with @option{-fwpa}.
7318 Disabled by default.
7320 @item -flto-compression-level=@var{n}
7321 This option specifies the level of compression used for intermediate
7322 language written to LTO object files, and is only meaningful in
7323 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7324 values are 0 (no compression) to 9 (maximum compression). Values
7325 outside this range are clamped to either 0 or 9. If the option is not
7326 given, a default balanced compression setting is used.
7329 Prints a report with internal details on the workings of the link-time
7330 optimizer. The contents of this report vary from version to version,
7331 it is meant to be useful to GCC developers when processing object
7332 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7334 Disabled by default.
7336 @item -use-linker-plugin
7337 Enables the extraction of objects with GIMPLE bytecode information
7338 from library archives. This option relies on features available only
7339 in @command{gold}, so to use this you must configure GCC with
7340 @option{--enable-gold}. See @option{-flto} for a description on the
7341 effect of this flag and how to use it.
7343 Disabled by default.
7345 @item -fcprop-registers
7346 @opindex fcprop-registers
7347 After register allocation and post-register allocation instruction splitting,
7348 we perform a copy-propagation pass to try to reduce scheduling dependencies
7349 and occasionally eliminate the copy.
7351 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7353 @item -fprofile-correction
7354 @opindex fprofile-correction
7355 Profiles collected using an instrumented binary for multi-threaded programs may
7356 be inconsistent due to missed counter updates. When this option is specified,
7357 GCC will use heuristics to correct or smooth out such inconsistencies. By
7358 default, GCC will emit an error message when an inconsistent profile is detected.
7360 @item -fprofile-dir=@var{path}
7361 @opindex fprofile-dir
7363 Set the directory to search the profile data files in to @var{path}.
7364 This option affects only the profile data generated by
7365 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7366 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7367 and its related options.
7368 By default, GCC will use the current directory as @var{path}
7369 thus the profile data file will appear in the same directory as the object file.
7371 @item -fprofile-generate
7372 @itemx -fprofile-generate=@var{path}
7373 @opindex fprofile-generate
7375 Enable options usually used for instrumenting application to produce
7376 profile useful for later recompilation with profile feedback based
7377 optimization. You must use @option{-fprofile-generate} both when
7378 compiling and when linking your program.
7380 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7382 If @var{path} is specified, GCC will look at the @var{path} to find
7383 the profile feedback data files. See @option{-fprofile-dir}.
7386 @itemx -fprofile-use=@var{path}
7387 @opindex fprofile-use
7388 Enable profile feedback directed optimizations, and optimizations
7389 generally profitable only with profile feedback available.
7391 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7392 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7394 By default, GCC emits an error message if the feedback profiles do not
7395 match the source code. This error can be turned into a warning by using
7396 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7399 If @var{path} is specified, GCC will look at the @var{path} to find
7400 the profile feedback data files. See @option{-fprofile-dir}.
7403 The following options control compiler behavior regarding floating
7404 point arithmetic. These options trade off between speed and
7405 correctness. All must be specifically enabled.
7409 @opindex ffloat-store
7410 Do not store floating point variables in registers, and inhibit other
7411 options that might change whether a floating point value is taken from a
7414 @cindex floating point precision
7415 This option prevents undesirable excess precision on machines such as
7416 the 68000 where the floating registers (of the 68881) keep more
7417 precision than a @code{double} is supposed to have. Similarly for the
7418 x86 architecture. For most programs, the excess precision does only
7419 good, but a few programs rely on the precise definition of IEEE floating
7420 point. Use @option{-ffloat-store} for such programs, after modifying
7421 them to store all pertinent intermediate computations into variables.
7423 @item -fexcess-precision=@var{style}
7424 @opindex fexcess-precision
7425 This option allows further control over excess precision on machines
7426 where floating-point registers have more precision than the IEEE
7427 @code{float} and @code{double} types and the processor does not
7428 support operations rounding to those types. By default,
7429 @option{-fexcess-precision=fast} is in effect; this means that
7430 operations are carried out in the precision of the registers and that
7431 it is unpredictable when rounding to the types specified in the source
7432 code takes place. When compiling C, if
7433 @option{-fexcess-precision=standard} is specified then excess
7434 precision will follow the rules specified in ISO C99; in particular,
7435 both casts and assignments cause values to be rounded to their
7436 semantic types (whereas @option{-ffloat-store} only affects
7437 assignments). This option is enabled by default for C if a strict
7438 conformance option such as @option{-std=c99} is used.
7441 @option{-fexcess-precision=standard} is not implemented for languages
7442 other than C, and has no effect if
7443 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7444 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7445 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7446 semantics apply without excess precision, and in the latter, rounding
7451 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7452 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7453 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7455 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7457 This option is not turned on by any @option{-O} option since
7458 it can result in incorrect output for programs which depend on
7459 an exact implementation of IEEE or ISO rules/specifications for
7460 math functions. It may, however, yield faster code for programs
7461 that do not require the guarantees of these specifications.
7463 @item -fno-math-errno
7464 @opindex fno-math-errno
7465 Do not set ERRNO after calling math functions that are executed
7466 with a single instruction, e.g., sqrt. A program that relies on
7467 IEEE exceptions for math error handling may want to use this flag
7468 for speed while maintaining IEEE arithmetic compatibility.
7470 This option is not turned on by any @option{-O} option since
7471 it can result in incorrect output for programs which depend on
7472 an exact implementation of IEEE or ISO rules/specifications for
7473 math functions. It may, however, yield faster code for programs
7474 that do not require the guarantees of these specifications.
7476 The default is @option{-fmath-errno}.
7478 On Darwin systems, the math library never sets @code{errno}. There is
7479 therefore no reason for the compiler to consider the possibility that
7480 it might, and @option{-fno-math-errno} is the default.
7482 @item -funsafe-math-optimizations
7483 @opindex funsafe-math-optimizations
7485 Allow optimizations for floating-point arithmetic that (a) assume
7486 that arguments and results are valid and (b) may violate IEEE or
7487 ANSI standards. When used at link-time, it may include libraries
7488 or startup files that change the default FPU control word or other
7489 similar optimizations.
7491 This option is not turned on by any @option{-O} option since
7492 it can result in incorrect output for programs which depend on
7493 an exact implementation of IEEE or ISO rules/specifications for
7494 math functions. It may, however, yield faster code for programs
7495 that do not require the guarantees of these specifications.
7496 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7497 @option{-fassociative-math} and @option{-freciprocal-math}.
7499 The default is @option{-fno-unsafe-math-optimizations}.
7501 @item -fassociative-math
7502 @opindex fassociative-math
7504 Allow re-association of operands in series of floating-point operations.
7505 This violates the ISO C and C++ language standard by possibly changing
7506 computation result. NOTE: re-ordering may change the sign of zero as
7507 well as ignore NaNs and inhibit or create underflow or overflow (and
7508 thus cannot be used on a code which relies on rounding behavior like
7509 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7510 and thus may not be used when ordered comparisons are required.
7511 This option requires that both @option{-fno-signed-zeros} and
7512 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7513 much sense with @option{-frounding-math}.
7515 The default is @option{-fno-associative-math}.
7517 @item -freciprocal-math
7518 @opindex freciprocal-math
7520 Allow the reciprocal of a value to be used instead of dividing by
7521 the value if this enables optimizations. For example @code{x / y}
7522 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7523 is subject to common subexpression elimination. Note that this loses
7524 precision and increases the number of flops operating on the value.
7526 The default is @option{-fno-reciprocal-math}.
7528 @item -ffinite-math-only
7529 @opindex ffinite-math-only
7530 Allow optimizations for floating-point arithmetic that assume
7531 that arguments and results are not NaNs or +-Infs.
7533 This option is not turned on by any @option{-O} option since
7534 it can result in incorrect output for programs which depend on
7535 an exact implementation of IEEE or ISO rules/specifications for
7536 math functions. It may, however, yield faster code for programs
7537 that do not require the guarantees of these specifications.
7539 The default is @option{-fno-finite-math-only}.
7541 @item -fno-signed-zeros
7542 @opindex fno-signed-zeros
7543 Allow optimizations for floating point arithmetic that ignore the
7544 signedness of zero. IEEE arithmetic specifies the behavior of
7545 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7546 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7547 This option implies that the sign of a zero result isn't significant.
7549 The default is @option{-fsigned-zeros}.
7551 @item -fno-trapping-math
7552 @opindex fno-trapping-math
7553 Compile code assuming that floating-point operations cannot generate
7554 user-visible traps. These traps include division by zero, overflow,
7555 underflow, inexact result and invalid operation. This option requires
7556 that @option{-fno-signaling-nans} be in effect. Setting this option may
7557 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7559 This option should never be turned on by any @option{-O} option since
7560 it can result in incorrect output for programs which depend on
7561 an exact implementation of IEEE or ISO rules/specifications for
7564 The default is @option{-ftrapping-math}.
7566 @item -frounding-math
7567 @opindex frounding-math
7568 Disable transformations and optimizations that assume default floating
7569 point rounding behavior. This is round-to-zero for all floating point
7570 to integer conversions, and round-to-nearest for all other arithmetic
7571 truncations. This option should be specified for programs that change
7572 the FP rounding mode dynamically, or that may be executed with a
7573 non-default rounding mode. This option disables constant folding of
7574 floating point expressions at compile-time (which may be affected by
7575 rounding mode) and arithmetic transformations that are unsafe in the
7576 presence of sign-dependent rounding modes.
7578 The default is @option{-fno-rounding-math}.
7580 This option is experimental and does not currently guarantee to
7581 disable all GCC optimizations that are affected by rounding mode.
7582 Future versions of GCC may provide finer control of this setting
7583 using C99's @code{FENV_ACCESS} pragma. This command line option
7584 will be used to specify the default state for @code{FENV_ACCESS}.
7586 @item -fsignaling-nans
7587 @opindex fsignaling-nans
7588 Compile code assuming that IEEE signaling NaNs may generate user-visible
7589 traps during floating-point operations. Setting this option disables
7590 optimizations that may change the number of exceptions visible with
7591 signaling NaNs. This option implies @option{-ftrapping-math}.
7593 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7596 The default is @option{-fno-signaling-nans}.
7598 This option is experimental and does not currently guarantee to
7599 disable all GCC optimizations that affect signaling NaN behavior.
7601 @item -fsingle-precision-constant
7602 @opindex fsingle-precision-constant
7603 Treat floating point constant as single precision constant instead of
7604 implicitly converting it to double precision constant.
7606 @item -fcx-limited-range
7607 @opindex fcx-limited-range
7608 When enabled, this option states that a range reduction step is not
7609 needed when performing complex division. Also, there is no checking
7610 whether the result of a complex multiplication or division is @code{NaN
7611 + I*NaN}, with an attempt to rescue the situation in that case. The
7612 default is @option{-fno-cx-limited-range}, but is enabled by
7613 @option{-ffast-math}.
7615 This option controls the default setting of the ISO C99
7616 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7619 @item -fcx-fortran-rules
7620 @opindex fcx-fortran-rules
7621 Complex multiplication and division follow Fortran rules. Range
7622 reduction is done as part of complex division, but there is no checking
7623 whether the result of a complex multiplication or division is @code{NaN
7624 + I*NaN}, with an attempt to rescue the situation in that case.
7626 The default is @option{-fno-cx-fortran-rules}.
7630 The following options control optimizations that may improve
7631 performance, but are not enabled by any @option{-O} options. This
7632 section includes experimental options that may produce broken code.
7635 @item -fbranch-probabilities
7636 @opindex fbranch-probabilities
7637 After running a program compiled with @option{-fprofile-arcs}
7638 (@pxref{Debugging Options,, Options for Debugging Your Program or
7639 @command{gcc}}), you can compile it a second time using
7640 @option{-fbranch-probabilities}, to improve optimizations based on
7641 the number of times each branch was taken. When the program
7642 compiled with @option{-fprofile-arcs} exits it saves arc execution
7643 counts to a file called @file{@var{sourcename}.gcda} for each source
7644 file. The information in this data file is very dependent on the
7645 structure of the generated code, so you must use the same source code
7646 and the same optimization options for both compilations.
7648 With @option{-fbranch-probabilities}, GCC puts a
7649 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7650 These can be used to improve optimization. Currently, they are only
7651 used in one place: in @file{reorg.c}, instead of guessing which path a
7652 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7653 exactly determine which path is taken more often.
7655 @item -fprofile-values
7656 @opindex fprofile-values
7657 If combined with @option{-fprofile-arcs}, it adds code so that some
7658 data about values of expressions in the program is gathered.
7660 With @option{-fbranch-probabilities}, it reads back the data gathered
7661 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7662 notes to instructions for their later usage in optimizations.
7664 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7668 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7669 a code to gather information about values of expressions.
7671 With @option{-fbranch-probabilities}, it reads back the data gathered
7672 and actually performs the optimizations based on them.
7673 Currently the optimizations include specialization of division operation
7674 using the knowledge about the value of the denominator.
7676 @item -frename-registers
7677 @opindex frename-registers
7678 Attempt to avoid false dependencies in scheduled code by making use
7679 of registers left over after register allocation. This optimization
7680 will most benefit processors with lots of registers. Depending on the
7681 debug information format adopted by the target, however, it can
7682 make debugging impossible, since variables will no longer stay in
7683 a ``home register''.
7685 Enabled by default with @option{-funroll-loops}.
7689 Perform tail duplication to enlarge superblock size. This transformation
7690 simplifies the control flow of the function allowing other optimizations to do
7693 Enabled with @option{-fprofile-use}.
7695 @item -funroll-loops
7696 @opindex funroll-loops
7697 Unroll loops whose number of iterations can be determined at compile time or
7698 upon entry to the loop. @option{-funroll-loops} implies
7699 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7700 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7701 small constant number of iterations). This option makes code larger, and may
7702 or may not make it run faster.
7704 Enabled with @option{-fprofile-use}.
7706 @item -funroll-all-loops
7707 @opindex funroll-all-loops
7708 Unroll all loops, even if their number of iterations is uncertain when
7709 the loop is entered. This usually makes programs run more slowly.
7710 @option{-funroll-all-loops} implies the same options as
7711 @option{-funroll-loops}.
7714 @opindex fpeel-loops
7715 Peels the loops for that there is enough information that they do not
7716 roll much (from profile feedback). It also turns on complete loop peeling
7717 (i.e.@: complete removal of loops with small constant number of iterations).
7719 Enabled with @option{-fprofile-use}.
7721 @item -fmove-loop-invariants
7722 @opindex fmove-loop-invariants
7723 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7724 at level @option{-O1}
7726 @item -funswitch-loops
7727 @opindex funswitch-loops
7728 Move branches with loop invariant conditions out of the loop, with duplicates
7729 of the loop on both branches (modified according to result of the condition).
7731 @item -ffunction-sections
7732 @itemx -fdata-sections
7733 @opindex ffunction-sections
7734 @opindex fdata-sections
7735 Place each function or data item into its own section in the output
7736 file if the target supports arbitrary sections. The name of the
7737 function or the name of the data item determines the section's name
7740 Use these options on systems where the linker can perform optimizations
7741 to improve locality of reference in the instruction space. Most systems
7742 using the ELF object format and SPARC processors running Solaris 2 have
7743 linkers with such optimizations. AIX may have these optimizations in
7746 Only use these options when there are significant benefits from doing
7747 so. When you specify these options, the assembler and linker will
7748 create larger object and executable files and will also be slower.
7749 You will not be able to use @code{gprof} on all systems if you
7750 specify this option and you may have problems with debugging if
7751 you specify both this option and @option{-g}.
7753 @item -fbranch-target-load-optimize
7754 @opindex fbranch-target-load-optimize
7755 Perform branch target register load optimization before prologue / epilogue
7757 The use of target registers can typically be exposed only during reload,
7758 thus hoisting loads out of loops and doing inter-block scheduling needs
7759 a separate optimization pass.
7761 @item -fbranch-target-load-optimize2
7762 @opindex fbranch-target-load-optimize2
7763 Perform branch target register load optimization after prologue / epilogue
7766 @item -fbtr-bb-exclusive
7767 @opindex fbtr-bb-exclusive
7768 When performing branch target register load optimization, don't reuse
7769 branch target registers in within any basic block.
7771 @item -fstack-protector
7772 @opindex fstack-protector
7773 Emit extra code to check for buffer overflows, such as stack smashing
7774 attacks. This is done by adding a guard variable to functions with
7775 vulnerable objects. This includes functions that call alloca, and
7776 functions with buffers larger than 8 bytes. The guards are initialized
7777 when a function is entered and then checked when the function exits.
7778 If a guard check fails, an error message is printed and the program exits.
7780 @item -fstack-protector-all
7781 @opindex fstack-protector-all
7782 Like @option{-fstack-protector} except that all functions are protected.
7784 @item -fsection-anchors
7785 @opindex fsection-anchors
7786 Try to reduce the number of symbolic address calculations by using
7787 shared ``anchor'' symbols to address nearby objects. This transformation
7788 can help to reduce the number of GOT entries and GOT accesses on some
7791 For example, the implementation of the following function @code{foo}:
7795 int foo (void) @{ return a + b + c; @}
7798 would usually calculate the addresses of all three variables, but if you
7799 compile it with @option{-fsection-anchors}, it will access the variables
7800 from a common anchor point instead. The effect is similar to the
7801 following pseudocode (which isn't valid C):
7806 register int *xr = &x;
7807 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7811 Not all targets support this option.
7813 @item --param @var{name}=@var{value}
7815 In some places, GCC uses various constants to control the amount of
7816 optimization that is done. For example, GCC will not inline functions
7817 that contain more that a certain number of instructions. You can
7818 control some of these constants on the command-line using the
7819 @option{--param} option.
7821 The names of specific parameters, and the meaning of the values, are
7822 tied to the internals of the compiler, and are subject to change
7823 without notice in future releases.
7825 In each case, the @var{value} is an integer. The allowable choices for
7826 @var{name} are given in the following table:
7829 @item struct-reorg-cold-struct-ratio
7830 The threshold ratio (as a percentage) between a structure frequency
7831 and the frequency of the hottest structure in the program. This parameter
7832 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7833 We say that if the ratio of a structure frequency, calculated by profiling,
7834 to the hottest structure frequency in the program is less than this
7835 parameter, then structure reorganization is not applied to this structure.
7838 @item predictable-branch-cost-outcome
7839 When branch is predicted to be taken with probability lower than this threshold
7840 (in percent), then it is considered well predictable. The default is 10.
7842 @item max-crossjump-edges
7843 The maximum number of incoming edges to consider for crossjumping.
7844 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7845 the number of edges incoming to each block. Increasing values mean
7846 more aggressive optimization, making the compile time increase with
7847 probably small improvement in executable size.
7849 @item min-crossjump-insns
7850 The minimum number of instructions which must be matched at the end
7851 of two blocks before crossjumping will be performed on them. This
7852 value is ignored in the case where all instructions in the block being
7853 crossjumped from are matched. The default value is 5.
7855 @item max-grow-copy-bb-insns
7856 The maximum code size expansion factor when copying basic blocks
7857 instead of jumping. The expansion is relative to a jump instruction.
7858 The default value is 8.
7860 @item max-goto-duplication-insns
7861 The maximum number of instructions to duplicate to a block that jumps
7862 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7863 passes, GCC factors computed gotos early in the compilation process,
7864 and unfactors them as late as possible. Only computed jumps at the
7865 end of a basic blocks with no more than max-goto-duplication-insns are
7866 unfactored. The default value is 8.
7868 @item max-delay-slot-insn-search
7869 The maximum number of instructions to consider when looking for an
7870 instruction to fill a delay slot. If more than this arbitrary number of
7871 instructions is searched, the time savings from filling the delay slot
7872 will be minimal so stop searching. Increasing values mean more
7873 aggressive optimization, making the compile time increase with probably
7874 small improvement in executable run time.
7876 @item max-delay-slot-live-search
7877 When trying to fill delay slots, the maximum number of instructions to
7878 consider when searching for a block with valid live register
7879 information. Increasing this arbitrarily chosen value means more
7880 aggressive optimization, increasing the compile time. This parameter
7881 should be removed when the delay slot code is rewritten to maintain the
7884 @item max-gcse-memory
7885 The approximate maximum amount of memory that will be allocated in
7886 order to perform the global common subexpression elimination
7887 optimization. If more memory than specified is required, the
7888 optimization will not be done.
7890 @item max-pending-list-length
7891 The maximum number of pending dependencies scheduling will allow
7892 before flushing the current state and starting over. Large functions
7893 with few branches or calls can create excessively large lists which
7894 needlessly consume memory and resources.
7896 @item max-inline-insns-single
7897 Several parameters control the tree inliner used in gcc.
7898 This number sets the maximum number of instructions (counted in GCC's
7899 internal representation) in a single function that the tree inliner
7900 will consider for inlining. This only affects functions declared
7901 inline and methods implemented in a class declaration (C++).
7902 The default value is 300.
7904 @item max-inline-insns-auto
7905 When you use @option{-finline-functions} (included in @option{-O3}),
7906 a lot of functions that would otherwise not be considered for inlining
7907 by the compiler will be investigated. To those functions, a different
7908 (more restrictive) limit compared to functions declared inline can
7910 The default value is 50.
7912 @item large-function-insns
7913 The limit specifying really large functions. For functions larger than this
7914 limit after inlining, inlining is constrained by
7915 @option{--param large-function-growth}. This parameter is useful primarily
7916 to avoid extreme compilation time caused by non-linear algorithms used by the
7918 The default value is 2700.
7920 @item large-function-growth
7921 Specifies maximal growth of large function caused by inlining in percents.
7922 The default value is 100 which limits large function growth to 2.0 times
7925 @item large-unit-insns
7926 The limit specifying large translation unit. Growth caused by inlining of
7927 units larger than this limit is limited by @option{--param inline-unit-growth}.
7928 For small units this might be too tight (consider unit consisting of function A
7929 that is inline and B that just calls A three time. If B is small relative to
7930 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7931 large units consisting of small inlineable functions however the overall unit
7932 growth limit is needed to avoid exponential explosion of code size. Thus for
7933 smaller units, the size is increased to @option{--param large-unit-insns}
7934 before applying @option{--param inline-unit-growth}. The default is 10000
7936 @item inline-unit-growth
7937 Specifies maximal overall growth of the compilation unit caused by inlining.
7938 The default value is 30 which limits unit growth to 1.3 times the original
7941 @item ipcp-unit-growth
7942 Specifies maximal overall growth of the compilation unit caused by
7943 interprocedural constant propagation. The default value is 10 which limits
7944 unit growth to 1.1 times the original size.
7946 @item large-stack-frame
7947 The limit specifying large stack frames. While inlining the algorithm is trying
7948 to not grow past this limit too much. Default value is 256 bytes.
7950 @item large-stack-frame-growth
7951 Specifies maximal growth of large stack frames caused by inlining in percents.
7952 The default value is 1000 which limits large stack frame growth to 11 times
7955 @item max-inline-insns-recursive
7956 @itemx max-inline-insns-recursive-auto
7957 Specifies maximum number of instructions out-of-line copy of self recursive inline
7958 function can grow into by performing recursive inlining.
7960 For functions declared inline @option{--param max-inline-insns-recursive} is
7961 taken into account. For function not declared inline, recursive inlining
7962 happens only when @option{-finline-functions} (included in @option{-O3}) is
7963 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7964 default value is 450.
7966 @item max-inline-recursive-depth
7967 @itemx max-inline-recursive-depth-auto
7968 Specifies maximum recursion depth used by the recursive inlining.
7970 For functions declared inline @option{--param max-inline-recursive-depth} is
7971 taken into account. For function not declared inline, recursive inlining
7972 happens only when @option{-finline-functions} (included in @option{-O3}) is
7973 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7976 @item min-inline-recursive-probability
7977 Recursive inlining is profitable only for function having deep recursion
7978 in average and can hurt for function having little recursion depth by
7979 increasing the prologue size or complexity of function body to other
7982 When profile feedback is available (see @option{-fprofile-generate}) the actual
7983 recursion depth can be guessed from probability that function will recurse via
7984 given call expression. This parameter limits inlining only to call expression
7985 whose probability exceeds given threshold (in percents). The default value is
7988 @item early-inlining-insns
7989 Specify growth that early inliner can make. In effect it increases amount of
7990 inlining for code having large abstraction penalty. The default value is 8.
7992 @item max-early-inliner-iterations
7993 @itemx max-early-inliner-iterations
7994 Limit of iterations of early inliner. This basically bounds number of nested
7995 indirect calls early inliner can resolve. Deeper chains are still handled by
7998 @item min-vect-loop-bound
7999 The minimum number of iterations under which a loop will not get vectorized
8000 when @option{-ftree-vectorize} is used. The number of iterations after
8001 vectorization needs to be greater than the value specified by this option
8002 to allow vectorization. The default value is 0.
8004 @item max-unrolled-insns
8005 The maximum number of instructions that a loop should have if that loop
8006 is unrolled, and if the loop is unrolled, it determines how many times
8007 the loop code is unrolled.
8009 @item max-average-unrolled-insns
8010 The maximum number of instructions biased by probabilities of their execution
8011 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8012 it determines how many times the loop code is unrolled.
8014 @item max-unroll-times
8015 The maximum number of unrollings of a single loop.
8017 @item max-peeled-insns
8018 The maximum number of instructions that a loop should have if that loop
8019 is peeled, and if the loop is peeled, it determines how many times
8020 the loop code is peeled.
8022 @item max-peel-times
8023 The maximum number of peelings of a single loop.
8025 @item max-completely-peeled-insns
8026 The maximum number of insns of a completely peeled loop.
8028 @item max-completely-peel-times
8029 The maximum number of iterations of a loop to be suitable for complete peeling.
8031 @item max-unswitch-insns
8032 The maximum number of insns of an unswitched loop.
8034 @item max-unswitch-level
8035 The maximum number of branches unswitched in a single loop.
8038 The minimum cost of an expensive expression in the loop invariant motion.
8040 @item iv-consider-all-candidates-bound
8041 Bound on number of candidates for induction variables below that
8042 all candidates are considered for each use in induction variable
8043 optimizations. Only the most relevant candidates are considered
8044 if there are more candidates, to avoid quadratic time complexity.
8046 @item iv-max-considered-uses
8047 The induction variable optimizations give up on loops that contain more
8048 induction variable uses.
8050 @item iv-always-prune-cand-set-bound
8051 If number of candidates in the set is smaller than this value,
8052 we always try to remove unnecessary ivs from the set during its
8053 optimization when a new iv is added to the set.
8055 @item scev-max-expr-size
8056 Bound on size of expressions used in the scalar evolutions analyzer.
8057 Large expressions slow the analyzer.
8059 @item omega-max-vars
8060 The maximum number of variables in an Omega constraint system.
8061 The default value is 128.
8063 @item omega-max-geqs
8064 The maximum number of inequalities in an Omega constraint system.
8065 The default value is 256.
8068 The maximum number of equalities in an Omega constraint system.
8069 The default value is 128.
8071 @item omega-max-wild-cards
8072 The maximum number of wildcard variables that the Omega solver will
8073 be able to insert. The default value is 18.
8075 @item omega-hash-table-size
8076 The size of the hash table in the Omega solver. The default value is
8079 @item omega-max-keys
8080 The maximal number of keys used by the Omega solver. The default
8083 @item omega-eliminate-redundant-constraints
8084 When set to 1, use expensive methods to eliminate all redundant
8085 constraints. The default value is 0.
8087 @item vect-max-version-for-alignment-checks
8088 The maximum number of runtime checks that can be performed when
8089 doing loop versioning for alignment in the vectorizer. See option
8090 ftree-vect-loop-version for more information.
8092 @item vect-max-version-for-alias-checks
8093 The maximum number of runtime checks that can be performed when
8094 doing loop versioning for alias in the vectorizer. See option
8095 ftree-vect-loop-version for more information.
8097 @item max-iterations-to-track
8099 The maximum number of iterations of a loop the brute force algorithm
8100 for analysis of # of iterations of the loop tries to evaluate.
8102 @item hot-bb-count-fraction
8103 Select fraction of the maximal count of repetitions of basic block in program
8104 given basic block needs to have to be considered hot.
8106 @item hot-bb-frequency-fraction
8107 Select fraction of the maximal frequency of executions of basic block in
8108 function given basic block needs to have to be considered hot
8110 @item max-predicted-iterations
8111 The maximum number of loop iterations we predict statically. This is useful
8112 in cases where function contain single loop with known bound and other loop
8113 with unknown. We predict the known number of iterations correctly, while
8114 the unknown number of iterations average to roughly 10. This means that the
8115 loop without bounds would appear artificially cold relative to the other one.
8117 @item align-threshold
8119 Select fraction of the maximal frequency of executions of basic block in
8120 function given basic block will get aligned.
8122 @item align-loop-iterations
8124 A loop expected to iterate at lest the selected number of iterations will get
8127 @item tracer-dynamic-coverage
8128 @itemx tracer-dynamic-coverage-feedback
8130 This value is used to limit superblock formation once the given percentage of
8131 executed instructions is covered. This limits unnecessary code size
8134 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8135 feedback is available. The real profiles (as opposed to statically estimated
8136 ones) are much less balanced allowing the threshold to be larger value.
8138 @item tracer-max-code-growth
8139 Stop tail duplication once code growth has reached given percentage. This is
8140 rather hokey argument, as most of the duplicates will be eliminated later in
8141 cross jumping, so it may be set to much higher values than is the desired code
8144 @item tracer-min-branch-ratio
8146 Stop reverse growth when the reverse probability of best edge is less than this
8147 threshold (in percent).
8149 @item tracer-min-branch-ratio
8150 @itemx tracer-min-branch-ratio-feedback
8152 Stop forward growth if the best edge do have probability lower than this
8155 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8156 compilation for profile feedback and one for compilation without. The value
8157 for compilation with profile feedback needs to be more conservative (higher) in
8158 order to make tracer effective.
8160 @item max-cse-path-length
8162 Maximum number of basic blocks on path that cse considers. The default is 10.
8165 The maximum instructions CSE process before flushing. The default is 1000.
8167 @item ggc-min-expand
8169 GCC uses a garbage collector to manage its own memory allocation. This
8170 parameter specifies the minimum percentage by which the garbage
8171 collector's heap should be allowed to expand between collections.
8172 Tuning this may improve compilation speed; it has no effect on code
8175 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8176 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8177 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8178 GCC is not able to calculate RAM on a particular platform, the lower
8179 bound of 30% is used. Setting this parameter and
8180 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8181 every opportunity. This is extremely slow, but can be useful for
8184 @item ggc-min-heapsize
8186 Minimum size of the garbage collector's heap before it begins bothering
8187 to collect garbage. The first collection occurs after the heap expands
8188 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8189 tuning this may improve compilation speed, and has no effect on code
8192 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8193 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8194 with a lower bound of 4096 (four megabytes) and an upper bound of
8195 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8196 particular platform, the lower bound is used. Setting this parameter
8197 very large effectively disables garbage collection. Setting this
8198 parameter and @option{ggc-min-expand} to zero causes a full collection
8199 to occur at every opportunity.
8201 @item max-reload-search-insns
8202 The maximum number of instruction reload should look backward for equivalent
8203 register. Increasing values mean more aggressive optimization, making the
8204 compile time increase with probably slightly better performance. The default
8207 @item max-cselib-memory-locations
8208 The maximum number of memory locations cselib should take into account.
8209 Increasing values mean more aggressive optimization, making the compile time
8210 increase with probably slightly better performance. The default value is 500.
8212 @item reorder-blocks-duplicate
8213 @itemx reorder-blocks-duplicate-feedback
8215 Used by basic block reordering pass to decide whether to use unconditional
8216 branch or duplicate the code on its destination. Code is duplicated when its
8217 estimated size is smaller than this value multiplied by the estimated size of
8218 unconditional jump in the hot spots of the program.
8220 The @option{reorder-block-duplicate-feedback} is used only when profile
8221 feedback is available and may be set to higher values than
8222 @option{reorder-block-duplicate} since information about the hot spots is more
8225 @item max-sched-ready-insns
8226 The maximum number of instructions ready to be issued the scheduler should
8227 consider at any given time during the first scheduling pass. Increasing
8228 values mean more thorough searches, making the compilation time increase
8229 with probably little benefit. The default value is 100.
8231 @item max-sched-region-blocks
8232 The maximum number of blocks in a region to be considered for
8233 interblock scheduling. The default value is 10.
8235 @item max-pipeline-region-blocks
8236 The maximum number of blocks in a region to be considered for
8237 pipelining in the selective scheduler. The default value is 15.
8239 @item max-sched-region-insns
8240 The maximum number of insns in a region to be considered for
8241 interblock scheduling. The default value is 100.
8243 @item max-pipeline-region-insns
8244 The maximum number of insns in a region to be considered for
8245 pipelining in the selective scheduler. The default value is 200.
8248 The minimum probability (in percents) of reaching a source block
8249 for interblock speculative scheduling. The default value is 40.
8251 @item max-sched-extend-regions-iters
8252 The maximum number of iterations through CFG to extend regions.
8253 0 - disable region extension,
8254 N - do at most N iterations.
8255 The default value is 0.
8257 @item max-sched-insn-conflict-delay
8258 The maximum conflict delay for an insn to be considered for speculative motion.
8259 The default value is 3.
8261 @item sched-spec-prob-cutoff
8262 The minimal probability of speculation success (in percents), so that
8263 speculative insn will be scheduled.
8264 The default value is 40.
8266 @item sched-mem-true-dep-cost
8267 Minimal distance (in CPU cycles) between store and load targeting same
8268 memory locations. The default value is 1.
8270 @item selsched-max-lookahead
8271 The maximum size of the lookahead window of selective scheduling. It is a
8272 depth of search for available instructions.
8273 The default value is 50.
8275 @item selsched-max-sched-times
8276 The maximum number of times that an instruction will be scheduled during
8277 selective scheduling. This is the limit on the number of iterations
8278 through which the instruction may be pipelined. The default value is 2.
8280 @item selsched-max-insns-to-rename
8281 The maximum number of best instructions in the ready list that are considered
8282 for renaming in the selective scheduler. The default value is 2.
8284 @item max-last-value-rtl
8285 The maximum size measured as number of RTLs that can be recorded in an expression
8286 in combiner for a pseudo register as last known value of that register. The default
8289 @item integer-share-limit
8290 Small integer constants can use a shared data structure, reducing the
8291 compiler's memory usage and increasing its speed. This sets the maximum
8292 value of a shared integer constant. The default value is 256.
8294 @item min-virtual-mappings
8295 Specifies the minimum number of virtual mappings in the incremental
8296 SSA updater that should be registered to trigger the virtual mappings
8297 heuristic defined by virtual-mappings-ratio. The default value is
8300 @item virtual-mappings-ratio
8301 If the number of virtual mappings is virtual-mappings-ratio bigger
8302 than the number of virtual symbols to be updated, then the incremental
8303 SSA updater switches to a full update for those symbols. The default
8306 @item ssp-buffer-size
8307 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8308 protection when @option{-fstack-protection} is used.
8310 @item max-jump-thread-duplication-stmts
8311 Maximum number of statements allowed in a block that needs to be
8312 duplicated when threading jumps.
8314 @item max-fields-for-field-sensitive
8315 Maximum number of fields in a structure we will treat in
8316 a field sensitive manner during pointer analysis. The default is zero
8317 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8319 @item prefetch-latency
8320 Estimate on average number of instructions that are executed before
8321 prefetch finishes. The distance we prefetch ahead is proportional
8322 to this constant. Increasing this number may also lead to less
8323 streams being prefetched (see @option{simultaneous-prefetches}).
8325 @item simultaneous-prefetches
8326 Maximum number of prefetches that can run at the same time.
8328 @item l1-cache-line-size
8329 The size of cache line in L1 cache, in bytes.
8332 The size of L1 cache, in kilobytes.
8335 The size of L2 cache, in kilobytes.
8337 @item min-insn-to-prefetch-ratio
8338 The minimum ratio between the number of instructions and the
8339 number of prefetches to enable prefetching in a loop with an
8342 @item prefetch-min-insn-to-mem-ratio
8343 The minimum ratio between the number of instructions and the
8344 number of memory references to enable prefetching in a loop.
8346 @item use-canonical-types
8347 Whether the compiler should use the ``canonical'' type system. By
8348 default, this should always be 1, which uses a more efficient internal
8349 mechanism for comparing types in C++ and Objective-C++. However, if
8350 bugs in the canonical type system are causing compilation failures,
8351 set this value to 0 to disable canonical types.
8353 @item switch-conversion-max-branch-ratio
8354 Switch initialization conversion will refuse to create arrays that are
8355 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8356 branches in the switch.
8358 @item max-partial-antic-length
8359 Maximum length of the partial antic set computed during the tree
8360 partial redundancy elimination optimization (@option{-ftree-pre}) when
8361 optimizing at @option{-O3} and above. For some sorts of source code
8362 the enhanced partial redundancy elimination optimization can run away,
8363 consuming all of the memory available on the host machine. This
8364 parameter sets a limit on the length of the sets that are computed,
8365 which prevents the runaway behavior. Setting a value of 0 for
8366 this parameter will allow an unlimited set length.
8368 @item sccvn-max-scc-size
8369 Maximum size of a strongly connected component (SCC) during SCCVN
8370 processing. If this limit is hit, SCCVN processing for the whole
8371 function will not be done and optimizations depending on it will
8372 be disabled. The default maximum SCC size is 10000.
8374 @item ira-max-loops-num
8375 IRA uses a regional register allocation by default. If a function
8376 contains loops more than number given by the parameter, only at most
8377 given number of the most frequently executed loops will form regions
8378 for the regional register allocation. The default value of the
8381 @item ira-max-conflict-table-size
8382 Although IRA uses a sophisticated algorithm of compression conflict
8383 table, the table can be still big for huge functions. If the conflict
8384 table for a function could be more than size in MB given by the
8385 parameter, the conflict table is not built and faster, simpler, and
8386 lower quality register allocation algorithm will be used. The
8387 algorithm do not use pseudo-register conflicts. The default value of
8388 the parameter is 2000.
8390 @item loop-invariant-max-bbs-in-loop
8391 Loop invariant motion can be very expensive, both in compile time and
8392 in amount of needed compile time memory, with very large loops. Loops
8393 with more basic blocks than this parameter won't have loop invariant
8394 motion optimization performed on them. The default value of the
8395 parameter is 1000 for -O1 and 10000 for -O2 and above.
8397 @item min-nondebug-insn-uid
8398 Use uids starting at this parameter for nondebug insns. The range below
8399 the parameter is reserved exclusively for debug insns created by
8400 @option{-fvar-tracking-assignments}, but debug insns may get
8401 (non-overlapping) uids above it if the reserved range is exhausted.
8403 @item ipa-sra-ptr-growth-factor
8404 IPA-SRA will replace a pointer to an aggregate with one or more new
8405 parameters only when their cumulative size is less or equal to
8406 @option{ipa-sra-ptr-growth-factor} times the size of the original
8412 @node Preprocessor Options
8413 @section Options Controlling the Preprocessor
8414 @cindex preprocessor options
8415 @cindex options, preprocessor
8417 These options control the C preprocessor, which is run on each C source
8418 file before actual compilation.
8420 If you use the @option{-E} option, nothing is done except preprocessing.
8421 Some of these options make sense only together with @option{-E} because
8422 they cause the preprocessor output to be unsuitable for actual
8426 @item -Wp,@var{option}
8428 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8429 and pass @var{option} directly through to the preprocessor. If
8430 @var{option} contains commas, it is split into multiple options at the
8431 commas. However, many options are modified, translated or interpreted
8432 by the compiler driver before being passed to the preprocessor, and
8433 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8434 interface is undocumented and subject to change, so whenever possible
8435 you should avoid using @option{-Wp} and let the driver handle the
8438 @item -Xpreprocessor @var{option}
8439 @opindex Xpreprocessor
8440 Pass @var{option} as an option to the preprocessor. You can use this to
8441 supply system-specific preprocessor options which GCC does not know how to
8444 If you want to pass an option that takes an argument, you must use
8445 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8448 @include cppopts.texi
8450 @node Assembler Options
8451 @section Passing Options to the Assembler
8453 @c prevent bad page break with this line
8454 You can pass options to the assembler.
8457 @item -Wa,@var{option}
8459 Pass @var{option} as an option to the assembler. If @var{option}
8460 contains commas, it is split into multiple options at the commas.
8462 @item -Xassembler @var{option}
8464 Pass @var{option} as an option to the assembler. You can use this to
8465 supply system-specific assembler options which GCC does not know how to
8468 If you want to pass an option that takes an argument, you must use
8469 @option{-Xassembler} twice, once for the option and once for the argument.
8474 @section Options for Linking
8475 @cindex link options
8476 @cindex options, linking
8478 These options come into play when the compiler links object files into
8479 an executable output file. They are meaningless if the compiler is
8480 not doing a link step.
8484 @item @var{object-file-name}
8485 A file name that does not end in a special recognized suffix is
8486 considered to name an object file or library. (Object files are
8487 distinguished from libraries by the linker according to the file
8488 contents.) If linking is done, these object files are used as input
8497 If any of these options is used, then the linker is not run, and
8498 object file names should not be used as arguments. @xref{Overall
8502 @item -l@var{library}
8503 @itemx -l @var{library}
8505 Search the library named @var{library} when linking. (The second
8506 alternative with the library as a separate argument is only for
8507 POSIX compliance and is not recommended.)
8509 It makes a difference where in the command you write this option; the
8510 linker searches and processes libraries and object files in the order they
8511 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8512 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8513 to functions in @samp{z}, those functions may not be loaded.
8515 The linker searches a standard list of directories for the library,
8516 which is actually a file named @file{lib@var{library}.a}. The linker
8517 then uses this file as if it had been specified precisely by name.
8519 The directories searched include several standard system directories
8520 plus any that you specify with @option{-L}.
8522 Normally the files found this way are library files---archive files
8523 whose members are object files. The linker handles an archive file by
8524 scanning through it for members which define symbols that have so far
8525 been referenced but not defined. But if the file that is found is an
8526 ordinary object file, it is linked in the usual fashion. The only
8527 difference between using an @option{-l} option and specifying a file name
8528 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8529 and searches several directories.
8533 You need this special case of the @option{-l} option in order to
8534 link an Objective-C or Objective-C++ program.
8537 @opindex nostartfiles
8538 Do not use the standard system startup files when linking.
8539 The standard system libraries are used normally, unless @option{-nostdlib}
8540 or @option{-nodefaultlibs} is used.
8542 @item -nodefaultlibs
8543 @opindex nodefaultlibs
8544 Do not use the standard system libraries when linking.
8545 Only the libraries you specify will be passed to the linker, options
8546 specifying linkage of the system libraries, such as @code{-static-libgcc}
8547 or @code{-shared-libgcc}, will be ignored.
8548 The standard startup files are used normally, unless @option{-nostartfiles}
8549 is used. The compiler may generate calls to @code{memcmp},
8550 @code{memset}, @code{memcpy} and @code{memmove}.
8551 These entries are usually resolved by entries in
8552 libc. These entry points should be supplied through some other
8553 mechanism when this option is specified.
8557 Do not use the standard system startup files or libraries when linking.
8558 No startup files and only the libraries you specify will be passed to
8559 the linker, options specifying linkage of the system libraries, such as
8560 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8561 The compiler may generate calls to @code{memcmp}, @code{memset},
8562 @code{memcpy} and @code{memmove}.
8563 These entries are usually resolved by entries in
8564 libc. These entry points should be supplied through some other
8565 mechanism when this option is specified.
8567 @cindex @option{-lgcc}, use with @option{-nostdlib}
8568 @cindex @option{-nostdlib} and unresolved references
8569 @cindex unresolved references and @option{-nostdlib}
8570 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8571 @cindex @option{-nodefaultlibs} and unresolved references
8572 @cindex unresolved references and @option{-nodefaultlibs}
8573 One of the standard libraries bypassed by @option{-nostdlib} and
8574 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8575 that GCC uses to overcome shortcomings of particular machines, or special
8576 needs for some languages.
8577 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8578 Collection (GCC) Internals},
8579 for more discussion of @file{libgcc.a}.)
8580 In most cases, you need @file{libgcc.a} even when you want to avoid
8581 other standard libraries. In other words, when you specify @option{-nostdlib}
8582 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8583 This ensures that you have no unresolved references to internal GCC
8584 library subroutines. (For example, @samp{__main}, used to ensure C++
8585 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8586 GNU Compiler Collection (GCC) Internals}.)
8590 Produce a position independent executable on targets which support it.
8591 For predictable results, you must also specify the same set of options
8592 that were used to generate code (@option{-fpie}, @option{-fPIE},
8593 or model suboptions) when you specify this option.
8597 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8598 that support it. This instructs the linker to add all symbols, not
8599 only used ones, to the dynamic symbol table. This option is needed
8600 for some uses of @code{dlopen} or to allow obtaining backtraces
8601 from within a program.
8605 Remove all symbol table and relocation information from the executable.
8609 On systems that support dynamic linking, this prevents linking with the shared
8610 libraries. On other systems, this option has no effect.
8614 Produce a shared object which can then be linked with other objects to
8615 form an executable. Not all systems support this option. For predictable
8616 results, you must also specify the same set of options that were used to
8617 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8618 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8619 needs to build supplementary stub code for constructors to work. On
8620 multi-libbed systems, @samp{gcc -shared} must select the correct support
8621 libraries to link against. Failing to supply the correct flags may lead
8622 to subtle defects. Supplying them in cases where they are not necessary
8625 @item -shared-libgcc
8626 @itemx -static-libgcc
8627 @opindex shared-libgcc
8628 @opindex static-libgcc
8629 On systems that provide @file{libgcc} as a shared library, these options
8630 force the use of either the shared or static version respectively.
8631 If no shared version of @file{libgcc} was built when the compiler was
8632 configured, these options have no effect.
8634 There are several situations in which an application should use the
8635 shared @file{libgcc} instead of the static version. The most common
8636 of these is when the application wishes to throw and catch exceptions
8637 across different shared libraries. In that case, each of the libraries
8638 as well as the application itself should use the shared @file{libgcc}.
8640 Therefore, the G++ and GCJ drivers automatically add
8641 @option{-shared-libgcc} whenever you build a shared library or a main
8642 executable, because C++ and Java programs typically use exceptions, so
8643 this is the right thing to do.
8645 If, instead, you use the GCC driver to create shared libraries, you may
8646 find that they will not always be linked with the shared @file{libgcc}.
8647 If GCC finds, at its configuration time, that you have a non-GNU linker
8648 or a GNU linker that does not support option @option{--eh-frame-hdr},
8649 it will link the shared version of @file{libgcc} into shared libraries
8650 by default. Otherwise, it will take advantage of the linker and optimize
8651 away the linking with the shared version of @file{libgcc}, linking with
8652 the static version of libgcc by default. This allows exceptions to
8653 propagate through such shared libraries, without incurring relocation
8654 costs at library load time.
8656 However, if a library or main executable is supposed to throw or catch
8657 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8658 for the languages used in the program, or using the option
8659 @option{-shared-libgcc}, such that it is linked with the shared
8662 @item -static-libstdc++
8663 When the @command{g++} program is used to link a C++ program, it will
8664 normally automatically link against @option{libstdc++}. If
8665 @file{libstdc++} is available as a shared library, and the
8666 @option{-static} option is not used, then this will link against the
8667 shared version of @file{libstdc++}. That is normally fine. However, it
8668 is sometimes useful to freeze the version of @file{libstdc++} used by
8669 the program without going all the way to a fully static link. The
8670 @option{-static-libstdc++} option directs the @command{g++} driver to
8671 link @file{libstdc++} statically, without necessarily linking other
8672 libraries statically.
8676 Bind references to global symbols when building a shared object. Warn
8677 about any unresolved references (unless overridden by the link editor
8678 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8681 @item -T @var{script}
8683 @cindex linker script
8684 Use @var{script} as the linker script. This option is supported by most
8685 systems using the GNU linker. On some targets, such as bare-board
8686 targets without an operating system, the @option{-T} option may be required
8687 when linking to avoid references to undefined symbols.
8689 @item -Xlinker @var{option}
8691 Pass @var{option} as an option to the linker. You can use this to
8692 supply system-specific linker options which GCC does not know how to
8695 If you want to pass an option that takes a separate argument, you must use
8696 @option{-Xlinker} twice, once for the option and once for the argument.
8697 For example, to pass @option{-assert definitions}, you must write
8698 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8699 @option{-Xlinker "-assert definitions"}, because this passes the entire
8700 string as a single argument, which is not what the linker expects.
8702 When using the GNU linker, it is usually more convenient to pass
8703 arguments to linker options using the @option{@var{option}=@var{value}}
8704 syntax than as separate arguments. For example, you can specify
8705 @samp{-Xlinker -Map=output.map} rather than
8706 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8707 this syntax for command-line options.
8709 @item -Wl,@var{option}
8711 Pass @var{option} as an option to the linker. If @var{option} contains
8712 commas, it is split into multiple options at the commas. You can use this
8713 syntax to pass an argument to the option.
8714 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8715 linker. When using the GNU linker, you can also get the same effect with
8716 @samp{-Wl,-Map=output.map}.
8718 @item -u @var{symbol}
8720 Pretend the symbol @var{symbol} is undefined, to force linking of
8721 library modules to define it. You can use @option{-u} multiple times with
8722 different symbols to force loading of additional library modules.
8725 @node Directory Options
8726 @section Options for Directory Search
8727 @cindex directory options
8728 @cindex options, directory search
8731 These options specify directories to search for header files, for
8732 libraries and for parts of the compiler:
8737 Add the directory @var{dir} to the head of the list of directories to be
8738 searched for header files. This can be used to override a system header
8739 file, substituting your own version, since these directories are
8740 searched before the system header file directories. However, you should
8741 not use this option to add directories that contain vendor-supplied
8742 system header files (use @option{-isystem} for that). If you use more than
8743 one @option{-I} option, the directories are scanned in left-to-right
8744 order; the standard system directories come after.
8746 If a standard system include directory, or a directory specified with
8747 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8748 option will be ignored. The directory will still be searched but as a
8749 system directory at its normal position in the system include chain.
8750 This is to ensure that GCC's procedure to fix buggy system headers and
8751 the ordering for the include_next directive are not inadvertently changed.
8752 If you really need to change the search order for system directories,
8753 use the @option{-nostdinc} and/or @option{-isystem} options.
8755 @item -iquote@var{dir}
8757 Add the directory @var{dir} to the head of the list of directories to
8758 be searched for header files only for the case of @samp{#include
8759 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8760 otherwise just like @option{-I}.
8764 Add directory @var{dir} to the list of directories to be searched
8767 @item -B@var{prefix}
8769 This option specifies where to find the executables, libraries,
8770 include files, and data files of the compiler itself.
8772 The compiler driver program runs one or more of the subprograms
8773 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8774 @var{prefix} as a prefix for each program it tries to run, both with and
8775 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8777 For each subprogram to be run, the compiler driver first tries the
8778 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8779 was not specified, the driver tries two standard prefixes, which are
8780 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8781 those results in a file name that is found, the unmodified program
8782 name is searched for using the directories specified in your
8783 @env{PATH} environment variable.
8785 The compiler will check to see if the path provided by the @option{-B}
8786 refers to a directory, and if necessary it will add a directory
8787 separator character at the end of the path.
8789 @option{-B} prefixes that effectively specify directory names also apply
8790 to libraries in the linker, because the compiler translates these
8791 options into @option{-L} options for the linker. They also apply to
8792 includes files in the preprocessor, because the compiler translates these
8793 options into @option{-isystem} options for the preprocessor. In this case,
8794 the compiler appends @samp{include} to the prefix.
8796 The run-time support file @file{libgcc.a} can also be searched for using
8797 the @option{-B} prefix, if needed. If it is not found there, the two
8798 standard prefixes above are tried, and that is all. The file is left
8799 out of the link if it is not found by those means.
8801 Another way to specify a prefix much like the @option{-B} prefix is to use
8802 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8805 As a special kludge, if the path provided by @option{-B} is
8806 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8807 9, then it will be replaced by @file{[dir/]include}. This is to help
8808 with boot-strapping the compiler.
8810 @item -specs=@var{file}
8812 Process @var{file} after the compiler reads in the standard @file{specs}
8813 file, in order to override the defaults that the @file{gcc} driver
8814 program uses when determining what switches to pass to @file{cc1},
8815 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8816 @option{-specs=@var{file}} can be specified on the command line, and they
8817 are processed in order, from left to right.
8819 @item --sysroot=@var{dir}
8821 Use @var{dir} as the logical root directory for headers and libraries.
8822 For example, if the compiler would normally search for headers in
8823 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8824 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8826 If you use both this option and the @option{-isysroot} option, then
8827 the @option{--sysroot} option will apply to libraries, but the
8828 @option{-isysroot} option will apply to header files.
8830 The GNU linker (beginning with version 2.16) has the necessary support
8831 for this option. If your linker does not support this option, the
8832 header file aspect of @option{--sysroot} will still work, but the
8833 library aspect will not.
8837 This option has been deprecated. Please use @option{-iquote} instead for
8838 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8839 Any directories you specify with @option{-I} options before the @option{-I-}
8840 option are searched only for the case of @samp{#include "@var{file}"};
8841 they are not searched for @samp{#include <@var{file}>}.
8843 If additional directories are specified with @option{-I} options after
8844 the @option{-I-}, these directories are searched for all @samp{#include}
8845 directives. (Ordinarily @emph{all} @option{-I} directories are used
8848 In addition, the @option{-I-} option inhibits the use of the current
8849 directory (where the current input file came from) as the first search
8850 directory for @samp{#include "@var{file}"}. There is no way to
8851 override this effect of @option{-I-}. With @option{-I.} you can specify
8852 searching the directory which was current when the compiler was
8853 invoked. That is not exactly the same as what the preprocessor does
8854 by default, but it is often satisfactory.
8856 @option{-I-} does not inhibit the use of the standard system directories
8857 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8864 @section Specifying subprocesses and the switches to pass to them
8867 @command{gcc} is a driver program. It performs its job by invoking a
8868 sequence of other programs to do the work of compiling, assembling and
8869 linking. GCC interprets its command-line parameters and uses these to
8870 deduce which programs it should invoke, and which command-line options
8871 it ought to place on their command lines. This behavior is controlled
8872 by @dfn{spec strings}. In most cases there is one spec string for each
8873 program that GCC can invoke, but a few programs have multiple spec
8874 strings to control their behavior. The spec strings built into GCC can
8875 be overridden by using the @option{-specs=} command-line switch to specify
8878 @dfn{Spec files} are plaintext files that are used to construct spec
8879 strings. They consist of a sequence of directives separated by blank
8880 lines. The type of directive is determined by the first non-whitespace
8881 character on the line and it can be one of the following:
8884 @item %@var{command}
8885 Issues a @var{command} to the spec file processor. The commands that can
8889 @item %include <@var{file}>
8891 Search for @var{file} and insert its text at the current point in the
8894 @item %include_noerr <@var{file}>
8895 @cindex %include_noerr
8896 Just like @samp{%include}, but do not generate an error message if the include
8897 file cannot be found.
8899 @item %rename @var{old_name} @var{new_name}
8901 Rename the spec string @var{old_name} to @var{new_name}.
8905 @item *[@var{spec_name}]:
8906 This tells the compiler to create, override or delete the named spec
8907 string. All lines after this directive up to the next directive or
8908 blank line are considered to be the text for the spec string. If this
8909 results in an empty string then the spec will be deleted. (Or, if the
8910 spec did not exist, then nothing will happened.) Otherwise, if the spec
8911 does not currently exist a new spec will be created. If the spec does
8912 exist then its contents will be overridden by the text of this
8913 directive, unless the first character of that text is the @samp{+}
8914 character, in which case the text will be appended to the spec.
8916 @item [@var{suffix}]:
8917 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8918 and up to the next directive or blank line are considered to make up the
8919 spec string for the indicated suffix. When the compiler encounters an
8920 input file with the named suffix, it will processes the spec string in
8921 order to work out how to compile that file. For example:
8928 This says that any input file whose name ends in @samp{.ZZ} should be
8929 passed to the program @samp{z-compile}, which should be invoked with the
8930 command-line switch @option{-input} and with the result of performing the
8931 @samp{%i} substitution. (See below.)
8933 As an alternative to providing a spec string, the text that follows a
8934 suffix directive can be one of the following:
8937 @item @@@var{language}
8938 This says that the suffix is an alias for a known @var{language}. This is
8939 similar to using the @option{-x} command-line switch to GCC to specify a
8940 language explicitly. For example:
8947 Says that .ZZ files are, in fact, C++ source files.
8950 This causes an error messages saying:
8953 @var{name} compiler not installed on this system.
8957 GCC already has an extensive list of suffixes built into it.
8958 This directive will add an entry to the end of the list of suffixes, but
8959 since the list is searched from the end backwards, it is effectively
8960 possible to override earlier entries using this technique.
8964 GCC has the following spec strings built into it. Spec files can
8965 override these strings or create their own. Note that individual
8966 targets can also add their own spec strings to this list.
8969 asm Options to pass to the assembler
8970 asm_final Options to pass to the assembler post-processor
8971 cpp Options to pass to the C preprocessor
8972 cc1 Options to pass to the C compiler
8973 cc1plus Options to pass to the C++ compiler
8974 endfile Object files to include at the end of the link
8975 link Options to pass to the linker
8976 lib Libraries to include on the command line to the linker
8977 libgcc Decides which GCC support library to pass to the linker
8978 linker Sets the name of the linker
8979 predefines Defines to be passed to the C preprocessor
8980 signed_char Defines to pass to CPP to say whether @code{char} is signed
8982 startfile Object files to include at the start of the link
8985 Here is a small example of a spec file:
8991 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8994 This example renames the spec called @samp{lib} to @samp{old_lib} and
8995 then overrides the previous definition of @samp{lib} with a new one.
8996 The new definition adds in some extra command-line options before
8997 including the text of the old definition.
8999 @dfn{Spec strings} are a list of command-line options to be passed to their
9000 corresponding program. In addition, the spec strings can contain
9001 @samp{%}-prefixed sequences to substitute variable text or to
9002 conditionally insert text into the command line. Using these constructs
9003 it is possible to generate quite complex command lines.
9005 Here is a table of all defined @samp{%}-sequences for spec
9006 strings. Note that spaces are not generated automatically around the
9007 results of expanding these sequences. Therefore you can concatenate them
9008 together or combine them with constant text in a single argument.
9012 Substitute one @samp{%} into the program name or argument.
9015 Substitute the name of the input file being processed.
9018 Substitute the basename of the input file being processed.
9019 This is the substring up to (and not including) the last period
9020 and not including the directory.
9023 This is the same as @samp{%b}, but include the file suffix (text after
9027 Marks the argument containing or following the @samp{%d} as a
9028 temporary file name, so that that file will be deleted if GCC exits
9029 successfully. Unlike @samp{%g}, this contributes no text to the
9032 @item %g@var{suffix}
9033 Substitute a file name that has suffix @var{suffix} and is chosen
9034 once per compilation, and mark the argument in the same way as
9035 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9036 name is now chosen in a way that is hard to predict even when previously
9037 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9038 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9039 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9040 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9041 was simply substituted with a file name chosen once per compilation,
9042 without regard to any appended suffix (which was therefore treated
9043 just like ordinary text), making such attacks more likely to succeed.
9045 @item %u@var{suffix}
9046 Like @samp{%g}, but generates a new temporary file name even if
9047 @samp{%u@var{suffix}} was already seen.
9049 @item %U@var{suffix}
9050 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9051 new one if there is no such last file name. In the absence of any
9052 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9053 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9054 would involve the generation of two distinct file names, one
9055 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9056 simply substituted with a file name chosen for the previous @samp{%u},
9057 without regard to any appended suffix.
9059 @item %j@var{suffix}
9060 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9061 writable, and if save-temps is off; otherwise, substitute the name
9062 of a temporary file, just like @samp{%u}. This temporary file is not
9063 meant for communication between processes, but rather as a junk
9066 @item %|@var{suffix}
9067 @itemx %m@var{suffix}
9068 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9069 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9070 all. These are the two most common ways to instruct a program that it
9071 should read from standard input or write to standard output. If you
9072 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9073 construct: see for example @file{f/lang-specs.h}.
9075 @item %.@var{SUFFIX}
9076 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9077 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9078 terminated by the next space or %.
9081 Marks the argument containing or following the @samp{%w} as the
9082 designated output file of this compilation. This puts the argument
9083 into the sequence of arguments that @samp{%o} will substitute later.
9086 Substitutes the names of all the output files, with spaces
9087 automatically placed around them. You should write spaces
9088 around the @samp{%o} as well or the results are undefined.
9089 @samp{%o} is for use in the specs for running the linker.
9090 Input files whose names have no recognized suffix are not compiled
9091 at all, but they are included among the output files, so they will
9095 Substitutes the suffix for object files. Note that this is
9096 handled specially when it immediately follows @samp{%g, %u, or %U},
9097 because of the need for those to form complete file names. The
9098 handling is such that @samp{%O} is treated exactly as if it had already
9099 been substituted, except that @samp{%g, %u, and %U} do not currently
9100 support additional @var{suffix} characters following @samp{%O} as they would
9101 following, for example, @samp{.o}.
9104 Substitutes the standard macro predefinitions for the
9105 current target machine. Use this when running @code{cpp}.
9108 Like @samp{%p}, but puts @samp{__} before and after the name of each
9109 predefined macro, except for macros that start with @samp{__} or with
9110 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9114 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9115 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9116 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9117 and @option{-imultilib} as necessary.
9120 Current argument is the name of a library or startup file of some sort.
9121 Search for that file in a standard list of directories and substitute
9122 the full name found. The current working directory is included in the
9123 list of directories scanned.
9126 Current argument is the name of a linker script. Search for that file
9127 in the current list of directories to scan for libraries. If the file
9128 is located insert a @option{--script} option into the command line
9129 followed by the full path name found. If the file is not found then
9130 generate an error message. Note: the current working directory is not
9134 Print @var{str} as an error message. @var{str} is terminated by a newline.
9135 Use this when inconsistent options are detected.
9138 Substitute the contents of spec string @var{name} at this point.
9141 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9143 @item %x@{@var{option}@}
9144 Accumulate an option for @samp{%X}.
9147 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9151 Output the accumulated assembler options specified by @option{-Wa}.
9154 Output the accumulated preprocessor options specified by @option{-Wp}.
9157 Process the @code{asm} spec. This is used to compute the
9158 switches to be passed to the assembler.
9161 Process the @code{asm_final} spec. This is a spec string for
9162 passing switches to an assembler post-processor, if such a program is
9166 Process the @code{link} spec. This is the spec for computing the
9167 command line passed to the linker. Typically it will make use of the
9168 @samp{%L %G %S %D and %E} sequences.
9171 Dump out a @option{-L} option for each directory that GCC believes might
9172 contain startup files. If the target supports multilibs then the
9173 current multilib directory will be prepended to each of these paths.
9176 Process the @code{lib} spec. This is a spec string for deciding which
9177 libraries should be included on the command line to the linker.
9180 Process the @code{libgcc} spec. This is a spec string for deciding
9181 which GCC support library should be included on the command line to the linker.
9184 Process the @code{startfile} spec. This is a spec for deciding which
9185 object files should be the first ones passed to the linker. Typically
9186 this might be a file named @file{crt0.o}.
9189 Process the @code{endfile} spec. This is a spec string that specifies
9190 the last object files that will be passed to the linker.
9193 Process the @code{cpp} spec. This is used to construct the arguments
9194 to be passed to the C preprocessor.
9197 Process the @code{cc1} spec. This is used to construct the options to be
9198 passed to the actual C compiler (@samp{cc1}).
9201 Process the @code{cc1plus} spec. This is used to construct the options to be
9202 passed to the actual C++ compiler (@samp{cc1plus}).
9205 Substitute the variable part of a matched option. See below.
9206 Note that each comma in the substituted string is replaced by
9210 Remove all occurrences of @code{-S} from the command line. Note---this
9211 command is position dependent. @samp{%} commands in the spec string
9212 before this one will see @code{-S}, @samp{%} commands in the spec string
9213 after this one will not.
9215 @item %:@var{function}(@var{args})
9216 Call the named function @var{function}, passing it @var{args}.
9217 @var{args} is first processed as a nested spec string, then split
9218 into an argument vector in the usual fashion. The function returns
9219 a string which is processed as if it had appeared literally as part
9220 of the current spec.
9222 The following built-in spec functions are provided:
9226 The @code{getenv} spec function takes two arguments: an environment
9227 variable name and a string. If the environment variable is not
9228 defined, a fatal error is issued. Otherwise, the return value is the
9229 value of the environment variable concatenated with the string. For
9230 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9233 %:getenv(TOPDIR /include)
9236 expands to @file{/path/to/top/include}.
9238 @item @code{if-exists}
9239 The @code{if-exists} spec function takes one argument, an absolute
9240 pathname to a file. If the file exists, @code{if-exists} returns the
9241 pathname. Here is a small example of its usage:
9245 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9248 @item @code{if-exists-else}
9249 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9250 spec function, except that it takes two arguments. The first argument is
9251 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9252 returns the pathname. If it does not exist, it returns the second argument.
9253 This way, @code{if-exists-else} can be used to select one file or another,
9254 based on the existence of the first. Here is a small example of its usage:
9258 crt0%O%s %:if-exists(crti%O%s) \
9259 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9262 @item @code{replace-outfile}
9263 The @code{replace-outfile} spec function takes two arguments. It looks for the
9264 first argument in the outfiles array and replaces it with the second argument. Here
9265 is a small example of its usage:
9268 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9271 @item @code{print-asm-header}
9272 The @code{print-asm-header} function takes no arguments and simply
9273 prints a banner like:
9279 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9282 It is used to separate compiler options from assembler options
9283 in the @option{--target-help} output.
9287 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9288 If that switch was not specified, this substitutes nothing. Note that
9289 the leading dash is omitted when specifying this option, and it is
9290 automatically inserted if the substitution is performed. Thus the spec
9291 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9292 and would output the command line option @option{-foo}.
9294 @item %W@{@code{S}@}
9295 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9298 @item %@{@code{S}*@}
9299 Substitutes all the switches specified to GCC whose names start
9300 with @code{-S}, but which also take an argument. This is used for
9301 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9302 GCC considers @option{-o foo} as being
9303 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9304 text, including the space. Thus two arguments would be generated.
9306 @item %@{@code{S}*&@code{T}*@}
9307 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9308 (the order of @code{S} and @code{T} in the spec is not significant).
9309 There can be any number of ampersand-separated variables; for each the
9310 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9312 @item %@{@code{S}:@code{X}@}
9313 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9315 @item %@{!@code{S}:@code{X}@}
9316 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9318 @item %@{@code{S}*:@code{X}@}
9319 Substitutes @code{X} if one or more switches whose names start with
9320 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9321 once, no matter how many such switches appeared. However, if @code{%*}
9322 appears somewhere in @code{X}, then @code{X} will be substituted once
9323 for each matching switch, with the @code{%*} replaced by the part of
9324 that switch that matched the @code{*}.
9326 @item %@{.@code{S}:@code{X}@}
9327 Substitutes @code{X}, if processing a file with suffix @code{S}.
9329 @item %@{!.@code{S}:@code{X}@}
9330 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9332 @item %@{,@code{S}:@code{X}@}
9333 Substitutes @code{X}, if processing a file for language @code{S}.
9335 @item %@{!,@code{S}:@code{X}@}
9336 Substitutes @code{X}, if not processing a file for language @code{S}.
9338 @item %@{@code{S}|@code{P}:@code{X}@}
9339 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9340 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9341 @code{*} sequences as well, although they have a stronger binding than
9342 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9343 alternatives must be starred, and only the first matching alternative
9346 For example, a spec string like this:
9349 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9352 will output the following command-line options from the following input
9353 command-line options:
9358 -d fred.c -foo -baz -boggle
9359 -d jim.d -bar -baz -boggle
9362 @item %@{S:X; T:Y; :D@}
9364 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9365 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9366 be as many clauses as you need. This may be combined with @code{.},
9367 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9372 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9373 construct may contain other nested @samp{%} constructs or spaces, or
9374 even newlines. They are processed as usual, as described above.
9375 Trailing white space in @code{X} is ignored. White space may also
9376 appear anywhere on the left side of the colon in these constructs,
9377 except between @code{.} or @code{*} and the corresponding word.
9379 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9380 handled specifically in these constructs. If another value of
9381 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9382 @option{-W} switch is found later in the command line, the earlier
9383 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9384 just one letter, which passes all matching options.
9386 The character @samp{|} at the beginning of the predicate text is used to
9387 indicate that a command should be piped to the following command, but
9388 only if @option{-pipe} is specified.
9390 It is built into GCC which switches take arguments and which do not.
9391 (You might think it would be useful to generalize this to allow each
9392 compiler's spec to say which switches take arguments. But this cannot
9393 be done in a consistent fashion. GCC cannot even decide which input
9394 files have been specified without knowing which switches take arguments,
9395 and it must know which input files to compile in order to tell which
9398 GCC also knows implicitly that arguments starting in @option{-l} are to be
9399 treated as compiler output files, and passed to the linker in their
9400 proper position among the other output files.
9402 @c man begin OPTIONS
9404 @node Target Options
9405 @section Specifying Target Machine and Compiler Version
9406 @cindex target options
9407 @cindex cross compiling
9408 @cindex specifying machine version
9409 @cindex specifying compiler version and target machine
9410 @cindex compiler version, specifying
9411 @cindex target machine, specifying
9413 The usual way to run GCC is to run the executable called @file{gcc}, or
9414 @file{<machine>-gcc} when cross-compiling, or
9415 @file{<machine>-gcc-<version>} to run a version other than the one that
9416 was installed last. Sometimes this is inconvenient, so GCC provides
9417 options that will switch to another cross-compiler or version.
9420 @item -b @var{machine}
9422 The argument @var{machine} specifies the target machine for compilation.
9424 The value to use for @var{machine} is the same as was specified as the
9425 machine type when configuring GCC as a cross-compiler. For
9426 example, if a cross-compiler was configured with @samp{configure
9427 arm-elf}, meaning to compile for an arm processor with elf binaries,
9428 then you would specify @option{-b arm-elf} to run that cross compiler.
9429 Because there are other options beginning with @option{-b}, the
9430 configuration must contain a hyphen, or @option{-b} alone should be one
9431 argument followed by the configuration in the next argument.
9433 @item -V @var{version}
9435 The argument @var{version} specifies which version of GCC to run.
9436 This is useful when multiple versions are installed. For example,
9437 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9440 The @option{-V} and @option{-b} options work by running the
9441 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9442 use them if you can just run that directly.
9444 @node Submodel Options
9445 @section Hardware Models and Configurations
9446 @cindex submodel options
9447 @cindex specifying hardware config
9448 @cindex hardware models and configurations, specifying
9449 @cindex machine dependent options
9451 Earlier we discussed the standard option @option{-b} which chooses among
9452 different installed compilers for completely different target
9453 machines, such as VAX vs.@: 68000 vs.@: 80386.
9455 In addition, each of these target machine types can have its own
9456 special options, starting with @samp{-m}, to choose among various
9457 hardware models or configurations---for example, 68010 vs 68020,
9458 floating coprocessor or none. A single installed version of the
9459 compiler can compile for any model or configuration, according to the
9462 Some configurations of the compiler also support additional special
9463 options, usually for compatibility with other compilers on the same
9466 @c This list is ordered alphanumerically by subsection name.
9467 @c It should be the same order and spelling as these options are listed
9468 @c in Machine Dependent Options
9474 * Blackfin Options::
9478 * DEC Alpha Options::
9479 * DEC Alpha/VMS Options::
9482 * GNU/Linux Options::
9485 * i386 and x86-64 Options::
9486 * i386 and x86-64 Windows Options::
9488 * IA-64/VMS Options::
9499 * picoChip Options::
9501 * RS/6000 and PowerPC Options::
9502 * S/390 and zSeries Options::
9507 * System V Options::
9512 * Xstormy16 Options::
9518 @subsection ARC Options
9521 These options are defined for ARC implementations:
9526 Compile code for little endian mode. This is the default.
9530 Compile code for big endian mode.
9533 @opindex mmangle-cpu
9534 Prepend the name of the cpu to all public symbol names.
9535 In multiple-processor systems, there are many ARC variants with different
9536 instruction and register set characteristics. This flag prevents code
9537 compiled for one cpu to be linked with code compiled for another.
9538 No facility exists for handling variants that are ``almost identical''.
9539 This is an all or nothing option.
9541 @item -mcpu=@var{cpu}
9543 Compile code for ARC variant @var{cpu}.
9544 Which variants are supported depend on the configuration.
9545 All variants support @option{-mcpu=base}, this is the default.
9547 @item -mtext=@var{text-section}
9548 @itemx -mdata=@var{data-section}
9549 @itemx -mrodata=@var{readonly-data-section}
9553 Put functions, data, and readonly data in @var{text-section},
9554 @var{data-section}, and @var{readonly-data-section} respectively
9555 by default. This can be overridden with the @code{section} attribute.
9556 @xref{Variable Attributes}.
9558 @item -mfix-cortex-m3-ldrd
9559 @opindex mfix-cortex-m3-ldrd
9560 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9561 with overlapping destination and base registers are used. This option avoids
9562 generating these instructions. This option is enabled by default when
9563 @option{-mcpu=cortex-m3} is specified.
9568 @subsection ARM Options
9571 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9575 @item -mabi=@var{name}
9577 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9578 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9581 @opindex mapcs-frame
9582 Generate a stack frame that is compliant with the ARM Procedure Call
9583 Standard for all functions, even if this is not strictly necessary for
9584 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9585 with this option will cause the stack frames not to be generated for
9586 leaf functions. The default is @option{-mno-apcs-frame}.
9590 This is a synonym for @option{-mapcs-frame}.
9593 @c not currently implemented
9594 @item -mapcs-stack-check
9595 @opindex mapcs-stack-check
9596 Generate code to check the amount of stack space available upon entry to
9597 every function (that actually uses some stack space). If there is
9598 insufficient space available then either the function
9599 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9600 called, depending upon the amount of stack space required. The run time
9601 system is required to provide these functions. The default is
9602 @option{-mno-apcs-stack-check}, since this produces smaller code.
9604 @c not currently implemented
9606 @opindex mapcs-float
9607 Pass floating point arguments using the float point registers. This is
9608 one of the variants of the APCS@. This option is recommended if the
9609 target hardware has a floating point unit or if a lot of floating point
9610 arithmetic is going to be performed by the code. The default is
9611 @option{-mno-apcs-float}, since integer only code is slightly increased in
9612 size if @option{-mapcs-float} is used.
9614 @c not currently implemented
9615 @item -mapcs-reentrant
9616 @opindex mapcs-reentrant
9617 Generate reentrant, position independent code. The default is
9618 @option{-mno-apcs-reentrant}.
9621 @item -mthumb-interwork
9622 @opindex mthumb-interwork
9623 Generate code which supports calling between the ARM and Thumb
9624 instruction sets. Without this option the two instruction sets cannot
9625 be reliably used inside one program. The default is
9626 @option{-mno-thumb-interwork}, since slightly larger code is generated
9627 when @option{-mthumb-interwork} is specified.
9629 @item -mno-sched-prolog
9630 @opindex mno-sched-prolog
9631 Prevent the reordering of instructions in the function prolog, or the
9632 merging of those instruction with the instructions in the function's
9633 body. This means that all functions will start with a recognizable set
9634 of instructions (or in fact one of a choice from a small set of
9635 different function prologues), and this information can be used to
9636 locate the start if functions inside an executable piece of code. The
9637 default is @option{-msched-prolog}.
9639 @item -mfloat-abi=@var{name}
9641 Specifies which floating-point ABI to use. Permissible values
9642 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9644 Specifying @samp{soft} causes GCC to generate output containing
9645 library calls for floating-point operations.
9646 @samp{softfp} allows the generation of code using hardware floating-point
9647 instructions, but still uses the soft-float calling conventions.
9648 @samp{hard} allows generation of floating-point instructions
9649 and uses FPU-specific calling conventions.
9651 The default depends on the specific target configuration. Note that
9652 the hard-float and soft-float ABIs are not link-compatible; you must
9653 compile your entire program with the same ABI, and link with a
9654 compatible set of libraries.
9657 @opindex mhard-float
9658 Equivalent to @option{-mfloat-abi=hard}.
9661 @opindex msoft-float
9662 Equivalent to @option{-mfloat-abi=soft}.
9664 @item -mlittle-endian
9665 @opindex mlittle-endian
9666 Generate code for a processor running in little-endian mode. This is
9667 the default for all standard configurations.
9670 @opindex mbig-endian
9671 Generate code for a processor running in big-endian mode; the default is
9672 to compile code for a little-endian processor.
9674 @item -mwords-little-endian
9675 @opindex mwords-little-endian
9676 This option only applies when generating code for big-endian processors.
9677 Generate code for a little-endian word order but a big-endian byte
9678 order. That is, a byte order of the form @samp{32107654}. Note: this
9679 option should only be used if you require compatibility with code for
9680 big-endian ARM processors generated by versions of the compiler prior to
9683 @item -mcpu=@var{name}
9685 This specifies the name of the target ARM processor. GCC uses this name
9686 to determine what kind of instructions it can emit when generating
9687 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9688 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9689 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9690 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9691 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9693 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9694 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9695 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9696 @samp{strongarm1110},
9697 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9698 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9699 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9700 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9701 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9702 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9703 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9704 @samp{cortex-a8}, @samp{cortex-a9},
9705 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9708 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9710 @item -mtune=@var{name}
9712 This option is very similar to the @option{-mcpu=} option, except that
9713 instead of specifying the actual target processor type, and hence
9714 restricting which instructions can be used, it specifies that GCC should
9715 tune the performance of the code as if the target were of the type
9716 specified in this option, but still choosing the instructions that it
9717 will generate based on the cpu specified by a @option{-mcpu=} option.
9718 For some ARM implementations better performance can be obtained by using
9721 @item -march=@var{name}
9723 This specifies the name of the target ARM architecture. GCC uses this
9724 name to determine what kind of instructions it can emit when generating
9725 assembly code. This option can be used in conjunction with or instead
9726 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9727 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9728 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9729 @samp{armv6}, @samp{armv6j},
9730 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9731 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9732 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9734 @item -mfpu=@var{name}
9735 @itemx -mfpe=@var{number}
9736 @itemx -mfp=@var{number}
9740 This specifies what floating point hardware (or hardware emulation) is
9741 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9742 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9743 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9744 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9745 with older versions of GCC@.
9747 If @option{-msoft-float} is specified this specifies the format of
9748 floating point values.
9750 @item -mfp16-format=@var{name}
9751 @opindex mfp16-format
9752 Specify the format of the @code{__fp16} half-precision floating-point type.
9753 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9754 the default is @samp{none}, in which case the @code{__fp16} type is not
9755 defined. @xref{Half-Precision}, for more information.
9757 @item -mstructure-size-boundary=@var{n}
9758 @opindex mstructure-size-boundary
9759 The size of all structures and unions will be rounded up to a multiple
9760 of the number of bits set by this option. Permissible values are 8, 32
9761 and 64. The default value varies for different toolchains. For the COFF
9762 targeted toolchain the default value is 8. A value of 64 is only allowed
9763 if the underlying ABI supports it.
9765 Specifying the larger number can produce faster, more efficient code, but
9766 can also increase the size of the program. Different values are potentially
9767 incompatible. Code compiled with one value cannot necessarily expect to
9768 work with code or libraries compiled with another value, if they exchange
9769 information using structures or unions.
9771 @item -mabort-on-noreturn
9772 @opindex mabort-on-noreturn
9773 Generate a call to the function @code{abort} at the end of a
9774 @code{noreturn} function. It will be executed if the function tries to
9778 @itemx -mno-long-calls
9779 @opindex mlong-calls
9780 @opindex mno-long-calls
9781 Tells the compiler to perform function calls by first loading the
9782 address of the function into a register and then performing a subroutine
9783 call on this register. This switch is needed if the target function
9784 will lie outside of the 64 megabyte addressing range of the offset based
9785 version of subroutine call instruction.
9787 Even if this switch is enabled, not all function calls will be turned
9788 into long calls. The heuristic is that static functions, functions
9789 which have the @samp{short-call} attribute, functions that are inside
9790 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9791 definitions have already been compiled within the current compilation
9792 unit, will not be turned into long calls. The exception to this rule is
9793 that weak function definitions, functions with the @samp{long-call}
9794 attribute or the @samp{section} attribute, and functions that are within
9795 the scope of a @samp{#pragma long_calls} directive, will always be
9796 turned into long calls.
9798 This feature is not enabled by default. Specifying
9799 @option{-mno-long-calls} will restore the default behavior, as will
9800 placing the function calls within the scope of a @samp{#pragma
9801 long_calls_off} directive. Note these switches have no effect on how
9802 the compiler generates code to handle function calls via function
9805 @item -msingle-pic-base
9806 @opindex msingle-pic-base
9807 Treat the register used for PIC addressing as read-only, rather than
9808 loading it in the prologue for each function. The run-time system is
9809 responsible for initializing this register with an appropriate value
9810 before execution begins.
9812 @item -mpic-register=@var{reg}
9813 @opindex mpic-register
9814 Specify the register to be used for PIC addressing. The default is R10
9815 unless stack-checking is enabled, when R9 is used.
9817 @item -mcirrus-fix-invalid-insns
9818 @opindex mcirrus-fix-invalid-insns
9819 @opindex mno-cirrus-fix-invalid-insns
9820 Insert NOPs into the instruction stream to in order to work around
9821 problems with invalid Maverick instruction combinations. This option
9822 is only valid if the @option{-mcpu=ep9312} option has been used to
9823 enable generation of instructions for the Cirrus Maverick floating
9824 point co-processor. This option is not enabled by default, since the
9825 problem is only present in older Maverick implementations. The default
9826 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9829 @item -mpoke-function-name
9830 @opindex mpoke-function-name
9831 Write the name of each function into the text section, directly
9832 preceding the function prologue. The generated code is similar to this:
9836 .ascii "arm_poke_function_name", 0
9839 .word 0xff000000 + (t1 - t0)
9840 arm_poke_function_name
9842 stmfd sp!, @{fp, ip, lr, pc@}
9846 When performing a stack backtrace, code can inspect the value of
9847 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9848 location @code{pc - 12} and the top 8 bits are set, then we know that
9849 there is a function name embedded immediately preceding this location
9850 and has length @code{((pc[-3]) & 0xff000000)}.
9854 Generate code for the Thumb instruction set. The default is to
9855 use the 32-bit ARM instruction set.
9856 This option automatically enables either 16-bit Thumb-1 or
9857 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9858 and @option{-march=@var{name}} options. This option is not passed to the
9859 assembler. If you want to force assembler files to be interpreted as Thumb code,
9860 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9861 option directly to the assembler by prefixing it with @option{-Wa}.
9864 @opindex mtpcs-frame
9865 Generate a stack frame that is compliant with the Thumb Procedure Call
9866 Standard for all non-leaf functions. (A leaf function is one that does
9867 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9869 @item -mtpcs-leaf-frame
9870 @opindex mtpcs-leaf-frame
9871 Generate a stack frame that is compliant with the Thumb Procedure Call
9872 Standard for all leaf functions. (A leaf function is one that does
9873 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9875 @item -mcallee-super-interworking
9876 @opindex mcallee-super-interworking
9877 Gives all externally visible functions in the file being compiled an ARM
9878 instruction set header which switches to Thumb mode before executing the
9879 rest of the function. This allows these functions to be called from
9880 non-interworking code. This option is not valid in AAPCS configurations
9881 because interworking is enabled by default.
9883 @item -mcaller-super-interworking
9884 @opindex mcaller-super-interworking
9885 Allows calls via function pointers (including virtual functions) to
9886 execute correctly regardless of whether the target code has been
9887 compiled for interworking or not. There is a small overhead in the cost
9888 of executing a function pointer if this option is enabled. This option
9889 is not valid in AAPCS configurations because interworking is enabled
9892 @item -mtp=@var{name}
9894 Specify the access model for the thread local storage pointer. The valid
9895 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9896 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9897 (supported in the arm6k architecture), and @option{auto}, which uses the
9898 best available method for the selected processor. The default setting is
9901 @item -mword-relocations
9902 @opindex mword-relocations
9903 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9904 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9905 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9911 @subsection AVR Options
9914 These options are defined for AVR implementations:
9917 @item -mmcu=@var{mcu}
9919 Specify ATMEL AVR instruction set or MCU type.
9921 Instruction set avr1 is for the minimal AVR core, not supported by the C
9922 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9923 attiny11, attiny12, attiny15, attiny28).
9925 Instruction set avr2 (default) is for the classic AVR core with up to
9926 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9927 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9928 at90c8534, at90s8535).
9930 Instruction set avr3 is for the classic AVR core with up to 128K program
9931 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9933 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9934 memory space (MCU types: atmega8, atmega83, atmega85).
9936 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9937 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9938 atmega64, atmega128, at43usb355, at94k).
9940 @item -mno-interrupts
9941 @opindex mno-interrupts
9942 Generated code is not compatible with hardware interrupts.
9943 Code size will be smaller.
9945 @item -mcall-prologues
9946 @opindex mcall-prologues
9947 Functions prologues/epilogues expanded as call to appropriate
9948 subroutines. Code size will be smaller.
9951 @opindex mtiny-stack
9952 Change only the low 8 bits of the stack pointer.
9956 Assume int to be 8 bit integer. This affects the sizes of all types: A
9957 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9958 and long long will be 4 bytes. Please note that this option does not
9959 comply to the C standards, but it will provide you with smaller code
9963 @node Blackfin Options
9964 @subsection Blackfin Options
9965 @cindex Blackfin Options
9968 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9970 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9971 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9972 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9973 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9974 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9975 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9976 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
9978 The optional @var{sirevision} specifies the silicon revision of the target
9979 Blackfin processor. Any workarounds available for the targeted silicon revision
9980 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9981 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9982 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9983 hexadecimal digits representing the major and minor numbers in the silicon
9984 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9985 is not defined. If @var{sirevision} is @samp{any}, the
9986 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9987 If this optional @var{sirevision} is not used, GCC assumes the latest known
9988 silicon revision of the targeted Blackfin processor.
9990 Support for @samp{bf561} is incomplete. For @samp{bf561},
9991 Only the processor macro is defined.
9992 Without this option, @samp{bf532} is used as the processor by default.
9993 The corresponding predefined processor macros for @var{cpu} is to
9994 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9995 provided by libgloss to be linked in if @option{-msim} is not given.
9999 Specifies that the program will be run on the simulator. This causes
10000 the simulator BSP provided by libgloss to be linked in. This option
10001 has effect only for @samp{bfin-elf} toolchain.
10002 Certain other options, such as @option{-mid-shared-library} and
10003 @option{-mfdpic}, imply @option{-msim}.
10005 @item -momit-leaf-frame-pointer
10006 @opindex momit-leaf-frame-pointer
10007 Don't keep the frame pointer in a register for leaf functions. This
10008 avoids the instructions to save, set up and restore frame pointers and
10009 makes an extra register available in leaf functions. The option
10010 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10011 which might make debugging harder.
10013 @item -mspecld-anomaly
10014 @opindex mspecld-anomaly
10015 When enabled, the compiler will ensure that the generated code does not
10016 contain speculative loads after jump instructions. If this option is used,
10017 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10019 @item -mno-specld-anomaly
10020 @opindex mno-specld-anomaly
10021 Don't generate extra code to prevent speculative loads from occurring.
10023 @item -mcsync-anomaly
10024 @opindex mcsync-anomaly
10025 When enabled, the compiler will ensure that the generated code does not
10026 contain CSYNC or SSYNC instructions too soon after conditional branches.
10027 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10029 @item -mno-csync-anomaly
10030 @opindex mno-csync-anomaly
10031 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10032 occurring too soon after a conditional branch.
10036 When enabled, the compiler is free to take advantage of the knowledge that
10037 the entire program fits into the low 64k of memory.
10040 @opindex mno-low-64k
10041 Assume that the program is arbitrarily large. This is the default.
10043 @item -mstack-check-l1
10044 @opindex mstack-check-l1
10045 Do stack checking using information placed into L1 scratchpad memory by the
10048 @item -mid-shared-library
10049 @opindex mid-shared-library
10050 Generate code that supports shared libraries via the library ID method.
10051 This allows for execute in place and shared libraries in an environment
10052 without virtual memory management. This option implies @option{-fPIC}.
10053 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10055 @item -mno-id-shared-library
10056 @opindex mno-id-shared-library
10057 Generate code that doesn't assume ID based shared libraries are being used.
10058 This is the default.
10060 @item -mleaf-id-shared-library
10061 @opindex mleaf-id-shared-library
10062 Generate code that supports shared libraries via the library ID method,
10063 but assumes that this library or executable won't link against any other
10064 ID shared libraries. That allows the compiler to use faster code for jumps
10067 @item -mno-leaf-id-shared-library
10068 @opindex mno-leaf-id-shared-library
10069 Do not assume that the code being compiled won't link against any ID shared
10070 libraries. Slower code will be generated for jump and call insns.
10072 @item -mshared-library-id=n
10073 @opindex mshared-library-id
10074 Specified the identification number of the ID based shared library being
10075 compiled. Specifying a value of 0 will generate more compact code, specifying
10076 other values will force the allocation of that number to the current
10077 library but is no more space or time efficient than omitting this option.
10081 Generate code that allows the data segment to be located in a different
10082 area of memory from the text segment. This allows for execute in place in
10083 an environment without virtual memory management by eliminating relocations
10084 against the text section.
10086 @item -mno-sep-data
10087 @opindex mno-sep-data
10088 Generate code that assumes that the data segment follows the text segment.
10089 This is the default.
10092 @itemx -mno-long-calls
10093 @opindex mlong-calls
10094 @opindex mno-long-calls
10095 Tells the compiler to perform function calls by first loading the
10096 address of the function into a register and then performing a subroutine
10097 call on this register. This switch is needed if the target function
10098 will lie outside of the 24 bit addressing range of the offset based
10099 version of subroutine call instruction.
10101 This feature is not enabled by default. Specifying
10102 @option{-mno-long-calls} will restore the default behavior. Note these
10103 switches have no effect on how the compiler generates code to handle
10104 function calls via function pointers.
10108 Link with the fast floating-point library. This library relaxes some of
10109 the IEEE floating-point standard's rules for checking inputs against
10110 Not-a-Number (NAN), in the interest of performance.
10113 @opindex minline-plt
10114 Enable inlining of PLT entries in function calls to functions that are
10115 not known to bind locally. It has no effect without @option{-mfdpic}.
10118 @opindex mmulticore
10119 Build standalone application for multicore Blackfin processor. Proper
10120 start files and link scripts will be used to support multicore.
10121 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10122 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10123 @option{-mcorea} or @option{-mcoreb}. If it's used without
10124 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10125 programming model is used. In this model, the main function of Core B
10126 should be named as coreb_main. If it's used with @option{-mcorea} or
10127 @option{-mcoreb}, one application per core programming model is used.
10128 If this option is not used, single core application programming
10133 Build standalone application for Core A of BF561 when using
10134 one application per core programming model. Proper start files
10135 and link scripts will be used to support Core A. This option
10136 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10140 Build standalone application for Core B of BF561 when using
10141 one application per core programming model. Proper start files
10142 and link scripts will be used to support Core B. This option
10143 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10144 should be used instead of main. It must be used with
10145 @option{-mmulticore}.
10149 Build standalone application for SDRAM. Proper start files and
10150 link scripts will be used to put the application into SDRAM.
10151 Loader should initialize SDRAM before loading the application
10152 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10156 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10157 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10158 are enabled; for standalone applications the default is off.
10162 @subsection CRIS Options
10163 @cindex CRIS Options
10165 These options are defined specifically for the CRIS ports.
10168 @item -march=@var{architecture-type}
10169 @itemx -mcpu=@var{architecture-type}
10172 Generate code for the specified architecture. The choices for
10173 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10174 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10175 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10178 @item -mtune=@var{architecture-type}
10180 Tune to @var{architecture-type} everything applicable about the generated
10181 code, except for the ABI and the set of available instructions. The
10182 choices for @var{architecture-type} are the same as for
10183 @option{-march=@var{architecture-type}}.
10185 @item -mmax-stack-frame=@var{n}
10186 @opindex mmax-stack-frame
10187 Warn when the stack frame of a function exceeds @var{n} bytes.
10193 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10194 @option{-march=v3} and @option{-march=v8} respectively.
10196 @item -mmul-bug-workaround
10197 @itemx -mno-mul-bug-workaround
10198 @opindex mmul-bug-workaround
10199 @opindex mno-mul-bug-workaround
10200 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10201 models where it applies. This option is active by default.
10205 Enable CRIS-specific verbose debug-related information in the assembly
10206 code. This option also has the effect to turn off the @samp{#NO_APP}
10207 formatted-code indicator to the assembler at the beginning of the
10212 Do not use condition-code results from previous instruction; always emit
10213 compare and test instructions before use of condition codes.
10215 @item -mno-side-effects
10216 @opindex mno-side-effects
10217 Do not emit instructions with side-effects in addressing modes other than
10220 @item -mstack-align
10221 @itemx -mno-stack-align
10222 @itemx -mdata-align
10223 @itemx -mno-data-align
10224 @itemx -mconst-align
10225 @itemx -mno-const-align
10226 @opindex mstack-align
10227 @opindex mno-stack-align
10228 @opindex mdata-align
10229 @opindex mno-data-align
10230 @opindex mconst-align
10231 @opindex mno-const-align
10232 These options (no-options) arranges (eliminate arrangements) for the
10233 stack-frame, individual data and constants to be aligned for the maximum
10234 single data access size for the chosen CPU model. The default is to
10235 arrange for 32-bit alignment. ABI details such as structure layout are
10236 not affected by these options.
10244 Similar to the stack- data- and const-align options above, these options
10245 arrange for stack-frame, writable data and constants to all be 32-bit,
10246 16-bit or 8-bit aligned. The default is 32-bit alignment.
10248 @item -mno-prologue-epilogue
10249 @itemx -mprologue-epilogue
10250 @opindex mno-prologue-epilogue
10251 @opindex mprologue-epilogue
10252 With @option{-mno-prologue-epilogue}, the normal function prologue and
10253 epilogue that sets up the stack-frame are omitted and no return
10254 instructions or return sequences are generated in the code. Use this
10255 option only together with visual inspection of the compiled code: no
10256 warnings or errors are generated when call-saved registers must be saved,
10257 or storage for local variable needs to be allocated.
10261 @opindex mno-gotplt
10263 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10264 instruction sequences that load addresses for functions from the PLT part
10265 of the GOT rather than (traditional on other architectures) calls to the
10266 PLT@. The default is @option{-mgotplt}.
10270 Legacy no-op option only recognized with the cris-axis-elf and
10271 cris-axis-linux-gnu targets.
10275 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10279 This option, recognized for the cris-axis-elf arranges
10280 to link with input-output functions from a simulator library. Code,
10281 initialized data and zero-initialized data are allocated consecutively.
10285 Like @option{-sim}, but pass linker options to locate initialized data at
10286 0x40000000 and zero-initialized data at 0x80000000.
10290 @subsection CRX Options
10291 @cindex CRX Options
10293 These options are defined specifically for the CRX ports.
10299 Enable the use of multiply-accumulate instructions. Disabled by default.
10302 @opindex mpush-args
10303 Push instructions will be used to pass outgoing arguments when functions
10304 are called. Enabled by default.
10307 @node Darwin Options
10308 @subsection Darwin Options
10309 @cindex Darwin options
10311 These options are defined for all architectures running the Darwin operating
10314 FSF GCC on Darwin does not create ``fat'' object files; it will create
10315 an object file for the single architecture that it was built to
10316 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10317 @option{-arch} options are used; it does so by running the compiler or
10318 linker multiple times and joining the results together with
10321 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10322 @samp{i686}) is determined by the flags that specify the ISA
10323 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10324 @option{-force_cpusubtype_ALL} option can be used to override this.
10326 The Darwin tools vary in their behavior when presented with an ISA
10327 mismatch. The assembler, @file{as}, will only permit instructions to
10328 be used that are valid for the subtype of the file it is generating,
10329 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10330 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10331 and print an error if asked to create a shared library with a less
10332 restrictive subtype than its input files (for instance, trying to put
10333 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10334 for executables, @file{ld}, will quietly give the executable the most
10335 restrictive subtype of any of its input files.
10340 Add the framework directory @var{dir} to the head of the list of
10341 directories to be searched for header files. These directories are
10342 interleaved with those specified by @option{-I} options and are
10343 scanned in a left-to-right order.
10345 A framework directory is a directory with frameworks in it. A
10346 framework is a directory with a @samp{"Headers"} and/or
10347 @samp{"PrivateHeaders"} directory contained directly in it that ends
10348 in @samp{".framework"}. The name of a framework is the name of this
10349 directory excluding the @samp{".framework"}. Headers associated with
10350 the framework are found in one of those two directories, with
10351 @samp{"Headers"} being searched first. A subframework is a framework
10352 directory that is in a framework's @samp{"Frameworks"} directory.
10353 Includes of subframework headers can only appear in a header of a
10354 framework that contains the subframework, or in a sibling subframework
10355 header. Two subframeworks are siblings if they occur in the same
10356 framework. A subframework should not have the same name as a
10357 framework, a warning will be issued if this is violated. Currently a
10358 subframework cannot have subframeworks, in the future, the mechanism
10359 may be extended to support this. The standard frameworks can be found
10360 in @samp{"/System/Library/Frameworks"} and
10361 @samp{"/Library/Frameworks"}. An example include looks like
10362 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10363 the name of the framework and header.h is found in the
10364 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10366 @item -iframework@var{dir}
10367 @opindex iframework
10368 Like @option{-F} except the directory is a treated as a system
10369 directory. The main difference between this @option{-iframework} and
10370 @option{-F} is that with @option{-iframework} the compiler does not
10371 warn about constructs contained within header files found via
10372 @var{dir}. This option is valid only for the C family of languages.
10376 Emit debugging information for symbols that are used. For STABS
10377 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10378 This is by default ON@.
10382 Emit debugging information for all symbols and types.
10384 @item -mmacosx-version-min=@var{version}
10385 The earliest version of MacOS X that this executable will run on
10386 is @var{version}. Typical values of @var{version} include @code{10.1},
10387 @code{10.2}, and @code{10.3.9}.
10389 If the compiler was built to use the system's headers by default,
10390 then the default for this option is the system version on which the
10391 compiler is running, otherwise the default is to make choices which
10392 are compatible with as many systems and code bases as possible.
10396 Enable kernel development mode. The @option{-mkernel} option sets
10397 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10398 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10399 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10400 applicable. This mode also sets @option{-mno-altivec},
10401 @option{-msoft-float}, @option{-fno-builtin} and
10402 @option{-mlong-branch} for PowerPC targets.
10404 @item -mone-byte-bool
10405 @opindex mone-byte-bool
10406 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10407 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10408 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10409 option has no effect on x86.
10411 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10412 to generate code that is not binary compatible with code generated
10413 without that switch. Using this switch may require recompiling all
10414 other modules in a program, including system libraries. Use this
10415 switch to conform to a non-default data model.
10417 @item -mfix-and-continue
10418 @itemx -ffix-and-continue
10419 @itemx -findirect-data
10420 @opindex mfix-and-continue
10421 @opindex ffix-and-continue
10422 @opindex findirect-data
10423 Generate code suitable for fast turn around development. Needed to
10424 enable gdb to dynamically load @code{.o} files into already running
10425 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10426 are provided for backwards compatibility.
10430 Loads all members of static archive libraries.
10431 See man ld(1) for more information.
10433 @item -arch_errors_fatal
10434 @opindex arch_errors_fatal
10435 Cause the errors having to do with files that have the wrong architecture
10438 @item -bind_at_load
10439 @opindex bind_at_load
10440 Causes the output file to be marked such that the dynamic linker will
10441 bind all undefined references when the file is loaded or launched.
10445 Produce a Mach-o bundle format file.
10446 See man ld(1) for more information.
10448 @item -bundle_loader @var{executable}
10449 @opindex bundle_loader
10450 This option specifies the @var{executable} that will be loading the build
10451 output file being linked. See man ld(1) for more information.
10454 @opindex dynamiclib
10455 When passed this option, GCC will produce a dynamic library instead of
10456 an executable when linking, using the Darwin @file{libtool} command.
10458 @item -force_cpusubtype_ALL
10459 @opindex force_cpusubtype_ALL
10460 This causes GCC's output file to have the @var{ALL} subtype, instead of
10461 one controlled by the @option{-mcpu} or @option{-march} option.
10463 @item -allowable_client @var{client_name}
10464 @itemx -client_name
10465 @itemx -compatibility_version
10466 @itemx -current_version
10468 @itemx -dependency-file
10470 @itemx -dylinker_install_name
10472 @itemx -exported_symbols_list
10474 @itemx -flat_namespace
10475 @itemx -force_flat_namespace
10476 @itemx -headerpad_max_install_names
10479 @itemx -install_name
10480 @itemx -keep_private_externs
10481 @itemx -multi_module
10482 @itemx -multiply_defined
10483 @itemx -multiply_defined_unused
10485 @itemx -no_dead_strip_inits_and_terms
10486 @itemx -nofixprebinding
10487 @itemx -nomultidefs
10489 @itemx -noseglinkedit
10490 @itemx -pagezero_size
10492 @itemx -prebind_all_twolevel_modules
10493 @itemx -private_bundle
10494 @itemx -read_only_relocs
10496 @itemx -sectobjectsymbols
10500 @itemx -sectobjectsymbols
10503 @itemx -segs_read_only_addr
10504 @itemx -segs_read_write_addr
10505 @itemx -seg_addr_table
10506 @itemx -seg_addr_table_filename
10507 @itemx -seglinkedit
10509 @itemx -segs_read_only_addr
10510 @itemx -segs_read_write_addr
10511 @itemx -single_module
10513 @itemx -sub_library
10514 @itemx -sub_umbrella
10515 @itemx -twolevel_namespace
10518 @itemx -unexported_symbols_list
10519 @itemx -weak_reference_mismatches
10520 @itemx -whatsloaded
10521 @opindex allowable_client
10522 @opindex client_name
10523 @opindex compatibility_version
10524 @opindex current_version
10525 @opindex dead_strip
10526 @opindex dependency-file
10527 @opindex dylib_file
10528 @opindex dylinker_install_name
10530 @opindex exported_symbols_list
10532 @opindex flat_namespace
10533 @opindex force_flat_namespace
10534 @opindex headerpad_max_install_names
10535 @opindex image_base
10537 @opindex install_name
10538 @opindex keep_private_externs
10539 @opindex multi_module
10540 @opindex multiply_defined
10541 @opindex multiply_defined_unused
10542 @opindex noall_load
10543 @opindex no_dead_strip_inits_and_terms
10544 @opindex nofixprebinding
10545 @opindex nomultidefs
10547 @opindex noseglinkedit
10548 @opindex pagezero_size
10550 @opindex prebind_all_twolevel_modules
10551 @opindex private_bundle
10552 @opindex read_only_relocs
10554 @opindex sectobjectsymbols
10557 @opindex sectcreate
10558 @opindex sectobjectsymbols
10561 @opindex segs_read_only_addr
10562 @opindex segs_read_write_addr
10563 @opindex seg_addr_table
10564 @opindex seg_addr_table_filename
10565 @opindex seglinkedit
10567 @opindex segs_read_only_addr
10568 @opindex segs_read_write_addr
10569 @opindex single_module
10571 @opindex sub_library
10572 @opindex sub_umbrella
10573 @opindex twolevel_namespace
10576 @opindex unexported_symbols_list
10577 @opindex weak_reference_mismatches
10578 @opindex whatsloaded
10579 These options are passed to the Darwin linker. The Darwin linker man page
10580 describes them in detail.
10583 @node DEC Alpha Options
10584 @subsection DEC Alpha Options
10586 These @samp{-m} options are defined for the DEC Alpha implementations:
10589 @item -mno-soft-float
10590 @itemx -msoft-float
10591 @opindex mno-soft-float
10592 @opindex msoft-float
10593 Use (do not use) the hardware floating-point instructions for
10594 floating-point operations. When @option{-msoft-float} is specified,
10595 functions in @file{libgcc.a} will be used to perform floating-point
10596 operations. Unless they are replaced by routines that emulate the
10597 floating-point operations, or compiled in such a way as to call such
10598 emulations routines, these routines will issue floating-point
10599 operations. If you are compiling for an Alpha without floating-point
10600 operations, you must ensure that the library is built so as not to call
10603 Note that Alpha implementations without floating-point operations are
10604 required to have floating-point registers.
10607 @itemx -mno-fp-regs
10609 @opindex mno-fp-regs
10610 Generate code that uses (does not use) the floating-point register set.
10611 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10612 register set is not used, floating point operands are passed in integer
10613 registers as if they were integers and floating-point results are passed
10614 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10615 so any function with a floating-point argument or return value called by code
10616 compiled with @option{-mno-fp-regs} must also be compiled with that
10619 A typical use of this option is building a kernel that does not use,
10620 and hence need not save and restore, any floating-point registers.
10624 The Alpha architecture implements floating-point hardware optimized for
10625 maximum performance. It is mostly compliant with the IEEE floating
10626 point standard. However, for full compliance, software assistance is
10627 required. This option generates code fully IEEE compliant code
10628 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10629 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10630 defined during compilation. The resulting code is less efficient but is
10631 able to correctly support denormalized numbers and exceptional IEEE
10632 values such as not-a-number and plus/minus infinity. Other Alpha
10633 compilers call this option @option{-ieee_with_no_inexact}.
10635 @item -mieee-with-inexact
10636 @opindex mieee-with-inexact
10637 This is like @option{-mieee} except the generated code also maintains
10638 the IEEE @var{inexact-flag}. Turning on this option causes the
10639 generated code to implement fully-compliant IEEE math. In addition to
10640 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10641 macro. On some Alpha implementations the resulting code may execute
10642 significantly slower than the code generated by default. Since there is
10643 very little code that depends on the @var{inexact-flag}, you should
10644 normally not specify this option. Other Alpha compilers call this
10645 option @option{-ieee_with_inexact}.
10647 @item -mfp-trap-mode=@var{trap-mode}
10648 @opindex mfp-trap-mode
10649 This option controls what floating-point related traps are enabled.
10650 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10651 The trap mode can be set to one of four values:
10655 This is the default (normal) setting. The only traps that are enabled
10656 are the ones that cannot be disabled in software (e.g., division by zero
10660 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10664 Like @samp{u}, but the instructions are marked to be safe for software
10665 completion (see Alpha architecture manual for details).
10668 Like @samp{su}, but inexact traps are enabled as well.
10671 @item -mfp-rounding-mode=@var{rounding-mode}
10672 @opindex mfp-rounding-mode
10673 Selects the IEEE rounding mode. Other Alpha compilers call this option
10674 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10679 Normal IEEE rounding mode. Floating point numbers are rounded towards
10680 the nearest machine number or towards the even machine number in case
10684 Round towards minus infinity.
10687 Chopped rounding mode. Floating point numbers are rounded towards zero.
10690 Dynamic rounding mode. A field in the floating point control register
10691 (@var{fpcr}, see Alpha architecture reference manual) controls the
10692 rounding mode in effect. The C library initializes this register for
10693 rounding towards plus infinity. Thus, unless your program modifies the
10694 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10697 @item -mtrap-precision=@var{trap-precision}
10698 @opindex mtrap-precision
10699 In the Alpha architecture, floating point traps are imprecise. This
10700 means without software assistance it is impossible to recover from a
10701 floating trap and program execution normally needs to be terminated.
10702 GCC can generate code that can assist operating system trap handlers
10703 in determining the exact location that caused a floating point trap.
10704 Depending on the requirements of an application, different levels of
10705 precisions can be selected:
10709 Program precision. This option is the default and means a trap handler
10710 can only identify which program caused a floating point exception.
10713 Function precision. The trap handler can determine the function that
10714 caused a floating point exception.
10717 Instruction precision. The trap handler can determine the exact
10718 instruction that caused a floating point exception.
10721 Other Alpha compilers provide the equivalent options called
10722 @option{-scope_safe} and @option{-resumption_safe}.
10724 @item -mieee-conformant
10725 @opindex mieee-conformant
10726 This option marks the generated code as IEEE conformant. You must not
10727 use this option unless you also specify @option{-mtrap-precision=i} and either
10728 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10729 is to emit the line @samp{.eflag 48} in the function prologue of the
10730 generated assembly file. Under DEC Unix, this has the effect that
10731 IEEE-conformant math library routines will be linked in.
10733 @item -mbuild-constants
10734 @opindex mbuild-constants
10735 Normally GCC examines a 32- or 64-bit integer constant to
10736 see if it can construct it from smaller constants in two or three
10737 instructions. If it cannot, it will output the constant as a literal and
10738 generate code to load it from the data segment at runtime.
10740 Use this option to require GCC to construct @emph{all} integer constants
10741 using code, even if it takes more instructions (the maximum is six).
10743 You would typically use this option to build a shared library dynamic
10744 loader. Itself a shared library, it must relocate itself in memory
10745 before it can find the variables and constants in its own data segment.
10751 Select whether to generate code to be assembled by the vendor-supplied
10752 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10770 Indicate whether GCC should generate code to use the optional BWX,
10771 CIX, FIX and MAX instruction sets. The default is to use the instruction
10772 sets supported by the CPU type specified via @option{-mcpu=} option or that
10773 of the CPU on which GCC was built if none was specified.
10776 @itemx -mfloat-ieee
10777 @opindex mfloat-vax
10778 @opindex mfloat-ieee
10779 Generate code that uses (does not use) VAX F and G floating point
10780 arithmetic instead of IEEE single and double precision.
10782 @item -mexplicit-relocs
10783 @itemx -mno-explicit-relocs
10784 @opindex mexplicit-relocs
10785 @opindex mno-explicit-relocs
10786 Older Alpha assemblers provided no way to generate symbol relocations
10787 except via assembler macros. Use of these macros does not allow
10788 optimal instruction scheduling. GNU binutils as of version 2.12
10789 supports a new syntax that allows the compiler to explicitly mark
10790 which relocations should apply to which instructions. This option
10791 is mostly useful for debugging, as GCC detects the capabilities of
10792 the assembler when it is built and sets the default accordingly.
10795 @itemx -mlarge-data
10796 @opindex msmall-data
10797 @opindex mlarge-data
10798 When @option{-mexplicit-relocs} is in effect, static data is
10799 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10800 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10801 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10802 16-bit relocations off of the @code{$gp} register. This limits the
10803 size of the small data area to 64KB, but allows the variables to be
10804 directly accessed via a single instruction.
10806 The default is @option{-mlarge-data}. With this option the data area
10807 is limited to just below 2GB@. Programs that require more than 2GB of
10808 data must use @code{malloc} or @code{mmap} to allocate the data in the
10809 heap instead of in the program's data segment.
10811 When generating code for shared libraries, @option{-fpic} implies
10812 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10815 @itemx -mlarge-text
10816 @opindex msmall-text
10817 @opindex mlarge-text
10818 When @option{-msmall-text} is used, the compiler assumes that the
10819 code of the entire program (or shared library) fits in 4MB, and is
10820 thus reachable with a branch instruction. When @option{-msmall-data}
10821 is used, the compiler can assume that all local symbols share the
10822 same @code{$gp} value, and thus reduce the number of instructions
10823 required for a function call from 4 to 1.
10825 The default is @option{-mlarge-text}.
10827 @item -mcpu=@var{cpu_type}
10829 Set the instruction set and instruction scheduling parameters for
10830 machine type @var{cpu_type}. You can specify either the @samp{EV}
10831 style name or the corresponding chip number. GCC supports scheduling
10832 parameters for the EV4, EV5 and EV6 family of processors and will
10833 choose the default values for the instruction set from the processor
10834 you specify. If you do not specify a processor type, GCC will default
10835 to the processor on which the compiler was built.
10837 Supported values for @var{cpu_type} are
10843 Schedules as an EV4 and has no instruction set extensions.
10847 Schedules as an EV5 and has no instruction set extensions.
10851 Schedules as an EV5 and supports the BWX extension.
10856 Schedules as an EV5 and supports the BWX and MAX extensions.
10860 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10864 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10867 Native Linux/GNU toolchains also support the value @samp{native},
10868 which selects the best architecture option for the host processor.
10869 @option{-mcpu=native} has no effect if GCC does not recognize
10872 @item -mtune=@var{cpu_type}
10874 Set only the instruction scheduling parameters for machine type
10875 @var{cpu_type}. The instruction set is not changed.
10877 Native Linux/GNU toolchains also support the value @samp{native},
10878 which selects the best architecture option for the host processor.
10879 @option{-mtune=native} has no effect if GCC does not recognize
10882 @item -mmemory-latency=@var{time}
10883 @opindex mmemory-latency
10884 Sets the latency the scheduler should assume for typical memory
10885 references as seen by the application. This number is highly
10886 dependent on the memory access patterns used by the application
10887 and the size of the external cache on the machine.
10889 Valid options for @var{time} are
10893 A decimal number representing clock cycles.
10899 The compiler contains estimates of the number of clock cycles for
10900 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10901 (also called Dcache, Scache, and Bcache), as well as to main memory.
10902 Note that L3 is only valid for EV5.
10907 @node DEC Alpha/VMS Options
10908 @subsection DEC Alpha/VMS Options
10910 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10913 @item -mvms-return-codes
10914 @opindex mvms-return-codes
10915 Return VMS condition codes from main. The default is to return POSIX
10916 style condition (e.g.@: error) codes.
10918 @item -mdebug-main=@var{prefix}
10919 @opindex mdebug-main=@var{prefix}
10920 Flag the first routine whose name starts with @var{prefix} as the main
10921 routine for the debugger.
10925 Default to 64bit memory allocation routines.
10929 @subsection FR30 Options
10930 @cindex FR30 Options
10932 These options are defined specifically for the FR30 port.
10936 @item -msmall-model
10937 @opindex msmall-model
10938 Use the small address space model. This can produce smaller code, but
10939 it does assume that all symbolic values and addresses will fit into a
10944 Assume that run-time support has been provided and so there is no need
10945 to include the simulator library (@file{libsim.a}) on the linker
10951 @subsection FRV Options
10952 @cindex FRV Options
10958 Only use the first 32 general purpose registers.
10963 Use all 64 general purpose registers.
10968 Use only the first 32 floating point registers.
10973 Use all 64 floating point registers
10976 @opindex mhard-float
10978 Use hardware instructions for floating point operations.
10981 @opindex msoft-float
10983 Use library routines for floating point operations.
10988 Dynamically allocate condition code registers.
10993 Do not try to dynamically allocate condition code registers, only
10994 use @code{icc0} and @code{fcc0}.
10999 Change ABI to use double word insns.
11004 Do not use double word instructions.
11009 Use floating point double instructions.
11012 @opindex mno-double
11014 Do not use floating point double instructions.
11019 Use media instructions.
11024 Do not use media instructions.
11029 Use multiply and add/subtract instructions.
11032 @opindex mno-muladd
11034 Do not use multiply and add/subtract instructions.
11039 Select the FDPIC ABI, that uses function descriptors to represent
11040 pointers to functions. Without any PIC/PIE-related options, it
11041 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11042 assumes GOT entries and small data are within a 12-bit range from the
11043 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11044 are computed with 32 bits.
11045 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11048 @opindex minline-plt
11050 Enable inlining of PLT entries in function calls to functions that are
11051 not known to bind locally. It has no effect without @option{-mfdpic}.
11052 It's enabled by default if optimizing for speed and compiling for
11053 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11054 optimization option such as @option{-O3} or above is present in the
11060 Assume a large TLS segment when generating thread-local code.
11065 Do not assume a large TLS segment when generating thread-local code.
11070 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11071 that is known to be in read-only sections. It's enabled by default,
11072 except for @option{-fpic} or @option{-fpie}: even though it may help
11073 make the global offset table smaller, it trades 1 instruction for 4.
11074 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11075 one of which may be shared by multiple symbols, and it avoids the need
11076 for a GOT entry for the referenced symbol, so it's more likely to be a
11077 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11079 @item -multilib-library-pic
11080 @opindex multilib-library-pic
11082 Link with the (library, not FD) pic libraries. It's implied by
11083 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11084 @option{-fpic} without @option{-mfdpic}. You should never have to use
11088 @opindex mlinked-fp
11090 Follow the EABI requirement of always creating a frame pointer whenever
11091 a stack frame is allocated. This option is enabled by default and can
11092 be disabled with @option{-mno-linked-fp}.
11095 @opindex mlong-calls
11097 Use indirect addressing to call functions outside the current
11098 compilation unit. This allows the functions to be placed anywhere
11099 within the 32-bit address space.
11101 @item -malign-labels
11102 @opindex malign-labels
11104 Try to align labels to an 8-byte boundary by inserting nops into the
11105 previous packet. This option only has an effect when VLIW packing
11106 is enabled. It doesn't create new packets; it merely adds nops to
11109 @item -mlibrary-pic
11110 @opindex mlibrary-pic
11112 Generate position-independent EABI code.
11117 Use only the first four media accumulator registers.
11122 Use all eight media accumulator registers.
11127 Pack VLIW instructions.
11132 Do not pack VLIW instructions.
11135 @opindex mno-eflags
11137 Do not mark ABI switches in e_flags.
11140 @opindex mcond-move
11142 Enable the use of conditional-move instructions (default).
11144 This switch is mainly for debugging the compiler and will likely be removed
11145 in a future version.
11147 @item -mno-cond-move
11148 @opindex mno-cond-move
11150 Disable the use of conditional-move instructions.
11152 This switch is mainly for debugging the compiler and will likely be removed
11153 in a future version.
11158 Enable the use of conditional set instructions (default).
11160 This switch is mainly for debugging the compiler and will likely be removed
11161 in a future version.
11166 Disable the use of conditional set instructions.
11168 This switch is mainly for debugging the compiler and will likely be removed
11169 in a future version.
11172 @opindex mcond-exec
11174 Enable the use of conditional execution (default).
11176 This switch is mainly for debugging the compiler and will likely be removed
11177 in a future version.
11179 @item -mno-cond-exec
11180 @opindex mno-cond-exec
11182 Disable the use of conditional execution.
11184 This switch is mainly for debugging the compiler and will likely be removed
11185 in a future version.
11187 @item -mvliw-branch
11188 @opindex mvliw-branch
11190 Run a pass to pack branches into VLIW instructions (default).
11192 This switch is mainly for debugging the compiler and will likely be removed
11193 in a future version.
11195 @item -mno-vliw-branch
11196 @opindex mno-vliw-branch
11198 Do not run a pass to pack branches into VLIW instructions.
11200 This switch is mainly for debugging the compiler and will likely be removed
11201 in a future version.
11203 @item -mmulti-cond-exec
11204 @opindex mmulti-cond-exec
11206 Enable optimization of @code{&&} and @code{||} in conditional execution
11209 This switch is mainly for debugging the compiler and will likely be removed
11210 in a future version.
11212 @item -mno-multi-cond-exec
11213 @opindex mno-multi-cond-exec
11215 Disable optimization of @code{&&} and @code{||} in conditional execution.
11217 This switch is mainly for debugging the compiler and will likely be removed
11218 in a future version.
11220 @item -mnested-cond-exec
11221 @opindex mnested-cond-exec
11223 Enable nested conditional execution optimizations (default).
11225 This switch is mainly for debugging the compiler and will likely be removed
11226 in a future version.
11228 @item -mno-nested-cond-exec
11229 @opindex mno-nested-cond-exec
11231 Disable nested conditional execution optimizations.
11233 This switch is mainly for debugging the compiler and will likely be removed
11234 in a future version.
11236 @item -moptimize-membar
11237 @opindex moptimize-membar
11239 This switch removes redundant @code{membar} instructions from the
11240 compiler generated code. It is enabled by default.
11242 @item -mno-optimize-membar
11243 @opindex mno-optimize-membar
11245 This switch disables the automatic removal of redundant @code{membar}
11246 instructions from the generated code.
11248 @item -mtomcat-stats
11249 @opindex mtomcat-stats
11251 Cause gas to print out tomcat statistics.
11253 @item -mcpu=@var{cpu}
11256 Select the processor type for which to generate code. Possible values are
11257 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11258 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11262 @node GNU/Linux Options
11263 @subsection GNU/Linux Options
11265 These @samp{-m} options are defined for GNU/Linux targets:
11270 Use the GNU C library instead of uClibc. This is the default except
11271 on @samp{*-*-linux-*uclibc*} targets.
11275 Use uClibc instead of the GNU C library. This is the default on
11276 @samp{*-*-linux-*uclibc*} targets.
11279 @node H8/300 Options
11280 @subsection H8/300 Options
11282 These @samp{-m} options are defined for the H8/300 implementations:
11287 Shorten some address references at link time, when possible; uses the
11288 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11289 ld, Using ld}, for a fuller description.
11293 Generate code for the H8/300H@.
11297 Generate code for the H8S@.
11301 Generate code for the H8S and H8/300H in the normal mode. This switch
11302 must be used either with @option{-mh} or @option{-ms}.
11306 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11310 Make @code{int} data 32 bits by default.
11313 @opindex malign-300
11314 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11315 The default for the H8/300H and H8S is to align longs and floats on 4
11317 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11318 This option has no effect on the H8/300.
11322 @subsection HPPA Options
11323 @cindex HPPA Options
11325 These @samp{-m} options are defined for the HPPA family of computers:
11328 @item -march=@var{architecture-type}
11330 Generate code for the specified architecture. The choices for
11331 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11332 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11333 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11334 architecture option for your machine. Code compiled for lower numbered
11335 architectures will run on higher numbered architectures, but not the
11338 @item -mpa-risc-1-0
11339 @itemx -mpa-risc-1-1
11340 @itemx -mpa-risc-2-0
11341 @opindex mpa-risc-1-0
11342 @opindex mpa-risc-1-1
11343 @opindex mpa-risc-2-0
11344 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11347 @opindex mbig-switch
11348 Generate code suitable for big switch tables. Use this option only if
11349 the assembler/linker complain about out of range branches within a switch
11352 @item -mjump-in-delay
11353 @opindex mjump-in-delay
11354 Fill delay slots of function calls with unconditional jump instructions
11355 by modifying the return pointer for the function call to be the target
11356 of the conditional jump.
11358 @item -mdisable-fpregs
11359 @opindex mdisable-fpregs
11360 Prevent floating point registers from being used in any manner. This is
11361 necessary for compiling kernels which perform lazy context switching of
11362 floating point registers. If you use this option and attempt to perform
11363 floating point operations, the compiler will abort.
11365 @item -mdisable-indexing
11366 @opindex mdisable-indexing
11367 Prevent the compiler from using indexing address modes. This avoids some
11368 rather obscure problems when compiling MIG generated code under MACH@.
11370 @item -mno-space-regs
11371 @opindex mno-space-regs
11372 Generate code that assumes the target has no space registers. This allows
11373 GCC to generate faster indirect calls and use unscaled index address modes.
11375 Such code is suitable for level 0 PA systems and kernels.
11377 @item -mfast-indirect-calls
11378 @opindex mfast-indirect-calls
11379 Generate code that assumes calls never cross space boundaries. This
11380 allows GCC to emit code which performs faster indirect calls.
11382 This option will not work in the presence of shared libraries or nested
11385 @item -mfixed-range=@var{register-range}
11386 @opindex mfixed-range
11387 Generate code treating the given register range as fixed registers.
11388 A fixed register is one that the register allocator can not use. This is
11389 useful when compiling kernel code. A register range is specified as
11390 two registers separated by a dash. Multiple register ranges can be
11391 specified separated by a comma.
11393 @item -mlong-load-store
11394 @opindex mlong-load-store
11395 Generate 3-instruction load and store sequences as sometimes required by
11396 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11399 @item -mportable-runtime
11400 @opindex mportable-runtime
11401 Use the portable calling conventions proposed by HP for ELF systems.
11405 Enable the use of assembler directives only GAS understands.
11407 @item -mschedule=@var{cpu-type}
11409 Schedule code according to the constraints for the machine type
11410 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11411 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11412 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11413 proper scheduling option for your machine. The default scheduling is
11417 @opindex mlinker-opt
11418 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11419 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11420 linkers in which they give bogus error messages when linking some programs.
11423 @opindex msoft-float
11424 Generate output containing library calls for floating point.
11425 @strong{Warning:} the requisite libraries are not available for all HPPA
11426 targets. Normally the facilities of the machine's usual C compiler are
11427 used, but this cannot be done directly in cross-compilation. You must make
11428 your own arrangements to provide suitable library functions for
11431 @option{-msoft-float} changes the calling convention in the output file;
11432 therefore, it is only useful if you compile @emph{all} of a program with
11433 this option. In particular, you need to compile @file{libgcc.a}, the
11434 library that comes with GCC, with @option{-msoft-float} in order for
11439 Generate the predefine, @code{_SIO}, for server IO@. The default is
11440 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11441 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11442 options are available under HP-UX and HI-UX@.
11446 Use GNU ld specific options. This passes @option{-shared} to ld when
11447 building a shared library. It is the default when GCC is configured,
11448 explicitly or implicitly, with the GNU linker. This option does not
11449 have any affect on which ld is called, it only changes what parameters
11450 are passed to that ld. The ld that is called is determined by the
11451 @option{--with-ld} configure option, GCC's program search path, and
11452 finally by the user's @env{PATH}. The linker used by GCC can be printed
11453 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11454 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11458 Use HP ld specific options. This passes @option{-b} to ld when building
11459 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11460 links. It is the default when GCC is configured, explicitly or
11461 implicitly, with the HP linker. This option does not have any affect on
11462 which ld is called, it only changes what parameters are passed to that
11463 ld. The ld that is called is determined by the @option{--with-ld}
11464 configure option, GCC's program search path, and finally by the user's
11465 @env{PATH}. The linker used by GCC can be printed using @samp{which
11466 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11467 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11470 @opindex mno-long-calls
11471 Generate code that uses long call sequences. This ensures that a call
11472 is always able to reach linker generated stubs. The default is to generate
11473 long calls only when the distance from the call site to the beginning
11474 of the function or translation unit, as the case may be, exceeds a
11475 predefined limit set by the branch type being used. The limits for
11476 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11477 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11480 Distances are measured from the beginning of functions when using the
11481 @option{-ffunction-sections} option, or when using the @option{-mgas}
11482 and @option{-mno-portable-runtime} options together under HP-UX with
11485 It is normally not desirable to use this option as it will degrade
11486 performance. However, it may be useful in large applications,
11487 particularly when partial linking is used to build the application.
11489 The types of long calls used depends on the capabilities of the
11490 assembler and linker, and the type of code being generated. The
11491 impact on systems that support long absolute calls, and long pic
11492 symbol-difference or pc-relative calls should be relatively small.
11493 However, an indirect call is used on 32-bit ELF systems in pic code
11494 and it is quite long.
11496 @item -munix=@var{unix-std}
11498 Generate compiler predefines and select a startfile for the specified
11499 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11500 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11501 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11502 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11503 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11506 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11507 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11508 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11509 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11510 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11511 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11513 It is @emph{important} to note that this option changes the interfaces
11514 for various library routines. It also affects the operational behavior
11515 of the C library. Thus, @emph{extreme} care is needed in using this
11518 Library code that is intended to operate with more than one UNIX
11519 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11520 as appropriate. Most GNU software doesn't provide this capability.
11524 Suppress the generation of link options to search libdld.sl when the
11525 @option{-static} option is specified on HP-UX 10 and later.
11529 The HP-UX implementation of setlocale in libc has a dependency on
11530 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11531 when the @option{-static} option is specified, special link options
11532 are needed to resolve this dependency.
11534 On HP-UX 10 and later, the GCC driver adds the necessary options to
11535 link with libdld.sl when the @option{-static} option is specified.
11536 This causes the resulting binary to be dynamic. On the 64-bit port,
11537 the linkers generate dynamic binaries by default in any case. The
11538 @option{-nolibdld} option can be used to prevent the GCC driver from
11539 adding these link options.
11543 Add support for multithreading with the @dfn{dce thread} library
11544 under HP-UX@. This option sets flags for both the preprocessor and
11548 @node i386 and x86-64 Options
11549 @subsection Intel 386 and AMD x86-64 Options
11550 @cindex i386 Options
11551 @cindex x86-64 Options
11552 @cindex Intel 386 Options
11553 @cindex AMD x86-64 Options
11555 These @samp{-m} options are defined for the i386 and x86-64 family of
11559 @item -mtune=@var{cpu-type}
11561 Tune to @var{cpu-type} everything applicable about the generated code, except
11562 for the ABI and the set of available instructions. The choices for
11563 @var{cpu-type} are:
11566 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11567 If you know the CPU on which your code will run, then you should use
11568 the corresponding @option{-mtune} option instead of
11569 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11570 of your application will have, then you should use this option.
11572 As new processors are deployed in the marketplace, the behavior of this
11573 option will change. Therefore, if you upgrade to a newer version of
11574 GCC, the code generated option will change to reflect the processors
11575 that were most common when that version of GCC was released.
11577 There is no @option{-march=generic} option because @option{-march}
11578 indicates the instruction set the compiler can use, and there is no
11579 generic instruction set applicable to all processors. In contrast,
11580 @option{-mtune} indicates the processor (or, in this case, collection of
11581 processors) for which the code is optimized.
11583 This selects the CPU to tune for at compilation time by determining
11584 the processor type of the compiling machine. Using @option{-mtune=native}
11585 will produce code optimized for the local machine under the constraints
11586 of the selected instruction set. Using @option{-march=native} will
11587 enable all instruction subsets supported by the local machine (hence
11588 the result might not run on different machines).
11590 Original Intel's i386 CPU@.
11592 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11593 @item i586, pentium
11594 Intel Pentium CPU with no MMX support.
11596 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11598 Intel PentiumPro CPU@.
11600 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11601 instruction set will be used, so the code will run on all i686 family chips.
11603 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11604 @item pentium3, pentium3m
11605 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11608 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11609 support. Used by Centrino notebooks.
11610 @item pentium4, pentium4m
11611 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11613 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11616 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11617 SSE2 and SSE3 instruction set support.
11619 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11620 instruction set support.
11622 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11623 instruction set support.
11625 AMD K6 CPU with MMX instruction set support.
11627 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11628 @item athlon, athlon-tbird
11629 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11631 @item athlon-4, athlon-xp, athlon-mp
11632 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11633 instruction set support.
11634 @item k8, opteron, athlon64, athlon-fx
11635 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11636 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11637 @item k8-sse3, opteron-sse3, athlon64-sse3
11638 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11639 @item amdfam10, barcelona
11640 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11641 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11642 instruction set extensions.)
11644 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11647 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11648 instruction set support.
11650 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11651 implemented for this chip.)
11653 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11654 implemented for this chip.)
11656 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11659 While picking a specific @var{cpu-type} will schedule things appropriately
11660 for that particular chip, the compiler will not generate any code that
11661 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11664 @item -march=@var{cpu-type}
11666 Generate instructions for the machine type @var{cpu-type}. The choices
11667 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11668 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11670 @item -mcpu=@var{cpu-type}
11672 A deprecated synonym for @option{-mtune}.
11674 @item -mfpmath=@var{unit}
11676 Generate floating point arithmetics for selected unit @var{unit}. The choices
11677 for @var{unit} are:
11681 Use the standard 387 floating point coprocessor present majority of chips and
11682 emulated otherwise. Code compiled with this option will run almost everywhere.
11683 The temporary results are computed in 80bit precision instead of precision
11684 specified by the type resulting in slightly different results compared to most
11685 of other chips. See @option{-ffloat-store} for more detailed description.
11687 This is the default choice for i386 compiler.
11690 Use scalar floating point instructions present in the SSE instruction set.
11691 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11692 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11693 instruction set supports only single precision arithmetics, thus the double and
11694 extended precision arithmetics is still done using 387. Later version, present
11695 only in Pentium4 and the future AMD x86-64 chips supports double precision
11698 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11699 or @option{-msse2} switches to enable SSE extensions and make this option
11700 effective. For the x86-64 compiler, these extensions are enabled by default.
11702 The resulting code should be considerably faster in the majority of cases and avoid
11703 the numerical instability problems of 387 code, but may break some existing
11704 code that expects temporaries to be 80bit.
11706 This is the default choice for the x86-64 compiler.
11711 Attempt to utilize both instruction sets at once. This effectively double the
11712 amount of available registers and on chips with separate execution units for
11713 387 and SSE the execution resources too. Use this option with care, as it is
11714 still experimental, because the GCC register allocator does not model separate
11715 functional units well resulting in instable performance.
11718 @item -masm=@var{dialect}
11719 @opindex masm=@var{dialect}
11720 Output asm instructions using selected @var{dialect}. Supported
11721 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11722 not support @samp{intel}.
11725 @itemx -mno-ieee-fp
11727 @opindex mno-ieee-fp
11728 Control whether or not the compiler uses IEEE floating point
11729 comparisons. These handle correctly the case where the result of a
11730 comparison is unordered.
11733 @opindex msoft-float
11734 Generate output containing library calls for floating point.
11735 @strong{Warning:} the requisite libraries are not part of GCC@.
11736 Normally the facilities of the machine's usual C compiler are used, but
11737 this can't be done directly in cross-compilation. You must make your
11738 own arrangements to provide suitable library functions for
11741 On machines where a function returns floating point results in the 80387
11742 register stack, some floating point opcodes may be emitted even if
11743 @option{-msoft-float} is used.
11745 @item -mno-fp-ret-in-387
11746 @opindex mno-fp-ret-in-387
11747 Do not use the FPU registers for return values of functions.
11749 The usual calling convention has functions return values of types
11750 @code{float} and @code{double} in an FPU register, even if there
11751 is no FPU@. The idea is that the operating system should emulate
11754 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11755 in ordinary CPU registers instead.
11757 @item -mno-fancy-math-387
11758 @opindex mno-fancy-math-387
11759 Some 387 emulators do not support the @code{sin}, @code{cos} and
11760 @code{sqrt} instructions for the 387. Specify this option to avoid
11761 generating those instructions. This option is the default on FreeBSD,
11762 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11763 indicates that the target cpu will always have an FPU and so the
11764 instruction will not need emulation. As of revision 2.6.1, these
11765 instructions are not generated unless you also use the
11766 @option{-funsafe-math-optimizations} switch.
11768 @item -malign-double
11769 @itemx -mno-align-double
11770 @opindex malign-double
11771 @opindex mno-align-double
11772 Control whether GCC aligns @code{double}, @code{long double}, and
11773 @code{long long} variables on a two word boundary or a one word
11774 boundary. Aligning @code{double} variables on a two word boundary will
11775 produce code that runs somewhat faster on a @samp{Pentium} at the
11776 expense of more memory.
11778 On x86-64, @option{-malign-double} is enabled by default.
11780 @strong{Warning:} if you use the @option{-malign-double} switch,
11781 structures containing the above types will be aligned differently than
11782 the published application binary interface specifications for the 386
11783 and will not be binary compatible with structures in code compiled
11784 without that switch.
11786 @item -m96bit-long-double
11787 @itemx -m128bit-long-double
11788 @opindex m96bit-long-double
11789 @opindex m128bit-long-double
11790 These switches control the size of @code{long double} type. The i386
11791 application binary interface specifies the size to be 96 bits,
11792 so @option{-m96bit-long-double} is the default in 32 bit mode.
11794 Modern architectures (Pentium and newer) would prefer @code{long double}
11795 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11796 conforming to the ABI, this would not be possible. So specifying a
11797 @option{-m128bit-long-double} will align @code{long double}
11798 to a 16 byte boundary by padding the @code{long double} with an additional
11801 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11802 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11804 Notice that neither of these options enable any extra precision over the x87
11805 standard of 80 bits for a @code{long double}.
11807 @strong{Warning:} if you override the default value for your target ABI, the
11808 structures and arrays containing @code{long double} variables will change
11809 their size as well as function calling convention for function taking
11810 @code{long double} will be modified. Hence they will not be binary
11811 compatible with arrays or structures in code compiled without that switch.
11813 @item -mlarge-data-threshold=@var{number}
11814 @opindex mlarge-data-threshold=@var{number}
11815 When @option{-mcmodel=medium} is specified, the data greater than
11816 @var{threshold} are placed in large data section. This value must be the
11817 same across all object linked into the binary and defaults to 65535.
11821 Use a different function-calling convention, in which functions that
11822 take a fixed number of arguments return with the @code{ret} @var{num}
11823 instruction, which pops their arguments while returning. This saves one
11824 instruction in the caller since there is no need to pop the arguments
11827 You can specify that an individual function is called with this calling
11828 sequence with the function attribute @samp{stdcall}. You can also
11829 override the @option{-mrtd} option by using the function attribute
11830 @samp{cdecl}. @xref{Function Attributes}.
11832 @strong{Warning:} this calling convention is incompatible with the one
11833 normally used on Unix, so you cannot use it if you need to call
11834 libraries compiled with the Unix compiler.
11836 Also, you must provide function prototypes for all functions that
11837 take variable numbers of arguments (including @code{printf});
11838 otherwise incorrect code will be generated for calls to those
11841 In addition, seriously incorrect code will result if you call a
11842 function with too many arguments. (Normally, extra arguments are
11843 harmlessly ignored.)
11845 @item -mregparm=@var{num}
11847 Control how many registers are used to pass integer arguments. By
11848 default, no registers are used to pass arguments, and at most 3
11849 registers can be used. You can control this behavior for a specific
11850 function by using the function attribute @samp{regparm}.
11851 @xref{Function Attributes}.
11853 @strong{Warning:} if you use this switch, and
11854 @var{num} is nonzero, then you must build all modules with the same
11855 value, including any libraries. This includes the system libraries and
11859 @opindex msseregparm
11860 Use SSE register passing conventions for float and double arguments
11861 and return values. You can control this behavior for a specific
11862 function by using the function attribute @samp{sseregparm}.
11863 @xref{Function Attributes}.
11865 @strong{Warning:} if you use this switch then you must build all
11866 modules with the same value, including any libraries. This includes
11867 the system libraries and startup modules.
11876 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11877 is specified, the significands of results of floating-point operations are
11878 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11879 significands of results of floating-point operations to 53 bits (double
11880 precision) and @option{-mpc80} rounds the significands of results of
11881 floating-point operations to 64 bits (extended double precision), which is
11882 the default. When this option is used, floating-point operations in higher
11883 precisions are not available to the programmer without setting the FPU
11884 control word explicitly.
11886 Setting the rounding of floating-point operations to less than the default
11887 80 bits can speed some programs by 2% or more. Note that some mathematical
11888 libraries assume that extended precision (80 bit) floating-point operations
11889 are enabled by default; routines in such libraries could suffer significant
11890 loss of accuracy, typically through so-called "catastrophic cancellation",
11891 when this option is used to set the precision to less than extended precision.
11893 @item -mstackrealign
11894 @opindex mstackrealign
11895 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11896 option will generate an alternate prologue and epilogue that realigns the
11897 runtime stack if necessary. This supports mixing legacy codes that keep
11898 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11899 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11900 applicable to individual functions.
11902 @item -mpreferred-stack-boundary=@var{num}
11903 @opindex mpreferred-stack-boundary
11904 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11905 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11906 the default is 4 (16 bytes or 128 bits).
11908 @item -mincoming-stack-boundary=@var{num}
11909 @opindex mincoming-stack-boundary
11910 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11911 boundary. If @option{-mincoming-stack-boundary} is not specified,
11912 the one specified by @option{-mpreferred-stack-boundary} will be used.
11914 On Pentium and PentiumPro, @code{double} and @code{long double} values
11915 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11916 suffer significant run time performance penalties. On Pentium III, the
11917 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11918 properly if it is not 16 byte aligned.
11920 To ensure proper alignment of this values on the stack, the stack boundary
11921 must be as aligned as that required by any value stored on the stack.
11922 Further, every function must be generated such that it keeps the stack
11923 aligned. Thus calling a function compiled with a higher preferred
11924 stack boundary from a function compiled with a lower preferred stack
11925 boundary will most likely misalign the stack. It is recommended that
11926 libraries that use callbacks always use the default setting.
11928 This extra alignment does consume extra stack space, and generally
11929 increases code size. Code that is sensitive to stack space usage, such
11930 as embedded systems and operating system kernels, may want to reduce the
11931 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11971 These switches enable or disable the use of instructions in the MMX,
11972 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, ABM or
11973 3DNow!@: extended instruction sets.
11974 These extensions are also available as built-in functions: see
11975 @ref{X86 Built-in Functions}, for details of the functions enabled and
11976 disabled by these switches.
11978 To have SSE/SSE2 instructions generated automatically from floating-point
11979 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11981 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11982 generates new AVX instructions or AVX equivalence for all SSEx instructions
11985 These options will enable GCC to use these extended instructions in
11986 generated code, even without @option{-mfpmath=sse}. Applications which
11987 perform runtime CPU detection must compile separate files for each
11988 supported architecture, using the appropriate flags. In particular,
11989 the file containing the CPU detection code should be compiled without
11994 This option instructs GCC to emit a @code{cld} instruction in the prologue
11995 of functions that use string instructions. String instructions depend on
11996 the DF flag to select between autoincrement or autodecrement mode. While the
11997 ABI specifies the DF flag to be cleared on function entry, some operating
11998 systems violate this specification by not clearing the DF flag in their
11999 exception dispatchers. The exception handler can be invoked with the DF flag
12000 set which leads to wrong direction mode, when string instructions are used.
12001 This option can be enabled by default on 32-bit x86 targets by configuring
12002 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12003 instructions can be suppressed with the @option{-mno-cld} compiler option
12008 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12009 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12010 data types. This is useful for high resolution counters that could be updated
12011 by multiple processors (or cores). This instruction is generated as part of
12012 atomic built-in functions: see @ref{Atomic Builtins} for details.
12016 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12017 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12018 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12019 SAHF are load and store instructions, respectively, for certain status flags.
12020 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12021 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12025 This option will enable GCC to use movbe instruction to implement
12026 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12030 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12031 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12032 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12036 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12037 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12038 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12039 variants) for single precision floating point arguments. These instructions
12040 are generated only when @option{-funsafe-math-optimizations} is enabled
12041 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12042 Note that while the throughput of the sequence is higher than the throughput
12043 of the non-reciprocal instruction, the precision of the sequence can be
12044 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12046 @item -mveclibabi=@var{type}
12047 @opindex mveclibabi
12048 Specifies the ABI type to use for vectorizing intrinsics using an
12049 external library. Supported types are @code{svml} for the Intel short
12050 vector math library and @code{acml} for the AMD math core library style
12051 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12052 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12053 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12054 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12055 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12056 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12057 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12058 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12059 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12060 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12061 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12062 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12063 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12064 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12065 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12066 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12067 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12068 compatible library will have to be specified at link time.
12070 @item -mabi=@var{name}
12072 Generate code for the specified calling convention. Permissible values
12073 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12074 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12075 ABI when targeting Windows. On all other systems, the default is the
12076 SYSV ABI. You can control this behavior for a specific function by
12077 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12078 @xref{Function Attributes}.
12081 @itemx -mno-push-args
12082 @opindex mpush-args
12083 @opindex mno-push-args
12084 Use PUSH operations to store outgoing parameters. This method is shorter
12085 and usually equally fast as method using SUB/MOV operations and is enabled
12086 by default. In some cases disabling it may improve performance because of
12087 improved scheduling and reduced dependencies.
12089 @item -maccumulate-outgoing-args
12090 @opindex maccumulate-outgoing-args
12091 If enabled, the maximum amount of space required for outgoing arguments will be
12092 computed in the function prologue. This is faster on most modern CPUs
12093 because of reduced dependencies, improved scheduling and reduced stack usage
12094 when preferred stack boundary is not equal to 2. The drawback is a notable
12095 increase in code size. This switch implies @option{-mno-push-args}.
12099 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12100 on thread-safe exception handling must compile and link all code with the
12101 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12102 @option{-D_MT}; when linking, it links in a special thread helper library
12103 @option{-lmingwthrd} which cleans up per thread exception handling data.
12105 @item -mno-align-stringops
12106 @opindex mno-align-stringops
12107 Do not align destination of inlined string operations. This switch reduces
12108 code size and improves performance in case the destination is already aligned,
12109 but GCC doesn't know about it.
12111 @item -minline-all-stringops
12112 @opindex minline-all-stringops
12113 By default GCC inlines string operations only when destination is known to be
12114 aligned at least to 4 byte boundary. This enables more inlining, increase code
12115 size, but may improve performance of code that depends on fast memcpy, strlen
12116 and memset for short lengths.
12118 @item -minline-stringops-dynamically
12119 @opindex minline-stringops-dynamically
12120 For string operation of unknown size, inline runtime checks so for small
12121 blocks inline code is used, while for large blocks library call is used.
12123 @item -mstringop-strategy=@var{alg}
12124 @opindex mstringop-strategy=@var{alg}
12125 Overwrite internal decision heuristic about particular algorithm to inline
12126 string operation with. The allowed values are @code{rep_byte},
12127 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12128 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12129 expanding inline loop, @code{libcall} for always expanding library call.
12131 @item -momit-leaf-frame-pointer
12132 @opindex momit-leaf-frame-pointer
12133 Don't keep the frame pointer in a register for leaf functions. This
12134 avoids the instructions to save, set up and restore frame pointers and
12135 makes an extra register available in leaf functions. The option
12136 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12137 which might make debugging harder.
12139 @item -mtls-direct-seg-refs
12140 @itemx -mno-tls-direct-seg-refs
12141 @opindex mtls-direct-seg-refs
12142 Controls whether TLS variables may be accessed with offsets from the
12143 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12144 or whether the thread base pointer must be added. Whether or not this
12145 is legal depends on the operating system, and whether it maps the
12146 segment to cover the entire TLS area.
12148 For systems that use GNU libc, the default is on.
12151 @itemx -mno-sse2avx
12153 Specify that the assembler should encode SSE instructions with VEX
12154 prefix. The option @option{-mavx} turns this on by default.
12157 These @samp{-m} switches are supported in addition to the above
12158 on AMD x86-64 processors in 64-bit environments.
12165 Generate code for a 32-bit or 64-bit environment.
12166 The 32-bit environment sets int, long and pointer to 32 bits and
12167 generates code that runs on any i386 system.
12168 The 64-bit environment sets int to 32 bits and long and pointer
12169 to 64 bits and generates code for AMD's x86-64 architecture. For
12170 darwin only the -m64 option turns off the @option{-fno-pic} and
12171 @option{-mdynamic-no-pic} options.
12173 @item -mno-red-zone
12174 @opindex mno-red-zone
12175 Do not use a so called red zone for x86-64 code. The red zone is mandated
12176 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12177 stack pointer that will not be modified by signal or interrupt handlers
12178 and therefore can be used for temporary data without adjusting the stack
12179 pointer. The flag @option{-mno-red-zone} disables this red zone.
12181 @item -mcmodel=small
12182 @opindex mcmodel=small
12183 Generate code for the small code model: the program and its symbols must
12184 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12185 Programs can be statically or dynamically linked. This is the default
12188 @item -mcmodel=kernel
12189 @opindex mcmodel=kernel
12190 Generate code for the kernel code model. The kernel runs in the
12191 negative 2 GB of the address space.
12192 This model has to be used for Linux kernel code.
12194 @item -mcmodel=medium
12195 @opindex mcmodel=medium
12196 Generate code for the medium model: The program is linked in the lower 2
12197 GB of the address space. Small symbols are also placed there. Symbols
12198 with sizes larger than @option{-mlarge-data-threshold} are put into
12199 large data or bss sections and can be located above 2GB. Programs can
12200 be statically or dynamically linked.
12202 @item -mcmodel=large
12203 @opindex mcmodel=large
12204 Generate code for the large model: This model makes no assumptions
12205 about addresses and sizes of sections.
12208 @node IA-64 Options
12209 @subsection IA-64 Options
12210 @cindex IA-64 Options
12212 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12216 @opindex mbig-endian
12217 Generate code for a big endian target. This is the default for HP-UX@.
12219 @item -mlittle-endian
12220 @opindex mlittle-endian
12221 Generate code for a little endian target. This is the default for AIX5
12227 @opindex mno-gnu-as
12228 Generate (or don't) code for the GNU assembler. This is the default.
12229 @c Also, this is the default if the configure option @option{--with-gnu-as}
12235 @opindex mno-gnu-ld
12236 Generate (or don't) code for the GNU linker. This is the default.
12237 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12242 Generate code that does not use a global pointer register. The result
12243 is not position independent code, and violates the IA-64 ABI@.
12245 @item -mvolatile-asm-stop
12246 @itemx -mno-volatile-asm-stop
12247 @opindex mvolatile-asm-stop
12248 @opindex mno-volatile-asm-stop
12249 Generate (or don't) a stop bit immediately before and after volatile asm
12252 @item -mregister-names
12253 @itemx -mno-register-names
12254 @opindex mregister-names
12255 @opindex mno-register-names
12256 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12257 the stacked registers. This may make assembler output more readable.
12263 Disable (or enable) optimizations that use the small data section. This may
12264 be useful for working around optimizer bugs.
12266 @item -mconstant-gp
12267 @opindex mconstant-gp
12268 Generate code that uses a single constant global pointer value. This is
12269 useful when compiling kernel code.
12273 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12274 This is useful when compiling firmware code.
12276 @item -minline-float-divide-min-latency
12277 @opindex minline-float-divide-min-latency
12278 Generate code for inline divides of floating point values
12279 using the minimum latency algorithm.
12281 @item -minline-float-divide-max-throughput
12282 @opindex minline-float-divide-max-throughput
12283 Generate code for inline divides of floating point values
12284 using the maximum throughput algorithm.
12286 @item -mno-inline-float-divide
12287 @opindex mno-inline-float-divide
12288 Do not generate inline code for divides of floating point values.
12290 @item -minline-int-divide-min-latency
12291 @opindex minline-int-divide-min-latency
12292 Generate code for inline divides of integer values
12293 using the minimum latency algorithm.
12295 @item -minline-int-divide-max-throughput
12296 @opindex minline-int-divide-max-throughput
12297 Generate code for inline divides of integer values
12298 using the maximum throughput algorithm.
12300 @item -mno-inline-int-divide
12301 @opindex mno-inline-int-divide
12302 Do not generate inline code for divides of integer values.
12304 @item -minline-sqrt-min-latency
12305 @opindex minline-sqrt-min-latency
12306 Generate code for inline square roots
12307 using the minimum latency algorithm.
12309 @item -minline-sqrt-max-throughput
12310 @opindex minline-sqrt-max-throughput
12311 Generate code for inline square roots
12312 using the maximum throughput algorithm.
12314 @item -mno-inline-sqrt
12315 @opindex mno-inline-sqrt
12316 Do not generate inline code for sqrt.
12319 @itemx -mno-fused-madd
12320 @opindex mfused-madd
12321 @opindex mno-fused-madd
12322 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12323 instructions. The default is to use these instructions.
12325 @item -mno-dwarf2-asm
12326 @itemx -mdwarf2-asm
12327 @opindex mno-dwarf2-asm
12328 @opindex mdwarf2-asm
12329 Don't (or do) generate assembler code for the DWARF2 line number debugging
12330 info. This may be useful when not using the GNU assembler.
12332 @item -mearly-stop-bits
12333 @itemx -mno-early-stop-bits
12334 @opindex mearly-stop-bits
12335 @opindex mno-early-stop-bits
12336 Allow stop bits to be placed earlier than immediately preceding the
12337 instruction that triggered the stop bit. This can improve instruction
12338 scheduling, but does not always do so.
12340 @item -mfixed-range=@var{register-range}
12341 @opindex mfixed-range
12342 Generate code treating the given register range as fixed registers.
12343 A fixed register is one that the register allocator can not use. This is
12344 useful when compiling kernel code. A register range is specified as
12345 two registers separated by a dash. Multiple register ranges can be
12346 specified separated by a comma.
12348 @item -mtls-size=@var{tls-size}
12350 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12353 @item -mtune=@var{cpu-type}
12355 Tune the instruction scheduling for a particular CPU, Valid values are
12356 itanium, itanium1, merced, itanium2, and mckinley.
12362 Generate code for a 32-bit or 64-bit environment.
12363 The 32-bit environment sets int, long and pointer to 32 bits.
12364 The 64-bit environment sets int to 32 bits and long and pointer
12365 to 64 bits. These are HP-UX specific flags.
12367 @item -mno-sched-br-data-spec
12368 @itemx -msched-br-data-spec
12369 @opindex mno-sched-br-data-spec
12370 @opindex msched-br-data-spec
12371 (Dis/En)able data speculative scheduling before reload.
12372 This will result in generation of the ld.a instructions and
12373 the corresponding check instructions (ld.c / chk.a).
12374 The default is 'disable'.
12376 @item -msched-ar-data-spec
12377 @itemx -mno-sched-ar-data-spec
12378 @opindex msched-ar-data-spec
12379 @opindex mno-sched-ar-data-spec
12380 (En/Dis)able data speculative scheduling after reload.
12381 This will result in generation of the ld.a instructions and
12382 the corresponding check instructions (ld.c / chk.a).
12383 The default is 'enable'.
12385 @item -mno-sched-control-spec
12386 @itemx -msched-control-spec
12387 @opindex mno-sched-control-spec
12388 @opindex msched-control-spec
12389 (Dis/En)able control speculative scheduling. This feature is
12390 available only during region scheduling (i.e.@: before reload).
12391 This will result in generation of the ld.s instructions and
12392 the corresponding check instructions chk.s .
12393 The default is 'disable'.
12395 @item -msched-br-in-data-spec
12396 @itemx -mno-sched-br-in-data-spec
12397 @opindex msched-br-in-data-spec
12398 @opindex mno-sched-br-in-data-spec
12399 (En/Dis)able speculative scheduling of the instructions that
12400 are dependent on the data speculative loads before reload.
12401 This is effective only with @option{-msched-br-data-spec} enabled.
12402 The default is 'enable'.
12404 @item -msched-ar-in-data-spec
12405 @itemx -mno-sched-ar-in-data-spec
12406 @opindex msched-ar-in-data-spec
12407 @opindex mno-sched-ar-in-data-spec
12408 (En/Dis)able speculative scheduling of the instructions that
12409 are dependent on the data speculative loads after reload.
12410 This is effective only with @option{-msched-ar-data-spec} enabled.
12411 The default is 'enable'.
12413 @item -msched-in-control-spec
12414 @itemx -mno-sched-in-control-spec
12415 @opindex msched-in-control-spec
12416 @opindex mno-sched-in-control-spec
12417 (En/Dis)able speculative scheduling of the instructions that
12418 are dependent on the control speculative loads.
12419 This is effective only with @option{-msched-control-spec} enabled.
12420 The default is 'enable'.
12422 @item -mno-sched-prefer-non-data-spec-insns
12423 @itemx -msched-prefer-non-data-spec-insns
12424 @opindex mno-sched-prefer-non-data-spec-insns
12425 @opindex msched-prefer-non-data-spec-insns
12426 If enabled, data speculative instructions will be chosen for schedule
12427 only if there are no other choices at the moment. This will make
12428 the use of the data speculation much more conservative.
12429 The default is 'disable'.
12431 @item -mno-sched-prefer-non-control-spec-insns
12432 @itemx -msched-prefer-non-control-spec-insns
12433 @opindex mno-sched-prefer-non-control-spec-insns
12434 @opindex msched-prefer-non-control-spec-insns
12435 If enabled, control speculative instructions will be chosen for schedule
12436 only if there are no other choices at the moment. This will make
12437 the use of the control speculation much more conservative.
12438 The default is 'disable'.
12440 @item -mno-sched-count-spec-in-critical-path
12441 @itemx -msched-count-spec-in-critical-path
12442 @opindex mno-sched-count-spec-in-critical-path
12443 @opindex msched-count-spec-in-critical-path
12444 If enabled, speculative dependencies will be considered during
12445 computation of the instructions priorities. This will make the use of the
12446 speculation a bit more conservative.
12447 The default is 'disable'.
12449 @item -msched-spec-ldc
12450 @opindex msched-spec-ldc
12451 Use a simple data speculation check. This option is on by default.
12453 @item -msched-control-spec-ldc
12454 @opindex msched-spec-ldc
12455 Use a simple check for control speculation. This option is on by default.
12457 @item -msched-stop-bits-after-every-cycle
12458 @opindex msched-stop-bits-after-every-cycle
12459 Place a stop bit after every cycle when scheduling. This option is on
12462 @item -msched-fp-mem-deps-zero-cost
12463 @opindex msched-fp-mem-deps-zero-cost
12464 Assume that floating-point stores and loads are not likely to cause a conflict
12465 when placed into the same instruction group. This option is disabled by
12468 @item -msel-sched-dont-check-control-spec
12469 @opindex msel-sched-dont-check-control-spec
12470 Generate checks for control speculation in selective scheduling.
12471 This flag is disabled by default.
12473 @item -msched-max-memory-insns=@var{max-insns}
12474 @opindex msched-max-memory-insns
12475 Limit on the number of memory insns per instruction group, giving lower
12476 priority to subsequent memory insns attempting to schedule in the same
12477 instruction group. Frequently useful to prevent cache bank conflicts.
12478 The default value is 1.
12480 @item -msched-max-memory-insns-hard-limit
12481 @opindex msched-max-memory-insns-hard-limit
12482 Disallow more than `msched-max-memory-insns' in instruction group.
12483 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12484 when limit is reached but may still schedule memory operations.
12488 @node IA-64/VMS Options
12489 @subsection IA-64/VMS Options
12491 These @samp{-m} options are defined for the IA-64/VMS implementations:
12494 @item -mvms-return-codes
12495 @opindex mvms-return-codes
12496 Return VMS condition codes from main. The default is to return POSIX
12497 style condition (e.g.@ error) codes.
12499 @item -mdebug-main=@var{prefix}
12500 @opindex mdebug-main=@var{prefix}
12501 Flag the first routine whose name starts with @var{prefix} as the main
12502 routine for the debugger.
12506 Default to 64bit memory allocation routines.
12510 @subsection M32C Options
12511 @cindex M32C options
12514 @item -mcpu=@var{name}
12516 Select the CPU for which code is generated. @var{name} may be one of
12517 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12518 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12519 the M32C/80 series.
12523 Specifies that the program will be run on the simulator. This causes
12524 an alternate runtime library to be linked in which supports, for
12525 example, file I/O@. You must not use this option when generating
12526 programs that will run on real hardware; you must provide your own
12527 runtime library for whatever I/O functions are needed.
12529 @item -memregs=@var{number}
12531 Specifies the number of memory-based pseudo-registers GCC will use
12532 during code generation. These pseudo-registers will be used like real
12533 registers, so there is a tradeoff between GCC's ability to fit the
12534 code into available registers, and the performance penalty of using
12535 memory instead of registers. Note that all modules in a program must
12536 be compiled with the same value for this option. Because of that, you
12537 must not use this option with the default runtime libraries gcc
12542 @node M32R/D Options
12543 @subsection M32R/D Options
12544 @cindex M32R/D options
12546 These @option{-m} options are defined for Renesas M32R/D architectures:
12551 Generate code for the M32R/2@.
12555 Generate code for the M32R/X@.
12559 Generate code for the M32R@. This is the default.
12561 @item -mmodel=small
12562 @opindex mmodel=small
12563 Assume all objects live in the lower 16MB of memory (so that their addresses
12564 can be loaded with the @code{ld24} instruction), and assume all subroutines
12565 are reachable with the @code{bl} instruction.
12566 This is the default.
12568 The addressability of a particular object can be set with the
12569 @code{model} attribute.
12571 @item -mmodel=medium
12572 @opindex mmodel=medium
12573 Assume objects may be anywhere in the 32-bit address space (the compiler
12574 will generate @code{seth/add3} instructions to load their addresses), and
12575 assume all subroutines are reachable with the @code{bl} instruction.
12577 @item -mmodel=large
12578 @opindex mmodel=large
12579 Assume objects may be anywhere in the 32-bit address space (the compiler
12580 will generate @code{seth/add3} instructions to load their addresses), and
12581 assume subroutines may not be reachable with the @code{bl} instruction
12582 (the compiler will generate the much slower @code{seth/add3/jl}
12583 instruction sequence).
12586 @opindex msdata=none
12587 Disable use of the small data area. Variables will be put into
12588 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12589 @code{section} attribute has been specified).
12590 This is the default.
12592 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12593 Objects may be explicitly put in the small data area with the
12594 @code{section} attribute using one of these sections.
12596 @item -msdata=sdata
12597 @opindex msdata=sdata
12598 Put small global and static data in the small data area, but do not
12599 generate special code to reference them.
12602 @opindex msdata=use
12603 Put small global and static data in the small data area, and generate
12604 special instructions to reference them.
12608 @cindex smaller data references
12609 Put global and static objects less than or equal to @var{num} bytes
12610 into the small data or bss sections instead of the normal data or bss
12611 sections. The default value of @var{num} is 8.
12612 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12613 for this option to have any effect.
12615 All modules should be compiled with the same @option{-G @var{num}} value.
12616 Compiling with different values of @var{num} may or may not work; if it
12617 doesn't the linker will give an error message---incorrect code will not be
12622 Makes the M32R specific code in the compiler display some statistics
12623 that might help in debugging programs.
12625 @item -malign-loops
12626 @opindex malign-loops
12627 Align all loops to a 32-byte boundary.
12629 @item -mno-align-loops
12630 @opindex mno-align-loops
12631 Do not enforce a 32-byte alignment for loops. This is the default.
12633 @item -missue-rate=@var{number}
12634 @opindex missue-rate=@var{number}
12635 Issue @var{number} instructions per cycle. @var{number} can only be 1
12638 @item -mbranch-cost=@var{number}
12639 @opindex mbranch-cost=@var{number}
12640 @var{number} can only be 1 or 2. If it is 1 then branches will be
12641 preferred over conditional code, if it is 2, then the opposite will
12644 @item -mflush-trap=@var{number}
12645 @opindex mflush-trap=@var{number}
12646 Specifies the trap number to use to flush the cache. The default is
12647 12. Valid numbers are between 0 and 15 inclusive.
12649 @item -mno-flush-trap
12650 @opindex mno-flush-trap
12651 Specifies that the cache cannot be flushed by using a trap.
12653 @item -mflush-func=@var{name}
12654 @opindex mflush-func=@var{name}
12655 Specifies the name of the operating system function to call to flush
12656 the cache. The default is @emph{_flush_cache}, but a function call
12657 will only be used if a trap is not available.
12659 @item -mno-flush-func
12660 @opindex mno-flush-func
12661 Indicates that there is no OS function for flushing the cache.
12665 @node M680x0 Options
12666 @subsection M680x0 Options
12667 @cindex M680x0 options
12669 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12670 The default settings depend on which architecture was selected when
12671 the compiler was configured; the defaults for the most common choices
12675 @item -march=@var{arch}
12677 Generate code for a specific M680x0 or ColdFire instruction set
12678 architecture. Permissible values of @var{arch} for M680x0
12679 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12680 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12681 architectures are selected according to Freescale's ISA classification
12682 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12683 @samp{isab} and @samp{isac}.
12685 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12686 code for a ColdFire target. The @var{arch} in this macro is one of the
12687 @option{-march} arguments given above.
12689 When used together, @option{-march} and @option{-mtune} select code
12690 that runs on a family of similar processors but that is optimized
12691 for a particular microarchitecture.
12693 @item -mcpu=@var{cpu}
12695 Generate code for a specific M680x0 or ColdFire processor.
12696 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12697 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12698 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12699 below, which also classifies the CPUs into families:
12701 @multitable @columnfractions 0.20 0.80
12702 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12703 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12704 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12705 @item @samp{5206e} @tab @samp{5206e}
12706 @item @samp{5208} @tab @samp{5207} @samp{5208}
12707 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12708 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12709 @item @samp{5216} @tab @samp{5214} @samp{5216}
12710 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12711 @item @samp{5225} @tab @samp{5224} @samp{5225}
12712 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12713 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12714 @item @samp{5249} @tab @samp{5249}
12715 @item @samp{5250} @tab @samp{5250}
12716 @item @samp{5271} @tab @samp{5270} @samp{5271}
12717 @item @samp{5272} @tab @samp{5272}
12718 @item @samp{5275} @tab @samp{5274} @samp{5275}
12719 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12720 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12721 @item @samp{5307} @tab @samp{5307}
12722 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12723 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12724 @item @samp{5407} @tab @samp{5407}
12725 @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}
12728 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12729 @var{arch} is compatible with @var{cpu}. Other combinations of
12730 @option{-mcpu} and @option{-march} are rejected.
12732 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12733 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12734 where the value of @var{family} is given by the table above.
12736 @item -mtune=@var{tune}
12738 Tune the code for a particular microarchitecture, within the
12739 constraints set by @option{-march} and @option{-mcpu}.
12740 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12741 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12742 and @samp{cpu32}. The ColdFire microarchitectures
12743 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12745 You can also use @option{-mtune=68020-40} for code that needs
12746 to run relatively well on 68020, 68030 and 68040 targets.
12747 @option{-mtune=68020-60} is similar but includes 68060 targets
12748 as well. These two options select the same tuning decisions as
12749 @option{-m68020-40} and @option{-m68020-60} respectively.
12751 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12752 when tuning for 680x0 architecture @var{arch}. It also defines
12753 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12754 option is used. If gcc is tuning for a range of architectures,
12755 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12756 it defines the macros for every architecture in the range.
12758 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12759 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12760 of the arguments given above.
12766 Generate output for a 68000. This is the default
12767 when the compiler is configured for 68000-based systems.
12768 It is equivalent to @option{-march=68000}.
12770 Use this option for microcontrollers with a 68000 or EC000 core,
12771 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12775 Generate output for a 68010. This is the default
12776 when the compiler is configured for 68010-based systems.
12777 It is equivalent to @option{-march=68010}.
12783 Generate output for a 68020. This is the default
12784 when the compiler is configured for 68020-based systems.
12785 It is equivalent to @option{-march=68020}.
12789 Generate output for a 68030. This is the default when the compiler is
12790 configured for 68030-based systems. It is equivalent to
12791 @option{-march=68030}.
12795 Generate output for a 68040. This is the default when the compiler is
12796 configured for 68040-based systems. It is equivalent to
12797 @option{-march=68040}.
12799 This option inhibits the use of 68881/68882 instructions that have to be
12800 emulated by software on the 68040. Use this option if your 68040 does not
12801 have code to emulate those instructions.
12805 Generate output for a 68060. This is the default when the compiler is
12806 configured for 68060-based systems. It is equivalent to
12807 @option{-march=68060}.
12809 This option inhibits the use of 68020 and 68881/68882 instructions that
12810 have to be emulated by software on the 68060. Use this option if your 68060
12811 does not have code to emulate those instructions.
12815 Generate output for a CPU32. This is the default
12816 when the compiler is configured for CPU32-based systems.
12817 It is equivalent to @option{-march=cpu32}.
12819 Use this option for microcontrollers with a
12820 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12821 68336, 68340, 68341, 68349 and 68360.
12825 Generate output for a 520X ColdFire CPU@. This is the default
12826 when the compiler is configured for 520X-based systems.
12827 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12828 in favor of that option.
12830 Use this option for microcontroller with a 5200 core, including
12831 the MCF5202, MCF5203, MCF5204 and MCF5206.
12835 Generate output for a 5206e ColdFire CPU@. The option is now
12836 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12840 Generate output for a member of the ColdFire 528X family.
12841 The option is now deprecated in favor of the equivalent
12842 @option{-mcpu=528x}.
12846 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12847 in favor of the equivalent @option{-mcpu=5307}.
12851 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12852 in favor of the equivalent @option{-mcpu=5407}.
12856 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12857 This includes use of hardware floating point instructions.
12858 The option is equivalent to @option{-mcpu=547x}, and is now
12859 deprecated in favor of that option.
12863 Generate output for a 68040, without using any of the new instructions.
12864 This results in code which can run relatively efficiently on either a
12865 68020/68881 or a 68030 or a 68040. The generated code does use the
12866 68881 instructions that are emulated on the 68040.
12868 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12872 Generate output for a 68060, without using any of the new instructions.
12873 This results in code which can run relatively efficiently on either a
12874 68020/68881 or a 68030 or a 68040. The generated code does use the
12875 68881 instructions that are emulated on the 68060.
12877 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12881 @opindex mhard-float
12883 Generate floating-point instructions. This is the default for 68020
12884 and above, and for ColdFire devices that have an FPU@. It defines the
12885 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12886 on ColdFire targets.
12889 @opindex msoft-float
12890 Do not generate floating-point instructions; use library calls instead.
12891 This is the default for 68000, 68010, and 68832 targets. It is also
12892 the default for ColdFire devices that have no FPU.
12898 Generate (do not generate) ColdFire hardware divide and remainder
12899 instructions. If @option{-march} is used without @option{-mcpu},
12900 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12901 architectures. Otherwise, the default is taken from the target CPU
12902 (either the default CPU, or the one specified by @option{-mcpu}). For
12903 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12904 @option{-mcpu=5206e}.
12906 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12910 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12911 Additionally, parameters passed on the stack are also aligned to a
12912 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12916 Do not consider type @code{int} to be 16 bits wide. This is the default.
12919 @itemx -mno-bitfield
12920 @opindex mnobitfield
12921 @opindex mno-bitfield
12922 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12923 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12927 Do use the bit-field instructions. The @option{-m68020} option implies
12928 @option{-mbitfield}. This is the default if you use a configuration
12929 designed for a 68020.
12933 Use a different function-calling convention, in which functions
12934 that take a fixed number of arguments return with the @code{rtd}
12935 instruction, which pops their arguments while returning. This
12936 saves one instruction in the caller since there is no need to pop
12937 the arguments there.
12939 This calling convention is incompatible with the one normally
12940 used on Unix, so you cannot use it if you need to call libraries
12941 compiled with the Unix compiler.
12943 Also, you must provide function prototypes for all functions that
12944 take variable numbers of arguments (including @code{printf});
12945 otherwise incorrect code will be generated for calls to those
12948 In addition, seriously incorrect code will result if you call a
12949 function with too many arguments. (Normally, extra arguments are
12950 harmlessly ignored.)
12952 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12953 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12957 Do not use the calling conventions selected by @option{-mrtd}.
12958 This is the default.
12961 @itemx -mno-align-int
12962 @opindex malign-int
12963 @opindex mno-align-int
12964 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12965 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12966 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12967 Aligning variables on 32-bit boundaries produces code that runs somewhat
12968 faster on processors with 32-bit busses at the expense of more memory.
12970 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12971 align structures containing the above types differently than
12972 most published application binary interface specifications for the m68k.
12976 Use the pc-relative addressing mode of the 68000 directly, instead of
12977 using a global offset table. At present, this option implies @option{-fpic},
12978 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12979 not presently supported with @option{-mpcrel}, though this could be supported for
12980 68020 and higher processors.
12982 @item -mno-strict-align
12983 @itemx -mstrict-align
12984 @opindex mno-strict-align
12985 @opindex mstrict-align
12986 Do not (do) assume that unaligned memory references will be handled by
12990 Generate code that allows the data segment to be located in a different
12991 area of memory from the text segment. This allows for execute in place in
12992 an environment without virtual memory management. This option implies
12995 @item -mno-sep-data
12996 Generate code that assumes that the data segment follows the text segment.
12997 This is the default.
12999 @item -mid-shared-library
13000 Generate code that supports shared libraries via the library ID method.
13001 This allows for execute in place and shared libraries in an environment
13002 without virtual memory management. This option implies @option{-fPIC}.
13004 @item -mno-id-shared-library
13005 Generate code that doesn't assume ID based shared libraries are being used.
13006 This is the default.
13008 @item -mshared-library-id=n
13009 Specified the identification number of the ID based shared library being
13010 compiled. Specifying a value of 0 will generate more compact code, specifying
13011 other values will force the allocation of that number to the current
13012 library but is no more space or time efficient than omitting this option.
13018 When generating position-independent code for ColdFire, generate code
13019 that works if the GOT has more than 8192 entries. This code is
13020 larger and slower than code generated without this option. On M680x0
13021 processors, this option is not needed; @option{-fPIC} suffices.
13023 GCC normally uses a single instruction to load values from the GOT@.
13024 While this is relatively efficient, it only works if the GOT
13025 is smaller than about 64k. Anything larger causes the linker
13026 to report an error such as:
13028 @cindex relocation truncated to fit (ColdFire)
13030 relocation truncated to fit: R_68K_GOT16O foobar
13033 If this happens, you should recompile your code with @option{-mxgot}.
13034 It should then work with very large GOTs. However, code generated with
13035 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13036 the value of a global symbol.
13038 Note that some linkers, including newer versions of the GNU linker,
13039 can create multiple GOTs and sort GOT entries. If you have such a linker,
13040 you should only need to use @option{-mxgot} when compiling a single
13041 object file that accesses more than 8192 GOT entries. Very few do.
13043 These options have no effect unless GCC is generating
13044 position-independent code.
13048 @node M68hc1x Options
13049 @subsection M68hc1x Options
13050 @cindex M68hc1x options
13052 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13053 microcontrollers. The default values for these options depends on
13054 which style of microcontroller was selected when the compiler was configured;
13055 the defaults for the most common choices are given below.
13062 Generate output for a 68HC11. This is the default
13063 when the compiler is configured for 68HC11-based systems.
13069 Generate output for a 68HC12. This is the default
13070 when the compiler is configured for 68HC12-based systems.
13076 Generate output for a 68HCS12.
13078 @item -mauto-incdec
13079 @opindex mauto-incdec
13080 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13087 Enable the use of 68HC12 min and max instructions.
13090 @itemx -mno-long-calls
13091 @opindex mlong-calls
13092 @opindex mno-long-calls
13093 Treat all calls as being far away (near). If calls are assumed to be
13094 far away, the compiler will use the @code{call} instruction to
13095 call a function and the @code{rtc} instruction for returning.
13099 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13101 @item -msoft-reg-count=@var{count}
13102 @opindex msoft-reg-count
13103 Specify the number of pseudo-soft registers which are used for the
13104 code generation. The maximum number is 32. Using more pseudo-soft
13105 register may or may not result in better code depending on the program.
13106 The default is 4 for 68HC11 and 2 for 68HC12.
13110 @node MCore Options
13111 @subsection MCore Options
13112 @cindex MCore options
13114 These are the @samp{-m} options defined for the Motorola M*Core
13120 @itemx -mno-hardlit
13122 @opindex mno-hardlit
13123 Inline constants into the code stream if it can be done in two
13124 instructions or less.
13130 Use the divide instruction. (Enabled by default).
13132 @item -mrelax-immediate
13133 @itemx -mno-relax-immediate
13134 @opindex mrelax-immediate
13135 @opindex mno-relax-immediate
13136 Allow arbitrary sized immediates in bit operations.
13138 @item -mwide-bitfields
13139 @itemx -mno-wide-bitfields
13140 @opindex mwide-bitfields
13141 @opindex mno-wide-bitfields
13142 Always treat bit-fields as int-sized.
13144 @item -m4byte-functions
13145 @itemx -mno-4byte-functions
13146 @opindex m4byte-functions
13147 @opindex mno-4byte-functions
13148 Force all functions to be aligned to a four byte boundary.
13150 @item -mcallgraph-data
13151 @itemx -mno-callgraph-data
13152 @opindex mcallgraph-data
13153 @opindex mno-callgraph-data
13154 Emit callgraph information.
13157 @itemx -mno-slow-bytes
13158 @opindex mslow-bytes
13159 @opindex mno-slow-bytes
13160 Prefer word access when reading byte quantities.
13162 @item -mlittle-endian
13163 @itemx -mbig-endian
13164 @opindex mlittle-endian
13165 @opindex mbig-endian
13166 Generate code for a little endian target.
13172 Generate code for the 210 processor.
13176 Assume that run-time support has been provided and so omit the
13177 simulator library (@file{libsim.a)} from the linker command line.
13179 @item -mstack-increment=@var{size}
13180 @opindex mstack-increment
13181 Set the maximum amount for a single stack increment operation. Large
13182 values can increase the speed of programs which contain functions
13183 that need a large amount of stack space, but they can also trigger a
13184 segmentation fault if the stack is extended too much. The default
13190 @subsection MeP Options
13191 @cindex MeP options
13197 Enables the @code{abs} instruction, which is the absolute difference
13198 between two registers.
13202 Enables all the optional instructions - average, multiply, divide, bit
13203 operations, leading zero, absolute difference, min/max, clip, and
13209 Enables the @code{ave} instruction, which computes the average of two
13212 @item -mbased=@var{n}
13214 Variables of size @var{n} bytes or smaller will be placed in the
13215 @code{.based} section by default. Based variables use the @code{$tp}
13216 register as a base register, and there is a 128 byte limit to the
13217 @code{.based} section.
13221 Enables the bit operation instructions - bit test (@code{btstm}), set
13222 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13223 test-and-set (@code{tas}).
13225 @item -mc=@var{name}
13227 Selects which section constant data will be placed in. @var{name} may
13228 be @code{tiny}, @code{near}, or @code{far}.
13232 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13233 useful unless you also provide @code{-mminmax}.
13235 @item -mconfig=@var{name}
13237 Selects one of the build-in core configurations. Each MeP chip has
13238 one or more modules in it; each module has a core CPU and a variety of
13239 coprocessors, optional instructions, and peripherals. The
13240 @code{MeP-Integrator} tool, not part of GCC, provides these
13241 configurations through this option; using this option is the same as
13242 using all the corresponding command line options. The default
13243 configuration is @code{default}.
13247 Enables the coprocessor instructions. By default, this is a 32-bit
13248 coprocessor. Note that the coprocessor is normally enabled via the
13249 @code{-mconfig=} option.
13253 Enables the 32-bit coprocessor's instructions.
13257 Enables the 64-bit coprocessor's instructions.
13261 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13265 Causes constant variables to be placed in the @code{.near} section.
13269 Enables the @code{div} and @code{divu} instructions.
13273 Generate big-endian code.
13277 Generate little-endian code.
13279 @item -mio-volatile
13280 @opindex mio-volatile
13281 Tells the compiler that any variable marked with the @code{io}
13282 attribute is to be considered volatile.
13286 Causes variables to be assigned to the @code{.far} section by default.
13290 Enables the @code{leadz} (leading zero) instruction.
13294 Causes variables to be assigned to the @code{.near} section by default.
13298 Enables the @code{min} and @code{max} instructions.
13302 Enables the multiplication and multiply-accumulate instructions.
13306 Disables all the optional instructions enabled by @code{-mall-opts}.
13310 Enables the @code{repeat} and @code{erepeat} instructions, used for
13311 low-overhead looping.
13315 Causes all variables to default to the @code{.tiny} section. Note
13316 that there is a 65536 byte limit to this section. Accesses to these
13317 variables use the @code{%gp} base register.
13321 Enables the saturation instructions. Note that the compiler does not
13322 currently generate these itself, but this option is included for
13323 compatibility with other tools, like @code{as}.
13327 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13331 Link the simulator runtime libraries.
13335 Link the simulator runtime libraries, excluding built-in support
13336 for reset and exception vectors and tables.
13340 Causes all functions to default to the @code{.far} section. Without
13341 this option, functions default to the @code{.near} section.
13343 @item -mtiny=@var{n}
13345 Variables that are @var{n} bytes or smaller will be allocated to the
13346 @code{.tiny} section. These variables use the @code{$gp} base
13347 register. The default for this option is 4, but note that there's a
13348 65536 byte limit to the @code{.tiny} section.
13353 @subsection MIPS Options
13354 @cindex MIPS options
13360 Generate big-endian code.
13364 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13367 @item -march=@var{arch}
13369 Generate code that will run on @var{arch}, which can be the name of a
13370 generic MIPS ISA, or the name of a particular processor.
13372 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13373 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13374 The processor names are:
13375 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13376 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13377 @samp{5kc}, @samp{5kf},
13379 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13380 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13381 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13382 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13383 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13384 @samp{loongson2e}, @samp{loongson2f},
13388 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13389 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13390 @samp{rm7000}, @samp{rm9000},
13391 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13394 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13395 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13397 The special value @samp{from-abi} selects the
13398 most compatible architecture for the selected ABI (that is,
13399 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13401 Native Linux/GNU toolchains also support the value @samp{native},
13402 which selects the best architecture option for the host processor.
13403 @option{-march=native} has no effect if GCC does not recognize
13406 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13407 (for example, @samp{-march=r2k}). Prefixes are optional, and
13408 @samp{vr} may be written @samp{r}.
13410 Names of the form @samp{@var{n}f2_1} refer to processors with
13411 FPUs clocked at half the rate of the core, names of the form
13412 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13413 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13414 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13415 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13416 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13417 accepted as synonyms for @samp{@var{n}f1_1}.
13419 GCC defines two macros based on the value of this option. The first
13420 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13421 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13422 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13423 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13424 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13426 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13427 above. In other words, it will have the full prefix and will not
13428 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13429 the macro names the resolved architecture (either @samp{"mips1"} or
13430 @samp{"mips3"}). It names the default architecture when no
13431 @option{-march} option is given.
13433 @item -mtune=@var{arch}
13435 Optimize for @var{arch}. Among other things, this option controls
13436 the way instructions are scheduled, and the perceived cost of arithmetic
13437 operations. The list of @var{arch} values is the same as for
13440 When this option is not used, GCC will optimize for the processor
13441 specified by @option{-march}. By using @option{-march} and
13442 @option{-mtune} together, it is possible to generate code that will
13443 run on a family of processors, but optimize the code for one
13444 particular member of that family.
13446 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13447 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13448 @samp{-march} ones described above.
13452 Equivalent to @samp{-march=mips1}.
13456 Equivalent to @samp{-march=mips2}.
13460 Equivalent to @samp{-march=mips3}.
13464 Equivalent to @samp{-march=mips4}.
13468 Equivalent to @samp{-march=mips32}.
13472 Equivalent to @samp{-march=mips32r2}.
13476 Equivalent to @samp{-march=mips64}.
13480 Equivalent to @samp{-march=mips64r2}.
13485 @opindex mno-mips16
13486 Generate (do not generate) MIPS16 code. If GCC is targetting a
13487 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13489 MIPS16 code generation can also be controlled on a per-function basis
13490 by means of @code{mips16} and @code{nomips16} attributes.
13491 @xref{Function Attributes}, for more information.
13493 @item -mflip-mips16
13494 @opindex mflip-mips16
13495 Generate MIPS16 code on alternating functions. This option is provided
13496 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13497 not intended for ordinary use in compiling user code.
13499 @item -minterlink-mips16
13500 @itemx -mno-interlink-mips16
13501 @opindex minterlink-mips16
13502 @opindex mno-interlink-mips16
13503 Require (do not require) that non-MIPS16 code be link-compatible with
13506 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13507 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13508 therefore disables direct jumps unless GCC knows that the target of the
13509 jump is not MIPS16.
13521 Generate code for the given ABI@.
13523 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13524 generates 64-bit code when you select a 64-bit architecture, but you
13525 can use @option{-mgp32} to get 32-bit code instead.
13527 For information about the O64 ABI, see
13528 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13530 GCC supports a variant of the o32 ABI in which floating-point registers
13531 are 64 rather than 32 bits wide. You can select this combination with
13532 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13533 and @samp{mfhc1} instructions and is therefore only supported for
13534 MIPS32R2 processors.
13536 The register assignments for arguments and return values remain the
13537 same, but each scalar value is passed in a single 64-bit register
13538 rather than a pair of 32-bit registers. For example, scalar
13539 floating-point values are returned in @samp{$f0} only, not a
13540 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13541 remains the same, but all 64 bits are saved.
13544 @itemx -mno-abicalls
13546 @opindex mno-abicalls
13547 Generate (do not generate) code that is suitable for SVR4-style
13548 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13553 Generate (do not generate) code that is fully position-independent,
13554 and that can therefore be linked into shared libraries. This option
13555 only affects @option{-mabicalls}.
13557 All @option{-mabicalls} code has traditionally been position-independent,
13558 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13559 as an extension, the GNU toolchain allows executables to use absolute
13560 accesses for locally-binding symbols. It can also use shorter GP
13561 initialization sequences and generate direct calls to locally-defined
13562 functions. This mode is selected by @option{-mno-shared}.
13564 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13565 objects that can only be linked by the GNU linker. However, the option
13566 does not affect the ABI of the final executable; it only affects the ABI
13567 of relocatable objects. Using @option{-mno-shared} will generally make
13568 executables both smaller and quicker.
13570 @option{-mshared} is the default.
13576 Assume (do not assume) that the static and dynamic linkers
13577 support PLTs and copy relocations. This option only affects
13578 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13579 has no effect without @samp{-msym32}.
13581 You can make @option{-mplt} the default by configuring
13582 GCC with @option{--with-mips-plt}. The default is
13583 @option{-mno-plt} otherwise.
13589 Lift (do not lift) the usual restrictions on the size of the global
13592 GCC normally uses a single instruction to load values from the GOT@.
13593 While this is relatively efficient, it will only work if the GOT
13594 is smaller than about 64k. Anything larger will cause the linker
13595 to report an error such as:
13597 @cindex relocation truncated to fit (MIPS)
13599 relocation truncated to fit: R_MIPS_GOT16 foobar
13602 If this happens, you should recompile your code with @option{-mxgot}.
13603 It should then work with very large GOTs, although it will also be
13604 less efficient, since it will take three instructions to fetch the
13605 value of a global symbol.
13607 Note that some linkers can create multiple GOTs. If you have such a
13608 linker, you should only need to use @option{-mxgot} when a single object
13609 file accesses more than 64k's worth of GOT entries. Very few do.
13611 These options have no effect unless GCC is generating position
13616 Assume that general-purpose registers are 32 bits wide.
13620 Assume that general-purpose registers are 64 bits wide.
13624 Assume that floating-point registers are 32 bits wide.
13628 Assume that floating-point registers are 64 bits wide.
13631 @opindex mhard-float
13632 Use floating-point coprocessor instructions.
13635 @opindex msoft-float
13636 Do not use floating-point coprocessor instructions. Implement
13637 floating-point calculations using library calls instead.
13639 @item -msingle-float
13640 @opindex msingle-float
13641 Assume that the floating-point coprocessor only supports single-precision
13644 @item -mdouble-float
13645 @opindex mdouble-float
13646 Assume that the floating-point coprocessor supports double-precision
13647 operations. This is the default.
13653 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13654 implement atomic memory built-in functions. When neither option is
13655 specified, GCC will use the instructions if the target architecture
13658 @option{-mllsc} is useful if the runtime environment can emulate the
13659 instructions and @option{-mno-llsc} can be useful when compiling for
13660 nonstandard ISAs. You can make either option the default by
13661 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13662 respectively. @option{--with-llsc} is the default for some
13663 configurations; see the installation documentation for details.
13669 Use (do not use) revision 1 of the MIPS DSP ASE@.
13670 @xref{MIPS DSP Built-in Functions}. This option defines the
13671 preprocessor macro @samp{__mips_dsp}. It also defines
13672 @samp{__mips_dsp_rev} to 1.
13678 Use (do not use) revision 2 of the MIPS DSP ASE@.
13679 @xref{MIPS DSP Built-in Functions}. This option defines the
13680 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13681 It also defines @samp{__mips_dsp_rev} to 2.
13684 @itemx -mno-smartmips
13685 @opindex msmartmips
13686 @opindex mno-smartmips
13687 Use (do not use) the MIPS SmartMIPS ASE.
13689 @item -mpaired-single
13690 @itemx -mno-paired-single
13691 @opindex mpaired-single
13692 @opindex mno-paired-single
13693 Use (do not use) paired-single floating-point instructions.
13694 @xref{MIPS Paired-Single Support}. This option requires
13695 hardware floating-point support to be enabled.
13701 Use (do not use) MIPS Digital Media Extension instructions.
13702 This option can only be used when generating 64-bit code and requires
13703 hardware floating-point support to be enabled.
13708 @opindex mno-mips3d
13709 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13710 The option @option{-mips3d} implies @option{-mpaired-single}.
13716 Use (do not use) MT Multithreading instructions.
13720 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13721 an explanation of the default and the way that the pointer size is
13726 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13728 The default size of @code{int}s, @code{long}s and pointers depends on
13729 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13730 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13731 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13732 or the same size as integer registers, whichever is smaller.
13738 Assume (do not assume) that all symbols have 32-bit values, regardless
13739 of the selected ABI@. This option is useful in combination with
13740 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13741 to generate shorter and faster references to symbolic addresses.
13745 Put definitions of externally-visible data in a small data section
13746 if that data is no bigger than @var{num} bytes. GCC can then access
13747 the data more efficiently; see @option{-mgpopt} for details.
13749 The default @option{-G} option depends on the configuration.
13751 @item -mlocal-sdata
13752 @itemx -mno-local-sdata
13753 @opindex mlocal-sdata
13754 @opindex mno-local-sdata
13755 Extend (do not extend) the @option{-G} behavior to local data too,
13756 such as to static variables in C@. @option{-mlocal-sdata} is the
13757 default for all configurations.
13759 If the linker complains that an application is using too much small data,
13760 you might want to try rebuilding the less performance-critical parts with
13761 @option{-mno-local-sdata}. You might also want to build large
13762 libraries with @option{-mno-local-sdata}, so that the libraries leave
13763 more room for the main program.
13765 @item -mextern-sdata
13766 @itemx -mno-extern-sdata
13767 @opindex mextern-sdata
13768 @opindex mno-extern-sdata
13769 Assume (do not assume) that externally-defined data will be in
13770 a small data section if that data is within the @option{-G} limit.
13771 @option{-mextern-sdata} is the default for all configurations.
13773 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13774 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13775 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13776 is placed in a small data section. If @var{Var} is defined by another
13777 module, you must either compile that module with a high-enough
13778 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13779 definition. If @var{Var} is common, you must link the application
13780 with a high-enough @option{-G} setting.
13782 The easiest way of satisfying these restrictions is to compile
13783 and link every module with the same @option{-G} option. However,
13784 you may wish to build a library that supports several different
13785 small data limits. You can do this by compiling the library with
13786 the highest supported @option{-G} setting and additionally using
13787 @option{-mno-extern-sdata} to stop the library from making assumptions
13788 about externally-defined data.
13794 Use (do not use) GP-relative accesses for symbols that are known to be
13795 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13796 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13799 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13800 might not hold the value of @code{_gp}. For example, if the code is
13801 part of a library that might be used in a boot monitor, programs that
13802 call boot monitor routines will pass an unknown value in @code{$gp}.
13803 (In such situations, the boot monitor itself would usually be compiled
13804 with @option{-G0}.)
13806 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13807 @option{-mno-extern-sdata}.
13809 @item -membedded-data
13810 @itemx -mno-embedded-data
13811 @opindex membedded-data
13812 @opindex mno-embedded-data
13813 Allocate variables to the read-only data section first if possible, then
13814 next in the small data section if possible, otherwise in data. This gives
13815 slightly slower code than the default, but reduces the amount of RAM required
13816 when executing, and thus may be preferred for some embedded systems.
13818 @item -muninit-const-in-rodata
13819 @itemx -mno-uninit-const-in-rodata
13820 @opindex muninit-const-in-rodata
13821 @opindex mno-uninit-const-in-rodata
13822 Put uninitialized @code{const} variables in the read-only data section.
13823 This option is only meaningful in conjunction with @option{-membedded-data}.
13825 @item -mcode-readable=@var{setting}
13826 @opindex mcode-readable
13827 Specify whether GCC may generate code that reads from executable sections.
13828 There are three possible settings:
13831 @item -mcode-readable=yes
13832 Instructions may freely access executable sections. This is the
13835 @item -mcode-readable=pcrel
13836 MIPS16 PC-relative load instructions can access executable sections,
13837 but other instructions must not do so. This option is useful on 4KSc
13838 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13839 It is also useful on processors that can be configured to have a dual
13840 instruction/data SRAM interface and that, like the M4K, automatically
13841 redirect PC-relative loads to the instruction RAM.
13843 @item -mcode-readable=no
13844 Instructions must not access executable sections. This option can be
13845 useful on targets that are configured to have a dual instruction/data
13846 SRAM interface but that (unlike the M4K) do not automatically redirect
13847 PC-relative loads to the instruction RAM.
13850 @item -msplit-addresses
13851 @itemx -mno-split-addresses
13852 @opindex msplit-addresses
13853 @opindex mno-split-addresses
13854 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13855 relocation operators. This option has been superseded by
13856 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13858 @item -mexplicit-relocs
13859 @itemx -mno-explicit-relocs
13860 @opindex mexplicit-relocs
13861 @opindex mno-explicit-relocs
13862 Use (do not use) assembler relocation operators when dealing with symbolic
13863 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13864 is to use assembler macros instead.
13866 @option{-mexplicit-relocs} is the default if GCC was configured
13867 to use an assembler that supports relocation operators.
13869 @item -mcheck-zero-division
13870 @itemx -mno-check-zero-division
13871 @opindex mcheck-zero-division
13872 @opindex mno-check-zero-division
13873 Trap (do not trap) on integer division by zero.
13875 The default is @option{-mcheck-zero-division}.
13877 @item -mdivide-traps
13878 @itemx -mdivide-breaks
13879 @opindex mdivide-traps
13880 @opindex mdivide-breaks
13881 MIPS systems check for division by zero by generating either a
13882 conditional trap or a break instruction. Using traps results in
13883 smaller code, but is only supported on MIPS II and later. Also, some
13884 versions of the Linux kernel have a bug that prevents trap from
13885 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13886 allow conditional traps on architectures that support them and
13887 @option{-mdivide-breaks} to force the use of breaks.
13889 The default is usually @option{-mdivide-traps}, but this can be
13890 overridden at configure time using @option{--with-divide=breaks}.
13891 Divide-by-zero checks can be completely disabled using
13892 @option{-mno-check-zero-division}.
13897 @opindex mno-memcpy
13898 Force (do not force) the use of @code{memcpy()} for non-trivial block
13899 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13900 most constant-sized copies.
13903 @itemx -mno-long-calls
13904 @opindex mlong-calls
13905 @opindex mno-long-calls
13906 Disable (do not disable) use of the @code{jal} instruction. Calling
13907 functions using @code{jal} is more efficient but requires the caller
13908 and callee to be in the same 256 megabyte segment.
13910 This option has no effect on abicalls code. The default is
13911 @option{-mno-long-calls}.
13917 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13918 instructions, as provided by the R4650 ISA@.
13921 @itemx -mno-fused-madd
13922 @opindex mfused-madd
13923 @opindex mno-fused-madd
13924 Enable (disable) use of the floating point multiply-accumulate
13925 instructions, when they are available. The default is
13926 @option{-mfused-madd}.
13928 When multiply-accumulate instructions are used, the intermediate
13929 product is calculated to infinite precision and is not subject to
13930 the FCSR Flush to Zero bit. This may be undesirable in some
13935 Tell the MIPS assembler to not run its preprocessor over user
13936 assembler files (with a @samp{.s} suffix) when assembling them.
13939 @itemx -mno-fix-r4000
13940 @opindex mfix-r4000
13941 @opindex mno-fix-r4000
13942 Work around certain R4000 CPU errata:
13945 A double-word or a variable shift may give an incorrect result if executed
13946 immediately after starting an integer division.
13948 A double-word or a variable shift may give an incorrect result if executed
13949 while an integer multiplication is in progress.
13951 An integer division may give an incorrect result if started in a delay slot
13952 of a taken branch or a jump.
13956 @itemx -mno-fix-r4400
13957 @opindex mfix-r4400
13958 @opindex mno-fix-r4400
13959 Work around certain R4400 CPU errata:
13962 A double-word or a variable shift may give an incorrect result if executed
13963 immediately after starting an integer division.
13967 @itemx -mno-fix-r10000
13968 @opindex mfix-r10000
13969 @opindex mno-fix-r10000
13970 Work around certain R10000 errata:
13973 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13974 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13977 This option can only be used if the target architecture supports
13978 branch-likely instructions. @option{-mfix-r10000} is the default when
13979 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13983 @itemx -mno-fix-vr4120
13984 @opindex mfix-vr4120
13985 Work around certain VR4120 errata:
13988 @code{dmultu} does not always produce the correct result.
13990 @code{div} and @code{ddiv} do not always produce the correct result if one
13991 of the operands is negative.
13993 The workarounds for the division errata rely on special functions in
13994 @file{libgcc.a}. At present, these functions are only provided by
13995 the @code{mips64vr*-elf} configurations.
13997 Other VR4120 errata require a nop to be inserted between certain pairs of
13998 instructions. These errata are handled by the assembler, not by GCC itself.
14001 @opindex mfix-vr4130
14002 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14003 workarounds are implemented by the assembler rather than by GCC,
14004 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14005 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14006 instructions are available instead.
14009 @itemx -mno-fix-sb1
14011 Work around certain SB-1 CPU core errata.
14012 (This flag currently works around the SB-1 revision 2
14013 ``F1'' and ``F2'' floating point errata.)
14015 @item -mr10k-cache-barrier=@var{setting}
14016 @opindex mr10k-cache-barrier
14017 Specify whether GCC should insert cache barriers to avoid the
14018 side-effects of speculation on R10K processors.
14020 In common with many processors, the R10K tries to predict the outcome
14021 of a conditional branch and speculatively executes instructions from
14022 the ``taken'' branch. It later aborts these instructions if the
14023 predicted outcome was wrong. However, on the R10K, even aborted
14024 instructions can have side effects.
14026 This problem only affects kernel stores and, depending on the system,
14027 kernel loads. As an example, a speculatively-executed store may load
14028 the target memory into cache and mark the cache line as dirty, even if
14029 the store itself is later aborted. If a DMA operation writes to the
14030 same area of memory before the ``dirty'' line is flushed, the cached
14031 data will overwrite the DMA-ed data. See the R10K processor manual
14032 for a full description, including other potential problems.
14034 One workaround is to insert cache barrier instructions before every memory
14035 access that might be speculatively executed and that might have side
14036 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14037 controls GCC's implementation of this workaround. It assumes that
14038 aborted accesses to any byte in the following regions will not have
14043 the memory occupied by the current function's stack frame;
14046 the memory occupied by an incoming stack argument;
14049 the memory occupied by an object with a link-time-constant address.
14052 It is the kernel's responsibility to ensure that speculative
14053 accesses to these regions are indeed safe.
14055 If the input program contains a function declaration such as:
14061 then the implementation of @code{foo} must allow @code{j foo} and
14062 @code{jal foo} to be executed speculatively. GCC honors this
14063 restriction for functions it compiles itself. It expects non-GCC
14064 functions (such as hand-written assembly code) to do the same.
14066 The option has three forms:
14069 @item -mr10k-cache-barrier=load-store
14070 Insert a cache barrier before a load or store that might be
14071 speculatively executed and that might have side effects even
14074 @item -mr10k-cache-barrier=store
14075 Insert a cache barrier before a store that might be speculatively
14076 executed and that might have side effects even if aborted.
14078 @item -mr10k-cache-barrier=none
14079 Disable the insertion of cache barriers. This is the default setting.
14082 @item -mflush-func=@var{func}
14083 @itemx -mno-flush-func
14084 @opindex mflush-func
14085 Specifies the function to call to flush the I and D caches, or to not
14086 call any such function. If called, the function must take the same
14087 arguments as the common @code{_flush_func()}, that is, the address of the
14088 memory range for which the cache is being flushed, the size of the
14089 memory range, and the number 3 (to flush both caches). The default
14090 depends on the target GCC was configured for, but commonly is either
14091 @samp{_flush_func} or @samp{__cpu_flush}.
14093 @item mbranch-cost=@var{num}
14094 @opindex mbranch-cost
14095 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14096 This cost is only a heuristic and is not guaranteed to produce
14097 consistent results across releases. A zero cost redundantly selects
14098 the default, which is based on the @option{-mtune} setting.
14100 @item -mbranch-likely
14101 @itemx -mno-branch-likely
14102 @opindex mbranch-likely
14103 @opindex mno-branch-likely
14104 Enable or disable use of Branch Likely instructions, regardless of the
14105 default for the selected architecture. By default, Branch Likely
14106 instructions may be generated if they are supported by the selected
14107 architecture. An exception is for the MIPS32 and MIPS64 architectures
14108 and processors which implement those architectures; for those, Branch
14109 Likely instructions will not be generated by default because the MIPS32
14110 and MIPS64 architectures specifically deprecate their use.
14112 @item -mfp-exceptions
14113 @itemx -mno-fp-exceptions
14114 @opindex mfp-exceptions
14115 Specifies whether FP exceptions are enabled. This affects how we schedule
14116 FP instructions for some processors. The default is that FP exceptions are
14119 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14120 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14123 @item -mvr4130-align
14124 @itemx -mno-vr4130-align
14125 @opindex mvr4130-align
14126 The VR4130 pipeline is two-way superscalar, but can only issue two
14127 instructions together if the first one is 8-byte aligned. When this
14128 option is enabled, GCC will align pairs of instructions that it
14129 thinks should execute in parallel.
14131 This option only has an effect when optimizing for the VR4130.
14132 It normally makes code faster, but at the expense of making it bigger.
14133 It is enabled by default at optimization level @option{-O3}.
14138 Enable (disable) generation of @code{synci} instructions on
14139 architectures that support it. The @code{synci} instructions (if
14140 enabled) will be generated when @code{__builtin___clear_cache()} is
14143 This option defaults to @code{-mno-synci}, but the default can be
14144 overridden by configuring with @code{--with-synci}.
14146 When compiling code for single processor systems, it is generally safe
14147 to use @code{synci}. However, on many multi-core (SMP) systems, it
14148 will not invalidate the instruction caches on all cores and may lead
14149 to undefined behavior.
14151 @item -mrelax-pic-calls
14152 @itemx -mno-relax-pic-calls
14153 @opindex mrelax-pic-calls
14154 Try to turn PIC calls that are normally dispatched via register
14155 @code{$25} into direct calls. This is only possible if the linker can
14156 resolve the destination at link-time and if the destination is within
14157 range for a direct call.
14159 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14160 an assembler and a linker that supports the @code{.reloc} assembly
14161 directive and @code{-mexplicit-relocs} is in effect. With
14162 @code{-mno-explicit-relocs}, this optimization can be performed by the
14163 assembler and the linker alone without help from the compiler.
14167 @subsection MMIX Options
14168 @cindex MMIX Options
14170 These options are defined for the MMIX:
14174 @itemx -mno-libfuncs
14176 @opindex mno-libfuncs
14177 Specify that intrinsic library functions are being compiled, passing all
14178 values in registers, no matter the size.
14181 @itemx -mno-epsilon
14183 @opindex mno-epsilon
14184 Generate floating-point comparison instructions that compare with respect
14185 to the @code{rE} epsilon register.
14187 @item -mabi=mmixware
14189 @opindex mabi=mmixware
14191 Generate code that passes function parameters and return values that (in
14192 the called function) are seen as registers @code{$0} and up, as opposed to
14193 the GNU ABI which uses global registers @code{$231} and up.
14195 @item -mzero-extend
14196 @itemx -mno-zero-extend
14197 @opindex mzero-extend
14198 @opindex mno-zero-extend
14199 When reading data from memory in sizes shorter than 64 bits, use (do not
14200 use) zero-extending load instructions by default, rather than
14201 sign-extending ones.
14204 @itemx -mno-knuthdiv
14206 @opindex mno-knuthdiv
14207 Make the result of a division yielding a remainder have the same sign as
14208 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14209 remainder follows the sign of the dividend. Both methods are
14210 arithmetically valid, the latter being almost exclusively used.
14212 @item -mtoplevel-symbols
14213 @itemx -mno-toplevel-symbols
14214 @opindex mtoplevel-symbols
14215 @opindex mno-toplevel-symbols
14216 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14217 code can be used with the @code{PREFIX} assembly directive.
14221 Generate an executable in the ELF format, rather than the default
14222 @samp{mmo} format used by the @command{mmix} simulator.
14224 @item -mbranch-predict
14225 @itemx -mno-branch-predict
14226 @opindex mbranch-predict
14227 @opindex mno-branch-predict
14228 Use (do not use) the probable-branch instructions, when static branch
14229 prediction indicates a probable branch.
14231 @item -mbase-addresses
14232 @itemx -mno-base-addresses
14233 @opindex mbase-addresses
14234 @opindex mno-base-addresses
14235 Generate (do not generate) code that uses @emph{base addresses}. Using a
14236 base address automatically generates a request (handled by the assembler
14237 and the linker) for a constant to be set up in a global register. The
14238 register is used for one or more base address requests within the range 0
14239 to 255 from the value held in the register. The generally leads to short
14240 and fast code, but the number of different data items that can be
14241 addressed is limited. This means that a program that uses lots of static
14242 data may require @option{-mno-base-addresses}.
14244 @item -msingle-exit
14245 @itemx -mno-single-exit
14246 @opindex msingle-exit
14247 @opindex mno-single-exit
14248 Force (do not force) generated code to have a single exit point in each
14252 @node MN10300 Options
14253 @subsection MN10300 Options
14254 @cindex MN10300 options
14256 These @option{-m} options are defined for Matsushita MN10300 architectures:
14261 Generate code to avoid bugs in the multiply instructions for the MN10300
14262 processors. This is the default.
14264 @item -mno-mult-bug
14265 @opindex mno-mult-bug
14266 Do not generate code to avoid bugs in the multiply instructions for the
14267 MN10300 processors.
14271 Generate code which uses features specific to the AM33 processor.
14275 Do not generate code which uses features specific to the AM33 processor. This
14278 @item -mreturn-pointer-on-d0
14279 @opindex mreturn-pointer-on-d0
14280 When generating a function which returns a pointer, return the pointer
14281 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14282 only in a0, and attempts to call such functions without a prototype
14283 would result in errors. Note that this option is on by default; use
14284 @option{-mno-return-pointer-on-d0} to disable it.
14288 Do not link in the C run-time initialization object file.
14292 Indicate to the linker that it should perform a relaxation optimization pass
14293 to shorten branches, calls and absolute memory addresses. This option only
14294 has an effect when used on the command line for the final link step.
14296 This option makes symbolic debugging impossible.
14299 @node PDP-11 Options
14300 @subsection PDP-11 Options
14301 @cindex PDP-11 Options
14303 These options are defined for the PDP-11:
14308 Use hardware FPP floating point. This is the default. (FIS floating
14309 point on the PDP-11/40 is not supported.)
14312 @opindex msoft-float
14313 Do not use hardware floating point.
14317 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14321 Return floating-point results in memory. This is the default.
14325 Generate code for a PDP-11/40.
14329 Generate code for a PDP-11/45. This is the default.
14333 Generate code for a PDP-11/10.
14335 @item -mbcopy-builtin
14336 @opindex mbcopy-builtin
14337 Use inline @code{movmemhi} patterns for copying memory. This is the
14342 Do not use inline @code{movmemhi} patterns for copying memory.
14348 Use 16-bit @code{int}. This is the default.
14354 Use 32-bit @code{int}.
14357 @itemx -mno-float32
14359 @opindex mno-float32
14360 Use 64-bit @code{float}. This is the default.
14363 @itemx -mno-float64
14365 @opindex mno-float64
14366 Use 32-bit @code{float}.
14370 Use @code{abshi2} pattern. This is the default.
14374 Do not use @code{abshi2} pattern.
14376 @item -mbranch-expensive
14377 @opindex mbranch-expensive
14378 Pretend that branches are expensive. This is for experimenting with
14379 code generation only.
14381 @item -mbranch-cheap
14382 @opindex mbranch-cheap
14383 Do not pretend that branches are expensive. This is the default.
14387 Generate code for a system with split I&D@.
14391 Generate code for a system without split I&D@. This is the default.
14395 Use Unix assembler syntax. This is the default when configured for
14396 @samp{pdp11-*-bsd}.
14400 Use DEC assembler syntax. This is the default when configured for any
14401 PDP-11 target other than @samp{pdp11-*-bsd}.
14404 @node picoChip Options
14405 @subsection picoChip Options
14406 @cindex picoChip options
14408 These @samp{-m} options are defined for picoChip implementations:
14412 @item -mae=@var{ae_type}
14414 Set the instruction set, register set, and instruction scheduling
14415 parameters for array element type @var{ae_type}. Supported values
14416 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14418 @option{-mae=ANY} selects a completely generic AE type. Code
14419 generated with this option will run on any of the other AE types. The
14420 code will not be as efficient as it would be if compiled for a specific
14421 AE type, and some types of operation (e.g., multiplication) will not
14422 work properly on all types of AE.
14424 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14425 for compiled code, and is the default.
14427 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14428 option may suffer from poor performance of byte (char) manipulation,
14429 since the DSP AE does not provide hardware support for byte load/stores.
14431 @item -msymbol-as-address
14432 Enable the compiler to directly use a symbol name as an address in a
14433 load/store instruction, without first loading it into a
14434 register. Typically, the use of this option will generate larger
14435 programs, which run faster than when the option isn't used. However, the
14436 results vary from program to program, so it is left as a user option,
14437 rather than being permanently enabled.
14439 @item -mno-inefficient-warnings
14440 Disables warnings about the generation of inefficient code. These
14441 warnings can be generated, for example, when compiling code which
14442 performs byte-level memory operations on the MAC AE type. The MAC AE has
14443 no hardware support for byte-level memory operations, so all byte
14444 load/stores must be synthesized from word load/store operations. This is
14445 inefficient and a warning will be generated indicating to the programmer
14446 that they should rewrite the code to avoid byte operations, or to target
14447 an AE type which has the necessary hardware support. This option enables
14448 the warning to be turned off.
14452 @node PowerPC Options
14453 @subsection PowerPC Options
14454 @cindex PowerPC options
14456 These are listed under @xref{RS/6000 and PowerPC Options}.
14458 @node RS/6000 and PowerPC Options
14459 @subsection IBM RS/6000 and PowerPC Options
14460 @cindex RS/6000 and PowerPC Options
14461 @cindex IBM RS/6000 and PowerPC Options
14463 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14470 @itemx -mno-powerpc
14471 @itemx -mpowerpc-gpopt
14472 @itemx -mno-powerpc-gpopt
14473 @itemx -mpowerpc-gfxopt
14474 @itemx -mno-powerpc-gfxopt
14476 @itemx -mno-powerpc64
14480 @itemx -mno-popcntb
14482 @itemx -mno-popcntd
14490 @itemx -mno-hard-dfp
14494 @opindex mno-power2
14496 @opindex mno-powerpc
14497 @opindex mpowerpc-gpopt
14498 @opindex mno-powerpc-gpopt
14499 @opindex mpowerpc-gfxopt
14500 @opindex mno-powerpc-gfxopt
14501 @opindex mpowerpc64
14502 @opindex mno-powerpc64
14506 @opindex mno-popcntb
14508 @opindex mno-popcntd
14514 @opindex mno-mfpgpr
14516 @opindex mno-hard-dfp
14517 GCC supports two related instruction set architectures for the
14518 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14519 instructions supported by the @samp{rios} chip set used in the original
14520 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14521 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14522 the IBM 4xx, 6xx, and follow-on microprocessors.
14524 Neither architecture is a subset of the other. However there is a
14525 large common subset of instructions supported by both. An MQ
14526 register is included in processors supporting the POWER architecture.
14528 You use these options to specify which instructions are available on the
14529 processor you are using. The default value of these options is
14530 determined when configuring GCC@. Specifying the
14531 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14532 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14533 rather than the options listed above.
14535 The @option{-mpower} option allows GCC to generate instructions that
14536 are found only in the POWER architecture and to use the MQ register.
14537 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14538 to generate instructions that are present in the POWER2 architecture but
14539 not the original POWER architecture.
14541 The @option{-mpowerpc} option allows GCC to generate instructions that
14542 are found only in the 32-bit subset of the PowerPC architecture.
14543 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14544 GCC to use the optional PowerPC architecture instructions in the
14545 General Purpose group, including floating-point square root. Specifying
14546 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14547 use the optional PowerPC architecture instructions in the Graphics
14548 group, including floating-point select.
14550 The @option{-mmfcrf} option allows GCC to generate the move from
14551 condition register field instruction implemented on the POWER4
14552 processor and other processors that support the PowerPC V2.01
14554 The @option{-mpopcntb} option allows GCC to generate the popcount and
14555 double precision FP reciprocal estimate instruction implemented on the
14556 POWER5 processor and other processors that support the PowerPC V2.02
14558 The @option{-mpopcntd} option allows GCC to generate the popcount
14559 instruction implemented on the POWER7 processor and other processors
14560 that support the PowerPC V2.06 architecture.
14561 The @option{-mfprnd} option allows GCC to generate the FP round to
14562 integer instructions implemented on the POWER5+ processor and other
14563 processors that support the PowerPC V2.03 architecture.
14564 The @option{-mcmpb} option allows GCC to generate the compare bytes
14565 instruction implemented on the POWER6 processor and other processors
14566 that support the PowerPC V2.05 architecture.
14567 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14568 general purpose register instructions implemented on the POWER6X
14569 processor and other processors that support the extended PowerPC V2.05
14571 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14572 point instructions implemented on some POWER processors.
14574 The @option{-mpowerpc64} option allows GCC to generate the additional
14575 64-bit instructions that are found in the full PowerPC64 architecture
14576 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14577 @option{-mno-powerpc64}.
14579 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14580 will use only the instructions in the common subset of both
14581 architectures plus some special AIX common-mode calls, and will not use
14582 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14583 permits GCC to use any instruction from either architecture and to
14584 allow use of the MQ register; specify this for the Motorola MPC601.
14586 @item -mnew-mnemonics
14587 @itemx -mold-mnemonics
14588 @opindex mnew-mnemonics
14589 @opindex mold-mnemonics
14590 Select which mnemonics to use in the generated assembler code. With
14591 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14592 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14593 assembler mnemonics defined for the POWER architecture. Instructions
14594 defined in only one architecture have only one mnemonic; GCC uses that
14595 mnemonic irrespective of which of these options is specified.
14597 GCC defaults to the mnemonics appropriate for the architecture in
14598 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14599 value of these option. Unless you are building a cross-compiler, you
14600 should normally not specify either @option{-mnew-mnemonics} or
14601 @option{-mold-mnemonics}, but should instead accept the default.
14603 @item -mcpu=@var{cpu_type}
14605 Set architecture type, register usage, choice of mnemonics, and
14606 instruction scheduling parameters for machine type @var{cpu_type}.
14607 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14608 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14609 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
14610 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
14611 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14612 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14613 @samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14614 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14615 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
14616 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14617 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14619 @option{-mcpu=common} selects a completely generic processor. Code
14620 generated under this option will run on any POWER or PowerPC processor.
14621 GCC will use only the instructions in the common subset of both
14622 architectures, and will not use the MQ register. GCC assumes a generic
14623 processor model for scheduling purposes.
14625 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14626 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14627 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14628 types, with an appropriate, generic processor model assumed for
14629 scheduling purposes.
14631 The other options specify a specific processor. Code generated under
14632 those options will run best on that processor, and may not run at all on
14635 The @option{-mcpu} options automatically enable or disable the
14638 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14639 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14640 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14641 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14643 The particular options set for any particular CPU will vary between
14644 compiler versions, depending on what setting seems to produce optimal
14645 code for that CPU; it doesn't necessarily reflect the actual hardware's
14646 capabilities. If you wish to set an individual option to a particular
14647 value, you may specify it after the @option{-mcpu} option, like
14648 @samp{-mcpu=970 -mno-altivec}.
14650 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14651 not enabled or disabled by the @option{-mcpu} option at present because
14652 AIX does not have full support for these options. You may still
14653 enable or disable them individually if you're sure it'll work in your
14656 @item -mtune=@var{cpu_type}
14658 Set the instruction scheduling parameters for machine type
14659 @var{cpu_type}, but do not set the architecture type, register usage, or
14660 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14661 values for @var{cpu_type} are used for @option{-mtune} as for
14662 @option{-mcpu}. If both are specified, the code generated will use the
14663 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14664 scheduling parameters set by @option{-mtune}.
14670 Generate code to compute division as reciprocal estimate and iterative
14671 refinement, creating opportunities for increased throughput. This
14672 feature requires: optional PowerPC Graphics instruction set for single
14673 precision and FRE instruction for double precision, assuming divides
14674 cannot generate user-visible traps, and the domain values not include
14675 Infinities, denormals or zero denominator.
14678 @itemx -mno-altivec
14680 @opindex mno-altivec
14681 Generate code that uses (does not use) AltiVec instructions, and also
14682 enable the use of built-in functions that allow more direct access to
14683 the AltiVec instruction set. You may also need to set
14684 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14690 @opindex mno-vrsave
14691 Generate VRSAVE instructions when generating AltiVec code.
14693 @item -mgen-cell-microcode
14694 @opindex mgen-cell-microcode
14695 Generate Cell microcode instructions
14697 @item -mwarn-cell-microcode
14698 @opindex mwarn-cell-microcode
14699 Warning when a Cell microcode instruction is going to emitted. An example
14700 of a Cell microcode instruction is a variable shift.
14703 @opindex msecure-plt
14704 Generate code that allows ld and ld.so to build executables and shared
14705 libraries with non-exec .plt and .got sections. This is a PowerPC
14706 32-bit SYSV ABI option.
14710 Generate code that uses a BSS .plt section that ld.so fills in, and
14711 requires .plt and .got sections that are both writable and executable.
14712 This is a PowerPC 32-bit SYSV ABI option.
14718 This switch enables or disables the generation of ISEL instructions.
14720 @item -misel=@var{yes/no}
14721 This switch has been deprecated. Use @option{-misel} and
14722 @option{-mno-isel} instead.
14728 This switch enables or disables the generation of SPE simd
14734 @opindex mno-paired
14735 This switch enables or disables the generation of PAIRED simd
14738 @item -mspe=@var{yes/no}
14739 This option has been deprecated. Use @option{-mspe} and
14740 @option{-mno-spe} instead.
14746 Generate code that uses (does not use) vector/scalar (VSX)
14747 instructions, and also enable the use of built-in functions that allow
14748 more direct access to the VSX instruction set.
14750 @item -mfloat-gprs=@var{yes/single/double/no}
14751 @itemx -mfloat-gprs
14752 @opindex mfloat-gprs
14753 This switch enables or disables the generation of floating point
14754 operations on the general purpose registers for architectures that
14757 The argument @var{yes} or @var{single} enables the use of
14758 single-precision floating point operations.
14760 The argument @var{double} enables the use of single and
14761 double-precision floating point operations.
14763 The argument @var{no} disables floating point operations on the
14764 general purpose registers.
14766 This option is currently only available on the MPC854x.
14772 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14773 targets (including GNU/Linux). The 32-bit environment sets int, long
14774 and pointer to 32 bits and generates code that runs on any PowerPC
14775 variant. The 64-bit environment sets int to 32 bits and long and
14776 pointer to 64 bits, and generates code for PowerPC64, as for
14777 @option{-mpowerpc64}.
14780 @itemx -mno-fp-in-toc
14781 @itemx -mno-sum-in-toc
14782 @itemx -mminimal-toc
14784 @opindex mno-fp-in-toc
14785 @opindex mno-sum-in-toc
14786 @opindex mminimal-toc
14787 Modify generation of the TOC (Table Of Contents), which is created for
14788 every executable file. The @option{-mfull-toc} option is selected by
14789 default. In that case, GCC will allocate at least one TOC entry for
14790 each unique non-automatic variable reference in your program. GCC
14791 will also place floating-point constants in the TOC@. However, only
14792 16,384 entries are available in the TOC@.
14794 If you receive a linker error message that saying you have overflowed
14795 the available TOC space, you can reduce the amount of TOC space used
14796 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14797 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14798 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14799 generate code to calculate the sum of an address and a constant at
14800 run-time instead of putting that sum into the TOC@. You may specify one
14801 or both of these options. Each causes GCC to produce very slightly
14802 slower and larger code at the expense of conserving TOC space.
14804 If you still run out of space in the TOC even when you specify both of
14805 these options, specify @option{-mminimal-toc} instead. This option causes
14806 GCC to make only one TOC entry for every file. When you specify this
14807 option, GCC will produce code that is slower and larger but which
14808 uses extremely little TOC space. You may wish to use this option
14809 only on files that contain less frequently executed code.
14815 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14816 @code{long} type, and the infrastructure needed to support them.
14817 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14818 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14819 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14822 @itemx -mno-xl-compat
14823 @opindex mxl-compat
14824 @opindex mno-xl-compat
14825 Produce code that conforms more closely to IBM XL compiler semantics
14826 when using AIX-compatible ABI@. Pass floating-point arguments to
14827 prototyped functions beyond the register save area (RSA) on the stack
14828 in addition to argument FPRs. Do not assume that most significant
14829 double in 128-bit long double value is properly rounded when comparing
14830 values and converting to double. Use XL symbol names for long double
14833 The AIX calling convention was extended but not initially documented to
14834 handle an obscure K&R C case of calling a function that takes the
14835 address of its arguments with fewer arguments than declared. IBM XL
14836 compilers access floating point arguments which do not fit in the
14837 RSA from the stack when a subroutine is compiled without
14838 optimization. Because always storing floating-point arguments on the
14839 stack is inefficient and rarely needed, this option is not enabled by
14840 default and only is necessary when calling subroutines compiled by IBM
14841 XL compilers without optimization.
14845 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14846 application written to use message passing with special startup code to
14847 enable the application to run. The system must have PE installed in the
14848 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14849 must be overridden with the @option{-specs=} option to specify the
14850 appropriate directory location. The Parallel Environment does not
14851 support threads, so the @option{-mpe} option and the @option{-pthread}
14852 option are incompatible.
14854 @item -malign-natural
14855 @itemx -malign-power
14856 @opindex malign-natural
14857 @opindex malign-power
14858 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14859 @option{-malign-natural} overrides the ABI-defined alignment of larger
14860 types, such as floating-point doubles, on their natural size-based boundary.
14861 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14862 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14864 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14868 @itemx -mhard-float
14869 @opindex msoft-float
14870 @opindex mhard-float
14871 Generate code that does not use (uses) the floating-point register set.
14872 Software floating point emulation is provided if you use the
14873 @option{-msoft-float} option, and pass the option to GCC when linking.
14875 @item -msingle-float
14876 @itemx -mdouble-float
14877 @opindex msingle-float
14878 @opindex mdouble-float
14879 Generate code for single or double-precision floating point operations.
14880 @option{-mdouble-float} implies @option{-msingle-float}.
14883 @opindex msimple-fpu
14884 Do not generate sqrt and div instructions for hardware floating point unit.
14888 Specify type of floating point unit. Valid values are @var{sp_lite}
14889 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14890 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14891 and @var{dp_full} (equivalent to -mdouble-float).
14894 @opindex mxilinx-fpu
14895 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14898 @itemx -mno-multiple
14900 @opindex mno-multiple
14901 Generate code that uses (does not use) the load multiple word
14902 instructions and the store multiple word instructions. These
14903 instructions are generated by default on POWER systems, and not
14904 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14905 endian PowerPC systems, since those instructions do not work when the
14906 processor is in little endian mode. The exceptions are PPC740 and
14907 PPC750 which permit the instructions usage in little endian mode.
14912 @opindex mno-string
14913 Generate code that uses (does not use) the load string instructions
14914 and the store string word instructions to save multiple registers and
14915 do small block moves. These instructions are generated by default on
14916 POWER systems, and not generated on PowerPC systems. Do not use
14917 @option{-mstring} on little endian PowerPC systems, since those
14918 instructions do not work when the processor is in little endian mode.
14919 The exceptions are PPC740 and PPC750 which permit the instructions
14920 usage in little endian mode.
14925 @opindex mno-update
14926 Generate code that uses (does not use) the load or store instructions
14927 that update the base register to the address of the calculated memory
14928 location. These instructions are generated by default. If you use
14929 @option{-mno-update}, there is a small window between the time that the
14930 stack pointer is updated and the address of the previous frame is
14931 stored, which means code that walks the stack frame across interrupts or
14932 signals may get corrupted data.
14934 @item -mavoid-indexed-addresses
14935 @item -mno-avoid-indexed-addresses
14936 @opindex mavoid-indexed-addresses
14937 @opindex mno-avoid-indexed-addresses
14938 Generate code that tries to avoid (not avoid) the use of indexed load
14939 or store instructions. These instructions can incur a performance
14940 penalty on Power6 processors in certain situations, such as when
14941 stepping through large arrays that cross a 16M boundary. This option
14942 is enabled by default when targetting Power6 and disabled otherwise.
14945 @itemx -mno-fused-madd
14946 @opindex mfused-madd
14947 @opindex mno-fused-madd
14948 Generate code that uses (does not use) the floating point multiply and
14949 accumulate instructions. These instructions are generated by default if
14950 hardware floating is used.
14956 Generate code that uses (does not use) the half-word multiply and
14957 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14958 These instructions are generated by default when targetting those
14965 Generate code that uses (does not use) the string-search @samp{dlmzb}
14966 instruction on the IBM 405, 440 and 464 processors. This instruction is
14967 generated by default when targetting those processors.
14969 @item -mno-bit-align
14971 @opindex mno-bit-align
14972 @opindex mbit-align
14973 On System V.4 and embedded PowerPC systems do not (do) force structures
14974 and unions that contain bit-fields to be aligned to the base type of the
14977 For example, by default a structure containing nothing but 8
14978 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14979 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14980 the structure would be aligned to a 1 byte boundary and be one byte in
14983 @item -mno-strict-align
14984 @itemx -mstrict-align
14985 @opindex mno-strict-align
14986 @opindex mstrict-align
14987 On System V.4 and embedded PowerPC systems do not (do) assume that
14988 unaligned memory references will be handled by the system.
14990 @item -mrelocatable
14991 @itemx -mno-relocatable
14992 @opindex mrelocatable
14993 @opindex mno-relocatable
14994 On embedded PowerPC systems generate code that allows (does not allow)
14995 the program to be relocated to a different address at runtime. If you
14996 use @option{-mrelocatable} on any module, all objects linked together must
14997 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14999 @item -mrelocatable-lib
15000 @itemx -mno-relocatable-lib
15001 @opindex mrelocatable-lib
15002 @opindex mno-relocatable-lib
15003 On embedded PowerPC systems generate code that allows (does not allow)
15004 the program to be relocated to a different address at runtime. Modules
15005 compiled with @option{-mrelocatable-lib} can be linked with either modules
15006 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15007 with modules compiled with the @option{-mrelocatable} options.
15013 On System V.4 and embedded PowerPC systems do not (do) assume that
15014 register 2 contains a pointer to a global area pointing to the addresses
15015 used in the program.
15018 @itemx -mlittle-endian
15020 @opindex mlittle-endian
15021 On System V.4 and embedded PowerPC systems compile code for the
15022 processor in little endian mode. The @option{-mlittle-endian} option is
15023 the same as @option{-mlittle}.
15026 @itemx -mbig-endian
15028 @opindex mbig-endian
15029 On System V.4 and embedded PowerPC systems compile code for the
15030 processor in big endian mode. The @option{-mbig-endian} option is
15031 the same as @option{-mbig}.
15033 @item -mdynamic-no-pic
15034 @opindex mdynamic-no-pic
15035 On Darwin and Mac OS X systems, compile code so that it is not
15036 relocatable, but that its external references are relocatable. The
15037 resulting code is suitable for applications, but not shared
15040 @item -mprioritize-restricted-insns=@var{priority}
15041 @opindex mprioritize-restricted-insns
15042 This option controls the priority that is assigned to
15043 dispatch-slot restricted instructions during the second scheduling
15044 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15045 @var{no/highest/second-highest} priority to dispatch slot restricted
15048 @item -msched-costly-dep=@var{dependence_type}
15049 @opindex msched-costly-dep
15050 This option controls which dependences are considered costly
15051 by the target during instruction scheduling. The argument
15052 @var{dependence_type} takes one of the following values:
15053 @var{no}: no dependence is costly,
15054 @var{all}: all dependences are costly,
15055 @var{true_store_to_load}: a true dependence from store to load is costly,
15056 @var{store_to_load}: any dependence from store to load is costly,
15057 @var{number}: any dependence which latency >= @var{number} is costly.
15059 @item -minsert-sched-nops=@var{scheme}
15060 @opindex minsert-sched-nops
15061 This option controls which nop insertion scheme will be used during
15062 the second scheduling pass. The argument @var{scheme} takes one of the
15064 @var{no}: Don't insert nops.
15065 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15066 according to the scheduler's grouping.
15067 @var{regroup_exact}: Insert nops to force costly dependent insns into
15068 separate groups. Insert exactly as many nops as needed to force an insn
15069 to a new group, according to the estimated processor grouping.
15070 @var{number}: Insert nops to force costly dependent insns into
15071 separate groups. Insert @var{number} nops to force an insn to a new group.
15074 @opindex mcall-sysv
15075 On System V.4 and embedded PowerPC systems compile code using calling
15076 conventions that adheres to the March 1995 draft of the System V
15077 Application Binary Interface, PowerPC processor supplement. This is the
15078 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15080 @item -mcall-sysv-eabi
15082 @opindex mcall-sysv-eabi
15083 @opindex mcall-eabi
15084 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15086 @item -mcall-sysv-noeabi
15087 @opindex mcall-sysv-noeabi
15088 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15090 @item -mcall-aixdesc
15092 On System V.4 and embedded PowerPC systems compile code for the AIX
15096 @opindex mcall-linux
15097 On System V.4 and embedded PowerPC systems compile code for the
15098 Linux-based GNU system.
15102 On System V.4 and embedded PowerPC systems compile code for the
15103 Hurd-based GNU system.
15105 @item -mcall-freebsd
15106 @opindex mcall-freebsd
15107 On System V.4 and embedded PowerPC systems compile code for the
15108 FreeBSD operating system.
15110 @item -mcall-netbsd
15111 @opindex mcall-netbsd
15112 On System V.4 and embedded PowerPC systems compile code for the
15113 NetBSD operating system.
15115 @item -mcall-openbsd
15116 @opindex mcall-netbsd
15117 On System V.4 and embedded PowerPC systems compile code for the
15118 OpenBSD operating system.
15120 @item -maix-struct-return
15121 @opindex maix-struct-return
15122 Return all structures in memory (as specified by the AIX ABI)@.
15124 @item -msvr4-struct-return
15125 @opindex msvr4-struct-return
15126 Return structures smaller than 8 bytes in registers (as specified by the
15129 @item -mabi=@var{abi-type}
15131 Extend the current ABI with a particular extension, or remove such extension.
15132 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15133 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15137 Extend the current ABI with SPE ABI extensions. This does not change
15138 the default ABI, instead it adds the SPE ABI extensions to the current
15142 @opindex mabi=no-spe
15143 Disable Booke SPE ABI extensions for the current ABI@.
15145 @item -mabi=ibmlongdouble
15146 @opindex mabi=ibmlongdouble
15147 Change the current ABI to use IBM extended precision long double.
15148 This is a PowerPC 32-bit SYSV ABI option.
15150 @item -mabi=ieeelongdouble
15151 @opindex mabi=ieeelongdouble
15152 Change the current ABI to use IEEE extended precision long double.
15153 This is a PowerPC 32-bit Linux ABI option.
15156 @itemx -mno-prototype
15157 @opindex mprototype
15158 @opindex mno-prototype
15159 On System V.4 and embedded PowerPC systems assume that all calls to
15160 variable argument functions are properly prototyped. Otherwise, the
15161 compiler must insert an instruction before every non prototyped call to
15162 set or clear bit 6 of the condition code register (@var{CR}) to
15163 indicate whether floating point values were passed in the floating point
15164 registers in case the function takes a variable arguments. With
15165 @option{-mprototype}, only calls to prototyped variable argument functions
15166 will set or clear the bit.
15170 On embedded PowerPC systems, assume that the startup module is called
15171 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15172 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15177 On embedded PowerPC systems, assume that the startup module is called
15178 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15183 On embedded PowerPC systems, assume that the startup module is called
15184 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15187 @item -myellowknife
15188 @opindex myellowknife
15189 On embedded PowerPC systems, assume that the startup module is called
15190 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15195 On System V.4 and embedded PowerPC systems, specify that you are
15196 compiling for a VxWorks system.
15200 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15201 header to indicate that @samp{eabi} extended relocations are used.
15207 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15208 Embedded Applications Binary Interface (eabi) which is a set of
15209 modifications to the System V.4 specifications. Selecting @option{-meabi}
15210 means that the stack is aligned to an 8 byte boundary, a function
15211 @code{__eabi} is called to from @code{main} to set up the eabi
15212 environment, and the @option{-msdata} option can use both @code{r2} and
15213 @code{r13} to point to two separate small data areas. Selecting
15214 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15215 do not call an initialization function from @code{main}, and the
15216 @option{-msdata} option will only use @code{r13} to point to a single
15217 small data area. The @option{-meabi} option is on by default if you
15218 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15221 @opindex msdata=eabi
15222 On System V.4 and embedded PowerPC systems, put small initialized
15223 @code{const} global and static data in the @samp{.sdata2} section, which
15224 is pointed to by register @code{r2}. Put small initialized
15225 non-@code{const} global and static data in the @samp{.sdata} section,
15226 which is pointed to by register @code{r13}. Put small uninitialized
15227 global and static data in the @samp{.sbss} section, which is adjacent to
15228 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15229 incompatible with the @option{-mrelocatable} option. The
15230 @option{-msdata=eabi} option also sets the @option{-memb} option.
15233 @opindex msdata=sysv
15234 On System V.4 and embedded PowerPC systems, put small global and static
15235 data in the @samp{.sdata} section, which is pointed to by register
15236 @code{r13}. Put small uninitialized global and static data in the
15237 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15238 The @option{-msdata=sysv} option is incompatible with the
15239 @option{-mrelocatable} option.
15241 @item -msdata=default
15243 @opindex msdata=default
15245 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15246 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15247 same as @option{-msdata=sysv}.
15250 @opindex msdata=data
15251 On System V.4 and embedded PowerPC systems, put small global
15252 data in the @samp{.sdata} section. Put small uninitialized global
15253 data in the @samp{.sbss} section. Do not use register @code{r13}
15254 to address small data however. This is the default behavior unless
15255 other @option{-msdata} options are used.
15259 @opindex msdata=none
15261 On embedded PowerPC systems, put all initialized global and static data
15262 in the @samp{.data} section, and all uninitialized data in the
15263 @samp{.bss} section.
15267 @cindex smaller data references (PowerPC)
15268 @cindex .sdata/.sdata2 references (PowerPC)
15269 On embedded PowerPC systems, put global and static items less than or
15270 equal to @var{num} bytes into the small data or bss sections instead of
15271 the normal data or bss section. By default, @var{num} is 8. The
15272 @option{-G @var{num}} switch is also passed to the linker.
15273 All modules should be compiled with the same @option{-G @var{num}} value.
15276 @itemx -mno-regnames
15278 @opindex mno-regnames
15279 On System V.4 and embedded PowerPC systems do (do not) emit register
15280 names in the assembly language output using symbolic forms.
15283 @itemx -mno-longcall
15285 @opindex mno-longcall
15286 By default assume that all calls are far away so that a longer more
15287 expensive calling sequence is required. This is required for calls
15288 further than 32 megabytes (33,554,432 bytes) from the current location.
15289 A short call will be generated if the compiler knows
15290 the call cannot be that far away. This setting can be overridden by
15291 the @code{shortcall} function attribute, or by @code{#pragma
15294 Some linkers are capable of detecting out-of-range calls and generating
15295 glue code on the fly. On these systems, long calls are unnecessary and
15296 generate slower code. As of this writing, the AIX linker can do this,
15297 as can the GNU linker for PowerPC/64. It is planned to add this feature
15298 to the GNU linker for 32-bit PowerPC systems as well.
15300 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15301 callee, L42'', plus a ``branch island'' (glue code). The two target
15302 addresses represent the callee and the ``branch island''. The
15303 Darwin/PPC linker will prefer the first address and generate a ``bl
15304 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15305 otherwise, the linker will generate ``bl L42'' to call the ``branch
15306 island''. The ``branch island'' is appended to the body of the
15307 calling function; it computes the full 32-bit address of the callee
15310 On Mach-O (Darwin) systems, this option directs the compiler emit to
15311 the glue for every direct call, and the Darwin linker decides whether
15312 to use or discard it.
15314 In the future, we may cause GCC to ignore all longcall specifications
15315 when the linker is known to generate glue.
15317 @item -mtls-markers
15318 @itemx -mno-tls-markers
15319 @opindex mtls-markers
15320 @opindex mno-tls-markers
15321 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15322 specifying the function argument. The relocation allows ld to
15323 reliably associate function call with argument setup instructions for
15324 TLS optimization, which in turn allows gcc to better schedule the
15329 Adds support for multithreading with the @dfn{pthreads} library.
15330 This option sets flags for both the preprocessor and linker.
15334 @node S/390 and zSeries Options
15335 @subsection S/390 and zSeries Options
15336 @cindex S/390 and zSeries Options
15338 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15342 @itemx -msoft-float
15343 @opindex mhard-float
15344 @opindex msoft-float
15345 Use (do not use) the hardware floating-point instructions and registers
15346 for floating-point operations. When @option{-msoft-float} is specified,
15347 functions in @file{libgcc.a} will be used to perform floating-point
15348 operations. When @option{-mhard-float} is specified, the compiler
15349 generates IEEE floating-point instructions. This is the default.
15352 @itemx -mno-hard-dfp
15354 @opindex mno-hard-dfp
15355 Use (do not use) the hardware decimal-floating-point instructions for
15356 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15357 specified, functions in @file{libgcc.a} will be used to perform
15358 decimal-floating-point operations. When @option{-mhard-dfp} is
15359 specified, the compiler generates decimal-floating-point hardware
15360 instructions. This is the default for @option{-march=z9-ec} or higher.
15362 @item -mlong-double-64
15363 @itemx -mlong-double-128
15364 @opindex mlong-double-64
15365 @opindex mlong-double-128
15366 These switches control the size of @code{long double} type. A size
15367 of 64bit makes the @code{long double} type equivalent to the @code{double}
15368 type. This is the default.
15371 @itemx -mno-backchain
15372 @opindex mbackchain
15373 @opindex mno-backchain
15374 Store (do not store) the address of the caller's frame as backchain pointer
15375 into the callee's stack frame.
15376 A backchain may be needed to allow debugging using tools that do not understand
15377 DWARF-2 call frame information.
15378 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15379 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15380 the backchain is placed into the topmost word of the 96/160 byte register
15383 In general, code compiled with @option{-mbackchain} is call-compatible with
15384 code compiled with @option{-mmo-backchain}; however, use of the backchain
15385 for debugging purposes usually requires that the whole binary is built with
15386 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15387 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15388 to build a linux kernel use @option{-msoft-float}.
15390 The default is to not maintain the backchain.
15392 @item -mpacked-stack
15393 @itemx -mno-packed-stack
15394 @opindex mpacked-stack
15395 @opindex mno-packed-stack
15396 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15397 specified, the compiler uses the all fields of the 96/160 byte register save
15398 area only for their default purpose; unused fields still take up stack space.
15399 When @option{-mpacked-stack} is specified, register save slots are densely
15400 packed at the top of the register save area; unused space is reused for other
15401 purposes, allowing for more efficient use of the available stack space.
15402 However, when @option{-mbackchain} is also in effect, the topmost word of
15403 the save area is always used to store the backchain, and the return address
15404 register is always saved two words below the backchain.
15406 As long as the stack frame backchain is not used, code generated with
15407 @option{-mpacked-stack} is call-compatible with code generated with
15408 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15409 S/390 or zSeries generated code that uses the stack frame backchain at run
15410 time, not just for debugging purposes. Such code is not call-compatible
15411 with code compiled with @option{-mpacked-stack}. Also, note that the
15412 combination of @option{-mbackchain},
15413 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15414 to build a linux kernel use @option{-msoft-float}.
15416 The default is to not use the packed stack layout.
15419 @itemx -mno-small-exec
15420 @opindex msmall-exec
15421 @opindex mno-small-exec
15422 Generate (or do not generate) code using the @code{bras} instruction
15423 to do subroutine calls.
15424 This only works reliably if the total executable size does not
15425 exceed 64k. The default is to use the @code{basr} instruction instead,
15426 which does not have this limitation.
15432 When @option{-m31} is specified, generate code compliant to the
15433 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15434 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15435 particular to generate 64-bit instructions. For the @samp{s390}
15436 targets, the default is @option{-m31}, while the @samp{s390x}
15437 targets default to @option{-m64}.
15443 When @option{-mzarch} is specified, generate code using the
15444 instructions available on z/Architecture.
15445 When @option{-mesa} is specified, generate code using the
15446 instructions available on ESA/390. Note that @option{-mesa} is
15447 not possible with @option{-m64}.
15448 When generating code compliant to the GNU/Linux for S/390 ABI,
15449 the default is @option{-mesa}. When generating code compliant
15450 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15456 Generate (or do not generate) code using the @code{mvcle} instruction
15457 to perform block moves. When @option{-mno-mvcle} is specified,
15458 use a @code{mvc} loop instead. This is the default unless optimizing for
15465 Print (or do not print) additional debug information when compiling.
15466 The default is to not print debug information.
15468 @item -march=@var{cpu-type}
15470 Generate code that will run on @var{cpu-type}, which is the name of a system
15471 representing a certain processor type. Possible values for
15472 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15473 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15474 When generating code using the instructions available on z/Architecture,
15475 the default is @option{-march=z900}. Otherwise, the default is
15476 @option{-march=g5}.
15478 @item -mtune=@var{cpu-type}
15480 Tune to @var{cpu-type} everything applicable about the generated code,
15481 except for the ABI and the set of available instructions.
15482 The list of @var{cpu-type} values is the same as for @option{-march}.
15483 The default is the value used for @option{-march}.
15486 @itemx -mno-tpf-trace
15487 @opindex mtpf-trace
15488 @opindex mno-tpf-trace
15489 Generate code that adds (does not add) in TPF OS specific branches to trace
15490 routines in the operating system. This option is off by default, even
15491 when compiling for the TPF OS@.
15494 @itemx -mno-fused-madd
15495 @opindex mfused-madd
15496 @opindex mno-fused-madd
15497 Generate code that uses (does not use) the floating point multiply and
15498 accumulate instructions. These instructions are generated by default if
15499 hardware floating point is used.
15501 @item -mwarn-framesize=@var{framesize}
15502 @opindex mwarn-framesize
15503 Emit a warning if the current function exceeds the given frame size. Because
15504 this is a compile time check it doesn't need to be a real problem when the program
15505 runs. It is intended to identify functions which most probably cause
15506 a stack overflow. It is useful to be used in an environment with limited stack
15507 size e.g.@: the linux kernel.
15509 @item -mwarn-dynamicstack
15510 @opindex mwarn-dynamicstack
15511 Emit a warning if the function calls alloca or uses dynamically
15512 sized arrays. This is generally a bad idea with a limited stack size.
15514 @item -mstack-guard=@var{stack-guard}
15515 @itemx -mstack-size=@var{stack-size}
15516 @opindex mstack-guard
15517 @opindex mstack-size
15518 If these options are provided the s390 back end emits additional instructions in
15519 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15520 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15521 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15522 the frame size of the compiled function is chosen.
15523 These options are intended to be used to help debugging stack overflow problems.
15524 The additionally emitted code causes only little overhead and hence can also be
15525 used in production like systems without greater performance degradation. The given
15526 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15527 @var{stack-guard} without exceeding 64k.
15528 In order to be efficient the extra code makes the assumption that the stack starts
15529 at an address aligned to the value given by @var{stack-size}.
15530 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15533 @node Score Options
15534 @subsection Score Options
15535 @cindex Score Options
15537 These options are defined for Score implementations:
15542 Compile code for big endian mode. This is the default.
15546 Compile code for little endian mode.
15550 Disable generate bcnz instruction.
15554 Enable generate unaligned load and store instruction.
15558 Enable the use of multiply-accumulate instructions. Disabled by default.
15562 Specify the SCORE5 as the target architecture.
15566 Specify the SCORE5U of the target architecture.
15570 Specify the SCORE7 as the target architecture. This is the default.
15574 Specify the SCORE7D as the target architecture.
15578 @subsection SH Options
15580 These @samp{-m} options are defined for the SH implementations:
15585 Generate code for the SH1.
15589 Generate code for the SH2.
15592 Generate code for the SH2e.
15596 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15597 that the floating-point unit is not used.
15599 @item -m2a-single-only
15600 @opindex m2a-single-only
15601 Generate code for the SH2a-FPU, in such a way that no double-precision
15602 floating point operations are used.
15605 @opindex m2a-single
15606 Generate code for the SH2a-FPU assuming the floating-point unit is in
15607 single-precision mode by default.
15611 Generate code for the SH2a-FPU assuming the floating-point unit is in
15612 double-precision mode by default.
15616 Generate code for the SH3.
15620 Generate code for the SH3e.
15624 Generate code for the SH4 without a floating-point unit.
15626 @item -m4-single-only
15627 @opindex m4-single-only
15628 Generate code for the SH4 with a floating-point unit that only
15629 supports single-precision arithmetic.
15633 Generate code for the SH4 assuming the floating-point unit is in
15634 single-precision mode by default.
15638 Generate code for the SH4.
15642 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15643 floating-point unit is not used.
15645 @item -m4a-single-only
15646 @opindex m4a-single-only
15647 Generate code for the SH4a, in such a way that no double-precision
15648 floating point operations are used.
15651 @opindex m4a-single
15652 Generate code for the SH4a assuming the floating-point unit is in
15653 single-precision mode by default.
15657 Generate code for the SH4a.
15661 Same as @option{-m4a-nofpu}, except that it implicitly passes
15662 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15663 instructions at the moment.
15667 Compile code for the processor in big endian mode.
15671 Compile code for the processor in little endian mode.
15675 Align doubles at 64-bit boundaries. Note that this changes the calling
15676 conventions, and thus some functions from the standard C library will
15677 not work unless you recompile it first with @option{-mdalign}.
15681 Shorten some address references at link time, when possible; uses the
15682 linker option @option{-relax}.
15686 Use 32-bit offsets in @code{switch} tables. The default is to use
15691 Enable the use of bit manipulation instructions on SH2A.
15695 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15696 alignment constraints.
15700 Comply with the calling conventions defined by Renesas.
15704 Comply with the calling conventions defined by Renesas.
15708 Comply with the calling conventions defined for GCC before the Renesas
15709 conventions were available. This option is the default for all
15710 targets of the SH toolchain except for @samp{sh-symbianelf}.
15713 @opindex mnomacsave
15714 Mark the @code{MAC} register as call-clobbered, even if
15715 @option{-mhitachi} is given.
15719 Increase IEEE-compliance of floating-point code.
15720 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15721 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15722 comparisons of NANs / infinities incurs extra overhead in every
15723 floating point comparison, therefore the default is set to
15724 @option{-ffinite-math-only}.
15726 @item -minline-ic_invalidate
15727 @opindex minline-ic_invalidate
15728 Inline code to invalidate instruction cache entries after setting up
15729 nested function trampolines.
15730 This option has no effect if -musermode is in effect and the selected
15731 code generation option (e.g. -m4) does not allow the use of the icbi
15733 If the selected code generation option does not allow the use of the icbi
15734 instruction, and -musermode is not in effect, the inlined code will
15735 manipulate the instruction cache address array directly with an associative
15736 write. This not only requires privileged mode, but it will also
15737 fail if the cache line had been mapped via the TLB and has become unmapped.
15741 Dump instruction size and location in the assembly code.
15744 @opindex mpadstruct
15745 This option is deprecated. It pads structures to multiple of 4 bytes,
15746 which is incompatible with the SH ABI@.
15750 Optimize for space instead of speed. Implied by @option{-Os}.
15753 @opindex mprefergot
15754 When generating position-independent code, emit function calls using
15755 the Global Offset Table instead of the Procedure Linkage Table.
15759 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15760 if the inlined code would not work in user mode.
15761 This is the default when the target is @code{sh-*-linux*}.
15763 @item -multcost=@var{number}
15764 @opindex multcost=@var{number}
15765 Set the cost to assume for a multiply insn.
15767 @item -mdiv=@var{strategy}
15768 @opindex mdiv=@var{strategy}
15769 Set the division strategy to use for SHmedia code. @var{strategy} must be
15770 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15771 inv:call2, inv:fp .
15772 "fp" performs the operation in floating point. This has a very high latency,
15773 but needs only a few instructions, so it might be a good choice if
15774 your code has enough easily exploitable ILP to allow the compiler to
15775 schedule the floating point instructions together with other instructions.
15776 Division by zero causes a floating point exception.
15777 "inv" uses integer operations to calculate the inverse of the divisor,
15778 and then multiplies the dividend with the inverse. This strategy allows
15779 cse and hoisting of the inverse calculation. Division by zero calculates
15780 an unspecified result, but does not trap.
15781 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15782 have been found, or if the entire operation has been hoisted to the same
15783 place, the last stages of the inverse calculation are intertwined with the
15784 final multiply to reduce the overall latency, at the expense of using a few
15785 more instructions, and thus offering fewer scheduling opportunities with
15787 "call" calls a library function that usually implements the inv:minlat
15789 This gives high code density for m5-*media-nofpu compilations.
15790 "call2" uses a different entry point of the same library function, where it
15791 assumes that a pointer to a lookup table has already been set up, which
15792 exposes the pointer load to cse / code hoisting optimizations.
15793 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15794 code generation, but if the code stays unoptimized, revert to the "call",
15795 "call2", or "fp" strategies, respectively. Note that the
15796 potentially-trapping side effect of division by zero is carried by a
15797 separate instruction, so it is possible that all the integer instructions
15798 are hoisted out, but the marker for the side effect stays where it is.
15799 A recombination to fp operations or a call is not possible in that case.
15800 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15801 that the inverse calculation was nor separated from the multiply, they speed
15802 up division where the dividend fits into 20 bits (plus sign where applicable),
15803 by inserting a test to skip a number of operations in this case; this test
15804 slows down the case of larger dividends. inv20u assumes the case of a such
15805 a small dividend to be unlikely, and inv20l assumes it to be likely.
15807 @item -mdivsi3_libfunc=@var{name}
15808 @opindex mdivsi3_libfunc=@var{name}
15809 Set the name of the library function used for 32 bit signed division to
15810 @var{name}. This only affect the name used in the call and inv:call
15811 division strategies, and the compiler will still expect the same
15812 sets of input/output/clobbered registers as if this option was not present.
15814 @item -mfixed-range=@var{register-range}
15815 @opindex mfixed-range
15816 Generate code treating the given register range as fixed registers.
15817 A fixed register is one that the register allocator can not use. This is
15818 useful when compiling kernel code. A register range is specified as
15819 two registers separated by a dash. Multiple register ranges can be
15820 specified separated by a comma.
15822 @item -madjust-unroll
15823 @opindex madjust-unroll
15824 Throttle unrolling to avoid thrashing target registers.
15825 This option only has an effect if the gcc code base supports the
15826 TARGET_ADJUST_UNROLL_MAX target hook.
15828 @item -mindexed-addressing
15829 @opindex mindexed-addressing
15830 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15831 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15832 semantics for the indexed addressing mode. The architecture allows the
15833 implementation of processors with 64 bit MMU, which the OS could use to
15834 get 32 bit addressing, but since no current hardware implementation supports
15835 this or any other way to make the indexed addressing mode safe to use in
15836 the 32 bit ABI, the default is -mno-indexed-addressing.
15838 @item -mgettrcost=@var{number}
15839 @opindex mgettrcost=@var{number}
15840 Set the cost assumed for the gettr instruction to @var{number}.
15841 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15845 Assume pt* instructions won't trap. This will generally generate better
15846 scheduled code, but is unsafe on current hardware. The current architecture
15847 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15848 This has the unintentional effect of making it unsafe to schedule ptabs /
15849 ptrel before a branch, or hoist it out of a loop. For example,
15850 __do_global_ctors, a part of libgcc that runs constructors at program
15851 startup, calls functions in a list which is delimited by @minus{}1. With the
15852 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15853 That means that all the constructors will be run a bit quicker, but when
15854 the loop comes to the end of the list, the program crashes because ptabs
15855 loads @minus{}1 into a target register. Since this option is unsafe for any
15856 hardware implementing the current architecture specification, the default
15857 is -mno-pt-fixed. Unless the user specifies a specific cost with
15858 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15859 this deters register allocation using target registers for storing
15862 @item -minvalid-symbols
15863 @opindex minvalid-symbols
15864 Assume symbols might be invalid. Ordinary function symbols generated by
15865 the compiler will always be valid to load with movi/shori/ptabs or
15866 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15867 to generate symbols that will cause ptabs / ptrel to trap.
15868 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15869 It will then prevent cross-basic-block cse, hoisting and most scheduling
15870 of symbol loads. The default is @option{-mno-invalid-symbols}.
15873 @node SPARC Options
15874 @subsection SPARC Options
15875 @cindex SPARC options
15877 These @samp{-m} options are supported on the SPARC:
15880 @item -mno-app-regs
15882 @opindex mno-app-regs
15884 Specify @option{-mapp-regs} to generate output using the global registers
15885 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15888 To be fully SVR4 ABI compliant at the cost of some performance loss,
15889 specify @option{-mno-app-regs}. You should compile libraries and system
15890 software with this option.
15893 @itemx -mhard-float
15895 @opindex mhard-float
15896 Generate output containing floating point instructions. This is the
15900 @itemx -msoft-float
15902 @opindex msoft-float
15903 Generate output containing library calls for floating point.
15904 @strong{Warning:} the requisite libraries are not available for all SPARC
15905 targets. Normally the facilities of the machine's usual C compiler are
15906 used, but this cannot be done directly in cross-compilation. You must make
15907 your own arrangements to provide suitable library functions for
15908 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15909 @samp{sparclite-*-*} do provide software floating point support.
15911 @option{-msoft-float} changes the calling convention in the output file;
15912 therefore, it is only useful if you compile @emph{all} of a program with
15913 this option. In particular, you need to compile @file{libgcc.a}, the
15914 library that comes with GCC, with @option{-msoft-float} in order for
15917 @item -mhard-quad-float
15918 @opindex mhard-quad-float
15919 Generate output containing quad-word (long double) floating point
15922 @item -msoft-quad-float
15923 @opindex msoft-quad-float
15924 Generate output containing library calls for quad-word (long double)
15925 floating point instructions. The functions called are those specified
15926 in the SPARC ABI@. This is the default.
15928 As of this writing, there are no SPARC implementations that have hardware
15929 support for the quad-word floating point instructions. They all invoke
15930 a trap handler for one of these instructions, and then the trap handler
15931 emulates the effect of the instruction. Because of the trap handler overhead,
15932 this is much slower than calling the ABI library routines. Thus the
15933 @option{-msoft-quad-float} option is the default.
15935 @item -mno-unaligned-doubles
15936 @itemx -munaligned-doubles
15937 @opindex mno-unaligned-doubles
15938 @opindex munaligned-doubles
15939 Assume that doubles have 8 byte alignment. This is the default.
15941 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15942 alignment only if they are contained in another type, or if they have an
15943 absolute address. Otherwise, it assumes they have 4 byte alignment.
15944 Specifying this option avoids some rare compatibility problems with code
15945 generated by other compilers. It is not the default because it results
15946 in a performance loss, especially for floating point code.
15948 @item -mno-faster-structs
15949 @itemx -mfaster-structs
15950 @opindex mno-faster-structs
15951 @opindex mfaster-structs
15952 With @option{-mfaster-structs}, the compiler assumes that structures
15953 should have 8 byte alignment. This enables the use of pairs of
15954 @code{ldd} and @code{std} instructions for copies in structure
15955 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15956 However, the use of this changed alignment directly violates the SPARC
15957 ABI@. Thus, it's intended only for use on targets where the developer
15958 acknowledges that their resulting code will not be directly in line with
15959 the rules of the ABI@.
15961 @item -mimpure-text
15962 @opindex mimpure-text
15963 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15964 the compiler to not pass @option{-z text} to the linker when linking a
15965 shared object. Using this option, you can link position-dependent
15966 code into a shared object.
15968 @option{-mimpure-text} suppresses the ``relocations remain against
15969 allocatable but non-writable sections'' linker error message.
15970 However, the necessary relocations will trigger copy-on-write, and the
15971 shared object is not actually shared across processes. Instead of
15972 using @option{-mimpure-text}, you should compile all source code with
15973 @option{-fpic} or @option{-fPIC}.
15975 This option is only available on SunOS and Solaris.
15977 @item -mcpu=@var{cpu_type}
15979 Set the instruction set, register set, and instruction scheduling parameters
15980 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15981 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15982 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15983 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15984 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15986 Default instruction scheduling parameters are used for values that select
15987 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15988 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15990 Here is a list of each supported architecture and their supported
15995 v8: supersparc, hypersparc
15996 sparclite: f930, f934, sparclite86x
15998 v9: ultrasparc, ultrasparc3, niagara, niagara2
16001 By default (unless configured otherwise), GCC generates code for the V7
16002 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16003 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16004 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16005 SPARCStation 1, 2, IPX etc.
16007 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16008 architecture. The only difference from V7 code is that the compiler emits
16009 the integer multiply and integer divide instructions which exist in SPARC-V8
16010 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16011 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16014 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16015 the SPARC architecture. This adds the integer multiply, integer divide step
16016 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16017 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16018 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16019 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16020 MB86934 chip, which is the more recent SPARClite with FPU@.
16022 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16023 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16024 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16025 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16026 optimizes it for the TEMIC SPARClet chip.
16028 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16029 architecture. This adds 64-bit integer and floating-point move instructions,
16030 3 additional floating-point condition code registers and conditional move
16031 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16032 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16033 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16034 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16035 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16036 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16037 additionally optimizes it for Sun UltraSPARC T2 chips.
16039 @item -mtune=@var{cpu_type}
16041 Set the instruction scheduling parameters for machine type
16042 @var{cpu_type}, but do not set the instruction set or register set that the
16043 option @option{-mcpu=@var{cpu_type}} would.
16045 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16046 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16047 that select a particular cpu implementation. Those are @samp{cypress},
16048 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16049 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16050 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16055 @opindex mno-v8plus
16056 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16057 difference from the V8 ABI is that the global and out registers are
16058 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16059 mode for all SPARC-V9 processors.
16065 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16066 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16069 These @samp{-m} options are supported in addition to the above
16070 on SPARC-V9 processors in 64-bit environments:
16073 @item -mlittle-endian
16074 @opindex mlittle-endian
16075 Generate code for a processor running in little-endian mode. It is only
16076 available for a few configurations and most notably not on Solaris and Linux.
16082 Generate code for a 32-bit or 64-bit environment.
16083 The 32-bit environment sets int, long and pointer to 32 bits.
16084 The 64-bit environment sets int to 32 bits and long and pointer
16087 @item -mcmodel=medlow
16088 @opindex mcmodel=medlow
16089 Generate code for the Medium/Low code model: 64-bit addresses, programs
16090 must be linked in the low 32 bits of memory. Programs can be statically
16091 or dynamically linked.
16093 @item -mcmodel=medmid
16094 @opindex mcmodel=medmid
16095 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16096 must be linked in the low 44 bits of memory, the text and data segments must
16097 be less than 2GB in size and the data segment must be located within 2GB of
16100 @item -mcmodel=medany
16101 @opindex mcmodel=medany
16102 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16103 may be linked anywhere in memory, the text and data segments must be less
16104 than 2GB in size and the data segment must be located within 2GB of the
16107 @item -mcmodel=embmedany
16108 @opindex mcmodel=embmedany
16109 Generate code for the Medium/Anywhere code model for embedded systems:
16110 64-bit addresses, the text and data segments must be less than 2GB in
16111 size, both starting anywhere in memory (determined at link time). The
16112 global register %g4 points to the base of the data segment. Programs
16113 are statically linked and PIC is not supported.
16116 @itemx -mno-stack-bias
16117 @opindex mstack-bias
16118 @opindex mno-stack-bias
16119 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16120 frame pointer if present, are offset by @minus{}2047 which must be added back
16121 when making stack frame references. This is the default in 64-bit mode.
16122 Otherwise, assume no such offset is present.
16125 These switches are supported in addition to the above on Solaris:
16130 Add support for multithreading using the Solaris threads library. This
16131 option sets flags for both the preprocessor and linker. This option does
16132 not affect the thread safety of object code produced by the compiler or
16133 that of libraries supplied with it.
16137 Add support for multithreading using the POSIX threads library. This
16138 option sets flags for both the preprocessor and linker. This option does
16139 not affect the thread safety of object code produced by the compiler or
16140 that of libraries supplied with it.
16144 This is a synonym for @option{-pthreads}.
16148 @subsection SPU Options
16149 @cindex SPU options
16151 These @samp{-m} options are supported on the SPU:
16155 @itemx -merror-reloc
16156 @opindex mwarn-reloc
16157 @opindex merror-reloc
16159 The loader for SPU does not handle dynamic relocations. By default, GCC
16160 will give an error when it generates code that requires a dynamic
16161 relocation. @option{-mno-error-reloc} disables the error,
16162 @option{-mwarn-reloc} will generate a warning instead.
16165 @itemx -munsafe-dma
16167 @opindex munsafe-dma
16169 Instructions which initiate or test completion of DMA must not be
16170 reordered with respect to loads and stores of the memory which is being
16171 accessed. Users typically address this problem using the volatile
16172 keyword, but that can lead to inefficient code in places where the
16173 memory is known to not change. Rather than mark the memory as volatile
16174 we treat the DMA instructions as potentially effecting all memory. With
16175 @option{-munsafe-dma} users must use the volatile keyword to protect
16178 @item -mbranch-hints
16179 @opindex mbranch-hints
16181 By default, GCC will generate a branch hint instruction to avoid
16182 pipeline stalls for always taken or probably taken branches. A hint
16183 will not be generated closer than 8 instructions away from its branch.
16184 There is little reason to disable them, except for debugging purposes,
16185 or to make an object a little bit smaller.
16189 @opindex msmall-mem
16190 @opindex mlarge-mem
16192 By default, GCC generates code assuming that addresses are never larger
16193 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16194 a full 32 bit address.
16199 By default, GCC links against startup code that assumes the SPU-style
16200 main function interface (which has an unconventional parameter list).
16201 With @option{-mstdmain}, GCC will link your program against startup
16202 code that assumes a C99-style interface to @code{main}, including a
16203 local copy of @code{argv} strings.
16205 @item -mfixed-range=@var{register-range}
16206 @opindex mfixed-range
16207 Generate code treating the given register range as fixed registers.
16208 A fixed register is one that the register allocator can not use. This is
16209 useful when compiling kernel code. A register range is specified as
16210 two registers separated by a dash. Multiple register ranges can be
16211 specified separated by a comma.
16214 @itemx -mdual-nops=@var{n}
16215 @opindex mdual-nops
16216 By default, GCC will insert nops to increase dual issue when it expects
16217 it to increase performance. @var{n} can be a value from 0 to 10. A
16218 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16219 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16221 @item -mhint-max-nops=@var{n}
16222 @opindex mhint-max-nops
16223 Maximum number of nops to insert for a branch hint. A branch hint must
16224 be at least 8 instructions away from the branch it is effecting. GCC
16225 will insert up to @var{n} nops to enforce this, otherwise it will not
16226 generate the branch hint.
16228 @item -mhint-max-distance=@var{n}
16229 @opindex mhint-max-distance
16230 The encoding of the branch hint instruction limits the hint to be within
16231 256 instructions of the branch it is effecting. By default, GCC makes
16232 sure it is within 125.
16235 @opindex msafe-hints
16236 Work around a hardware bug which causes the SPU to stall indefinitely.
16237 By default, GCC will insert the @code{hbrp} instruction to make sure
16238 this stall won't happen.
16242 @node System V Options
16243 @subsection Options for System V
16245 These additional options are available on System V Release 4 for
16246 compatibility with other compilers on those systems:
16251 Create a shared object.
16252 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16256 Identify the versions of each tool used by the compiler, in a
16257 @code{.ident} assembler directive in the output.
16261 Refrain from adding @code{.ident} directives to the output file (this is
16264 @item -YP,@var{dirs}
16266 Search the directories @var{dirs}, and no others, for libraries
16267 specified with @option{-l}.
16269 @item -Ym,@var{dir}
16271 Look in the directory @var{dir} to find the M4 preprocessor.
16272 The assembler uses this option.
16273 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16274 @c the generic assembler that comes with Solaris takes just -Ym.
16278 @subsection V850 Options
16279 @cindex V850 Options
16281 These @samp{-m} options are defined for V850 implementations:
16285 @itemx -mno-long-calls
16286 @opindex mlong-calls
16287 @opindex mno-long-calls
16288 Treat all calls as being far away (near). If calls are assumed to be
16289 far away, the compiler will always load the functions address up into a
16290 register, and call indirect through the pointer.
16296 Do not optimize (do optimize) basic blocks that use the same index
16297 pointer 4 or more times to copy pointer into the @code{ep} register, and
16298 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16299 option is on by default if you optimize.
16301 @item -mno-prolog-function
16302 @itemx -mprolog-function
16303 @opindex mno-prolog-function
16304 @opindex mprolog-function
16305 Do not use (do use) external functions to save and restore registers
16306 at the prologue and epilogue of a function. The external functions
16307 are slower, but use less code space if more than one function saves
16308 the same number of registers. The @option{-mprolog-function} option
16309 is on by default if you optimize.
16313 Try to make the code as small as possible. At present, this just turns
16314 on the @option{-mep} and @option{-mprolog-function} options.
16316 @item -mtda=@var{n}
16318 Put static or global variables whose size is @var{n} bytes or less into
16319 the tiny data area that register @code{ep} points to. The tiny data
16320 area can hold up to 256 bytes in total (128 bytes for byte references).
16322 @item -msda=@var{n}
16324 Put static or global variables whose size is @var{n} bytes or less into
16325 the small data area that register @code{gp} points to. The small data
16326 area can hold up to 64 kilobytes.
16328 @item -mzda=@var{n}
16330 Put static or global variables whose size is @var{n} bytes or less into
16331 the first 32 kilobytes of memory.
16335 Specify that the target processor is the V850.
16338 @opindex mbig-switch
16339 Generate code suitable for big switch tables. Use this option only if
16340 the assembler/linker complain about out of range branches within a switch
16345 This option will cause r2 and r5 to be used in the code generated by
16346 the compiler. This setting is the default.
16348 @item -mno-app-regs
16349 @opindex mno-app-regs
16350 This option will cause r2 and r5 to be treated as fixed registers.
16354 Specify that the target processor is the V850E1. The preprocessor
16355 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16356 this option is used.
16360 Specify that the target processor is the V850E@. The preprocessor
16361 constant @samp{__v850e__} will be defined if this option is used.
16363 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16364 are defined then a default target processor will be chosen and the
16365 relevant @samp{__v850*__} preprocessor constant will be defined.
16367 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16368 defined, regardless of which processor variant is the target.
16370 @item -mdisable-callt
16371 @opindex mdisable-callt
16372 This option will suppress generation of the CALLT instruction for the
16373 v850e and v850e1 flavors of the v850 architecture. The default is
16374 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16379 @subsection VAX Options
16380 @cindex VAX options
16382 These @samp{-m} options are defined for the VAX:
16387 Do not output certain jump instructions (@code{aobleq} and so on)
16388 that the Unix assembler for the VAX cannot handle across long
16393 Do output those jump instructions, on the assumption that you
16394 will assemble with the GNU assembler.
16398 Output code for g-format floating point numbers instead of d-format.
16401 @node VxWorks Options
16402 @subsection VxWorks Options
16403 @cindex VxWorks Options
16405 The options in this section are defined for all VxWorks targets.
16406 Options specific to the target hardware are listed with the other
16407 options for that target.
16412 GCC can generate code for both VxWorks kernels and real time processes
16413 (RTPs). This option switches from the former to the latter. It also
16414 defines the preprocessor macro @code{__RTP__}.
16417 @opindex non-static
16418 Link an RTP executable against shared libraries rather than static
16419 libraries. The options @option{-static} and @option{-shared} can
16420 also be used for RTPs (@pxref{Link Options}); @option{-static}
16427 These options are passed down to the linker. They are defined for
16428 compatibility with Diab.
16431 @opindex Xbind-lazy
16432 Enable lazy binding of function calls. This option is equivalent to
16433 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16437 Disable lazy binding of function calls. This option is the default and
16438 is defined for compatibility with Diab.
16441 @node x86-64 Options
16442 @subsection x86-64 Options
16443 @cindex x86-64 options
16445 These are listed under @xref{i386 and x86-64 Options}.
16447 @node i386 and x86-64 Windows Options
16448 @subsection i386 and x86-64 Windows Options
16449 @cindex i386 and x86-64 Windows Options
16451 These additional options are available for Windows targets:
16456 This option is available for Cygwin and MinGW targets. It
16457 specifies that a console application is to be generated, by
16458 instructing the linker to set the PE header subsystem type
16459 required for console applications.
16460 This is the default behavior for Cygwin and MinGW targets.
16464 This option is available for Cygwin targets. It specifies that
16465 the Cygwin internal interface is to be used for predefined
16466 preprocessor macros, C runtime libraries and related linker
16467 paths and options. For Cygwin targets this is the default behavior.
16468 This option is deprecated and will be removed in a future release.
16471 @opindex mno-cygwin
16472 This option is available for Cygwin targets. It specifies that
16473 the MinGW internal interface is to be used instead of Cygwin's, by
16474 setting MinGW-related predefined macros and linker paths and default
16476 This option is deprecated and will be removed in a future release.
16480 This option is available for Cygwin and MinGW targets. It
16481 specifies that a DLL - a dynamic link library - is to be
16482 generated, enabling the selection of the required runtime
16483 startup object and entry point.
16485 @item -mnop-fun-dllimport
16486 @opindex mnop-fun-dllimport
16487 This option is available for Cygwin and MinGW targets. It
16488 specifies that the dllimport attribute should be ignored.
16492 This option is available for MinGW targets. It specifies
16493 that MinGW-specific thread support is to be used.
16497 This option is available for mingw-w64 targets. It specifies
16498 that the UNICODE macro is getting pre-defined and that the
16499 unicode capable runtime startup code is choosen.
16503 This option is available for Cygwin and MinGW targets. It
16504 specifies that the typical Windows pre-defined macros are to
16505 be set in the pre-processor, but does not influence the choice
16506 of runtime library/startup code.
16510 This option is available for Cygwin and MinGW targets. It
16511 specifies that a GUI application is to be generated by
16512 instructing the linker to set the PE header subsystem type
16515 @item -mpe-aligned-commons
16516 @opindex mpe-aligned-commons
16517 This option is available for Cygwin and MinGW targets. It
16518 specifies that the GNU extension to the PE file format that
16519 permits the correct alignment of COMMON variables should be
16520 used when generating code. It will be enabled by default if
16521 GCC detects that the target assembler found during configuration
16522 supports the feature.
16525 See also under @ref{i386 and x86-64 Options} for standard options.
16527 @node Xstormy16 Options
16528 @subsection Xstormy16 Options
16529 @cindex Xstormy16 Options
16531 These options are defined for Xstormy16:
16536 Choose startup files and linker script suitable for the simulator.
16539 @node Xtensa Options
16540 @subsection Xtensa Options
16541 @cindex Xtensa Options
16543 These options are supported for Xtensa targets:
16547 @itemx -mno-const16
16549 @opindex mno-const16
16550 Enable or disable use of @code{CONST16} instructions for loading
16551 constant values. The @code{CONST16} instruction is currently not a
16552 standard option from Tensilica. When enabled, @code{CONST16}
16553 instructions are always used in place of the standard @code{L32R}
16554 instructions. The use of @code{CONST16} is enabled by default only if
16555 the @code{L32R} instruction is not available.
16558 @itemx -mno-fused-madd
16559 @opindex mfused-madd
16560 @opindex mno-fused-madd
16561 Enable or disable use of fused multiply/add and multiply/subtract
16562 instructions in the floating-point option. This has no effect if the
16563 floating-point option is not also enabled. Disabling fused multiply/add
16564 and multiply/subtract instructions forces the compiler to use separate
16565 instructions for the multiply and add/subtract operations. This may be
16566 desirable in some cases where strict IEEE 754-compliant results are
16567 required: the fused multiply add/subtract instructions do not round the
16568 intermediate result, thereby producing results with @emph{more} bits of
16569 precision than specified by the IEEE standard. Disabling fused multiply
16570 add/subtract instructions also ensures that the program output is not
16571 sensitive to the compiler's ability to combine multiply and add/subtract
16574 @item -mserialize-volatile
16575 @itemx -mno-serialize-volatile
16576 @opindex mserialize-volatile
16577 @opindex mno-serialize-volatile
16578 When this option is enabled, GCC inserts @code{MEMW} instructions before
16579 @code{volatile} memory references to guarantee sequential consistency.
16580 The default is @option{-mserialize-volatile}. Use
16581 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16583 @item -mtext-section-literals
16584 @itemx -mno-text-section-literals
16585 @opindex mtext-section-literals
16586 @opindex mno-text-section-literals
16587 Control the treatment of literal pools. The default is
16588 @option{-mno-text-section-literals}, which places literals in a separate
16589 section in the output file. This allows the literal pool to be placed
16590 in a data RAM/ROM, and it also allows the linker to combine literal
16591 pools from separate object files to remove redundant literals and
16592 improve code size. With @option{-mtext-section-literals}, the literals
16593 are interspersed in the text section in order to keep them as close as
16594 possible to their references. This may be necessary for large assembly
16597 @item -mtarget-align
16598 @itemx -mno-target-align
16599 @opindex mtarget-align
16600 @opindex mno-target-align
16601 When this option is enabled, GCC instructs the assembler to
16602 automatically align instructions to reduce branch penalties at the
16603 expense of some code density. The assembler attempts to widen density
16604 instructions to align branch targets and the instructions following call
16605 instructions. If there are not enough preceding safe density
16606 instructions to align a target, no widening will be performed. The
16607 default is @option{-mtarget-align}. These options do not affect the
16608 treatment of auto-aligned instructions like @code{LOOP}, which the
16609 assembler will always align, either by widening density instructions or
16610 by inserting no-op instructions.
16613 @itemx -mno-longcalls
16614 @opindex mlongcalls
16615 @opindex mno-longcalls
16616 When this option is enabled, GCC instructs the assembler to translate
16617 direct calls to indirect calls unless it can determine that the target
16618 of a direct call is in the range allowed by the call instruction. This
16619 translation typically occurs for calls to functions in other source
16620 files. Specifically, the assembler translates a direct @code{CALL}
16621 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16622 The default is @option{-mno-longcalls}. This option should be used in
16623 programs where the call target can potentially be out of range. This
16624 option is implemented in the assembler, not the compiler, so the
16625 assembly code generated by GCC will still show direct call
16626 instructions---look at the disassembled object code to see the actual
16627 instructions. Note that the assembler will use an indirect call for
16628 every cross-file call, not just those that really will be out of range.
16631 @node zSeries Options
16632 @subsection zSeries Options
16633 @cindex zSeries options
16635 These are listed under @xref{S/390 and zSeries Options}.
16637 @node Code Gen Options
16638 @section Options for Code Generation Conventions
16639 @cindex code generation conventions
16640 @cindex options, code generation
16641 @cindex run-time options
16643 These machine-independent options control the interface conventions
16644 used in code generation.
16646 Most of them have both positive and negative forms; the negative form
16647 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16648 one of the forms is listed---the one which is not the default. You
16649 can figure out the other form by either removing @samp{no-} or adding
16653 @item -fbounds-check
16654 @opindex fbounds-check
16655 For front-ends that support it, generate additional code to check that
16656 indices used to access arrays are within the declared range. This is
16657 currently only supported by the Java and Fortran front-ends, where
16658 this option defaults to true and false respectively.
16662 This option generates traps for signed overflow on addition, subtraction,
16663 multiplication operations.
16667 This option instructs the compiler to assume that signed arithmetic
16668 overflow of addition, subtraction and multiplication wraps around
16669 using twos-complement representation. This flag enables some optimizations
16670 and disables others. This option is enabled by default for the Java
16671 front-end, as required by the Java language specification.
16674 @opindex fexceptions
16675 Enable exception handling. Generates extra code needed to propagate
16676 exceptions. For some targets, this implies GCC will generate frame
16677 unwind information for all functions, which can produce significant data
16678 size overhead, although it does not affect execution. If you do not
16679 specify this option, GCC will enable it by default for languages like
16680 C++ which normally require exception handling, and disable it for
16681 languages like C that do not normally require it. However, you may need
16682 to enable this option when compiling C code that needs to interoperate
16683 properly with exception handlers written in C++. You may also wish to
16684 disable this option if you are compiling older C++ programs that don't
16685 use exception handling.
16687 @item -fnon-call-exceptions
16688 @opindex fnon-call-exceptions
16689 Generate code that allows trapping instructions to throw exceptions.
16690 Note that this requires platform-specific runtime support that does
16691 not exist everywhere. Moreover, it only allows @emph{trapping}
16692 instructions to throw exceptions, i.e.@: memory references or floating
16693 point instructions. It does not allow exceptions to be thrown from
16694 arbitrary signal handlers such as @code{SIGALRM}.
16696 @item -funwind-tables
16697 @opindex funwind-tables
16698 Similar to @option{-fexceptions}, except that it will just generate any needed
16699 static data, but will not affect the generated code in any other way.
16700 You will normally not enable this option; instead, a language processor
16701 that needs this handling would enable it on your behalf.
16703 @item -fasynchronous-unwind-tables
16704 @opindex fasynchronous-unwind-tables
16705 Generate unwind table in dwarf2 format, if supported by target machine. The
16706 table is exact at each instruction boundary, so it can be used for stack
16707 unwinding from asynchronous events (such as debugger or garbage collector).
16709 @item -fpcc-struct-return
16710 @opindex fpcc-struct-return
16711 Return ``short'' @code{struct} and @code{union} values in memory like
16712 longer ones, rather than in registers. This convention is less
16713 efficient, but it has the advantage of allowing intercallability between
16714 GCC-compiled files and files compiled with other compilers, particularly
16715 the Portable C Compiler (pcc).
16717 The precise convention for returning structures in memory depends
16718 on the target configuration macros.
16720 Short structures and unions are those whose size and alignment match
16721 that of some integer type.
16723 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16724 switch is not binary compatible with code compiled with the
16725 @option{-freg-struct-return} switch.
16726 Use it to conform to a non-default application binary interface.
16728 @item -freg-struct-return
16729 @opindex freg-struct-return
16730 Return @code{struct} and @code{union} values in registers when possible.
16731 This is more efficient for small structures than
16732 @option{-fpcc-struct-return}.
16734 If you specify neither @option{-fpcc-struct-return} nor
16735 @option{-freg-struct-return}, GCC defaults to whichever convention is
16736 standard for the target. If there is no standard convention, GCC
16737 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16738 the principal compiler. In those cases, we can choose the standard, and
16739 we chose the more efficient register return alternative.
16741 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16742 switch is not binary compatible with code compiled with the
16743 @option{-fpcc-struct-return} switch.
16744 Use it to conform to a non-default application binary interface.
16746 @item -fshort-enums
16747 @opindex fshort-enums
16748 Allocate to an @code{enum} type only as many bytes as it needs for the
16749 declared range of possible values. Specifically, the @code{enum} type
16750 will be equivalent to the smallest integer type which has enough room.
16752 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16753 code that is not binary compatible with code generated without that switch.
16754 Use it to conform to a non-default application binary interface.
16756 @item -fshort-double
16757 @opindex fshort-double
16758 Use the same size for @code{double} as for @code{float}.
16760 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16761 code that is not binary compatible with code generated without that switch.
16762 Use it to conform to a non-default application binary interface.
16764 @item -fshort-wchar
16765 @opindex fshort-wchar
16766 Override the underlying type for @samp{wchar_t} to be @samp{short
16767 unsigned int} instead of the default for the target. This option is
16768 useful for building programs to run under WINE@.
16770 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16771 code that is not binary compatible with code generated without that switch.
16772 Use it to conform to a non-default application binary interface.
16775 @opindex fno-common
16776 In C code, controls the placement of uninitialized global variables.
16777 Unix C compilers have traditionally permitted multiple definitions of
16778 such variables in different compilation units by placing the variables
16780 This is the behavior specified by @option{-fcommon}, and is the default
16781 for GCC on most targets.
16782 On the other hand, this behavior is not required by ISO C, and on some
16783 targets may carry a speed or code size penalty on variable references.
16784 The @option{-fno-common} option specifies that the compiler should place
16785 uninitialized global variables in the data section of the object file,
16786 rather than generating them as common blocks.
16787 This has the effect that if the same variable is declared
16788 (without @code{extern}) in two different compilations,
16789 you will get a multiple-definition error when you link them.
16790 In this case, you must compile with @option{-fcommon} instead.
16791 Compiling with @option{-fno-common} is useful on targets for which
16792 it provides better performance, or if you wish to verify that the
16793 program will work on other systems which always treat uninitialized
16794 variable declarations this way.
16798 Ignore the @samp{#ident} directive.
16800 @item -finhibit-size-directive
16801 @opindex finhibit-size-directive
16802 Don't output a @code{.size} assembler directive, or anything else that
16803 would cause trouble if the function is split in the middle, and the
16804 two halves are placed at locations far apart in memory. This option is
16805 used when compiling @file{crtstuff.c}; you should not need to use it
16808 @item -fverbose-asm
16809 @opindex fverbose-asm
16810 Put extra commentary information in the generated assembly code to
16811 make it more readable. This option is generally only of use to those
16812 who actually need to read the generated assembly code (perhaps while
16813 debugging the compiler itself).
16815 @option{-fno-verbose-asm}, the default, causes the
16816 extra information to be omitted and is useful when comparing two assembler
16819 @item -frecord-gcc-switches
16820 @opindex frecord-gcc-switches
16821 This switch causes the command line that was used to invoke the
16822 compiler to be recorded into the object file that is being created.
16823 This switch is only implemented on some targets and the exact format
16824 of the recording is target and binary file format dependent, but it
16825 usually takes the form of a section containing ASCII text. This
16826 switch is related to the @option{-fverbose-asm} switch, but that
16827 switch only records information in the assembler output file as
16828 comments, so it never reaches the object file.
16832 @cindex global offset table
16834 Generate position-independent code (PIC) suitable for use in a shared
16835 library, if supported for the target machine. Such code accesses all
16836 constant addresses through a global offset table (GOT)@. The dynamic
16837 loader resolves the GOT entries when the program starts (the dynamic
16838 loader is not part of GCC; it is part of the operating system). If
16839 the GOT size for the linked executable exceeds a machine-specific
16840 maximum size, you get an error message from the linker indicating that
16841 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16842 instead. (These maximums are 8k on the SPARC and 32k
16843 on the m68k and RS/6000. The 386 has no such limit.)
16845 Position-independent code requires special support, and therefore works
16846 only on certain machines. For the 386, GCC supports PIC for System V
16847 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16848 position-independent.
16850 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16855 If supported for the target machine, emit position-independent code,
16856 suitable for dynamic linking and avoiding any limit on the size of the
16857 global offset table. This option makes a difference on the m68k,
16858 PowerPC and SPARC@.
16860 Position-independent code requires special support, and therefore works
16861 only on certain machines.
16863 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16870 These options are similar to @option{-fpic} and @option{-fPIC}, but
16871 generated position independent code can be only linked into executables.
16872 Usually these options are used when @option{-pie} GCC option will be
16873 used during linking.
16875 @option{-fpie} and @option{-fPIE} both define the macros
16876 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16877 for @option{-fpie} and 2 for @option{-fPIE}.
16879 @item -fno-jump-tables
16880 @opindex fno-jump-tables
16881 Do not use jump tables for switch statements even where it would be
16882 more efficient than other code generation strategies. This option is
16883 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16884 building code which forms part of a dynamic linker and cannot
16885 reference the address of a jump table. On some targets, jump tables
16886 do not require a GOT and this option is not needed.
16888 @item -ffixed-@var{reg}
16890 Treat the register named @var{reg} as a fixed register; generated code
16891 should never refer to it (except perhaps as a stack pointer, frame
16892 pointer or in some other fixed role).
16894 @var{reg} must be the name of a register. The register names accepted
16895 are machine-specific and are defined in the @code{REGISTER_NAMES}
16896 macro in the machine description macro file.
16898 This flag does not have a negative form, because it specifies a
16901 @item -fcall-used-@var{reg}
16902 @opindex fcall-used
16903 Treat the register named @var{reg} as an allocable register that is
16904 clobbered by function calls. It may be allocated for temporaries or
16905 variables that do not live across a call. Functions compiled this way
16906 will not save and restore the register @var{reg}.
16908 It is an error to used this flag with the frame pointer or stack pointer.
16909 Use of this flag for other registers that have fixed pervasive roles in
16910 the machine's execution model will produce disastrous results.
16912 This flag does not have a negative form, because it specifies a
16915 @item -fcall-saved-@var{reg}
16916 @opindex fcall-saved
16917 Treat the register named @var{reg} as an allocable register saved by
16918 functions. It may be allocated even for temporaries or variables that
16919 live across a call. Functions compiled this way will save and restore
16920 the register @var{reg} if they use it.
16922 It is an error to used this flag with the frame pointer or stack pointer.
16923 Use of this flag for other registers that have fixed pervasive roles in
16924 the machine's execution model will produce disastrous results.
16926 A different sort of disaster will result from the use of this flag for
16927 a register in which function values may be returned.
16929 This flag does not have a negative form, because it specifies a
16932 @item -fpack-struct[=@var{n}]
16933 @opindex fpack-struct
16934 Without a value specified, pack all structure members together without
16935 holes. When a value is specified (which must be a small power of two), pack
16936 structure members according to this value, representing the maximum
16937 alignment (that is, objects with default alignment requirements larger than
16938 this will be output potentially unaligned at the next fitting location.
16940 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16941 code that is not binary compatible with code generated without that switch.
16942 Additionally, it makes the code suboptimal.
16943 Use it to conform to a non-default application binary interface.
16945 @item -finstrument-functions
16946 @opindex finstrument-functions
16947 Generate instrumentation calls for entry and exit to functions. Just
16948 after function entry and just before function exit, the following
16949 profiling functions will be called with the address of the current
16950 function and its call site. (On some platforms,
16951 @code{__builtin_return_address} does not work beyond the current
16952 function, so the call site information may not be available to the
16953 profiling functions otherwise.)
16956 void __cyg_profile_func_enter (void *this_fn,
16958 void __cyg_profile_func_exit (void *this_fn,
16962 The first argument is the address of the start of the current function,
16963 which may be looked up exactly in the symbol table.
16965 This instrumentation is also done for functions expanded inline in other
16966 functions. The profiling calls will indicate where, conceptually, the
16967 inline function is entered and exited. This means that addressable
16968 versions of such functions must be available. If all your uses of a
16969 function are expanded inline, this may mean an additional expansion of
16970 code size. If you use @samp{extern inline} in your C code, an
16971 addressable version of such functions must be provided. (This is
16972 normally the case anyways, but if you get lucky and the optimizer always
16973 expands the functions inline, you might have gotten away without
16974 providing static copies.)
16976 A function may be given the attribute @code{no_instrument_function}, in
16977 which case this instrumentation will not be done. This can be used, for
16978 example, for the profiling functions listed above, high-priority
16979 interrupt routines, and any functions from which the profiling functions
16980 cannot safely be called (perhaps signal handlers, if the profiling
16981 routines generate output or allocate memory).
16983 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16984 @opindex finstrument-functions-exclude-file-list
16986 Set the list of functions that are excluded from instrumentation (see
16987 the description of @code{-finstrument-functions}). If the file that
16988 contains a function definition matches with one of @var{file}, then
16989 that function is not instrumented. The match is done on substrings:
16990 if the @var{file} parameter is a substring of the file name, it is
16991 considered to be a match.
16994 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16995 will exclude any inline function defined in files whose pathnames
16996 contain @code{/bits/stl} or @code{include/sys}.
16998 If, for some reason, you want to include letter @code{','} in one of
16999 @var{sym}, write @code{'\,'}. For example,
17000 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17001 (note the single quote surrounding the option).
17003 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17004 @opindex finstrument-functions-exclude-function-list
17006 This is similar to @code{-finstrument-functions-exclude-file-list},
17007 but this option sets the list of function names to be excluded from
17008 instrumentation. The function name to be matched is its user-visible
17009 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17010 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17011 match is done on substrings: if the @var{sym} parameter is a substring
17012 of the function name, it is considered to be a match. For C99 and C++
17013 extended identifiers, the function name must be given in UTF-8, not
17014 using universal character names.
17016 @item -fstack-check
17017 @opindex fstack-check
17018 Generate code to verify that you do not go beyond the boundary of the
17019 stack. You should specify this flag if you are running in an
17020 environment with multiple threads, but only rarely need to specify it in
17021 a single-threaded environment since stack overflow is automatically
17022 detected on nearly all systems if there is only one stack.
17024 Note that this switch does not actually cause checking to be done; the
17025 operating system or the language runtime must do that. The switch causes
17026 generation of code to ensure that they see the stack being extended.
17028 You can additionally specify a string parameter: @code{no} means no
17029 checking, @code{generic} means force the use of old-style checking,
17030 @code{specific} means use the best checking method and is equivalent
17031 to bare @option{-fstack-check}.
17033 Old-style checking is a generic mechanism that requires no specific
17034 target support in the compiler but comes with the following drawbacks:
17038 Modified allocation strategy for large objects: they will always be
17039 allocated dynamically if their size exceeds a fixed threshold.
17042 Fixed limit on the size of the static frame of functions: when it is
17043 topped by a particular function, stack checking is not reliable and
17044 a warning is issued by the compiler.
17047 Inefficiency: because of both the modified allocation strategy and the
17048 generic implementation, the performances of the code are hampered.
17051 Note that old-style stack checking is also the fallback method for
17052 @code{specific} if no target support has been added in the compiler.
17054 @item -fstack-limit-register=@var{reg}
17055 @itemx -fstack-limit-symbol=@var{sym}
17056 @itemx -fno-stack-limit
17057 @opindex fstack-limit-register
17058 @opindex fstack-limit-symbol
17059 @opindex fno-stack-limit
17060 Generate code to ensure that the stack does not grow beyond a certain value,
17061 either the value of a register or the address of a symbol. If the stack
17062 would grow beyond the value, a signal is raised. For most targets,
17063 the signal is raised before the stack overruns the boundary, so
17064 it is possible to catch the signal without taking special precautions.
17066 For instance, if the stack starts at absolute address @samp{0x80000000}
17067 and grows downwards, you can use the flags
17068 @option{-fstack-limit-symbol=__stack_limit} and
17069 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17070 of 128KB@. Note that this may only work with the GNU linker.
17072 @cindex aliasing of parameters
17073 @cindex parameters, aliased
17074 @item -fargument-alias
17075 @itemx -fargument-noalias
17076 @itemx -fargument-noalias-global
17077 @itemx -fargument-noalias-anything
17078 @opindex fargument-alias
17079 @opindex fargument-noalias
17080 @opindex fargument-noalias-global
17081 @opindex fargument-noalias-anything
17082 Specify the possible relationships among parameters and between
17083 parameters and global data.
17085 @option{-fargument-alias} specifies that arguments (parameters) may
17086 alias each other and may alias global storage.@*
17087 @option{-fargument-noalias} specifies that arguments do not alias
17088 each other, but may alias global storage.@*
17089 @option{-fargument-noalias-global} specifies that arguments do not
17090 alias each other and do not alias global storage.
17091 @option{-fargument-noalias-anything} specifies that arguments do not
17092 alias any other storage.
17094 Each language will automatically use whatever option is required by
17095 the language standard. You should not need to use these options yourself.
17097 @item -fleading-underscore
17098 @opindex fleading-underscore
17099 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17100 change the way C symbols are represented in the object file. One use
17101 is to help link with legacy assembly code.
17103 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17104 generate code that is not binary compatible with code generated without that
17105 switch. Use it to conform to a non-default application binary interface.
17106 Not all targets provide complete support for this switch.
17108 @item -ftls-model=@var{model}
17109 @opindex ftls-model
17110 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17111 The @var{model} argument should be one of @code{global-dynamic},
17112 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17114 The default without @option{-fpic} is @code{initial-exec}; with
17115 @option{-fpic} the default is @code{global-dynamic}.
17117 @item -fvisibility=@var{default|internal|hidden|protected}
17118 @opindex fvisibility
17119 Set the default ELF image symbol visibility to the specified option---all
17120 symbols will be marked with this unless overridden within the code.
17121 Using this feature can very substantially improve linking and
17122 load times of shared object libraries, produce more optimized
17123 code, provide near-perfect API export and prevent symbol clashes.
17124 It is @strong{strongly} recommended that you use this in any shared objects
17127 Despite the nomenclature, @code{default} always means public ie;
17128 available to be linked against from outside the shared object.
17129 @code{protected} and @code{internal} are pretty useless in real-world
17130 usage so the only other commonly used option will be @code{hidden}.
17131 The default if @option{-fvisibility} isn't specified is
17132 @code{default}, i.e., make every
17133 symbol public---this causes the same behavior as previous versions of
17136 A good explanation of the benefits offered by ensuring ELF
17137 symbols have the correct visibility is given by ``How To Write
17138 Shared Libraries'' by Ulrich Drepper (which can be found at
17139 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17140 solution made possible by this option to marking things hidden when
17141 the default is public is to make the default hidden and mark things
17142 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17143 and @code{__attribute__ ((visibility("default")))} instead of
17144 @code{__declspec(dllexport)} you get almost identical semantics with
17145 identical syntax. This is a great boon to those working with
17146 cross-platform projects.
17148 For those adding visibility support to existing code, you may find
17149 @samp{#pragma GCC visibility} of use. This works by you enclosing
17150 the declarations you wish to set visibility for with (for example)
17151 @samp{#pragma GCC visibility push(hidden)} and
17152 @samp{#pragma GCC visibility pop}.
17153 Bear in mind that symbol visibility should be viewed @strong{as
17154 part of the API interface contract} and thus all new code should
17155 always specify visibility when it is not the default ie; declarations
17156 only for use within the local DSO should @strong{always} be marked explicitly
17157 as hidden as so to avoid PLT indirection overheads---making this
17158 abundantly clear also aids readability and self-documentation of the code.
17159 Note that due to ISO C++ specification requirements, operator new and
17160 operator delete must always be of default visibility.
17162 Be aware that headers from outside your project, in particular system
17163 headers and headers from any other library you use, may not be
17164 expecting to be compiled with visibility other than the default. You
17165 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17166 before including any such headers.
17168 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17169 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17170 no modifications. However, this means that calls to @samp{extern}
17171 functions with no explicit visibility will use the PLT, so it is more
17172 effective to use @samp{__attribute ((visibility))} and/or
17173 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17174 declarations should be treated as hidden.
17176 Note that @samp{-fvisibility} does affect C++ vague linkage
17177 entities. This means that, for instance, an exception class that will
17178 be thrown between DSOs must be explicitly marked with default
17179 visibility so that the @samp{type_info} nodes will be unified between
17182 An overview of these techniques, their benefits and how to use them
17183 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17189 @node Environment Variables
17190 @section Environment Variables Affecting GCC
17191 @cindex environment variables
17193 @c man begin ENVIRONMENT
17194 This section describes several environment variables that affect how GCC
17195 operates. Some of them work by specifying directories or prefixes to use
17196 when searching for various kinds of files. Some are used to specify other
17197 aspects of the compilation environment.
17199 Note that you can also specify places to search using options such as
17200 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17201 take precedence over places specified using environment variables, which
17202 in turn take precedence over those specified by the configuration of GCC@.
17203 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17204 GNU Compiler Collection (GCC) Internals}.
17209 @c @itemx LC_COLLATE
17211 @c @itemx LC_MONETARY
17212 @c @itemx LC_NUMERIC
17217 @c @findex LC_COLLATE
17218 @findex LC_MESSAGES
17219 @c @findex LC_MONETARY
17220 @c @findex LC_NUMERIC
17224 These environment variables control the way that GCC uses
17225 localization information that allow GCC to work with different
17226 national conventions. GCC inspects the locale categories
17227 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17228 so. These locale categories can be set to any value supported by your
17229 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17230 Kingdom encoded in UTF-8.
17232 The @env{LC_CTYPE} environment variable specifies character
17233 classification. GCC uses it to determine the character boundaries in
17234 a string; this is needed for some multibyte encodings that contain quote
17235 and escape characters that would otherwise be interpreted as a string
17238 The @env{LC_MESSAGES} environment variable specifies the language to
17239 use in diagnostic messages.
17241 If the @env{LC_ALL} environment variable is set, it overrides the value
17242 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17243 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17244 environment variable. If none of these variables are set, GCC
17245 defaults to traditional C English behavior.
17249 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17250 files. GCC uses temporary files to hold the output of one stage of
17251 compilation which is to be used as input to the next stage: for example,
17252 the output of the preprocessor, which is the input to the compiler
17255 @item GCC_EXEC_PREFIX
17256 @findex GCC_EXEC_PREFIX
17257 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17258 names of the subprograms executed by the compiler. No slash is added
17259 when this prefix is combined with the name of a subprogram, but you can
17260 specify a prefix that ends with a slash if you wish.
17262 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17263 an appropriate prefix to use based on the pathname it was invoked with.
17265 If GCC cannot find the subprogram using the specified prefix, it
17266 tries looking in the usual places for the subprogram.
17268 The default value of @env{GCC_EXEC_PREFIX} is
17269 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17270 the installed compiler. In many cases @var{prefix} is the value
17271 of @code{prefix} when you ran the @file{configure} script.
17273 Other prefixes specified with @option{-B} take precedence over this prefix.
17275 This prefix is also used for finding files such as @file{crt0.o} that are
17278 In addition, the prefix is used in an unusual way in finding the
17279 directories to search for header files. For each of the standard
17280 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17281 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17282 replacing that beginning with the specified prefix to produce an
17283 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17284 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17285 These alternate directories are searched first; the standard directories
17286 come next. If a standard directory begins with the configured
17287 @var{prefix} then the value of @var{prefix} is replaced by
17288 @env{GCC_EXEC_PREFIX} when looking for header files.
17290 @item COMPILER_PATH
17291 @findex COMPILER_PATH
17292 The value of @env{COMPILER_PATH} is a colon-separated list of
17293 directories, much like @env{PATH}. GCC tries the directories thus
17294 specified when searching for subprograms, if it can't find the
17295 subprograms using @env{GCC_EXEC_PREFIX}.
17298 @findex LIBRARY_PATH
17299 The value of @env{LIBRARY_PATH} is a colon-separated list of
17300 directories, much like @env{PATH}. When configured as a native compiler,
17301 GCC tries the directories thus specified when searching for special
17302 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17303 using GCC also uses these directories when searching for ordinary
17304 libraries for the @option{-l} option (but directories specified with
17305 @option{-L} come first).
17309 @cindex locale definition
17310 This variable is used to pass locale information to the compiler. One way in
17311 which this information is used is to determine the character set to be used
17312 when character literals, string literals and comments are parsed in C and C++.
17313 When the compiler is configured to allow multibyte characters,
17314 the following values for @env{LANG} are recognized:
17318 Recognize JIS characters.
17320 Recognize SJIS characters.
17322 Recognize EUCJP characters.
17325 If @env{LANG} is not defined, or if it has some other value, then the
17326 compiler will use mblen and mbtowc as defined by the default locale to
17327 recognize and translate multibyte characters.
17331 Some additional environments variables affect the behavior of the
17334 @include cppenv.texi
17338 @node Precompiled Headers
17339 @section Using Precompiled Headers
17340 @cindex precompiled headers
17341 @cindex speed of compilation
17343 Often large projects have many header files that are included in every
17344 source file. The time the compiler takes to process these header files
17345 over and over again can account for nearly all of the time required to
17346 build the project. To make builds faster, GCC allows users to
17347 `precompile' a header file; then, if builds can use the precompiled
17348 header file they will be much faster.
17350 To create a precompiled header file, simply compile it as you would any
17351 other file, if necessary using the @option{-x} option to make the driver
17352 treat it as a C or C++ header file. You will probably want to use a
17353 tool like @command{make} to keep the precompiled header up-to-date when
17354 the headers it contains change.
17356 A precompiled header file will be searched for when @code{#include} is
17357 seen in the compilation. As it searches for the included file
17358 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17359 compiler looks for a precompiled header in each directory just before it
17360 looks for the include file in that directory. The name searched for is
17361 the name specified in the @code{#include} with @samp{.gch} appended. If
17362 the precompiled header file can't be used, it is ignored.
17364 For instance, if you have @code{#include "all.h"}, and you have
17365 @file{all.h.gch} in the same directory as @file{all.h}, then the
17366 precompiled header file will be used if possible, and the original
17367 header will be used otherwise.
17369 Alternatively, you might decide to put the precompiled header file in a
17370 directory and use @option{-I} to ensure that directory is searched
17371 before (or instead of) the directory containing the original header.
17372 Then, if you want to check that the precompiled header file is always
17373 used, you can put a file of the same name as the original header in this
17374 directory containing an @code{#error} command.
17376 This also works with @option{-include}. So yet another way to use
17377 precompiled headers, good for projects not designed with precompiled
17378 header files in mind, is to simply take most of the header files used by
17379 a project, include them from another header file, precompile that header
17380 file, and @option{-include} the precompiled header. If the header files
17381 have guards against multiple inclusion, they will be skipped because
17382 they've already been included (in the precompiled header).
17384 If you need to precompile the same header file for different
17385 languages, targets, or compiler options, you can instead make a
17386 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17387 header in the directory, perhaps using @option{-o}. It doesn't matter
17388 what you call the files in the directory, every precompiled header in
17389 the directory will be considered. The first precompiled header
17390 encountered in the directory that is valid for this compilation will
17391 be used; they're searched in no particular order.
17393 There are many other possibilities, limited only by your imagination,
17394 good sense, and the constraints of your build system.
17396 A precompiled header file can be used only when these conditions apply:
17400 Only one precompiled header can be used in a particular compilation.
17403 A precompiled header can't be used once the first C token is seen. You
17404 can have preprocessor directives before a precompiled header; you can
17405 even include a precompiled header from inside another header, so long as
17406 there are no C tokens before the @code{#include}.
17409 The precompiled header file must be produced for the same language as
17410 the current compilation. You can't use a C precompiled header for a C++
17414 The precompiled header file must have been produced by the same compiler
17415 binary as the current compilation is using.
17418 Any macros defined before the precompiled header is included must
17419 either be defined in the same way as when the precompiled header was
17420 generated, or must not affect the precompiled header, which usually
17421 means that they don't appear in the precompiled header at all.
17423 The @option{-D} option is one way to define a macro before a
17424 precompiled header is included; using a @code{#define} can also do it.
17425 There are also some options that define macros implicitly, like
17426 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17429 @item If debugging information is output when using the precompiled
17430 header, using @option{-g} or similar, the same kind of debugging information
17431 must have been output when building the precompiled header. However,
17432 a precompiled header built using @option{-g} can be used in a compilation
17433 when no debugging information is being output.
17435 @item The same @option{-m} options must generally be used when building
17436 and using the precompiled header. @xref{Submodel Options},
17437 for any cases where this rule is relaxed.
17439 @item Each of the following options must be the same when building and using
17440 the precompiled header:
17442 @gccoptlist{-fexceptions}
17445 Some other command-line options starting with @option{-f},
17446 @option{-p}, or @option{-O} must be defined in the same way as when
17447 the precompiled header was generated. At present, it's not clear
17448 which options are safe to change and which are not; the safest choice
17449 is to use exactly the same options when generating and using the
17450 precompiled header. The following are known to be safe:
17452 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17453 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17454 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17459 For all of these except the last, the compiler will automatically
17460 ignore the precompiled header if the conditions aren't met. If you
17461 find an option combination that doesn't work and doesn't cause the
17462 precompiled header to be ignored, please consider filing a bug report,
17465 If you do use differing options when generating and using the
17466 precompiled header, the actual behavior will be a mixture of the
17467 behavior for the options. For instance, if you use @option{-g} to
17468 generate the precompiled header but not when using it, you may or may
17469 not get debugging information for routines in the precompiled header.